Hans Henrich Hock, Brian D.

Joseph
Language History, Language Change, and Language Relationship
Hans Henrich Hock, Brian D. Joseph

Language History,
Language Change,
and Language
Relationship
An Introduction to Historical and Comparative
Linguistics

3rd revised edition

ISBN 978-3-11-060969-1
e-ISBN (PDF) 978-3-11-061328-5

Library of Congress Control Number: 2019935654

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available on the Internet at http://dnb.dnb.de.

© 2019 Walter de Gruyter GmbH, Berlin/Boston

Typesetting: Dörlemann Satz, Lemförde
Printing and binding: CPI books GmbH, Leck

www.degruyter.com
Preface to the Third Edition
It is gratifying and exciting to bring out another edition of this book. Particularly
in the case of a textbook, being able to do so means that the book has been suc-
cessful enough, and that the publisher and our readers have been happy enough
with it, to justify making updates and adjustments in content and presentation.
Moreover, on a personal level, this is something new for both of us – other books
of ours may have been reprinted or reissued or have had second editions, but a
third is a first for us, so to speak. So we have undertaken this revision with great
relish.
The job of reworking the material for this new edition has allowed us to make
some needed corrections and rewordings but also to reassess the content. We have
been able to decide what content to keep as is, what to eliminate, and what to
update. No chapter was left untouched in the process, each chapter contains at
least some minor changes, and some parts have been substantially revised. Parts
of Chapter 5, on analogy, for instance, have been deleted and other parts reor-
dered; and similarly Chapter 7, on semantic change, has in part received some
serious revision. Most important, however, is that the field of historical linguis-
tics, like language itself, has not stood still. Every year in the past decade has
witnessed new research and new challenges to consider and respond to. A new
discussion, § 7, on the regularity of sound change and its importance for general
historical-comparative linguistics has been added to Chapter 4, in reaction to
published critiques and misperceptions of this vital, foundational principle. A
good part of the section on the origin of language (Chapter 17, § 5) has been rewrit-
ten, reflecting advances in our understanding of this fascinating but probably
never-to-be-solved area of investigation. And just about all of Chapter 18 is new,
responding to new interpretations of archaeological and linguistic evidence,
some of them highly controversial, and also considering the challenges presented
by new advances in genomics.
In addition, the Notes and References have been updated. As before, we invite
readers to use them to follow up on topics that they find particularly interesting.
In short, we hope that you, our readers, have as good a time in reading and
working with this revision as we have had in producing it!

Champaign, Illinois and Columbus, Ohio

November 2018
Hans Henrich Hock ([email protected])
and Brian D. Joseph ([email protected])

https://doi.org/10.1515/9783110613285-202
Contents
Chapter 1: Introduction 1
1 Language keeps changing 1
2 Types of linguistic change 6
3 Language relationship 12
4 A word of caution, or “Long live the speaker” 14
5 A note on transcription and terminology 16

Appendix to Chapter 1: Phonetics, phonetic symbols, and other symbols 18

Chapter 2: The discovery of Indo-European 29

1 Language relationship 29
2 Proto-Indo-European 32
3 The Indo-European languages 35
4 Abbreviations of Indo-European language names 55

Chapter 3: Writing: Its history and its decipherment 57

1 Introduction 57
2 History of writing 57
3 The decipherment of ancient scripts 85
4 The phonetic interpretation of written records 91
5 Writing in the rest of the world 95

Chapter 4: Sound change 104

1 Introduction 104
2 Grimm’s Law 104
3 From Grimm’s Law to Verner’s Law 108
4 The regularity hypothesis and the neogrammarians 113
5 Some types of sound change 116
6 Why sound change? 132
7 The regularity of sound change redux 139

Chapter 5: Analogy and change in word structure 141

1 Introduction 141
2 Relatively systematic analogy 142
3 Sporadic or non-systematic analogy 151
4 Analogy and phonology: Rule-governed, regular analogy 162
5 Hypercorrection – an interdialectal form of analogy 164
6 Morphological change 166
VIII Contents

Chapter 6: Syntactic change 170

1 Introduction 170
2 Questionable “syntactic” changes 171
3 Syntax, analogy, or both? 172
4 Me revisited, or the critics’ revenge 175
5 A successful major shift: Word order in English and related
languages 181
6 Conclusion 188

Chapter 7: Semantic change 189

1 Introduction 189
2 The inherent fuzziness of meaning – polysemy, semantic overlap,
metaphor 191
3 Synonymy and homonymy 196
4 The relationship between sound and meaning 198
5 Factors responsible for semantic change 200
6 The effects of semantic change 211
7 Conclusion 222

Chapter 8: Lexical borrowing 223

1 Introduction 223
2 The substance of borrowing 225
3 Nativization, or how do you deal with a word once you have
borrowed it? 228
4 “Hyper-foreignization” – A further effect of
borrowing 239
5 Why borrow? Motivations for borrowing strategies 240
6 The effects of borrowing 253

Chapter 9: L exical change and etymology

The study of words 260
1 Introduction 260
2 Coinage 264
3 Proper names: A case study in lexical origins 270
4 Coinage in argots, jargons, and slang 278

Chapter 10: Language, dialect, and standard 284

1 Introduction 284
2 Language and dialect 285
3 Social dialects 290
 Contents IX

4 Discontinuous dialects – Professional jargons and related forms of

speech 291
5 Standard languages 293
6 Diglossia 300
7 Dialect borrowing 303

Chapter 11: Dialect geography and dialectology 308

1 Introduction 308
2 Patterns of dialect interaction 310
3 Focal, transition, and relic areas 312
4 Dialectology as a diagnostic tool 313
5 Isoglosses and the problem of defining regional dialects 315
6 Migration and dialect leveling 319

Chapter 12: Language spread, link languages, and bilingualism 322

1 Introduction: Link languages and their sources 322
2 Interference and interlanguage 327
3 Code switching and code mixing 332
4 Substratum 335
5 Koinés 339
6 Outlook 341

Chapter 13: Convergence: Dialectology beyond language boundaries 343

1 Introduction: Convergence defined 343
2 Convergence illustrated – Kupwar 346
3 The Balkans 350
4 South Asia 356
5 Europe 363
6 The dialectology of convergence areas 363

Chapter 14: P
 idgins, creoles, and related forms of language 366
1 Introduction: Foreigner Talk, “Tarzanian”, and other simplified
forms of speech 366
2 Pidgins defined 371
3 Pidgin origins 374
4 Trade Jargons and other pidgin-like languages 381
5 Creoles 384
6 Decreolization and African American Vernacular English 388

Chapter 15: Language death 392

X Contents

Chapter 16: C
 omparative method: Establishing language relationship 398
1 Introduction 398
2 Chance similarities, onomatopoeia, and “nursery words” 402
3 Similarities due to linguistic contact 404
4 Systematic, recurrent correspondences 406
5 Shared idiosyncrasies 407
6 Reconstruction 408
7 What can we reconstruct and how confident are we of our
reconstructions? 412
8 Language families other than Indo-European 416

Chapter 17: P
 roto-World? The question of long-distance genetic
relationships 426
1 Introduction 426
2 Longer-distance comparison 429
3 Are there any unrelated languages? 436
4 Lexical mass comparison: Can it establish “Proto-World”? 437
5 The origin of Language 443

Chapter 18: L inguistic palaeontology: Historical linguistics, history, and

prehistory 447
1 Introduction 447
2 Preliminary I: The term “Aryan” 449
3 Preliminary II: How many PIEs? 449
4 Religion, society, and law 451
5 Material culture 456
6 Economy 457
7 Horses, wheels, and more 461
8 Time and original home (“Urheimat”) of PIE 466
9 Genetics, genomics, and “race” 474
10 Ideology vs. comparative linguistics and ­linguistic
palaeontology 480
11 Conclusions and outlook 485

Chapter notes and suggested readings 486

References 513
Language index 543
General index 552
Chapter 1: Introduction
glam.our, glam.or (glam´ǝr) n. [Scot. var. of grammar (with sense of gramarye), popularized
by Sir Walter Scott; orig. esp. in cast the glamour, to cast an enchantment] 1. orig., a magic
spell or charm 2. seemingly mysterious and elusive fascination or allure, as of some person,
object, scene, etc.; bewitching charm: the current sense
(Webster’s New World Dictionary of the American Language, Second College Edition, 1970)

verve (vûrv) n. 1. Energy and enthusiasm in the expression of ideas and especially in artistic
endeavor: The play lacks verve. 2. Vitality; liveliness; vigor. 3. Rare. Aptitude; talent. [French,
from Old French, fancy, fanciful expression, from Latin verba, plural of verbum, word …
(The American Heritage Dictionary of the English Language, 1969)

1 Language keeps changing

Hey, it’s far out, man.
I don’t know if you can, I don’t know uh like how many of you can dig how many people
there are, man.
Like I was rappin’ to the fuzz, hunh, right, can you dig it?
Man, there’s supposed to be a million and a half people here by tonight. Can you dig that?
New York State Thruway is closed, man, hunhhunh.
Yeah, lotta freaks, hunhunh.

These are the words that Arlo Guthrie used at the end of his song “Coming into
Los Angeles”, bantering with the masses of young people who were gathered at
Woodstock in August 1969 for the most famous rock festival of the time.
When we wrote the first edition of this book, the event was still relatively fresh
on our minds, but we were also aware that Guthrie’s language was already quite
removed from the youth language of the late 1990s. At the same time, through
our kids we would have been able to come up with current counterparts for some
of Guthrie’s expressions, such as awesome or cool for far out. Now we are at the
awkward stage where even our kids aren’t up-to-date on slang. We know that rap
nowadays most commonly refers to a type of music, but we can’t be sure whether
in Rap (or elsewhere), the word rap is still used to mean ‘talk’ or ‘speak’. Hitting
the internet for suggestions on current equivalents of Guthrie’s usage isn’t par-
ticularly helpful. Some sites suggest 50 (five-oh) for Guthrie’s fuzz, i. e. the police;
but how can we know whether this is actually widely used or still current? The
general rule is that once slang expressions appear in writing or on the internet,
they might as well be dead.
Language change, however, is not limited to slang. It affects all areas of lan-
guage use, even the staid scholarly world, as the following example illustrates.

https://doi.org/10.1515/9783110613285-001
2 Introduction

The American Oriental Society is a learned society dedicated to the lan-

guages, literatures, and cultures of all of Asia, from Israel and Palestine to China
and Japan. Yet at a meeting in the 1990s, the proposal was made to rename the
organization: It was argued that its name suggests an interest only in the Far East.
The proposal was greeted with disbelief, even outrage, and was quickly voted
down.
This minor incident, in a scholarly society most people have never even heard
of, puts into stark relief the fact that language keeps changing inexorably and that
even scholars dedicated to the study of language, literature, and culture can be
caught unawares by change.
Evidently, unbeknownst to both sides, the meaning of the word oriental has
changed for many speakers of modern American English. When the Society was
founded in 1842, the word was used to refer to the area just east of Europe in which
the sun rises – the ancient Near East of Israel, Babylonia, and adjacent areas.
And scholars grounded in Latin knew that the word oriental is derived from Latin
oriens ‘rising (sun)’. (In this regard note that the name of the German counterpart
of the Society uses the term morgenländisch ‘connected with the “morning land”,
the east’.) It was only by extension that the word was used to refer to areas even
farther east, eventually to all of Asia.
Since the founding of the Society, several things happened. To become a com-
petent scholar of oriental studies it is no longer necessary to know Latin and thus
to be able to know the “literal” meaning of oriental. As a consequence, the term
oriental cannot be interpreted literally anymore: from the perspective of America,
the sun rises in the Atlantic Ocean or, possibly, in Europe, not the Near East. A
perhaps even more important element is this: Even though we can look up the
established meaning of words in a dictionary, we learn most meanings from the
way words are used around us, without any formal instruction as to what is the
“real” meaning of a word. As a result, we have to figure out most meanings for
ourselves. In the process we may not always get things right (from the perspective
of the older generation). This is especially true for meaning, since words normally
have a fairly broad range of meanings. Consider the word reader. We all know
what the word basically means, or at least we believe so and act accordingly. But
put the word into different contexts and you find that its exact connotations can
differ considerably. For instance, the expression Johnny is a good reader conveys
very different messages, depending on whether Johnny is in first grade or in a
poetry-reading class in college. We deal with such issues by assuming that there is
a prototypical or core meaning and that other meanings, such as the use of reader
in the context of a poetry-reading class, are transferred or more peripheral. But,
again, for most words we have to figure out the core meaning ourselves, and in
the process we may go wrong.
 Language keeps changing 3

Something like this seems to have happened to the word oriental, such that a
sizable number of speakers of American English have determined for themselves
that the prototypical or core meaning of the word refers to East Asia – the area
of Asia and its inhabitants that differs most prototypically from Europe and its
inhabitants, and from the European-descended majority population of North
America. This change in interpretation, however, is a fairly recent development.
For a large number of American scholars, the word continues to mean just about
the same as Asian, at least in scholarly contexts.
But things are even more complex: It is quite doubtful that most scholars
are so highly ensconced in their ivory towers that they do not know that oriental
means ‘East Asian’ in every-day usage. Rather, it seems that they consider this
usage secondary or peripheral, “colloquial” rather than “scholarly”, if not down-
right “slangy”. And instead of simply failing to understand why some of their col-
leagues wanted to change the name of their Society, they probably were outraged
at the intrusion of the non-scholarly connotation that oriental had acquired.
What is interesting about this reaction is the use of the term “slangy” in ref-
erence to a usage considered less correct, i. e., less prestigious. This is a common
reaction to linguistic change, especially among those who believe that language
must remain pure and unchanged, and that change will somehow reduce not only
the purity of language but also our ability to speak and even think clearly. Whole
books are written – and have been written for centuries – warning of impending
doom, prophesying that our language will go to the dogs if certain usages disap-
proved by the authors run rampant.
Interestingly, these critics often do not agree with each other on which usages
should be disapproved. Many American critics still inveigh against the use of hope-
fully in expressions like Hopefully, it will rain and advise the use of something like
I hope it will rain instead. Their British colleagues generally are amused by this bit
of linguistic conservatism and find the use of hopefully quite compact and handy.
The targets of disapproval may also change over time. Up to about the 1960s the
use of data as a singular mass noun, rather than a (historically correct) plural
form of datum, was subject to continued criticism; but nobody objected to the
singular use of agenda, originally plural of agendum ‘to be dealt with’. Today, as
the singular use of data has been accepted by most educated speakers of English,
its singular use generally is no longer an issue. Instead, the debate centers around
words like media and criteria, historically plurals of medium and criterion, respec-
tively, which are undergoing similar changes from plural to singular use.
Most judgmental statements of this sort come from non-linguists. But lin-
guists have not always been free of such prejudices, either. Up to about the 1870s,
most historical linguists subscribed to the idea that language change is tanta-
mount to decay. Their initiation to linguistics included a thorough study of the
4 Introduction

classical European languages, Latin and Greek, and they had been persuaded
that these glorious tongues of classical antiquity were the most perfect on earth,
and that the modern languages were but poor shadows of them. This view was
consonant with traditional Christian and pre-Christian beliefs according to which
an original, perfect, and idyllic garden of Eden or “Golden Age” gave way to an
ever-worsening fall from grace, to ever-increasing decay and depravity.
Anyone who has studied Greek and Latin knows that the word structure (or
“morphology”) of these languages is “richer” than that of most modern Euro-
pean languages. Thus, Latin nouns had six different “cases”, forms of nouns
whose choice depended on the syntactic context. There was a nominative for the
subject of the sentence (caesar ‘Caesar’), a genitive to indicate possession (caesa-
ris ‘of Caesar’), a dative which among other things indicated the recipient of a gift
(caesari ‘to, for Caesar’), an accusative for the object of the sentence (caesarem
‘Caesar’), an ablative to indicate the source of an action, including the agent of
the passive (caesare ‘from, by Caesar’), and a vocative for addressing a person or,
more rarely, a thing (caesar ‘O Caesar’). In contrast, Modern French nouns have
one invariant form (César). The situation is similar in English and many other
modern languages. This reduction in morphological richness was considered
simply another manifestation of general human sloth and depravity.
Other linguists have claimed that the morphological reduction really is an
improvement: Not having to memorize four, five, six, or even more cases for each
noun simplifies the language and thereby makes it more efficient and easier to
learn.
On the surface, the second view is more appealing, especially in an age that
worships the notion of progress. However, “progress” is as much a subjective
notion as is “decay”. More than that, we have no objective way of telling whether
a language with a richer case system is more difficult than one with a “poorer”
one. True, learning the six different nominal cases of Latin, or the four cases of
German may appear difficult to speakers of English who are used to just two such
forms (base form [wolf] vs. genitive [wolf’s]), but when they go to Germany they
find, often much to their surprise, that “even the children speak German” and get
their cases right most of the time.
Linguists are not surprised: One of the few generally accepted beliefs in lin-
guistics is that all children manage to learn their own native language with equal
ease and efficiency and that, by extension, all languages are equally “simple” or
“complex”.
This does not mean that as individual users of language, linguists are com-
pletely free of personal preferences or even prejudices, especially as regards
ongoing changes in grammar and usage; and they may do their best to stem par-
ticular changes which they consider undesirable. Objectively, however, they are
 Language keeps changing 5

fully aware that over the long haul, these attempts at stemming the tide of change
are ineffectual: Language changes inexorably. But interestingly, in the process
it does not go to the dogs. We are probably as capable, if we try hard enough, to
express our ideas clearly and effectively as our linguistic ancestors were – when
they tried hard enough. At any rate, we have not started barking as yet.
The extent to which language changes can be seen more clearly by extending
our horizon beyond the five decades that separate Arlo Guthrie’s language from
the present. Compare for instance the Lord’s Prayer as it was translated into Old
English about a thousand years ago with a more modern version.

Old English (ca. 950 A.D. The original text has been slightly simplified.)
Fader urer ðu arð in heofnum, sie gehalgad noma ðin, to-cymeð ric ðin, sie willo ðin suæ is
in heofne ond in eorðo, hlaf userne oferwistlic sel us todæg ond forgef us scylda usra suæ uœ
forgefon scyldum usum, ond ne inlæd usih in costunge, ah gefrig usich from yfle.

Modern English (ca. 1985)

Our father who is in heaven, may your name be sacred. Let your kingdom come. May your will
be fulfilled just as much on earth as it is in heaven. Give us today our daily bread. And forgive
us our transgressions, as we forgive those who transgress against us. And do not lead us into
temptation, but deliver us from sin.

Both in Old English and Modern English, of course, other translations are possible
and have in fact been produced. But the difference between the two versions cited
here goes much beyond individual word choices; it concerns the entire language.
So much so, in fact, that Old English is at least as “foreign” to a speaker of Modern
English as, say, Modern German. Nevertheless, Old and Modern English are in
some sense the “same” language, with Modern English resulting from centuries
of linguistic change taking place in structure and vocabulary. Linguistic change
evidently affected not just meaning and usage (as in our earlier examples), but
pronunciation, grammar, and everything else.
And, it does not stop. Even in the relatively short span between the second
edition of this book in 2009 and the current edition, new words and usages have
entered the language. The modifier ish, with the meaning ‘approximately’, was
until recently used mostly wih adjectives as an “honest, God-fearing” suffix as
in reddish ‘somewhat red’. In current usage it has taken a new life as something
like an independent word that can be used in exchanges like this one: Are you
hungry? – Yeah, ish (meaning ‘I’m kind of hungry’).
6 Introduction

2 Types of linguistic change

Sound change. What perhaps most saliently distinguishes the Old English (OE)
translation of the Lord’s Prayer from Modern English (Mod. Engl.) counterpart is
that the two are pronounced very differently, as reflected in part by differences
in spelling. In some words, the differences are minor, as in the word for ‘father’
where the major difference is Old English d : Mod. Engl. th. In others, the differ-
ence is much greater, as in OE hlaf ‘bread’ (with a vowel similar to the a in Mod.
Engl. father) vs. the corresponding Mod. Engl. loaf, which lacks the initial h and
has a different vowel. In addition, of course, loaf has changed its meaning, no
longer referring to ‘bread’ in general, but to a certain quantity of bread. (This
is the reason that the word cannot be used in the modern version of the Lord’s
Prayer.)
Linguists use the term sound change to refer to such changes in pronunci-
ation; and like many other technical terms, it is defined more narrowly than in
ordinary usage, to refer only to certain types of change in the pronunciation of
words. Sound change understood in this sense has been claimed to be completely
regular, in the sense that all words that can undergo a given shift in pronunciation
do so. Recent research suggests that this claim may be an overstatement; still,
sound change does turn out to be overwhelmingly regular.
It is this overwhelming regularity of sound change that makes it possible
for us to trace words through history, or across different, but related languages,
and to be confident that we are really comparing different versions of the same
word, rather than words which just happen to sound similar. Understanding how
sound change operates often helps us explain otherwise strange facts about lan-
guage. For instance, English has a number of words pronounced with initial n but
preceded in spelling by a “silent” k, such as knee, knight, and knave. The k of these
words was pronounced in early English (as in OE cneo, cniht, cnafa) and is still
pronounced in the related German (e. g. Knie ‘knee’). It was lost in English by a
regular sound change that operated in word-initial position before n. The modern
spelling with k reflects an earlier stage when the k was still pronounced.

Analogy. Sound change is not the only change that may affect pronunciation.
Words often change their pronunciation under the influence of other words, or
by analogy with them. For instance, the early Modern English plural of cow was
kine, a form still found in nineteenth-century poetry. The present-day plural cows
came about in the seventeenth century under the analogy of the most common,
productive mode of plural formation, as in pig : pig-s, horse : horse-s. Unlike sound
change, analogy is not normally regular; and Modern English has retained many
irregular plural forms such as men, women, children, feet.
 Types of linguistic change 7

Analogy comes in many different shapes, some of which may be exploited

for creating new words. One of these is a process called blending which produces
something like a compromise between two competing words. An example of
an ordinary, unconscious blending is feets, a form frequently used by children
as a compromise between the “correct” plural feet (insisted on by grown-ups)
and a more regularized analogical foots, comparable to our earlier cows. More
conscious applications of the same process are found in words like brunch, a
compromise between breakfast and lunch, transceiver, which is simultaneously
a transmitter and a receiver, or modem, a modulator-demodulator. A similar
explanation may hold true for irregardless, a form used in a variety of English
that is commonly considered “nonstandard”, “vernacular”, or even “ungram-
matical”. The word may owe its existence to a blending of irrespective and
regardless.
Analogy, combined with sound change, may have profound effects on word
structure (also known as morphology). Thus, Old English was similar to Latin
in having a fairly “rich” nominal case system. For instance, the word for ‘stone’
had the following inflection, with three different case endings, each distinct for
singular and plural:

singular plural
nominative/accusative: stān stānas
dative: stāne stānum
genitive: stānes stāna

Modern English has just two cases:

base form stone stones

genitive stone’s stones’

The primary factor underlying the change from Old English to Modern English is
sound change, which eliminated all of the vowels in the suffixes, as well as the m
of the dative plural. But if only sound change had applied we would have a system
that differs from the one actually found:

nominative/accusative: stone stones

dative: stone stone
genitive: stones stone

As can be readily seen, this hypothetical system does a very poor job of keeping
the different case forms apart. The same form is used for the nominative/accusa-
8 Introduction

tive and dative singular and also for the dative and genitive plural; and another
form is employed for the nominative/accusative plural and genitive singular.
At this point, analogy stepped in and put some greater order into the system,
by extending the -s of the nominative/accusative plural to all other forms of the
plural. As a consequence, neither singular nor plural distinguished between
nominative/accusative and dative, making this distinction superfluous in English
grammar. If the development had been carried to its logical conclusion, the gen-
itive forms would have been affected, too, and Modern English would no longer
distinguish the genitive from the base form. But as we just noted, analogy does not
apply with the same regularity as sound change; and the English genitive singular
merrily retained its -s. (The different s-forms were then differentiated in writing,
but not in pronunciation, by a judicious use of an apostrophe.)

Semantic change. As we have seen earlier, words may also change their meaning.
This type of change, referred to as semantic change, is notoriously unpredict-
able and “fuzzy”, probably because of the way in which we readily stretch and
extend the meaning of words to cover new situations. (Recall the different uses of
the word reader mentioned in section 1.) One of the consequences of the fuzziness
of semantic change is that semantic flip-flops may occur. As noted earlier, OE hlaf
‘bread’ corresponds to Mod. Engl. loaf through sound change, but the modern
word designates a narrower semantic range, namely a certain quantity of bread.
Exactly the opposite happened in the case of the modern word bread. This word
can be traced back to OE bread (probably different in pronunciation); but the
meaning of the Old English word was more narrow: ‘(bread) crumb, morsel’. One
of the most remarkable flip-flops of this kind is the relatively recent slang use of
bad or sick to mean their exact opposite, ‘excellent, cool, etc.’.
Semantic change can lead to many other, quite radical and unexpected
results. Perfect examples for such changes are the words glamour and verve.
Consider first the word glamour. The ultimate source of the word is Greek
gráphein ‘to scratch’. By a fairly mundane change the verb came to mean ‘write’
after the advent of writing. (This semantic shift has parallels in many other lan-
guages, such as Latin scribere, English write, Sanskrit likh-, all of which originally
meant ‘scratch’.) Once gráphein had changed its meaning, nouns derived from
it, such as grámma and grammatikḗ came to refer to the products of writing: a
letter of the alphabet, a letter of communication, or letters ‘learning in general’,
as in Arts and Letters, Doctor of Letters, etc. From Greek the word entered Latin as
grammatica, with roughly the same range of meanings.
In Old French, a new derivative was created, grammaire, whose meaning
underwent a certain expansion, referring not only to ‘(Latin) grammar’ and ‘phil-
ological learning’, but to all traditional learning, including the occult sciences of
 Types of linguistic change 9

alchemy and astrology. The latter meaning was especially prevalent among the
“unlettered”, for whom the art of writing conjured up images of wizards poring
over books on alchemy, magic, and the supernatural. After all, being able to read
and write was an esoteric phenomenon in a society where literacy was limited to
a small portion of the population.
The word was borrowed as gramer into Middle English, with the same range of
meanings as in Old French and, again, with magical connotations mainly among
the “unlettered”. Along the way, the two meanings of the word, the educated one
of ‘(Latin) grammar and learning in general’ and the popular magical interpreta-
tion, came to be differentiated in writing, giving rise to Mod. Engl. grammar vs.
gramary(e).
The form glamour is in origin a Scots English variant of gramary(e) and was
introduced into modern standard English by Sir Walter Scott with the meaning
‘magic; magic charm’. The later development to ‘charm’ and related meanings
reflects a common metaphorical extension from ‘magic charm’, similar to the
more transparent metaphorical use of words like bewitching. In fact, the word
charm, and the word enchanting as well as, exhibit the same development; the
more original meanings are preserved in expressions like cast a charm on someone
and the somewhat archaic enchantress ‘female sorcerer’.
Along the way, the word grammar lost a lot of its earlier glamour, as it were,
and increasingly was used to refer to instruction in linguistic structure, often with
emphasis on “correctness”. Its earlier, more general meaning remains in fixed
expressions like grammar school, a school which was intended to inculcate not
just grammar in the modern sense, but learning in general.
The word verve, too, derives from a word dear to linguists, namely Latin verba,
plural of verbum ‘word’. (Along a different path this word furnished Mod. Engl.
verb, likewise an important element in the vocabulary of linguists.) The Latin
word could by extension also refer to general sayings or proverbs and to some-
thing like ‘mere, empty words’.
The early French outcome of verba was verve, whose meanings, ranging
from ‘proverb’ to ‘verbosity’, can easily be explained as specializations of the
Latin meanings. In later medieval French, verve came to be used in the meaning
‘caprice, fantasy’, possibly an extension from ‘verbosity’ via something like
‘verbal exuberance’.
From the later medieval connotations it is only a short step to the modern
French meaning ‘enthusiasm, vitality, etc.’; and it is with this meaning that the
word was borrowed from French into English.
10 Introduction

Syntactic change. In addition to their pronunciation, morphology, and meaning,

words can also change in their syntax – the way they are put together into sen-
tences, and how these sentences, in turn, are related to each other. Take for
instance the order of subject, verb, and object in sentences like The dog (subject)
bit (verb) the man (object). In Modern English the normal order places the verb
between the subject and the object, not only in sentences like The dog bit the
man but also in more unusual sentences like The man bit the dog. By contrast,
“verb-initial structures” like Bit the man the dog or “verb-final” ones like The man
the dog bit do not qualify as well-formed complete English sentences (although
they may be part of well-formed sentences, as in The man the dog bit is getting
rabies shots).
Old English had a much greater freedom of word order and the Old English
counterpart of Bit the man the dog was perfectly acceptable. The most unmarked
sentence structure, however, would have been closest to the verb-final type The
man the dog bit.
The change from the Old English verb-final syntax to the modern verb-medial
pattern was partly shared by German, a closely related language. However, in
German the development stopped in midstream, as it were: Only those verb forms
which have personal endings (such as Engl. third person has vs. non-third person
have) appear in medial position. If the verb consists of more than one word (as in
Engl. has gone : have gone), the elements without personal endings stay at the end
of the sentence. To make matters even more complicated, German verb-medial
syntax is limited to main clauses, dependent clauses are verb-final. This accounts
for the curious mismatch in word order between English and similar languages
(such as French) with more or less solidly medial verbs on one hand, and German
on the other:

English German
He loves his wife Er liebt seine Frau
She has loved her husband Sie hat ihren Mann geliebt
that the children love their parents dass die Kinder ihre Eltern lieben

It is this mismatch which is in large measure responsible for the notorious diffi-
culties experienced by speakers of languages like English when confronted with
German.

Change resulting from language contact. The developments illustrated so far

are the major linguistic changes that affect languages under all circumstances. A
number of other changes take place only when different languages (or dialects)
are in contact with each other.
 Types of linguistic change 11

The most common development in contact situations is borrowing, the

adoption or adaptation of words from one language into another. Modern English
is full of such borrowings; note such culinary expressions as curry (from Tamil,
a language of South India), frankfurter (German), bagel (Yiddish), paté (French),
enchilada (Spanish), spaghetti (Italian), pita (Greek), kimchee (Korean). Many
other languages, or more accurately, their speakers are much more reluctant to
adopt foreign words to this large extent. In fact, languages like Chinese have
severe limitations in this regard.
Extended bilingual or multilingual contact between languages can lead to an
increase in structural similarities as well. Many areas of the world are notori-
ous for developments of this sort, for instance the Balkans, South Asia, South-
east Asia, and the indigenous languages of the northwestern United States and
adjacent areas of Canada. The idea that structural elements should diffuse from
one language to another often meets with disbelief. People are prepared to accept
lexical borrowings, since these go hand in hand with the borrowing of the objects,
ideas, or concepts expressed by the words; but why would languages borrow, say,
a “passive” construction – if earlier they were perfectly happy without it?
Actually, structural elements usually do not diffuse through borrowing, but
are the cumulative results of a different, very common phenomenon. This is the
transfer of pronunciation and other aspects of linguistic structure that leads to
the “accent” with which people speak a foreign language, even after long years
of residence in the country where the language is natively spoken. In fact, the
development does not seem to be entirely unilateral. True, we tend only to notice
the accent with which the foreigner speaks our language; but when that foreigner
returns home after many years abroad, those who remained behind will notice a
certain accent in his or her speech as well. This is often considered an affectation,
and to some extent it may be. But to a certain degree the “foreign” accent in one’s
native language is as natural and normal as the accent in one’s second, non-na-
tive language.
Under certain conditions, language contact can lead to yet other results.
One of the most striking is the development of pidgins, languages with minimal
linguistic structure and, even more important, with minimal vocabulary. The
lexicon of such languages has been estimated to contain no more than 1,000 to
2,000 words. Many ideas which find compact single-word expressions in “normal”
languages therefore have to be rendered by circumlocutions. It is said that the
word piano is rendered in one of the pidgins as something like big fellow box you
fight him he cry. A large number of pidgins arose in the wake of European impe-
rialist-colonialist expansion around the world, when speakers of European lan-
guages came in contact with speakers of languages that were totally unfamiliar to
them.
12 Introduction

Even such extreme developments can be explained (at least in part) by prin-
ciples commonly observable elsewhere. Sentences with extreme structural and
lexical simplifications are frequently used in one of two contexts: (a) with babies
who are not yet able to speak the full form of the language, (b) with speakers
whose language we do not understand, and who do not understand ours. Exam-
ples are expressions like Baby want (to) seep? = Does the baby/do you want to
sleep? and the famous Me Tarzan – You Jane = I am Tarzan, you are Jane. Though
there are considerable differences in details, especially in friendliness or rude-
ness of intonation, both types of discourse share one thing beyond structural and
lexical simplification: They are used in situations where we believe that we cannot
successfully communicate in ordinary language and that, therefore, we must sim-
plify our language, on the assumption that we will somehow get through – if we
simplify enough and, in the case of foreigners, speak slowly and loudly enough.

3 Language relationship
Language change, thus, is not only pervasive but also takes many different forms.
More than that, even within a single type of linguistic change, such as sound
change, there are many different possible subtypes. In the preceding section,
for instance, we have seen that English lost initial k before n (as in knee), while
German did not. Similarly, we have observed that English developed a solidly
“verb-medial” syntactic pattern, while German stopped in midstream, as it were.
Such differences in development are very common, especially when lines of com-
munication are attenuated or even broken.
We can observe many examples of this phenomenon in the different varieties
of English, especially comparing American and British English. The differences
affect pronunciation, vocabulary, word formation, and even syntax. Thus, British
English generally “drops” the r in words like cart and car, while American English
generally does not; American English “drops” the h of herb, while British English
does not. Where British English has lorry, bonnet, boot, American English offers
truck, hood, trunk. Differences between British and American English can some-
times lead to genuine misunderstandings. It is said that at a joint-staff meeting
of the Allied Command during the Second World War, a British officer proposed
that an important matter be tabled, whereupon his American counterpart became
angry. He interpreted the word the American way, as a near-synonym of shelve
(i. e. ‘delay’), whereas the intended meaning was the British one of placing the
matter on the table for immediate discussion. One of the most commonly encoun-
tered differences in word formation concerns the past participle of the verb get.
 Language relationship 13

Whereas American English makes a distinction between I have gotten a letter from
home ‘I have received a letter…’ and I’ve got a letter from home ‘I have a letter …’,
British English uses got in both contexts. In the area of syntax, questions like Did
you give Mary the hat? can perfectly acceptably be answered in British English
by saying I gave her it yesterday, whereas such a structure would be considered
unacceptable in (standard) American English.
Divergent changes of this type, if continuing over a long enough period, can
be pervasive enough to change what originally were different varieties of the same
language into effectively different languages, much as a millennium of linguistic
changes has effectively turned Old and Modern English into different languages.
In fact, it is through such divergent developments that Latin, the language of
the Roman Empire, came to be differentiated into the Romance languages (Por-
tuguese, Spanish, French, Italian, Romanian, etc.). Similar developments can be
observed in northern India, whose modern languages are the differentiated out-
comes of Sanskrit. Linguistic relationships of this type have been known for a long
time. Many other relationships had long been suspected, but it was not before the
end of the eighteenth century that linguists began to establish some of these rela-
tionships beyond a reasonable doubt. Some modern linguists even believe that
they can establish that all human languages are related to each other; and their
claims have received a fair amount of attention in the popular press. The claims,
if correct, would add another important element to the continuing discussion of
the question of human origins. At this point, most historical linguists still con-
sider the issue controversial, but the ensuing debate has introduced an element
of excitement – as well as acrimony – into an otherwise rather staid profession.
What distinguishes some of the relationships that were established later from
the Romance “language family” is this: Latin, the ancestor from which the
Romance languages have “descended”, is historically attested. There could there-
fore be no doubt that the Romance languages are “daughter” languages sprung
from the same “mother”.
The Germanic languages (English, German, Dutch, Frisian, Norwegian,
Danish, Swedish, Icelandic, etc.) exhibit a degree of similarity with each other
that is not substantially different from the one between the Romance languages.
Note especially the striking similarities in such an idiosyncratic pattern as Engl.
good : better : best, Germ. gut : besser : best, Icel. góður : betri : bezt-. However,
no mother language is historically attested from which the Germanic languages
might have developed as daughters. Claiming that the Germanic languages are
nevertheless related, therefore, requires the assumption of a linguistic ancestor
which happened to be spoken before the arrival of literacy.
In the case of Germanic, this was actually not a major problem, since in the
Middle Ages the languages were still quite similar to each other. It was much more
14 Introduction

difficult to accept that languages as disparate as Latin, Sanskrit, and the Germanic
languages might be related to each other. In this case there was both an absence of
a known linguistic ancestor and a much greater degree of differentiation.
The linguistic relationship of Latin, Sanskrit, and Germanic now is firmly
established, too; and so are similar relationships between many other languages
of the world. A method called comparative reconstruction has made it possi-
ble to develop at least some ideas on the structure and vocabulary even of unat-
tested linguistic ancestor languages. Linguists and prehistorians even have drawn
on reconstructed vocabulary to draw inferences about the culture and society of
the speakers of reconstructed languages.
The results of reconstruction clearly are only theories or hypotheses, and as
such they are subject to revision as new evidence is considered or as old evidence
is reconsidered. Nevertheless, an attempt to test the method on the Romance
languages (where we can compare the reconstructed ancestor language with the
attested Latin) suggests that the method can yield amazingly accurate results.

4 A word of caution, or “Long live the speaker”

The work of historical and comparative linguists has yielded interesting and
illuminating results and can explain many things that are otherwise “strange”,
such as the “silent” k in Engl. knee, the differences in word order between English
and German, or the amazing similarity in the idiosyncratic pattern of Engl.
good : better : best, Germ. gut : besser : best, Icel. góður : betri : bezt-. But most
language users are neither historians nor linguists. In many cases, their behavior
goes counter to historical and comparative evidence.
For instance, linguists believe that the attempts of self-appointed critics to
stem the tide of linguistic change must ultimately be futile – linguistic change
takes place inexorably, whether we like it or not. And yet, the actions of critics
can have an effect.
A most remarkable example is found in Icelandic, a language which like
Chinese today greatly restricts the number of foreign words that it adopts. At an
earlier period, a lot of foreign words were employed quite freely, such as dedicera
from Latin dedicare ‘dedicate’, or borgmeistari from Germ. Bürgermeister ‘mayor’.
Linguistic critics in Iceland deplored such borrowings, arguing that they alienated
Icelanders from their own language and tradition. And, apparently because they
struck a sympathetic chord among most Icelanders, the critics won out: Words
such as dedicera and borgmeistari were eliminated from the language, and severe
restrictions were placed on the adoption of new words.
 A word of caution, or “Long live the speaker” 15

As a less “global” example, consider a process known as popular etymology

or folk etymology, which gives words a new, historically inaccurate etymology.
For instance, French andier ‘metal support for a fireplace grid’ was borrowed
into English first as andire (with various spelling variations). With substitution
of English for French pronunciation, this word can hardly be argued to be an
unacceptable word in English (acceptable words such as endeavo(u)r, Andrew,
Andover are very similar in structure). Nevertheless, for some reason andire must
have been felt to be “strange”, and various attempts were made to make it more
like a “recognizable” English word. The first attempt was the replacement of the
final -ire by -iron, no doubt because the object designated by the word is made of
metal. Further developments along the same lines have led to the less established
variants endiron (presumably because it is located at the two ends of the fireplace
grid) and handiron (probably simply because of the similarity of sound). As a con-
sequence of these processes, the word now has a shape that suggests a derivation
very different from the historically correct one.
Spelling pronunciations, such as the British “sounding out” of the initial h of
herb are historically similarly inaccurate. The word belongs to a set of borrowings
from French which includes hono(u)r and hour and whose written initial h origi-
nally was not pronounced, because French itself did not pronounce it. (The situ-
ation is similar to that of the k in knee: The spelling with h is an “echo” of the fact
that an h was once pronounced – in Latin.) But because the word is spelled with
an h, and because an h in initial position usually is sounded out (as in he, him,
hymn), some speakers pronounced the h of herb as well, and in British English
their pronunciation eventually prevailed.
In this case the spelling pronunciation is not completely inaccurate from a
historical perspective, since Latin, the ancestor of French, actually pronounced
an h here. In other cases, spelling pronunciation may lead to historically incorrect
results. For instance, during the Renaissance people in western Europe became
aware that many words borrowed from Latin with the letter t were ultimately from
Greek where they were written with the letter θ (theta), transcribed as th and,
in English, pronounced the same as the th of words like thing. This discovery
led to the reintroduction of th in many words, such as Dorothy and Thomas. In
some of these, spelling pronunciation then led to the introduction of the th sound
(Dorothy), while in others, the old t pronunciation remained (Thomas). In at least
one word, Anthony, British and American English went different ways: British
English has retained the pronunciation t, while American English has opted for
the spelling pronunciation th.
So far so good, especially since by the time that these spelling changes
occurred, the Greeks had come to pronounce θ as th, just as in English. However,
in the process some innocent bystanders were re-spelled and, eventually, re-pro-
16 Introduction

nounced as well – words which were of genuine Latin, not of Greek origin, and
which never had a θ in them. This is how the word author (Middle English autour,
from Lat. auctor via Fr. auteur) got its th spelling and pronunciation.
Historical linguists cannot – and should not – ignore such “ahistoric” behav-
ior. Speakers’ attitudes, whatever their historical or linguistic justification, play a
significant role in language change. Even if speakers of British English are made
aware of the fact that the initial h of herb is historically as incorrect as – heaven
forfend – sounding out the initial h of words like honour or hour, they will adopt
an h-less pronunciation for herb only at the risk of being considered Cockney by
their peers, and Americans would run into a similar problem if they pronounced
Anthony with t. Similarly, insisting on saying andire, instead of andiron, endiron,
or handiron, will hardly be appreciated as a more correct pronunciation; it is most
likely to be met with confusion, if not utter lack of comprehension.

5 A note on transcription and terminology

In this introductory chapter, words have been cited in the form they are usually
written (or transliterated), and phonetic interpretations have been given approx-
imately, using English spelling conventions. However, English spelling is noto-
riously “non-phonetic”: especially as far as vowels are concerned, there is little
consistency in the relation between spelling and pronunciation. English speakers
themselves are keenly aware of this and can cite expressions such as
Though the rough cough ploughed him through and through
as examples of the inconsistency of English spelling: Here, the sequence ough
designates five different pronunciations. Yet further phonetic variants of ough
spellings are found in hiccough and lough.
Consider further the case of the past tenses of the verbs read and lead: For
read, we have read, presumably to distinguish it from the color red but as a con-
sequence spelled the same way as the present-tense form. On the other hand, for
lead we have led, which is nicely differentiated from the present-tense form, as
well as from the name of the mineral, lead; but the latter is not distinguished from
the present-tense lead, even though sounding different.
George Bernard Shaw, an ardent supporter of spelling reform, is said to have
proposed the spelling ghoti for fish to demonstrate the vagaries of English spell-
ing, with gh as in rough or cough, o as in women, ti as in nation. This, of course, is
an exaggeration, since gh never has the value f in initial position, and ti does not
designate the sound of sh except before vowels; English spelling is not quite that
“crazy”. Still, as long as we are having fun, we can expand on Shaw’s ghoti by
 A note on transcription and terminology 17

proposing ghubtoti as the spelling for fetish, where the additional u has the value
as in bury, and bt as in debt, doubt, and subtle.
Even if we ignore such exaggerations, the fact remains that English spelling
is not an ideal way of transcribing speech sounds, especially of languages very
different from English. In the remainder of this book, a standardized method of
transcription is followed, using symbols widely used among linguists (especially
historical linguists). Readers not familiar with these symbols should consult the
Appendix to this Introduction, which also lists a few additional, non-phonetic
symbols used in this book.
While a familiarity with phonetics is essential for the study of most forms
of language, there is one class of languages for which it is not. These are sign
languages (also referred to as signed languages), found all over the world as a
means of communication between the deaf (see Chapter 16, § 8), and also used by
speakers of “oral” languages in communicating with deaf people. There are many
misconceptions about sign languages. But increasingly intensive research shows
no significant differences in structure, in complexity, or in expressivity between
sign languages and oral languages; and some aspects of linguistic change in sign
languages are beginning to be known. Where appropriate, we therefore add brief
remarks on sign languages to the discussion of linguistic change.
Most of the examples in this book come from members of the Indo-European
language family, including its ancestor, “Proto-Indo-European”. Of all the known
language families, this one has been most thoroughly researched and, in addi-
tion, is most familiar to the authors of this book. Chapter 2, following the Appen-
dix on Phonetics, provides an account of the discovery of Indo-European, infor-
mation on the reconstructed Proto-Indo-European language and the symbols and
terminology used to describe it and its early descendants, a brief overview of the
members of the Indo-European family, and a list of abbreviations for the names
of Indo-European languages. Chapter 3 surveys the origin and history of writing
and the decipherment of ancient scripts which give us access to early texts that
otherwise would have been lost to us. Our discussion of language change begins
with Chapter 4, on Sound Change, and continues in the subsequent chapters.
Appendix to Chapter 1:
Phonetics, phonetic symbols, and other
symbols
Scientific phonetics [is] the indispensable foundation
of all study of language … above all, of historical grammar.
(Henry Sweet, Collected Papers)

Henry Sweet was a nineteenth and early-twentieth century scholar of Old

English and speech science who founded the modern discipline of phonetics,
the scientific study of speech sounds. He reportedly was the model for Henry
Higgins, the male lead character in George Bernard Shaw’s play Pygmalion,
upon which the musical My Fair Lady is based. Just as Henry Higgins needed to
know about speech in order to “improve” the speech of his Cockney-speaking
subject, Eliza Doolittle, so we must, as Sweet suggests, understand the phonetic
basis of language, before we can get seriously into the examination of language
change.
Speech sounds are wave-like variations in air pressure produced by vocal
organs, especially the tongue, interfering with the flow of air from the lungs
through the vocal tract. In this regard they are not substantially different from the
sound made by slapping your hand on a table or breaking a stick, which also sets
air in motion and creates sound waves. However, while we perceive the slapping
or breaking noise as a single, undifferentiated event, we perceive speech sounds
as differentiated into discrete units. Thus, we hear a word such as peer as con-
sisting of three speech sounds, a p, an ee, and an r; similarly, we hear wreath as
consisting of a r, an ee, and a th-sound (as if it were spelled reeth).
But what do we mean by such symbols as p, ee, or th? Does p designate the
initial sound of peer, pieced, pill, etc., does it have the same value as in philosophy,
and what about the “silent letter” in pterodactyl or pneumonia? Does ee stand
for the vowel sound in Engl. peer, geese, feet etc., or for the vowel in German leer
(which sounds close to Engl. lair)? Which th-sound do we mean, the one in thing,
the one in this, or the one in Thomas or Thames? And how do we distinguish the
two similar, and yet different, sounds in thing and this?
These examples confirm what we saw at the end of the preceding chapter:
English spelling is not an ideal way of transcribing speech sounds. But even if
other languages may have more consistent spelling systems, these systems are
accurate only for the language for which they were designed. No ordinary spelling
system can be expected to be usable for all languages.
 A note on transcription and terminology 19

In linguistics, by contrast, we need a system that can be used for any lan-
guage. This is especially true for historical linguistics, which by its very nature
deals with the relationship between different languages or different stages of the
same language, each of which has its own phonetic and orthographic peculiari-
ties.
In addition, we need a firm grasp of the nature of speech sounds, such as the
question of the distinction between the two th-sounds in thing and this. Without
such an understanding we cannot understand sound change. Without under-
standing sound change we cannot establish that, say, OE hlāf and Mod. E loaf are
really different versions of the same word, not just words that happen to sound
similar. And without establishing such identifications across time and space, the
whole enterprise of historical and comparative linguistics comes to a grinding
halt.
In what follows, therefore, we develop a vocabulary for talking about the
nature of speech sounds, and introduce an inventory of symbols that we can
use to represent them. You will recognize many of the symbols from the English
alphabet, though some of the specific values we assign might be unfamiliar. Other
symbols are adaptations of familiar ones, often with “diacritical” marks above or
below the symbol. (Some of the symbols differ from the IPA system, but they are
well established in historical-comparative linguistics.)
If at this point you find the large number of phonetic terms and symbols quite
bewildering, you are in good company. Just about everybody who has studied lin-
guistics has had the same initial reaction. As you see the terms used throughout
this book, you will become more and more familiar with them, and at the end of
the book you should find, much to your surprise, that they have become second
nature.
Along the way, you may treat what follows as a source of reference which you
can return to whenever needed. Summaries of the terms and symbols are given in
Tables 1–3, which are placed at the end of this appendix for easy reference.

Analyzing speech sounds. As noted earlier, speech sounds are produced by

various speech organs obstructing the air that is being forced out of the lungs and
through the vocal tract. Differences between speech sounds result from differ-
ences in the type of obstruction (the manner of articulation), in the organ of
speech that makes the obstruction (the articulator), and the place in the vocal
tract where the obstruction is made (the place of articulation). Most important
among these are manner and place of articulation. Figure 1 provides a view of the
vocal tract, the major articulators and places of articulation, and the names used
to designate them.
20 Appendix to Chapter 1

Fig. 1: Major articulators and places of articulation

Stops and places of articulation. The most radical articulation, which is also
easiest to observe, consists of completely blocking the airflow at one of the places
of articulation. Sounds produced in this manner are called stops. Stops are classi-
fied by their place of articulation as labial (p, b), dental (t, d), palatal (č and ǰ,
similar to the initial and final sounds of Engl. church and judge), and velar (k and
g as in get). A glottal stop occurs marginally in emphatic speech (before initial
vowel, as in ʔoff with his head) or in dialectal British English for written t(t) in
words such as little, got. In Semitic and many other languages, the glottal stop is a
regular speech sound, and so is the uvular stop q. Yet other types of stop are found
in other languages. One of these will make an occasional appearance in the chapter
on writing; this is the Semitic “pharyngeal” stop, articulated somewhere between
the uvula and the glottis. Together with the glottal and uvular stop, the pharyngeal
stop forms a class of sounds that in Semitic linguistics is referred to as “gutturals”.

Nasals Although nasals like n and m sound very different from stops, they are
articulatorily closely related – as anyone who has had a cold or hay fever can
readily understand. Nasals essentially are oral stops of the type b, d, g, except that
the opening to the nasal passage is left open. This permits air to escape through
the nasal passage, producing a nasal resonance that turns the oral stop into a
nasal. If a cold or hay fever interferes with the passage of air through the nose, the
difference between oral b, d etc. and nasal m, n etc. disappears, and nasals seem
to turn into “dasals”.
 Appendix to Chapter 1 21

In addition to labial m and dental n, English has a velar nasal ŋ, often written
ng, which occurs in sing and sink.

Fricatives, sibilants, and voicing. A less radical obstruction than for stops and
nasals narrows the air passage sufficiently to produce a friction noise at the place
of articulation. Not surprisingly, the resulting sounds are called fricatives. The
English labial fricatives (f, v) are actually labiodental, articulated with the upper
teeth against the lower lip; fricatives involving both lips (“bilabial”) occur in
some languages, but the general tendency is for labial fricatives to be labiodental.
Dental fricatives are found in Engl. thing and this. The initial sound of this is tran-
scribed as ð; for the initial sound of thing two transcriptions are employed: θ and
þ. The transcription þ is traditional in talking about early Germanic languages, θ
is used elsewhere. A glottal fricative is the h of Engl. horse.
Closely allied to the fricatives are the sibilants (s, z, and the sounds in share
and measure, transcribed š and ž). Sibilants differ from ordinary fricatives by
being articulated with some additional friction noise. The exact location and
manner in which that hissing noise is produced may vary from speaker to speaker;
even the primary place of articulation may vary – as long as the acoustic effect
comes close to the expectations of other speakers of the language.
The sibilants s and z are ideal for beginners to understand something that we
have glossed over so far; that is the difference between such pairs as s and z, p
and b – or the initial θ and ð of Engl. thing and this which we talked about earlier.
The difference is one of “voicing”, which is present in sounds like z, b, and ð, and
absent in their counterparts s, p, and θ. Voicing is a kazoo-like effect produced
by the vocal folds (more commonly known as vocal cords), two membranes in
the glottis, deep down in the vocal tract and therefore difficult to observe. Fortu-
nately, the sibilants s and z provide an easy means to get around this difficulty:
Hold your fingers over your ears while you first articulate a z and then an s; you
should feel a strong buzzing vibration during the z which should be absent in s.
The source for the buzz in the voiced sound z is the kazoo-like vibration of the
vocal folds. Voiceless sounds like s are articulated without this vibration, with
the vocal folds at rest, in an open position.

Affricates and aspirates. Many languages have a set of complex sounds which
are called affricates. These begin as stops; but in contrast to ordinary stops, the
stoppage of the airflow is released into a fricative or sibilant that is produced at
the same place of articulation, as in German zu, pronounced [tsū].
Articulatorily, aspirated consonants can be considered a special type of
affricates, with the stop released into something like the glottal fricative h. In
English, voiceless stops tend to be aspirated, especially in initial position, before
22 Appendix to Chapter 1

an accented vowel; but aspiration does not serve to distinguish stops from each
other in the same way as voicing does. In many early Indo-European languages
aspiration plays a much more important role. For instance, Sanskrit makes a dis-
tinction between phala- ‘fruit’, pala- (a unit of weight), bhala (an interjection),
and bala- ‘strength’.

Liquids. If the obstruction of the airflow is reduced even more than for fricatives
and sibilants, the result will be various r- and l-sounds. Especially the r-sounds
exhibit a great amount of variation, both within and across languages. While
Scots English, Spanish, and many other languages have a trilled r, the r of Amer-
ican English is rather similar to w (with which it is often confused by children),
and many varieties of French, German, and other languages have a uvular or
velar R. But differences in r-sounds usually do not distinguish different words.
In most cases it will be sufficient to use a single symbol, r. There is less variation
in l-sounds, but some languages (including conservative Spanish) distinguish
between a dental and a palatal. Since l-sounds are articulated with “lateral” (i. e.,
side) contact of the tongue, they are commonly called laterals.
It is difficult to find a common articulatory or acoustic feature that would
define r-sounds and laterals as a group. But children often have great difficulties in
distinguishing them in the early stages of learning their first language; and many
languages have only one or the other of the two classes of sounds. To express this
affiliation between r-sounds and laterals, it is convenient to use the term liquid.

Semivowels, palatalized consonants, and labiovelars. The consonants with

the least amount of obstruction to the airflow are so close to the vowels that they
are often referred to as semivowels. English examples of semivowels are the
initial consonants of yes and woman.
Just as fricatives and sibilants can combine with stops to produce affricates
and aspirates, so semivowels can combine with stops and other consonants into
special types of speech sounds. A y-like element frequently combines with conso-
nants to produce palatalized consonants. The term “palatalized” is confusing,
because it sounds very similar to “palatal”. But in linguistic change, palatalized
consonants are closely related to palatals (see Chapter § 5.1.1); and the term “pala-
talized” is too deeply ingrained in linguistic terminology to be abandoned anyway.
A rarer combination is that of a w- like element with a velar consonant to form a
labiovelar. This, too, is a confusing term, since it might suggest that the lips
touch the velum, a physical impossibility even for the most acrobatically inclined.
The “labio” element merely refers to the fact that the lips are rounded while the
back of the tongue articulates against the velum. As we will see in Chapter 2, labi-
ovelar stops are a feature of Proto-Indo-European.
 Appendix to Chapter 1 23

A summary of the major consonant symbols that are needed for this book is
given in Table 1 with examples mainly drawn from English. Parentheses indicate
symbols that are used only rarely. The table does not include aspirates, palatal-
ized consonants, and labiovelars; these consonants are designated by diacritics
that modify the symbols in Table 1. See Table 3 for the most important of these
diacritics.

Vowels. The speech sounds examined up to this point are jointly referred to as
consonants. They differ in significant ways from the vowels (such as a, i, u):
– Vowels normally are the center, the “syllabic peak”, of the syllable (as in con-
sonant); consonants usually are not.
– Vowels are formed with the least amount of obstruction of the airflow; in fact,
the articulator does not even touch the place of articulation.
– Vowels further differ from consonants in that they are all produced in a very
limited space, the velar area; and in that small area, a large number of differ-
ent vowels can be articulated.

Vowels therefore have to be characterized by terms that are different from

those used for the consonants. The most basic classification is in terms of rela-
tive tongue position, on a front-to-back scale (front, central, and back), and
a high-to-low scale (high, mid, and low). See Table 2 for illustrations of these
distinctions.
In addition, many languages make a distinction in terms of “rounding”:
round vowels are articulated with lip rounding, unround vowels are not. The
general tendency is for front vowels to be unround (e. g. the vowel in Engl. sit)
and for back vowels to be rounded (as in Engl. soot); but languages like German
and French are well known for their front rounded vowels, often called “umlaut
vowels”, and the difficulties these sounds cause for speakers of languages like
English (which do not have them) are notorious.
The symbols used for the basic vowels are given in Table 2 with examples
mainly from English. Symbols in parentheses may occasionally be used for finer
distinctions; for most purposes the unparenthesized symbols are sufficient.
The low back vowels in the lower right corner of Table 2 may cause some
difficulties, for this is the area of the vowel spectrum where different varieties
of English show the greatest amount of variation. Many varieties of American
English, for instance, pronounce caught and law with the same vowel as father.
In the Midwest this tends to be the low central vowel [a]; but in other areas it may
be closer to [ɔ]. Such differences in pronunciation do not affect the use of our pho-
netic terms and symbols; on the contrary, it is through these terms and symbols
that we can characterize the differences.
24 Appendix to Chapter 1

Two of the vowels and the symbols used for them have special names; these
are the central vowels ǝ, called schwa (a term from Hebrew grammar which came
into English via German) and ɨ, called “barred i” (because the symbol ɨ has a “bar”
across it). The vowel ǝ occurs in most varieties of English corresponding to the
final “a” of sofa; ɨ is found in many varieties of American English in the pronun-
ciation of the adverb just (as in I just can’t wait till I get to find out more about how
languages change).

Diphthongs. Vowels can combine with each other or with semivowels to form
more complex “syllabic peaks”, as in Engl. my pronounced [may]. The name for
such complex structures is diphthong (literally, ‘double sound’).

Vowel length and nasalization. For vowels, we need to use diacritics much more
frequently than for consonants. In addition to relative tongue position and round-
ing, vowels often are distinguished in terms of length, as in Engl. sit vs. seat,
with short vs. long vowel. English long vowels tend to be slightly diphthongal
(e. g., seat may be transcribed as [siyt]); but in many other languages, length is the
major or only relevant distinction.
Vowels also may be nasalized by opening the passage to the nasal cavity, just
as for nasals like m and n. Nasalized vowels are a well-known feature of French;
but they are found in many other languages, including Portuguese and Hindi.
Length and nasalization are indicated by diacritic symbols modifying the basic
vowel symbols. These diacritics are given in Table 3.

Other diacritics and their phonetic values. In the course of examining conso-
nants and vowels we have introduced several diacritics (for aspiration, palatali-
zation, labiovelars; and for vowel length and nasalization). A few other diacritics
are used in this book.

Retroflex and alveolar consonants. Where English and most other European
languages have just one class of consonants articulated at or near the front teeth,
most languages of South Asia distinguish between two classes. One of these, the
dental class, is articulated at the same place as Engl. θ and ð, the other much
farther back than Engl. t and d, close to the area where Am. Engl. r is pronounced.
This second group is distinguished from the dentals by the name “retroflex”
(because the tongue is flexed backward in their articulation). Retroflex conso-
nants are marked by a subscript dot, as in ṭ and ḍ. Some South Asian languages
add a third class of consonants, right in between the dentals and retroflexes.
These are referred to as “alveolars” and are marked by a subscript line, as in ṯ
and ḏ. (English “dentals” actually are alveolars, too, being usually articulated just
 Appendix to Chapter 1 25

behind the dental area; but since we do not need to distinguish them from other
consonants in the same general area, no special symbols or terms are needed.)

Consonants marked by superscript diacritics. In transcriptions of Sanskrit you

may occasionally find the symbol ṁ (an m with superscript dot) which designates
a complex nasal “glide” or transition sound between a vowel and a following con-
sonant. Sanskrit examples also may contain the symbol ś (an s with superscript
accent mark); this is the conventional symbol to transcribe the palatal voiceless
sibilant of Sanskrit. Superscript accent marks are even more commonly used for
Proto-Indo-European reconstructions, where they indicate a class of “palatal-pre-
velar” consonants (see Chapter 2, § 2).

Syllabic nasals and liquids. At the beginning of the vowel section we observed
that consonants differ from vowels by usually not forming the “syllabic peak”.
Occasionally, consonants may do so anyway, especially the nasals and liquids. In
fact, English has such syllabic nasals and liquids in words like bottle, button,
bottom, but they are hidden by the spelling and, complicating things even more,
they may in super-careful speech be pronounced with a ǝ-vowel + non-syllabic
liquid, as in [bɔtǝl] or [batǝl]. Syllabic liquids are indicated by a subscript circle,
as in [bǝtn̥ ] = button; they are a prominent feature of Proto-Indo-European (see
Chapter 2, § 2).

Accent or stress: In many languages, words of more than one syllable contain
one syllable that is more prominent than the rest, in terms of loudness, pitch, or
some other feature. This syllable is conventionally referred to as bearing accent
or stress. Where necessary in this book, the accented syllable is designated by a
superscript acute accent mark, as in áccent or, in phonetic transcription, [ǽksent].
Stress can be the sole phonetic characteristic distinguishing between two words,
as in Engl. áccent (noun) vs. accént beside áccent (verb).

Conclusion. We now have the tools to describe and transcribe the various sounds
of different languages that you will encounter in the rest of the book. In many
cases we simply cite words in the form they are usually written (or transliterated).
This is especially the case if the actual pronunciation of a word or phrase is not
at issue; but many writing and transliteration systems are close enough to our
transcription that a phonetic interpretation is not necessary even when talking
about sound change. Where phonetic interpretations are crucial, though, famili-
arity with our phonetic terminology and phonetic alphabet will be indispensable,
as Sweet would put it.
26 Appendix to Chapter 1

Before we can begin to apply our terminology and phonetic alphabet in exam-
ining language change, however, we need to become familiar with a few addi-
tional special symbols that are employed in this book. These are given in Table 4.

Tab. 1: Consonants

Labial Dental Palatal Velar Uvular Glottal

Stops vl1 p t č k (q3) ʔ

pet Tom church kid bottle4
vd2 b d ǰ g
bet dumb judge go

Fricatives vl. f (Φ5) θ/þ6 (ç) x (χ3) h

fat thorn Germ.ich Scots.loch hand
vd. v (β5) ð (j̵) ɣ (ɦ)
vat that Span. lago behold
(Span. debe)

Sibilants vl. s š
see she
vd. z ž
zeal measure

Affricates7 vl. pf ts (kx)

Germ. Pferd Germ. Katze

Liquids r, l (λ) (R)

red, lead (conservative Span. ella) (Germ. rechts)

Nasals m n ñ ŋ
mow no Span. señor sing

Semivowels w y
woo you

[Notes: 1voiceless; 2voiced; 3used in many non-Indo-European languages; 4in Scots English and
many British urban dialects; 5bilabial fricatives); 6þ is traditionally used in Germanic, θ else-
where; 7in languages with a contrast between affricates and stop + fricative, the fricative element
may be raised to avoid confusion.
General note: In addition to velar, uvular, and glottal stops and fricatives, other back consonants
are possible; the Semitic languages, for instance, have a set of stops and fricatives (called gut-
turals) articulated between the uvula and the glottis, including ʕ (stop) and ħ or ḥ (fricative); in
addition, note ḫ = [x].]
 Appendix to Chapter 1 27

Tab. 2: Vowels

Front Central Back

unround 1
round round unround2

High i (I3) ü ɨ u (U3) (ı)

seat, sit Fr.rue (just [adverb]) boot, foot

Mid e (ɛ3) ö ǝ o
bate, bet Fr. feu sofa boat

Low æ a ɔ
bat father (caught, law)

[Notes: 1same position as i, e, but lips rounded; 2same position as u, but lips unrounded; 3these
symbols are used occasionally to indicate slightly lower vowels, as in [sit] = sit vs. [sīt] = seat.]

Tab. 3: Diacritics

Aspiration: ph, bh, etc. (In languages with contrast between aspirates and stop + h, the h-ele-
ment of aspirates may be raised to avoid confusion.]
Palatalization and labiovelars: ty, dy, etc. and kw, gw, etc.
“Dottings” and other diacritics: The languages of South Asia have a contrast between pure
dental consonants (articulated at the same location as Engl.[θ]) and “retroflex” consonants (for
which the tongue tip points back, roughly as in Am. Engl. [r]). The latter consonants are distin-
guished from the former by subscript dots, as in ṭ, ḍ, ṇ. (In the Semitic languages, subscript
dots indicate “emphatic consonants”.) The Dravidian languages of Southern India distinguish
yet another, intermediate series, called alveolar, which is transcribed by means of subscript
hyphens, as in ṯ, ḏ, ṉ. For Sanskrit, an ṁ with superscript dot is used to indicate a special nasal
glide. Accent marks over consonants indicate palatal or “palatal/prevelar” pronunciation, as in
Sanskrit ś, or Proto-Indo-European ḱ, ǵ, etc.
Nasalized vowels: ã, õ, etc. (Fr. en [ã] ‘in’ etc.)
Long vs. short vowels: ī vs. ĭ, ā vs. ă, ū vs. ŭ, etc. (roughly as in seat vs. sit, etc.). (Vowel short-
ness is indicated only when necessary; ordinarily, short vowels are left unmarked.)
“Syllabic” nasals/liquids: l̥, r̥ , n̥ , m̥ (bottle, button, bottom)
(Syllabic liquids and nasals behave like vowels in constituting the most prominent part of a
syllable. In English, these sounds may in very slow, careful pronunciation be pronounced with a
ǝ-vowel + nonsyllabic liquid, as in [bɔtǝl]. See also Chapter 2, § 2.)
Accent or stress: indicated by an acute accent mark over the vowel in the stressed syllable, e. g.
áccent or, in phonetic transcription, [ǽksent].
28 Appendix to Chapter 1

Tab. 4: Other special symbols

“Arrows”: To indicate sound changes, unshafted arrows are used. For instance, a > b means
“a changes to b”; and b < a means “b results from a”.
For analogical replacements, shafted arrows are used (→ and ←).
To indicate borrowings, double-shafted arrows are used (⇒ and ⇐).
Asterisks: A preposed asterisk indicates a reconstructed form (one not actually attested but
hypothesized to have existed – see Chapter 16); a postposed asterisk designates any other
hypothetical form.
Brackets: Where it is necessary to distinguish phonetic transcriptions from customary spelling,
the transcription is placed in square brackets, [ ]; angle brackets, < >, may be used to focus on
the written form, as distinct from its phonetic value.
Chapter 2: The discovery of Indo-European
The Sanscrit language, whatever be its antiquity, is of a wonderful structure; more perfect than the Greek, more
copious than the Latin, and more exquisitely refined than either, yet bearing to both of them a stronger affinity, both
in the roots of verbs and in the forms of grammar, than could possibly have been produced by accident; so strong
indeed, that no philologer could examine them all three, without believing them to have sprung from some common
source, which, perhaps, no longer exists: there is a similar reason, though not quite so forcible, for supposing
that both the Gothick and the Celtick, though blended with a very different idiom, had the same origin with the
Sanscrit; and the old Persian might be added to the same family, if this were the place for discussing any question
concerning the antiquities of Persia.
(Sir William Jones, Third Anniversary Discourse, on the Hindus, Royal Asiatic Society, 1786)

1 Language relationship
The preceding remarks, made by Sir William Jones at the close of the eighteenth
century, present a turning point in our understanding of historical and compar-
ative linguistics in general. Scholars were especially impressed by Jones’s sug-
gestion that the similarities between certain languages of India and Europe are
too great to be attributed to chance and can only be explained by assuming that
the languages are related by descent from a common ancestor. This suggestion
stimulated a veritable explosion of research, resulting in an increasing confidence
that not only the disparate and far-flung languages mentioned by Jones, but many
others as well (see § 3 below) are members of a family of closely related languages
and descended from a common, unattested ancestor.
In the early nineteenth century, the family of languages came to be called
Indo-European, because its known members extended from India in the east to
Europe in the west; and the common ancestor is now referred to as Proto-Indo-Eu-
ropean. (Since then, another member of the family – Tocharian – has been discov-
ered in Chinese Turkestan, farther east than the northwestern part of India where
Sanskrit was first spoken, but the name “Indo-European” has stuck.)
The idea that languages might be related to each other through descent from a
common ancestor was not entirely without precedent. The similarities among the
Romance languages could be, and were, easily explained as resulting from their
descent from a known common ancestor, Latin. (One of the earliest to propose
such an account was the famous poet and scholar Dante.) Note for example the
similarities in the word correspondences below:

https://doi.org/10.1515/9783110613285-002
30 The discovery of Indo-European

Latin Italian Spanish French Romanian

‘one’ ūnus uno un un [ö̃] un
‘two’ duo due dos deux [dö] doi
‘three’ trēs tre tres trois [tRwa] trei
‘fish’ piscis pesce pez poisson peşte
‘heart’ cor cuore corazón cœur (înimă1)
‘winter’ hiberno inverno invierno hiver iarnă
(1The Romanian word for ‘heart’ comes from a different source: Lat. anima ‘soul’. Cases like
this, where one or more related languages exhibit a non-cognate word, are not unusual – or
unexpected – given the fact that vocabulary, too, is subject to change.)

The situation is similar for the medieval and modern languages of northern India
whose structure and vocabulary were routinely derived by indigenous grammar-
ians from their attested common ancestor, Sanskrit. The following correspond-
ences for the numerals ‘one’ to ‘three’ may illustrate the ease with which a rela-
tionship can be established.

Sanskrit Bengali Hindi Marathi Gujarati Kashmiri

‘one’ ēkaḥ ēk ēk ēk ek akh
‘two’ dvau dui dō dōn be zɨʔ
‘three’ trayaḥ tin tīn tīn t(r)ǝṇ trɨʔ
neut. trīṇi

However, attempts at longer-range comparisons, going beyond easily recognized

families like Romance, were less successful. In Europe, it was commonly assumed
that all languages ultimately were descended from Hebrew, since that, presuma-
bly, was the language spoken by Adam and Eve in the Garden of Eden. Similarly in
India, where Sanskrit had exerted a strong influence on the Dravidian languages
of the south, it was assumed that these languages were descended from Sanskrit.
We now know that both of these interpretations are erroneous. And that knowl-
edge is founded on the vastly improved understanding of historical and compar-
ative linguistics that was stimulated by Jones’s pioneering remarks.
Although Jones also made some very wrong guesses about language relation-
ships, his comments came at an opportune moment in history, when earlier Euro-
pean attitudes, limited by religious, social, and cultural isolation, were giving way
to new modes of thinking.
During the medieval period western Europe scholars had become linguisti-
cally parochial, limiting their attention almost exclusively to Latin. But, although
Latin was the common currency of all scholars in western Europe, the form of
the language that was used had become far removed from the Classical Latin of
 Language relationship 31

Caesar and Cicero – indirectly mirroring the great distance between medieval and
Classical Latin thought and life.
In the late thirteenth century, with the Renaissance (lit. ‘rebirth’) and its great
reawakening of interest in the literary and philosophical traditions of Classical
Latin antiquity, scholars attempted to reform Latin usage by making it conform
more closely to the language of the great classical authors. But in the process,
a profound change in linguistic thinking occurred. Scholars like Dante realized
that there were vast differences between the Latin language that was learned in
schools and the Italian, Spanish, French etc. vernaculars that children learned
in early childhood. They began to account for the differences by determining the
linguistic changes that differentiated the vernaculars from the ancestral Latin lan-
guage. Moreover, they began to advocate an increased use of the vernaculars. As
a consequence, grammars of Italian and Spanish were published in the fifteenth
century, to be followed by grammars of French, Polish, and other languages in
the sixteenth century.
The Renaissance emphasis of returning ad fontes ‘to the sources [of western
European culture]’ had implications far beyond linguistic scholarship. Religious
reformers like Luther, Zwingli, and Calvin attempted to return to the early Chris-
tian idea that the Gospel should be preached in the language of the people and
therefore began to translate the Bible into the vernacular languages. The resulting
translations expanded the cultivation of western European vernacular languages;
and this cultivation, in turn, encouraged closer examination of the earlier, medi-
eval forms of these languages.
The Renaissance received an unexpected boost from the Ottoman Turks’ con-
quest of the Byzantine empire. Greek scholars began to flee to western Europe;
and an early trickle of westward migration turned to a flood with the fall of Con-
stantinople in 1453. The result was a vast increase in awareness of classical Greek
antiquity and its vehicle, Classical Greek. And Greek was added to Latin as a lan-
guage to be studied in school by anyone who wanted to be considered properly
educated.
The Ottoman conquest of Constantinople also had other, more indirect,
effects on western linguistic awareness. It seriously interrupted the trade routes
with India and China, sources of highly treasured silks and spices. Portugal there-
fore began to look for alternative routes around the coast of Africa. And in 1492,
Columbus set sail for what he believed to be India, to discover what turned out
to be an entirely “new world” – the Americas. The voyages by the Portuguese
and Spanish explorers, and their later rivals from France, the Netherlands, and
the British Isles, set in motion not only an expansion of European power across
the world but also the subjugation, even destruction, of non-European societies.
(Some of the linguistic effects of these developments are taken up in Chapter 14.)
32 The discovery of Indo-European

The travels of European explorers and missionaries also led to an increase in

familiarity with languages beyond Western Europe’s cultural horizon.
One of the results was an ever expanding set of comparative word lists, with
at least some attempts at establishing families of related languages on the basis
of vocabulary similarities.
Some of these attempts were rather naive from our present perspective. For
instance, one scholar classified languages in terms of their words for ‘god’, group-
ing Latin and Greek closely together because of their correspondence deus : theós,
and Slavic with Iranian because of bogŭ : baga.
At first glance, these correspondences appear to be excellent, both in form
and in meaning. But after more than a century of working out the sound corre-
spondences in inherited words, we know that the similarities in deus : theós are
accidental. Lat. deus is from the root underlying Gk. Zeus (genitive Dios), and if
Greek theós had a good Latin cognate, it would have to have an initial f. The sim-
ilarity of bogŭ : baga results from secondary contact between Slavic and Iranian
and does not reflect close linguistic relationship. In fact, Iranian is most closely
related to Sanskrit, and that language has dēva, a fine cognate of Lat. deus.
Some attempts at establishing relationship, however, were remarkably
mature, especially those by Sajnovics (1770) and Gyármathi (1799) to relate Hun-
garian to Finnish.
In the remainder of this chapter we will take a closer look at Indo-European,
since this is the most thoroughly investigated language family, as well as the
source for most of the examples in this book. Brief discussions of other language
families are found in Chapter 16.

2 Proto-Indo-European
What makes it possible to rule out correspondences like Lat. deus : Greek theós
is the fact that Jones’s famous “discovery” of 1786 brought about a sea change in
the way scholars think about language relationship. Jones inspired a dramatic
outburst in comparative research on the Indo-European languages; and, most
important, that research soon went beyond the stage of superficial comparison
of individual words and began to examine recurrent and systematic correspond-
ences in hundreds, even thousands, of words, including words like the following:
 Proto-Indo-European 33

Sanskrit Greek Latin Old English Old Church Slavic

‘one’ ēkaḥ (heis) ūnus ān edinĭ
‘two’ dvau duō duo twā dŭva
‘three’ trayaḥ treis trēs þrīe trĭje
‘that’ tad to (is-)tud þæt to
‘father’ pitā patēr pater fæder (otĭcŭ)
‘foot’ pad- pod- ped- fōt (noga)
‘blood, gore’ krūra- kreas cruor [k-] hrēaw krŭvĭ
(As in the earlier Romance example, some languages display non-cognate words for particu-
lar lexical items. These are put in parentheses. The parenthesized is- of Lat. istud is an added
element that originally reinforced the demonstrative force of tud.)

One of the most important developments was a discovery by the Danish scholar
Rask (1818) which was popularized by the German linguist Grimm (1819) and
therefore came to be known as Grimm’s Law. As Rask observed, the consonant
system of the Germanic languages differed from the systems of most of the other
Indo-European languages in a manner that was remarkably systematic. For
instance, where other languages have p, t, k early Germanic offers the fricatives
f, þ, h, as in the boldface consonants in the above examples. (These systematic
correspondences are discussed in further detail in Chapter 4.)
The regularities summed up by Grimm’s Law encouraged scholars to look
for similar regularities in other correspondences between various Indo-European
languages. As a result of this work historical linguists were able to tackle another
task with much greater confidence and vastly improved results, namely the com-
parative reconstruction of the Proto-Indo-European parent language.
The language thus reconstructed had a very rich system of consonants,
which by the end of the nineteenth century was classified as follows.

Labial Dental Palatal Velar Labiovelar

voiceless p t ḱ k kw
voiceless aspirated ph th ḱh kh kwh
voiced b d ǵ g gw
voiced aspirated bh dh ǵh gh gwh
fricative/sibilant s
nasals m n
liquids r, l

In addition, there were the semivowels y and w.

The vowel system is less “exotic”. With the exception of ǝ, all vowels come in
long and short varieties:
34 The discovery of Indo-European

i/ī u/ū
e/ē ǝ o/ō
a/ā

In addition to the vowels, Proto-Indo-European had a set of syllabic nasals and

liquids (also called syllabic resonants or sonorants), variants of consonantal
(nonsyllabic) nasals and liquids which, like the vowels, were the center or syllabic
peak of their respective syllables. These are commonly written as follows:
l̥, r̥ , n̥ , m̥

The relationship between syllabic and nonsyllabic sonorants (such as m̥ : m)

closely mirrors the relationship between syllabic i and u and their nonsyllabic
semivowel counterparts y and w, respectively. For instance, the Sanskrit root kr̥ -
‘do, make’ appears with syllabic sonorant before a consonant (as in kr̥ -ta- ‘done’),
but with nonsyllabic sonorant before a vowel (as in kr-āṇa- ‘doing’), just as prati-
‘in return, re-’ appears with syllabic i before a consonant (prati-vadati ‘replies’),
but with nonsyllabic y before a vowel (praty-avadat ‘replied’).
Proto-Indo-European was reconstructed with a fairly rich system of noun and
verb inflection. In the nouns, eight cases were distinguished with the following
major functions. Examples are given from Sanskrit. (Adjectives and pronouns had
essentially the same inflectional system.)

Nominative: Subject case (dēvas ‘god’)

Accusative: Direct object case (dēvam)
Instrumental: ‘With’ or ‘by’ case, i. e., accompaniment, instrument, agent
of passive (dēvēna)
Dative: Indirect object, beneficiary of an action (dēvāya)
Ablative: ‘From’ case, i. e., source (dēvāt)
Genitive: ‘Of’ case: possession, modification of other nouns (dēvasya)
Locative: Location in place and time (dēvē)
Vocative: Address case, used to address or call persons (dēva)

In addition to case, nouns were inflected for number, distinguishing not only sin-
gular and plural, but also a “dual” (for pairs of persons or things, such as dēvau
‘two gods’). Nouns also distinguished three “genders”: masculine, feminine, and
neuter. As in modern German and French, and unlike modern English, there was
no straightforward relationship between biological sex and grammatical gender.
True, most nouns for males and females were masculine and feminine respec-
tively, but some some exceptions are found. Many nouns for inanimate things
could be masculine, feminine, or neuter.
 The Indo-European languages 35

Verbs were inflected for person (“first”, “second”, and “third”) and number
(“singular”, “dual”, and “plural”). Verb inflection further distinguished between
different “tenses”, “moods”, and “voices”. In the “tense” system, we can distin-
guish between a present and three formations that tend to have past tense value,
the “imperfect”, the “aorist”, and the “perfect”. Corresponding to our future,
Proto-Indo-European used the present tense, modal formations such as the sub-
junctive, or formations indicating a desire to do something. The following mood
distinctions were postulated for Proto-Indo-European: “indicative” (unmarked),
“imperative” (for commands), “optative” (‘should, could, might do something’),
and “subjunctive” (‘shall, can be expected to do something’). In voice, there was
a distinction between active and “medio-passive”, the latter expressing a range of
functions, including “reflexive” (‘hurt oneself’) and passive (‘is hurt’). All of these
reconstructed categories are based on correspondences between the different
Indo-European languages, in just as compelling a way as a sound corresponence
such as t : t : t : þ : t leads to a reconstruction *t.

3 The Indo-European languages

The focus of Jones’s famous statement was the three classical languages – Greek,
Latin, and Sanskrit. This should not come as a surprise. As we have seen earlier,
the rediscovery of Greek in western Europe had led to the inclusion of Greek,
beside Latin, as a language to be studied by anyone aspiring to an education.
And Jones, a judge in the administrative system of the British East India Company,
certainly was an educated man. He added Sanskrit to the other two classical lan-
guages in the course of his duties as judge in India. What made him and other
Britishers do so was a set of fortuitous events and decisions quite unconnected
with language or linguistics.
In the eighteenth century, Britain began to establish its Indian colonial
empire by setting up the East India Company for trading purposes in the area
around Calcutta. From the start, the British took political control of the terri-
tory that they dominated by adopting the administrative system of the Mughal
empire, which once had held sway over most of India but was now in decline. The
Mughals were Muslims, and like all traditional Islamic rulers believed that Islamic
law applied only to Muslims, and that non-Muslims had to be governed by their
own laws. In India, where most non-Muslims are Hindus, this meant that legal
disputes between Hindus had to be settled by their own law. And formal Hindu
law was handed down in Sanskrit. Fortunately for comparative linguistics, when
the British took over the Mughal administrative system, they also took over the
36 The discovery of Indo-European

pattern of separate codes of justice – Islamic law for Muslims, and Hindu, i. e. San-
skrit, law for Hindus. But wanting to be able to administer these laws effectively,
they felt it necessary to acquire first-hand knowledge of the languages in which
the laws were handed down – Arabic for Muslim law, Sanskrit for Hindu law. In
addition, Jones and his fellow administrators had to learn Persian, since this was
the administrative language of the Mughals.
It is to Jones’s great credit that he was not satisfied with learning Sanskrit
and Persian merely for administrative reasons, but that he drew on his classical
education, compared the structures of the three classical languages that he had
mastered (Greek, Latin, and Sanskrit), and came to the conclusion that their sim-
ilarities point to descent from a common ancestor. More than that, he suggested
that even some modern languages might have to be added to this family: Persian,
Celtic, and “Gothick” – a term by which he probably meant Germanic, the family
that includes English, German, as well as Gothic.
Jones’s proposal was adopted by the German philosopher, poet, and philolo-
gist Friedrich Schlegel, and with Schlegel’s modifications it set in motion a linguis-
tic field that provided substantial proof for Jones’s hunch, added more languages
to the mix than Jones was willing to consider (see § 3 below), and established the
relationship of a vast and far-flung group of languages ranging from India in the
east to Europe in the west. (As often happens in the history of the sciences, the
field did not always develop in a straight line. Schlegel’s proposal, for instance,
that the ancestor of Greek, Latin, and Sanskrit was not “some … source, which,
perhaps, no longer exists” but rather Sanskrit itself set the field back for the better
part of the 19th century.)
Since Jones’s and Schlegel’s time, comparative linguists have established
beyond any doubt that Persian, Celtic, and Germanic are in fact relatives of Greek,
Latin, and Sanskrit, as are many other languages, some of which were discovered
or deciphered much later. The remainder of this section provides a brief outline
of the major languages that have been shown to belong to the Indo-European lan-
guage family, of their history, and – where appropriate – of the manner in which
they were discovered.
The map below gives an approximate indication of the geographic location
of Indo-European languages at an early period (ca. 1000 BC). The map is only
approximate and does not account for prehistoric or historic migrations. Thus,
the British Isles are shown as Celtic territory, since Celtic is the Indo-European
language for which we have the earliest attestations there. The later arrival of the
Germanic Anglo-Saxons is not accounted for. Similarly, Indo-Aryan is placed into
the northwest portion of South Asia. Its later expansion to the east and south is
ignored.
 The Indo-European languages 37

Approximate geographical distribution of Indo-European languages (ca. 1000 BC)

3.1. Celtic. At an early period, the westernmost group of the Indo-European lan-
guages consisted of the Celtic languages (William Jones’s “Celtick”). While today
confined to the western periphery of Europe and the British isles, the Celts once
ranged over a vast territory. The names Bavaria, Belgrade, Bohemia, London, and
Louvain are all considered Celtic in origin. In late prehistoric times, Bohemia and
all of Southern Germany were Celtic. It was apparently from this home base that
Celtic peoples expanded westward into Gaul (today’s France) and Belgium where
they had long been settled when Julius Caesar conducted his famous campaign
to conquer Gaul. Celts likewise had settled in northern Spain and the British Isles.
In all of this vast territory, the Celts appear to have been the vanguard, as it were,
of Indo-European westward expansion. They made great contributions in metal-
lurgy; for instance the Germanic words for ‘iron’ were borrowed from Celtic, no
doubt together with the art of using the metal. Their cultural importance can be
further gauged by the fact that much of the early Germanic naming system was
borrowed from Celtic, and so was the word for kingdom.
The movement of the Celts was not limited to the west. In the fourth century
BC, Celts crossed the Alps, settled in northern Italy, laid siege to Rome, and exacted
tribute from the city of Rome. (When the Romans complained about the large
amount of booty they had to put up to “pay off” the invaders, the Celtic leader
Brennus is said to have uttered Vae victis ‘woe to the defeated’, thrown his sword
38 The discovery of Indo-European

on the scale, and demanded that the weight of the sword be added to the gold and
treasures already on the scale.) A century later, the Celts invaded the Balkan pen-
insula, thrust southward into Greece, and threatened the sanctuary of the famous
oracle at Delphi. Some even settled in Anatolia, where they were known as Gala-
tians, the people addressed in one of Paul’s epistles in the New Testament.
By the time of late antiquity, the power of the Celts had greatly diminished,
partly perhaps because they had overextended themselves, but no doubt also
because in much of western Europe they had become integrated into the Roman
empire and adopted Roman customs and Latin speech. Only a small number of
Celtic inscriptions are found on the continent – in upper Italy, Gaul, and neigh-
boring parts of Spain, plus a few inscriptions in the Balkans. These inscriptions,
written in the Greek alphabet, the Roman alphabet, or other, regional off-shoots
of the Greek writing system, date from about the fourth century BC to the third
century AD.
The richest attestations come from the “Insular Celtic” of the British Isles: Old
Irish in Ireland and spreading into Scotland (attested from ca. 400 AD), Welsh in
Wales (ca. 8th c. AD), and Cornish (now extinct) in Cornwall (9th c. AD). Breton
(9th c. AD), now spoken in Britanny, on the western coast of France, originally was
an insular Celtic dialect as well, whose speakers fled the British Isles to escape the
onslaught of the Anglo-Saxons. Old Irish is the ancestor both of modern Irish and
Scots Gaelic, the two surviving members of the “Goidelic” branch of Insular Celtic.
Welsh, Breton, and the extinct Cornish form the “British” or “Brythonic” branch.
The earliest Old Irish texts consist of short inscriptions on memorial stones,
giving the genitive form of the name of the person in whose memory the stone
was erected. They tell us a fair amount about early Irish names and the phonetics
of Old Irish, and more than we need to know about the case endings of the geni-
tive. Beyond that they are interesting mainly for the special, and quite unusual,
script in which they were written, the so-called Ogham characters (see Chapter 3).
Later Insular Celtic texts (including those of Breton) are all in Roman script. The
modern Irish script of Ireland is a regional, “Insular” modification of the Roman
script which was used also for medieval English.
Although Cornish, as noted, is now extinct, there have been recent attempts
to revive it. Modern Irish, Scots Gaelic, Welsh, and Breton have been able to main-
tain their spoken tradition, but centuries of discrimination in favor of the official
or unofficial state language (English or French) have led to a great reduction in the
number of speakers. To varying degrees, these languages are all now considered
to be endangered.
There is some uncertainty whether Pictish, originally spoken in the area of
today’s Scotland and replaced early on by Gaelic, is part of the Celtic languages,
part of the larger Indo-European language family, or even Indo-European at all.
 The Indo-European languages 39

3.2. Italic (Latin). To the south and east of the original position of the Celts is
the group of Italic languages. The most richly attested of these is Latin, whose
earliest attestations, brief inscriptions, begin about the sixth century BC. (Liter-
ary texts start to appear in the third century BC.) Originally the speech of a small
area around Rome, Latin became the dominant language of western Europe,
first through conquest and domination of the rest of Italy, followed by the far-
flung military conquests of the Roman Empire, and eventually the spiritual con-
quest by Roman Christianity of Europe from the Baltic to the Mediterranean, and
from Poland and Hungary to Ireland and Iceland. The Roman Empire’s policy
of encouraging conquered peoples to acquire Roman citizenship by adopting
Roman customs, including the Latin language, led to the adoption of Latin as the
language first of Italy, and then of most of the rest of Southern Europe, as well as
parts of the Balkans.
The spoken Latin that spread in this way, slightly different from the written
Latin of Caesar and Cicero, became the source of the modern Romance languages,
including Portuguese, Spanish, Catalan, French, Italian, Romantsch (the fourth
official language of Switzerland, beside Italian, French, and German), as well as
Romanian. Romanian originally was connected with the rest of the Romance lan-
guages through Dalmatian, spoken in present-day Croatia and Bosnia-Hercego-
vina; but the language eventually was replaced by varieties of Slavic, as well as
by Italian. The last speaker of Dalmatian is reported to have been killed in 1898,
in a mine explosion.
Three of the Romance languages spread beyond their original territories and
became world languages as a result of the “age of discovery” and the subsequent
period of western imperialism. Spanish and Portuguese have large numbers of
native speakers outside the Iberian peninsula, especially in Latin America. French
has become the official state language of a number of former colonies, particu-
larly in Africa; it also is one of Canada’s two official languages and the dominant
language in the province of Quebec.
Although the Romance languages clearly developed out of spoken Latin by
undergoing divergent linguistic changes, we cannot trace these developments
directly. Latin remained the written medium for many centuries after the Romance
languages began to develop. The earliest texts clearly written in Romance come
from the ninth century AD and are limited to the French area.
Latin, in fact, for a long time was the dominant written language throughout
all of western Europe, not just in Romance territory. More than that, it also was
the dominant spoken language of the educated. As a consequence, a scholar, say,
from Finland, could attend a university in France and communicate with other
scholars entirely in Latin, without having to learn the local language. Especially
in the Romance area, the coexistence of spoken Latin with the emerging Romance
40 The discovery of Indo-European

vernaculars had profound effects on the vernaculars, especially in their vocab-

ulary, where forms inherited from Latin, such as Spanish dueño ‘lord’ (< Latin
dominus) coexist with borrowings from Latin, such as dominar ‘to dominate’.
Before Latin spread throughout Italy, it coexisted with several other “Italic”
languages. The two best-known among these are Oscan (from about the fourth
century BC) and Umbrian (third century BC). Both of these, attested in relatively
short inscriptions, were replaced by Latin in about the first century BC.
Ancient Italy was home to a number of other languages. Some of these are
Indo-European, but not part of Italic; others are of controversial origin. None of
these is attested sufficiently well to be of great help for comparative or historical
linguistics. Perhaps the most fascinating of these languages is Etruscan. At an
early period, the Etruscans held much vaster territory and had a much higher
degree of civilization than the Romans. In fact, they are commonly believed to
have been the major source for the early transmission of ancient Greek culture –
including the art of writing – to the Roman world. They have left behind more
than nine thousand inscriptions, in a script which can be perfectly read, in the
sense that we understand the phonetic values of its letters. But like the Old Irish
Ogham inscriptions, most of the texts are short and simply contain names which
do not tell us very much about the structure and vocabulary of the language; and
the relationship of Etruscan to other languages remains a mystery.

3.3. Germanic. As the Celts moved westward, they were closely followed by the
Germanic peoples, who settled Southern Germany and Austria. At the dawn of
history Germanic peoples were found throughout most of present-day Germany
and Austria, present-day Belgium and the Netherlands, as well as Scandinavia. But
in Scandinavia they remained confined to the more southern and western areas.
During the early historic period, the Germanic peoples rivaled – perhaps even
surpassed – the Celts in mobility and in the vast area over which they range. The
Goths, the Burgundians, and the Vandals, originating in the Scandinavian area,
migrated as far east as the Caspian Sea, temporarily made common cause with
the Iranian tribe of the Alans and with the dreaded Huns, then turned west again,
to besiege and – temporarily – conquer parts of the Roman Empire. Ostro-Goths
(‘East Goths’) established a short-lived empire in Italy; Visi-Goths (‘West Goths’)
ruled over parts of Spain; the Burgundians eventually settled in a part of France
which even today bears their name – Burgundy; and the notorious Vandals, after
sacking Rome and wreaking havoc all over the empire, established a kingdom
in North Africa. In the meantime, a southern expansion brought the Alemannic
tribes into Switzerland, and the Frankish tribes into present-day France, whose
name is derived from that of the Franks. Although they had help from such
non-Germanic peoples as the Alans and Huns, the migrating Germanic tribes
 The Indo-European languages 41

bore much of the responsibility for the collapse of the Western Roman Empire.
Ironically, it was a Germanic ruler over the French – Charlemagne – who, having
been crowned emperor by the Pope of Rome, reestablished the Western Roman
Empire. And as ruler over much of Germany, he incorporated that country into the
Empire – something which none of the emperors of the original Roman Empire
had been able to accomplish.
The extended invasions of the Germanic peoples left a rich legacy of bor-
rowings in the Romance languages. Not surprisingly, these borrowings include
martial terms such as Span. guerra, Fr. guerre ‘war’ from *werra- (compare Engl.
war) or Fr. gonfalon ‘battle flag’ from *gund-fano. Many personal names were
borrowed, too, presumably because of the prestige or power of Germanic rulers.
Examples are Fr. Henri, Charles, Span. Enrique, Carlos which correspond to the
native Germanic names reflected in Germ. Heinrich, Karl. (The English versions
of the names, Henry and Charles, do not continue the original Germanic forms of
the names; they are borrowings from French that came to England in 1066 and
reflect the power and prestige of the Norman invaders. History, in a way, repeated
itself; and the names were “recycled” to a Germanic language.) Curiously, the
borrowings from Germanic also included the word for ‘soap’: Fr. savon, Old Span.
xabon, Mod. Span. jabón.
These great migrations – or barbarian invasions, depending on your perspec-
tive – generally did not lead to a geographic expansion of Germanic languages;
but many other, slightly later migrations did. One of these was the migration of
the Angles, Saxons, and Jutes into England under the leadership of two chief-
tains whose names have come to us as Hengest and Horsa. (The names of these
two gentlemen can be roughly translated as Stallion and Stud.) Pushing back the
indigenous Celtic population into the marginal areas of Wales and Cornwall, they
settled in the southern part of the island up to Northumberland.
In the meantime, the Scandinavians, too, began to migrate. They settled
Iceland and, as Vikings and Varangians, roamed far and wide. They probably
sailed to the North American continent. They certainly laid siege to and plun-
dered many a Mediterranean harbor and kingdom, including Sicily. They invaded
and settled a part of France which came to be named Normandy in their honor,
soon adopting a regional variant of French as their language. Pursuing an eastern
route, the Swedish Varangians took up trade with the Iranians in Southern Russia
and established the first royal dynasty of Russia, the house of Rus. The quintes-
sentially “Russian” names Oleg and Olga in reality reflect Varangian domination;
they are derived from earlier forms of the modern Scandinavian names Helge and
Helga. Finally, known as “Danes”, Scandinavians from Denmark, Norway, and
Iceland engaged in extended warfare with the Anglo-Saxons, furnished some of
the Anglo-Saxons’ rulers (e. g. Canute), and settled down in the so-called Danelaw,
42 The discovery of Indo-European

an extensive strip on the east coast of England, stretching from Yorkshire down to
the vicinity of London. After settling and intermarrying with the Anglo-Saxons,
the Danes exerted considerable influence on the development of the English lan-
guage, especially in terms of extensive contributions to the vocabulary, including
such words as give, get, take; skin, skirt, sky, egg, and even the pronouns they, their,
them.
Soon after the Danish threat had abated, the Vikings that had settled in Nor-
mandy and had, as we have seen, become Francophone, conquered England in
1066 and left a lasting mark on the English language. Words such as beef, pork,
mutton; court, royal, justice, and even the second part of the function word because
owe their origin to French influence. The immense French influence on English
vocabulary may also have set the precedent for a tradition which treasures a huge
vocabulary and, consequently, encourages continued borrowing. Compared to
most other languages, English seems to be almost voracious in adopting words
from all the languages of the world.
Like French, Spanish, and Portuguese, English became an international lan-
guage as the result of colonialism and imperialism. The 400 (+) million native
speakers in North America and the Caribbean, Australia, New Zealand, and South
Africa vastly outnumber the 60 (+) million speakers in the United Kingdom plus
4 million in the Republic of Ireland. But in addition, English is used as a fluent
means of communication by about another 500 million speakers in former British
and American colonies around the world, including India, Singapore, the Phil-
ippines, Kenya, and Nigeria. The impact of this immense spread of English on
the languages of the world – as well as on English itself – is only beginning to be
researched and understood. New standard varieties of English – the “New Eng-
lishes” – are emerging, such as South Asian English which is used by millions of
South Asians to communicate with each other, not just with native speakers of
English. In the process, English is being indigenized, in vocabulary, structure,
and pronunciation. We will take a closer look at some of these developments in
Chapter 12.
Developments like these could hardly have been foreseen when the Ger-
manic languages first appeared on the historical scene. The earliest attestations
of Germanic come in brief inscriptions beginning about the first century AD. The
inscriptions are written in a special offshoot of the alphabets used in Greece and
Rome, the runes. (For the script, see Chapter 3, § 2.6.) The language of these texts
is virtually identical to the reconstructed Proto-Germanic ancestor.
The oldest extensive text is a Gothic Bible translation produced by the Gothic
bishop Wulfilas (‘Little Wolf’) in the fourth century, which has come down to us
in fragmentary form. Manuscripts have been found in Germany, Italy, and even
Egypt; and new finds are still being made. The earliest discovery, made in the
 The Indo-European languages 43

Ruhr area of Germany, attracted the attention of sixteenth-century German and

Dutch scholars who, in the spirit of the Renaissance, studied the manuscript in
their endeavor to unravel the history of their language. Because of its early attes-
tation and the relative richness of the text, the Gothic Bible translation has ever
since been a valuable source of information on early Germanic sound and word
structure. But because it often closely followed the word order of the Greek orig-
inal, it is less useful in the area of syntax. Most varieties of Gothic seem to have
died out quite early; but in the sixteenth century, the Dutch traveler van Busbecq
brought back a few vocabulary specimens of a remnant of Gothic from the Crimea,
in southern Ukraine. We do not know if this last remnant of the language survived
much longer. No modern descendants of Gothic are known.
The earliest documents of the other Germanic languages appear still later
than the older runic inscriptions and Gothic: seventh century AD for Old English,
thirteenth century for Old Frisian (spoken in the coastal areas of present-day
Netherlands and Germany), tenth century for Old Low Frankish (the ancestor of
modern Dutch and Flemish), ninth century for Old Saxon (the ancestor of “Low”
German in northern Germany), eighth century for Old High German (in the south-
ern, mountainous area of Germany, hence the term “High” German), twelfth
century for Old Norse (mainly Old Icelandic, with embedded verse passages that
go back to about the ninth century AD).
The major modern Germanic languages are English (see above); German (the
official language of Germany, Austria, Liechtenstein, and parts of Switzerland and
Belgium); Dutch (in the Netherlands) and the closely related Flemish (in Belgium),
as well as a colonial offshoot, Afrikaans, spoken in South Africa; the Scandinavian
languages Danish, Norwegian, and Swedish; and Icelandic and Faroese.

3.4. Slavic. The Slavic languages, to the east of Germanic, now cover a vast terri-
tory, especially if we include Russian, the lingua franca of the former Soviet Union
and the major language of the Russian Federation. Prehistorically, the languages
occupied a smaller region. Recall that Bohemia originally was Celtic territory. The
place names of much of the Russian core area, from St. Petersburg to Moscow,
have been claimed to be Baltic in origin. Most of present-day Russia is colonial
territory, with a large variety of non-Indo-European indigenous languages, many
of which belong to the Uralic and Altaic language families.
From around the fifth century AD, Slavic speakers begin to push into the
Balkan peninsula, engaging in protracted warfare with the Eastern Roman Empire
(whose capital was Constantinople). The word slave, a phonetic variant of Slav,
bears testimony to the warfare. In ancient times, prisoners of war, if permitted to
live, became slaves. The fact that for a long period in the history of the East Roman
Empire, most prisoners of war were Slavs made it possible to reinterpret the word
44 The discovery of Indo-European

Slav as meaning ‘slave’ and to use it in this new meaning even when referring to
prisoners from other ethnic groups.
The situation in the northwestern Slavic area is especially complicated.
Over the centuries, large portions of Poland and eastern Germany alternated
in seesaw fashion between Slavic and German. Even today, two Slavic speech
islands, Upper and Lower Sorbian, are located in eastern Germany. Many places
in Germany (almost as far west as Kiel and Hamburg) bear names that are of
Slavic origin, including the cities of Berlin, Leipzig, and Dresden. On the other
hand, the city of Gdańsk in present-day Poland was for many centuries, up to
1945, essentially a German city, Danzig. The name Danzig, in turn, is a German-
ization of the Polish name of the city, Gdańsk, which clearly is older, suggesting
that the city originally was Polish. But the story doesn’t end there. A yet earlier
form of the name, Gŭdansk-, contains Guda-, which has been plausibly connected
with Guta-, the name the Goths used for themselves. We can therefore conclude
that this was once Gothic, i. e. Germanic, territory. But, again, this does not pre-
clude the possibility that at an even earlier time, the area had been Slavic or even
Baltic.
Some evidence suggests that Slavic had strong prehistoric contacts with
Iranian in present-day southern Russia. Note for instance the shared word for
‘god’: Old Persian baga : Old Church Slavic bogŭ. It is possible, therefore, that the
original homeland of the Slavs was close to southern Russia. But given the mobil-
ity of the Slavs in historical and even prehistoric times, we cannot be certain.
The literary attestation of Slavic begins in the ninth century AD with a Bible
translation commissioned by the ruler of Moravia in what is the present-day
Czech Republic and produced by two brothers, Cyril and Methodius, who hailed
from Thessalonica, a city in what is now northern Greece. Although the local
Slavic then spoken in Thessalonica must already have differed from the Slavic
of Moravia, it was apparently close enough for the two brothers to use their own
variety of the language. But while in this respect they can be said to have compro-
mised, they did not compromise on another issue. Faced with considerable lin-
guistic differences between Slavic and the original Greek text, they devised a new
writing system, based on the Greek alphabet, but with letters added to accurately
transcribe the sounds of their native language. (The “Cyrillic” writing system, for
which see Chapter 3, § 2.6, was a later development.)
Ironically, Moravia, for which the Bible translation had been produced,
became part of Western Roman Christianity and consequently accepted the Latin
Bible translation, as well as the Latin (or Roman) alphabet. Cyril and Methodius’s
translation came to be adopted by Eastern Orthodox Slavic Christianity, and the
“Old Church Slavic” of their translation became a language of liturgy and learn-
ing comparable to Latin in western Christianity. And like Latin in the Romance
 The Indo-European languages 45

area, Old Church Slavic was used as the major written language in much of South
and East Slavic, masking the developments taking place in the regional forms of
Slavic.
Although exhibiting dialectal features belonging to South Slavic, the Old
Church Slavic texts come fairly close to the ancestral, Proto-Slavic language. Like
Gothic, they are therefore extremely useful for comparative linguistic research.
But, again like Gothic, they often closely follow the word order of the Greek origi-
nal and therefore are not as helpful as far as syntax is concerned.
The Slavic languages are usually divided into three groups: West Slavic (Czech,
Slovak, Polish, Sorbian), South Slavic (Bulgarian, Macedonian, Slovenian, and
the languages formerly known as Serbo-Croatian and now distinguished along
national lines as Bosnian, Croatian, Montenegrin, and Serbian), and East Slavic
(Belorussian, Ukrainian, Russian, Ruthenian). What used to be called Serbo-Cro-
atian is one of several languages which are “divided against themselves” because
of religious and ethnic differences between their speakers. Although its varie-
ties have virtually identical grammars, they are used by different communities:
Serbian by Eastern Orthodox Christians, Croatian by Roman Catholics, Bosnian
by Muslims. The difference in religion is partly mirrored by differences in script;
e. g., Serbians use Cyrillic, Croats the Roman alphabet. Where Serbians tend to
be fairly open to borrowings from other languages, Croats prefer to create new
words from their own native resources. (On this issue see Chapter 8, § 5.2.) While
the government of the former Yugoslavia emphasized the essential unity of Ser-
bo-Croatian and linguists who are not emotionally involved would agree, nation-
alists dwell on the differences of Serbian, Croatian, Montenegrin, and Bosnian,
and consider these varieties fundamentally different languages. At present, the
nationalists have the upper hand, demonstrating the importance of speakers’ atti-
tudes which we talked about in Chapter 1.

3.5. Baltic. The Baltic languages include Lithuanian (from 16th c. AD), Latvian
(16th c.), and the now extinct Old Prussian (ca. 14th c.). They were originally
spoken over a much larger territory than today. During the historical period,
much of Baltic gave way to Slavic, and a relentless campaign of forced German-
ization led to the demise of Old Prussian in the seventeenth century. Ironically,
the names Prussia and Prussian were taken over by the Germans and through a
series of events came to be associated with an important regional and dynastic
element in German politics. (Such secondary associations of traditional names
are not unusual. Compare for instance the case of Frankish : French and Bur-
gundian : Burgundy. Similarly, modern Greeks use the term Romeikos, literally
‘Roman’, to refer to themselves and their language, echoing the fact that Greece
once was the center of the Eastern Roman Empire. See also Chapter 9, § 3.1.)
46 The discovery of Indo-European

Though attested relatively late, Baltic has been rather conservative in certain
aspects of its linguistic structure, especially in the area of noun inflection. Thus,
of the original eight cases reconstructed for Proto-Indo-European (see § 2 above),
the early classical languages Sanskrit, Latin, and Greek preserved eight, six, and
five, respectively. And as we have seen earlier, modern languages like French and
English have reduced the number of cases even more. Modern Lithuanian and
Latvian preserve seven of the original eight cases. Similarly, among the classi-
cal languages, the Proto-Indo-European distinction singular : dual : plural was
fully preserved only in Sanskrit; Greek showed the dual in moribund form; and
Latin had lost it as a grammatical category. Modern European languages such as
English, German, and French likewise have no trace of the dual. Again, Lithua-
nian preserves the dual number. In spite of their relatively late attestations, the
Baltic languages thus provide valuable information for comparative Indo-Euro-
pean linguistics.
This does not mean that Baltic is archaic in every respect. For instance, the
Indo-European triple gender distinction masculine : feminine : neuter has in Lith-
uanian and Latvian been reduced to one between masculine and feminine; only
Old Prussian preserved the neuter gender. Even the case system is not entirely
archaic, for Lithuanian and Latvian actually added three cases to the inherited
system. The development of these cases, an “illative” (specifying the direction into
something), an “allative” (the direction toward something), and an “adessive”
(the location near something) represents a structural convergence with the neigh-
boring Uralic languages (such as Estonian) which have an abundance of such
locational and directional cases. Structural convergence of this sort is not unusual
under conditions of extensive bilingualism (see Chapter 13). One suspects, there-
fore, that the Baltic Indo-European and Uralic languages have been in prolonged
bilingual contact.
There has been a continuing controversy as to whether Baltic and Slavic form
a special, “Balto-Slavic” subgroup of Indo-European which underwent enough
common developments in their prehistory to set them off from the rest of the
Indo-European languages. To some extent the debate has been fueled by nation-
alism. Slavic scholars and their sympathizers tend to argue for “Balto-Slavic”,
with the implicit assumption, perhaps, that Slavic was the senior partner in this
relationship. Baltic scholars and their sympathizers, who are fearful of Slavic
attempts at domination, tend to emphatically reject the relationship. Such intru-
sion of personal prejudice into linguistics unfortunately is not as rare as it should
be. But if backed up by evidence, prejudice does not necessarily invalidate the
force of an argument.
In the case of Baltic and Slavic, strong arguments have been mustered by
both sides. It may well be that many of the similarities shared by Baltic and Slavic
 The Indo-European languages 47

reflect not just a period of common prehistory, but the fact that they were neigh-
bors from Proto-Indo-European times to the present and thus kept influencing
each other for millennia, both in structure and in vocabulary.

3.6. Albanian. To the south of Baltic and Slavic, on the west coast of the Balkan
peninsula in present-day Albania and neighboring areas of Greece, Kosovo, Mac-
edonia, and Montenegro, we find Albanian, attested very late (from the 15th
century AD). It had been subject for many centuries to the influences of neigh-
boring languages, including Greek, Latin and its descendants, as well as South
Slavic. As a result, a large proportion of its vocabulary is of foreign origin, though
a sufficient core of indigenous vocabulary, inherited from Proto-Indo-European,
remains for comparative linguistic work.
Not much is known about the historical antecedents of Albanian or whether it
has any close relatives within the Indo-European language family. According to a
traditional view, modern Albanian is descended from Illyrian, a language spoken
in ancient times in the north of modern Albania. But what little is known about
Illyrian argues against close affiliation with Albanian.
Another language mentioned as a possible ancestor of Albanian is Thracian,
the language of ancient Thrace in the southeastern part of the Balkan peninsula.
Although some linguistic evidence may support this affiliation, the geographic
separation between Thrace and Albania causes difficulties. Most important,
however, like many other early Indo-European languages of the Balkans, includ-
ing Illyrian, Thracian is attested much too sparsely to permit successful compar-
ative linguistic work.

3.7. Greek. The Greek language, in the extreme south of the Balkan peninsula,
was until recently believed to have been first recorded about 800 BC. Some time
prior to that date, alphabetic writing had been developed from Semitic sources in
Asia Minor (see Chapter 3, § 2.5). The introduction of the alphabet was a techno-
logical innovation which some scholars believe made it possible for the Homeric
epics, the Iliad and the Odyssey, to be given a more permanent written codifi-
cation, even though they had been successfully handed down for centuries in
an oral tradition (see Chapter 3, § 2.1 on the oral transmission of texts), and the
earliest manuscripts of the epics date from much later. Still, the introduction of
the alphabet was a well-documented historical event, so it was generally believed
that no Greek texts older than the ninth century BC would ever be found.
All of this changed dramatically in 1952, when it was discovered that the
non-alphabetic “Linear B” script of Bronze-Age Mycenaean times was used to
write an early form of Greek, called Mycenaean Greek (see Chapter 3, § 3.3). This
discovery pushes back our knowledge of Greek to a time between about 1400 and
48 The discovery of Indo-European

the twelfth century BC, a period when some of the linguistic changes that differen-
tiate Greek from the rest of the Indo-European languages had not yet taken place.
The decipherment of Linear B has made it possible to confirm certain hypotheses
about these changes and to disconfirm others, in addition to raising new issues
which still await full resolution.
The exact reasons for the end of Mycenaean literacy, which necessitated a
“reinvention” of literacy in the ninth century, are still under dispute. A fair amount
of evidence suggests that Mycenaean society collapsed under the onslaught of the
so-called Sea People, who also wreaked havoc on Egypt and other areas of the
mainland. But the identity of the Sea People is uncertain. Another possibility is
that the invaders were the Doric or “West Greek” tribes. Both Greek tradition and
the geographic distribution of these tribes (see Chapter 11, § 5) suggest that they
were late intruders. But instead of destroying the Mycenaean civilization, they
may have simply filled the void left behind by the Sea People. In short, we really
do not know what caused the downfall of the civilization.
In historical times, Greek is characterized by a strong differentiation of dia-
lects, each associated with one of the many city states of ancient Greece and jeal-
ously guarding its identity. The major early Greek dialects are: Attic and Ionic
(closely related), Arcadian and Cypriot (also closely related), Aeolic (with several
subdialects), and the Doric or “West Greek” dialects. Attic (the speech of Athens)
and Ionic were the major literary languages of classical Greek society. Doric, with
many regional dialects, covered a large area of Greece. Of the many dialects of
Doric, those of Corinth and Sparta were politically important. Arcadian, Cypriot,
and the Aeolic dialects were politically and, to a large degree, geographically
more marginal.
From the time of the Persian wars, a variant form of Attic, with some influ-
ence from Ionic and other dialects, began to emerge. As the “Koiné” (from Gk.
hē koinē glôssa ‘the common language’), it became the common language of the
Greek empire established by Alexander the Great and eventually replaced virtu-
ally all the ancient dialects. (See Chapter 12, § 5.) As a consequence, nearly all the
modern Greek dialects are descended from the Koiné, though one dialect, Tsako-
nian, spoken in the interior of the Peloponnesus, is believed to have developed
directly from Laconian (an ancient Doric dialect).

3.8. Anatolian. Anatolia, to the east of Greece, is the home of a large number
of ancient languages, many of which were written in the cuneiform (“wedge-
shaped”) script of ancient Mesopotamia. In the early part of the twentieth century
it was shown that one of these languages, Hittite, is Indo-European, even though
in its structure and vocabulary it differs considerably from the other early attested
Indo-European languages. (See Chapter 3, § 3.1.) Its oldest texts come from the
 The Indo-European languages 49

seventeenth century BC and are among the earliest texts in any Indo-European
language; the latest texts date from about 1200 BC.
Beside Hittite, several other, fairly closely related languages have been found
in Anatolia, including Palaic and Luwian (roughly contemporary with Hittite),
Lydian (6th–4th c. BC), and Lycian (5th–4th c. BC). Due to their geographical loca-
tion, Hittite and its relatives are referred to as Anatolian.
Because they are attested so early, Hittite and the other early Anatolian
languages could be expected to yield important information for comparative
Indo-European linguistics. To some extent, this expectation has not gone unful-
filled, especially for certain issues in sound and word structure. But many other
aspects of Anatolian linguistic structure have created more problems than they
have solved. For instance, Hittite has a verb ‘to have’, whereas the comparative
evidence of the other Indo-European languages shows that Proto-Indo-European
expressed the notion ‘I have no money’ by something like ‘Of me (there) is no
money’ or ‘For me (there) is no money’. Although in the case of ‘have’, Anato-
lian most certainly innovated, in other areas the rest of Indo-European may have
been more innovative. The problem is that Indo-Europeanists do not yet agree on
the interpretation of many of the differences between Anatolian and the rest of
Indo-European.

3.9. Armenian. Yet farther to the east, in the Caucasus region, we find Armenian.
Though attested from a rather early period (5th c. AD), it is much less archaic in
vocabulary and structure than Baltic and Slavic. Partly this is due to strong pre-
historic influence from Iranian. In fact, the large amount of Iranian words taken
into Armenian, such as hazar ‘1000’ : Old Iranian *hazahra, Mod. Pers. hazār,
led early researchers to consider Armenian an Iranian dialect. We now know
that Armenian is historically quite distinct from Iranian, or any other attested
Indo-European language group, for that matter. Another important source for the
“changed” character of Armenian may be convergence with non-Indo-European,
Caucasic languages, such as Georgian.
Recent hypotheses about the nature of Proto-Indo-European dispute the tra-
ditional view that the structure of Armenian underwent profound changes that
differentiated it from the rest of the Indo-European languages. Instead, some
of the special characteristics of Armenian, such as the existence of a series of
“glottalized” consonants (with glottal-stop coarticulation), are considered highly
archaic, direct inheritances from Proto-Indo-European. This “glottalic” view of
Indo-European is still a matter of controversy (see Chapter 16, § 7). One of the dif-
ficulties with the assumption that the glottalized consonants of Armenian are an
archaism is that these consonants can also be explained as a regional innovation,
a convergence of Armenian with the neighboring non-Indo-European languages.
50 The discovery of Indo-European

In this regard, note that an Iranian language, Ossetic, displays the same “glot-
talic” sound system as Armenian and most of the other languages of the Cauca-
sus; but in the case of Ossetic, glottalized consonants clearly are a late, regional
development.
Old Armenian texts begin in the fifth century AD and are composed in a liter-
ary language that combines elements from a number of different dialects. It was
written in a special script said to be devised by the Christian priest Mesrop. Some
scholars believe the script was developed out of an earlier northern Iranian script
with some influence from the Greek alphabet; others think that it is overwhelm-
ingly of Greek origin. As in the case of Gothic and Old Church Slavic, the earliest
text was a Bible translation. Like Old Church Slavic, the language of that transla-
tion soon became a literary standard, and this “Classical Armenian” remained in
use into the nineteenth century.
During the twelfth century, a Middle Armenian language acquired currency
at the court of Cilicia. During this period, a major sound change began to dif-
ferentiate Western and Eastern Armenian, which has given rise to two modern
standard languages, West and East Armenian. East Armenian is the language of
the Republic of Armenia, while West Armenian is used by Armenians hailing from
Turkey who, as the result of persecution, are now largely dispersed in countries
such as Syria, Lebanon, Egypt, and even the United States. Armenian enclaves
are also found in Iran. The division between the East and West Armenian lit-
erary languages masks a much more profound diversity of local and regional
dialects.

3.10. Indo-Iranian. Iranian and Indo-Aryan, at the southeastern periphery

of the attested Indo-European languages, though distinct subgroups, are more
closely related to each other than any other subgroups of Indo-European. The
Indo-Iranian subgroup is therefore the most uncontroversial subdivision of the
Indo-European family. The close relationship between Iranian and Indo-Aryan is
manifested by a large number of common innovations in phonetic structure, as
well as in vocabulary. For instance, the fairly rich vowel system of Proto-Indo-Eu-
ropean given in § 2 above has been reduced to an early Indo-Iranian system with
just three vowel distinctions: ī̆ : ā̆ : ū̆. More than that, Iranians and Indo-Aryans
even agreed on a name that they used to refer to themselves – ārya. (The name
Iran originally is the genitive plural of this word, designating the country as the
land ‘of the Aryans’.) In the nineteenth century the term Aryan was extended to
designate all Indo-Europeans, under the false assumption that the Irish word
Eire is cognate with ārya; and ill-founded theories about the racial identity of
these Aryans were propounded, which ultimately were used to justify the racist
excesses of the Nazis (see Chapter 18). For these reasons, the term Aryan fell into
 The Indo-European languages 51

disfavor for a while, and Indo-Aryan was commonly referred to as Indic (German
indisch), especially among German scholars. The preference of Indian scholars for
the name Indo-Aryan eventually led to the readoption of the term.

3.10.1. Iranian. The two major Old Iranian languages are Avestan and Old Persian.
Of these, Avestan is attested in much earlier – and much more archaic – texts,
dating back to at least the seventh century BC. These are the sacred writings of
Zoroastrianism, hymns composed by the founder of that religion, Zarathushtra,
handed down orally for a long time, and put into written form quite late by persons
no longer fully competent in the language. (The oldest extant manuscripts come
from about the thirteenth century.) Unlike Old Persian, Avestan seems to have
come from a more eastern part of Iranian, close to Indo-Aryan. Because of its early
attestation and, perhaps, because of its greater proximity to Indo-Aryan, Avestan
is very close to the oldest form of Indo-Aryan, Vedic Sanskrit (see below). In fact,
the two languages are so similar that it is possible to change an entire Avestan
hymn into an acceptable Vedic hymn by merely adjusting the pronunciation.
Some general information about Zoroastrianism was available from Greek
sources, which also are responsible for the form of the name Zoroaster that came
to be known to the West. But the texts of the religion were unknown. It was only
in the seventeenth and eighteenth centuries that a few manuscripts came into
western hands, unfortunately in a script that nobody could decipher. A young
Frenchman, Anquetil du Perron, was deeply impressed by these manuscripts and
determined to learn to read them. To do so, he joined the French military and
sailed to India in 1754. Taking leave he set out on his own to Bombay, where he
contacted Zoroastrian priests who, like other Zoroastrians or “Parsis”, had fled
Persia at the time of the Muslim conquest. The priests at first regarded him with
great suspicion, but after repeated efforts on his part to learn from them, they
eventually opened up, teaching him their rituals, as well as the language and
script of their sacred texts. Armed with some manuscripts that they had given
him, he returned to France in 1761 and busied himself with the study of the texts
and their language. In 1771 he finally published a three-volume translation of the
Avestan texts. Although many of his readings turned out to be flawed, his work
opened the way to an understanding of this important ancient Iranian language.
Subsequent research, informed by a fuller understanding of related Iranian lan-
guages and of Sanskrit, has greatly improved our understanding of Avestan so
that, unlike William Jones, we are now able to include Avestan among the most
ancient Indo-European languages.
Unlike Avestan, Old Persian can be dated confidently, being attested in rock
inscriptions by the great Persian kings (Darius I, Xerxes I, and Artaxerxes II and
III) who lived in the sixth to fourth centuries BC. The inscriptions had been noticed
52 The discovery of Indo-European

for a long time by travelers, but the cuneiform script of the inscriptions could not
be read until its nineteenth-century decipherment (see Chapter 3, § 3.1). Although
not as archaic as Avestan, the language of the inscriptions is sufficiently old to
establish beyond any doubt what William Jones had only considered a possibility,
namely that Persian is related to Sanskrit, Greek, Latin, and the other Indo-Euro-
pean languages.
At an early period, Iranian dialects were spoken over a vast territory, ranging
from Iran to the Hindukush, at some times even into what is now the Xinjiang
province of China, and into the steppes of Southern Russia. Iranian tribes in
Southern Russia, commonly known as Scythians, are probably responsible for
the early Iranian borrowings in Slavic mentioned in § 3.4.
Modern Iranian languages include Modern Persian (or Farsi), Kurdish,
Pashto (in Afghanistan), and Ossetic (in the Caucasus Mountains). Like Arme-
nian, Ossetic has a structure that bears great similarities to the structures of the
non-Indo-European languages of the Caucasus. In the case of Ossetic, it is clear
that these similarities result from secondary convergence. (See also § 3.9.)

3.10.2. Indo-Aryan (Sanskrit). Early Indo-Aryan is represented by Sanskrit,

whose earliest texts, the Vedas, go back at least to about 1500 BC. Like the Avestan
texts, the Vedas were transmitted orally; but their oral transmission continued
just about to the present day. Thanks to an elaborate “back-up” system (including
a highly developed formal grammar), the texts have come down to us with only
minor changes.
Classical Sanskrit, attested in texts from about the fifth century BC to the
present, was mainly a language of the educated. From an early period it coexisted
with Prakrit, a later form of Indo-Aryan that was the language of the common
people. This coexistence is mirrored in a literary convention of Classical Sanskrit
drama, where Sanskrit is reserved for the educated protagonists (mainly male),
while Prakrit is used by the uneducated (including most women).
At an early time, Indo-Aryan appears to have been limited to the extreme
northwest of South Asia. Settlement of the Ganges valley, especially of the more
eastern area of Bengal, is believed to have taken place only slowly, over the course
of several centuries. In historical times, we can see an even farther expansion
of Sanskrit, both in India (where it becomes the sacred language and, for many
centuries, the literary language even of the Dravidian south) and outside. A
major vehicle for the spread of Sanskrit outside India was Mahayana Buddhism
which expanded, via Central Asia, as far east as China and Japan. An early form
of Prakrit, Pali, experienced a more moderate spread to Sri Lanka and parts of
Southeast Asia as the vehicle of Theravada Buddhism. Unlike Sanskrit, however,
it did not survive on the Indian mainland.
 The Indo-European languages 53

As Sanskrit, and Indo-Aryan in general, spread within South Asia, its speakers
came into contact with many speakers of non-Indo-European languages. In India,
the two most important groups of such languages are Dravidian and Munda. The
Dravidian literary languages, Tamil, Malayalam, Kannada, and Telugu, are all
spoken in the south of the subcontinent. But “tribal” Dravidian languages are
found as far north as the central mountain range of India and even in present-day
Pakistan. The Munda languages are part of an Austro-Asiatic family, which
includes members like Mon and Khmer in Southeast Asia. All Munda speakers
belong to “tribal” societies and, like Dravidian – and Indo-Aryan – “tribals”, live
in relatively inaccessible areas in the central mountains.
There is some controversy over when the Indo-Aryans first came into contact
with Dravidian and Munda speakers. Some linguists argue for prehistoric contact
between Indo-Aryan and Dravidian, attributing features of Indo-Aryan structure
to convergence with Dravidian. Others have proposed that Indo-Aryan speak-
ers first came into contact with Munda speakers. One scholar has postulated an
unknown northwestern language which supposedly influenced Indo-Aryan. The
available evidence may simply be too limited to decide between these hypotheses.
(See also Chapter 13.) There is no doubt, however, that Indo-Aryan, Dravidian,
and Munda eventually all came to structurally converge with each other through
multilingual contact extending over several millennia.
Like Latin in the western world, Sanskrit functioned as a scholarly lingua
franca of India long after the emergence of regional languages that had descended
from it. And like Latin, Sanskrit has slowly been losing ground. Unlike Latin, San-
skrit has remained in spoken form to the present day. (In fact, after hearing two
American colleagues gossip in Sanskrit at a scholarly meeting, H. H. Hock went
to do research in India on spoken Sanskrit and learned to speak Sanskrit, too.)
During the past two decades, however, the spoken use of Sanskrit has decreased
so dramatically that it may become extinct within a generation or two.
The most widely spoken modern Indo-Aryan language is Hindi-Urdu. Like the
former Serbo-Croatian, it is a language that is “divided against itself” because of
religious and political differences. Hindi is used by Hindus and other non-Mus-
lims; it is also one of the two official link languages of the Republic of India. Urdu
today is primarily the language of Muslims, and the official state language of the
Islamic nation of Pakistan. Hindi draws on Sanskrit sources for its religious and
cultural terminology, as well as to translate English technical terminology. Urdu
uses Persian and Arabic sources for the same purposes.
The two forms of language differ most markedly in their highly literary and
intellectual varieties. There is virtually no difference in their everyday forms of
use. During the Indian independence struggle, leaders like Gandhi advocated the
use of a common, non-sectarian “Hindustani” language, based on these everyday
54 The discovery of Indo-European

varieties. The subsequent political separation of British India into Islamic Paki-
stan and secular India led to an increased polarization between Urdu and Hindi,
putting an end to attempts at promoting Hindustani. (See also Chapter 8, § 5 and
Chapter 12, § 1.)
Other modern Indo-Aryan languages are Bengali, Gujarati, Marathi, Panjabi,
as well as Sinhala (in Sri Lanka). Yet another modern Indo-Aryan language is
Romani, the language of the “Gypsies” or Dom, an Indo-Aryan tribe that migrated
in the Middle Ages from central India via northern India, first to Central Asia, from
there to much of Asia and the Balkans, and then to virtually all of Europe.

3.10.3. Indo-Iranians in the ancient Near East – the Mitanni. Documents from
Anatolia and other parts of the ancient Near East (ca. 15th c. BC), containing
names and other words of Indo-Iranian origin, show that an Indo-Iranian group
(often called the Mitanni) had migrated to this area of the world. Among these are
passages in a treatise which suggest that the Indo-Iranians brought with them an
improved method of horse training that may have added to the military prowess
of the Hittites. Some of the words contained in these passages are phonetically
closer to the earliest attested Indo-Aryan than to Old Iranian. It has therefore been
suggested that the people in question really were Indo-Aryans, not Iranians or
speakers of an as yet undivided Indo-Iranian. Assuming that they indeed were
Indo-Aryans, a number of questions arise: How did these Indo-Aryans get to the
Near East? Were they just an isolated tribe that strayed from the India-bound path
of the rest of Indo-Aryan? Or did perhaps all of the Indo-Aryans meander through
the Ancient Near East before settling in India? Although the last-mentioned sce-
nario is rather unlikely, we have no hard evidence to answer these questions.

3.11. Tocharian. The easternmost branch of Indo-European, if we disregard the

later eastward expansion of Indo-Aryan, consists of a group of two closely related
languages, Tocharian A and B. Documents in these languages (going back to
the sixth to eighth centuries AD) were found in Chinese Turkestan in the late
nineteenth and early twentieth century, and the languages were recognized as
Indo-European in the early part of this century.
Like Armenian, Tocharian appears to have been greatly affected by outside
influence. It is probably due to this influence that, again like Armenian, Tocha-
rian is markedly less archaic than Baltic and Slavic, even though both of these
languages are attested later. This is especially true for Tocharian sound struc-
ture, whose historical derivation from Proto-Indo-European is extremely complex.
Some aspects of its verb inflection, however, seem to preserve archaic elements.
 Abbreviations of Indo-European language names 55

3.12. Other Indo-European languages. In addition to the subgroups of Indo-Eu-

ropean briefly outlined above, ancient Europe and the Near and Middle East offer
evidence for a number of other languages that seem to belong to the Indo-Euro-
pean language family, such as Phrygian (Asia Minor, 5th c. BC), Thracian (Asia
Minor to eastern Balkans, inscriptions 5th c. BC, attested mainly in citations in
ancient Greek and Roman writers), or Venetic (Northern Italy, 2nd to 1st c. BC).
Many such languages seem to have been used in the central unmarked area of
the map above, between Slavic, Albanian, and Greek. However, their attestations
are too fragmentary to be helpful for historical and comparative linguistic work.

4 Abbreviations of Indo-European language

names
In the remainder of this book it will be useful to abbreviate the names of lan-
guages from which examples are cited. For many languages, different stages are
distinguished by prefixes, such as “Modern”, “New”, or “Old”. A key to these
abbreviations is given below, first of prefixes, then of the various language names.

Prefixes:
Class. = Classical N = New
Mod. = Modern P = Proto- (= reconstructed)
O = Old pre- = an earlier stage, without indication of the exact loca-
M = Middle tion in time
(The most usual designations of the old, middle, and modern stages of a given language are O,
M, and N (for ‘new’), as in OE = Old English, ME = Middle English, NE = New or Modern English.)

Language names:
Alb. = Albanian Arm. = Armenian
Att. = Attic (Greek) Av(est). = Avestan
BCMS = Bosnian-Croatian-Montenegrin-Serbian BS = Balto-Slavic
Bulg. = Bulgarian Celt. = Celtic
E(ngl.) = English Fr. = French
G(erm.) = German Gaul. = Gaulish (Celtic)
Gk. = Greek (usually ancient Gk.) Gmc. = Germanic
Go(th). = Gothic HG = High German
IAr. = Indo-Aryan Icel. = Icelandic
IE = Indo-European Ir. = Irish
Iran. = Iranian It(al). = Italian
56 The discovery of Indo-European

Lat. = Latin Latv. = Latvian

Li(th). = Lithuanian N = Norse
Norw. = Norwegian OCS = Old Church Slavic
Pers. = Persian Pol. = Polish
Port. = Portuguese Rom. = Romance
Rmi = Romani Rmn. = Romanian
Ru(ss). = Russian SCr. = Serbo-Croatian
Serb. = Serbian Skt. = Sanskrit
Slav. = Slavic Sp(an). = Spanish
Swed. = Swedish Toch. = Tocharian
W = Welsh
Chapter 3: Writing: Its history and its
decipherment
Alpha es et O
‘You are the beginning and end’ (from the old Christmas carol In dulci jubilo)

1 Introduction
The great advances of historical and comparative linguistics in the nineteenth and
twentieth centuries would have been impossible without the availability, interpre-
tation, and in many cases decipherment of written documents.
It is, of course, fairly obvious that written texts are important for historical
linguistics, for they are the only source for earlier language stages (such as the
Old English version of the Lord’s Prayer cited in Chapter 1). But their significance
goes much farther. For instance, some of the early Indo-European languages, such
as Hittite and Tocharian, are only attested in written form; no modern descend-
ants have survived. Without written texts of these languages, our knowledge of
Indo-European would be greatly diminished. Moreover, both the Hittite and the
Tocharian texts yielded their information only because the scripts in which they
were written were already known from their use in other languages. The Hittites
had adapted the so-called cuneiform script of ancient Mesopotamia, and the
Tocharians an offshoot of scripts used in India.
This, however, is not where the importance of understanding written texts
ends. The cuneiform script of ancient Mesopotamia had long died out, together
with the civilizations that employed it. It was only the decipherment of cuneiform
script by nineteenth-century scholars that made it possible to read this script and
the documents in which it was used, and to understand how these texts were
pronounced.
In this chapter we take a closer look at the nature of writing, written texts,
their origin, development, interpretation, and decipherment.

2 History of writing
2.1. Oral traditions. In the modern industrialized world, writing has become such
an essential component of all of our activities that we find it hard to imagine a
world without writing. But even in the western world we do not need to go back far

https://doi.org/10.1515/9783110613285-003
58 Writing: Its history and its decipherment

before we come to periods in which writing was limited to a small elite, so much
so that literacy itself could be considered something like magic by the common
people. (See Chapter 1.) And if we go back far enough we come to a time, some-
where before the fourth millennium BC, when nobody knew how to read and write.
This does not mean that human beings were condemned to living without
the benefits that we can so easily derive from reading: access to the wisdom and
also the follies of earlier generations. Preliterate societies have their own ways of
handing down such information, within the oral medium. Many readers will be
familiar with the system of Griots in West Africa, bards who committed the history
of their society to memory and who thus maintained a continuity of tradition. It
was this system, and its amazing accuracy, that made it possible for Alex Haley to
reestablish the link with his own African ancestors as reported in his book Roots.
We tend to disbelieve the accuracy of such orally transmitted texts, because
preoccupation with written transmission and prejudice against rote learning has
greatly diminished our ability, or willingness, to learn texts accurately by heart.
In societies where rote learning – at least of important things – is treasured, we
can even today observe feats of memorization that we find hard to believe. During
this century an eminent expert in indigenous Indian grammar is said to have set
aside a few years to memorize the entire Rāmāyaṇa, one of the two great epics of
India which is about three times the length of the Iliad and Odyssey combined.
But his internalization of the text went far beyond an ability to recite the text
from beginning to end, or any particular section. If asked where a particular word
occurred within the text, he was able to recite every single passage in which the
word occurred. Put differently, he had memorized the text in such a fashion that
he could operate on it more or less in the same fashion as one would on a text
stored in computer memory.
This feat of memorization should not come as a surprise in India, a country
which until very recently preserved all of its early sacred literature entirely in oral
form, along with supporting texts on the function, meaning, and even grammar
of these texts – over a period of between 3500 and 2500 years. To accomplish
this task without serious lapses, an extremely intricate support system was devel-
oped. One of the systems was an amazingly sophisticated grammatical tradition.
Perhaps most interesting, however, is the early development of an elaborate back-
up system through which the texts were memorized in two, three, or even more
different versions. One of these operated as follows: Let a, b, c … stand for the
words of the original text; the back-up, then, goes abba, bccb, … (Using an English
translation of an early Sanskrit hymn, we can illustrate this method as follows:
The original text is I invoke Agni …; the back-up text runs I invoke invoke I invoke
Agni Agni invoke …) Rules of conversion made it possible to restore the original
text from any of the back-up versions.
 History of writing 59

Modern scholarship has been able to show that there had been some textual
corruption early in the Indian oral tradition. However, once the back-up systems
had been put in place, the texts were handed down with such amazing accuracy
that from one end of the South Asian subcontinent to the other, a particular text
would exhibit no significant variations, except for differences attributable to the
fact that the pronunciation is to some extent “colored” by the native language of
the reciters.

2.2. Forerunners of writing. Oral traditions often employ non-oral “props” to

increase their accuracy, or for contexts in which oral communication is not possi-
ble. Some, fairly elementary props continue to be used even today; others disap-
peared with the arrival of literacy.
A very wide-spread phenomenon is the use of mnemonic devices, which help
in remembering numbers or other information that can be quantified or is in some
way related to counting. A very simply device of tallying numerical information
is presented in Illustration 1. This device, of course, is still in wide use today. But
evidence for its use appears to go back at least 10,000 years.

Ill. 1: Mnemonic counting device

More complex is the system of rosary beads, employed to keep track of sequences
of prayers in traditional Christianity, but also in other religions such as Buddhism.
Counters of greater size and different material or color may mark the beginning
or end of different prayer sequences, while other counters help keep track of indi-
vidual prayers within the sequence. The cultural importance of rosaries in medi-
eval Europe can be gauged from the fact that the modern English word bead is
derived from an Old English word bede ‘prayer’. When people keeping track of
their prayers on the counters of the rosary were asked what they were doing, they
might well have said I’m counting my beads, meaning ‘… my prayers’. But seeing
counters being manipulated, those asking the question were free to interpret the
word beads as referring to the counters, not to prayers.
The most elaborate system of mnemonic devices seems to be presented by
the quipus of the pre-Spanish Inca empire of Peru and adjacent areas of South
America. (See Illustration 2.) Quipus were primarily used for recording numerical
information by means of a combination of different string colors and knots. The
often complex statistical information thus stored “triggered” memorized recita-
tions concerning particular transactions, such as the inflow and outflow of goods
60 Writing: Its history and its decipherment

in the imperial treasury. Secondarily, apparently, the quipus were also used as
mnemonic devices for other types of texts, such as historical accounts, that like-
wise had been committed to memory. However, like the strokes in Illustration 1,
the quipus provided an accurate record only of numerical information. They
could not be used for accurate, verbatim transcriptions of non-numerical texts.

Ill. 2: Section of a quipu bundle

Especially significant is the appearance during the ninth to sixth millennia BC of

clay tokens in the ancient Near East, which came in different sizes, shapes, and
designs and were used for identifying goods exchanged in trade, and different
quantities of such goods. (See Illustration 3.) It has been claimed that these clay
tokens played a very important role in the eventual development of full writing in
this area, since they bear a strong resemblance to some of the symbols used in the
early stages of Near Eastern writing.

Ill. 3: Clay tokens of the Ancient Near East

 History of writing 61

Beside mnemonic devices, preliterate societies seem to have employed pic-

torial representations to convey messages. Many such representations have
been found all over the world. However, lacking information on the context for
these representations, their purpose, or the language of those who produced
them, the interpretation of these representations usually is quite uncertain.
A common joke among prehistorians, commenting on earlier practices, is that
“When in doubt, give a mystical interpretation.” Although such an interpretation
may be correct for some representations, a fair amount of evidence suggests that
more mundane interpretations are more appropriate. First, in the Ancient Near
East, where we can follow the development of writing in greatest detail, writing
was initially employed for keeping track of commercial transactions. A perfect
parallel is found in the case of the Inca quipus. Secondly, we have direct evidence
that at least some pictorial representations had rather mundane purposes. Con-
sider Illustration 4, a note sent in the nineteenth century by an Ojibwa girl to her
boyfriend. The message of this note is quite clear and was confirmed by the boy’s
reaction: He did indeed come to the girl’s place as requested. The message con-
veyed can be roughly paraphrased as follows:

From the girl of the Bear Totem [see sign on upper left] to the boy of the Mudpuppy Totem
[sign on lower left]: Take the path which leads toward the lakes [right]. After it is joined by
a path from the right, a path goes off to the left, leading to two teepees with three Chris-
tian girls in them [see crosses]. I’m in the left one and want you to come [see waving hand
symbol in left teepee].

Ill. 4: Letter of an Ojibwa girl to her boyfriend

(adapted from G. Mallery (1893): Picture Writing of the American Indians [10th Annual Report of
the Bureau of Ethnology])

Highly successful as it may have been, however, the representation in Illustration

4 does not convey a specific linguistic message and in this sense is not much dif-
ferent from the secondary use of the quipus to keep track of non-numerical infor-
mation. It merely conveys certain ideas, and its interpretation depends heavily
on knowing enough about the cultural and other contexts in which it was sent.
62 Writing: Its history and its decipherment

Without knowing the context, it would be impossible to come up with the correct
interpretation simply by looking at Illustration 4; one might just as well assume
that it conveyed some deep mystical meaning. More than that, the message of
Illustration 4 can be paraphrased in many different ways and, evidently, in many
different languages – not just Ojibwa, but also English. In short, unlike writing, it
fails to express a specific linguistic message.
A similar, but more elaborate, use of pictorial symbols is found in the Ancient
Near East – cylinder seals that were rolled into wet clay to produce an impression
identifying a person or clan. These seals continued to be used when writing was
introduced (much like modern seals and sealing wax) and even could incorporate
writing into the pictorial representation. See Illustration 5. Like the clay tokens
in Illustration 3, these seals may have been an important precursor of writing in
the area.

Impression Cylinder
(side view)

Ill. 5: Ancient Near Eastern cylinder seal and impression

(adapted from E. Chiera (1938): They Wrote on Clay)

2.3. The development of writing in the Ancient Near East. As noted earlier, the
development of writing, from the early, pre-writing stages of mnemonic devices
and pictorial descriptions, via early attempts to combine and expand them, to full
writing can be best observed in the Ancient Near East of Mesopotamia, especially
in the civilization of ancient Sumer. In other areas of the world, such as ancient
Egypt, China, or Meso-America, the early history of writing is shrouded in mystery.
Some scholars have proposed that, in fact, writing originated only once in
the “Old World” of Eurasia and Northern Africa, in the place where we can see
it develop most clearly – Mesopotamia. According to this view, the fact that we
lack specific evidence for such a development elsewhere in the “Old World” is
no accident: Writing outside Mesopotamia was not an independent invention,
but resulted from the spread of writing – or the idea of writing – to the rest of the
world. This view appeared to receive further support from the fact that the earliest
attestations of writing in Mesopotamia date back to about 3100 BC. Writing in
 History of writing 63

Egypt (to the west) and in ancient Elam (to the east) was considered to come in
somewhat later (about 3000 and 2900 BC, respectively). Written records appear
even later in the Indus Valley (about 2400 BC), and yet later in China (commonly
dated as beginning about 1300 BC, but see below). That is, the farther from Mes-
opotamia, the later the appearance of writing.
This diffusionist view of the origin of writing has in recent years lost much
of its earlier persuasiveness. The most important reason for a change in perspec-
tive is the increasing realization that the characters used by the Meso-American
Mayans in their inscriptions represent genuine writing, recounting not just astro-
nomical information (as had earlier been believed) but also the chronicles of royal
families and their kingdoms.
Scholars now are ready to accept at least three different, independent origins
of writing: Mesopotamia, China, and Meso-America. On the other hand, the
writing systems of Egypt and Elam are generally considered diffused from Meso-
potamia. (The Indus Valley writing still awaits successful decipherment. Recent
research, however, suggests independent origin; and some researchers even ques-
tion whether it was a writing system.)
Whatever the merits of these scholarly debates, one should note that the tra-
ditionally posited time difference between the earliest appearance of writing in
Mesopotamia, Egypt, and Elam is exceedingly small, given the time depth we are
talking about. In fact, recent research suggests that Elamitic writing developed at
just about the same time on the Iranian plateau as it did in Mesopotamia, that is,
about 3300–2900 BC. And as work on Chinese proto-history continues, the date for
the development of Chinese writing gets pushed back farther and farther, currently
to at least the seventeenth century BC by conservative estimates (see § 5.1 below).
Even in their earliest stages, the different writing systems of the Old World
exhibit significant differences in detail; see Illustration 6. And the observable
similarities (as in the symbol for ‘sun’) can easily be explained in terms of the
properties of the designated objects. The possibility of independent development,
therefore, cannot be ruled out.

Ill. 6: Some pictograms in early Sumerian, Egyptian, and Chinese writing

64 Writing: Its history and its decipherment

What is more important is to look at the conditions under which writing

appears, not only in Mesopotamia, but also in Egypt, Meso-America, and else-
where. As it turns out, in all of these areas writing was introduced at a time when
large states had developed, with a centralized system of collecting revenue. The
revenue, in turn, had to be distributed to the king’s court, the nobles or warriors,
and the priests; and some of it had to be kept as a reserve for the general popu-
lace in time of need. In Mesopotamia and Egypt, it was clearly for the purpose of
keeping track of these transactions that writing was developed. And it is proba-
bly no accident that under very similar conditions, the Inca empire developed its
elaborate system of quipus.
Meso-America appears to be an exception. Even the earliest attested inscrip-
tions deal with calendrical information and events at the royal court. But perhaps
the inscriptions do not tell the whole story. The word used to refer to writing, tz’ib,
literally means ‘paint’, which suggests that an earlier phase of writing consisted of
“painting” symbols on paper-like material, and that the documents have been lost
to us because of the perishability of the material. At present we do not know what
kinds of texts were written in the early “painted” form of writing; but it is at least
possible that these texts dealt with commercial matters. (Evidence for a continued
tradition of “painted writing” is found in several late codices that survived the
organized effort of Spanish colonialists to destroy all native documents because
of their “heathen” nature.)
At any rate, the commercial origin of writing in Mesopotamia becomes very
clear when we look at early transitional phases, between the pre-writing stage
and full writing. An example is found in Illustration 7. As this example shows, the
intermediate phase tends to combine mnemonic and pictorial devices to produce
more specific messages than earlier precursors of writing. Still, the messages are
quite limited. This shortcoming was partly compensated for by the addition of
cylinder seal impressions (see Illustration 5 above) which identified the persons
involved in the transaction. (Compare the similar use of the totem symbols in
Illustration 4.) Even then, a message like the one in Illustration 7 does not tell us
what exactly happened to the 54 cattle it refers to.

Ill. 7: Example of proto-writing in Ancient Near East

 History of writing 65

In spite of their shortcomings, these early attempts at increasing the power

of graphic symbols must have been successful enough to stimulate the develop-
ment of even better representations. This development was accomplished by an
increasingly conscious – and also increasingly confident, even daring – expan-
sion of the original code. To illustrate these developments more vividly, most of
the following examples are drawn from English, as if English were now in the
process of developing writing out of pre-writing.
One of the methods to expand the code was the principle of semantic trans-
ference, the use of symbols for easily depicted objects to designate semantically
related concepts that are abstract and thus more difficult to draw. See Illustra-
tion 8 in which the easily drawn symbol for ‘sun’ is used to refer to the seman-
tically related ‘day’, a notion that is quite difficult to draw. See also the genuine
Sumerian examples for ‘God’, ‘sky’, and ‘star’ in Illustration 6 above. In some
cases, of course, the use of the same symbol for two different notions might lead to
confusion (as in the case of ‘sun’ and ‘day’). To remedy this situation, the symbols
sometimes were secondarily differentiated; see again Illustration 8, as well as the
genuine Sumerian example in Illustration 9, where ‘king’ is differentiated from
‘man’ by a stylized crown.

Ill. 8: Semantic transference and secondary symbol differentiation

Ill. 9: Symbol differentiation in Sumerian

A similar expansion was achieved through phonetic transference, the use

of an easily drawn symbol to designate a phonetically similar or identical word
that is difficult to render graphically. Illustration 10 gives a genuine example from
Sumerian, an ingenious solution of the otherwise nearly insoluble problem of
how to depict the notion ‘life’.
66 Writing: Its history and its decipherment

Ill. 10: Phonetic transference

Yet other ways were devised to expand the code. One of these was the use of
phonetic indicators to help differentiate which of several possible interpreta-
tions of a given symbol was intended. Phonetic indicators, thus, are used very
much like the statement “sounds like” in the game of charades. For instance,
in the Ancient Near East, there was a sexist metaphor equating two women
with strife, discord, or quarrel. As a consequence, the notions ‘strife’, ‘discord’,
‘quarrel’ could be designated by a picture of two women facing each other, as in
the upper part of Illustration 11. To indicate which of these readings was intended,
the picture of a cord might be added to narrow down the choice to ‘discord’, as in
the lower part of Illustration 11. (See Illustration 18 further below for a different
phonetic indicator used with the same base symbol, and compare the real exam-
ples from Egyptian and Mayan in Illustrations 13 and 19 below.)

Ill. 11: Phonetic indicators

An alternative solution consisted in the introduction of semantic indicators,

in which the meaning of subsidiary symbols narrows down the range of possible
interpretations. See Illustration 12, where the addition of an anchor narrows down
the choice of possible readings to ‘sea’. As the real Ancient Egyptian example in
Illustration 13 shows, phonetic and semantic indicators may be employed simul-
taneously. Here we have phonetic transfer from a conventionalized pictograph for
pr ‘house’ to pr- ‘go out’, plus addition of a phonetic indicator r and a semantic
indicator for ‘MOTION’.

Ill. 12: Semantic indicators

 History of writing 67

Ill. 13: Semantic and phonetic indicators in Ancient Egyptian

An even more powerful means of expanding the code is known as the rebus
principle, after a parlor game in which words or component parts of words are
expressed by pictures of objects, or by other symbols that have (nearly) the same
sound. (An English example would be a picture of a bee followed by the symbol 4,
spelling out before.) Combining existing characters in their phonetic values made
it possible to express longer and more complex words, as in Illustration 14. On the
left side, a sequence of a kneeling person = kneel [nīl] plus the by now well-known
symbol for ‘sun’ [sǝn] spells out the name Neilson. On the right side, the combina-
tion of the symbols for ‘sun’ and ‘day’ is employed to spell out both Sunday and
the identical-sounding ice cream concoction sundae. This method was especially
useful in expressing foreign names. While such names tend to be meaningful in
their original language, they often are “meaningless” in another language and
therefore difficult to represent in terms of meaningful symbols. (Consider for
instance the name Mississippi, which quite appropriately means ‘big river’ in the
Algonquian language from which it was adopted, but which in English is just an
arbitrary combination of sounds.)

Ill. 14: Expansion of the code through the rebus principle

The increasing use of the rebus principle brought about an increasing pho-
neticization of the writing system, in contrast to the early logographic system
which mainly focused on the semantics of words. (Sometimes the term picto-
graphic is used for the early system, because its symbols generally are pictorial
representations. But the term logographic is more useful, since writing systems
may be logographic without using pictorial representations. A perfect example of
such a writing system is that of Chinese; see § 5.1 below.)
Ultimately the most powerful and, at the same time, most daring method was
the principle of acrophony and similar arbitrary phonetic reductions. Nowadays,
68 Writing: Its history and its decipherment

the related process of acronymy is used mainly as a means of abbreviation, gen-

erally by drawing on the initial letters of words, as in US = United States. Occa-
sionally, it may use the first syllables or parts of syllables, as in sitcom = situation
comedy. In any event, acronymy usually is based on spelling. At the early period
in the development of writing that we are concerned with, acrophony operated
on the initial syllables of words as they are pronounced, or on parts of these syl-
lables. Moreover, the purpose of the process was simply to create a more flexible
code for spelling out words phonetically, without regard for the original semantics
of the symbols employed. And to accomplish this goal, other, non-acrophonic
“mutilations” of words might be used to create new phonetic symbols.
Illustration 15 exemplifies various different routes that may be taken in this
process. One is to abstract the initial consonant plus the following vowel as the
essential phonetic value of the writing symbol. Alternatively, the final consonant
plus the preceding vowel might be abstracted in this way. Even more daringly, the
precise quality of the vowel may be ignored, so that a symbol like the one for ‘day’
may stand for [d] plus any vowel or even no vowel at all. And again, the same may
be done for a final consonant plus preceding vowel. Yet another daring step is to
ignore the difference in voicing between, say, [t] and [d] and to permit the same
symbol to stand for both [tri] and [dri].

Ill. 15: Creation of phonetic symbols by acrophony and related processes

If pushed to its logical conclusion, the principle of acrophony can lead to

the development of a syllabary, in which, ideally, written symbols consistently
designate particular phonetic syllables, no matter which word they might occur
in. In many writing systems, symbols actually express parts of syllables, so that
complete syllables have to be spelled out by combinations of “syllabic” signs.
(Compare Illustration 16 for some examples.) Whatever the details, at this stage,
the writing system has become completely phoneticized.
 History of writing 69

Ill. 16: Development of syllabary

In the Sumerian tradition, the syllables or parts of syllables designated by syl-

labic symbols could be of the shape consonant + vowel (CV), vowel + consonant
(VC), consonant + vowel + consonant (CVC), or just plain vowel (V). See Illustra-
tion 16 for some hypothetical examples. In Egyptian, on the other hand, syllabic
symbols consistently were of the shape CV (Illustration 17). Moreover, as in the
case of the symbol ‘day’ = [dV] of Illustration 15, the “V” of these syllabic symbols
could be any vowel, or even no vowel at all.

Ill. 17: Egyptian syllabary (or abjad)

70 Writing: Its history and its decipherment

While from one perspective the Egyptian method of phonetic writing is still
syllabic, it is easy to see how this writing system might be reinterpreted as a “con-
sonantal alphabet”, a system very much like our alphabet, except that it writes
only consonants and has no special symbols for vowels. The term abjad, the
Arabic word based on the initial letters of the Arabic writing system (ʔabǰad), is
now used as a designation for this type of writing system. In fact, there has been
some controversy over whether the Egyptian writing system should be considered
to be a syllabary or an abjad. Whatever the resolution of this controversy, the
Egyptian type of writing was very important for the development of writing. (See
§ 2.4 below.)
In the traditional cultures of the ancient Near East and Egypt (and also of
Meso-America), syllabic writing never completely replaced the earlier logographic
systems with or without phonetic or semantic indicators. The two coexisted, just
as in the hypothetical examples of Illustration 18 or the real Egyptian example
in 13. Note also Illustration 19, which shows a similar coexistence of logographs
with or without phonetic indicators and syllabic writing in another traditional
writing system, that of the Mayans. (What complicates things in Mayan writing
is that phonetic indicators are combined with the logographic base symbol into
a complex symbol whose parts are difficult to distinguish for the beginner, and
the syllabic symbols are combined in a similar fashion.) Readers familiar with
Japanese may find a parallel in that language, too, in that Chinese logographic
symbols are used side by side with syllabic writing. Even languages like English
exhibit traces of mixed writing, as in the combination of logographic numeral
symbols with something like phonetic indicators in expressions such as 1st, 2nd,
3rd, 4th.

Ill. 18: Mixed system of logographic and phonetic writing

 History of writing 71

Ill. 19: Mixed system in Mayan

(adapted from M. D. Coe (1992): Breaking the Maya code.)

In Mesopotamia, things got even more complicated. At a fairly early period,

Semitic peoples such as the Assyrians and Babylonians replaced the Sumerians
and took over their writing. As they did so, they retained many symbols with the
old Sumerian values, whether syllabic or logographic. This meant that in order
to read and write it was necessary to learn Sumerian; and grammars of Sumerian
were prepared for this purpose.
At the same time, the Babylonians and Assyrians began using symbols with
the phonetic values of their own languages. Symbols thus could have their log-
ographic meaning, the phonetic value of Sumerian, or the phonetic value of
Semitic languages like Babylonian. Yet another layer of complications was added
when the Indo-European Hittites adopted the writing system. As a consequence,
the symbol on the (a)-side of Illustration 20 below could have both its Sumerian
value ē ‘house’ and the Assyrian value bītu ‘house’. In many cases, the range of
phonetic values of particular symbols became even more formidable, as in item
(b) of Illustration 20.

Ill. 20: Multiple values of symbols in Mesopotamian tradition

With an increasing development away from the early pictographic system toward
a logographic or even phonetic one, there was less and less motivation in the
tradition of Mesopotamia to draw pleasing or realistic pictures. Symbols became
72 Writing: Its history and its decipherment

increasingly simplified and standardized. But more than that, in order to make
the process of writing easier, the orientation of symbols was changed in many
cases (see stage II in Illustration 21). Most noticeable is the fact that slowly, and by
no means affecting all symbols at the same time, the shape of the letters began to
change radically in response to the material and implements used for writing (see
stages III and IV). In the tradition of Mesopotamia, most writing was produced
on wet clay tablets which, after they had been covered with writing, were either
sun-dried or, for more permanent storage, baked into brick. The “stylus” with
which the clay tablets were engraved was wedge-formed (see Illustration 22) and,
unless scribes were very careful, would leave a wedge impression at the insertion
point, followed by a narrow groove. With the increasing trend away from pic-
tography, there was no longer any need to avoid the wedge; and in keeping with
a general tendency to make writing simpler, lines were made straight, and the
resulting shapes were rearranged for yet further simplification. The wedge shape
characteristic of the resulting system is responsible for the name cuneiform (lit.
‘wedge-form’).

Ill. 21: Development of cuneiform characters

Ill. 22: Stylus used for cuneiform writing

The developments just discussed may appear extreme, especially if one considers
that the ancient Egyptian script in Illustration 17, referred to as hieroglyphic,
remained essentially pictorial. However, this pictorial script was retained only
in monumental inscriptions. Side by side with it there arose another, simplified
system which, like the cuneiform script, was greatly influenced by the imple-
ments and material used for writing (a narrow ink brush on papyrus). The latter
system of writing is referred to as demotic. A modern analogue to this influence
of writing material and instrument is the perhaps short-lived development of rel-
atively grainy characters on dot-matrix printers, where the lines and curves of
written characters are decomposed into dots or “pixels”. Here, too, the medium
 History of writing 73

of writing, a pixel-oriented graphics system on the computer plus the mechanics

of the dot-matrix printer, changed the shape of characters. A further parallel can
be seen in the shape of the early Germanic runes; see § 2.6 below.

2.4. The origination of full syllabaries and consonantal alphabets. As noted

earlier, in the traditions of Mesopotamia and Egypt, the step toward a fully pho-
netic, syllabic (or even alphabetic or abjad) system of writing was never com-
pleted. The reason, no doubt, was the fact that there was a more or less unbro-
ken cultural tradition; and logographic, beside syllabic writing was part of that
tradition. (Another reason may have been that the complex system that resulted
suited the scribes very well, as it was virtually impenetrable to the majority of the
population, thus assuring the small class of scribes continued employment and
a good income.)
Full phonetic writing, in the form of syllabaries or even abjad forms of alpha-
bets developed on the periphery of Mesopotamia and ancient Egypt, among
peoples influenced by these great civilizations, but not beholden to their tradi-
tions. (The term “full phonetic writing” is perhaps a little too strong. Just as in
modern alphabetic systems, a few logographic symbols remained, similar to, say,
the $ or £ of English.)
By a process sometimes referred to as stimulus diffusion, these societies
adopted the idea of writing, especially that of syllabic or abjad writing, as well as
the “look” of the characters employed; but the symbols they created, as well as
their phonetic values, were quite independent. (A similar development took place
more recently in the case of the Cherokee syllabary, whose symbols, such as
A and H, are inspired by the shape of the alphabet used for English, but whose
phonetic values differ greatly: A, for instance, designates the syllable go, and H,
mi. See § 5.3 below.)
Two of these writing systems were especially important historically. The Old
Persian one provided the starting point for the decipherment of the cuneiform
script and thus, ultimately, for understanding of the Ancient Near Eastern texts –
whether Sumerian, Semitic Assyrian and Babylonian, or Indo-European Old
Persian and Hittite. The other system, used by various South and West Semitic
peoples, became the source for a large number of modern writing systems. These
include the consonantal alphabets or abjads of Hebrew and Arabic and the alpha-
betic script of ancient Greece which, unlike its Semitic ancestor and relatives,
designates both vowels and consonants. From this script, in turn, are derived
the Roman alphabet, used in western Europe (and its “Insular” offshoot, used
for medieval English and still employed for Gaelic), the runic script of early Ger-
manic, and the Cyrillic alphabet used in large parts of eastern Europe and adja-
cent parts of Asia.
74 Writing: Its history and its decipherment

Let us begin with a brief look at the Old Persian syllabary. As in many sylla-
baries, the symbols of this script must be read with different values, sometimes
with and sometimes without an inherent vowel; and long vowels are written as
combinations of two short vowels (see Illustration 24). Ideally, the script would
have had a different symbol for each consonant + vowel combination; but as Illus-
tration 23 shows, only the set of consonant characters with an inherent a-vowel
is complete. These are the “base consonants”, which are also used to designate
consonants without an inherent vowel. Missing characters for other consonant
+ vowel combinations are produced by placing the i- or u-vowel symbol after the
base consonants.

Ill. 23: Old Persian syllabary

 History of writing 75

Ill. 24: The name of King Xerxes in the Old Persian syllabary

Various explanations have been advanced for the evidently inconsistent character
of the script. The most plausible is that at a certain point, the decision was made
not to wait any longer for the completion of the script, but to go ahead and use
whatever had been developed by that time, not unlike an early release of software
before all the “bugs” are worked out. Presumably, the great rulers of the mighty
Persian empire were getting impatient with the slow progress of the scholars who
were working on the writing system and decided to go ahead and use an incom-
plete “prototype” version in order to proclaim their great deeds to the world and
to posterity.
Even though using characters with very different shape, the Old Persian syl-
labary shows certain similarities to the Egyptian system, at least as far as the base
consonants are concerned: In both cases, symbols stand for a consonant ± vowel.
The systems differ in that the Old Persian symbols stand for C(a), while the Egyp-
tian ones indicate a consonant ± any vowel.
This more radical development becomes important in the writing systems of
South and West Semitic peoples in the area of Syria/Palestine in the early part of
the second millennium BC. At an early period, a large variety of writing systems
is found, some with characters similar to cuneiform, others with more pictorial
symbols reminiscent of the Egyptian hieroglyphs. These facts suggest that the
inhabitants of this centrally located area were influenced by both the Ancient
Near Eastern and the Egyptian traditions.
In their phonetic character, however, these writing systems were most similar
to the Egyptian syllabary which, as noted earlier, was inherently open to reinter-
pretation as an abjad or consonantal alphabet. Although there may still be some
controversy over whether this reinterpretation took place in Egyptian, it is quite
certain that it was completed in the new systems of the South and West Semitic
peoples.
The development of a writing system which indicates only consonants was
possible and made sense because of the word structure of the Semitic languages
and of the distantly related Ancient Egyptian. In these languages, the basic build-
ing block of words is the “root”, a configuration of (generally) three consonants
76 Writing: Its history and its decipherment

which carry the basic lexical meaning, such as √KTB ‘write’. By insertion of dif-
ferent vowel patterns into this “consonantal skeleton” (with or without prefixes
or suffixes), different types of words are created, such as KaTaB ‘he wrote’, KāTiB
‘writer’, kit(ā)b ‘book’, mi-KTaB ‘letter’ etc. Given enough context, it is quite
predictable which of the various forms of √KTB must be intended and what,
therefore, the vocalism of that word must be. It is thus possible to write without
specifying the vowels. (Even in English this is marginally possible, as in f y cn rd
ths y cn bcm a gd scrtry, an advertisement for a shorthand system seen on many
subways in the US, where only the word a causes difficulties, since it contains no
consonant.)
In the Semitic languages (and the related Egyptian) the acrophonic princi-
ple therefore could and did lead to characters which spell a particular consonant
without further specification of the following vowel (if any). Hence: p (V) = pa, pi,
pu, p etc. See Illustration 17 above for part of the Egyptian syllabary or abjad and
Illustration 25 below for the two first symbols of the early Semitic abjad of Canaan,
and the further developments of these symbols in Arabic and Hebrew, the two
best-known modern Semitic languages.

Ill. 25: Developments of Semitic consonantal signs

Illustration 25 further shows a number of other features of Canaanite writing.

First, although greatly simplified and conventionalized in their shape, the letters
are pictographic in origin. Thus, the letter ʔaleph is a stylized picture of an ox; and
beth represents the ground plan of a house. The selection of these symbols is not
just due to chance: ʔaleph and beth mean ‘ox’ and ‘house’, respectively. Moreover,
ʔaleph and beth are the names of the two letters, and the phonetic values of the
letters are identical to the initial sounds of the letter names/pictographs, i. e., ʔ
and b. A similar acrophonic relation between letter name and speech sound holds
true for the rest of the writing symbols.
The great advantage of this system over the traditional mixed systems is (a)
its much smaller number of symbols (20 to 30 vs. 600 to 700 in cuneiform and
hieroglyphic script), and (b) the easy manner in which the phonetic value of each
 History of writing 77

symbol can be derived from the corresponding letter name. Memorizing a rela-
tively small number of letter names in a fixed order therefore made it possible to
learn how to write in about a week, rather than the years required to master the
traditional scripts of Egypt and the Ancient Near East.
Note that while a number of letters, in addition to and , are clearly picto-
graphic in origin, other letters, such as = ḥēth, are not. Moreover, some letters
are quite patently secondary modifications of other characters, created to increase
the number of speech sounds that could be distinguished. Compare for instance
= hē beside = ḥēth, where one is no doubt derived from the other by addition
or deletion of strokes, although the direction of derivation is uncertain. Develop-
ments of this type reflect the same tendency that we observed earlier in the often
daring methods of expanding the code during the early stages of the development
of writing. Once people have discovered a good thing, they usually find ways to
make it even better.
There is yet another parallel to earlier developments. Once writing symbols
become conventionalized, their pictorial origin becomes irrelevant and their
shapes change to suit the writing materials, the scribes’ convenience, or even
more artistic, calligraphic concerns. This accounts for the change of early to ,
to , and the further developments in later Semitic.

2.5. The development of the alphabet. At some time during the early ninth
century BC (or perhaps even earlier), the consonantal alphabet of the Semitic
Phoenicians was adopted by the Greeks through commercial contacts in Asia
Minor. Superficially, the writing systems of the two peoples look very similar, as
can be seen in Illustration 26. The difference in language, however, combined
with the way in which the writing system was memorized and recited, brought
about a major change – from an abjad system that clearly distinguished only con-
sonants to a fully alphabetical one which had distinct symbols both for the con-
sonants and for the vowels. Thus, whereas in the Semitic languages, the symbol
in column (a) of Illustration 26 had the value of a consonant, transcribed [ʔ], the
corresponding Greek symbol (column (a‘)) represents a vowel, [a].
What made this change possible is the fact that the Greeks took over from the
Phoenicians not only the letters of the writing system, but also (a) the order in
which the letters were recited, (b) the letter names, and (c) the acrophonic prin-
ciple according to which the first sound of the letter name designates the speech
sound denoted by the letter.
Lacking the consonant [ʔ] in their language, the Greeks omitted that sound
in the name of the first letter of the writing system, [ʔaleph], and pronounced
the name as [alpha]. And presto, given the acrophonic principle, they acquired a
letter which designated a vowel, not a consonant!
78 Writing: Its history and its decipherment

Ill. 26: Semitic abjad vs. Greek alphabet

Similar developments led to the development of the vowel letters for e and o. The
former developed out of a letter designating an h-sound in Semitic, the latter,
from a letter transcribing the “pharyngeal” stop [ʕ] and lacking any counterpart
in Greek. See the two columns on the left of Illustration 27 below.

Ill. 27: The development of selected letters of the alphabet

The remaining vowels of the early Greek alphabet derive from characters which
could, under certain circumstances, be used as vowels in Semitic. One of these
was the letter yod whose basic value, given the acrophonic principle, was y; but
 History of writing 79

between consonants it was realized as i. Since Greek at this point lacked the sound
y, but had an i, the symbol was naturally appropriated to designate the vowel,
and the name of the letter changed to iota. The case of Semitic waw, with base
value w but inter-consonant realization as u, is even more interesting. In this case,
early Greek had both sounds. So the question arose: In what value should waw
be adopted? The answer was to have it both ways, by splitting the letter in two,
one ( ) having the value u, the other, , later simplified to , a doubled version of
Greek υ, being used to designate w. Interestingly, an almost exactly parallel devel-
opment occurred when the Roman alphabet was adopted for writing the early Ger-
manic languages. The Roman alphabet used a single symbol (V) for both u and w.
Germanic users of the alphabet seem to have felt the need for a clearer distinction,
and V was doubled to VV or W to indicate w – hence the modern English name
of the letter, double-u. (Just as the other letters, the symbol , , soon changed
direction, becoming , .)
Within Greek, the development continued. When Ionic, an important dialect
of early Greek, lost its initial “aitches”, the letter Η which previously had the value
h acquired a vowel value, namely the long front vowel ē [ε̄ ]; and this value was
adopted in many of the other Greek dialects. In West Greek, by contrast, the letter
retained its value h, and it was this pronunciation that was adopted in Latin.
Once this precedent was set, a new letter was devised for the corresponding
long back vowel, Ό̄ , by opening up the lower part of the old letter and changing
it to . For these developments, see the last three rows of Illustration 27. In both
these developments and in the earlier “splitting” of waw we can see the same
creative processes at work that we have encountered several times already: Once
people have recognized a good thing, they like to make it even better.
In addition to these changes, there was a general tendency for the letters to
change “direction”, presumably to make them easier to write. For instance, while
in early Greek, the letter B faced to the left, as it had done in Semitic, later on,
it faced to the right. In this case, the change in letter orientation appears to be
connected with a change in the direction of writing. While the Semitic languages
were written from right to left (and still are), the Greek alphabet soon changed to
being written from left to right. (At an intermediate stage, the direction of writing
shifted in alternate lines. Starting, say, from the right in the first line and ending
on the left, the writer continued at the left side of the next line and went to the
right, much as one plows furrows with a team of oxen. The technical term for this
writing therefore is boustrophedon, lit. ‘in the manner of oxen turning’.)
Though the starting point for the development of a full alphabet can be
characterized as “sheer dumb luck”, the alphabet turned out extremely useful
for Greek and other Indo-European languages. The word structure of Indo-Euro-
pean is quite different from Semitic: Vowels are not just modifiers of essentially
80 Writing: Its history and its decipherment

consonantal roots, but may be primary meaning carriers, as in Engl. a, or Gk. ei,
pronounced [ē], ‘you (sing.) are’. A purely consonantal script would find it very
difficult to express structures of this type.

2.6. A note on the further fate of the alphabet. From the Greeks, specifically
from West Greek dialects, the alphabet spread to Italy. One of the languages for
which it came to be used was Etruscan. In the process, certain changes took place
in the value of individual letters, much as what happened in the earlier adoption
of the Phoenician writing system by the Greeks, except that the alphabetic char-
acter of the system remained unchanged. One of the modifications was the use of
the letter gamma to designate the voiceless velar stop, [k], presumably because
Etruscan made no distinction between voiced [g] and voiceless [k].
As the Romans adopted the alphabet from the Etruscans, they introduced
further modifications. Some of these were motivated by considerations of writing
ease, others by the structure of their own language, Latin. One change, introduced
only after some time, was “splitting” the original letter gamma into two – (pho-
netically [k]) and (= [g]) with an added stroke – because, unlike Etruscan, Latin
did make a distinction between voiceless [k] and voiced [g]. (See Illustration 27.)
Another innovation resulted from the fact that Latin had a sound f that was absent
in Greek and therefore lacked a corresponding character in the Greek alphabet.
At the same time, West Greek dialects had a combination = wh, designating a
sound combination absent in Latin. And just as Engl. wh can sound similar to f to
speakers lacking this sound combination, so the Romans identified = wh with
their own f and used it to transcribe that sound. But since occurred only in com-
bination with , the latter symbol was soon felt to be redundant, and was used
by itself to designate f. In both of these developments we see the same ingenuity
at work as in earlier Semitic vs. or early Greek vs. .
Other (direct or indirect) offshoots of the Greek alphabet include the runic
writing system (see Illustration 28), and the Cyrillic alphabet. The runic alphabet,
usually named futhark after its first six letters, is found in early Germanic inscrip-
tions from the early AD period to at least the eighth century. Eventually it gave way
to the Roman alphabet, but some of its characters, especially the single letter þ for
the voiceless sound designated in Modern English by th, was retained in many of
the early Germanic scripts and is still used in Modern Icelandic.
Early Roman reports tell us that the Germanic people used runes inscribed on
small wooden chips or tablets for oracular purposes, and great magical powers
were commonly ascribed to the runes. But the extant inscriptions are on stone or
metal, and, to the extent that they are complete or that we can interpret them,
they convey rather mundane messages. For instance, one of the most celebrated
inscriptions, found on a golden drinking horn, states (in transcription): ek
 History of writing 81

Ill. 28: Older Germanic runes

­hlewagastiz holitijaz (or holtingaz) horna tawido, which translates as ‘I, Hlewa-
gasti of Holt (or: Holting) made the horn’.
There must, however, be something to the Roman reports about the use
of runes on wood. First, this would explain the shape of the runes. Observant
readers may have noticed that only vertical or diagonal strokes are used in runic
writing; there are no horizontal strokes. If we assume that runes commonly were
written on wood, we can explain this peculiarity. Vertical or diagonal strokes that
cut through the wood grain leave marks that remain legible, while horizontal
strokes, along with the grain of the wood, are quickly filled up again and become
invisible. The fact that no early wooden inscriptions have been preserved can be
explained by the fact that wood is more perishable than metal and stone.
The Cyrillic alphabet is said to have been developed by the great Slavic Apos-
tles, Cyril and Methodius, with the specific idea of devising an alphabet that ade-
quately transcribes all the distinctive sounds of Slavic. Actually, there is some
question as to whether Cyril and Methodius really invented the Cyrillic alphabet.
Scholars now believe that they invented a similar alphabet, called Glagolitic,
which was in early use among the Southern Slavs. Whatever the details, however,
the creator or creators of the Cyrillic alphabet exhibited the same ingenuity as the
Greek and Semitic peoples before them. For instance, the Greek letter B, at this
point probably pronounced as a bilabial fricative [β], was split into two letters,
Б and B, to designate the two distinctive sounds b and v, respectively. The writing
82 Writing: Its history and its decipherment

system betrays its origin from the major literary tradition of Greek in which the
letter H was used for ē, which had come to be pronounced as [i], by using a symbol
derived from it as a vowel sign – И designating [i]. (The actual shape has changed
slightly to differentiate this letter from the sign H, evidently a later development
of earlier N, used to designate [n].)
Just as the Roman alphabet became the common currency of the part of medi-
eval Europe that embraced Roman Catholicism, so the Cyrillic alphabet became
the property of the more eastern, mainly Slavic, parts of Europe that adhered to
the Eastern Orthodox variety of Christianity. Most important was its use in Russia,
for through Russian domination over a large variety of non-Slavic peoples it has
come to be used for the languages of many of these peoples, too – as usual, with
appropriate modifications in response to linguistic differences.
There are many other offshoots of the Greek and Roman alphabets. One of
these is the Morse Code which substitutes different sequences of dots and dashes
(or short and long beeps) for the letters of the alphabet, as in . . . – – – . . . = SOS.
A curious parallel is found in the Ogham script of very early Irish texts. Here
lines of different lengths and numbers are drawn in one direction or another at
the edge of a stone memorial (or across the edge); see Illustration 29. Scholars of
writing are agreed that in principle, the symbols are a code version of the Roman
alphabet, very much in the same way as the Morse Code. What is less clear is the
principle for the different order in which the characters are arranged, compared
to the traditional Roman alphabet. One intriguing explanation starts with the
observation that the traditional names of the letters originally are tree names. In
the early Irish law texts, trees are divided into four classes which can be glossed
as ‘ordinary trees’, ‘chieftain trees’, ‘shrub trees’, and ‘bramble trees’. Interest-
ingly, the letter names in the four groups of the Ogham alphabet follow the same
system of classification, the first group having names of the ordinary tree family,
the second of the chieftain tree family, and so on.

Ill. 29: The Ogham alphabet

 History of writing 83

It is tempting to speculate that the use of tree names for letter names is a
cultural parallel to the early Germanic use of wood for runic writing. In fact, the
parallel can be extended further.
In the later, specifically Scandinavian period of runic writing, there existed
an alternative system of “feather runes” as in Illustration 30 below. In this case
we can discern a clear motivation for the code. Like the Ogham alphabet, the
runic alphabet is divided into groups. In early Germanic, these are the three
rows of characters in Illustration 28 above. Now, the number of left branches of
feather runes indicates the letter group in the runic alphabet, the number of right
branches, the position of the letter within that group.

Ill. 30: Scandinavian “feather runes”

The similarities between feather runes and Ogham symbols are remarkable enough
to suggest that one of the systems may have been influenced by the other. And the
fact that the feather runes can be clearly motivated as a code based on an ordered
alphabet makes it likely that the Irish Ogham symbols arose in a similar fashion,
except that in this case the underlying alphabet is not independently attested.
There is much independent evidence to indicate cultural exchange between
early Celtic and Germanic, with Celtic generally the donor and Germanic the
receiver (see Chapter 8, § 2 for an example). One might suspect that the parallel
between Ogham symbols and feather runes owes it origin to the same early Celtic
influence on Germanic. However, chronological and geographical considerations
make it difficult to substantiate this view. Perhaps the problem lies in the fact
that we are dealing with a later, more regionally limited phenomenon, not with
an instance of the above-mentioned early interaction between all of Celtic and all
of Germanic. In this regard, note the use of the term rūn- both in early Irish and
in Germanic – and only in these languages – to refer to secret wisdom that can be
conveyed by Ogham or runic characters. In Germanic the meaning of the term is
further extended to refer to the characters themselves.
Note, incidentally, that here we do have evidence in traditional lore that the
runes were associated with magic. But that magic may well be nothing more
84 Writing: Its history and its decipherment

remarkable than the glamour of this book’s introductory chapter – the power
attributed to literacy in a society where writing was limited to a chosen few.
So far, we only have traced the history of the Greek and Roman capital
letters. Lower-case letters are a later development and seem to have first arisen
in cursive writing. The original form of many of the letters required several strokes
of the pen (or whatever other writing instrument was used). Thus, three strokes
were needed for the letter A. Cursive writing favors fewer strokes. Illustration 31
shows how replacing the three strokes needed for the letter A by one gave rise to
the shape of the letter which we now classify as lower case. Similarly, replacing
the two strokes needed to write the letter T led to the (lower-case) “Insular” form
of t, found in medieval English and still used in Gaelic.

Ill. 31: Development of lower-case, cursive letters

Designating these early cursive letters as lower case actually required a further
step. It seems that the continued use of the original letter forms in more impor-
tant public inscriptions led to the reinterpretation of these letters as “more impor-
tant”, too. Consequently, the cursive letters were reinterpreted as less important.
By mixing the two types of letters, it was then possible to indicate the importance
of some words or even whole passages by writing them with the “more important”
letters or, to make things easier, by just letting them begin with one of these letters.
And once the distinction between capital and lower-case letters was introduced, it
was possible to put the distinction to new uses, such as the English use of capitals
to indicate proper names. (Note incidentally that the distinction between capital
and lower-case letters is by no means universal; many writing systems, such as
the indigenous systems of South Asia, do not make it.)
 The decipherment of ancient scripts 85

3 The decipherment of ancient scripts

While the literary traditions of Arabic, Hebrew, Greek, and Latin have to some
extent survived to the present day, together with their writing systems, the scripts
of ancient Mesopotamia, Persia, and Egypt, as well as many others around the
world, died out together with the cultures that employed them. Our knowledge
of these scripts results from the work of many scholars who succeeded in deci-
phering them.
Deciphering a script draws on techniques very similar to code breaking. And
like code breaking, it requires some idea of what is being expressed. It also helps
to know something about the language expressed by the writing system. But even
without this information, it is possible to make certain guesses on the nature of
the writing system, and these may help in the ultimate decipherment.
Given even relatively small amounts of texts, it is relatively easy to determine
whether a script is logographic, a syllabary, or alphabetic. Alphabetic scripts
usually do not have more than fifty different symbols; syllabaries may range from
about thirty to about 150; logographic systems tend to have thousands of charac-
ters. However, more specific clues are needed to identify the precise values of the
individual characters.

3.1. The decipherment of the cuneiform scripts. The first of the ancient scripts
to be deciphered was that of Old Persia. For a long time, travelers had returned
with reports about inscriptions found in the ruins of the ancient Persian empire
and about great inscriptions on isolated rocks from the same era. The tempta-
tion was great to attribute them to the rulers of that empire, whose names were
familiar from Greek sources: Darius, Xerxes, Cyrus, and Artaxerxes. However, the
cuneiform script of the inscriptions was uninterpretable.
The first steps in the direction of decipherment were made in 1788 by the
German scholar Carsten Niebuhr and, building on his findings, in 1802 by the
Danish scholar Frederik Münter. Their work suggested that the great inscriptions
contained three parallel texts, in three different writing systems. While all of
them were in cuneiform, the number of discrete symbols in the first text was
small enough to suggest an alphabetic system; that in the second was compat-
ible with a syllabary; while the third text had to be logographic. Münter also
plausibly argued that the first text had to be in the language of the Persian
empire, (Old) Persian, and must therefore be relatively close to Avestan, the
only ancient Iranian language known at that time. (See Chapter 2, § 3.10.1 on
the (re-)discovery of Avestan.) Finally, he suggested that certain recurring
symbol sequences mean ‘king’ and ‘king of kings’. Most of these suggestions
turned out to be accurate, except for the suggestion that the first text was
86 Writing: Its history and its decipherment

alphabetic. We now know that it is written in the Old Persian syllabary given in
Illustration 24.
The first actual decipherment of the Old Persian script was accomplished in
the same year as Münter’s work, 1802, by Georg Friedrich Grotefend, a German
high school teacher with virtually no knowledge of Avestan or other relevant lan-
guages, but with a lot of experience in breaking secret scripts. His initial assump-
tions about the nature of the text were very similar to those of Niebuhr and Münter.
Beyond that, he suggested specific readings for the first text, not only of titles like
‘king’ and ‘king of kings’, but also of the names of the kings. Many details of his
readings had to be revised. For instance, like Münter, Grotefend assumed that the
Old Persian text was written in an alphabet; he therefore analyzed the sequence of
symbols referring to Xerxes as xšharša, instead of the correct syllabic xa-ša-ya-a-
ra-ša-a (see Illustration 23 above). Nevertheless, his work opened the door for the
successful interpretation of the Old Persian inscriptions, which was undertaken
by an international succession of scholars, including Rasmus Rask (of Denmark),
Christian Lassen (of Germany), Henry Rawlinson (of Britain), and Jules Oppert
(of France).
Excavations in Assyria, undertaken in the early forties of the nineteenth
century, yielded a large number of inscriptions whose writing could be identi-
fied as identical to the third text of the great inscriptions of ancient Persia. Since
the message of these inscriptions had by now been sufficiently identified in the
Old Persian portions of the royal Persian inscriptions, the task of deciphering the
script was made easier. Still, it took nearly a decade before the British scholar
Rawlinson determined that the symbols of the script could have a broad range of
values (see Illustration 20 above) and that the texts could be interpreted under the
assumption that they were in a Semitic language.
Scholars like Rawlinson further discovered that in addition to the ancient
Semitic language(s) expressed in these inscriptions, there was a layer of vocab-
ulary that belonged to a very different language. This language came to be iden-
tified as Sumerian. And it became possible to learn more about this language
from a somewhat unexpected source. As noted earlier, reading and writing in
ancient Mesopotamia meant being literate in Sumerian, as well as in Assyrian
or Babylonian. In order to facilitate the learning of Sumerian, lists of grammati-
cal paradigms, dictionaries, and bilingual Sumerian-Babylonian texts had been
composed. As fragments of these documents became available, the knowledge of
Sumerian increased; and several grammars of Sumerian are now available.
In the early part of this century, excavations near Ankara (Turkey) led to the
discovery of the Hittite state archives which yielded a large number of written
records. The script used in these documents was the same as that employed for
Assyrian, Babylonian, and Sumerian. But large portions of the text clearly pre-
 The decipherment of ancient scripts 87

sented a different language – Hittite. The credit for having successfully identified
this language as Indo-European goes to the Czech scholar Bedřich Hrozný, who in
1915, during the First World War, published a paper on the “Solution of the Hittite
problem” in the journal of the German Oriental Society. The passage that was
instrumental in this finding is said to have been the following:

nu ninda-an e-iz-za-at-te-ni wa-a-tar-ma e-ku-ut-te-ni

Drawing on his experience with the often clumsy nature of cuneiform syllabary
writing, Hrozný was able to interpret the text as spelling out the message

nu nindan ezzatteni watar-ma ekutteni

Only one word in this passage actually was known, the logogram ninda ‘bread’.
However, it stood to reason that ‘bread’ is ‘eaten’ and that therefore there might
be a word meaning ‘eat’ in this passage, occurring near the word for ‘bread’. Of
the two words flanking ninda, ezzatteni contained an element ezz- [ets-] which
looked amazingly similar to the Indo-European root for ‘eat’, *ed- found in Lat. edō
‘I eat’, as well as in Engl. eat. Once this identification had been made, it became
possible to interpret nearly all the other elements of the sentence as Indo-Euro-
pean as well. Thus, the initial nu could be identified with *nu, the word for ‘now’
found widely in the Indo-European languages, including Engl. now. Similarly,
watar could be identified with Engl. water and its cognates in the other Indo-Eu-
ropean languages. Even the eku- of the last word can be identified as the element
ēb- found in the Latin word ēbrius ‘drunk’, the source of the modern English bor-
rowing inebriated. Given these identifications, then, it was possible to give a per-
fectly sensible interpretation to the hitherto nearly incomprehensible passage as
meaning

‘Now you eat bread, and you will drink water’

As in the case of Grotefend’s decipherment of Old Persian, many of Hrozný’s inter-

pretations have since been revised. But his identification of Hittite as Indo-Euro-
pean has proved to be correct.

3.2. The decipherment of ancient Egyptian. In 1799, during Napoleon’s expedi-

tion into Egypt, workmen digging fortifications near Rosetta came across a stone
bearing inscriptions in three different scripts. The lowest one of these was in a
language well-known to scholars – Greek. The topmost was in hieroglyphic char-
acters, a writing system which from ancient times had been known as the script of
88 Writing: Its history and its decipherment

ancient Egypt and which prominently figured on virtually all of the archaeological
remains. (The script in the middle turned out to be a cursive variant of the hiero-
glyphs; its decipherment played a less important role than that of the hieroglyphs
and the script is therefore ignored in the following discussion.)
Early attempts at interpreting the hieroglyphic texts operated under the
assumption that their script was ideographic, expressing in pictures complex
philosophical ideas; and fantastic interpretations were proposed which, by hind-
sight, were nothing short of outrageous. At this point it became clear that the
inscription found by Napoleon’s workmen, called the Rosetta Stone, provided the
key to deciphering the hieroglyphs, much as the Old Persian inscriptions were the
key for the decipherment of the cuneiform script of Mesopotamia.
The Greek text of the Rosetta Stone contained a number of well-known names,
including Ptolemaios (= Ptolemy) and Alexandros (= Alexander). Some of these
could be identified in the hieroglyphic part of the inscription. However, not much
else could be done, especially since the hieroglyphic portion had been heavily
damaged. Along the way, however, scholars identified Coptic, a language now
surviving as a liturgical language among Egyptian Christians, as a late descend-
ant of Egyptian. Drawing on such earlier work, a young Frenchman, François
Champollion, who at the age of eleven had decided to become the decipherer
of the hieroglyphs, succeeded in doing so in a series of publications beginning
in 1813. By demonstrating that certain suffixal elements (e. g. -f ‘he’, -s ‘she’) in
the hieroglyphic text recurred in contexts where the Greek text had personal pro-
nouns he was able to identify the meanings of these elements. This made it pos-
sible to identify these elements with Coptic suffixes of similar meaning and thus
to determine an approximate pronunciation. As a consequence, the number of
symbols with identifiable values increased, and substituting the same values in
other passages made it possible to read additional words. Along this route, then,
he was able in 1822 and 1824 to propose a decipherment of the hieroglyphs which,
in spite of later revisions in detail, made it possible to read the texts of ancient
Egypt and to understand the nature of the writing system.

3.3. Other decipherments. A number of other writing systems of the ancient

Near East and neighboring areas, as well as the script of the Mayan kingdoms
of Meso-America have been deciphered by employing similar methods. In the
ancient world, these include Elamite, the language of the non-Indo-European
predecessors of the ancient Persian rulers, as well Hieroglyphic Luwian and a
number of other languages more or less closely related to Hittite.
The perhaps most amazing decipherment, in 1952, led to the discovery that
centuries before the appearance of alphabetic writing in ancient Greece, Greek
had been written in a very different, syllabic script called “Linear B”. In 1900,
 The decipherment of ancient scripts 89

the British archaeologist Arthur Evans had discovered in Knossos, on the island
of Crete, the remains of the archives of the Minoan civilization. In the documents
found there, Evans was able to distinguish three phases: (i) An early stage (about
2000–1600 BC) using a pictorial writing system; (ii) a later stage partly over-
lapping with the preceding one (about 1700–1550 BC) in which the characters
were simplified to their barest outlines, whence the name Linear A; and (iii) a
third variety, Linear B, also starting around 1700 BC, whose characters, though
similar to Linear A, differed in many important details. Linear B died out around
the twelfth century BC, together with the Minoan civilization, apparently under
the onslaught of newcomers. Although possibly Greeks too, the new arrivals did
not take over the script they encountered. As a consequence, the Greeks returned
to illiteracy for some 300 years, until contact with the Phoenicians reintroduced
writing (see § 2.5 above).
The decipherment of Linear B and the identification of its language as an
early, regional form of Greek came as a great surprise even to the decipherers. The
common opinion had favored just about any language other than Greek, including
Etruscan or even some ancient relative of Basque. There was little motivation for
looking to Greek as a possible key to breaking the code. Things were made even
more difficult by the fact that not even a trace of a bilingual text could be found.
Only one thing was certain: The number of distinct characters was too large for
the script to be alphabetic, and too small to be logographic; it had to be a sylla-
bary. Attempts were made to use the computers then available to help in breaking
the code. Exhaustive lists of symbols were established, as well as of the various
combinations into which they entered. It was possible to show that certain symbol
groups recurred, suggesting possible lexical units. Moreover, some symbol groups
recurred only partially, with different symbols following them under what seemed
to be syntactically different conditions. This suggested that the language of Linear
B had a system of roots followed by something like inflectional endings. In addi-
tion, attempts were made to randomly assign different phonetic values to different
symbols and to examine whether the resulting structures bore any resemblances
to forms of known languages.
The breakthrough came in 1952 when a British architect and amateur lin-
guist, Michael Ventris, was able to demonstrate that certain phonetic substitu-
tions yielded results which could be read as place names in ancient Crete, such
as ko-no-so, which would be a rather standard way of spelling the name Knossos
(ancient Greek Knōsós) in a syllabary. Substituting the same phonetic values in
other contexts made it possible to read other words, and slowly it became clear
that these words – and their structure – did not just sound similar to place names
and other words of ancient Greek, they were ancient Greek. For instance, beside
ko-no-so, a putative adjective form of the same word ko-no-si-jo ‘of Knossos’ could
90 Writing: Its history and its decipherment

be identified, and this corresponded magnificently to the Greek adjective Knōsío-.

Because of its association with Mycenae on the Greek mainland and to distin-
guish it from the earlier Minoan tradition, this form of Greek has been named
Mycenaean.
As in all the other decipherments, further work suggested modifications of
Ventris’s original interpretations. But again, as in the other successful decipher-
ments, the basic breakthrough had been made.
As a postscript it might be mentioned that attempts to decipher Linear A by
means of the same methods so far have not yielded any satisfactory results. One
scholar, Cyrus Gordon, claims that Linear A was used to write a Semitic language;
but his claim does not seem to have been accepted so far in the scholarly commu-
nity. Perhaps the language written in that script has left no known descendants.
The situation is similar for many other scripts, ancient and relatively modern,
which like Linear A has so far resisted satisfactory decipherment, in spite of
repeated claims by different scholars that they have been able to break the code.
The problem with such claims is that they are usually subscribed to only by those
who propose them.
The most recent breakthrough, the decipherment of the Mayan script, is
important for a number of reasons. First, it took place in the face of preconceived
notions that, except for numerical information, largely of calendrical and astro-
nomical nature, the Mayan hieroglyphs did not present real writing, but were
“ideographic”, expressing complex ideas in picture form, much like the pictorial
representation in Illustration 4 above. These preconceptions, similar to pre-deci-
pherment views on the Egyptian hieroglyphs, had the backing of the most author-
itative scholars in the field, who ostracized and ridiculed occasional attempts
(generally by amateurs or linguists) at interpreting the Mayan hieroglyphs as
representing genuine writing. Underlying this attitude were two factors. One was
the theory by the then greatest authority on writing, the American scholar I. J.
Gelb, that true writing only originated once, in the Ancient Near East, and that
all other forms of writing spread from this original source through diffusion. (As
Mayan scholars realized, there was no credible evidence for cultural diffusion
from Eurasia to Meso-America.) The second factor, closely allied to the first, was
a Eurocentric, or more correctly “Old-World-centric” view which linked the devel-
opment of writing to “civilization”, which believed “primitive” people only to be
able to imitate the writing of “civilized” people, and which – of course – consid-
ered the indigenous peoples of the Americas to be “primitive” compared to the
“civilized” people of the “Old World” of Eurasia and Northern Africa.
The credit for having laid a solid foundation for the eventual decipherment of
the Mayan script goes to a 1952 article by the Russian scholar Yu. V. Knorosov, who
showed that like the traditional scripts of the Ancient Near East and Egypt, Mayan
 The phonetic interpretation of written records 91

writing represented a mixed code with logographic characters with or without

phonetic indicators as well as syllabic signs; see Illustration 19 above. Although
many details still need to be worked out, a loosely organized international team
of linguists, archaeologists, and specialists in writing systems has since then suc-
ceeded in unraveling many of the mysteries of Mayan writing and, in the process,
has shown that the Mayan hieroglyphs are indeed true writing, not just a quaint,
“primitive” ideographic code. Moreover, although much of the information con-
veyed by Mayan writing is indeed of calendrical and astronomical nature, there
are also reports of historical events and personalities. Finally, recent research
suggests that Mayan writing continues an older tradition of the Olmecs, an earlier
Meso-American civilization probably speaking a different language.
The significance of the Mayan decipherment is obvious. First, it casts serious
doubts on such preconceived notions as “Old World” = “civilized” vs. “New World”
= “primitive”. Clearly, the development of advanced civilizations which find it
useful to develop writing is possible among all human beings, in all areas of the
world. Secondly, the fact that the Olmecs and Mayans evidently developed true
writing independently from the “Old World” has brought about serious rethinking
about the development of writing in the “Old World”. Whereas Gelb had argued
for a single origination of writing, in ancient Sumer, and for diffusion of writing
from Sumer to the rest of Eurasia and North Africa, we now have to admit the
possibility that writing originated independently, in a number of different socie-
ties, as these societies developed economic, social, and cultural structures that
were conducive to the development of writing as a more effective means of record
keeping and communication.

4 The phonetic interpretation of written records

One of the questions that arises in dealing with ancient scripts is “How do we know
how the symbols of this system were pronounced?” In fact, given the vagaries of
English spelling (see Chapter 1, § 5), we might have great doubts as to whether
it would ever be possible to assign consistent – or even inconsistent – phonetic
values to written symbols. Fortunately, many of the “inconsistent” English spell-
ings are reflexes of earlier linguistic stages. (Compare the initial k of words like
knight, or the gh of the same word, both of which reflect sounds – [k] and [x] – that
were actually pronounced in earlier English.) In the earliest stages of writing in a
particular language, there is no earlier history which might motivate such histor-
ical spellings. In this regard, at least, early writing systems tend to be free of the
inconsistencies of languages like English.
92 Writing: Its history and its decipherment

Even for modern English it would be possible to come up with pretty good
phonetic interpretations – if, say, as the result of some major disaster, English
had died out and a scholar of the twenty-third century were trying to decipher
documents dug up from the ruins of our society. In fact, putting ourselves into the
position of that scholar may serve to demonstrate how people go about interpret-
ing the phonetic values of ancient scripts.

4.1. Determining the nature of the script. Investigators would very quickly be
able to determine that there are two different sets of written characters in English,
which rarely combine with each other. On one hand there are characters like b,
r, e, on the other, 1, 2, 3, as well as $, £, etc. The fact that members of the set 1,
2, 3 appear consecutively at the top or bottom of pages would make it easy to
recognize these as logographs designating numerals. And the fact that symbols
like $ and £ combine mainly with these numerals suggests that these symbols are
logographs, too, referring to some kind of “operators” or “classifiers” related to
the numerals.
These and other considerations would suggest that only the members of the
set b, r, e, etc. qualify as writing symbols. Moreover, the limited number of char-
acters would suggest that the writing is alphabetic.
Within the set of writing symbols, it would be possible to isolate a subset e, a,
i, u, o and another subset s, d, r, t, etc., which differ from each other in terms of their
combinability: Members of the second set can enter into more complex combina-
tions (such as str, sts, rts), while e, a, i, u, o are more limited in their combinations
(we find ei, ie, ai, ia, oi, io, etc., but combinations like eia or oae would be difficult
to locate). A scholar familiar with general tendencies in linguistic structure would
be able to conclude from this and other, similar information that the set e, a, i, u, o
designates vowels, while the set s, d, r, t, etc. characterizes consonants. Moreover,
an observant investigator would note that one letter, y, sometimes behaves like a
vowel (as in my), and sometimes as a consonant (as in your).

4.2. Beginning to crack the code. The major task, now, would be to assign spe-
cific values to the vowels and consonants of the script. And this is where things
get to be much more difficult. The fact that across human languages, dentals tend
to be the most frequent consonants may suggest that the most frequent English
consonants, s, d, r, and t, are dentals. Perhaps it would be possible to make a few
similar guesses concerning other sets of symbols. But this still leaves open the
question of which consonant symbol designates which consonant, not to speak
of the values of the vowel symbols. The fact that the same sounds may be spelled
in many different ways in English would only add to the problems faced by the
investigator.
 The phonetic interpretation of written records 93

A way to deal with the latter problem is to look for evidence that might estab-
lish that certain spellings, though using different symbols, refer to the same
sounds. One area of evidence helpful in this regard consists of misspellings or
variant spellings, such as nite (for night), insure beside ensure, plough beside plow.
Another area which an experienced investigator might look to is poetic lan-
guage, which frequently draws on phonetic similarities or identities as the foun-
dation for creating poetic lines. In English, poetic texts are relatively easily iso-
lated, since each line is treated as if it were a paragraph. Moreover, even cursory
examination would show that in many poetic texts, the words at the end of neigh-
boring lines (or of alternate lines) tend to end in the same spelling. This would
suggest that English uses the principle of end rhyme. If in poetic texts we find
not only end rhymes like ring : king, but also rite : night etc., this evidence would
provide further support, in addition to variant spellings like night : nite, for the
phonetic equivalence of the spellings ite and ight, at least in some words. And
given enough patience or a good enough computer, it would be possible to estab-
lish a large set of such “equivalent” spellings.
Of course, English poetry offers occasional examples of “eye rhymes”, words
that are permitted to rhyme because they are spelled the same, even though
sounding different; e. g. bomb : womb : comb. But the special nature of these
rhymes can be established by noting that, say, womb elsewhere rhymes with
groom, doom, bloom, etc., while comb rhymes with home, gnome, etc.; but groom
etc. cannot rhyme with home etc.
In this manner it is possible to establish sets of letter combinations likely to
express the same pronunciation and to contrast them with other sets with dif-
ferent pronunciation. However, this still leaves open the exact pronunciations
expressed by these spellings.

4.3. Establishing phonetic values. To more firmly establish the phonetic identity
of given spellings it is necessary to draw on evidence beyond the writing system
and its nature. Such evidence may come from at least two types of sources.
We may find in our texts statements by indigenous grammarians about the
nature of their sound system and the relationship between sound and spelling.
Statements of this sort may be highly accurate, but they may also be quite vague
and unhelpful. In English, for instance, depending on the texts we might find,
the difference between the vowels in bit and bite might be described phonetically
accurately as [i] vs. [ay] or in a phonetically misleading manner as “short i” vs.
“long i”. The situation is similar in the pre-modern world. On the one hand are
the Sanskrit phoneticians who made very detailed and accurate phonetic obser-
vations and thus were able to distinguish consonants such as b, bh as voiced from
voiceless consonants such as p, ph. On the other hand there is the traditional
94 Writing: Its history and its decipherment

western approach, where for instance in German, b, d, g are called “soft” vs.
“hard” p, t, k, without any phonetically verifiable definition of these terms.
The investigator may be lucky enough to come across modern English texts
that provide an accurate description, not just of the sound system, but also of the
many different ways in which the sounds of that system are spelled. But what if
such texts have not survived or if the quality of the texts that have survived is not
very high?
In many cases, another approach is available by examining borrowings:
For instance Engl. strike has been borrowed as Streik in German and as στράικ
(= straik) in Greek. Now, in both languages, the borrowed words are spelled with
an orthographic diphthong, ei in German and ai in Greek (where it is specifically
characterized as a diphthong by the accent over the α). Evidence of this sort sug-
gests that Engl. strike is pronounced with a diphthong. Given enough patience
it would further be possible to determine that in one or more of the borrowing
languages the “fit” between spelling and pronunciation is much closer than in
English. If in addition one or more of the languages offers texts by indigenous
grammarians that provide reliable information on their sound system and its rela-
tion to spelling, the investigator can begin to assign specific phonetic values to
particular English spellings. For instance, if it can be determined that German ei
regularly spells a diphthong [ai], then the spelling Streik suggests that Engl strike
is pronounced with a similar diphthong, [ai].
In this way, a good amount of knowledge about the pronunciation of written
records may be gained. However, some details can be known only in those rare
cases where indigenous grammarians’ descriptions are very meticulous and reli-
able. Elsewhere we may well miss out on such fine details as whether a t is really
dental (as in French or Spanish), or post-dental (as in English).
More than that, a fair number of cases remain where even the experts cannot
agree on the phonetic interpretation of certain letters or letter combinations. This
is the case, for instance, for the Old English “digraphs”, eo, ea, etc., which by some
are claimed to have been diphthongs, by others to have been monophthongs with
a phonetic value intermediate between the sounds normally designated in Old
English by the individual vowel letters. Fortunately, such cases of uncertainty are
rather rare in most of the early Indo-European languages, which form the basis
for the most extensive – and intensive – study of linguistic change. Moreover, the
occasional cases of uncertainty cause no major difficulties – as long as we do not
draw on them to build grandiose theories about the nature of language change.
 Writing in the rest of the world 95

5 Writing in the rest of the world

Up to this point, we have been focusing mainly on writing systems in the Middle
East and Europe, though there have been references to writing elsewhere, espe-
cially to the Mayan hieroglyphs and the significance of their decipherment for
theories on the origin of writing. Writing, however, has a much wider distribution,
both in time and in space. Examining the development of writing elsewhere in the
world therefore can provide a valuable balance.
In particular, writing has a long history in East Asia, where there are three
major writing traditions: Chinese, Japanese, and Korean. The logographic Chinese
system and the phonetically based Korean system present characteristics that
allow them to be called unique; they are therefore discussed in some detail. The
Japanese system, on the other hand, is not examined here, since in effect it com-
bines the two different approaches of Chinese and Korean. The East Asian systems
we examine confirm some of the principles of the development of writing systems
that we have seen before, but they also provide new insights. The case is similar
for Brahmi, the writing system from which all the indigenous writing systems of
India are historically derived.
In addition, a brief look at some indigenous writing systems in America and
Africa can shed important light on how such systems can emerge.

5.1. The Chinese system. Chinese writing is the oldest known system in East
Asia, appearing in the fourteenth century BC, during the late Shang dynasty, with
a few small texts coming from slightly earlier periods. The system used at the time
of first attestation is already considerably elaborated, suggesting an extended
period of earlier development which, depending on one’s estimate, may have
lasted for 400 to 700 years.
The earliest Chinese writing is found in so-called oracle texts, responses by
oracular interpreters to questions posed about future events (similar to our horo-
scopes). Most of the texts were written on animal bones and tortoise shells.
The writing system at that stage bears strong similarities to the early systems
of the Near East. Pictorial symbols are used as logographs, with the possible
addition of phonetic and semantic indicators. And as in ancient Mesopota-
mia, the shape of the symbols, traditionally called “characters”, changed over
time, in response to the materials that were used for writing. Illustration 32
gives an example of an early stage of these changes. The change in letter shape
was greatly accelerated once brush and ink came to be used. It has been esti-
mated that one cannot guess the meaning of a single modern character from
its form, in contrast to the transparent relation between meaning and form
in Ancient Chinese. In this respect, too, the development of Chinese writing
96 Writing: Its history and its decipherment

is similar to that in the Ancient Near East and its later abjad and alphabetic
successors.

Ill. 32: Changes in a Chinese character

But as anyone familiar with Chinese can readily tell, in one important respect
Chinese writing followed a very different course from the writing systems farther
west. Even to the present day, Chinese writing has remained essentially logo-
graphic (with the possible addition of phonetic and semantic indicators). The
main reason for this different development also is quite obvious to anyone who
has any familiarity with Chinese. As the common wisdom goes, Chinese is a
“monosyllabic language”; that is, most Chinese words and other meaningful ele-
ments consist of just one syllable. Although there are exceptions, generally words
are syllables – and syllables are words. In contrast to other languages, therefore,
it makes no sense in Chinese to distinguish between word symbols (logographs)
and syllabic symbols; there is no difference to begin with. Since the overall
structure of Chinese has remained pretty much the same as it was in ancient
times, it is not surprising that the writing system has retained its essentially
logographic nature to the present day. In fact, Chinese writing is the only major
logographic system now in use, with a set of between 1,850 and 4,000 distinct
characters.
Using a logographic system with so many different symbols has definite dis-
advantages in the modern world. For instance, for telegraphic transmissions, a
special numerical code had to be devised for the representation of each graph,
and Chinese typewriters were enormously complex, and therefore slow and
clumsy to use. The twentieth century saw several attempts to radically reform
Chinese writing by adopting, and adapting, the Roman alphabet of the west.
The efforts at introducing an alphabetic system, however, have not succeeded.
One reason is that logography is felt to be an integral part of Chinese culture and
identity. But there are other, more practical reasons as well.
The most important practical reason is that the logographic system has the
advantage of bridging gaps of communication within Chinese. Although it is cus-
 Writing in the rest of the world 97

tomary to speak of “the” Chinese language, in fact there are numerous Chinese
languages (often, but erroneously, called dialects) which in their spoken form
are mutually unintelligible. Use of a common logographic system makes it possi-
ble for these speakers to communicate in writing, no matter how differently they
may pronounce the characters. As a parallel in societies with alphabetic writing,
compare the use of numerical symbols, such as 2, which can be read by anyone,
no matter whether the symbol is pronounced [tū] (English), [dü] (Albanian),
[kaksi] (Finnish), [-wili] (Swahili), or [iki] (Turkish).
In fact, the use of Chinese writing to communicate across different languages
is not limited to Chinese; speakers of several other East Asian languages employ
Chinese characters for the same purposes, including the Japanese and the (South)
Koreans. This is because Chinese writing spread far beyond its original bounda-
ries, along with many other aspects of Chinese culture and civilization, as well
as Chinese words. Even where different writing systems developed, they did so
under the influence of Chinese writing and generally came to coexist with Chinese
characters which continued to be used for a massive number of words borrowed
from Chinese.

5.2. Writing in Korea. The most striking case in point is Korean. From about
600 AD, the people of Korea began to use an adaptation of the Chinese writing
system to write their own language. Korean, however, is a language very differ-
ent from Chinese. Whereas Chinese is “monosyllabic”, Korean words tend to be
complex, with roots followed by strings of suffixes. Chinese logography was thus
not particularly well suited for writing Korean, although South Koreans still use
it to write Chinese words borrowed into Korean. To write native Korean words,
Chinese writing has generally given way to a system said to have been designed
specifically for Korean by an enlightened monarch, King Sejong, who ruled in
Korea from 1418 until 1450.
The system that King Sejong invented (or perhaps, had a committee of schol-
ars invent under his direction) was completed in 1444 and has come to be called
Han’gŭl (meaning ‘great script’). In principle it is an alphabet, with 28 letters,
but it differs from ordinary alphabets in its organization. There are no separate
symbols for all of the distinctive sounds of the language. Rather, there is a set of
consonantal and vocalic “base symbols”, and a set of phonetically based dia-
critics that differentiate, or “derive”, the symbols for other sounds from these
base symbols. For example, Korean has a set of “lax” stops (produced with
relatively reduced muscle tension), which can be somewhat imperfectly tran-
scribed as [t], [č], etc., and a corresponding set of aspirated “tense” stops (pro-
duced with relatively greater muscle tension), such as [th] or [čh]. Han’gŭl base
symbols are used to represent the lax stops, while the tense stops are written by
98 Writing: Its history and its decipherment

modifying the corresponding base symbols by means of a superscript horizon-

tal line (which in the case of [th] is connected to the base symbol by a vertical);
see Illustration 33.

Ill. 33: Some Han’gŭl symbols

The Han’gŭl system is a remarkable achievement showing considerable originality

and linguistic insight. Contrast, for instance, the lack of phonetic sophistication
in a system such as our writing system, with the vowels (a, e, i, o, u) scattered hel-
ter-skelter through the alphabet, the consonants similarly sprinkled without pho-
netic rhyme or reason (voiced b followed by voiceless c and separated by a long
interval from its own voiceless partner p). While there have been some changes in
spelling conventions since its invention, Han’gŭl’s place in the history and typol-
ogy of writing systems in the world is secure – as is its place in Korean life. Until
recently, for instance, the ninth of October was a Korean national holiday, cele-
brating the invention of Han’gŭl!

5.3. Brahmi and the writing systems of India. Different explanations have
been proposed for the sources that may have inspired the creation Han’gŭl. Was
it a writing system used outside Korea, and if so, which of the many candidates
should we take seriously? Since Han’gŭl does not bear any convincing similarities
to outside writing systems, this explanation does not have much to recommend
it. Or was there a more indirect inspiration, the fact that together with its script,
China must also have transmitted to Korea its knowledge of phonetics which
then could provide the basis for the amazing phonetic sophistication of Han’gŭl?
This is the more attractive hypothesis, since it makes it possible to attribute the
phonetic sophistication of Han’gŭl to a prior sophistication in phonetics – while
(most) other writing systems lack such sophistication because they were created
without prior familiarity with phonetics.
Attentive readers may have noted the parenthetical “most” in the last sen-
tence of the preceding paragraph. As it turns out, there is in fact another tradition
of writing (other than modern phonetic transcription systems) that likewise is
phonetically sophisticated. This is the Brahmi of the India of Emperor Ashoka
(3rd c. BC) and its later descendants, Devanagari (used for Hindi, Marathi, and
now generally for Sanskrit), Tamil script (for Tamil, of course), and all the other
indigenous writing systems of the Indian subcontinent.
 Writing in the rest of the world 99

Here, too, different explanations for the origin of the script have been pro-
posed, and the precedent (or, if you will, post-cedent) of Han’gŭl may help in
choosing between these and thus provide a first step to unraveling the mystery of
Brahmi’s origin.
The explanations offered so far range from derivation from the West Semitic
writing systems that gave birth to the alphabet, to descent from the writing of
the Indus Civilization (ca. 3000–1700 BC), to stimulus diffusion inspired by the
writing of the Persian Empire, but without directly copying any of the symbols of
that writing.
While the first set of explanations may seem to be exciting from a western
perspective, because it links Brahmi to our own writing systems, it suffers from
the same problems as the similar attempts at outside derivation for Korean.
The second explanation would be especially interesting to those in India who
would like to situate themselves in an unbroken tradition from the beginnings of
the Indus civilization to the present. However, the explanation is highly problem-
atic at this stage of our knowledge. First, the Indus script has not been successfully
deciphered; deriving Brahmi from the Indus script thus would amount to explain-
ing the unexplained by the even less explained. Second, there is absolutely no
evidence of writing in India for some 1000 years after the demise of the Indus Civi-
lization. Even the latest phases of the Indus Civilization no longer employ writing.
And the elaborate system of preserving the Vedic texts by means of multiple text
versions (see § 1 above) makes sense only in an oral tradition, not in a written one.
The third account looks most promising, especially if we assume that Emperor
Ashoka (or his father or grandfather) commissioned Brahmins who were spe-
cialists in the long and sophisticated tradition of Sanskrit phonetics to devise a
writing system for his proclamations, for this would explain the great phonetic
sophistication of the Brahmi system. As in the case of Han’ğul, we can account
for this highly unusual sophistication in terms of a pre-existing sophistication in
phonetics. Moreover, just as the Korean writing system was indirectly inspired by
prior knowledge of another writing system – that of Chinese – so it has been pro-
posed that Brahmi was a similar product of stimulus diffusion. Under this view,
some of the scholars and artisans of the Persian Empire found refuge in India after
the destruction of the Empire by Alexander the Great, bringing along with them
the idea of writing, as well as a tradition of using that writing for monumental
rock inscriptions proclaiming the great deeds of the emperors.
The phonetic sophistication of Brahmi (and its descendants) shows itself
most obviously in the phonetically inspired order in which the characters are
arranged – vowels first (in short-long pairs, where applicable); then the stop con-
sonants (including the nasals) ordered according to place of articulation, starting
from the velar region; and so on. See Illustration 34.
100 Writing: Its history and its decipherment

(Some of the nasals are not used at this point as yet. – The vowels ē, ō are placed after the other
vowels because they function like diphthongs.)

Ill. 34: The vowels and stop consonants of Brahmi

Interestingly, while Brahmi and its descendants at first sight look very much like
alphabets, with distinct symbols for vowels and consonants, they share with
Han’gŭl the fact that they draw vowels and consonants together into single, syl-
labic symbols, by employing diacritics for vowels that can be attached to conso-
nants. Moreover, all consonant symbols, unless specially modified, denote the
consonant plus a short ă. See Illustration 35, using the letter = pa.

Ill. 35: Some consonant-vowel combinations in Brahmi

What is especially striking is the manner in which the vowel symbols are attached.
The high vowels are placed highest, the back vowels the lowest, and the a-vow-
els somewhere in between. While this does not correspond to the articulatory
differences between the vowels, it does capture their acoustic or auditory effect.
High vowels do in fact sound higher, and back vowels lower, with a-vowels in
between. Since the Sanskrit phoneticians had developed a theory focusing on
auditory phonetics, it is likely that this “iconic” placement of the vowel diacritics
is not just accidental.
 Writing in the rest of the world 101

As in other parts of the world, different writing materials, the effects of cursive
writing, and different ideas of aesthetics led to regional differentiations of the
writing system. These developments can be illustrated focusing on the develop-
ment of Devanagari, the most widely used writing system of the north, and Tamil
script, used for the southern Indian Tamil language.
See above all the development of the Brahmi letter = ka in these two scripts,
documented in the left column of Illustration 36. The original Brahmi character
required two strokes (a); drawing together the two strokes led to the next stage
(b); subsequent developments involve incorporation of a top line (from which
letters were suspended), hence the Devanagari character (c); and yet further mod-
ifications account for the Tamil outcome (d). The other columns of Illustration 36
document changes in other characters, but skip the intermediate stage(s). In some
cases it is relatively easy to see the historical relationship (e. g. for a), in others it
will take a hefty dose of ingenuity to see such a relationship (e. g. i). In the latter
case, it is only because of the very rich corpus of intervening stages that we are
able to trace the development.

Ill. 36: Developments from Brahmi to Devanagari and Tamil script

Over the course of two-and-a-half millennia, Brahmi thus came to be differen-

tiated in such a remarkable fashion that, without knowing the earliest and the
intermediate stages, the relationship between its descendants would be quite dif-
ficult to see.
In fact, the diversification of Brahmi goes even farther; for the script spread
beyond the Indian subcontinent, both north and southeast. In the north, a stage
of writing intermediate between Brahmi and Devanagari was adapted to writing
Tibetan, first for Buddhist texts and then as a general means of written communi-
cation. The spread of writing to the southeast, including Thailand, Kampuchea,
and Java (Indonesia), was propelled both by the expansion of Buddhism from East
and South India and by the spread of trade relations with the same area.
102 Writing: Its history and its decipherment

5.4. Writing elsewhere. While the development of Mayan writing of Meso-Amer-

ica shows that writing can develop independently in different parts of the world,
and yet with remarkably similar results (if we disregard differences in the shape
of the letters), Korean and Brahmi demonstrate that writing can also spread by
stimulus diffusion, and that the results of the spread can differ remarkably from
the original.
There are several similar examples in nineteenth-century North America and
Africa of the creation of writing systems under stimulus diffusion. Particularly
well known in North America is the syllabary that Sequoyah (1770?-1843) devel-
oped for writing his native Cherokee (an Iroquoian language). He had learned of
writing from contact with Europeans, especially English speakers, although he
could not read or write English; and he set out to develop a writing system for
Cherokee. After experimenting with a logographic system, he hit on the idea of
a syllable-based graphic representation, and as the basis for his syllabary took
some of his symbols from English books that he had seen. Since he did not read
English, he was not bound by the alphabetic values of the letters in English, but
could use them in completely novel syllabic values; for instance, H stood for [mi],
h for [ni], and M for [lu]. Other symbols were simply invented. The result was an
initial syllabary of some 200 signs which Sequoyah later simplified to 85. The
syllabary was used for publishing books and newspapers in Cherokee, and it is
still in use today.
Syllabaries were developed to write other Native American languages also,
some by missionaries in the nineteenth century and some possibly native adapta-
tions (via stimulus diffusion). One that has achieved rather wide use was invented
by the English missionary J. Evans, around 1840, to write Cree, an Algonquian
language spoken across much of Canada but mainly in the western provinces.
This syllabic script utilizes geometric symbols such as small circles, angles, tri-
angles, some oriented in a variety of positions, some together with various dia-
critical marks, to spell all the consonant-vowel syllables of Cree; the symbol >, for
instance, stands for [po], and a superscript dot marks all syllables with an a-vowel.
Cree syllabics are still used, and in fact their use spread in Canada; Evans’s syl-
labic script was adapted, for example, for the writing of Slave, an Athabaskan
language spoken in Canada’s Northwest Territories, and the Eastern Arctic Eskimo
spoken on Baffin Island, also in the far north of Canada (a Latin-based alphabet is
used for other Eskimo dialects, including Greenlandic Eskimo).
Nineteenth-century Africa witnessed the creation of several similar indige-
nous writing systems, especially in West Africa. One assumes that these systems,
too, arose under stimulus diffusion, but the role of contact with Europeans is not
always clear. Noteworthy is the Vai system, developed prior to 1848 by Bukele, a
speaker of Vai. As with Cherokee, the Vai system was at first logographic but soon
 Writing in the rest of the world 103

developed a syllabic basis, with some syllabic signs being taken from logograms
for monosyllabic words. Other syllabic systems in West Africa that emerged in
roughly the same period, such as the Mende syllabary invented by Kisimi Kamala
or the Toma syllabary used in part of Liberia, seem to have been stimulated by
the Vai system.
An interesting parallel to King Sejong’s invention of Han’gŭl is found in the
Bamum syllabary of Cameroon, which was invented by a local ruler named Njoya
(with some help apparently from a European missionary). Bamum is especially
interesting since it developed some alphabetic principles, thus showing the inde-
pendent development of an alphabet out of a syllabary, parallel to the develop-
ment of alphabetic writing in Semitic.
In spite of all their differences, these writing systems show patterns of devel-
opment that are remarkably similar to developments in the Near East and Korea
and thus bear testimony to the fact that, given similar circumstances, human
beings tend to respond in very similar ways.
Chapter 4: Sound change
Etymology is a science in which consonants count for little,
and vowels, for nothing at all.
(Statement attributed to Voltaire, probably apocryphal.)

Sound change, in so far as it takes place mechanically, takes place

according to laws that admit no exceptions.
(Osthoff and Brugmann, 1878)

1 Introduction
The early days of comparative Indo-European linguistics concentrated heavily on
studying the similarities and differences in word structure in the Indo-European
languages. This line of investigation was a continuation of earlier scholarship
which predated comparative Indo-European linguistics. Special attention was
given to attempts to derive all noun and verb endings from earlier independent
words which were said to have fused with the preceding noun and verb stems.
For instance, the -s- appearing in forms like Gk. lū́-s-ō ‘I will loosen’, a marker of
future tense, was claimed to be related to the s of the root es- ‘be’. Similarly, the
-dēdun of the Gothic past tense (as in nasidēdun ‘they saved’) was considered to
derive from the root underlying modern Engl. do. Some of the proposed ideas have
some merit, such as the derivation of -dēdun; but even this is still controversial.
Many other ideas turned out to be premature and, in hindsight, naive.
They were naive especially because they were proposed without a proper
understanding of linguistic change, particularly of the way in which sound
change operates.

2 Grimm’s Law
A major breakthrough in comparative Indo-European linguistics came when
the Danish scholar Rasmus Rask and, following him, the German linguist Jacob
Grimm, began to take a closer look at the relationship between the Germanic lan-
guages and the rest of Indo-European. Recall that William Jones, in his famous
pronouncement of 1786, had hedged his bets as to whether Germanic (designated
by the term Gothick) was related to Sanskrit or not:

https://doi.org/10.1515/9783110613285-004
 Grimm’s Law 105

… there is a similar reason, though not quite so forcible, for supposing that … the Gothick …,
though blended with a very different idiom, had the same origin with the Sanscrit …

Jones’s reason was that Germanic looked very different from the classical lan-
guages, Greek, Latin, and Sanskrit, especially in the way it was pronounced. For
instance, where the classical languages had voiced stops, as in Gk. édomai, Lat.
edō, Skt. ádmi ‘eat’, the Germanic languages had voiceless ones, as in Engl. eat,
Goth. itan, or even sibilants, as in Germ. essen. At the same time, some Germanic
words seemed to preserve the voiced stops of the other Indo-European languages,
such as Engl. day, corresponding to Lat. dies ‘day’; but again, German differed by
offering a voiceless stop in its cognate, Tag ‘day’. It was perhaps this inconsist-
ency in the way Germanic corresponded to the classical Indo-European languages
that led Jones to talk about Germanic being “blended with a very different idiom”.
The purpose of Rask’s and Grimm’s work was to elucidate more clearly the
relationship between Germanic and the classical Indo-European languages and
to show that Germanic was in fact part of the Indo-European language family. To
this end, Rask and Grimm conducted thorough investigations into the nature of
precisely those aspects which appeared to make Germanic quite “alien”, namely
the differences in pronunciation.
The result of the work, published in 1818 and 1819, was twofold. First, the
work succeeded in establishing once and for all that the Germanic languages are
part of Indo-European. Secondly, it did so by providing a brilliant account for the
differences between Germanic and the classical languages in terms of a set of
amazingly systematic sound changes, and a similar set of sound changes differ-
entiating German from the rest of Germanic.
To simplify matters, let us concentrate on the sound changes differentiating
all of Germanic from the rest of Indo-European. The discovery of this set of system-
atic changes has been so influential in the development of historical linguistics
that the name soon attached to it, Grimm’s Law, has become a stock expression
for everyone interested in language change and linguistic relationship. The name
actually is a misnomer. The credit for discovering the systematic correspondences
between Germanic and the classical languages must go to Rask. However, Grimm
was so successful in formulating the changes – and in marketing them – that he
received the recognition of having the “law” named after him, at least outside
the German-speaking countries. (Note expressions like Grimm’s Law, Fr. le loi de
Grimm.) In German, the law is more commonly known as the (First) Germanic
Sound Shift to distinguish it from a similar wholesale remaking of the Germanic
stop system in Old High German, often referred to as the Second or High German
Sound Shift (for which see Chapter 11, § 5).
106 Sound change

Having talked so much about Grimm’s Law, let us see how it operates. Let us
begin with a brief look at the differences between Germanic, represented here by
Gothic and Old English, and the classical Indo-European languages, concentrat-
ing on the initial consonants; see example (1). (In some cases, the initial conso-
nant is preceded by a prefixed element. Such elements are put in parentheses.)

(1) Classical Indo-European languages Germanic languages

Greek Latin Sanskrit Gothic Old English
a. patḗr pater pitā́ fadar fæder ‘father’
treîs trēs trayas þreis þrī ‘three’
(he-)katón centum [k-] śatám hund hund ‘hundred’
b. déka decem dáśa taihun tēon ‘ten’
geúomai gustus ǰōṣ- kiusan cēosan ‘taste, test,
choose’
c. phérō ferō bharāmi baira beoru ‘I carry’
(é-)thē-ka fē-c-ī (a-)dhā-m (ga-)dē-þ-s dǣd ‘put/do;
deed’
kheúō fu-n-d-ō ho-tar- giutan gēotan ‘pour’

As these examples show, change is not limited to Germanic. Especially in the last
three items (set (c)) we notice some major differences between the initial conso-
nants of Greek, Latin, and Sanskrit. Still, the greatest differences separate Ger-
manic from the rest of Indo-European.
Starting with a reconstruction of Proto-Indo-European (PIE) that postulated
voiceless stops for set (a), voiced ones for set (b), and voiced aspirated ones for
set (c), Grimm accounted for the different look of Germanic by postulating three
sweeping and highly systematic sound changes, affecting whole classes of sounds
at the same time:

(2) Grimm’s Law:

i. PIE voiceless stops become voiceless fricatives;
ii. PIE voiced stops become voiceless stops;
iii PIE voiced aspirates become voiced stops or fricatives (depending on
the context in which they occur).

Change (i) accounts for the differences in set (a) of (1) above, e. g. Gk treîs, Lat.
trēs, Skt. trayas corresponding to Goth. þreis, OE þrī ‘three’. Change (ii) explains
correspondences like Gk. déka, Lat. decem, Skt. dáśa : Goth taihun, OE tēon ‘ten’.
And change (iii) derives Goth. (ga-)dē-þ-s, OE dǣd ‘deed’ from the PIE root *dhē
‘put, make’ underlying Gk. (é-)thē-ka, Lat. fē-c-ī, Skt. (a-)dhā-m.
 Grimm’s Law 107

What is especially remarkable is that these changes apply not just to a few
words. Their effects recur in hundreds of other words. Grimm’s Law, thus, is not
only phonetically highly systematic, by affecting all classes of stop consonants,
but it also is lexically systematic, by applying to so many words.
This dual systematicity greatly impressed other Indo-Europeanists and
inspired a massive outburst of research on sound change, compensating for its
neglect in earlier Indo-Europeanist studies.
Since Rask’s and Grimm’s times, many similar systematic sound changes
have been found in many other areas of the world. For instance, among the early
Indo-European languages, Armenian had a similar sweeping sound shift; see the
initial consonants in the examples in (3) below. (Some of the Armenian conso-
nants underwent further changes, such as original *p > h.)

(3) Armenian sound shift

Greek Latin Sanskrit Armenian
a. patḗr pater pitā́ hayr ‘father’
tó (is-)tud tad thē ‘that’
téttares quattuor čatvāri čhorkh ‘four’
b. déka decem dáśa tasn ‘ten’
gunḗ ǰani/gnā- kin ‘woman’
c. phérō ferō bharāmi berem ‘I carry’
(é-)thē-ka fē-c-ī (a-)dhā-m dir ‘put, place; make, do’
thermós fornax gharma- ǰerm ‘hot, heat; oven’

A similar change, but restricted to the voiceless stops, differentiated Southern

Bantu languages from the other Bantu languages; see example (4). In this case,
different languages exhibit different outcomes, but they share the fact that they
systematically change all voiceless stops of the ancestral Proto-Bantu language.
Moreover, the fricative outcomes of *p and *k in Sotho can easily be related to
the aspirates of Xhosa; see § 5.4 below. (The development of *t to Sotho r is more
difficult to explain and therefore is not discussed here.)

(4) Southern Bantu sound shift

Proto-Bantu Xhosa Sotho
kama khama xama ‘wring’
tatu thathu raru ‘three’
pa pha Φa ‘give’

Another parallel to Grimm’s Law, affecting voiceless stops, has been observed
in the “Chipewyan consonant shift” of Athapaskan. That such changes need not
108 Sound change

be limited to ancient and/or “exotic” languages is shown by the British English

dialect of Liverpool, in which voiceless stops are shifting toward fricatives, as in
lock which, via [lɔkh] with heavily aspirated velar stop, is changing to [lɔx] with
velar fricative.
Though clearly interesting and impressive, changes such as those in (3) and
(4) were discovered much later, too late to create the same excitement as Grimm’s
Law. But the very fact that they were discovered is ultimately the result of the
interest in sound change generated by Rask’s and Grimm’s discoveries.
In recent times, doubts have been voiced about both Grimm’s Law and the
Armenian consonant shift. Some scholars have claimed that the Armenian and
Proto-Germanic consonant systems are archaic, close to the one of Proto-In-
do-European, and that the systems of the other Indo-European languages are the
result of innovating changes. This is the so-called “glottalic” view of Proto-In-
do-European mentioned in the discussion of Armenian and Iranian (Ossetic) in
Chapter 2. As noted in that discussion, the glottalic view is highly controversial.
Many linguists are not convinced by the cogency of the arguments and evidence
marshalled in its support. Even if the theory should eventually prevail, this does
not diminish the importance of the work of Rask, Grimm, and their successors and
its significance for the development of historical linguistic thinking and meth-
odology. For ease of exposition, the subsequent discussion ignores the glottalic
theory and proceeds under the assumption that the traditional, Grimm’s Law,
view is correct. (The issue is taken up again in Chapter 16.)

3 From Grimm’s Law to Verner’s Law

While Rask and Grimm, as well as their contemporaries, were highly impressed by
the systematicity of the sound changes they had discovered, they did not expect the
changes to be absolutely regular. They were too much influenced by the idealism
of the Romantic movement to believe that human beings were capable of behaving
with complete, exceptionless regularity, as if they were automata or machines.
In fact, correspondences such as Lat. dies ‘day’ : Engl. day suggested that
Grimm’s Law did not apply in all words. Even more examples could be found
in which Grimm’s Law and other changes separating Germanic from the rest
of Indo-European applied inconsistently, affecting most sounds, but leaving
one or two sounds unaffected. Some examples of such apparent exceptions to
Grimm’s Law are given in (5), where Latin represents the classical Indo-European
languages, and Old English, Germanic. (Note that Lat. c designates the sound
[k].) The first two examples are especially interesting, since here [k] and [p] have
 From Grimm’s Law to Verner’s Law 109

undergone Grimm’s Law in the early part of the words, changing to the voiceless
fricatives h [x] and [f]; but voiceless stops occurring toward the end of the word,
marked in boldface, do not exhibit the change. One might toy with the idea that,
having applied Grimm’s Law once or twice within a given word, the speakers of
early Germanic got tired and therefore did not change voiceless stops occurring
later in the word. But the other two examples show that even voiceless stops not
preceded by other Indo-European voiceless stops in the same word may fail to
undergo the change. That is, the exceptions seem to be completely random.

(5) Latin Old English

captus hæft ‘captured, prisoner’
piscis fisc ‘fish’
spuō spīwan ‘spew, spit’
stō standan ‘stand’

In addition to such words in which Grimm’s Law failed to apply (or applied only
partially), there were a number of other words in which there was a change, but
the outcome of the change was different from the one predicted by Grimm’s Law.
Instead of being reflected by the expected voiceless fricatives, Indo-European
voiceless stops came out as voiced. Compare the examples in (6). Here again, it
seemed impossible to come up with any generalization about the words in which
such exceptional outcomes are found. True, the examples in (6a) all refer to close
family relatives; but so does (6b). More than that, within one and the same para-
digm (= the set of inflected forms of a given word) we find some forms exhibiting
outcomes conforming to Grimm’s Law, whereas others have exceptional voiced
outcomes. Compare examples (7a) vs. (7b), where the classical Indo-European
languages and Germanic are respectively represented by Sanskrit and English.
Such alternations within the same paradigm are now commonly called paradig-
matic alternations.

(6) Latin Sanskrit Gothic Old English

a. pater pitā́ fadar fæder ‘father’
māter mātā́ mōdor ‘mother’
b. frāter bhrātā́ brōþar brōþor ‘brother’

(7) Sanskrit OEngl.

a. vártate weorþan ‘turn, become’ (pres.)
vavárta wearþ (past sing.)
b. vavr̥ túr wurdon (past plur.)
vr̥ taná- (ge)worden (past participle)
110 Sound change

As time progressed, scholars discovered that some forms originally considered

exceptional were simply irrelevant. For instance, closer examination of the evi-
dence and of the regular sound correspondences between the Indo-European
languages showed that the similarities between Engl. day and Lat. dies are acci-
dental. The two forms are not really cognate. A genuine English cognate of the
root in Lat. dies is, believe it or not, the first element in the word Tuesday. The root
underlying both of these is PIE *dy(e)u-, originally meaning ‘sky, light, day’. The
Latin word represents a fairly straightforward outcome. The Germanic word pre-
sents no difficulties from the perspective of sound change either, since its initial
t is precisely what we would expect by Grimm’s Law. The meaning of the word
represents a more special development, but that development has parallels else-
where in Indo-European. In the meaning ‘sky’, *dy(e)u- became the name of the
God of the sky, often with the addition of the word ‘father’; compare Skt. dyaúṣ
pitā́, Gk. Zeús Patḗr, Lat. Juppiter. It is this name of the Sky God that underlies the
Tues of Engl. Tuesday. (See also Chapter 18, § 4.3.)
In addition, scholars realized that some apparent exceptions to Grimm’s Law
are in reality borrowings. Thus, Latin has a word pondus ‘weight’ which in pro-
nunciation and meaning is close to Engl. pound, OE pund. If the two words were
inherited from Proto-Indo-European, the English word would be a double excep-
tion to Grimm’s Law, since neither the voiceless stop nor the voiced one would have
changed. In fact, however, OE pund is an early borrowing from Latin, made at a time
when Grimm’s Law had run its course and therefore did not affect any new words.
Other borrowings from that early period include street, OE strǣt, from Lat. (via)
strāta ‘a prepared road’, and cheap, OE cēap ‘bargain, purchase’, from Latin caupō
‘merchant’. These are part of a large amount of vocabulary exchanged between
Latin and Germanic during the time of the Roman expansion into Germanic terri-
tory and the slightly later “return visits” of the Germanic tribes. Among the words
going from Germanic to Latin and its descendants is the word for ‘soap’: PGmc.
*saipōn-, hence Engl. soap, Germ. Seife : Fr. savon, Span. jabon, etc. (See Chapter 2.)
A better understanding of etymological relationships and of borrowings thus
managed to clear away some of the apparent exceptions to Grimm’s Law. But by
its very nature, this approach produced only piecemeal solutions. A much more
sweeping solution, which has not received the recognition it deserves, is C. Lott­
ner’s discovery in 1862 that exceptions of the type (5) above, far from being totally
random, exhibit an amazing degree of regularity: The unshifted voiceless stops all
occur after a voiceless fricative, whether that fricative is the original PIE sibilant s
or the result of Grimm’s Law (such as the f of OE hæft).
Lottner’s insight suggests that part one of Grimm’s Law as originally stated
needs to be revised, so as to systematically exempt voiceless stops after Germanic
voiceless fricatives (or after their equally voiceless PIE sources).
 From Grimm’s Law to Verner’s Law 111

This still left the exceptions in (6) and (7), and these were much more diffi-
cult to explain. It was only in 1877 that the Danish linguist Karl Verner found a
solution which showed that these, too, were not really irregular but exhibited a
regularity of their own. The reason for the long wait was that the regularity of
these forms could not be accounted for by modifying Grimm’s Law; it required
a law of its own. Moreover, the conditions under which the law applied were far
from obvious if one restricted one’s horizon to Germanic. Rather, it was neces-
sary to look to other languages, mainly Greek and Sanskrit, for an explanation.
And if that were not enough, one had to attribute the change at least in part to a
conditioning factor considered quite unlikely to bring about voicing, namely the
location of the Indo-European stress or accent. Once all of these elements were
brought together, however, the solution was so clear, so obvious, and so “neat”
that no doubt many scholars asked themselves, “Why couldn’t I have thought of
that?” But they didn’t, and the change responsible for the voiced outcomes came
to be called Verner’s Law.
To see how Verner’s Law works, consider again the forms in (6) and (7) and
note that the voiced outcomes are found only in those forms in which the PIE
voiceless stops occur between vowels or between r and vowel, and where the
syllable preceding the stop is not accented in Sanskrit (which preserves the PIE
accent placement). Elsewhere, the voiceless stop occurs.
Now, as example (8) shows, this distinction between voiced and voiceless
outcomes is not restricted to PIE voiceless stops; it is also found in the reflexes of
PIE *s. (The r found in Old English goes back to an earlier *z.) Verner’s Law, thus,
can be said to affect all Germanic fricatives, whether they reflect original *s or
result from PIE voiceless stops by Grimm’s Law.

(8) Sanskrit Old English

ǰṓṣate cēosan ‘taste, choose’ (pres.)
ǰuǰṓṣa cēas (past sg.)
ǰuǰuṣúr curon (past pl.)
ǰuṣaná- (ge)coren (past participle)

Keeping in mind these various factors, as well as some others which it would take
too long to exemplify, Verner’s Law can be formulated as follows:
112 Sound change

(9) Verner’s Law:

Proto-Germanic fricatives (including sibilant s) become voiced if the follow-
ing three conditions are met:
i. They are not initial
ii. What precedes and follows them is voiced, and
iii. The PIE accent is not on the immediately preceding syllable.

Before we can proceed to show how Verner’s Law operated in relation to Grimm’s
Law, we need to mention one other change. After Verner’s Law ceased to operate,
the accent shifted to the root syllable of the word which, in most cases, coincides
with the initial syllable. It was this change that obscured the accentual condition
of Verner’s Law and, consequently, made it so difficult to recognize.
If we let GL stand for Grimm’s Law, VL for Verner’s Law, and AS for the early
accent shift to the initial or root syllable, we can illustrate the way these three
processes interacted. As example (10) shows, only the order GL before VL before
AS will yield the right results. Other sequences fail to do so. See the unsuccessful
derivations in (10’) and (10’’), where the incorrect forms are marked by a following
asterisk.

(10) PIE pǝtḗr bhrā́tēr

GL faþér bróþer
VL faðér inapplicable
AS fáðer bróþer
Old English outcomes fæder brōþor
(10’) PIE pǝtḗr bhrā́tēr
VL inapplicable inapplicable
GL faþér bróþer
AS fáþer bróþer
Old English outcomes fæþer* brōþor
(10’’) PIE pǝtḗr bhrā́tēr
AS pǝ́ tēr (bhrā́tēr)
GL fáþer bróþer
VL inapplicable inapplicable
Old English outcomes fæþer* brōþor

Situations like these, where only one sequence of changes will yield the correct
results establish what linguists call a relative chronology: Even when we
cannot be sure about the “absolute” chronology (i. e. when the changes took place
in historical time), we are at least able to demonstrate their relative ordering.
When looking at demonstrations of the type (10)–(10’’), non-linguists often get
 The regularity hypothesis and the neogrammarians 113

the feeling that linguists are just playing a shell game, imposing their own view on
history. In fact, however, it is the history of the language that imposes the solution
on linguists. If history had been different, the outcomes would be different, and a
different relative chronology would suggest itself.
The influence of Verner’s Law on historical linguistics was profound. The fact
that the law was conditioned by phonetic factors previously not considered even
remotely relevant stimulated the linguistic community to pay much greater atten-
tion to fine phonetic details that had not been examined in earlier studies. And
this closer look at the factors that condition sound change has greatly enriched
our understanding of language history. This is not to say that all the after-effects
of Verner’s Law were beneficial. There was, as in many other cases, a certain band-
wagon effect that resulted in a large variety of attempts at explaining historical
developments in terms of accentual differences – even in cases where there simply
was no evidence for such differences. But these misuses of accentual explanations
do not diminish the significance – and correctness – of Verner’s Law.

4 The regularity hypothesis and the

neogrammarians
In addition to stimulating greater attention to the phonetic conditions of sound
change, Verner’s Law provided the final stepping stone toward solving the appar-
ent exceptions to Grimm’s Law. If allowances were made for the operation of
certain additional changes, such as analogy (see Chapter 5), it appeared that the
nearly sixty years of scholarly endeavor since Rask and Grimm published their
findings had succeeded in establishing that Grimm’s Law was not just amazingly
sweeping and systematic. It seemed to have, in fact, operated with complete reg-
ularity, affecting every single word it could affect at the time that it took place.
Developments like these suggested a very different meaning for the slogan
“Sound change is regular”, which had been voiced by a number of earlier schol-
ars. Now a group of young linguists working at the University of Leipzig (Germany)
felt justified in raising this slogan to an axiom of historical linguistics, with the
assumption that “regular” meant not just “overwhelmingly regular”, but abso-
lutely regular. (See the second motto at the beginning of this chapter.) Another
way of putting it was to say that sound change “operates blindly”. Because this
new view on the nature of sound change had been proposed by a group of young
linguists at Leipzig, the group and its later followers came to be known as the
neogrammarians (“Junggrammatiker” in German). Their claim, in turn, has
been labeled the regularity hypothesis.
114 Sound change

To understand what is meant by the claim that sound change is absolutely

regular, it is necessary to understand what is meant by the term sound change.
Several other factors also need to be taken into consideration in order to interpret
the claim. Without these modifications, the claim is meaningless.
Sound change, in the sense that the term is used here, refers to change in
sounds conditioned only by phonetic factors or, more accurately, but less ele-
gantly, change in sounds not conditioned by non-phonetic factors. Clearly there
are all kinds of changes that could affect the pronunciation of words, but were
not conditioned just by phonetic facts. For instance, as mentioned in Chapter 1,
the early Modern English plural of cow, kine, changed to cows on the analogy of
other, more regular plurals, such as pig : pig-s, horse : horse-s. Changes of this type
were known to be quite irregular. For instance, swine did not change to swows*,
or wine to wows*.
Similarly, we find that tabooed words often are deliberately distorted in polite
company, making it possible to use the tabooed word without “actually saying it”.
Examples in English are expressions like gosh, darn, doggone …; blasted, blessed,
bleeding …; shoot; friggin’ instead of well-known taboo words which are better
left unprinted. Here again, it is quite clear that the changes in question affect the
pronunciation of the words, but that they are by no means regular. Thus, rod is
not replaced by rosh, ham by harn, muddy by masted, hit by hoot, or tucking by
triggin’. (On taboo distortion see also Chapter 7.)
It was believed that processes like analogy and taboo distortion do not take
place with the regularity of, say, Grimm’s Law because they heavily depend on
the mental association of forms with each other or on meaning, whereas sound
change, properly understood, operates without any regard to such mental asso-
ciations or meanings – or even the effect that it might have on our ability to com-
municate such mental associations or meanings.
There is a lot of empirical evidence that sound change does, in fact, take
place “blindly”, without consideration of the havoc it may create on our ability
to communicate. Thus, in Southern U.S. English, pen and pin have come to be
pronounced identically, by a process that regularly changed [ɛ] to [i] before nasal.
As a consequence it is now necessary in these dialects to resort to circumlocutions
like writing [pin] vs. needle or stick [pin] in order to properly distinguish the two
words. Similarly, in many varieties of American English, can and can’t are not dis-
tinguished, except in very slow, deliberate speech, necessitating speakers to ask
things like Do you mean you are able to or not? In New York English, the vowel of
words like can and man has changed to such an extent that the common woman’s
name Ann sounds like the relatively rare man’s name Ian. It is reported that chil-
dren, being more familiar with the name Ann, asked their parents why they gave
their baby brother, Ian, a girl’s name. Similarly, in some varieties of the English
 The regularity hypothesis and the neogrammarians 115

in and around Chicago, John has acquired a pronunciation that outsiders hear as
Jan, again resulting in all kinds of confusion. (More on the New York and Chicago
changes in § 5.4 below.)
Changes of this sort are not restricted to modern English; they have taken
place at all stages of the language. Compare for instance cleave ‘stick to’ and cleave
‘chop, split’. The second of these two words goes back to OE cleofan, is related to
regional Germ. klieben ‘chop, split’, and derives from a PIE root *glewbh-, while
the first reflects OE cleofian, is related to Germ. kleben ‘stick’, and goes back to
PIE *gleybh-. Since ‘stick to’ and ‘chop, split’ convey meanings that are just about
diametrically opposed, the use of the two words must have led to a lot of confu-
sion. In modern English, this confusion is to a large extent resolved by avoiding
the use of cleave in the meaning ‘stick to’. But this change took place only after
sound change made it impossible to distinguish the two words. As in all the other
examples above, there is no evidence that speakers tried to block the changes in
mid-stream, in order to avoid possible confusion.
In addition to understanding properly what is meant by the term sound
change, it is further necessary to be aware of a lot of “fine print”. For instance,
in the natural sciences the expression “absolutely regular” would mean that a
particular change takes place under the same conditions, anywhere, and at any
time that it has a chance to do so. In the regularity hypothesis, this can hardly be
the intended meaning. For even a moment’s reflection will tell us that Grimm’s
Law took place at some point between Proto-Indo-European and Germanic, and
that it took place only at that point, and only in Germanic (although some other
languages, such as Armenian, may have had similar changes). If Grimm’s Law
were not restricted this way, we should expect all the other Indo-European lan-
guages – in fact, all the languages of the world – to have had the same change.
The change also should have applied again and again, so that a d going back to
earlier *dh by part (iii) of Grimm’s Law, would next undergo part (ii) of the same
law and become t, only to undergo part (i) and turn into þ. As a consequence, PIE
*dhē- should not have stopped at the stage represented by Mod. Engl. deed, but
should have changed further to teet*, and then to theeth*. The regularity hypoth-
esis, therefore, is a statement about particular sound changes as historical events,
limited by place, time, as well as language (or even dialect).
One final restriction on the regularity hypothesis must be mentioned: The
neogrammarians were keenly aware that certain types of change which do not
easily qualify as analogy or the like, nevertheless are notoriously irregular. These
prominently include the following two processes: (i) “metathesis”, the transpo-
sition of sounds, as in OE þrit(t)ig > Mod. Engl. thirty; and (ii) “dissimilation”, as
in Engl. col(o)nel > [kǝrnǝl], where the first of two [l] sounds has changed to [r]
so as to become “dissimilar” to the second. The neogrammarians made several
116 Sound change

attempts to account for the irregularity of these changes. Perhaps the best among
these is the claim that dissimilation and metathesis are similar to speech errors, a
lapse in some special control faculty, perhaps the same faculty that we put to the
test in tongue twisters. (See also § 5.5 below.)
From the time it was formulated, the neogrammarian regularity hypothesis
ran into strong opposition. Even so, the hypothesis was widely accepted by most
historical linguists. Recent research has raised questions about many of the neo-
grammarians’ assumptions and has suggested that sound change is not always
regular. But even this research confirms that much of sound change is so close
to regular that the neogrammarian hypothesis can still be accepted as a general
guideline.
Even if we may have to give up the notion that sound change is absolutely
regular, in favor of the more modest proposition that it is overwhelmingly regular,
the regularity hypothesis has proved enormously fruitful in historical linguis-
tics. It challenges linguists to look more carefully at linguistic change in order to
explain apparent irregularities. And any closer investigation is bound to yield new
and interesting results – in any field of inquiry. In the field of historical linguistics,
the regularity hypothesis certainly has done just that. (For further discussion,
including the importance of the regularity hypothesis for comparative-historical
linguistics, see § 7 at the end of this chapter.)

5 Some types of sound change

Except for the restriction discovered by Lottner, Grimm’s Law is a set of changes
that take place across the board, and no special phonetic conditions are discern-
ible that might motivate them. Verner’s Law, by contrast, depends on a lot of
conditions. Why the most celebrated of these, the accentual one, should have
induced voicing of original voiceless fricatives is still open to question. However,
the fact that the voicing takes place in a medial voiced environment seems to
make intuitive sense. The fricative “assimilates” to the voicing of its surroundings.
Here, then, we have a perfect case of the context in which the change takes place
in some way triggering the change. Processes of this sort can be called condi-
tioned changes.

5.1. Assimilation, weakening, loss. Over the 200-odd years that modern histori-
cal linguistics has been practiced, a large number of conditioned types of changes
have been observed. By far the most common of these are changes which in some
ways ease the process of pronunciation. This, however, should not be taken to
 Some types of sound change 117

suggest that all sound change leads to phonetic simplification. Some changes
consist of the addition of new sounds, a phenomenon that could hardly be con-
sidered simplification; see § 5.2 below. Others appear to be neutral as regards
simplicity; see § 5.3. Moreover, there clearly must be limits on the extent to which
simplification can progress. If phonetic simplicity were permitted to run its full
course, it would change all words to something like [ǝ], a simple central vowel
without any complex distinctions of vowel position (high, mid, low; front, central,
back; etc.), to say nothing of the effort of producing a large variety of different
consonants. But how would we convey with this one, maximally simple utterance
the plethora of different meanings that we are able to express through our more
“complicated” words? Human language requires a certain degree of complexity
to successfully communicate meaning, variation, and creativeness. (See also § 6
below.)
Nevertheless, it is true that changes which seem to ease pronunciation make
up the bulk of regular sound change. That these changes have not, over the long
history of human language, led to the ultimate stage of simplification, [ǝ], sug-
gests that language has enough resilience, as it were, to counteract the ravages
of simplificatory change and to keep reintroducing enough “complications”,
whether by sound change or other changes, to retain its functionality.

5.1.1. Assimilation. One type of simplificatory change, briefly mentioned in the

preceding section, is assimilation. As the name suggests, the process consists of
making the pronunciation of a given sound more similar to that of another sound,
thus simplifying the articulatory gestures required to pronounce the word. The
sound triggering the change may precede or follow, but usually it is an immediate
neighbor of the sound that is changing.
The word assimilation itself provides a perfect illustration of the process
it designates. It is derived from the Latin elements ad- ‘to’ and similis ‘similar’,
which combined to form the basis of Lat. assimilare ‘to make similar to’, in which
the d of ad so successfully assimilated to the following s that it acquired the same
articulation. A less radical example of assimilation is found in the formation of
English plurals. Although we write the plural ending as -s, no matter where it
occurs, in pronunciation it assimilates to preceding voiced sounds by becoming
voiced; compare bed-s [bedz] vs. bets [bets].
Several subtypes of assimilation occur frequently enough, with interesting
results, to have been given special names. One of these is umlaut, a process
common in early Germanic; see example (11). Umlaut typically involves the assim-
ilation of a root vowel to a vowel occurring in a suffix; see Stage II in example (11).
Moreover, typically some of the suffix vowels triggering the change are subse-
quently lost, so that the conditions for the vowel change are no longer fully recov-
118 Sound change

erable; see Stage III. The example in (11) further illustrates a common outcome
of umlaut. If the entire suffix is lost, the vowel change produced by umlaut may
take over the function of the original suffix, in this case, the function of indicat-
ing plurality. Many of the “irregular” plurals of Modern English owe their origin
to umlaut; compare foot : feet, tooth : teeth, mouse : mice, louse : lice, man : men,
woman : women.

(11) Stage I Stage II Stage III

(= pre-Old Engl. I) (= pre-Old Engl. II) (= Old English)
kū-z kū-z cū [kū] ‘cow’
kū-iz kǖ-iz cȳ [kǖ] ‘cows’

A process in some ways similar to umlaut is palatalization. Like umlaut, the

change typically proceeds in three stages: A stage without assimilation is followed
by a stage in which a class of consonants, mainly the dentals or velars, assimi-
lates to following front vowels producing a palatalized consonant (indicated by a
superscript y), and a final stage at which some of the triggers for the change are
lost. At that point, the palatalized consonant tends to become a palatal which,
in turn, has a strong tendency to change into a sibilant or fricative. Compare the
example in (12). Traces of the palatalization in (12) are found in many English
borrowings from French. And because French palatalization was a conditioned
process, which took place only before front vowel, many of the words borrowed
from French alternate between the original velar and the sibilant reflex of pala-
talization, depending on the vowel that originally followed. Compare alternations
of the type electric [-k] : electricity [-s-]. As noted in § 3 above, such systematic
changes in the pronunciation of words, depending on the context in which they
occur, are commonly referred to as paradigmatic alternations.

(12) Stage I Stage II Stage III Further changes

(= Latin) (= Proto- (= Proto-
Romance I) Romance II)
faciat [fakiat] fakyya fakya > fačǝ > Mod. Fr. fasse [fas(ǝ)]
‘would make’

Many words borrowed with s resulting from French palatalization underwent

another process of English palatalization, if the s was followed by a suffix-initial
front vowel, as in electrician. (The vowel triggering this second round of palatali-
zation has since ceased to be pronounced.)
Before turning our attention to other simplificatory processes, let us briefly
mention a type of change which often is considered assimilatory, namely final
 Some types of sound change 119

devoicing. As the name suggests, this process involves the devoicing of final con-
sonants. The starting point for the change seems to lie in utterance-final position
where even languages like standard English, not otherwise known to have final
devoicing, exhibit a slight degree of devoicing. In many other languages, such as
German and Russian, the change goes farther and leads to a complete “merger”
of voiced stops and fricatives with their voiceless counterparts. Moreover, the
change is not confined to utterance-final position but applies word-finally, as
well. Compare the example in (13).

(13) OHG Mod. Germ (NHG)

tag Tag [tāk] ‘day’
vs. taga Tage [tāgǝ] ‘days’
stab Stab [štāp] ‘staff, stick’
vs. staba Stäbe [štɛ̄ bǝ] ‘staffs, sticks’
etc.

Assimilatory changes can also be observed in sign languages. A type of assimi-

lation that is very common in American Sign Language has come to be known as
symmetry. For instance, the sign for ‘angry’ once consisted of one hand moving
away from the waist; now it is articulated instead with both hands moving in sym-
metry. A slightly different example of the change is found in the development of
the sign for ‘final’. This sign used to involve the index finger of one hand striking
the little finger of the other; now, however, little finger strikes against little finger.

5.1.2. Weakening. A process that even more clearly simplifies pronunciation is

weakening (also referred to as lenition), a relaxation, reduction, or even total
omission of the articulatory gestures required to make particular speech sounds.
Weakening occurs most commonly in a medial voiced environment (just like
Verner’s Law), but may be found in other contexts as well. In modern English it is
especially the voiceless dental stop [t] that is liable to be weakened. In American
English, for instance, we find better pronounced as [beṙǝr] (with an r-like flap),
[bedǝr] (with voicing), or in some areas apparently even [beØǝr] (with loss of the t
indicated by the symbol Ø). Many varieties of British English have the pronuncia-
tion [beʔǝ(r)] (with glottal stop). What is common to all of these pronunciations is
a relaxation in the gestures required to make a voiceless [t] in intervocalic voiced
environment. In the case of [ṙ], contact is made less firmly, resulting in a flap; in
[d], the switch from voicing to voicelessness and back to voicing has been elimi-
nated; [ʔ] omits the oral closure required to make a [t], but leaves something like an
“echo” of that closure in the shape of the glottal stop; and the most radical solution
is represented by Ø – a complete omission of all gestures required to articulate [t].
120 Sound change

Alert readers may have noticed that the voicing in [bedǝr] could also be inter-
preted as a simple case of assimilation of voiceless [t] to its voiced surroundings.
Intervocalic voicing is an area in which the two processes, assimilation and weak-
ening, overlap. But perhaps there is more to it. One could argue that assimilation
in general is simply a special case of weakening, in that the articulatory gestures
required to pronounce sounds differently are relaxed, leading to more similar pro-
nunciations.
In some languages, weakening can be quite sweeping, affecting all intervo-
calic stops. This is the case in the western Romance languages. See for instance
the Spanish examples in (14), where intervocalic Latin [p, t, k] become voiced
fricatives and where [d, g] are lost altogether.

(14) Latin Spanish

lupus lobo [β] ‘wolf’
status estado [ð] ‘state’
locus luego [γ] ‘place’
habere haber [β] ‘have’
videre ve(e)r [Ø] ‘see’
legere leer [Ø] ‘read’

5.1.3. Loss. The loss of speech sounds is not limited to the contexts that typically
exhibit weakening, but occurs frequently in other environments as well. As we
already have seen in Chapter 1, English lost initial [k] before nasal, as in OE cnyht
> Mod. Engl. knight [Øn-]. A context especially liable to undergo loss is the end
of words; compare (15). The reason for this presumably is the fact that our voice
often “trails off” at the end of utterances, both in intonation (which goes down
to a fairly low pitch) and in the precise articulation of speech sounds. Like final
devoicing, the results may subsequently be generalized to all word-final positions.

(15) Old English Mod. English

singan singØØ
stānas stones [-nØz]

A repeated process of loss in final syllables is responsible for the fact that English
has lost most of the inflectional endings of Old English. Old English had endings
to differentiate four different noun cases (nominative, genitive, dative, and accu-
sative) and to distinguish these cases in two different numbers (e. g. dative sg.
stān-e ‘to the stone’ : dative pl. stān-um ‘to the stones’). Of these different endings,
only two have remained in Modern English, both sounding identical: the plural
marker -s and the genitive marker -s. Here, then, loss may be said to have simpli-
 Some types of sound change 121

fied not only pronunciation but the whole inflectional system of English. (Note,
however, that analogy played a role, too, in this development. See Chapter 5, § 4,
as well as the brief discussion in Chapter 1.)
Sometimes, loss of a sound is compensated for by lengthening of the preced-
ing vowel, where lengthening maintains the timing of the structure from which
the sound is lost. For example, Engl. tooth derives from PIE *dont- (as in Gk.
o-dónt-), via PGmc. *tanþ- which changed into OE tōþ with loss of the nasal n and
with compensatory lengthening of the preceding vowel, hence OE ō. There is
an interesting parallel in sign languages. When an original compound symbol of
American Sign Language is reduced through loss of one of the component signs,
the remaining sign is lengthened through repetition. For instance, ‘orange’, orig-
inally a compound of ‘slice’ and ‘yellow’, now is formed without the element
‘yellow’ and with repetition of the sign for ‘slice’.

5.2. Epenthesis, the gain or insertion of speech sounds. Although loss is a very
widespread phenomenon and, as we have just seen, can have far-reaching effects
on the structure of languages, some sound changes have the opposite phonetic
effect – they introduce speech sounds. This type of change is generally referred
to as epenthesis.
A common subtype of epenthesis consists of the insertion of vowels before
word-initial consonant groups or into such groups elsewhere. A well-known
example is the process of prothesis in early Spanish and French, which inserted
an [e] in front of s + stop clusters. Compare Lat. spata ‘sword’ : Span. espada, Fr.
épée. As these two words show, epenthesis in one context of a given word does
not prevent weakening or even loss in others. Note especially the French word in
which the s which had triggered the prothesis of e was lost by a later weakening
change.
Not only vowels may be inserted, but consonants as well. This is an especially
common phenomenon between nasals and following liquids, as in OE þunØrian
‘to thunder’ > þundrian, whence Mod. Engl. thunder. The motivation for this
change seems to be as follows. Nasals are pronounced with the same articula-
tion as voiced stops, except that the passage to the nose is left open, permitting
nasal resonance to be audible. Switching from the nasal to the following non-na-
sal liquid requires a delicate timing in the adjustment of articulatory gestures.
Ideally, the change from stop to liquid should take place at the same time as the
change from nasal to non-nasal. Epenthetic developments as in þunrian > þun-
drian result if the two gestures are not properly timed, i. e., if speakers switch
too early from nasal to non-nasal, producing a stretch of oral stop articulation.
Compare the schematic presentation in Illustration 1. Here a solid horizontal line
indicates the presence of a particular articulation, a broken line, its absence. If in
122 Sound change

the original sequence n + r the stop articulation is held out longer than the nasal
articulation (see the circled part of on the right hand of Illustration 1), the result
is an interval of a non-nasal – i. e., oral – stop d. (The vertical lines in Illustration
1 indicate the boundaries between the sounds.)

Illustration 1: Consonant epenthesis as the result of wrong timing

The very common phenomenon of aspiration, as in the change from Proto-Bantu

tatu ‘three’ to Xhosa thathu (example (4) above), can likewise be considered an
example of wrong timing; see Illustration 2. If the switch from voiceless (stop) to
voiced (vowel) is not timed correctly, an interval of voiceless articulation results,
without the stop to “support” it; and the [h]-like hissing noise of that interval is
what we call aspiration.

Illustration 2: Aspiration as the result of wrong timing

5.3. Acoustically or auditorily conditioned changes. The types of sound

change we have looked at so far are best explained as being triggered by artic-
ulatory factors. However, when we learn our first language, no one tells us how
to articulate. No one says, “Now, in order to make a [t] you press your tongue tip
firmly against the back of the teeth, make sure not to let your glottis vibrate, and
then release the closure.” We have to find out for ourselves how to articulate the
 Some types of sound change 123

sounds we hear. And in the process we may make mistakes. Glaring mistakes are
usually corrected over time. But less obvious deviations may persist. Moreover,
misunderstanding the phonetic output of others is not limited to children. Adults,
too, may mis-hear and consequently mispronounce words they are not familiar
with. Many speakers of American English, for instance, pronounce the abbrevi-
ation etc. as [ekseterǝ], instead of the correct form [etseterǝ]; another common
example is aestetic for aesthetic.
It is therefore not surprising that we can find occasional examples of sound
changes which appear to result from such misunderstandings. An example is the
substitution of uvular [r] for trilled (post-)dental [r]. The substitution has been
reported to be frequent among Spanish children; but in most Spanish dialects,
children are corrected and told to use [r]. In rural dialects of Puerto-Rican Spanish,
[r] has caught on; and the second part of the name Puerto Rico is now pronounced
[Riko]. In French, a similar substitution has become effectively the norm, except
in theatrical stage pronunciation, where [r] is still preferred. Note however that
[r] has a strong tendency to weaken toward a voiced or voiceless velar (or uvular)
fricative or “scrape”. The usual Modern French pronunciation of a word like rouge
has an initial voiced velar fricative [γ]. And the rural pronunciation of Puerto Rico
commonly is [puelto xiko], with voiceless velar fricative.

5.4. Structurally conditioned changes, “chain shifts”. Although the neogram-

marians firmly believed that sound change is conditioned only by purely phonetic
factors, research in the latter part of this century, especially by the French scholar
André Martinet, has shown that changes may to some extent be triggered by more
abstract considerations, such as the structure of phonetic systems. The most
famous examples of this type are represented by so-called chain shifts in vowel
systems. To illustrate the structural motivation of such chain shifts, let us look
at two concrete examples of change in present-day American English. (Similar
changes are found in many other languages and dialects.)
Certain lower middle-class dialects in New York, Chicago, and many other
large urban areas of the northeastern part of the United States, exhibit the diph-
thongization of [æ] to something like [æǝ] under certain conditions (which need
not detain us here). The diphthong tends to be strongly nasalized, and its initial
element, [æ], tends to rise to mid-vowel or even high-vowel position, yielding
something like [eǝ] or [iǝ]. As a consequence, words like Ann tend to be pro-
nounced [æǝn], [eǝn], or even [iǝn] sounding just like Ian (as noted in § 4, this is
another example the “blind” application of sound change).
Now, as long as nothing else happens, the changes in question may not
appear particularly remarkable. But something else did happen, both in New York
and in Chicago, even if that “something else” was different in the two cities. In
124 Sound change

both cities, there was a system-based reaction to the fact that old [æ] was vacating
its position as a low front vowel and thus introducing a certain imbalance in the
vowel system. In Chicago, the old central vowel [a] began to shift to the position
vacated by [æ], thus rebalancing the system. Hence the pronunciation of words
like John [ǰan] becomes sufficiently similar to that of words like Jan [ǰæn] in other
dialects to confuse people not familiar with this dialect. In New York, on the other
hand, the imbalance is redressed by the fact that the vowel [a] begins to follow
the example of old [æ], by diphthongizing and moving up toward the position of
[u], as in the change of coffee [kafi] to [koǝfi] or even [kuǝfi]. Both of these chain
shifts are outlined in Illustration 3, where the arrows marked with the numeral 1
indicate the initial change, the raising of old [æ] toward the position of [i] in [iǝ];
and the arrows marked by 2 represent the follow-up changes, of old [a] toward [æ]
in Chicago and toward the position of [u] in [uǝ] in New York.

Illustration 3: Chicago and New York chain shifts

Chain shifts can lead to major rearrangements of phonetic systems. For instance,
the change in New York, if carried to its logical conclusion, would eliminate the
low vowels [æ] and [a] from the system. The capacity of chain shifts to bring about
such major rearrangements has led scholars to suspect that similar sweeping
rearrangements of phonetic systems in earlier or even prehistoric times, such as
Grimm’s Law, may likewise have resulted from chain shifts, even if the details of
these shifts may escape us. For Grimm’s Law, for instance, it is possible to cook up
three or four different scenarios, all of them chain shifts. (Some of them may be
more likely than others, but which of them actually took place remains anybody’s
guess.)
The traditional interpretation, going back in spirit to the time of Grimm and
Rask, assumes that the voiceless stops changed first by becoming aspirates (see
the Xhosa example in (4) above and also Illustration 4 below). Under this view,
the aspirates further changed into fricatives by the following steps. In aspirates
with turbulent aspiration, the [h]-like hissing noise of aspiration may assimilate
to the position of the preceding stop, producing affricates, so that [th] > [ts], [ph]
> [pΦ], etc. Thus in many varieties of modern Indo-Aryan, words like phūl ‘flower’
are pronounced [pΦūl]. Affricates, in turn, may be simplified, losing their stop
 Some types of sound change 125

element. This is found in other varieties of the same Indo-Aryan languages, where
[pΦūl] ‘flower’ is realized as [Φūl] or with further change as [fūl]. At this point,
then, the fricative stage attested in Proto-Germanic (as well as Sotho) has been
reached. Changes of voiceless stops to voiceless fricatives are also observed in
Northern Dravidian languages (in Central India and present-day Pakistan), and
in Hungarian and other members of the Uralic family (in Eastern Europe and
adjacent parts of Asia). The whole series of developments, from (aspirated) voice-
less stop through affricate to fricative, has been observed in a change that is still
unfolding in the British English dialect of Liverpool, with words like lock changing
to [lɔkh] > [lɔkx] > [lɔx]; see § 2 above.
Once this complex set of developments has been set in motion, the position
of plain voiceless stops has been vacated. And just as old [a] started to fill the
position vacated by the diphthongization and raising of old [æ] in Chicago, so – it
is claimed – the voiced stops begin to move into the position of the old voiceless
stops in early Germanic. But this change leaves the position of the old voiced stops
empty, and so the old aspirates move to fill that position. Compare Illustration 4,
which ignores the further development of the voiceless aspirates ph, th, kh toward
f, þ, x. This kind of shift, where sounds are “dragged” into a vacated position is
commonly referred to as a drag chain.

Illustration 4: Grimm’s Law as a chain shift

One of the other proposed chain-shift explanations is very similar, except that
it reverses the order of events: The voiced aspirates are considered to shift first,
toward the position of the voiced stops. To avoid merging with the voiced aspi-
rates, the voiced stops move toward the voiceless stops. And these change their
articulation to voiceless aspirate to escape merger with the old voiced stops. For
obvious reasons, this type of shift is called a push chain.
In the absence of relevant historical evidence, these scenarios must remain
speculative, and a choice between them is not possible on purely empirical
grounds. At the same time, some kind of chain-shift no doubt is responsible for
Grimm’s Law. It is hardly conceivable that some speakers of Proto-Indo-European
woke up one fine morning to discover that their entire stop system had mysteri-
126 Sound change

ously changed over night, making their speech radically different from that of
their fellow Indo-Europeans and branding them as Germanic “oddballs”.
As something like a postscript to this section, it might be mentioned that the
Great English Vowel Shift, too, resulted from some kind of chain shift. This
change radically transformed the English vowel system and is largely responsible
for the multiple phonetic values attached to English vowel letters. As a conse-
quence i can denote both [i] and [ay], depending on whether it originally desig-
nated a short or long vowel, and the vowel letters a, e, and i are pronounced [ey],
[ī], and [ay], in contrast to most other European languages which have [a], [e], and
[i] (long or short, depending on the language). Examples are given in Illustration 5.

Old Engl. Mid. Engl. Mod. Engl. Old Engl. Mid. Engl. Mod.Engl.
bītan bīten bite [ay] vs. biten biten bitten [i]
hūs hūs house [aw] sungen sungen sung [ǝ]
hē hē he [ī] better better better [e]
dōm dōm doom [ū] etc.
dǣd dǣd deed [ī]
stān stɔ̄ n stone [ow]
nama nāme name [ey]

Illustration 5: The Great English Vowel Shift

As in the case of Grimm’s Law, opinions differ as to how the change unfolded. The
most widely accepted hypothesis assumes a drag chain, with the high long vowels
ī and ū changing first, becoming diphthongs – most likely [ǝy] and [ǝw] respec-
tively. The high-vowel positions vacated in this way then were filled by the long
mid vowels, whose emptied positions, in turn, attracted the long low vowels, and
so on. Compare the simplified presentation in Illustration 5a, which distinguishes
two phases, one prior to Shakespeare, the second post-Shakespearean and affect-
ing the outputs ē and ǣ of the pre-Shakespearean phase.

Illustration 5a: The Great English Vowel Shift as a chain shift

 Some types of sound change 127

Instead of a drag chain, some linguists postulate a push chain, where the
change was initiated by a general raising of the vowels putting pressure on the
highest vowels. Since these could not be raised any further, they diphthongized
instead.
In this case, empirical evidence makes it possible to decide in favor of the
drag chain. Spelling variation and testimony by contemporary observers show
that only the shifts on the left side of Illustration 5a had been completed by the
time of Shakespeare. Old ǣ and ā, which by now had become ē and ǣ respec-
tively, lagged behind and reached their modern positions only in the post-Shake-
speare period; compare the right side of Illustration 5a. The fact that these two
low vowels lagged behind is precisely what we would expect in a drag chain. If the
shift had been a push chain, one would expect them to have been in the vanguard
of the change.

5.5. Fast, furious, and faulty speech: Typically sporadic changes. While the
types of sound change examined in the preceding sections by and large exhibit
the regularity postulated by the neogrammarians, a few changes are notoriously
irregular or sporadic.
Consider for instance words like Engl. ma’am or bye. The first of these is
patently derived from madam; but just as patently, the change involved is not
a regular change. For instance, we do not say A’am for Adam. Moreover, madam
still coexists with ma’am. Regular sound change supposedly does not leave such
unchanged residue. The expression bye is derivable from good bye, which itself
is derived from God be with ye (with good substituted for God for taboo reasons).
And again, the changes that link God or good be with ye to good bye and bye are
isolated, limited to just this expression.
Irregular shortening developments of this type are rather frequent in forms of
address and formulas of greeting and leave-taking, i. e., in expressions of verbal
politeness. Compare further It. mona (as in Mona Lisa) < Madon(n)a ‘my lady’;
the polite second-person address forms Skt. bhavat < bhagavat ‘(your) lordship’
and Span. usted < vuestra merced ‘your grace’; and the German greeting Mo(ǝ)ŋ <
Morgen < Guten Morgen ‘good morning’. While developments like Morgen < Guten
Morgen may be considered something like ellipsis (see Chapter 5), reductions like
Mo(ǝ)ŋ < Morgen cannot be explained in this manner. Like ma’am they seem to be
clear examples of sporadic sound change, and thus an acute embarrassment to
the regularity hypothesis.
Note however that reduced pronunciations of the type Mo(ǝ)ŋ are not limited
to politeness expressions. They are a common phenomenon in fast or allegro
speech and other forms of less than carefully monitored speech. In fast speech,
German speakers are just as likely to say mo(ǝ)ŋ for the adverb morgen ‘tomorrow’
128 Sound change

as for the expression (Guten) Morgen. In fact, fast speech is notorious for its exten-
sive and pervasive reduction of phonological structure. Even sound sequences
that would not be permissible in careful or lento speech occur quite freely in
fast speech, as in English [ŋaygow], with initial velar nasal, for careful Can I go?
In general, we filter out such highly reduced forms and pretend that only the
lento forms exist. And because we, as speakers, filter out allegro forms, linguis-
tic change generally operates on these, and not on allegro forms. The fact that
politeness expressions are frequent exceptions can be explained as follows. While
society expects us to be polite, we may not necessarily want to lose too much time
over it. Even sticklers for etiquette may find excessively lengthy politeness expres-
sions in bad taste. As a consequence we tend to use the shorter forms furnished
by fast speech (as well as ellipsis).
Similar extensive, even excessive, reductions are commonly found in expres-
sions like you know when we use them – much to the dismay of self-anointed
critics – as speech fillers or in order to reassure ourselves that the addressee is still
listening. Reductions of you know may range from the fairly innocuous [y(ǝ)now]
to things like [nyǝ] or even [yow]. Here it is the relatively subordinate semantic or
communicative value of the expression that is responsible for the phonological
reduction.
One suspects that similar factors are responsible for the very common pho-
nological reduction of clitics. These are a special class of words with the follow-
ing characteristics. They are typically function words and thus, like the type you
know, of reduced communicative significance. Probably as a consequence, they
do not bear an accent of their own. As a result, they differ from “well-behaved”,
“normal” words which do bear accent. Furthermore, unlike normal words they
cannot occur by themselves, and must therefore “lean on” another word, called
the host. (The name clitic is derived from the Greek root kli- ‘to lean on’.)
Elements of this type take something of an intermediate position between full
words and affixes. Examples of English clitics are the ’ve of forms like I’ve, you’ve
and the ’s of forms like John’s got the flu, Mary’s at work. As can be readily seen,
these elements cannot be pronounced by themselves (except by linguists who
have learned to pronounce all kinds of things that ordinary speakers don’t). They
have to lean on a preceding host. In fact, if there is a slight break in the utterance,
separating the host from the element in question, the clitic cannot occur and the
full form must be used instead. Compare unacceptable Mary – ’s at work* with
acceptable Mary – is at work.
What is relevant in the present context is that all of these English clitics have
undergone a large variety of weakenings or reductions. Compare the reduced
forms ’ve and ’s with their corresponding full, non-clitic forms have, and has or is.
Phonological reductions of this type are very common in clitics.
 Some types of sound change 129

If we try to generalize, we may say that the different forms of irregular reduc-
tion and weakening processes we have examined above originate in speech that
is “downgraded”, either because it is less than carefully monitored, or because it
is communicatively of minor importance.
In addition to downgrading, we may also “upgrade” our speech. For instance,
although glottal stops do not occur in the lento speech of most varieties of English,
they are not uncommon in speech that expresses anger or other forms of strong
psychological affect. While in examples like shut [ʔ]up already, such glottal stops
are a rather transitory phenomenon, in some expressions they have become insti-
tutionalized. In the U.S. military, for instance, the command attention usually is
pronounced with a glottal stop instead of the final n. In the absence of a conven-
tional spelling for glottal stops, this variant pronunciation is commonly spelled
(at)ten(s)hut. The spelling nope may hide a similar glottal-stop pronunciation
[noʔ]. (Spelling and/or the absence of [ʔ] from the inventory of “normal” English
speech sounds may be responsible for the fact that some speakers may actually
pronounce attenshut and nope with a final dental or labial stop.)
Upgrading is not limited to angry speech. In English, expressions like [mma(r)
vǝlǝs] or [biyūtiful] for normal marvel(l)ous [ma(r)vǝlǝs] or beautiful [byūtiful]
serve to express the fact that the speaker feels that something is especially ‘mar-
velous’ or ‘beautiful’.
In Modern English, the expressive consonant doubling, or gemination, in
expressions like [mma(r)velǝs] is a fairly transitory phenomenon, presumably
because the normal language does not have phonetic geminates. (Written double
consonants, as in lass, are pronounced the same as single consonants, as in gas.)
Modern Italian, however, has geminates and, interestingly, expressive gemination
appears in mammà, a word for ‘mother’ which like its cousins in other European
languages (Fr. maman, Germ. Mama, or Engl. mama) belongs to the affective
vocabulary of nursery talk, the form of language used by adults with very young
children and modeled on the babbling of early childhood.
Like the reductions of “downgraded” speech, expressive gemination or glot-
tal-stop insertion affects only individual words and leaves most words unaffected.
Thus, while there is an English nope, there is no gope* for go. Moreover, changed
nope coexists with unchanged no, just as ma’am coexists with madam. Affective
changes, thus, are just as irregular or sporadic as the effects of downgrading.
Moreover, both types of sporadic change play a marginal role in language change.
A much more significant role is played by a group of sporadic changes that
were recognized by the neogrammarians as systematic exceptions to their regular-
ity hypothesis. The two most prominent of these changes are known by the names
dissimilation and metathesis. Before trying to explain their irregularity, it is useful
to take a closer look at the changes.
130 Sound change

Dissimilation is the logical opposite of assimilation. Where assimilation

makes different sounds more similar to each other, dissimilation makes similar
sounds more different. An English example, cited earlier in this chapter, is colonel
pronounced as if it were written cornel, where the first of the two [l]-sounds dis-
similates to [r]. An even more radical dissimilatory step is found in the common
non-standard pronunciation of library as if it were written libary. Here the first of
two [r]-sounds dissimilates by simply disappearing. Examples like contrary, with
both [r]-sounds retained, show that the change is in fact sporadic, not regular.
Note however that some instances of dissimilation have all the appearance
of having once been completely regular. For instance, in the early history of San-
skrit, all but the last of several aspirated stops in a given word lost their aspiration
by dissimilation; see example (16). Like Grimm’s and Verner’s Laws, this change
has received a great amount of attention in historical linguistics; and like these
changes it was named after its discoverer – Grassmann’s Law.

(16) Proto-Indo-European Sanskrit

bhudh-ye-toy budh-ya-tē ‘is awake’
bhe-bhowdh-e bu-bōdh-a ‘was awake’
dhi-dhē-ti da-dhā-ti ‘puts, places’

Metathesis consists of the transposition of sounds within a given word. An

example in modern non-standard English is the pronunciation of ask as [æks] or
[aks]. This pronunciation, actually, has hoary antecedents. Already in Old English
we find the ancestor of this word spelled both as ascian and as acsian. Metathesis
has been quite frequent in English vowel + liquid or liquid + vowel combinations.
But the examples in (17) show that the process was far from regular: Many words
that might have qualified did not undergo the change. Note that English seems to
have been especially “metathesis-happy” in vowel + liquid combinations. Some
of the changes, such as brid > bird, took place already in Old English dialects, and
metatheses occurred even in the perhistory of Old English, as in PGmc. *brestan- >
OE berstan ‘burst’ (which, to make things interesting, fluctuated between bersten
and bresten in Middle English).

(17) a. OE brid NE bird

þrit(t)ig thirty
vs. þrysce thrush
b. beorht bright
vs. word word
 Some types of sound change 131

Metathesis frequently goes beyond individual words and affects whole utterances.
In such cases it has received a special name, spoonerism, after an English cleric
who was famous for his often amusing transpositions, such as Let me sew you to
your sheets instead of the intended Let me show you to your seats.
Spoonerisms suggest that dissimilation and metathesis have a great affinity to
speech errors, in the sense of “faulty” phonetic production. This impression is
reinforced by incidents such as the following. In the early seventies, an announcer
on a radio station in Champaign (Illinois) attempted to say … in rural areas. What
actually came out was something like … in [rūǝl], uh, [rūlǝl], uh, [rūrǝl] areas – I
always have problems with that word. Evidently, the sequence of three liquids, [r …
r … l], caused the announcer considerable difficulties and resulted in two different
dissimilations. Difficult sequences of this type, of course, are the foundation for
tongue twisters, such as Peter Piper picked a peck of pickled peppers. Dissimila-
tions and metatheses are especially frequent when people are tired or drunk (or
both), i. e., when their ability for monitoring their speech production is dimin-
ished. (In addition, of course, tired and drunk speech also is full of reductions
comparable to those of fast speech.)
It may very well be that the only thing distinguishing speech errors like these
from historically attested dissimilations and metatheses is that they remain tem-
porary mistakes, while changes as in bryde > bird, for some reason, caught on and
became a permanent feature of the language.
In addition to dissimilations and metatheses, “faulty speech” abounds in
distant assimilations, such as heroic pouplets for heroic couplets. In fact, tongue
twisters like Peter Piper picked a peck of pickled peppers normally are cleverly
constructed such that our choice of dissimilating the repeated [p]s in sequences
like Peter Piper is balanced by the assimilative influence of the [k]s of words like
peck and pickle.
Like dissimilation and metathesis, distant assimilation frequently catches on
in the historical development of languages. For instance, Old French had the verb
cercher [serčer] ‘search, look for’, from which Engl. search was borrowed. The
expected Modern French outcome is [serše]. Instead we find chercher [šerše] with
distant assimilation of the initial [s] to the later [š], as in the famous expression
cherchez la femme. But again, like dissimilation and metathesis, distant assimila-
tion normally is a sporadic phenomenon.
132 Sound change

6 Why sound change?

6.1. Early views. Ordinary human beings, not influenced by what linguists think
on this matter, generally view all linguistic change, including sound change, as
a matter of decay. This view, in fact, was shared by many of the pioneers of his-
torical linguistics. Only occasionally do we find a dissenting voice, such as that
of Grimm, who attributed his famous sound shift, not to decay, but to an equally
dubious idea – the Germanic people’s “drive for freedom”. (If a “drive for freedom”
were a meaningful factor, we would expect Americans – and others – to fall all
over themselves trying to implement Grimm’s Law.)
Since early Indo-Europeanists came to historical linguistics from a back-
ground in – and great love for – classical languages, it is understandable why
they reacted the way they did. As already noted in Chapter 1, the great classical
languages had five noun cases (Greek), six cases (Latin), or even eight (Sanskrit),
whereas the number of different case endings is vastly reduced in the modern
languages. The majority of Romance languages have no case distinctions what-
soever in their nouns, English only marks genitives by separate endings, and
even a language like German, which nominally preserves four cases, does not
distinguish the different cases as clearly as the classical languages. If you have
grown up loving rich case systems, the modern languages do indeed look as if
they had undergone massive decay. Of course, the premise here is that you love
cases. Some linguists have claimed that those many different case endings of the
classical languages are a useless burden on people’s memory and that languages
like English have improved by getting rid of this unnecessary baggage. This view,
again, rests on an a priori assumption, namely that case endings somehow are
bad for you. Linguists nowadays find it difficult to be comfortable with either
view, considering that the speakers of languages such as English are quite happy
without all those case endings, while speakers of modern “case-rich” languages
such as Finnish or Turkish are just as happy with them. This insight, incidentally,
goes back to the neogrammarians and is one of their many great contributions to
linguistics.
Decay and progress are not the only motivations proposed for linguistic
change. As noted above, Grimm attributed his famous changes to the Germanic
people’s “drive for freedom” which supposedly led to a more energetic pronun-
ciation. As a consequence, the voiceless stops became heavily aspirated stops
which, by later changes, turned into fricatives. Other scholars operated with the
same idea of more energetic pronunciation, but attributed it to factors such as a
move to higher elevations which made it necessary to breathe more vigorously,
or a change in diet which required more energetic chewing. These explanations
and many others like them failed just as much as the notions decay and progress.
 Why sound change? 133

First, we have no evidence suggesting that the Germanic people lived in a rela-
tively mountainous area at the time of Grimm’s Law, or that they changed to a
different diet. Secondly, given our much broader knowledge of linguistic change,
we can say for certain that there is no correlation whatsoever between climate or
diet and linguistic change. (For instance, we don’t find people in mountainous
areas embracing Grimm’s Law; in fact, Liverpool, where a similar change is taking
place, is not known for mountainous terrain.)

6.2. Neogrammarian explanations. Although many individual neogrammari-

ans continued to cling to such older ideas about the reasons for linguistic change,
their major theorists attempted to give different explanations. In these they were
strongly influenced by the work of contemporary phoneticians. Relying on vastly
improved instruments, the phoneticians had come to realize that all human
speech is full of low-level deviations from an idealized norm. No two utterances
are ever completely alike; there is no such thing as an exact repetition (except by
mechanical means). The feeling that there is a norm results only from the fact
that variation follows the classical bell-shaped curve pattern (see the top part
of Illustration 6), with most of the deviations staying fairly close to the norm,
so close in fact that only a trained phonetician can observe them. Building on
this empirically well-established foundation, the neogrammarians proposed that
sound change results from a shift of the idealized “target” within the area of
variation; see the lower part of Illustration 6.

Illustration 6: Sound change as deviation from an idealized norm

134 Sound change

While at first quite appealing, this explanation runs into serious difficulties
once we examine it more closely. The very idea expressed by a bell-shaped curve
is that the deviations from the norm cancel each other out and thereby confirm
the idea of the norm. Why, then, are we to assume that all of a sudden the rules
of the game no longer apply and there is a cumulative deviation in the direction
of a new target?
One might suppose that a certain direction is built into linguistic change in
so far as it leads to simplification. Assimilation, weakening, and loss, the three
most common types of change, certainly can be argued to reduce the amount of
effort required to speak. And the fact that speakers of English find it very diffi-
cult to pronounce the word-initial [kn-] in foreign words like knish or names like
Knut might be considered to corroborate the view that the change of earlier initial
[kn-] to [n-] was a genuine simplification. Similarly, speakers of English, German,
and many other languages find initial [sr-] difficult to pronounce, as in Srinagar,
the name of the capital of Kashmir (in northern India). And lo and behold, PIE
*sr- was eliminated in Germanic by changing to *str-, as in Engl. stream, Germ.
strömen ‘to stream’ vs. Skt. sravati ‘flows’, all containing the PIE root *sr(e)u- ‘to
stream, to flow’. So again, sound change simplified matters, didn’t it? It even has
been claimed that the replacement of trilled [r] by [r] (see § 5.3 above) was a sim-
plification, not just an acoustically based misidentification. And from the per-
spective of those who have it, [r] is in fact simpler than [r].
But, those who have trilled [r] find [r] difficult. And those who have neither
find both sounds difficult. Similarly, speakers of languages that tolerate initial
sr- (Kashmiris, for instance) have no difficulties with this combination and might
consider str- more “complex”. In fact, for cases like [r] vs. [R], or sr- vs. str- it is dif-
ficult to come up with any objective evidence that supports the view that the new
pronunciation is any easier than the old one – except the circular argument that
otherwise the change would not have taken place. For [kn-] vs. [n-] it is much easier
to consider [n-] a simpler structure. Nevertheless, speakers of languages that tol-
erate initial kn-, such as German, have no difficulties at all. Here as elsewhere, the
maxim holds that “even the children speak the language” which has the suppos-
edly more difficult sounds or combinations of sounds. (See also Chapter 1, § 1.)
Even if we dismiss the notion of simplification, it might be claimed that pro-
cesses like assimilation have a built-in directionality and thus would motivate
a cumulative deviation from the norm, as in Illustration 6. After all, one sound
assimilates in the direction of another.
But assimilation comes in many different degrees and varieties. For instance,
if we are given a sequence tm and told to assimilate, we can a priori go into at least
the following different directions: pm, bm, mm with various degrees of assimila-
tion of the first sound to the second; tn, dn, nn with a similar variety in assimi-
 Why sound change? 135

lation of the second to the first; or even tp, db, tt, dd, pp, bb with both sounds
assimilating to each other. Even closely related languages may choose different
paths. For instance, some of the early descendants of Sanskrit changed tm to tp,
others to tt, and yet others to pp, as in. Skt. ātman- ‘self’ : atpan-, attan-, appan-.
Realizing the difficulties with this approach, the neogrammarians came up
with a second explanation. Children learn the basics of their first language
without any instruction, simply by imitating the speech of their elders. In the
process, they may misperceive the norms of their elders and come up with differ-
ent norms of their own.
This explanation, too, seems plausible at first. In fact, even today it can claim
many adherents. But the same problem arises as in the case of the first explana-
tion: Why should the deviations be cumulative, in one direction? In fact, when
we examine early stages of child language we find a great degree of variation,
both for individual children and across different children. Recent research shows
that although early child language deviations and linguistic change show certain
similarities, there are also considerable differences. We only need to look at what
commonly happens to the children of immigrants to convince ourselves that the
effect of parents’ input and of deviations in early language learning are minimal
at best. No matter what the original language of the parents, or the children’s early
attempts to learn it, once children are socialized into peer groups, they quickly
adopt the speech of their peer group. As a consequence, British parents who
proudly maintain their accent for the rest of their lives in America, find – much
to their horror – that their children speak with a “broad midwestern accent”, a
“Southern twang”, or what not, depending on the speech of their peers.
Sensing that this explanation does not provide satisfactory answers either,
some of the neogrammarians proposed that sound change originates as devia-
tions in the idiolect, or individual speech variety, of a prestigious person. Here,
of course, we must again ask why the deviations of such a speaker should be con-
sistent. Now, in some cases, they might result from a speech defect. For instance,
it has been claimed that the French change of trilled [r] to uvular [r] originated
with Louis XIV, who could not articulate [r]. His great prestige supposedly was
responsible for the adoption of the change by other speakers. This claim receives
some support in the fact that the change apparently spread to many urban speak-
ers of German, along with many patterns of behavior that emanated from the court
of Louis XIV. However, the change [r] > [r] has been observed in many other areas
of the world, including isolated rural areas of northern Germany which hardly
were influenced by the court of Louis XIV. In fact, in Germany, the uvular pro-
nunciation has been observed as early as about 1600, well before Louis XIV. Most
important, the idea that change might originate with some prestigious person is
just a thought experiment. There is no empirical evidence whatsoever that behind
136 Sound change

every one of the thousands and thousands of sound changes that have occurred
in human language there has been a famous person.

6.3. Labov and the social motivation of change. The fundamental difficulty
with all three of the explanations proposed by the neogrammarians is that they
are based on thought experiments, not on the observation of changes as they
actually take place. The reason is that the neogrammarians firmly believed that
sound change is unobservable. They came to this conclusion by the following line
of reasoning. Sound change takes place “blindly”, without regard for its effects on
the structure of words or our ability to communicate; the fact that speakers make
no attempts to remedy these effects until the change has run its course indicates
that sound change is unobservable to them.
So far, so good. But then the neogrammarians made a grave mistake: They
assumed that sound change is unobservable not only to speakers, but to linguists
as well. For some reason the neogrammarians failed to realize that the phoneti-
cians had no difficulties in observing the low-level variation in human speech
which ordinary speakers were not aware of. As a consequence, the neogrammar-
ians made no attempts to observe sound change in progress.
Meanwhile, a number of linguists had serious reservations about many of
the neogrammarians’ views, including the belief that sound change and analogy
differ fundamentally from each other, one being “mechanical” and regular, the
other, based on mental associations and irregular. They argued that instead, the
two types of change were fundamentally the same and differed from each other
only in degree. In the hope of finding empirical evidence for this view, they began
to investigate sound changes in progress. By hindsight, some of their results were
quite revealing and would at least have required some serious rethinking about
the nature of change. However, the number of scholars pursuing this “unortho-
dox” line of inquiry was small, much smaller than the orthodox followers of the
neogrammarians.
It was not before the mid-1960s that a major change took place, in response
to a series of detailed empirical investigations by the American scholar William
Labov which were presented clearly and forcefully enough to catch the attention
of most historical linguists.
Like his “unorthodox” predecessors, Labov found that the neogrammarians’
views on sound change were in serious need of revision. Sound change is observ-
able, at least by trained linguists. As sound change takes place, it may be condi-
tioned not just by phonetic factors, but also by such factors as word structure and
meaning. Even more significant, during its propagation, sound change exhibits a
lot of irregularity. It is only in its final outcome that sound change is overwhelm-
ingly regular.
 Why sound change? 137

Interesting as these findings may be, Labov came up with an even more
radical proposal. Sound change and, in fact, all linguistic change is ultimately
motivated not by purely linguistic factors, but by social considerations.
This claim is most strikingly supported by Labov’s study of a recent sound
change on Martha’s Vineyard, an island off the coast of Massachusetts. If we
simply consider the “input” and “output” of the change, there is nothing much
remarkable about it: The vowel [a] was centralized to the position of the mid-cen-
tral vowel [ǝ] in the diphthongs [ay] and [aw], as in right [rayt] > [rǝyt] or rout
[rawt] > [rǝwt]. However, the manner in which the change unfolded is quite
remarkable.
Labov found that at the earliest stage, only a few words exhibited a variation
between [a] and a slightly more centralized variant, only in the diphthong [ay] if
followed by voiceless sounds, and only in the speech of a few individuals.
Somewhere along the way, the variant with centralization was perceived by
speakers as a symbol of identity, differentiating “islanders” from the “mainland-
ers”. (There has been a long tradition of animosity of Martha’s Vineyarders toward
the mainland of Massachusetts, occasionally leading to attempts to secede from
the Commonwealth of Massachusetts.)
When it had come to be perceived as socially relevant, the centralized variant
began to get generalized along a number of different parameters, including the
following:
– the number of speakers using the variable in their speech
– the number of words exhibiting the variant
– the phonetic contexts in which it occurred, including an extension of the var-
iable to the diphthong [aw]
– the degree of centralization (from a slightly centralized [a] toward a fully
mid-central [ǝ])

Moreover, Labov found an interesting correlation between attitude toward the

mainland and the degree to which the centralized variant was adopted. Those
with the most polarized attitude toward the mainland tended to have the highest
degree of centralization, in the largest number of words, and in the broadest range
of phonetic contexts. On the other hand, those with a positive attitude toward the
mainland exhibited the change to a much lesser degree, if at all.
Based on this study and others that he conducted at roughly the same time
Labov proposed that sound change (and all other change, as well) proceeds in the
following manner:
(i) The starting point for sound change is the inherent variability of human
speech. (This is where Labov agreed with the neogrammarians and the nine-
teenth-century phoneticians.)
138 Sound change

(ii) For reasons that perhaps must remain a mystery, a particular variable is
interpreted by a certain group as socially significant. At this point, the variable
ceases to be a “mere performance” variant and takes on not only social, but also
linguistic significance.
(iii) Under the pressure of its social significance or “marking”, the variable
gets generalized to new contexts, in terms of both social and linguistic parame-
ters. (On the role that male : female differences can play in social marking and the
extent to which a change may be generalized, see Chapter 11, § 1.) What makes it
possible for the generalization to continue is the fact that the new pronunciation
does not immediately replace the old one, but that old and new pronunciation
coexist with each other for some time. The variation between old and new pro-
nunciation, then, can be extended to new forms, much along the lines of analog-
ical change. If, say, we have a variation [aw] : [ǝw] in the word house, then this
variation can be extended to, say, mouse or louse.
(iv) If, as usually happens, the process of generalization continues long
enough and without anything to disturb it, the eventual outcome may be a regular
sound change, which affects all instances of the sound, and all speakers in the
speech community.
This view of sound change as socially conditioned has since then been con-
firmed by a number of other studies. It also explains a number of things about
language change which otherwise would be difficult to account for.
One of these is the fact, noted earlier, that even if a language “decides” to have
a specific type of change such as assimilation, the direction of assimilation cannot
be predicted on purely linguistic grounds. This is to be expected under Labov’s
view of linguistic change. The low-level variation of human speech includes a
large variety of small-scale assimilations, going in many different directions.
Which of these is chosen as socially significant is, from the linguistic perspective,
quite arbitrary.
Another aspect of linguistic change explained by Labov’s view is the fact that
there appear to be changes which are moving extremely slowly, so much so that
there can be some legitimate doubt as to whether they will ever reach completion.
One of these, noted already in the early 20th century, is the English change of long
[ū] (as in boot) to short [u] (as in foot) found in many varieties of English (and
difficult to localize geographically). Let us refer to this change as oo-shortening.
Unlike the centralizing change on Martha’s Vineyard (which was completed in
about three generations), this change seems to have been going on for several cen-
turies but still shows no sign of coming to completion. Even now, oo-shortening
is affecting only a few lexical items, and variability is limited to just a few words
(such as roof, room, root), while many others only have the long-vowel pronunci-
ation (such as food, mood, groom, groove).
 The regularity of sound change redux 139

Interestingly, the difference in social connotations between, say, [rūt] and

[rut] (for written root) is minimal. At best we can say that [rūt] may strike some
people as perhaps a little too formal, while others consider [rut] a little too infor-
mal. This is a far cry from the strong social connotations associated with central-
ization on Martha’s Vineyard. If social marking is indeed the driving force behind
linguistic change, then the differences between oo-shortening and the centraliza-
tion of Martha’s Vineyard make eminent sense. The fact that centralization took
place quite rapidly reflects the strong social motivation of the change, while the
slow-moving nature of oo-shortening is explained by its weak social motivation.
As a final illustration of the explanatory power of Labov’s view of linguistic
change, let us take up the issue of typically regular vs. typically sporadic sound
change. As noted earlier, sporadic changes are very similar to speech errors and,
as such, would be highly noticeable. Regular changes, by contrast, may start as
rather minor deviations from an idealized norm, so much so that if somebody said
to us, “Hey, you didn’t make your [a] in house low enough”, we would not even
understand what the person is saying – except, of course, if we are linguists. (But,
as noted in Chapter 1, most speakers are not linguists!)
If these arguments are on the right track, then we can explain the difference
between regular and sporadic sound change. The fact that most sporadic changes
are noticeable as speech errors makes it difficult for them to be seized upon as
marker of social identity. After all, who would want to be identified with a speech
error?
Labov’s new view of linguistic change has proved its usefulness in many
other areas. But perhaps its greatest importance lies in the fact that it provides a
satisfactory answer to the question, “Why sound change?”

7 The regularity of sound change redux

The hypothesis that sound change is (overwhelmingly) regular has recently been
challenged again. For instance, in an attempt to propose etymologies for prehis-
toric mushrooms based on words for ‘foot’ and ‘leg’, somebody has rejected the
regularity hypothesis, with the argument that it “wrongly” rejects “obvious” rela-
tionships such as Lat. caput : Mod. Germ. Kopf, both meaning ‘head’. (Recall that
by Grimm’s Law – who can forget that? – Lat. c- [k] would have to correspond to
Germ. h, and Germ. k- to Lat. g.) Closer examination shows that the meaning ‘head’
of Mod. Germ. Kopf is an innovation. In Old High German the word meant ‘cup’; it
is only in Middle High German that it came to acquire the meaning ‘head’, in addi-
tion to its older meaning ‘cup’. Considering caput and Kopf to be cognates, thus,
140 Sound change

not only violates the regularity principle, it also ignores the evidence of history. In
fact, it also fails to capture an interesting part of Medieval soldiers’ slang, shared
with French and Italian, in which heads were metaphorically equated with pots,
to be smashed in battle; see Chapter 9, § 4.1.
The overwelming regularity of sound change is, in fact, a boon to histori-
cal and comparative linguistics. It is only because of this regularity that we can
establish linguistic relationships such as between the different Indo-European
languages (or even between different stages of the same language). Consider the
easily recognizable similarities in actual Indo-European correspondences such
as (18). If sound change did apply in “random” fashion, i. e., differently for each
different word, in each different language, we might expect correspondences such
as (19). And under those circumstances, claiming that the languages are related
would clearly be preposterous.

(18) Actual early Indo-European correspondences

PIE Sanskrit Greek Latin Old English
treyes trayaḥ treis trēs þrīe ‘three’
tod tad to (is-)tud þæt ‘that’
pǝtḗr pitā patēr pater fæder ‘father’
pod/pōd pad- pod- ped- fōt ‘foot’
krew/krū- krūra- kreas cruor [k-] hrēaw ‘blood, gore’
duō dvau duō duo twā ‘two’
dhē- dhā- thē fē- dǣ-d ‘put, do, deed’

(19) “Virtual” early Indo-European correspondences, with “random” applica-

tion of sound changes that are possible in human languages
PIE Sanskrit Greek Latin Old English
treyes tray deis thrīh ðē ‘three’
tod thaz ti (is-)duy þoh ‘that’
pǝtḗr piyā farēl vaser pæle ‘father’
pod/pōd fiz- pit- ber- fōs ‘foot’
krew/krū- hiva- khrā xōr griu ‘blood, gore’
duō va tā ðō bū ‘two’
dhē- fē- dī thā- hi- ‘put, do, deed’

Moreover, it is the regularity of sound change that allows us to draw a clear dis-
tinction between sound change and other types of change that are in fact irreg-
ular, especially analogical change. Thus, overall, the regularity of sound change
proves to be a foundational concept in historical linguistics, a tool that clarifies
our view and our understanding of language history in case after case after case.
Chapter 5: A
 nalogy and change in word
structure
“I never heard of ‘Uglification’,” Alice ventured to say. “What is it?” The Gryphon lifted
up both its paws in surprise. “Never heard of uglifying!” it exclaimed. “You know what to
beautify is, I suppose?” “Yes,” said Alice doubtfully: “it means – to – make – anything –
prettier.” “Well, then,” the Gryphon went on, “if you don’t know what to uglify is, you are
a simpleton.”
(Lewis Carroll, Alice’s Adventures in Wonderland.)

1 Introduction
As noted in the preceding chapter, early historical linguists believed that linguis-
tic change is tantamount to decay, a falling away from a pristine stage at which
language was perfect and wonderful.
While phonetic deviations were considered the result of slovenly speech, the
major reason for decay in linguistic structure was thought to be “false analogy”.
Ancient Greek and Latin grammar and linguistic philosophy had introduced and
popularized the notion analogy as a designation for structural pattern or regu-
larity. “False” analogy, then, lay in permitting a word to deviate from the “true”
or “proper” pattern. For instance, in Early Modern English, the “true” pattern of
making a plural of the word cow consisted of a vowel change (reflecting the sound
change of umlaut) and the addition of an ending [-n]; hence cow, plural kine.
When the plural form was replaced by the form we use today, cow-s, the word
was permitted to follow the “incorrect” or “false” analogy or pattern of words like
pig : pig-s, horse : horse-s. False analogy was considered characteristic of late,
decaying languages.
One of the great achievements of the neogrammarians, generally overshad-
owed by the fame of their regularity hypothesis, was the insistence that such
notions as decay and false analogy are inappropriate in historical linguistics.
Reconstructed languages and their early offshoots should not be considered any
more perfect than later, or even modern, descendants. There is no indication
whatsoever that speakers of ancient Greek (or, we might add, Old English) were
able to communicate any more effectively than speakers of Byzantine or Modern
Greek (or of Modern English). Similarly, the neogrammarians argued, all linguis-
tic phenomena encountered in observable history must be accepted as possible
in reconstructed Proto-Indo-European as well, or in its early descendants. This
view, which liberated historical linguistics from earlier, prescientific ideas, has

https://doi.org/10.1515/9783110613285-005
142 Analogy and change in word structure

since become known as the uniformitarian hypothesis, a term borrowed from

geology where the notion of uniformitarianism brought about a similar revolution
in thinking.
Consequently, changes of the type cow : kine → cow-s could no longer be
claimed as limited to late, “decaying” languages, but had to be accepted as a pos-
sibility at all stages of linguistic development. This change in perspective led to
the rejection of the term “false analogy” and its replacement by the simple term
analogy.
Note, however, that as a consequence of this relabeling, the term analogy
underwent a considerable change in meaning, from “pattern” or “regularity” to
something like “change in phonetic shape conditioned by non-phonetic factors
in other lexical items, such as word structure, syntactic function, and semantics”.
Analogical change, as defined now, tends to introduce greater phonetic similarity
between semantically, formally, or functionally similar linguistic forms.

2 Relatively systematic analogy

The neogrammarians considered all analogy to be essentially irregular or spo-
radic. Still, they recognized that two analogical processes are considerably more
systematic than others: four-part analogy and leveling. Four-part analogy is
the process that gave rise to Mod. Engl. cow-s. Leveling is responsible for the dis-
appearance in English of many of the paradigmatic alternations created by
Verner’s Law (see Chapter 4, § 3). Recall that one of the results of Verner’s Law
was an alternation between s and r, as on the left side of (1) below. This alter-
nation has been eliminated in Modern English, where s (now pronounced [z]) has
been generalized throughout; see the right side of (1). Both of these processes are
sensitive to very general aspects of word structure, or morphology. At the same
time, both processes, especially four-part analogy, can have profound effects on
morphology. These are taken up in § 4 below.)

(1) OE frēosan (present) : Mod.Engl. freeze

frēas (past sing.) : froze
fruron (past plur.) : froze
(ge)froren (past participle) : frozen

2.1. Leveling. Leveling can be defined as the complete or partial elimination of

paradigmatic alternations that do not seem to signal important differences in
meaning or function. This elimination takes place within a paradigm, i. e., in the
 Relatively systematic analogy 143

set of inflected forms of a given word (or in a subset of such forms), such as the
verbal paradigm in (1). The motivation for leveling has been plausibly expressed
in the slogan one meaning – one form, so that each difference in form in princi-
ple is balanced out by a difference in meaning.
The development in (1) is a perfect example of leveling. As noted earlier, the
Old English paradigmatic alternation was the result of Verner’s Law (Chapter 4,
§ 3), which like all other regular sound change operated without regard for the
complications that it might introduce in word structure (see Chapter 4, § 4). As
long as the alternation was still in place, the word for ‘freeze’ had (at least) two
different variants, one with s, the other with r. The distinction between s and r,
however, did not correlate with any significant difference in meaning or func-
tion, since s is found both in the present and in the singular of the past, while
the other forms of the past have r. Eliminating the alternation, thus, did not
entail sacrificing any important distinctions. Moreover, it served beautifully to
bring the various forms of the verb ‘freeze’ closer to the ideal of “one meaning –
one form”.
The situation is a little different as far as the root vowels are concerned. In
Old English, these vowels differed for all of the four different forms: ēo (whatever
its precise pronunciation) in the present, ēa in the past singular, u in the past
plural, and o in the past participle. Modern English clearly shows some effects of
leveling, in that the three past forms have the same root vowel [ō]. But the present
tense has escaped this leveling and uses a different root vowel, [ī]. How are we to
explain this incomplete or partial leveling?
The answer becomes clear if we ask ourselves, What would have happened if
the leveling had affected the entire paradigm? Clearly, in that case, there would be
no formal distinction between present and past. Leveling, then, must have been
blocked so that the important distinction between present and past tense was not
lost. On the other hand, the vowel alternations in the three past tense forms did
not signal any important distinctions, since all of the forms had the same tense
value.
Leveling is a fairly systematic process. For instance, most of the s : r alter-
nations created by Verner’s Law were eliminated in English. Words which once
exhibited the alternation include Mod. Engl. lose, choose, rise. However, the sys-
tematicity of leveling is a far cry from the regularity of sound change. For instance,
the verb rise had undergone leveling in the prehistory of English, while lose,
freeze, and choose were changed after the Old English period. Even today, one
verb, the notoriously “irregular” verb ‘be’ has escaped leveling in the past tense,
retaining the s : r alternation in was : were. Other, less obvious relics are for-
lorn, whose second element is the old past participle of lose, and rear, originally
a derived form of rise meaning ‘make rise, make grow up’.
144 Analogy and change in word structure

English is not the only language that has leveled out most of the effects of Vern-
er’s Law. German did likewise. But the direction of leveling was different. While in
the dialects underlying Standard English the sibilant was generalized, in German
it was r. Compare example (2), the German counterpart of the English example (1).
The difference between English and German suggests that even if languages start
out with essentially the same alternation, they may differ as to the direction of
leveling. At the same time, it is remarkable that English consistently extended the
sibilant alternant, while German equally consistently generalized the r.

(2) OHG friosan (present) : NHG frieren ‘freeze’

frōs (past sg.) : fror
frurun (past pl.) : froren
(gi)froran (past partic.) : gefroren

Things are even more complex, however. Some dialects of English appear to have
done the same thing as German. For instance in the Newfoundland dialect of
English, some people say frore instead of Stand. Engl. froze.
It appears, then, that the direction of leveling is unpredictable, at least across
different languages and dialects.
In fact, there are further complications. In many cases, it may be perfectly
clear whether a given alternation is significant or not; however, there are cases
where different speakers evidently had different ideas about this matter. For
instance, the effects of umlaut have largely been eliminated in English. Only a
small set of “irregular” forms preserves them, such as tooth : teeth, goose : geese,
foot : feet. In German, by contrast, umlaut is still very much alive, as in the pres-
ent-tense verb paradigm in (3). More than that, in the nouns it has actually been
extended (by four-part analogy, see the next section). Compare example (4) and
note that this extension of umlaut has affected many other nouns. It has been
plausibly argued that this extension serves to make the plural forms more clearly
distinct from the singular. We must therefore conclude that the effects of umlaut
were considered significant in German, in contradistinction to English where they
were considered insignificant.

(3) OHG NHG

sg. 1 faru fahre ‘go’ (compare Engl. fare)
2 ferist fährst
3 ferit fährt
pl. 1 faram(es) fahren
2 faret fahrt
3 farant fahren
 Relatively systematic analogy 145

(4) OHG NHG

sg. pl. sg. pl.
gast gest-i Gast Gäst-e ‘guest’
boum boum-a Baum Baum-e* → Bäum-e ‘tree’

In general, leveling affects what we might call the phonetic representation of

words and affixes. And in so doing it tends to undo the effects of sound change.
Leveling thus confirms the somewhat paradoxical maxim:

Sound change, though inherently regular, creates irregularities in the morphology. Analogy,
though inherently irregular, attempts to undo the effects of sound change and thus to make
morphology more regular.

(Many linguists believe this to be true about all of analogy. The statement,
however, is most saliently true about leveling. Other analogical processes may be
triggered by factors that have no relation to sound change.)
In some cases, the effects of leveling can go beyond the phonetic representa-
tion of words and affixes. This is especially so when, as the result of other lin-
guistic change, a given morphological category exhibits an alternation between
a “Ø-affix” (i. e., the absence of an affix) and an affix with phonetic content. Con-
sider, for instance, the example of English plurals. In Old English, the nominative
plural forms of the most productive masculine and neuter nouns were as in the
left column of (5), with -as in the masculines and -Ø in the neuters. As gender dis-
tinctions became irrelevant in English noun inflection, the two endings came to
be interpreted as variants of a single plural affix, realized as -(e)s in some nouns,
-Ø in others. In Modern English, the phonetically full ending -(e)s generally was
leveled out, at the expense of the Ø-ending; compare the right column of (5). This
development is not surprising, given that the singular : plural distinction can be
considered important. Significantly, however, its effect goes much beyond ordi-
nary leveling: In examples like (1) and (2), individual lexical items are affected,
whereas in (5), the effect is on the overall morphology.

(5) Old English Modern English

sg. pl. sg. pl.
masc. stān stān-as stone stone-s
neut. word word-Ø word word* → word-s

However, just like the German extension of the umlaut pattern in certain noun
plurals in (4), the development in (5) can also be explained in terms of the concept
of four-part analogy (see the next section). This seems to be true for all cases in
146 Analogy and change in word structure

which leveling affects the overall morphology of a given language. Leveling and
four-part analogy thus are not always clearly distinguishable from each other.
In some cases they may “cooperate”; and it is this cooperation, it seems, which
makes it possible for leveling to have a general effect on morphology.
As a kind of postscript, it might be mentioned that the earlier English plural
alternation between -(e)s and -Ø was not always leveled out in favor of -(e)s. A
few sets of English nouns referring to animals have kept the Ø-plural and, in fact,
extended it to words which originally had an s-plural. Two major sets of nouns
can be distinguished, a set of words referring to animals that are or may be hunted
(especially deer) – the “hunt” type – and another set consisting mainly of words
for (different types of) fish (such as the word fish itself, as well as many other
words such as haddock) – the “fish” type. What complicates matters and at the
same time makes things more interesting is that there is a fair amount of vacilla-
tion between the two sets. Finally, there are a few words such as sheep and swine
which are not easily classified.
For the “hunt” type it may be significant that deer, the word for the quin-
tessential object of the hunt, had a Ø-plural even in Old English. But most other
words had original s-plurals (or other forms marked by a plural suffix). In fact,
even in present-day usage, many of the hunt words may have alternative s-plurals
when they are not used in the context of hunting; thus beside expressions such
as they were hunting (wild) fowl or boar we may get references to barnyard fowls
or five boars. (The Old English plurals of fowl and boar were fugl-as and bār-as,
from which the modern fowl-s and boar-s can be derived by regular changes.) The
reason for the somewhat surprising extension of the Ø-ending for hunt words may
have been a social one. One suspects that the starting point for the Ø-plural was
the word for ‘deer’, in which the Ø-plural was inherited, and that this form of the
plural was generalized among the medieval and early modern British gentry as a
grammatical marker associated with the hunt, one of the favorite activities of that
class and one from which the commoners tended to be excluded. It is the same
social setting that sported such hunt-related expressions as an exultation of larks,
a pride of lions, a pack of wolves. Further support for this account of our Ø-plurals
comes from the fact that the words deer and fowl have been semantically special-
ized to refer to animals or birds that are hunted, while OE deor and fugal simply
meant ‘animal’ and ‘bird’, just as their modern German cognates, Tier and Vogel
[f-]. (For another meaning of fowl, see below.) A similar semantic development
can be seen in the word hound, now generally a hunting dog, but in Old English
meaning ‘dog’ in general. (Again, German has preserved the original meaning in
the cognate Hund ‘dog’.)
For the “fish” type, it may be relevant that a fundamental distinction was
made in traditional Roman Catholic society between meat, which could be eaten
 Relatively systematic analogy 147

on most days of the week, and fish, which had to be eaten on Fridays instead
of meat. (Recall that up to the sixteenth century, Roman Catholicism was the
dominant religion in all of England.) This distinction is reflected in fixed, idio-
matic expressions such as neither fish nor flesh (where flesh is used in the older
meaning ‘meat’) or neither fish nor fowl, which contrast fish with other types of
meat that can be consumed on days other than Fridays. Now, when talking about
things such as meat and fish as prepared food, as in Are we having meat or fish
tonight?, we regularly use the singular form in a collective sense, even for words
that normally have s-plurals and even for words that may belong to the hunt type.
Thus we would normally say We are having chicken or (game) fowl for dinner; an
expression like We are having chickens for dinner would suggest that we would
have to consume uncooked individual chickens or that we would have to do the
cooking ourselves. One suspects that because of its special significance in tradi-
tional Roman Catholic society, the word fish and other words referring to different
types of fish extended the collective singular form fish to contexts where a plural
form might have been appropriate, such as They caught six fish. This extension
evidently was a slow process, for the old s-plural (reflecting OE fisc-as) persisted
into the seventeenth century.
Vacillation between the two sets, then, may be explained as resulting from
the fact that words such as fowl can be used both to refer to an animal of the hunt
and to a particular type of meat. Those who use the term fowl only (or mainly) as
a word for food will treat the word as belonging to the fish category, while those
who are familiar with the word in broader contexts, including the hunt, may treat
it as belonging to the hunt category.
Finally, the Ø-plurals sheep and swine correspond to Old English Ø-plurals
and may therefore simply be archaic retentions. The retention of the Ø-plural
could perhaps be motivated by the fact that sheep and swine are thought of not so
much as individuals, but as collections of animals; however, related words such
as lamb and pig have s-plurals, even though they could just as easily be thought
of as collections of animals.

2.2. Four-part analogy: The process designated four-part analogy involves the
remaking of a morphologically “derived” formation on the model of another, gen-
erally more productive derivational pattern by means of an analogy which can be
expressed by a proportion involving four parts:
a : a’
::
b : X (X → b’)
or “a is to a’ as b is to X (solving for X, X → b’)”. The process may also be charac-
terized as a proportional analogy; but that term is used in a broader meaning,
148 Analogy and change in word structure

to include processes such as backformation and hypercorrection; see later in this

section as well as § 6 below.
Four-part analogy is the process that gave rise to the regular s-plural of
Mod. Engl. cow; see (6). The proportion in (6) reflects the fact that most English
plurals are morphologically derived from the corresponding singular by the addi-
tion of the plural marker -s. Proportions like the one in (6) have characteristics
very similar to mathematical equations. The new form, cow-s, therefore can be
obtained by solving for X, on the derived – plural – side of the equation.

(6) Engl. stone : stones

arm : arms
… : …
cow : X, X → cows (replacing earlier kine)

Now, not all imaginable proportions lead to analogical replacements. Some fail
to do so simply out of “inertia”. For instance, words like Mod. Engl. tooth : teeth,
goose : geese, foot : feet have not changed their plurals to tooths*, gooses*, foots*.
Given the fact that analogy normally is not regular, the failure of certain forms to
undergo a possible analogical replacement should not come as a surprise. Still, if
somebody were to say tooths or the like, we would be able to understand what she
or he is saying – even if we might consider the form to be wrong.
Our reaction would be very different if we heard someone say something like
thang as a “past tense” of thing. True, we can set up a neat proportion of the type
ring : rang = thing : X. But there is no morphological relation between ring, a verb,
and thing, a noun. This lack of relationship makes the proportion meaningless.
The derivational pattern giving rise to rang as the past tense of ring simply cannot
be extended to thing. This shows that to be meaningful, four-part analogy has to
operate on forms that are morphologically related. (The vernacular pronunciation
[θæŋ] of thing reflects sound change, not analogy; [θæŋ] does not mean some-
thing like ‘a thing of the past’.)
Morphological relationship, however, is not enough. If someone were to say
that a skunk roke or rought to high heaven, our reaction would be the same as
for thang – a complete lack of comprehension. Now, roke and rought can easily
be motivated by proportions of the type speak : spoke = reek : X or seek : sought
= reek : X. More than that, if we reversed the proportions, as in reek : reeked =
speak : X and solved for X = speaked, the situation would be quite different. True,
as in the case of foots, people might take exception to our “bad grammar”, but
they would have no difficulty understanding us. The reason for this difference in
reaction is that patterns of the type speak : spoke, seek : sought are not productive
in English, while the type reek : reeked is. Four-part analogy, thus, has a greater
 Relatively systematic analogy 149

chance for success if it extends a productive pattern of morphological deriva-

tion.
This is not to say that patterns that are less productive cannot be generalized.
For instance, the verb dive made its past tense with the productive suffix -(e)d as
early as Old English. Even so, Modern English permits dove as an alternative form
of dived. The model for this development must have been drive : drove. But this
pattern of forming past tenses is not productive in English.
Some developments of this type, which run counter to expectation, are diffi-
cult to explain. For others, however, it is possible to think of an explanation.
Consider for instance the case of Engl. bring : brought : brought. In many
vernacular forms of English, this pattern is replaced by bring : brang : brung
or bring : brung : brung, on the model of verbs like sing : sang : sung or
fling : flung : flung. Both of these model patterns are “irregular” and non-pro-
ductive. Still, the patterns sing : sang : sung, fling : flung : flung are found
with a fair number of other verbs and thus are less irregular than the pattern
bring : brought : brought, which is limited to just one word. Replacing brought by
brang or brung, thus, makes the inflection of bring more regular. True, it is not as
regular as a potential bring : bringed : bringed; but that might be considered going
too far, considering that this is an irregular verb and that even words like bring
and fling have resisted complete regularization.
Even cases where a productive pattern is replaced by a non-productive one
can sometimes be motivated. For instance, the original Modern English plural of
dwarf is the completely regular form dwarf-s, but in many varieties of American
English, the plural has changed to dwarv-es, whose formation follows a much
more irregular pattern.
Α likely model for this change is the semantically closely related word elf,
whose plural elves has been irregular for many centuries. And the reason for the
extension of this irregular pattern to dwarf may be that both words belong to the
vocabulary of fairy tales and similar types of stories, in which elves make a fre-
quent appearance and in which at least one other word ending in f and making an
irregular plural is commonly used, namely wolf. The irregular plurals in -ves, thus,
may have been felt to be characteristic of fairy-tale or fantasy characters, much as
the Ø-plural of words like deer appears to have been interpreted as characteristic
of hunt-related terms.
Though productivity plays a major role in four-part analogy, it is by no means
clear what makes a particular type of formation productive and others, non-pro-
ductive. In this regard, note that the English s-plural has not always been the
single most productive form of making plurals. In Middle English it competed
with another plural formation, using the suffix -(e)n; and in some Middle English
words, this suffix actually replaced the s-ending. Thus, while the Old English
150 Analogy and change in word structure

plural of the word shoe had an s-plural, Middle English shows both s(c)hoes and
s(c)hoon or s(c)hoen. In Modern English, by contrast, the n-plural remains only in
a few relics: oxen, children, brethren, and early Mod. Engl. kine.
Some scholars have attributed the eventual victory of the s-plural to outside
influence. After 1066, French exerted a great influence on English, especially in the
area of the lexicon. And the normal, productive plural marker of French is -s, which
in medieval times was still pronounced as [s]. However, the idea that the s-plural
of French lent a helping hand in the eventual victory of the English s-plural runs
into difficulties – the victory took place much later, at a time when the importance
and potential influence of French had greatly diminished. During the time of great-
est French influence, the s-plural had serious competition from the n-plural. More
than that, even if the s-plural did become productive due to French influence, the
Middle English productivity of the n-plural cannot be attributed to outside influ-
ence. The issue of how productivity arises, therefore, remains an open question.
The notions productivity and morphological relatedness, important as they
may be, are not sufficient to make four-part analogy successful. Consider the reac-
tion you might have if somebody talked to you about a Chinee who will be coming
to visit. Your reaction would probably be very much the same as in the case of
thang (as a “past tense” of the noun thing) or the skunk roke to high heaven.
As a matter of fact, though currently not acceptable as an English word,
Chinee is found in eighteenth-century travel descriptions. And it is perfectly pos-
sible to explain Chinee on the basis of a productive plural : singular proportion,
as in (7). In contrast to thang, this proportion cannot be simply ruled out on the
grounds that there is no morphological relatedness, for the word Chinese clearly
can be used in the sense of a plural, as in The Chinese are a people in East Asia.

(7) employee-s [-īz] : employee [-ī]

Chinese [-īz] : X.

The fact that the word Chinee nevertheless causes great difficulties can be explained
by observing that we have solved the equation on the “wrong” side – not on the
morphologically derived side, but on the basic side. To be fully successful, four-
part analogy evidently needs to be solved on the derived side of the proportion.
This does not mean that developments of the type (7) are impossible. Even
though much rarer than “well-behaved” developments, they are sufficiently
common to have received a name of their own, backformation.
Examples of fully successful backformations are Mod. Engl. pea, sherry, and
orate. The ancestor of Mod. Engl. pea is early Mod. Engl. pease, a mass noun like
rice. The word survives in the nursery rhyme Pease porridge hot …. Now, a mass
of pease consists of individual pieces. Moreover, pease happens to end in [-z],
 Sporadic or non-systematic analogy 151

the variant of the plural marker -s that would be appropriate after vowel. It was
therefore possible to reinterpret the word as a plural, referring to a plurality of
individual pieces. Given this reinterpretation, the pattern beans : bean (etc.) =
peas(e) : X made it possible to create a new singular, pea.
The word sherry came about in very much the same way. The word entered
English as referring to the fortified wine coming from the city of Xerez (de la Fron-
tera), now Jerez de la Frontera, which at that time was pronounced [šere(t)s], and
was nativized as sherries. (This form is found for instance in Shakespeare’s Henry
IV, Part 2.) The rest is history, as it were.
Orate and many (but not all) other verbs ending in -ate are backformations
from “agent nouns” like orator, created on the model speak-er [-ǝ(r)] : speak =
orat-or [-ǝ(r)] : X. Verbs of this sort are motivated by the fact that nouns like orator
clearly designate someone who engages in a particular activity, but as the result
of historical accident there was no morphologically related verb to express that
activity. Backformation solved the problem by furnishing a handy verb from
which the agent noun can be considered to be derived.
Unlike leveling, four-part analogy has a clear and very strong impact on mor-
phology. As we have seen, it can lead to the integration of particular words into
more productive morphological patterns, sometimes even less productive ones.
And, as noted in § 2.1, in combination with leveling it can bring about an exten-
sion of particular affixes (as in example (5)), or even the elimination of affixes (as
in the case of deer, fish, fowl).
Combined with the effects of sound change, especially the very common phe-
nomenon of loss in final syllables, four-part analogy (with or without the help
of leveling) can have far-reaching repercussions for morphology. These are dis-
cussed in fuller detail in § 4 below.

3 Sporadic or non-systematic analogy

As we have seen in the preceding section, leveling and four-part analogy (if oper-
ating in the right direction or in favor of productive forms) can be highly sys-
tematic, even if not regular in the manner of sound change. The reason for this
systematicity no doubt lies in the fact that both types of change operate within
well-defined and fairly general parameters. Leveling works within paradigms,
and most paradigms accommodate more than one or two words. In fact, produc-
tive paradigms may work for hundreds of words. Four-part analogy operates with
derivational patterns, especially productive ones which, again, may accommo-
date hundreds of words.
152 Analogy and change in word structure

Most other types of analogical influence on linguistic form are not condi-
tioned by such well-defined and general parameters. Instead, by their very nature
they tend to affect only one or two words at a time. This does not mean that they
are rare. In fact, they are quite common. But their effect usually is much more
“helter-skelter” than that of four-part analogy and leveling.

3.1. Blending, contamination, and similar processes. Consider for instance

the phenomenon of blending, also referred to as portmanteau. Example (8), a
citation from Lewis Carroll’s Through the Looking-Glass, provides an excellent
definition of the process.

(8) Humpty Dumpty to Alice:

Well, slithy means “lithe and slimy”. “Lithe” is the same as “active”. You
see it’s like a portmanteau – there are two meanings packed into one
word.

As in the slithy of example (8), blending often is a deliberate process, employed to

coin new words, often with a wisp of humor (at least when they are first coined).
Compare the examples in (9). In many cases, the results of blending express some-
thing like a compromise between “competing” words, as in the examples in (9a).
For instance, the word brunch designates a meal that combines properties of both
breakfast and lunch. But blendings may also “telescope” the components of a
compound expression into a single, more neatly packaged unit. Thus, telethon in
(9b) is a telephone marathon, an American fundraising event calling for telephone
pledges, and long and exhausting like a marathon race. Such “packaging” makes
blends especially useful for advertising purposes, as in (9c).

(9) Blending Sources

a. chortle chuckle : snort (Lewis Carroll)
brunch breakfast : lunch
broccoflower broccoli : cauliflower
motel motor : hotel
boatel a boat converted into a hotel
b. Dixiecrat a Dixie Democrat (i. e., from the US South or Dixieland)
exercycle exercise (bi)cycle
telecast television newscast
bik(e)athon bike marathon
telethon telephone marathon
c. Toyotathon a Toyota marathon sale
Croisan’wich a croissant sandwich
 Sporadic or non-systematic analogy 153

Blending may also occur unconsciously, as a kind of compromise between actu-

ally competing forms. For instance, the first form in (10), frequently heard among
children, is a compromise or cross between the correct form feet of adult lan-
guage and the regularized form foots. Similarly, the early Modern English plural
form kine appears to be a compromise between the phonetically regular outcome
of OE cy [kǖ], kye (preserved in British English dialects), and an n-plural of the
type oxen, *cow-(e)n; see the second example in (10). Other similar non-deliberate
blendings are given in the remainder of (10).

(10) feets feet X foot-s

kine kye X *cow-(e)n
children child-er (OE cildru) X *child-en
irregardless irrespective X regardless
near miss near hit X miss (in aviation)

The last two of these blendings are commonly considered substandard or incor-
rect, and a fair amount of ink has been spilled on near miss as being a logically
flawed recent development. Whatever its logical shortcomings, near miss has
been around for quite some time: It was used, for instance, in an Allied news
report on the bombing of the French city of Caen at the end of World War II.
Blends even may affect whole utterances, at least in one-time speech errors.
For instance, the following sentence was heard on National Public Radio (7 Feb-
ruary 1991): Thomas’s death will be missed. Evidently, the person who scripted this
line conflated the following two sentences into one: Thomas will be missed and
Thomas’s death will be regretted/mourned.
A process very similar to blending, and sometimes difficult to distinguish
from it, is contamination. Before trying to define what is meant by the term, let
us take a brief look at an example.
Latin had two adjectives whose meanings were polar opposites, gravis ‘heavy’
and levis ‘light’. (Such words are called antonyms.) In the form of Latin which,
as “Proto-Romance”, underlies the modern Romance languages, gravis changed
to grevis, hence OFr. grief, OSpan. grieve, Ital. greve ‘heavy’. (Original Latin gravis
would have yielded OFr. gref, OSpan. grave, Ital. grave. Mod. Fr. grave ‘heavy’
clearly is a later borrowing from Latin, and the grave of Modern Italian and
Spanish is no doubt likewise.) What has happened here is that gravis has been
“contaminated” by levis by adopting the pronunciation grevis, whose e is closer
to the e of levis.
Changes of this sort are very common in antonyms, as well as in numerals.
Both antonyms and numerals are often uttered in close succession (with perhaps
a short conjunction intervening), as in Is it heavy or light? or One two three. One
154 Analogy and change in word structure

possible interpretation, therefore, is that contamination starts out as a distant

assimilation, a speech error similar to Spoonerisms (see Chapter 4, § 5.5). The fact
that the assimilation catches on would be due to the close semantic relationship
between the words in question. A different interpretation, not necessarily incom-
patible with the preceding, is that antonyms and numerals form psychologically
closely related sets and that their adjacency in mental storage might favor some
kind of formal assimilation, irrespective of whether they occur next to each other
in actual speech.
Some additional examples of contamination may be found in (11). As these
examples show, contamination differs from blending in that the two words or
their meanings are not telescoped into one but remain as distinct words, albeit
phonetically more similar ones. For instance, Engl. eleven, whose n results from
the influence of ten, does not designate something like ‘101/2’, but just plain ‘11’.

(11) Engl. male : female ⇐ Fr. male : femelle

Engl. ten : eleven PGmc. *tehun : *ainlif
(compare Germ. zehn : elf)
Dial.Gk. hepta : hoktō Standard Gk. hepta : oktō ‘7’ : ‘8’

A nearly opposite development is observed in cases like U.S. Armed Forces Engl.
niner for nine, used to distinguish this numeral from five under poor communi-
cation conditions. Similar circumstances have given rise to Germ. eins, zwo, drei
for eins, zwei, drei, and Juno, Julei for Juni, Juli. In some cases, such differentia-
tions involve selection of a dialectal or archaic variant (Germ. zwo); Germ. Julei
with -[ai] may have been modeled on Engl. July, and Juno is the name of a Roman
Goddess, boldly substituted for the similar sounding Juni to avoid confusion. In
cases like U.S. Armed Forces Engl. niner, the differentiation seems to result from
deliberate distortion. As examples like niner vs. five show, in contrast to contam-
ination, distortions or substitutions of this type are not limited to neighboring
numerals.
The examples of blending and contamination so far examined would certainly
entitle us to believe that neither process has much chance of being systematic. In
each case, the analogical process involves just two lexical items – in stark contrast
to leveling and four-part analogy which potentially affect hundreds of words.
Nevertheless, some blendings have broader effects. For instance, once forms
like bik(e)athon and telethon in (9b) have arisen, it is possible to reanalyse them as
containing a suffix -(a)thon. Four-part analogy, then, can extend this suffix to new
forms, such as rentathon or saleathon. (Expressions like this are special favorites
of the advertising industry, as in (9c) above. A recent TV commercial by one car
maker pokes fun at the competitor’s saleathon by staging a fake thonathon.)
 Sporadic or non-systematic analogy 155

Other processes, very similar to blending and contamination, can lead to

patterns that are quite productive. Consider for instance the English onomatopo-
etic set bang, bash, batter; clang, clap, clash, clatter; crack, crash; smack, smash,
smatter; …. There are at least thirty-two words of this sort in current use; and
several others sprang up during the history of English, only to die again. If the
written tradition of English can be trusted, the set started out from very humble
beginnings. Only one word, clatter, is attested as early as the eleventh century,
and the thirteenth century adds dash. It is in the fourteenth century that the
pattern begins to take off, with the appearance of batter, clap, crack, patter, rap,
ram, rap, rattle, slash, smatter. And some words, such as whang and wham, have
made their appearance as late as the nineteenth and twentieth centuries.
The exact process by which this set propagated itself is not entirely certain.
It is tempting to set up proportions of the type bang : batter = clang : X. But the
problem is that words of this type do not seem to have any morphological struc-
ture to speak of. What, for instance, would be the meaning or function of b- or of
-ang or -atter? Proportions of this type would not be much better than sing : sang =
thing : X. Rather, what seems to be going on is some kind of more vague influence
of forms on each other, similar to what we find in blending and contamination.
Some scholars have referred to very similar phenomena as rhyming forma-
tion: Words that agree in part of their phonetic representation and are similar
in meaning often get to be more similar by coming to rhyme with each other or,
alternatively, to alliterate, i. e., to acquire identical initial consonants. While end
rhyme is a very common principle of poetry, alliteration can serve a similar poetic
function, as for instance in early Germanic.
Sometimes, this rhyming principle operates simply by combining compati-
ble pre-existing words that happen to rhyme or alliterate, such as Engl. wining
and dining, or rest and relaxation, bed and board, rock and roll. In most cases,
such expressions are quite transparent. But note occasional examples like kit and
caboodle. What’s a caboodle?
A more interesting type is represented by such things as helter-skelter,
hoity-toity, hocuspocus, hodgepodge, hurdy-gurdy, hurly-burly, nitwit, pell-mell. In
some words of this type, it is possible to account for the origin of one or, even more
rarely, both of the rhyming components. For instance, in pêle-mêle, the French
source of pell-mell, it is possible to etymologize the second element as mêle ‘mix’.
But the source of the other part of the word is wrapped in mystery. In many other
words of this type, the entire etymology is uncertain.
In English, expressions of this type are fairly common if they involve rhyme.
Pure alliterating examples are much more difficult to find. Perhaps words like
know-nothing, go-getter, two-timer, tattle-tale belong here. Much more common
is the type flim-flam, riff-raff, zig-zag, mish-mash, pitter-patter; drip-drop, flip-
156 Analogy and change in word structure

flop, sing-song; hee-haw, gew-gaw. Here we find not only repetition of the initial
consonants, but also the final ones (if there are any). In addition, most of these
words exhibit one of a very limited set of vowel alternations, which recur in such
sets as sing : sang or sing : song. Linguists refer to these alternations as ablaut
or apophony.
There is thus a strong general tendency toward rhyming formation. Moreo-
ver, words of the type drip-drop, pitter-patter are remarkably similar to the words
in sets like bang : bash : batter. Not only are they onomatopoetic, some of the
words agree in the phonetic shape of their rhyme as well. Compare for instance
pitter-patter and batter, shatter, etc. Finally there are rhyming pairs in other ono-
matopoetic words, such as thump : bump, drip : blip.
Given these facts we are perhaps entitled to explain sets like bang : bash : batter
etc. as resulting from rhyming formation.
In some languages, rhyming patterns of this type may become generalized, to
such an extent that we can talk about a morphologically productive pattern and
invoke four-part analogy (or some kind of rule) to account for the propagation
of the pattern. Many varieties of English have acquired just such a pattern from
Yiddish. Compare such expressions as school-shmool, linguistics-shminguistics, or
nice-shmice. Here the second part of the rhyming word has a completely arbitrary
and standardized initial consonant combination, and the connotations of the com-
bination are completely predictable. For instance, school-shmool means something
like ‘school – who cares?’ or ‘school – who needs it?’ Moreover, given the right
circumstances, the pattern can be extended to just about any noun or adjective.
Sign languages, too, offer sporadic instances where one sign affects the form of
another because of a close relationship in meaning. For example, in American Sign
Language the sign for ‘patient’ (adject.) used to be formed with the head nodding
down and an index finger drawn against the lips. Now it is formed by a downward
movement of the hand over the lips – the hand movement telescopes the origi-
nal separate gestures of the head and of the hand into a single sign. In addition,
however, another development has taken place. The sign for ‘patient’ is no longer
formed with the index finger, but with the hand in the shape of a fist and the back
of the thumb touching the lips. This change has been attributed to contamination
by the semantically related sign for ‘suffer’ which employs the fist shape.

3.2. Other sporadic processes. While blending and contamination involve at

least two words at a time, other analogical developments tend to affect single
words and thus would normally be highly sporadic.
One of these is reanalysis or reinterpretation of morphological structure.
If for instance a child responds to the admonishment Behave! by saying But I’m
being have, he or she evidently has misunderstood the morphology of behave.
 Sporadic or non-systematic analogy 157

Instead of interpreting the expression as a single lexical item, the child has (re-)
interpreted it as being morphologically composite, parallel to the many other
commands that grown-ups tend to direct at children, such as Be quiet! or Be nice!
Given this interpretation, the response But I’m being have is no different from
responses like But I’m being quiet.
One wonders how many reinterpretations of this sort take place in the early
stages of first-language learning. That this is a common phenomenon is suggested
by recurrent anecdotes. One that has been so widely told and retold as to have
entered English folklore is the following. When asked why she kept calling her
bear Gladly, a child answered Oh you silly, don’t you know? It’s Gladly, my cross-
eyed bear. Evidently the child had misunderstood the beginning words of the
well-known church hymn, Gladly, my cross I’d bear. Other children are reported
to have said things like Lead us not into Penn Station for the passage Lead us not
into temptation in the Lord’s Prayer.
Most such reinterpretations of early childhood do not make it into adult
speech or become part of the language in general. This is, of course, not surpris-
ing, given Labov’s observation that linguistic change takes place in the post-child-
hood social setting of peer groups. (See Chapter 4, § 6.3.)
Nevertheless, some reinterpretations have caught on. This seems to be espe-
cially common in the area of “linking” or liaison phenomena. In English, for
instance, the indefinite article has two forms, an which is used before vowel, and
a which occurs elsewhere. Now, the n of an generally is phonetically linked to
the following vowel, so that things like an apple are pronounced as if they were
written a napple. This fact is often drawn on for the purpose of punning, as in He’s
a nice man : He’s an ice man.
What is important for present purposes is that liaison can create ambigui-
ties, especially in rarer, less well-known words. Should the sequence a-n-vowel be
morphologically resolved as a plus a word beginning in n + vowel, or as an plus a
word beginning in a vowel? Evidently, such uncertainties have in some cases led
to reanalyses. For instance, the two English words napkin and apron ultimately
both derive from a French word stem nape ‘(table) cloth’. While napkin has faith-
fully preserved the original initial n, the combination a-n-apron must have been
reinterpreted as an + apron. Even common words may be affected by reanalysis.
Witness the common English expression A whole nother one, instead of “correct”
A whole other one, with nother reanalyzed in the combination another.
Just like blending and contamination, reanalysis can sometimes lead to new
morphological patterns. In fact, we already have seen one example of such rea-
nalysis, the -(a)thon of forms like bik(e)athon, telethon, etc. While the starting
point for this particular reanalysis lies in blendings, reinterpretation may affect
much more mundane morphological structures. For instance, the -ician of Engl.
158 Analogy and change in word structure

beautician and mortician arose by a reanalysis of forms like electrician, mathema-

tician, logician as containing a suffix -ician indicating a skilled professional. As a
matter of historical fact, of course, words like electrician are to be analysed as elec-
tric plus a suffix -ian which, among other things, indicates a skilled professional.
Words like beautician and mortician therefore were attacked as “incorrect” when
they first were used, because there is no beautic* or mortic* from which they could
be derived. The fact that they have become part of the general English vocabulary
confirms the observation in Chapter 1 that attempts by purists to stem the tide of
linguistic change by and large are ineffectual, and that the ultimate authority lies
with the speaker.
The role of linguistically unsophisticated speakers is even more noticeable
in two closely related analogical processes, “recomposition” and “folk etymol-
ogy”, especially in the latter. To understand the nature of these processes it is
useful to observe that sound change often obscures the structure of compounds.
For instance, Mod. Engl. lord derives from OE hlāf-weard by a series of regular
sound changes, and lady goes back to hlāf-dige. The Old English words literally
meant ‘bread-warden’ and ‘bread-kneader’. Given the importance of bread as a
staple of medieval diet, those who controlled the production and distribution of
bread in an extended household wielded considerable power. This accounts for
the fact that the meanings of these words quickly acquired much more powerful
and “noble” connotations.
Similarly, daisy derives from MEngl. daies-ei(e) ‘day’s eye’ = ‘(little) sun’,
and boatswain (a petty officer on a boat), pronounced [bōsn̥ ], comes from MEngl.
bot(e)swayn(e) ‘young man on the boat’.
Many “faded” compounds of this type have remained unchanged, effec-
tively becoming morphologically non-composite lexical items. Others have been
renewed, made more transparent, by recomposition. Compare for instance the
examples in (12). In the word hussy of (12a), sound change obscured the outcome
of OE hūswīf ‘house-wife’ beyond recognition; and the meaning of the word
changed drastically, too (on this see Chapter 7). The sound changes involved were
of a sort that affect longer words which contain internal consonant combinations
of the type [sw], [tsw], etc. These changes would naturally fail to apply to OE
hūs ‘house’ and wīf ‘woman, wife’ when they occurred by themselves, and these
words simply developed into Mod. Engl. house and wife. Modern English, then,
in effect undid the results of the sound changes that led to hussy, by recombining
the independent words house and wife. The examples in (12b) may perhaps be
explained along the same lines. Here, too, the result of sound change as it applies
in longer words appears to have been undone by recombining the component
parts in the shape that they received as independent words. However, the exam-
ples in (12b) can alternatively be explained as spelling pronunciations. The com-
 Sporadic or non-systematic analogy 159

ponent parts are spelled the same as when they are used as independent words.
The pronunciation associated with the spelling of the independent words then
may have been introduced into the compounds.

(12) Early form of the compound Outcome by sound Recomposition

change
a. OE hūswīf hussy1 housewife
b. OE tōweard toward [tɔrd] toward [tuwɔrd]
OE forhēafod forehead [fɔrɨd] forehead [fɔrhɛd]
(1In British English, this word was until recently pronounced as if written huzzif or
huzzive.)

Examples of the type (12) represent cases where the results of recomposition are
historically correct. For instance, the elements in Mod. Engl. housewife are his-
torically the same as those in OE hūswīf. As noted in Chapter 1, however, most
speakers are not linguists. They are therefore not aware of what the historically
correct etymology of a given word may be. All they are concerned with is that a
particular form “looks like it ought to be a compound”, but is not easily recogniz-
able as such.
There are quite a few such words in English. Note for instance cranberry or
raspberry, which are clearly compounds containing the word berry as their second
element and in this respect resemble fully transparent compounds like blackberry
and blueberry. But what is cran- or rasp- (pronounced [raz/ræz])?
Many words of this type go on their merry way without being remade. Others,
however, are subjected to attempts to make them more transparent. For instance,
Old English had a compound brȳd-guma ‘man of the bride’, consisting of brȳd
‘bride’ and guma ‘man’. By regular sound change, this compound would have
come out as something like bridegum or bridgum. If OE guma ‘man’ had survived
into Modern English, this outcome would have caused no great difficulties, except
perhaps a need for recomposition. However, the word guma has been lost from the
English vocabulary. Still, the word brid(e)gum looks like a compound, containing
the word bride as its first element. But what about the second part? What does it
mean? Faced with this quandary, English speakers at a certain point replaced the
expected gum by a word that sounds similar and has a recognizable meaning,
namely groom. The outcome, bridegroom, has the advantage of being a transpar-
ent compound. And if the original meaning of groom, ‘attendant, servant’, isn’t as
appropriate as ‘man’ (the original meaning of guma), well, that’s too bad. At least
the word has an “etymology”.
Historical linguists refer to such historically incorrect etymologies as folk
etymology or popular etymology.
160 Analogy and change in word structure

Folk etymology, because it does not depend on knowledge of the history of

words, is not necessarily limited to words that originally were compounds. It is
sufficient that in some way they “look” like compounds and therefore should also
“behave” like compounds. For instance, the word asparagus, a borrowing from
Greek via Latin, can be considered a compound in that it is a bit “too long” to be
an ordinary, uncompounded word. In some varieties of English, the word con-
sequently has received a popular etymology, being remade into sparrow-grass.
Those who do not speak this variety of English will find this particular popular
etymology amusing or even ridiculous. But this is a common reaction to folk ety-
mology when it first arises. Those who do use the word sparrow-grass are perfectly
convinced of its semantic appropriateness and will justify their belief by saying
things like You see, it has long stalks like grass, and sparrows and other birds hop
around in it. (As a matter of historical fact, sparrow grass has been attested since
the seventeenth century.)
In fact, many words that we use in Modern English without feeling that they
are “amusing” or even “ridiculous” attempts at folk etymology, do owe their
present shape to this process. Some additional examples are given in (13).

(13) Modern English word Source

andiron (also handiron, endiron) Fr. andier ‘firedog’
breaker ‘water cask on lifeboat’ Sp. bareca ‘barrel’
buckaroo Sp. vaquero ‘cowboy’, with some help
from Engl. buck (?)
buckwheat Dutch boekweite ‘beech wheat’
carry-all Fr. carriole ‘small carriage’
chaise lounge Fr. chaise longue ‘long chair; couch’

In some cases, popular etymology has given rise to further developments. A case
in point is gridiron – like andiron above (see Chapter 1, § 4), an incomplete or
partial popular etymology, with substitution of the recognizable English word
iron for the final element of a French borrowing, which in this case was gredyre.
Once gridiron had come into use, the question must have arisen as to what the
first element, grid, might mean. Evidently, somebody made the inference that this
part is what distinguishes ‘gridirons’ from other ‘irons’ and thus carries the basic
meaning in the word gridiron. Something like backformation, then, made it pos-
sible to extract grid from gridiron and to use it as an independent word.
The development of grid from gridiron can alternatively be explained by another
process, ellipsis, the elimination or deletion of what is considered redundant ver-
biage. The fact that gridirons are prototypically made of iron may have made the
element iron appear to be redundant. As a consequence, it could be deleted.
 Sporadic or non-systematic analogy 161

Ordinarily, ellipsis is quite limited in its application, as can be illustrated

with the following additional example. With the advent of the women’s libera-
tion movement, males opposed to equality for women often were referred to as
male chauvinists. This expression contained a metaphorical extension of the word
chauvinist which earlier had been employed to refer to an overly ardent nation-
alist, a jingoist. In the context male chauvinist, however, the word began to be
reinterpreted as meaning a ‘male supremacist’. As a consequence of this reinter-
pretation, male chauvinist could be considered redundant, being equivalent to
saying male male supremacist. At this point, ellipsis eliminated the redundancy
by deleting the redundant element male – and in the process establishing a new
meaning for chauvinist.
Under special circumstances, ellipsis can go much farther. For instance, the
original Old English word of negation was ne, as in ic ne wāt ‘I don’t know’. This
ordinary mode of negation could be reinforced by the hyperbolic use of either wiht
‘something, anything’ or nāwiht ‘nothing, not anything’; see example (14), Stage I.
(Evidently, at this stage the phenomenon of negative spread, as in non-stand-
ard Mod. Engl. I don’t know nothing, was not yet considered unacceptable in the
standard language.) As time progressed, the hyperbolic force of (nā)wiht began
to fade. First the word became an all-purpose emphasizer with negation, as in
Stage II. Over time, even the emphatic force began to fade, and the form nāwiht
came to be interpreted as part of a composite, “discontinuous” marker of nega-
tion ne … nāwiht; see Stage III. But once ordinary negation was expressed by two
words, ne and nāwiht, the stage was set for ellipsis to come in and to eliminate the
seeming redundancy. The result was that ne, the word that originally had been the
marker of negation, was deleted, and not, the reflex of originally hyperbolic nāwiht
became the only marker of negation. See Stage IV. The result of this development,
however, was not limited to effectively replacing the lexical item ne by nāwiht > not.
It affected the entire syntax of negation, not only in the sample sentences of (14),
but in all sentences: Whereas in Old English, the particle of negation preceded the
verb, in the English of Shakespeare it follows. (Modern English has introduced
further changes through the introduction of the “helping verb” do.)

(14) Stage I ic ne wāt ‘I don’t know’ : ic ne wāt (nā)wiht ‘I don’t

know (no)thing’
Stage II ic ne wāt ‘I don’t know’ : ic ne wāt (nā)wiht ‘I don’t
know at all’
Stage III ic ne wāt nāwiht ‘I don’t know’
Stage IV I wot(e) not ‘I don’t know’ (Shakespeare)
→ present-day I don’t know (by further changes, including obsolescence
of wot(e))
162 Analogy and change in word structure

While sweeping developments of this type may not be particularly common, the
specific changes exemplified in (14) have parallels in a number of other modern
European languages, including German, the Scandinavian languages, and many
nonstandard varieties of French. Standard French, however, has not yet reached
Stage IV; instead, we find a complex coexistence between Stage I/II forms with
the original negation ne, the discontinuous negation of Stage III, ne … pas, and
the Stage-IV elliptical pas. (The original meaning of Fr. pas is ‘step’, as in OFr. je
ne vais pas ‘I don’t go a step’.)

4 Analogy and phonology: Rule-governed, regular

analogy
All of the analogical processes that we have looked at so far conform to the neo-
grammarian view that analogy is inherently irregular, in contradistinction to
regular sound change. Under certain special conditions, however, clearly ana-
logical developments can unfold with the same regularity (and speed) as sound
change. Some of these developments were known to the neogrammarians, but
they did not realize – or appreciate – the regularity of the changes.
Many examples of regular analogy are quite complex in their analysis and
presentation. One example, however, from the history of English, can be pre-
sented without going into difficult technical details. This is a change that has
taken place in certain varieties of British English, involving original word-final r.
The common wisdom is that British English has lost final r, so that words
like matter are pronounced [mætǝ]. And in certain varieties of British English this
may be true. In others, however, word-final r was linked to the following word, if
that word began in a vowel and followed without any break, as in the matter is [ðǝ
mætǝ riz]. Being linked to the beginning of the next word, however, the r was not
phonetically final and thus did not qualify for the final r-loss.
The result was a word-final variation between r and Ø; see (15). Significantly,
this variation was different from, say, the variation that may trigger leveling. In
cases such as OE freosan : freas : fruron : (ge)froren, the variation was restricted
to particular words with particular meanings (such as the Old English word for
‘freeze’), of a particular phonetic shape (root-final s in the present), in a particu-
lar grammatical category (“irregular” verb). In the case of Brit. Engl. word-final
r, the variation would be found in all words, irrespective of meaning, phonetic
form, or grammatical category. All that needed to be known was that the variation
occurred word-finally, with r found before any vowel-initial following word, and Ø
found elsewhere. Instead of laboriously memorizing for every word that it had this
 Analogy and phonology: Rule-governed, regular analogy 163

variation, speakers would be better off formulating a general rule that captured
the alternation.

(15) the butter [-ǝ] was : the butter [-ǝr] is

the matter [-ǝ] was : the matter [-ǝr] is
… : …

From the historical perspective, the correct rule would have to be an “r-deletion”,
which can be informally expressed as follows:

r-deletion: Word-final [r] → Ø

In this formulation, then, the forms with [r] are considered basic, and the forms
without [r], derived.
However, here again we need to remember that most speakers are not lin-
guists and therefore have no particular motivation to be “historically correct”.
There is nothing to prevent them from taking the forms without [r] as being
more basic, and the ones with [r], as derived. In fact, this might be considered a
much better analysis. Forms with [r] occur only before words with initial vowel.
By contrast, forms without [r] occur much more frequently, in all other con-
texts, including if nothing follows. They can therefore be considered to be more
basic.
Of course, if we go in for this analysis, then the rule formulation above won’t
do. Instead, we have to reformulate the rule as an “r-insertion”:

r-insertion: Insert [r] between word-final vowel and word-initial vowel

This formulation might be felt to be additionally motivated by a general tendency

in languages to break up groups of neighboring vowels by the insertion of some
kind of consonant.
Both rules will perfectly well account for the [r] : Ø alternation in (15) and, in
that sense, their effects could not be distinguished from each other. However, the
formulation of r-insertion makes much broader claims. Not only should words
like butter and matter exhibit an alternation between [r] before vowel and Ø else-
where; all other words in final vowel should do likewise. Otherwise, r-insertion
wouldn’t be much of a rule.
Faced with this situation, speakers have a choice: Either forget about r-inser-
tion and try r-deletion instead, or fix things up so that r-insertion works.
The varieties of British English under discussion (that is, their speakers)
evidently opted for the second alternative. As a consequence, all words in final
164 Analogy and change in word structure

vowel, whether originally r-less or r-ful, insert an [r] if they precede a vowel-initial
word. Compare the examples in (16).

(16) the idea [-ǝ] was : the idea [-ǝr] is

Australia [-ǝ] was : Australia [-ǝr] is
… : …
just like: the butter [-ǝ] was : the butter [-ǝr] is
the matter [-ǝ] was : the matter [-ǝr] is
… : …

Significantly, in these varieties of English, the change brought about by this refor-
mulation of the rule has taken place with the same degree of regularity and
speed as regular sound change. As suggested earlier, the reason seems to be that,
like sound change, the change operates without any regard to specific meaning,
phonetic shape, or grammatical function (except that speakers have to know
where words begin and end). An added factor is that it is a rule that is being
reinterpreted and generalized. In a manner of speaking, the regularity of the rule
begets the regularity of the change.
Whatever the explanation for the regularity of the change, developments like
this suggest that the neogrammarian distinction between regular sound change
and irregular analogy may have been too strict.

5 H
 ypercorrection – an interdialectal form of
analogy
Certain American English dialects present phenomena that are superficially
similar to the British rule of r-insertion. Speakers of originally r-less dialects such
as those of New England (some areas), New York, and the old South, often pro-
nounce an [r] in words like paw, saw, or sofa. This “intrusive r”, however, seems
to owe its origin to a very different process, namely hypercorrection. Unlike
the other analogical processes that we have examined so far, hypercorrection cru-
cially is motivated by the relationship between different dialects or languages – or
rather by the relationship between these as perceived by their speakers.
In many cases, speakers focus on differences in prestige. Speakers of less
prestigious dialects try to imitate a more prestigious one by adaptations in their
pronunciation. This is no doubt the case with the intrusive r of American English.
It is found with speakers who are switching from their old r-less pronunciation to
the r-ful pronunciation that is more prestigious in the United States. If speakers of
 Hypercorrection – an interdialectal form of analogy 165

these dialects want to use the pronunciation which is more standard in American
English, they will quite naturally stick [r]s into words like pore [pɔ̄ ], sore [sɔ̄ ], and
better [beṙǝ]. So far, so good. Evidently, however, speakers do not always know
where to draw the line. In many cases they go overboard and add an [r] in final
position to words like paw, saw, and sofa, or even word-internally in words like
wash [wɔrš] or popcorn popper = [pɔrpkɔrn pɔrpǝr]. Presumably because of its
different origin, the intrusive r of American English differs from British English
r-insertion by being a fairly sporadic phenomenon, with a lot of variation between
different speakers and even for individual speakers.
A similar, and somewhat related, phenomenon is observed in some of the less
prestigious dialects of New York, where words like pearl or earn are pronounced
with [oy], instead of the more standard American [ǝr]. When speakers of these
dialects try to use the more standard pronunciation, they often substitute [ǝr] for
[oy] not only in words like pearl and earl, but even in words that are pronounced
with [oy] in the standard dialect. Example (17) illustrates one of the results of this
development. Moreover, it serves to show that hypercorrection is very similar to
four-part analogy by operating with a proportional schema. It differs from four-
part analogy by operating across dialects.

(17) NY dial. pearl [poyl] : pearl [pǝrl]

earn [oyn] : earn [ǝrn]
oil [oyl] : x = [ǝrl] (just like earl)

Another case of hypercorrection is found in many vernacular varieties of English.

As the result of a variety of sound changes and analogical developments, English
at a certain stage had two competing forms of the so-called gerund, a form in
-ing (as in going) and a form in -en (as in goen). At a later stage, Standard English
leveled out the form in -ing at the expense of -en. Many nonstandard dialects
generalized -en, instead. This difference has since become one of the major fea-
tures distinguishing standard from nonstandard English or formal from informal
usage, and the use of the form in -en is often referred to as “dropping one’s gs”. As
speakers who “drop their gs” try to speak the prestige dialect, they replace their
-en by -ing. And again, in many cases they go too far and extend this substitution
to words like taken (as in I have taking it).
166 Analogy and change in word structure

6 Morphological change
Examples like the ellipsis in (14) show that analogy can have profound effects
on the structure of languages. This particular development had a multiplicity of
effects. First, like a number of other analogical processes (especially blending), it
resulted in lexical change. (This phenomenon is discussed further in Chapter 9.)
It also brought about, as its most far-reaching effect, a syntactic change. (See
the following chapter.)
But along the way it also affected the morphology, by creating a discontinu-
ous marker of negation, ne … nāwiht. This effect on the morphology is in fact the
more usual characteristic of analogical change; and some linguists have referred
to analogy as morphological change. However, as we will see in this section, mor-
phological change – understood as change in morphological systems – results
from a complex interplay between analogical change and sound change, at times
involving even syntax.
Many historical linguists would consider the interplay between sound change
and four-part analogy to be the prototypical vehicle for morphological change. An
example we have looked at before (Chapter 1, § 2, and Chapter 4, § 5.1) involves the
English case system.
To recapitulate: Old English had four different cases, with each case poten-
tially differentiated between singular and plural. Moreover, it had, depending on
one’s count, between six and fourteen different inflectional classes, in which the
same cases were distinguished, but by means of different suffixes. Compare for
instance the two paradigms on the left side of example (18). Modern English has
only two cases: nominative and genitive; the suffix used to distinguish the gen-
itive from the nominative is phonetically identical to the plural marker -s; and
nothing has remained of the different inflectional classes. Compare the right side
of (18).

(18) Old English Modern English

sing. pl. sing. pl.
Nom. stān stānas
Acc. stān stānas stone stones [-z]
Dat. stāne stānum
Gen. stānes stāna(na) stone’s [-z] stones’ [-z]

Nom. caru cara

Acc. care cara care cares [-z]
Dat. care carum
Gen. care cara/carena care’s [-z] cares’ [-z]
 Morphological change 167

In Chapter 1 we noted that the primary factor underlying the change from the Old
English to the Modern English case system is sound change. Both vowels and
nasals were regularly lost in final syllables. For the top paradigm in (18), these
changes would have yielded the following outcomes:

(18’) Expected Modern English

sing. plur.
Nom. stone stones
Acc. stone stones
Dat. stone stone
Gen. stone(’)s stone

The usual view is that at this point four-part analogy stepped in. The identity of
the nominative, accusative, and dative forms of the singular was extended to the
plural, yielding a nominative/accusative/dative plural form stones. At the same
time, the genitive plural was analogically remade to have the same final -s as both
the genitive singular and the new nominative/accusative/dative plural. (The dif-
ferent “s-forms” were differentiated in writing, but not in pronunciation, by a judi-
cious use of apostrophes.) The modern inflection of care can then be explained as
having adopted the pattern of stone : stones etc. by four-part analogy.
One can easily imagine that if this interplay between sound change and four-
part analogy continues unabated, English might eventually lose all remaining
traces of affixal morphology. But other developments may prevent English from
reaching such a stage.
First, sound change is not the only process that can trigger morphological
innovations. As we have seen in § 3, the analogical processes of blending, reinter-
pretation, and ellipsis can introduce new morphological structures which then
can be generalized by four-part analogy. Compare the cases of tele-thon, beaut-
ician, and ne … (nā)wiht. And, as these examples show, the result may increase
morphological complexity, rather than decrease it.
Borrowing, too, can enrich the morphology of a language. In fact, one of the
major factors that prevented Modern English from losing all traces of morphol-
ogy is the rich morphology introduced by massive borrowings from Latin and
Greek, often via French. Consider the case of -able. This affix came into English
through words like debatable and capable and originally was limited to being
used with borrowed root elements, such as debate and cap-. But patterns such
as debate : debatable = do : X (= doable) led to the extension of the affix to native
words, such that, given the proper occasion, nearly every transitive verb can now
be extended by -able to indicate that it is possible to engage in the action indi-
cated by the verb.
168 Analogy and change in word structure

The extent to which borrowing has influenced English morphology can be

gauged from examples such as (19), where only the suffix –ly in (19d) is of English
origin (and as we will see, even this suffix is an innovation in English).

(19) a. anti-dis-establish-ment-ari-an-ism
b. anti-dis-establish-ment-ari-an-ist
c. anti-dis-establish-ment-ari-an-ist-ic-al
d. anti-dis-establish-ment-ari-an-ist-ic-al-ly

New English affixes were even introduced from purely native sources; and these,
too, counteracted the general tendency of English toward becoming a language
with minimal morphology. For instance, in Old English compounds of the type
frēond-līc, literally ‘having the body [= nature] of a friend’, the second element
-līc slowly lost its original force and came to be interpreted as an affix for deriving
adjectives from nouns. Once reinterpreted in this manner, it was extended by four-
part analogy to many other nouns and acquired a certain productivity, reflected
in Modern English adjectives like friendly and heavenly. Along a similar route, the
adverbial Old English form of this reinterpreted adjective suffix, -līc(e), became
even more productive. It is the source of the adverbial ending -ly (as in merrily,
quickly, gently) which, through four-part analogy, can be extended to almost any
adjective of Modern English.
Even syntax, mainly in the form of the syntax of clitics, has contributed to
Modern English morphology. A common traditional view of the genitive ’s (or s’)
of Modern English is the one summarized above. The suffix originated in the gen-
itive singular of certain noun classes and was extended from there to the genitive
plural, as well as to the genitive singular and plural of other noun classes. This
account works well if we limit ourselves to single nouns, but it fails to account for
the placement of genitive ’s in structures like (20), where it follows not the noun
to which it belongs – the head noun, but an “appendage” to that noun: an of-gen-
itive in (20a) and a relative clause in (20b).

(20) a. The present Queen of England’s castle …

b. The man I saw yesterday’s dog …

In earlier English, the s would have had to be attached to the head noun, as one
would expect an inflectional suffix to do. For instance, if we wanted to refer to
an imagined plurality of English queens, we could say The queens of England
have been strong women, but if we were to say The queen of Englands have been …
people would doubt our ability to speak English – or our sanity. But structures
 Morphological change 169

with the genitive suffix ’s in the same position are patently unacceptable in
Modern English; see (20’).

(20’) a. The present Queen’s of England castle …

b. The man’s I saw yesterday dog …

The placement of Modern English ’s could presumably be explained as an ana-

logical extension from the head noun to the total “noun phrase” composed of the
head noun + appendage. But historical evidence shows that the development was
more complex. In early Modern English we find structures of the type (20“) with a
full word, his, following the noun phrase. (The example in (20“b) is a citation from
Shakespeare’s Twelfth Night.) This his no doubt is the genitive of the third person
singular pronoun; except that at this stage it could be used both for the mascu-
line and for the neuter. The existence of such structures suggests that Modern
English ’s results from something like a blending of the original genitive ending
-s with ’s, a clitic form of his with clitic reduction (see Chapter 4, § 5.5). In the
process of blending, the syntactic character of ’s as an element following entire
noun phrases was generalized. And subsequently, the use of ’s was extended from
masculines and neuters to feminines, which earlier had used the appropriate fem-
inine form her or hir; see (20“c) from Lyly’s Euphues. The effective transformation
of a clitic to an affix observed in Engl. ’s, in fact, is a rather common phenomenon.
(See below.)

(20”) a. The man I saw yesterday his dog …

b. … a sea-fight ’gainst the Count his gallies
c. Lucilla hir company

It can easily be imagined that developments of this type, as well as of the type
quick-ly or do-able, could lead to a considerable increase in morphological com-
plexity – if they continue unabated. However, this expectation must be balanced
against the extensive morphological simplification that we have seen before and
that also manifests itself in the fact that English quite commonly derives nouns
from verbs or verbs from nouns without any morphological changes. Consider
such things as crown (noun) → crown (verb), or walk (verb) → walk (noun) → walk
(verb) as in walk a dog.
In sum, then, it must be admitted that the interplay between simplification
and what some have called “complexification” is itself a fairly complex phenom-
enon.
Chapter 6: Syntactic change
Why did the chicken cross the road?
(Popular riddle)

1 Introduction
The preceding two chapters have illustrated two major areas of change that affect
general linguistic structure – sound change, which for obvious reasons affects
the sound structure or phonology, and analogy, which typically affects the mor-
phology. But as we have seen in the preceding chapter, analogical change can
interact with syntax, too.
For instance, ellipsis in earlier English structures of the type Ic ne wāt nāwiht
‘I do not know (anything)’ yielded the Shakespearean type I wot(e) not with the
negation following, rather than preceding, the verb. See Chapter 5, § 3.2. And
section 4 of the same chapter showed that syntactic constituents which become
clitics can wind up as morphological affixes.
In fact, syntax can even interact with phonology. Recall that clitics, besides
being syntactic constituents, have special phonological properties (see Chapter 4,
§ 5.5). One of these phonological properties, namely that they cannot occur by
themselves, has clear syntactic consequences, in that they need a “host” to lean
on and are therefore dependent in their syntactic behavior on the behavior of
their host.
Because of these multiple interactions it is not always easy to determine
where syntax begins and where morphology or phonology ends. Even linguists
are not always in agreement on this matter.
To further complicate matters, much uncertainty exists among non-linguists
who – as noted in Chapter 1 – constitute the vast majority of language users. For
them, syntax is often synonymous with style. For instance, English-speaking chil-
dren frequently say things like (1a); and just as frequently, adults correct them by
responding with something like (1b).

(1) a. Me and Charlie went to the movies.

b. Don’t say “Me and Charlie”, say “Charlie and I”.

In making such comments, adults are not properly distinguishing syntax from
style. According to prescriptive standard grammar, it is syntactically incorrect
to use the form me for the subject of the sentence, even if it is conjoined with

https://doi.org/10.1515/9783110613285-006
 Questionable “syntactic” changes 171

another subject (Charlie, in this case). The correct form is I. At the same time, it
is not considered polite to talk about yourself first; and for this – purely stylistic,
not grammatical – reason the first person pronoun should follow, not precede,
Charlie. Put differently, in the standard form of English, both (1’) and (1’’) are
syntactically correct, by having the right case on the pronoun; but stylistically,
(1“) would be preferred. When adults say things like (1b) they confuse these issues
and, as it turns out, wind up confusing the children as well. But more on that in
§ 5 below. (Interestingly, a recent popular book on the English language by a well-
known linguist labels sentences such as (1’) ungrammatical. It may be true that
structures like these do sound a little odd, but that is perhaps only because they
are used so rarely.)

(1’) I and Charlie went to the movies.

(1”) Charlie and I went to the movies.

2 Questionable “syntactic” changes

Confusion between syntax and style, combined with an insufficient understand-
ing of syntax, lies at the foundation of many of the dire warnings by self-appointed
critics about usages such as Hopefully, it will rain, or the use of data as a singular
noun. (Recall the discussion of these structures in Chapter 1.)
Let us begin with data. To elaborate on a point made in Chapter 1, § 1, like
many other English nouns in -a, this is an original neuter plural of Greek and/or
Latin origin. Specifically, data is the plural of Lat. datum ‘something given (that
can be used as the basis for discussion)’. Even Shakespeare confessed to knowing
only “a little Latin, and less Greek”; and since Shakespeare’s time, the study of
Latin has become even more limited. It is therefore not surprising that the average
English speaker would not know, without being expressly told, that data is a
plural form in Latin. Moreover, data normally come collectively as a set; the need
to talk about an individual datum arises rarely, indeed. Speakers may therefore be
forgiven if they reanalyze data as a singular mass noun, and consequently use it
with a singular verb (as in (2a)), not with the plural verb in (2b) which is required
under a plural interpretation of data.

(2) a. The data presented here shows that Professor Boondoggle’s analysis is
wrong.
b. The data presented here showØ that Professor Boondoggle’s analysis is
wrong.
172 Syntactic change

Critics commonly refer to structures like (2a) as syntactically incorrect. But that
assessment is highly questionable. The syntax of (2a) is impeccable, once data is
reinterpreted as a singular mass noun. If there is a mistake, it does not lie in the
syntax, but in the morphological analysis of data. Moreover, it is a mistake only
in comparison with the traditional analysis of data as plural. But, then, cows once
was a “mistake” for kine which, in turn, was a “mistake” for even earlier kye < OE
cȳ. (See Chapter 4, § 5.11, and Chapter 5, §§ 2.2 and 3.1.) More than that, the same
critics who object to data as singular have no difficulties with a number of other
original plural expressions that are now routinely used as singulars. Consider
such terms as linguistics or politics, which are clearly marked as a plural by their
final -s, or the expression United States, also clearly marked as plural and even in
its original meaning a plurality of States.
As noted in Chapter 1, data is now used so widely that critics generally have
lost interest. This does not mean, however, that the battle is over. Even historical
linguists who are naturally open-minded about linguistic change may person-
ally prefer to treat data as plural. In the meantime, the critics are directing their
attention to new a-plurals that are undergoing a similar reanalysis, such as media
(originally plural of medium, as in information medium) and criteria (originally
plural of criterion). On the other hand, similar reinterpretations seem to be escap-
ing the critics’ attention, a fact which reinforces the impression that the critics are
inconsistent, and their dire warnings, ultimately, quite ineffectual.
Two examples of such reinterpretations that seem to have slipped by the
critics are stigmata, originally plural of stigma, used by educated English speak-
ers as a singular in reference to the crucifixion wounds on Christ’s hands or a
psychologically based medical condition with similar markings; and schemata,
originally plural of schema, now similarly used among certain linguistic theorists
as a technical term construed as singular. Even some words derived from Latin
plurals in simple -a are routinely used as singular, such as agenda, originally the
plural of Lat. agendum ‘(something) to be dealt with’.

3 Syntax, analogy, or both?

The case of hopefully, discussed briefly in Chapter 1, § 1, looks more promising
as a genuine case of syntactic change, at least at first sight. This is because those
who decry the use of hopefully in structures such as (3a) resort to the following
syntactic argument. The word hopefully behaves syntactically similar to words
like fortunately, unfortunately, regrettably, and luckily (see example (4a)) and like
these words expresses a speaker’s attitude to what is being reported in the rest of
 Syntax, analogy, or both? 173

the sentence. These correct (read: traditionally established) attitudinal adverbs

all correspond to structures of the type (4b) which contain the corresponding
adjective. For hopefully, however, parallel structures like (3b) are unacceptable,
as indicated by the postposed asterisk.

(3) a. Hopefully, it will rain tomorrow.

b. It is hopeful that it will rain tomorrow.*
c. We are hopeful that it will rain tomorrow.

(4) a. Fortunately, it will rain tomorrow.

b. It is fortunate that it will rain tomorrow.
c. We are fortunate that it will rain tomorrow.

Under closer examination, this syntactic argument against hopefully quickly

evaporates. First, underlying the argument is the hidden assumption that the
b-versions in (3) and (4) are more basic, and that the a-versions can be justified
only if they can be derived from the b-versions. But this assumption is arbitrary.
It would be just as easy to choose the c-structures of (3) and (4) as the basis for
derivation. And as the examples show, such a derivation would permit (3a).
Secondly, at least one attitudinal adverb established in traditional English
usage, happily, cannot be explained by derivation from a b-version, but only from
a c-version; see (5). This might suggest that syntactically there is nothing wrong
with (3a).

(5) a. Happily, it will rain tomorrow.

b. It is happy that it will rain tomorrow.*
c. We are happy that it will rain tomorrow.

Things get even more complex if we consider additional examples, such as pre-
sumably and actually. See examples (6)–(7). Here one, or the other, or both, of
the putative base structures is either strange (indicated by a question mark) or
unacceptable (characterized by an asterisk). Still, the adverbs are well established
in traditional usage, presumably even among the critics who inveigh against
hopefully. For (6) it is possible to come up with another, alternative “base struc-
ture”, as in d., which “works”, but is quite different from b. and c. For (7), yet a
different base structure could be postulated, see (7d). But the need for such ever-
new, ever-different base structures casts serious doubts on the whole syntactic
approach.
174 Syntactic change

(6) a. Presumably, it will rain tomorrow.

b. It is presumable that it will rain tomorrow.?
c. We are presumable that it will rain tomorrow. *
d. It can be presumed that it will rain tomorrow.

(7) a. Actually, it will rain tomorrow.

b. It is actual that it will rain tomorrow. *
c. We are actual that it will rain tomorrow. *
d. It is an actual fact that it will rain tomorrow.

Nevertheless, the use of hopefully in (3a) clearly is an innovation and must be

motivated by something. If a “hard-core” syntax approach fails to provide a satis-
factory explanation, what does?
An answer is possible, but unfortunately also quite complex: The adverb
hopefully has been around for a long time in such structures as (8). In this context
its use was parallel to that of other attitudinal adverbs, such as happily; see (9a).
Now, happily also occurred in structures like (9b). The coexistence of structures
like (9a) and (9b), and the existence of structures like (8) then could give rise to
the pattern in (3a) above by simple four-part analogy, see (10). But unlike ordinary
four-part analogy, this one operates on syntactic structures, not just words. In this
sense, then, the development that probably gave rise to hopefully is syntactic,
after all.

(8) “Tomorrow it will rain,” Mary said hopefully.

(9) a. “Tomorrow it will rain,” John said happily.

b. Happily, it will rain tomorrow.

(10) “Tomorrow it will rain,” John said happily : Happily, it will rain
tomorrow.
“Tomorrow it will rain,” Mary said hopefully. : X

Attitudinal adverbs like fortunately and sadly presumably arose in a similar

manner. For some of these, the analogy may have been motivated by the pattern
(4b) : (4a). Others no doubt require different patterns. Given the relatively non-sys-
tematic nature of analogical change, it is not surprising that the developments –
and their outcomes – likewise are relatively non-systematic and therefore defy a
straightforward syntactic account.
The use of hopefully in (3a) can thus be explained as the result of a rather
ordinary type of linguistic change. This should put to rest the critics’ claim that it
 Me revisited, or the critics’ revenge 175

is syntactically anomalous. But one suspects that the real reason for the critics’
objections to hopefully is that they consider it newfangled and therefore undesir-
able. It is too early to tell whether the usage will prevail against these feelings.
But many other innovations have caught on, in spite of the critics. And, as noted
in Chapter 1, the usage is now generally accepted in British English. These facts
suggest that, hopefully, the new use of hopefully is here to stay.

4 Me revisited, or the critics’ revenge

For a number of the developments discussed in the preceding sections it is ques-
tionable whether they really constitute syntactic change. Moreover, cases like
data and hopefully involve just one or two words at a time. Generally, however,
syntactic change is highly systematic, affecting whole classes of words and their
syntactic behavior. This is true, at least, for its eventual outcome. However, just
like sound change (see § 6.3 of Chapter 4), syntactic change may exhibit a lot of
irregularity during its propagation. And, again as in the case of sound change,
some developments may propagate rapidly, others very slowly, depending on
their social value.
The issue of social evaluation once again raises the specter of the critics,
or more specifically, of prescriptive grammarians who in the seventeenth and
eighteenth centuries tried to establish the ground rules for a standard variety of
English.
Through the time of Shakespeare, English had been relatively free of pre-
scriptive regulation. One reason for the new prescriptivism must have been the
increasing emphasis of the Renaissance on using and cultivating the vernaculars
instead of Latin (Chapter 2, § 1) and the feeling that for the vernaculars to compete
with Latin, they had to be provided with a prescriptive grammar that could rival
that of Latin. Another factor is that French had been giving way to English as the
language of Parliament, the King’s court, and the courts of justice; but English
was not ideally suited to this new task, since it exhibited high variability, both
geographically and socially, which made communication between different
groups difficult. Under the circumstances, the development of a common, stand-
ard language was highly desirable. (Most European languages underwent similar
standardizing developments at roughly the same time, and for roughly the same
reasons. On this, see Chapter 10.)
The grammarians who busied themselves with laying down rules for a stand-
ard English language were faced with the near-total absence of an English gram-
matical tradition and an overwhelming presence of a Latin grammatical tradition.
176 Syntactic change

As a consequence, they tended to formulate rules that were better founded in

Latin grammar than in the realities of English usage.
For instance, in the area of morphology they set up English nominal para-
digms with six cases, for no other reason than that Latin distinguished six cases;
compare the singular paradigm in (13). But while in Latin the case forms were dis-
tinguished by different endings, English had just two overtly distinguished case
forms: boy and boy’s. The rest of the paradigm had to be eked out by combinations
of preposition + noun. But note that English has many other such combinations,
e. g. with the boy, which were not incorporated into the paradigm because they
lacked Latin counterparts. Paradigms such as the one on the left of (11), thus,
were quite arbitrary and uninformative as far as the structure of English goes.
Nevertheless, paradigms of this type became an integral part of English school
grammar, much to the dismay of generations of school children who, after labori-
ously learning them by heart, could only say O boy! The tradition came to an end
only in the middle of the twentieth century, when linguists finally were able to
persuade school teachers that paradigms of this sort, while appropriate for Latin,
are meaningless for English.

(11) English Latin

Nominative the boy puer
Genitive the boy’s pueri
Dative to the boy puero
Accusative the boy puerum
Ablative from the boy puero
Vocative O boy! puer!

The effect of Latin grammar was not limited to the teaching of morphology. Syntax,
too, was affected. And while Latin influence on morphology mainly resulted in
the minor annoyance of unnecessarily complicating grammatical description, in
syntax the influence had far-reaching effects.
For instance, Latin had a rule according to which doubled negative particles
within the same sentence cancel each other out and, in fact, may create a strong
positive, as in (12). From the Old English period, English had inherited a very dif-
ferent rule, namely that doubled negatives reinforce each other, as in (13).

(12) Nullus non venit

no-one not came
‘Someone (certainly) came.’
 Me revisited, or the critics’ revenge 177

(13) a. þe on land Dena laðra nænig … sceððan ne-meahte (Beowulf)

whom in the land of none of the injure not might
the Danes enemies
‘… whom none of the enemies might injure in the land of the Danes.’
b. I cannot go no further (Shakespeare, As You Like It 2.4.9)

Influenced by the Latin model, the grammarians of the new English standard
inveighed against the traditional use of double negation and promoted the Latin
rule that double negatives cancel each other out.
In this particular case, the grammarians slowly won out. Structures like
(13b) are ungrammatical in Modern Standard English in the meaning intended
by Shakespeare, even though they appear quite often in nonstandard varieties;
in Modern Standard, (13b) is possible only in the meaning ‘I can (certainly) go
further’.
The success of the grammarians may have been aided by two factors. One was
an appeal to logic: If negation is equated with -X, and a positive statement with
+X, then simple mathematics will tell you that -(-X) = + X. Secondly, and perhaps
even more significant, adopting the Latin rule of double negation provided an
easy, and very effective, way of distinguishing standard from vernacular and of
linguistically marginalizing speakers of the vernacular. In Modern English, sen-
tences like I don’t want to give nothing to nobody, nohow, no time are clearly ver-
nacular and “uneducated”. Educated, upper-class speech instead uses structures
like I don’t want to give anything to anybody, under any conditions, ever. (More on
such distinctions between vernacular and standard in Chapter 10.)
In other areas, the Latin-influenced efforts of the prescriptive grammarians
were much less successful. For instance, English had inherited from its earliest
attested stages a rule that permitted relative pronouns and particles to be fronted
into clause-initial position without “pied-piping” their prepositions along with
them. As a result, the prepositions could remain stranded later in the clause. An
Early Modern English example is given in (14).

(14) … that the old carlot once was master of

(Shakespeare, As You Like It 3.5.108)

Latin did not permit structures of this type. So the schoolmaster-grammarians

began to inveigh against this inherited English usage too and declared that
prepositions should not be permitted to end a sentence. But in this case their
success was much more limited than in the case of double negation. Even Winston
Churchill, a man hardly known as a knee-jerk liberal, is said to have poked fun at
the rule. “This,” he said, “is something up with which I will not put.”
178 Syntactic change

Let’s now return to the Me and Charlie went to the movies of example (1) above.
Here, too, the grammarians tried their hand at legislation – and still do. But the
result has been very mixed, and still is. In fact, to some extent the attempts at
prescriptive legislation appear to have backfired.
The issue at hand concerns the case marking of pronouns. Just as in many
other languages, English case marking has survived longer in the pronouns than
in the nouns. Thus, English pronouns distinguish between a nominative case (I)
and an objective case (me); nouns do not. In that sense, pronouns are somewhat
anomalous and therefore vulnerable to developments that might tidy up the sit-
uation.
In the second-person pronoun, you, these developments have been carried
to their logical conclusion, and the distinction between nominative and objective
has been lost. In the other pronouns, the developments have been less radical.
To understand these developments it is necessary to look at the system of
case marking that early Modern English inherited. Simplifying things a little, this
system can be characterized by the following rules:
(a) Subject pronouns are in the nominative case;
(b) Pronouns that are the predicates of equational sentences of the type X is Y
containing the verb ‘be’ also are marked nominative;
(c) Pronouns that are the objects of verbs (other than the verb ‘be’) or of preposi-
tions are in the objective case.
Traces of this earlier system are clearly present in the early Modern English of
Shakespeare, as in (15). But side by side with it we find signs of an innovative
system with different case marking conventions; compare (15’).

(15) a. Why, so I do … (Twelfth Night 1.1.18)

b. A sister! You are she (Twelfth Night 5.1.326)
c. I saw him hold acquaintance with the waves (Twelfth Night 1.2.16)
Say I do speak with her, my lord, what then? (Twelfth Night 1.4.23)

(15’) a. Mistress, dispatch you with your safest haste

And get you from our court.
Me, uncle? (As You Like It 1.3.41–42)
b. That’s me, I warrant you (Twelfth Night 2.5.79)
c. You know my father hath no child but I (As You Like It 1.2.18)
Let fortune go to hell for it, not I (Merchant of Venice 3.2.21)
… all debts are cleared between you and I (Merchant of Venice 3.2.321)

In examples like (15’c) the nominative case of the pronoun does not occur directly
after the verb or preposition. This suggests that rule (c) above is getting relaxed,
 Me revisited, or the critics’ revenge 179

requiring objective marking only on pronouns that are directly preceded by the
verb or preposition. Let us refer to these pronouns as adjacent to the verb or
preposition.
Examples like (15’a) and (15’b) are more difficult to interpret. One possibility
is that, as in French, the objective case of the pronoun is beginning to be used as
an emphatic form. This interpretation would account for both (15’a) and (15’b).
But the type (15’b) is also amenable to another analysis. The objective case here
is the result of an extension of rule (c) so that it applies after all verbs, including
the verb ‘be’ (at least if the pronoun is adjacent to the verb). If carried to its logical
conclusion, this development would eliminate the need for rule (b) and, in that
sense, simplify the grammar of English.
The development of case marking in vernacular or untutored Modern English,
uninfluenced by the rules of the prescriptivists, suggests that two of these accounts
are especially appropriate: the notion that adjacency plays a role in case marking
and the explanation of (15’b) as reflecting an extension of rule (c) and incipient
loss of rule (b). Interestingly, however, the results are somewhat different. Con-
sider the examples in (15’’).

(15’’)a. I went to the movies (subject pronoun adjacent to the verb)

Me and Charlie went to the (subject pronoun not adjacent to the verb)
movies
b. It’s me
c. They saw me
They saw me and Charlie
They gave it to me
They gave it to me and Charlie

Whereas in early Modern English, adjacency affected the case marking of object
pronouns, in untutored Modern English it applies to subject pronouns. Contrast
the two examples under (15’’a). And while in early Modern English the “default”
case marking of pronouns not adjacent to the verb, was the nominative, as
in (15’c), in Modern English it seems to be the objective case, as in the second
example of (15’’a).
Now, the rules for Latin case marking are very similar to rules (a)–(c) above.
Given what we have seen so far, it is not surprising that the grammarians insisted
that Standard English should follow these rules, and not the new rule systems
underlying (15’) and (15’’).
In the case of structures like (15’’b), the critics’ success has been quite mixed.
Many speakers of Standard English feel that It’s I is overly formal, even stilted,
and prefer to say It’s me, at least in informal, more friendly or intimate, contexts.
180 Syntactic change

As for structures like (15’’a), we have seen in the beginning of this chapter
that present-day critics tend to correct expressions like Me and Charlie went to
the movies, by insisting on Charlie and I …, confusing syntax and style. Moreover,
in doing so, they provide no reliable grammatical guidelines for correctness. As
a consequence, there is nothing to prevent the poor target of such corrections to
interpret them as generally prohibiting sequences like me and Charlie and requir-
ing instead the general use of structures like Charlie and I.
The result is that speakers come up with hypercorrect sentences such as
They saw Charlie and I and They gave it to Charlie and I. These structures seem
to comply with the demand to say Charlie and I, not Me and Charlie, yet they
obviously violate rule (c) above which requires object pronouns to be in the objec-
tive case in Standard English. Critics shudder at such sentences, and so do many
other, more liberal speakers of Standard English. But most of the critics don’t
realize their own role in bringing about such structures – by not providing proper
guidelines as to when one should say Charlie and I and when Charlie and me.
To do so requires making a proper distinction between syntax and usage. The
usage issue is very simple: It is considered impolite to talk about yourself first,
so you should mention Charlie first and then yourself. The syntactic issue is a bit
more complex, but can be explained in fairly simple terms, too: When you say Me
and Charlie went to the movies you are basically saying that you went and Charlie
went. But, except for Cookie Monster on Sesame Street, Tarzan, and speakers of
pidgins (see Chapter 14), no English speaker would say Me went to the movies,
everybody would say I went to the movies. That’s why you should say Charlie and
I went to the movies. On the other hand, nobody would say He saw I or They gave
it to I, people use me instead. So, why don’t you also say He saw Charlie and me
and They gave it to Charlie and me?
While structures such as He saw Charlie and I are superficially similar to
Shakespeare’s Let fortune go to hell for it, not I and all debts are cleared between
you and I in (15’c), it is more likely that they are hypercorrections than direct
descendants of the Shakespearean constructions for children are constantly sub-
jected to admonishments of the type Don’t say “me and Charlie”; say “Charlie and
I”, without any guidelines as to when Charlie and I is appropriate. Nevertheless,
just like morphological hypercorrections, these hypercorrect uses of the nomina-
tive pronoun have the potential of becoming accepted as normal. There are some
indications that this has happened in some varieties of English.
 A successful major shift: Word order in English and related languages 181

5 A
 successful major shift: Word order in English
and related languages
Attentive readers may by now have realized that syntactic change differs mark-
edly from most forms of sound change and analogical/morphological change.
It does not just affect individual words or classes of words, not even individual
sentences, but the patterning of a large number of sentences. For instance, the
developments in pronoun case marking were not limited to the sentences cited
above, but instead affected all sentences containing subject and object pro-
nouns. In order to trace syntactic change it is therefore necessary to examine
the fate of abstract patterns for sentences, patterns whose structural make-up
may vary considerably. Moreover, by their very nature, such sentence pat-
terns are quite complex; and to discuss how they are put together requires
using fairly extensive and specialized terminology. This is especially true if we
examine more complex syntactic changes than the ones we looked at in earlier
sections.
For these reasons let us look at only one example of a complex syntactic shift.
This is a sequence of changes which significantly altered major word order from
early Germanic to Modern English. The example has been chosen because, among
the various more complex syntactic changes that can be observed, word order
changes are most easily illustrated.
Assume you wanted to say in Modern English that a chicken crossed the road.
And assume you are interested only in stating the facts – no questions asked, no
commands, and no passive. You wouldn’t have much of a choice, would you? The
most natural way of stating the message would be as in (16a), with the subject (in
small caps) preceding the verb (in boldface) which, in turn, precedes the object
(in italics). For some speakers (16b) would be acceptable, too, but clearly more
“marked”, with particular emphasis on the road. Many other speakers would
prefer to express such an emphasis by saying something like It’s the road that the
chicken crossed, or they would use a passive The road was crossed by the chicken.
(See § 6 below.) Other permutations of (16a) would be entirely unacceptable, such
as (16c)–(16 f).

(16) a. the chicken crossed the road [Basic, “unmarked” order]

b. the road the chicken crossed [“Marked” order; the road is “in
relief”
≠ The road the chicken crossed
(was wide)]
c. the chicken the road crossed *
182 Syntactic change

d. the road crossed the chicken *

e. crossed the road the chicken *
f. crossed the chicken the road *

In this respect, Modern English differs markedly from the majority of the early
Indo-European languages, as well as from Old English, especially the very archaic
stage of Old English found in the famous epic Beowulf. In these languages, any
of the six different orders in (16) would be acceptable; see (17). (To save space,
the Old English nouns are given without preceding demonstrative pronouns, the
forerunners of the modern definite article.)

(17) Old English (Beow.) Sanskrit

a. henn oferēode rāde kukkuṭaḥ atarat mārgam
b. rāde henn oferēode mārgam kukkuṭaḥ atarat
c. henn rāde oferēode kukkuṭaḥ mārgam atarat
d. rāde oferēode henn mārgam atarat kukkuṭaḥ
e. oferēode rāde henn atarat mārgam kukkuṭaḥ
f. oferēode henn rāde atarat kukkuṭaḥ mārgam

Latin
a. pullus transiit viam [Marked]
b. viam pullus transiit [Marked]
c. pullus viam transiit [Basic]
d. viam transiit pullus [Marked]
e. transiit viam pullus [Marked]
f. transiit pullus viam [Marked]

Moreover, the basic, “unmarked” order is not (17a), but (17c), with the subject (S)
before the object (O), which in turn precedes the verb (V). Other orders convey
special connotations. For instance, in (17b), (17e), and (17 f) the first word is in
relief. It may simply be emphasized, it may be treated as the topic of the rest of
sentence, or it may be in some other way under special focus. Placing constituents
in a position on the extreme left of clauses to convey such connotations is such a
widespread phenomenon in languages other than English that the position has
received a special name, namely topic. Patterns of the type (17a) and (17d), with a
constituent following the verb are rarer than the others. They, too, convey special
connotations. For instance, (17d) might be used to place special emphasis or focus
both on the road and on the chicken.
Ignoring Old English, Sanskrit, and Latin patterns of the type (17a) and (17d)
in which a constituent follows the verb, we can diagram the differences in sen-
 A successful major shift: Word order in English and related languages 183

tence patterns between Modern English and the earlier Indo-European languages
as follows.

Basic structure Structure with TOPIC

Modern English SVO (TOPIC SV …) [in marginal use]
Early Indo-European SOV TOPIC S … V [productive pattern]

Fig. 1: Word order differences between Modern English and early Indo-European

How, then, did English change from the early basic SOV pattern in (17c) to its
modern SVO pattern in (16a)?
To understand this development, it is necessary to consider sentences with
complex verbs, such as Engl. had crossed, consisting of an auxiliary (had) and a
main verb that carries the main lexical meaning (crossed). In Modern English, the
basic order of these two is as given in (18a), with the auxiliary preceding the main
verb. This contrasts with the basic order of Sanskrit and Latin, in which the main
verb precedes the auxiliary; see (18b) and (18c). (The Sanskrit and Latin struc-
tures corresponding to (18a) are construed as passives; but this does not affect
the argument.)

(18) a. the chicken had crossed the road

b. kukkuṭena mārgaḥ tīrṇaḥ āsīt
c. (a) pullo via transita erat

Given the parallelism between Beowulfian Old English and Sanskrit/Latin in (17),
we might expect a similar parallelism as regards the order of auxiliary and main
verb. In fact, we do find structures of the type (19a), with the main verb followed
by the auxiliary, both placed at the end of their clause. But in structures with
complex verbs, Beowulf tends to prefer a different order, given in (19b), with the
auxiliary in second position, but the main verb stranded at the end of the clause.
(Note that the verb ‘go’, contained in the Old English word for ‘cross’, is an irregu-
lar verb which makes its past tense and past participle from different roots.)

(19) a. henn rāde ofergangen hæfde

b. henn hæfde rāde ofergangen

This placement of the auxiliary in second position clearly is an innovation. The

innovation may have been motivated by the fact that the early Germanic auxil-
iaries had properties characteristic of clitics (see Chapter 4, § 5.5). For instance,
one of the two auxiliaries of early Germanic, the verb ‘be’, underwent clitic reduc-
184 Syntactic change

tion similar to what we find in the Modern English auxiliaries. Just as in Modern
English, has and is often occur in the reduced form ’s (compare He’s come; She’s
here), so PGmc. *ist had been reduced to is in Old English.
There is a general tendency in the languages of the world for clitic auxiliaries
to go to the second position of the clause. Perhaps this is because the first position
of the clause tends to be the topic, which can be expected to be accented and thus
to serve as the host for the clitic auxiliary, somewhat like a magnet. Whatever the
explanation, we must accept that at the time of Beowulf, auxiliaries had begun
to move into second position. We can diagram this development as in Figure 2.

Fig. 2: First stage of word order shift

Syntactic changes, however, do not happen all of a sudden. We don’t wake up one
fine morning and discover, much to our surprise, that our syntax has changed!
Rather, change in syntax has many of the properties of sound change as observed
by Labov (Chapter 4, § 6.3). There is a lot of variation between the old and the new
pattern, but slowly the innovative pattern gains ground, and eventually the old
pattern may disappear entirely.
In the case of our syntactic change, patterns of the type (19a) persisted
throughout the history of Old English. They did so especially in dependent
clauses. This may be because dependent clauses did not make as much use of the
clause-initial topic position that served as the host for clitic auxiliaries. Whatever
the explanation, dependent clauses lagged behind in the development through-
out the history of Old English, not just as regards the movement of auxiliaries to
second position, but in the later changes as well.
In a complex verb like OE ofergangen hæfde or hæfde … ofergangen, only one
element usually is inflected for person, number, and the like, namely the auxil-
iary. Thus if we wanted to say that several chickens had crossed the road, the verb
would have to take the form ofergangen hæfdon or hæfdon … ofergangen. A verb
which thus inflects for person, number, etc. is referred to as a finite verb.
Now, the original motivation for the auxiliary to move into second position
was that it was a clitic. However, as we have just seen, auxiliaries also are finite
verbs. This made it possible to reinterpret the movement to second position as
conditioned, not by the cliticness of the auxiliaries, but by their finite status.
Compare Figure 3:
 A successful major shift: Word order in English and related languages 185

Fig. 3: The second stage in the development – Reinterpretation

Such a reinterpretation would be very similar to the British English reinterpreta-

tion of r-deletion as r-insertion discussed in Chapter 5, § 5. And just as in the latter
case, both the old and the new interpretation would perfectly well account for the
situation for which they were originally intended. But recall that the reinterpreted
rule of r-insertion made much broader predictions than the old r-deletion, for it
could apply not only to originally r-ful words like matter, but also to originally
r-less ones, such as idea. And once it was applied to the latter, the rule radically
changed British English phonology.
The situation is similar for the reformulation in Figure 3. Whereas in struc-
tures of the type (19a/b), the auxiliary is the finite verb and the main verb is non-fi-
nite, in the Old English structures of the type (17), the main verb, oferēode, is
finite. (For instance, if several chickens had crossed the road, we would have had
to say oferēodon.)
If the formulation in Figure 3 is to remain valid, finite main verbs like oferēode
likewise must move into second position. And this is what happened in post-Beo­
wulf Old English, which would have increasingly expressed the idea ‘The chicken
crossed the road’ as (20b), not (20a).

(20) a. henn rāde oferēode

b. henn oferēode rāde

Toward the end of the Old English period a further development set in. The
“stranded” main verb of structures like (19b) = (21a) began to line up immediately
after the second-position auxiliary; see (21b). The reason for this development
seems to be that auxiliary and main verb functionally belong together, as compo-
nent parts of a morphologically complex, but functionally simple verb.

(21) a. henn hæfde rāde ofergangen

b. henn hæfde ofergangen rāde
c. the chicken had crossed the road

The resulting structure (21b) looks remarkably similar to its Modern English
counterpart, given in (21c). One might be tempted to believe that the late Old
English stage represented in (21b) is completely identical to Modern English and
186 Syntactic change

that, therefore, the change from SOV to SVO had been completed by this time.
Things are a little more complex, however. In late Old English, both (21a) and
(21b) were still grammatical, whereas the counterpart of (21a), The chicken had
the road crossed is not acceptable in Modern English, at least not in the meaning
‘the chicken had crossed the road’. More than that, not only did (21a) and (21b)
both continue to be grammatical in late Old English, (20a) continued to coexist
with (20b), and (19a) with (19b). That is, the change toward SVO had by no means
been concluded.
In addition, Modern English differs from Old English not just in having SVO,
but also by increasingly disfavoring structures with initial topic. (Recall that
structures such as It’s the road the chicken crossed provide a handy alternative.)
Moreover, to the extent that it still tolerates structures with topic, Modern English
usually does not place the verb directly after the topic. For instance, it is impos-
sible in Modern English to say The road crossed the chicken in the sense of ‘The
chicken crossed the road’, with emphasis or some other kind of prominence on
the road. If we can place the road in initial position at all, we have to place the
subject immediately after it, and the verb has to follow the subject, as in The road
the chicken crossed. That is, Modern English generally requires that the subject
precede its verb in simple declarative statements. Only traces of the older pattern,
with verb before subject, remain, such as Out of the cave came a tiger and espe-
cially There is a chicken on the road.
Finally, recall that in Old English, dependent clauses tended to lag behind
in the development from SOV toward SVO. Whereas main clauses increasingly
favored the (b) versions of examples (19), (20), and (21), dependent clauses tended
to favor the older (a) patterns. In this respect, too, syntax has changed on the way
toward Modern English, for now there is no longer any major difference in word
order between main clauses and dependent clauses; both types of clauses have
SVO.
The developments outlined so far took place not only in English, but with
certain variations, in the majority of the other European languages. German,
Dutch, and Frisian, however, participated only in the first two stages of the devel-
opment. As a consequence, they place all main-clause finite verbs into the second
position (whether they are auxiliaries or main verbs), but leave non-finite main
verbs stranded in final position. Compare the German examples in (22a) and
(22b). Moreover, the fact that dependent clauses lagged behind was reinterpreted
in these languages as being syntactically significant. Dependent clauses gener-
alized the older, verb-final patterns at the expense of the innovated second-po-
sition structures and, as a consequence, came to systematically differ from main
clauses. This accounts for the ordering of elements in (22c) and (22d). (To simplify
matters, the remainder of this discussion concentrates on German.)
 A successful major shift: Word order in English and related languages 187

(22) a. Das Huhn überquerte die Strasse ‘the chicken crossed the road’
b. Das Huhn hatte die Strasse ­überquert ‘the chicken had crossed the
road’
c. dass Das Huhn die Strasse überquerte ‘that the chicken crossed the
road’
d. dass Das Huhn die Strasse überquert ‘that the chicken had crossed
hatte the road’

These divergent developments are responsible for the differences between English
and German word order noted in § 2 of Chapter 1.
German differs from English in another important respect. In contrast to
English, structures with initial topic are still productive, at least in main clauses.
And in main-clause structures, the finite verb still directly follows the topic, while
the subject is placed after the finite verb; see (23). This retention of topic struc-
tures, however, is not limited to German. Many other European languages like-
wise have retained topic structures, although the degree to which such structures
are used may differ considerably.

(23) a. Die Strasse überquerte das Huhn

b. Die Strasse hatte Das Huhn überquert

Even in German, structures like (22) are subject to certain restrictions. For instance,
the fact that expressions like die Strasse and das Huhn do not distinguish nomi-
native from accusative case may make structures like (23) ambiguous. Instead of
interpreting them as meaning that ‘the chicken crossed the road’ or ‘the chicken
had crossed the road’, it would be possible to understand them to mean ‘the road
crossed the chicken’ or ‘the road had crossed the chicken’. In the present case,
such an interpretation is rather unlikely, simply because roads don’t normally
cross chickens. But in sentences like Die Mutter liebt das Kind, there could be
genuine confusion as to whether we should translate this as The child loves the
mother or The mother loves the child. In sentences of this type, therefore, the
topic construction is generally avoided – if there is no context that might help to
disambiguate. (This is especially true in written texts, which lack most of the into-
national clues of spoken language and where readers cannot ask for clarification.)
If enough disambiguating context is present, however, sentences of this type can
be – and are – used.
188 Syntactic change

6 Conclusion
Although many of the changes discussed in this chapter, if they are syntactic at
all, have relatively minor consequences, the extended example in § 5 shows that
syntactic change can be at least as sweeping and general as Grimm’s Law or the
Great English Vowel Shift, and that its effects on the structure of the language can
be at least as great as the morphological changes which ideally can turn a lan-
guage from having minimal morphology to one with rich morphology, and back
again to minimal morphology.
In fact, in English there seems to be a certain connection between the mor-
phological development toward minimal morphology and the loss of free word
order. While in German, ambiguities of the type Die Mutter liebt das Kind can be
considered somewhat minor annoyances, the total loss of nominative/accusative
case distinctions in English nouns would have led to systematic ambiguities in all
structures of this type – unless English eliminated free order and adopted the con-
vention that the subject must necessarily precede the verb directly in unmarked
declarative sentences. This is probably a major reason for why Modern English
can topicalize only by moving the direct object in front of the subject, but leaving
the order of subject and verb untouched, as in The child, the mother loves.
As noted earlier, even this relic of topicalization is beginning to fade, and
many speakers of English, especially in the American Midwest, are uncomforta-
ble with such structures. But since there are great communicative advantages to
placing a topic element in sentence-initial position, they use alternative devices
for accomplishing this task. One of these is the passive construction, as in The
child is loved by the mother. Another one employs periphrases, such as As for the
child, the mother loves her, or It is the child the mother loves.
In the development of these alternatives, the same principle seems to be at
work as in the case of morphological change. Loss or attrition in one component of
the grammar tends to be compensated for in another component. In this manner,
the communicative capability of language is maintained, and something like a
“steady-state dynamic equilibrium” prevails.
Chapter 7: Semantic change
“When I use a word,” Humpty Dumpty said, in a rather scornful tone,
“it means just what I choose it to mean – neither more nor less.”
(Lewis Carroll, Through the Looking-Glass.)

1 Introduction
The preceding three chapters have been devoted to changes in linguistic struc-
ture – the topic which probably interests linguists most. The majority of speak-
ers, however, are not linguists, and linguistic structure is something they hardly
ever think about. There is a good reason for this. In order to use linguistic struc-
ture effectively we have to place its knowledge safely below the level of conscious-
ness. We can see this when we first learn a new language and are still consciously
trying to get the grammar and pronunciation right. Uttering even a single, simple
sentence can be agony.
But our difficulties extend beyond grammar and pronunciation. We also need
to make sure that the utterances we create have meaning and, moreover, that
they mean what we want them to mean. For instance, a German learning English
will have to know that public viewing – a made-up, pseudo-English expression in
German – does not mean ‘a public watching of an event on a large screen’ in real
English, but rather ‘the display of a deceased person before the funeral’. That
is, we need to pay attention to semantics, the meanings associated with mor-
phemes, words, and collocations of words.
For non-linguists such semantic oddities are highly fascinating. Linguists,
by contrast, find lexical semantics extremely elusive and therefore difficult to
deal with, because meaning is inherently fuzzy and non-systematic. They greatly
prefer to deal with the much more “orderly” structure of language.
At the same time, the very fuzziness and lack of systematicity of semantics
is an essential component of language, if we consider that through language we

https://doi.org/10.1515/9783110613285-007
190 Semantic change

want to convey a virtual infinity of different meanings and that we have to do so

with a finite number of sounds and a larger, but still finite number of words.
To some extent we make up for this limitation by conveying meaning not
through individual speech sounds, but through combinations of sounds. Thus,
the sounds [s], [a], [w], [n], and [d] by themselves have no meaning, but put these
meaningless sounds together in that order and you have the word [sawnd] = sound.
In this way we are able to use a limited number of speech sounds (roughly between
25 and 110) to “spell out” a large number of different words (somewhere between at
least 1000 and the amount of words contained in the Oxford English Dictionary).
An important result of this indirect encoding of meaning through combinations of
meaningless speech sounds is that there is no direct relationship between meaning
and the sounds which spell out the words that convey meaning. In other words, the
relationship between meaning and form is arbitrary. (See § 4 below.)
Words by themselves, however, are not sufficient to convey meaning. Putting
words together helter-skelter, as in dog bite kid or kid dog bite, is not sufficient;
we need to put the words together according to syntactic rules, together with the
proper morphological machinery (where appropriate), as in The dog was biting
the kid vs. A kid bit a dog, and so on. Doing so permits us to greatly expand the
messages and meanings that we want to convey.
Ultimately, however, the number of new, or even old, ideas that we may want
to convey through language is virtually without limit, exceeding the capacities
of phonology and linguistic structure. Take for instance our favorite linguistic
term sound change. Phonology helps us to “spell out” the words, morphology
permits us to put the words together into a compound, and syntax enables us to
use the term in meaningful sentences. But none of this accounts for the seman-
tics of sound change. As we have seen, the term does not refer to just any change
in sounds, but is limited to changes conditioned by phonetic or phonological
factors. Moreover, sound here refers not just to any sound, but to speech sound;
and change refers not just to any change, but to linguistic change. Put differently,
we are playing fast and loose with the semantics in order to create a handy term
that conveys a meaning which otherwise would have to be expressed as some-
thing like linguistic change in the phonological structure and only conditioned by
phonetic or phonological factors.
What makes it possible for us to do this is the fuzziness of lexical semantics.
There isn’t just one meaning for any given word but a range of meanings, such that
for instance sound can refer to any sound or noise, including speech sounds. And
context determines which of the different meanings is appropriate in a particular
context. Without this semantic fuzziness we would be confronted by the impos-
sible task of having to learn or create a virtual infinity of lexical items with the
very limited means that phonology provides us for distinguishing between them.
 The inherent fuzziness of meaning – polysemy, semantic overlap, metaphor 191

2 The inherent fuzziness of meaning – polysemy,

semantic overlap, metaphor
As the preceding discussion has shown, natural languages have to permit a great
amount of semantic fuzziness and the flexibility that this fuzziness makes avail-
able. Consider for example the expression in (1) which, depending on the context
or reference, may have a variety of different interpretations. If John is in kindergar-
ten, his just barely being able to sound out written words may merit the response
in (1). But if John is in his teens or older, with normal abilities, this would not be
a probable interpretation. He would have to be able to do something more spec-
tacular, such as excel in a speed reading class, in a poetry reading contest, or the
like. The expression in (1), however, is able to convey all of these interpretations,
by being sufficiently vague or “fuzzy”. This vagueness or fuzziness of meaning is
referred to as polysemy.

(1) John reads very well

In many cases we are not even aware of such polysemy. But if we think about it,
we tend to say that there is a true or core meaning and that other meanings are
transferred or extended. For instance, the core meaning of read might be some-
thing like ‘comprehend the meaning of written symbols’. Meanings such as ‘com-
prehend the meaning conveyed by written symbols’ might be a first extension of
the core meaning. A further extension might be ‘sound out written symbols’. Yet
another extension would be ‘sound out a poem for an audience’. This is very much
what we find in dictionaries, as in the following excerpt from the entry for read in
the American Heritage Dictionary of the English Language (first edition). But note
that even the most comprehensive dictionary can only list a small subset of the
total range of meanings that can be related to each other in this way.

read … 1. To comprehend or take in the meaning of (something written or printed). 2. To

utter or render aloud (something written or printed). 3. To have the knowledge of (a lan-
guage) necessary to understand printed or written material …

This relationship between the different meanings in a word like read can be
modeled as in Figure 1, with a core meaning surrounded by a set of extended
meanings arranged in concentric circles. However, this diagram greatly simplifies
matters. For instance, a derivation of the type read1 ‘comprehend the meaning of
written symbols’ → read2 ‘comprehend the meaning conveyed by written symbols’
→ read3 ‘sound out written symbols’ is highly unlikely. Rather, read2 and read3
are both extensions of read1, but they go in different directions. And read4 ‘sound
192 Semantic change

out a poem for an audience’ would be an extension of read3. A more realistic, but
also more unwieldy, model therefore would look more like Figure 2, with a core
meaning surrounded by concentric amoeba-like extensions.

Fig. 1: Core and extended meanings as concentric circles

Fig. 2: More realistic model of core and extended meanings

Examples like read readily show that the range of meanings of a given word can
vary considerably. Depending on the circumstances, that range may be broader
or narrower.
Moreover, because the meanings of words ordinarily are polysemous and
extend over a larger area of references, the meanings of different words may
overlap, as in Figure 3. Thus, for at least one of the interpretations of (1), we can
 The inherent fuzziness of meaning – polysemy, semantic overlap, metaphor 193

also say something like (2). (Although more realistic, models of the type illus-
trated in Figure 2 are rather unwieldy, as noted above. This is especially true for
relationships between the meanings of different words. Figure 3 therefore uses the
simpler, concentric-circle model.)

(2) John recites very well

Fig. 3: Semantic overlap

The fuzziness in meaning resulting from semantic extension and semantic overlap
can be much more extensive than the example of read and recite suggests. Con-
sider the following, much more complex cases.
What is the meaning of animal, either taken by itself or contrasted with plant
or other words? Are bacteria animals or plants? What about corals? Insects? Birds?
Human beings? Speakers may disagree greatly with each other. But even more
significant, they may disagree with themselves. For instance, the following sen-
tences might be uttered by the same person, at different times, without any feeling
of contradiction. In some of these, animal is used in a fairly restricted sense as a
near-synonym of mammal, in others it has a more general, scientific, meaning,
conforming to dictionary definitions such as ‘Any organism of the kingdom Ani-
malia, distinguished from plants by … locomotion, fixed structure … and non-pho-
tosynthetic metabolism’ (American Heritage Dictionary of the English Language,
first edition). But other factors play a role, too, for the way many people use the
word, such as land-based vs. air- or water-based, human vs. non-human. And
expressions like (7) and (8) illustrate a similar vacillation for fish and its relation
to animal.
194 Semantic change

(3) This powder kills noxious insects, but is harmless to humans and animals.
(4) It’s incorrect to call bacteria “bugs”, because bacteria are plant-like, but
bugs or insects are animals.
(5) Birds and human beings are two-legged animals.
(6) Noah gathered pairs of all the birds and animals on his ark.
(7) Whales aren’t really fish, they’re animals.
(8) Jonah was swallowed by a great fish, probably a whale.

What about the word star? Again, speakers can accept a wide range of different,
often contradictory meanings conveyed by this word, as in (9)–(12). And here,
again, one of the contributing elements is a difference between scientific and ordi-
nary use.

(9) Venus is the brightest star in the early night sky.

(10) Venus is a planet, not a star.
(11) Stars shine at night.
(12) The sun is a star.

It is tempting to model the relationship between these two different interpre-

tations of star as an ellipse, with two core meanings as its foci, see Figure 4.
However, even those who insist on considering the sun a (fixed) star would find
it very strange if in broad sunlight somebody said to them Look at that beautiful
starlight. (See the cartoon at the beginning of this chapter.)

Fig. 4: A semantic ellipse with two foci

Moreover, an elliptic model would do nothing for the apparently multifocal

semantic range of animal. A much more appropriate model probably would be
one which considers the broad semantic range of the scientific or dictionary defi-
nition to be the true or basic meaning of animal. Within that range, the meaning
‘mammal’ would hold a special status, similar but not identical to a core meaning,
 The inherent fuzziness of meaning – polysemy, semantic overlap, metaphor 195

since we tend to think of mammals as the most prototypical animals. For

obvious reasons, the notion ‘human being’ would at least partly overlap that of
‘mammal’. The meanings ‘fish’ and ‘bird’ are clearly less close to the prototypical
‘mammal’, ‘reptile’ even less; and ‘insect’ would be quite peripheral. An attempt
to model these relationships is found in Figure 5.
This model of semantic relationship differs considerably from the models that
we looked at earlier. The models in Figures 1–3 operate with the assumption of an
increasing semantic broadening or expansion of a core meaning. In the case of
animal, by contrast, a broad range of meanings undergoes a variety of semantic
narrowings; and among these narrower meanings one can be considered proto-
typical. Such tendencies toward semantic narrowing plus development of a proto-
typical quasi-core meaning are quite typical of inherently broad cover terms such
as animal, France, Asia, or Orient. (See § 5.6 below for further illustration, and
recall the discussion in Chapter 1 about the American Oriental Society.)

Fig. 5: A semantic range with prototypical center plus other more peripheral meanings

The major vehicle for expanding the range of meanings of a given word is meta-
phor. The term metaphor is most commonly used in reference to the often daring
or arcane expansion of meanings in poetic language, such as, say, arrows for the
sun’s rays or mother of the waters for the ocean. But it is a much more widespread
196 Semantic change

phenomenon, which we constantly draw on to creatively expand the power of

words.
The effects may be relatively subtle, as in the ever-increasing semantic range
of read. But they may approach the boldness of poetic metaphor. For instance,
the word clear originally referred to visible objects such as liquids or air, but then,
by metaphorical extension, came to be used also about invisible objects, such as
statements and arguments. It is true, metaphorical extensions such as that of clear
may not strike us as particularly daring; that is because the metaphoric link, like
linguistic structure, has been pushed below the level of consciousness. But it is
not difficult to bring it back to consciousness. This for instance happens if some-
body answers the question Is everything clear? with something like Clear as mud.
Metaphor comes in many different guises, some based on perceived similar-
ity, others on contiguity or physical association. Examples of the former are the
just-mentioned arrows and clear. Consider further the metaphorical extension of
animal in expressions like John’s a real animal. Examples of extension by conti-
guity are the different meanings of read, or the use of hands to refer to laborers.
(Further illustrations are found in later sections.)

3 Synonymy and homonymy

The need to use a finite number of words to express a virtual infinity of meanings
has further consequences. One of these is that absolute synonymy, i. e., com-
plete identity in meaning between phonetically different expressions, tends to be
avoided and is consequently rare. Having more than one linguistic form to express
exactly the same range of meanings would simply not be economical. True, the-
sauruses are full of supposed synonyms, intended to help us avoid excessive rep-
etition. But uncritical use of a thesaurus can have strange, even ridiculous, conse-
quences because the listed alternatives are only partial, not absolute, synonyms,
exhibiting semantic overlap, not identity.
Thus, although the meanings of subconscious and unconscious overlap to a
considerable degree, even a cursory look at examples like (13) shows that they
cannot be used completely interchangeably. Their synonymy, thus, is only a
partial one.

(13) She knocked him _______

There is yet another consequence. In order to make it possible for words to clearly
and unambiguously signal different meanings, there is a tendency to avoid homon-
 Synonymy and homonymy 197

ymy, the use of phonetically identical words with divergent references. This is true
at least when the two words can occur in the same context and can create undesira-
ble confusion. Thus, there is no difficulty about Engl. read [red] (past tense of read)
and red (the color). But in varieties of American English in which can and can’t are
pronounced identically (as something like [kæǝn], see Chapter 4, § 4), problems do
arise, and people will ask questions such as Is she able or not?
In principle, homonymy is fundamentally different from polysemy, but the
two phenomena are not always easy to distinguish. For instance, are the two
expressions ear of corn (etc.) and ear of animals just homonymous, or are they a
single, polysemous word? What about reader, the noun corresponding to the read
of (1) above, compared to the British English university rank of Reader = roughly
‘Associate Professor’? And what about the two highlighted words in (14)?

(14) The Air Marshal of the People’s Republic declared martial law today.

For literate speakers of English, the spellings of marshal and martial would guar-
antee that the words are considered mere homonyms, in spite of their identical
pronunciation as [maršǝl]. But what about speakers that are not literate? As for
Reader, many speakers might consider the word a specialized use of the normal
word reader, because Readers read out their lectures. Historically, this is no doubt
how the word acquired its meaning. But do all Readers read out their lectures?
Don’t some of them lecture without a written text? And what about the British uni-
versity rank of Lecturer = roughly ‘Assistant Professor’? Does a Reader lecture any
less than a Lecturer and a Lecturer read any less than a Reader? As for ear and ear,
American speakers might see a relationship, since ears of corn for them refer to
‘cobs of the maize plant’. And such cobs stick out from the plant in such a manner
that with some imagination they could be compared to the ears of human beings
or animals. In British English, such an interpretation would be preposterous. The
word corn commonly designates a plant such as ‘wheat’, whose ears are of a very
different shape and therefore could hardly be considered even remotely similar to
human or animal ears.
It is possible to argue that the tendency to avoid both absolute synonymy and
excessive homonymy springs from a common principle. That principle, in fact, is
identical to the one invoked earlier as the motivation for the analogical process of
leveling, which in Chapter 5, § 2.1 was characterized by the slogan one meaning –
one form. In the present case, this means that if there are two different forms,
then we expect there to be two different meanings; and if there is no formal dis-
tinction, then we expect no semantic distinction, either. But just as in the case of
leveling, our response to violations of the principle will be selective, based on an
evaluation of the significance of the violation.
198 Semantic change

4 The relationship between sound and meaning

As noted earlier, the relationship between meaning and form is arbitrary,
because the indirect encoding of meaning through combinations of inherently
meaningless speech sounds in effect entails that. This can be seen from the fact
that different languages use phonetically very different expressions to express
(roughly) the same meanings. See for instance the words for ‘dog’ in (15a). The
converse is true, too – different languages may assign very different meanings to
forms that are phonetically identical or quite similar, as shown by (15b).

(15) a. Engl. dog, Germ. Hund [hunt], Fr. chien [šyɛ̃ ], Span. perro [per̄o], It. cane
[kane], Lith. šuo, Finnish koira, Hindi kuttā, Tamil kūraṇ, Arabic kalb,
Amharic wuššā
b. Arabic kalb ‘dog’ : Germ. Kalb ‘calf’
Turkish beter ‘worse’ : Engl. better
Turkish alt ‘bottom’ : Germ. alt ‘old’
Turkish kar ‘snow’ : French car ‘because’, Engl. car
Sanskrit yunaǰmi ‘I join’ : Engl. you nudge me
Gk. hén ‘one’ : Engl. hen
Mod. Gk. míti ‘nose’ : Engl. meaty
Cree tanse ‘how?’ : Germ. tanze ‘I dance’
Albanian nis ‘begin’ : Ukrainian nis ‘nose’

Note especially cases such as Arabic kalb : Germ. Kalb, where both words refer
to a type of animal to be sure, but to very different animals (‘dog’ vs. ‘calf’) and
pairs like Turkish beter : Engl. better, where the meanings literally are opposite,
‘worse’ vs. ‘better’.
Further evidence for the arbitrariness of the meaning-sound relationship can
be seen in the fact that sound change and/or other changes may alter the shape of
words beyond recognition, without affecting their meaning. For instance, believe
it or not, Germ. Hund, Fr. chien, It. cane, and Lith. šuo all are reflexes of the same
proto-language word, PIE *ḱuon-.
In spite of these facts, however, ordinary speakers, without training in lin-
guistics, tend to believe that the relation between word and meaning is in and
of itself meaningful, not arbitrary. Note for example the statement in (16) or the
story, possibly apocryphal, of the German-speaking traveler from Tyrol (Austria)
who, upon coming to Italy, was struck by how senseless it was of the Italians to
say cavallo for the animal (the horse) that everyone knows is called a Pferd! Lin-
guists may consider such statements and stories naive and irrelevant. But here as
elsewhere we should keep in mind that most speakers are not linguists and, as
 The relationship between sound and meaning 199

seen below, their opinions do matter, by providing the motivation for linguistic
changes – whether linguists like it or not.

(16) It’s very clear why pigs are called pigs – they’re very dirty animals.

The feeling that the sound-meaning relationship is not arbitrary is especially

prevalent and appropriate as regards onomatopoeia, i. e., words intended to
capture sounds of nature through similar speech sounds. Compare for instance
the English animal sounds in (17).

(17) rooster: cockadoodledoo

young bird: peep, cheep, chirp
dog: arf, bow-wow
cow: moo

The special relationship between meaning and sound in onomatopoeia becomes

especially obvious if we consider possible alternatives. Forms like peep, cheep,
and chirp simply won’t do as imitations of the rooster’s call; nor will moo work for
the dog’s bark, or bow-wow for the sound of cows. Further, it would be strange to
refer to the sounds of young birds as choop, or to the sound of cows as mee.
Examples like cheep vs. moo illustrate a very general tendency of onomato-
poeia to use high front vowels to refer to the sounds of young or small animals and
human beings, and back vowels (or at least, vowels that are not high and front)
to depict the sounds of adult or large creatures. This tendency is grounded in the
fact that the sounds emitted by small creatures tend to be higher pitched and as a
consequence acoustically closer to high front vowels.
The relation between high front vowels and ‘small’ may be exploited in areas
that do not involve onomatopoeia, such as teeny vs. tiny, leetle vs. little, where
the forms with ee [ī] have front vowels higher than the other forms and designate
something even smaller. Such non-onomatopoetic relationships between sound
and meaning are referred to as sound symbolism.
Another area in which non-linguists commonly find an especially salient
non-arbitrary relationship between meaning and form is taboo, i. e., words which
are considered forbidden or at least impolite under certain social circumstances.
Here it is not so much the meaning that is found unacceptable, but the linguistic
expression of that meaning. In English, for instance, expressions like (18a) are
acceptable in most circles, even if they may appear excessively severe. But the
nearly synonymous (18b) is not, either in this form or without an object, used as a
general expletive. The fact that the offensiveness of (18b) lies in its linguistic form
becomes especially clear when we contrast it with the expressions in (18c), where
200 Semantic change

a simple alteration in phonetic form renders the expression much less offensive.
Evidently, the different distortions in (18c) permit speakers to have their cake and
eat it too. They can utter the tabooed expression, but forestall reprimand by being
able to claim that they did not “really” use it. (See also Chapter 4, § 4.)

(18) a. (May) God punish them by sending them to hell

b. (May) God damn them
c. dang, darn, gosh darn, doggone, dad burn(ed) …

While ordinary speakers (i. e. non-linguists) may be convinced that there is a

clear connection between sound and meaning, at least in onomatopoeia, sound
symbolism, and taboo, linguists are just as firmly convinced that even here the
relation is to some degree arbitrary. To support this view they can point out that
other languages often employ different onomatopoetic means to express the same
animal sound. Compare for instance (17’). Similarly, languages – or their speak-
ers – may disagree over what is taboo. While there is a great tendency for words
for excrement, sexual activities, or the name of the divinity to be taboo, in some
societies it is the name of one’s elder sister, the names of deceased persons and
everything which sounds like these names, or yet other words. Even so, the belief
of linguistically naive speakers that the sound-meaning relationship in such
words is not arbitrary is sufficiently powerful to bring about linguistic change.

(17’) German French Hindi

rooster: kikeriki cocorico kuk(a)ṛũkũ
dog: wauwau tou-tou bhõ-bhõ

5 Factors responsible for semantic change

In large measure, semantic change results from the same factors that are respon-
sible for the fuzziness of meaning, namely the semantic relation between words,
and the relationship that we perceive between meaning and form. The effects of
these, as well as other factors, are discussed in the following sections.

5.1. Metaphor. The major vehicle through which words acquire new or broader
meanings is metaphor.
Metaphoric language comes in many different shapes, which in traditional
rhetorical theory have been classified in various subtypes. One of these is the
designation of a thing or person by means of its most salient part, as in (19).
 Factors responsible for semantic change 201

The extended meaning in (19a) results from the fact that in a traditional setting,
employers would consider the hands to be the most important part of laborers.
(Their minds, obviously, are of little concern.) Similarly, the meaning in (19b)
reflects the fact that the most important part of a table is the board on top. (Legs
are there merely to support the top board.)

(19) a. Engl. hands = ‘laborers’

compare Span. braceros ‘laborers’ : brazo ‘arm’
b. Engl. board = ‘table’
c. Engl. wheels = ‘car’

A very similar broadening of meaning can be accomplished through the desig-

nation of a group of things or persons by means of a word referring to something
with which that group is habitually associated. Compare (20). The developments
in (20c,d), presupposing a development as in (19b), show that metaphorical
extensions can build on other, previous metaphorical extensions. In principle
there is no limit on how far such chains of metaphorical extensions can go.

(20) a. The pulpit = ‘the clergy’

b. The bar = ‘the legal profession (which pleads its case at the bar separat-
ing the jury etc. from the rest of the court)’;
c. The board (of directors) = ‘those who sit around a table in order to make
decisions’
d. (Bed and) board = ‘the food served at a table’

A large number of metaphoric uses are motivated by what can be called social
factors, employing a fairly broad definition of the term social. For instance, we
may choose to give our claims greater impact by exaggeration or to mitigate their
force through understatement. Compare (21) and (22), where the examples under
(a) illustrate current extended metaphorical uses, while the (b) examples illus-
trate original metaphors which have lost their metaphorical flavor.

(21) a. Engl. I’m terribly, awfully, frightfully sorry

I’m sorely disappointed
This is the most unique experience I’ve ever had
b. Germ. sehr ‘very’, cognate of Engl. sore.
Engl. very < ME verai ‘true, truly’ (⇐ OFr. verai ‘true’)

(22) a. Engl. I feel a little, a bit under the weather

The danger is not inconsiderable
b. Give me a couple of oranges (= a few, several oranges)
202 Semantic change

Other socially motivated metaphors include euphemism and paronomasia or

punning. Euphemisms may be used to avoid words that are under taboo (see
below), but also to avoid a large range of other unpleasant connotations. The
examples in (23) may serve to illustrate the range of possible developments. Politi-
cians are notoriously adept at employing euphemisms in order to “pretty up” their
actions and ideas. The use of euphemism in totalitarian states has been carica-
tured in the Newspeak of George Orwell’s 1984, with expressions such as joycamp
(a forced-labor camp) and Ministry of Love (a prison in which dissenters are tor-
tured). Some euphemisms, such as pacified village, may never succeed in becom-
ing part of the regular vocabulary. But words such as mortician and beautician
have. (See Chapter 5, § 3.2 on these two words.) Paronomasia, on the other hand,
does not normally seem to make lasting contributions to the lexicon. In pres-
ent-day American English, it is a favorite device for creating names of business
establishments, see (24). Note however that something very much like parono-
masia played a great role in the early development of writing. (Recall for instance
the Sumerian use of the symbol for ti ‘arrow’ to refer to ti ‘life’; Chapter 3, § 2.3.)

(23) deceased, loved one = ‘dead body, corpse’

mortician = ‘undertaker’
beautician = ‘hair dresser’
pre-owned automobile = ‘used car’
pacified village = ‘village which has been forced to side with “us”’
liberation = ‘killing or putting into “reeducation camps” people who
sided with “them”’
ethnic cleansing = ‘killing or expelling “unwanted” ethnic groups’
final solution = ‘“elimination” of Jews, Gypsies (i. e., Roma and Sinti),
and other “undesirables”’

(24) The Mane Event, Head Hunters, Headmasters, Cost Cutters, From Hair to
Eternity, Shear Delight (all names for hair cutting salons)
Cuttin’ Corners, The Mower’s Edge, The Lawn Ranger (names of lawn
mower servicing and lawn mowing businesses)
Chin’s Wok’n Roll Cafe, The Great Impasta, Lox Stock & Bagel, Wiener
Lose, Snaks Park Avenue (names for restaurants, delis)
The Daily Grind, Common Grounds (names of coffee houses)
The Laser’s Edge (name for an establishment that produces laser-printed
resumés)

The use of metaphor is especially widespread in certain types of language use.

Beside the language of politicians, with its heavy reliance on euphemisms, the
 Factors responsible for semantic change 203

most notorious areas of such language use are argots, jargons, and slang.
These are discussed more fully in Chapters 9 and 10.
As noted in § 2 above, in many cases the metaphoric link is pushed below the
level of consciousness. In cases like clear, the link can easily be brought back to
the surface. But in other cases the relationship has become much more tenuous.
Compare for instance the relationship between reader and the British university
rank of Reader (§ 3 above).
This common tendency for the metaphoric link to become attenuated is
referred to as semantic fading. If fading goes far enough, the link between origi-
nal and derived meaning becomes severed, and synonymy results in homonymy,
as in the case of Mod. Fr. pas ‘not’ and pas ‘step’ (Chapter 5, § 3.2). Similarly, there
is no obvious semantic connection between very and veritable, in spite of their
historical relationship (as they ultimately go back to forms derived from Lat. vērus
‘true’).

5.2. Taboo. Taboo likewise tends to lead to frequent vocabulary renewal. This
can be gauged by the large number of lexical replacements for tabooed words,
especially ones considered most objectionable. In addition to the examples in
(18) above, see the set in (25a) which serves to avoid bloody, a word which is
considered taboo in British English. (25b) illustrates the fact that words for ‘toilet’
are subject to similar constant lexical renewal. The Victorian era is said to have
been especially notorious for the degree to which at least some speakers placed
all kinds of words under taboo because of their – often marginal – sexual conno-
tations. Some of the taboo-induced replacements of that era have been retained,
and those in (25c) are often cited as examples. But modern speakers usually are
no longer aware that these result from taboo. In the case of white meat and dark
meat, the retention may be due to the fact that the linguistic distinction made by
the terms corresponds to a widespread distinction in taste preference.

(25) a. Brit. Engl. blasted, bleeding, blighted, blooming, blessed, etc.

b. bathroom, powderroom, ladies’/men’s room, lavatory, WC, loo, etc.
c. chest ‘breast’
limb ‘leg’
white meat ‘(chicken) breast’
dark meat ‘(chicken) legs and thighs’
drumstick ‘(chicken) leg’

The examples in (18) and (25) show that there may be different lexical reactions
to taboo. While (25b) and (25c) simply replace the tabooed words without any
consideration for the way they are pronounced, the replacements in (18) and (25a)
204 Semantic change

sound similar to the words under taboo. In many cases, the result is a pre-existing
word which simply gets to be used in a new meaning. In others, such as gosh, the
result is a totally new word. Words of the latter type make it useful to distinguish
between taboo-induced replacement and taboo-induced deformation.
In spite of the social restrictions against them, tabooed words may be remark-
ably persistent. Some of the “Anglo-Saxon four-letter” words considered most
offensive can be traced back to Old English and beyond. While it is not appropri-
ate to use these tabooed words in polite company, there are many other areas of
language use in which they may be quite appropriate (such as among adolescents
or in the military). And sometimes, breaking the taboo may be considered appro-
priate, as a sign of deep anger. Even persons who would never permit themselves
to utter such words certainly know them. Thus, a famous lexicographer is said
to have been approached by an elderly lady who congratulated him on his new
dictionary, but added, “You naughty man; there are a lot of naughty words in your
dictionary.” Whereupon the lexicographer replied, “You naughty lady; you knew
precisely where to look.”
Lexical replacement is not necessarily limited to the tabooed words. “Inno-
cent bystanders”, words that happen to be mere homonyms, may be affected
as well – or instead. For instance, American English has generally replaced ass
by donkey and cock by rooster because of their homonymy with words under
strong taboo. In the case of the perhaps most heavily tabooed word of English,
the process of replacement seems to have gone even farther, eliminating all inde-
pendent words with short vowel in the context between [f] and [k]; see the data in
(26), where dates in parentheses indicate the last attestation cited in the Oxford
English Dictionary. Significantly perhaps, most of these last dates come from the
Victorian era.

(26) fuk (a sail) (1529)

fac ‘factotum’ (1841)
feck ‘effect, efficiency’ (1887) (now only in feckless)
fack/feck (one of the stomachs of a ruminant) (1887)
feck(s)/fack(s) ‘(in) faith, (in) fact’ (1891)

The effects of taboo-induced replacement and deformation can be quite far-reach-

ing. Many Polynesian and Micronesian societies, for instance, have the tradition
that upon the death of a person, his or her name becomes subject to taboo. More-
over, since names tend to be compounds of ordinary words (of the type ‘noble
warrior’, ‘slim-waisted’, or ‘dances with wolves’), their component parts like-
wise may be subject to taboo. Even derivational and inflectional affixes may be
affected. The result has been such a large turnover in the lexicon that it is difficult
 Factors responsible for semantic change 205

to establish the degree of relationship between different languages on the basis

of vocabulary correspondences.
Deformation due to taboo likewise can create difficulties in comparative
linguistics. Note for instance the word for ‘tongue’ in Indo-European. Being the
organ of speech, the tongue was in earlier, more “primitive” times considered to
be mystically identical with speech and therefore, like speech, a supernatural
force. Giving a name to somebody or something would give one power over that
person or thing. We see echoes of this identification of ‘tongue’ and ‘speech’ in
such expressions as the English tongue or speaking in tongues, as well as in our
word language which, via French, derives from Lat. lingua ‘tongue, language’;
note also Arab. lisan ‘tongue, language’.
Below are given the words for ‘tongue’ from most of the (ancient) Indo-Eu-
ropean languages, together with the forms from which they could be derived by
regular sound change, if they had developed only according to sound change.
As can be seen, although there clearly was a word for ‘tongue’ in Proto-Indo-
European, which probably contained the sounds [n̥ ǵhwā/ū-], taboo-induced
deformation has taken its toll to such an extent that the phonetic nature of
the reconstructed word cannot be determined with greater precision. For
instance, the question whether there was an initial consonant and what the
nature of that consonant was cannot be resolved on the basis of the extant
evidence.

(27) Indo-European ‘tongue’:

Lat. lingua, OLat. dingua : *dn̥ ǵ(h)wā- (with some help from
lingō ‘lick’?)
Oscan fangva : *dhn̥ ǵ(h)wā-
OIr. tengae : *tn̥ ǵ(h)wā-t-
Engl. tongue : *dn̥ ǵhwā-
OCS językŭ : *Øn̥ ǵ(h)ū-
Lith. liežuvis : *Øn̥ ǵ(h)ū- (?) (with help from liež-
‘lick’)
Avest. hizvā : *siǵ(h)wā
Skt. ǰihvā : *ǵ(h)iǵhwā

5.3. Onomatopoeia. Even if there is a certain arbitrariness to onomatopoeia (see

§ 4 above), its purpose is to mirror sounds of the real world. If sound change
makes a given onomatopoetic expression too different from the real-world sound
it is to depict, one of two things can happen. Either the word ceases to be ono-
matopoetic, or its pronunciation is remade so as to restore the fit between speech
sounds and the real-world acoustic effect.
206 Semantic change

Middle English had a verb pīpen ‘to chirp, make the sound of little birds’
whose [ī] very nicely approximated the acoustic impression created by the sounds
of little birds. As a result of the Great Vowel Shift (see Chapter 4, § 5.4), pīpen
turned into Mod. Engl. pipe, pronounced [payp]. The verb is still listed in the dic-
tionaries; but it is rarely used, except perhaps in the expressions to pipe up and to
pipe down. In ordinary language it has been replaced by words like peep, cheep,
chirp whose sounds more closely mirror the chirping of little birds. Similarly, Clas-
sical Greek had a word bē [bɛ̄ ] to depict the sound made by sheep. Regular sound
change would have turned the word into Mod. Gk. [vi], a far cry, we might say,
from the sound of sheep. The word has been productively replaced by [bɛ] which,
given the vagaries of Modern Greek spelling, is written μπε = mpe.
Onomatopoeia also may be responsible for the creation of new linguistic
expressions. For instance, among the words for ‘dog’ in example (15), the Hindi
one is onomatopoetic in origin.
Interestingly, the Hindi word is not the only onomatopoetic replacement of
an earlier word for ‘dog’ in the history of Hindi. The Proto-Indo-European form
*ḱuon- underlying the Germ. Hund, Fr. chien, and Ital. cane of (15) above came
out as śvan- in Sanskrit. A Modern Hindi descendant of this word is sōnhā, whose
meaning has become specialized to ‘wild dog’. Within the history of Sanskrit, a
new, onomatopoetic word was created, namely kurkura-, literally no doubt ‘the
one that snarls, growls, or barks, i. e. makes the sound [kurkur]’. (Compare in this
regard the German word for ‘growl, snarl’, knurren.) The Modern Hindi reflex of
this form is kūkar, whose meaning tends to be specialized as ‘puppy’. The normal
Hindi word, kuttā, cannot be traced to any Sanskrit antecedents and thus must
reflect an even later instance of onomatopoeia. Similarly, Mod. Engl. cur is derived
from MEngl. curre, a shortened form of cur-dogge ‘growling, snarling dog’, whose
cur may be from Scand. kurra ‘growl, snarl’. Similar developments are found
outside Indo-European, as in Tamil kurai ‘to bark’ : kūraṇ ‘dog’. (On Engl. hound
see § 6.1 below.)
For another effect of onomatopoetic considerations, see the discussion of
Engl. bang, clang; bash, clash, crash, smash; batter, clatter, smatter, shatter in
Chapter 5, § 3.1.

5.4. Avoidance of excessive homonymy. Although there is a great amount

of homonymy in language, mere homonymy does not cause any great difficul-
ties. Engl. read (pres.) and reed sound the same, and so do read (past) and red.
However, reed and read hardly ever have a chance of occurring in the same context
and of being mistaken, one for the other. And even if, with some imagination, we
can think up a context that might be ambiguous, such as This is a good [rīd], sen-
tences of this sort are hardly ever uttered out of context. Thus, if the sentence is
 Factors responsible for semantic change 207

uttered by someone pointing at a book (or in a context where someone has just
mentioned a book), we can be sure that the word read is meant. If the reference is
to something like the mouthpiece of a saxophone or plants on a lake shore, then
reed must be intended.
Some instances of homonymy, however, can create genuine confusion with
potentially quite undesirable results. For instance, Old English had two verbs,
lǣtan ‘let, permit’ and lettan ‘hinder, prevent’, both of which became Mod. Engl.
let through regular sound change. Now, assume someone had robbed us and were
running down the street. If we called out, Let that man, nobody would be sure
whether we meant ‘stop that man’ or ‘let/permit that man (to run through)’. Sim-
ilarly, Lat. cattus ‘cat’ and gallus ‘rooster’ both came out as gat in Southwestern
French. The resulting ambiguity could be disastrous in a rural society, for it makes
quite a difference whether the gat reported to have entered the hen house is a
rooster or a cat.
In cases of such “excessive homonymy”, one of the two homonymous words
soon gets replaced. For instance, English no longer uses let in the meaning ‘hinder’
or ‘prevent’, except in a few opaque relics, such as without let or hindrance or let
ball (in tennis). (But note that let ball frequently is made more transparent by
folk-etymological change to net ball.) Similarly, Southwestern French gat ‘rooster’
was replaced by a variety of other words, such as the dialectal word for ‘vicar’.
(See also § 5.6 below, as well as Chapter 4, § 4 for cleave ‘stick to’ vs. cleave ‘chop,
split’ and Southern U.S. Engl. pen [pin] and pin [pin].)

5.5. Avoidance of synonymy, semantic differentiation. As we have seen,

language is designed to convey the maximum amount of meanings through a
minimum amount of lexical items. One would expect, then, that complete syn-
onymy – where two phonetically distinct words would express exactly the same
range of meanings – is highly disfavored.
This expectation has been met, for where other types of linguistic change
could give rise to complete synonymy, we see that languages – or more accurately,
their speakers – time and again seek ways to remedy the situation by differentiat-
ing the two words semantically. For instance, analogical processes do not always
lead to the complete replacement of the old form by the new. In many cases, the
two forms continue to coexist. In all cases, however, their meanings get differen-
tiated, so that the new form almost invariably will take on the more productive,
basic meaning, while the old form persists in more specialized, marginal func-
tions. Examples are given in (28).
208 Semantic change

(28) New form Old form

brothers brethren
older elder
orientate orient (verb)
housewife hussy

Borrowing can lead to similar doublets, and again we see that they tend to be
semantically differentiated. The only difference from analogical examples such
as (28) is that the direction of differentiation cannot be predicted. In some cases
the native word remains more basic, in others, the borrowing wins out, and in yet
others it is difficult to tell which is more basic. Compare for instance the examples
in (29).

(29) Native word Borrowed word

kingly royal (from French)
cow beef (from French)
shirt skirt (from Scandinavian)

Even other semantically induced changes, such as onomatopoeia, can lead to

semantic differentiation. This is no doubt the reason for the semantic differenti-
ation of Hindi sōnhā ‘wild dog’ vs. kūkar ‘puppy’ vs. kuttā ‘dog’ (see § 5.3 above).
Such developments are all in keeping with the one meaning – one form slogan
we have now seen at work repeatedly.

5.6. Reinterpretation. Another major reason for shifts in meaning is re­inter­

pretation. A fairly simple case of reinterpretation is found in the develop-
ment of earlier Engl. bead ‘prayer’ to ‘counter on a rosary; bead’; see Chapter 3,
§ 2.2.
Another, more complex, case is the one of oriental, discussed in Chapter 1.
Originally the term referred to the areas just to the east of the Graeco-Roman
world, also known as Anatolia or the Levant, in ancient times also as Asia. In
fact, just like Anatolia and Levant, the term orient originally referred to the rising
(sun) and then, by metaphorical extension, to the east where the sun rises. (See
§ 6.2 below.) Unlike Anatolia and Levant, but like Asia, the term slowly came to be
used in a broader range of meanings, so that ultimately it could refer to the entire
continent of Asia. As we have seen in § 2, words with such a broad semantic range
tend to acquire a quasi-core meaning, based on a prototypical understanding of
the term. This is no doubt what happened with oriental. East Asians, being most
clearly different from Europeans, have come to be considered prototypical orien-
tals, not only by Europeans and European Americans, but also by East Asians or
 Factors responsible for semantic change 209

their descendants in countries like the United States. The final step in the develop-
ment lies in the reinterpretation of this prototypical meaning as the core meaning
of the word.
Interestingly, a similar development is affecting the word Asian. However, in
this case the change is running into active opposition from citizens of other Asian
countries such as India or Sri Lanka (or their descendants in countries like the
United States) who do not look kindly at the prospect of being left without a name
for the continent they share with the East Asians.
Reinterpretation is not always easily distinguishable from some of the minor,
more sporadic analogical changes, especially popular etymology. For instance,
through sound change the Old English words wēod ‘plant’ and wǣd(e) ‘garment’
both became Mod. Engl. weed. The resulting homonymy apparently was too great.
In the meaning ‘garment’ the word generally was replaced by other lexical items
(such as garment); weed, by and large, survived only in the meaning ‘undesirable
plant’. However, relics of ‘garment’ remained in the rare term weed ‘a token of
mourning, such as a black armband’ and the somewhat more common expression
widow’s weeds ‘a widow’s mourning clothes’. Most speakers of Modern English
familiar with these terms no doubt have reinterpreted weed in these expressions
as a specialized use of the “victorious” word weed ‘undesirable plant’, with the
rationalization that mourning clothes or cloths are less colorful and fashionable,
and more “weedy” than others. This rationalization can be explained as the result
of reinterpretation. But for those speakers who only know the term widow’s weeds,
the development can equally well be attributed to folk etymology, motivated by
the fact that the word weed has not survived in the meaning ‘garment’ outside
the compound. (Compare the similar case of Engl. bride-gum* → bridegroom in
Chapter 5, § 3.2.)
Reinterpretations often reflect changes in culture and society. For instance,
Old French had a word marechal, a borrowing from the word māre-skalk found
in the Frankish speech of their Germanic overlords. The original meaning of the
word was ‘a stable hand (skalk) in charge of horses (māre)’. Now, horses were
very important war equipment in medieval times. Hence, the word marechal was
reinterpreted as ‘somebody in charge of important war equipment’. Further exten-
sions and reinterpretations along similar lines led to increasingly loftier connota-
tions: ‘somebody in charge of horses and horsemen’→ ‘somebody in charge of the
cavalry’→ ‘a (high) military officer’, and so on.
Similarly, Gk. presbúteros originally designated an ‘older (person)’. In a
society governed by older and presumably wiser persons, the meaning of the word
came to be extended to ‘older person who is a community leader’, whence it could
be reinterpreted as simply meaning ‘community leader’. As a consequence, now
even younger persons can become ‘presbyters’. Compare also NE elder, an earlier
210 Semantic change

form of the comparative of old, but now commonly used to refer to ‘community
leaders’, especially of religious communities, without regard to age.
Reinterpretation can proceed in many different, often contradictory direc-
tions. Thus, Old English had the words cnafa ‘boy’ and cniht ‘servant’. While the
connotations of the former were relatively neutral or even positive, those of cniht
were relatively lowly. In the Modern English reflexes, knave ‘villain’ and knight
‘nobleman’, the connotations are just about reversed.
The shifts in meaning of cnafa and cniht can be explained in terms of
well-precedented extensions and reinterpretations. In medieval society, cnafas
> knaves ‘boys’ tended to be apprentices, and apprentices were treated like serv-
ants or even serfs. And since servants and serfs do not necessarily like the treat-
ment meted out to them by their masters, they may act in ways that their masters
perceive as “uppity, insolent, not-to-be-trusted”; hence the modern meaning of
knave. The word cniht acquired its lofty meanings in a slightly different, but con-
temporary context. In medieval warfare, noblemen often took some of their serv-
ants with them into battle. In this context, the word cniht could be reinterpreted as
referring to a lower-rank warrior and subsequently, because war was considered a
noble enterprise, even to a lower-rank nobleman. (See also § 6.1 below.)
The role of cultural factors in reinterpretation can be seen in the development
of the Modern English word write. As noted in Chapter 3, § 2.6, in early Germanic
runic writing, letters were mainly produced by means of scratching or engraving
into wood. As a consequence, the verb *wrītan- ‘scratch’ could be used appro-
priately to refer to the act of writing. The arrival of Christianity introduced not
only a new alphabet (the Roman one), but also new writing materials such as
parchment, and letters were no longer scratched into the material but applied
to it in ink, by means of a quill. In some of early Germanic, these differences
were apparently considered too great for the old verb still to be appropriate, and
the Latin verb, scrībō, was borrowed; compare Mod. Germ. schreiben ‘write’. In
English and Icelandic, on the other hand, the similarities outweighed the differ-
ences. Both the runes and the Roman letters served to put words into written form.
As a consequence, the word wrītan was reinterpreted as referring to this shared
process. This led to a certain complication, in that now there was a question as
to whether wrītan ‘write’ was the same word as wrītan ‘scratch’. In English, the
difficulty resolved itself as wrītan ‘scratch’ became obsolete. Icelandic, however,
preserves ríta ‘scratch’ beside ríta ‘write’; but because of their semantic remote-
ness, the two words probably have become simple homonyms.
Interestingly, semantic reinterpretations such as the one of Germanic *wrītan
owe their existence not so much to active processes like metaphoric extension,
but rather to inertia. The old term simply continues to be used even though the
activity or phenomenon designated by it has undergone significant change.
 The effects of semantic change 211

5.7 Other linguistic changes. As seen in § 5.5, semantic differentiation can be

triggered by analogy, borrowing, and even semantically induced changes. Non-se-
mantic linguistic changes have led to many other types of semantic change.
For instance, the fact that American English speakers can at least entertain
the idea that ear (of corn) and ear (of a human being or animal) might be the same
word (§ 3 above) was made possible by sound change. Originally, the two words
were quite different: ear (listening organ) goes back to PGmc. *auzō, cognate
with Lat. auris ‘ear’while the ear of corn reflects *ahuz-, cognate with Lat. acer
‘sharp, pointed’. Similarly, as noted in Chapter 5 (§ 3.2), Mod. Engl. daisy derives
from MEngl. daies-ei(e) ‘day’s eye’, a metaphor for ‘sun’, much like sunflower
in present-day English. As sound change obscured the compound character of
daisy, the sense of metaphor was irretrievably lost (to all, of course, but historical
linguists).
Borrowing often is accompanied by semantic change. For instance, Lat. com-
prehendere, meaning ‘grasp, include’ could, like Mod. Engl. grasp, be used met-
aphorically to mean ‘understand’. The English borrowing comprehend is gener-
ally used only in the originally metaphorical meaning. But in the absence of the
originally basic meaning, the word no longer feels like a metaphor. (The original
meaning is preserved in comprehensive.) Similarly, the German word Angst has
been adopted in English as a technical term of Freudian psychology. In German,
the Freudian use is merely an extension of the basic meaning of Angst, namely
‘fear, anxiety, anguish’.
Sometimes, the special connotations associated with borrowings give rise to
linguistic change. For instance, Japanese has borrowed many English words for
non-traditional, “western-value” concepts, such as gārufurendo ⇐ girl friend. The
morally somewhat loose connotations of such words then led to the coinage of
new terms with similar connotations, made up from borrowed words, such as
rabu hoteru ‘love hotel’ = a place where people have illicit sexual relations. Some-
thing similar happened in American English advertising, where the fashionable
connotations associated with French borrowings spawned a whole set of “qua-
si-French” expressions like le car and le bag. (See Chapter 8, § 4 for similar “qua-
si-foreign” forms in English.)

6 The effects of semantic change

If we look at the semantic changes so far examined, we can observe a wide variety
of effects. The sound structure of words may be affected by taboo-induced defor-
mation and onomatopoetic considerations. Taboo and the avoidance of excessive
212 Semantic change

homonymy can lead to vocabulary loss. Such loss can, of course, likewise result
from changes in society and culture which render particular words unnecessary.
For instance, the word thill ‘either of a pair of shafts or poles between which an
animal is hitched to pull a wagon’ has effectively ceased to be used, except perhaps
by the few individuals that still drive animal-drawn wagons. Many other similar
words became obsolete with the introduction of the internal combustion engine.
(Other words, however, such as car or wheel have survived through inertia, albeit
with different meanings.)
Of course, what most saliently gets affected by semantic change is the
meaning of words, including their connotations. Original homonyms may some-
times become, or threaten to become, polysemous variants of the same word,
such as ear and ear. Just about the exact opposite may occur, too, as in the case
of PGmc. *wrītan ‘scratch/write’ : Mod. Icel. ríta ‘scratch’ and ríta ‘write’. And so
forth.
In some cases, whole fields of words undergo similar semantic changes, such
as those words referring to animal-drawn wagons and chariots that were retained
by inertia after the introduction of the internal combustion engine. Similarly,
when Britain changed from an absolute form of monarchy to a parliamentary one,
words like king, queen, prince, princess, court all acquired connotations appropri-
ate to the new political context.

6.1. Social attitudes and change in connotations: Especially noteworthy are the
semantic developments of words like OFr. marechal, OE cniht, Gk. presbúteros on
one hand, and OE cnafa on the other (see § 5.6 above). While the words of the first
set have acquired connotations that are considerably more favorable, the conno-
tations of cnafa have become much less favorable. Developments of the former
type are referred to as melioration, those of the latter, as pejoration.
Both types of development are quite common and tell us a lot about social
attitudes. Pejoration, for instance, has time and again affected words referring to
young, innocent, or defenseless people. Just consider the sources for the English
words silly and daft, as well as the semantically similar Germ. albern ‘silly’, Fr.
niais ‘stupid’.
The word silly ultimately derives from OE sǣlig which had the meaning
‘happy, blessed, blissful’, a meaning preserved in Middle English (as in þurh
seli martirdom ‘through blessed martyrdom’; 13th c.). Middle English also offers
extended meanings that come closer to the modern semantics of silly, such as Vp
an seli asse he rod ‘he rode on a humble (or simple) donkey’ (13th c.) Humility and
simplicity, however, are often equated with feebleness; and expressions like seely
Idiotes (16th c.) illustrate another unfortunate, but common development – the
extension of words meaning ‘weak’ or ‘feeble’ to mean ‘feeble-minded’ or ‘stupid’.
 The effects of semantic change 213

Brit. Engl. daft ‘crazy’ developed along very similar lines from ME dafte ‘gentle,
mild’, OE (ge)dæfte ‘mild, meek’. Similarly, Germ. albern ‘silly’ is derived from
MHG alware ‘simple’, which itself comes from OHG alawāri ‘kind, gentle’. French
niais can be derived from Lat. *nīdax ‘nestling’, via ‘helpless’, ‘simple’, ‘foolish’.
Some of the connotations of Engl. simple point in the same direction; and simple-
ton only has the pejorative meaning.
Interestingly, the association of ‘simple’ or ‘foolish’ with ‘young’, ‘helpless’, or
‘delicate’ sometimes can lead to reverse developments of melioration. An example
is found in Engl. nice, a borrowing from Old French ni(s)ce ‘stupid, foolish’ which,
in turn, reflects Lat. nēscius ‘unknowing, ignorant’. In earlier English, the word
still preserves the meaning ‘ignorant, foolish’. The modern meaning seems to
have developed via the meanings ‘shy, bashful’ and hence ‘delicate, dainty’.
A different development, also common in words referring to the powerless, is
seen in examples like OE cnafa : NE knave. For parallels see the following exam-
ples.
The word boy may be used by, say, a stereotypical Southern U.S. sheriff, in
addressing a black man even if that man is in his eighties. Similarly, Gk. paîs and
Lat. puer, whose literal meaning is ‘boy, child’, were also used to designate ‘serv-
ants’ and ‘slaves’. Developments of this type are not necessarily limited to western
or Indo-European languages. In Kharia (a Munda language in India), for instance,
kɔn-ghɛr ‘young man’ likewise has come to be used to mean ‘servant, slave’.
Note also common, mean, originally ‘common; normal, average’; villain, orig-
inally ‘belonging to the villa (i. e. the landlord’s mansion) or to the village (hence,
peasant)’; varlet, a relative of valet, both from French and originally meaning
‘young servant’ (ultimately derived from Celt. *wasso- < *upo-sto- ‘standing by,
waiting on’). Interestingly, the word vassal, derived from the same source as
varlet, acquired more positive connotations when it was used to refer to feudato-
ries of a king or prince.
Similarly churl is from OE ceorl ‘(free) man’ → ‘commoner’ → ‘person lowest
in the social order’ → ‘peasant’ → ‘rude person’; and dial. Engl. carl, a borrowing
from ON karl ‘man’, changed to ‘peasant, serf’ → ‘rude person’. Here, too, we find
that a related word, the Frankish variant Karl, gave rise to the much more lofty
French name Charles (whence the English name) = Lat. Carolus, Germ. Karl. And
because of its association with the emperor Charlemagne, Carolus Magnus, Karl
der Große, this word was borrowed with the meaning ‘king’ into various Slavic
languages; compare South Slavic kraly ‘king’.
Sexist attitudes are reflected in the fate of many words referring to women.
Shakespeare’s quean ‘loose woman’ reflects OE cwene ‘woman’; NE hussy is
earlier hūswīf ‘housewife’; NE whore, Germ. Hure are related to Goth. hōrs, Lat.
cārus ‘dear (one)’. Other examples are: wench, originally ‘girl, young woman’;
214 Semantic change

mistress ‘a kept woman’, originally a femininized form of mister which itself

reflects earlier master; madam ‘proprietress of a brothel’, originally the same
as the respectful term of address madam(e), variant ma’am, from Fr. madame
‘my lady’. Other words for ‘loose woman’, many of them of uncertain origin,
are harlot, slattern, strumpet, trollop. Note also Fr. putain, Sp. puta ‘prostitute’
from PRom. *pūtta ‘girl’; NHG Dirne ‘prostitute’ from OHG thiorna ‘maid(en)’;
Avestan ǰahi(kā)- ‘woman’ → ‘prostitute’; Hindi raṇḍī ‘widow’ → ‘prostitute’;
Telugu tottu ‘female servant, slave’ → ‘prostitute’; Arab. xalīlā ‘dear (fem).’ →
‘prostitute’.
Significantly, there are few counterparts for men. In fact, the notion ‘loose
man’ strikes many people as anomalous, if not ridiculous, and words like stud,
which come close to denoting something of this sort, tend to be used even with
some degree of admiration. A rare example of a male-reference word with nega-
tive connotations is pimp, but even this word is used generally about men who
are connected with females of sexually “questionable character”. The word pimp
is generally considered to be of uncertain origin. However, the Oxford English
Dictionary (OED) gives a phonetically very similar form pimping (adj.), meaning
‘small, trifling, insignificant, … petty, mean, sickly’. The OED compares this word
to Dutch pimpel ‘weak little man’, Germ. pimpelig ‘effeminate, sickly, puling’,
which, it adds, “implies a stem pimp”. In addition, there is a German form Pimpf
‘young male’ which looks remarkably similar (although the difference between
initial [p] and final [pf] causes difficulties). Perhaps these similarities are simply
accidental; but if they are not, then it is possible that Engl. pimp originally meant
‘young boy’, and that its current meaning came about through pejoration, via
an intermediate meaning ‘weak, effeminate man’. Even though etymology might
favor an original meaning of weakness, pimps of course are anything but weak in
the way they hold power over their women.
Similar examples of pejoration in terms referring to young, innocent persons,
young males, and females of all ages are found in language after language and,
apparently, at all stages of human language.
Developments of this type show a rather unpleasant side of the human char-
acter, which glorifies strength and power and holds in contempt the weak, the
gentle, and the female. Occasionally, however, we get a glimpse of the underdog’s
getting even, such as Engl. surly, originally sirly ‘lordly, masterful’, hence ‘impe-
rious, domineering’, and then ‘rude, uncivil’.
At the same time, words designating warriors and rulers, such as OFr. mare-
chal : Engl. marshal, Engl. knight, Gk. presbúteros, Engl. elder, tend to acquire
positive connotations, as we have seen before. Note also Fr. Charles, Lat. Carolus,
Germ. Karl, South Slav. kral. This was especially the case in medieval and early
modern society, when the lords of the manor considered themselves greatly supe-
 The effects of semantic change 215

rior to the common people and reserved for themselves the right to hunt and to
indulge in elegant living. It is this context that may have given rise to Ø-plurals
for animals that are hunted, such as deer and fowl, and to cultivating such witty
expressions as an exaltation of larks, a pride of lions, a pack of wolves. (Chapter 5,
§ 2.1.) Here, too, French borrowings such as beef, veal, mutton, venison, as well
as dine acquired their more elegant, lofty connotations compared with the corre-
sponding Anglo-Saxon terms cow, calf, sheep, deer, and eat. (See Chapter 8.) And
this is the context as well for the special development of Engl. hound, the cognate
of Germ. Hund, Fr. chien, Ital. cane (§§ 4 and 5.3 above), to designate not just any
dog, but a dog used for hunting.
Sometimes we can see both pejoration and melioration in succession, as
social factors cause words to change from one sphere to the other. For instance,
Old French had a word ber/barun ‘man’ which, like churl above, could be used to
refer to ‘common men’ or even ‘servants’. As these became ‘servants’ and ‘vassals’
of the king, their status became elevated to king’s barons, and eventually they
could be The Great Barons who were members of the Great Council, the House
of Lords. But in the right context, there is once again pejoration, in the phrase
robber baron.

6.2. Sporadic vs. systematic effects: Because of its inherently fuzzy nature,
meaning can be expected to change in a fuzzy, non-systematic manner. Most of
the changes examined so far clearly are sporadic, as expected, generally affecting
individual words. For instance, while some tabooed words undergo deformation,
others are replaced by euphemisms. Yet others remain unaffected themselves,
but induce replacements of innocent homonyms. Metaphor generally affects indi-
vidual words. Sound symbolism operates to change Engl. tiny to teeny, little to
leetle, but fails to affect small, whose rounded back vowel does not conform to
the correlation “high vowel : small”. Note the similar difficulty with Engl. big,
whose high front vowel is in conflict with the expected correlation “non-front
vowel : big, large”. Moreover, the direction of semantic change may differ in con-
temporary varieties of the same language, as in the case of table ‘put on the table
for immediate discussion’ (Brit. Engl.) vs. ‘shelve’ (Am. Engl.); see Chapter 1.
More systematic, sweeping developments are observable in the medieval and
early modern developments of terms associated with war, nobility, and the activi-
ties of the nobility. Here, whole semantically definable fields of words underwent
similar meliorative developments. But as we have seen, at roughly the same time
that terms such as knight undergo melioration, knave develops negative conno-
tations. And the ancestor of Mod. Engl. baron successively underwent both pejo-
ration and melioration. Although more sweeping than other semantic changes,
changes in semantic fields, thus, are far from fully systematic.
216 Semantic change

This does not mean that there are no systematic semantic changes at all. But
such changes tend to be restricted to fairly narrowly confined and more or less
self-contained subparts of the lexicon or to lexical items whose use is intimately
tied up with linguistic structure.

6.2.1. Cardinal-point systems. An example of a self-contained subpart of the

lexicon is presented by the words for the cardinal points or directions in early
Indo-European: East (E), west (W), north (N), and south (S). Two major tendencies
for naming the cardinal points can be observed. One refers to the rising sun or
dawn to designate E, the midday sun for S, and the setting sun or evening for W;
see (30) below. In this system, terms for N are derived from various sources, such
as Gk. boréas named after a northerly wind, or Lat. septentrio from the constella-
tion also referred to as ursa minor ‘the Little Dipper’, which contains Polaris, the
northern beacon in the night sky.

(30) a. East: Source

Gk. anatolḗ (hence Anatolia) anatéllō ‘rise’
Lat. oriens (hence Orient) orīri ‘to rise’
Ital. levante (hence Levant) levare ‘to rise’
OCS vŭstokŭ vŭstekǫ ‘rise’
(hence Vladivostok ‘ruler
of the east’)
Gk. héōs
Gmc. aust- (hence Engl. east) PIE *(a)wes- ‘dawn’
Avest. ušastara-
b. South:
Gk. mesēmbría meso- ‘mid’ + (h)ēméra ‘day’
Lat. meridies medio- ‘mid’ + dies ‘day’
Gmc. sunþa- (Engl. south) sunnō ‘sun’ (‘towards the sun’)
c. West:
Lat. occidens occido ‘fall down, sink’
Ital. ponente porre ‘put, set down’
Gmc. west- PIE *wes- ‘stay, spend the night’ (?)
OCS zapadŭ za-padǫ ‘fall down, sink’
Gk. hespéra = ‘evening’
Lith. vakarai plural of vakaras ‘evening’
Avest. daoštara- daoš- ‘evening, dark’

Beside this relatively loose system of designations, another, more systematic one
is found. The system is fully operative in Sanskrit and Old Irish, but traces are
 The effects of semantic change 217

found in Germanic and Welsh words for N and/or S. In this system, orientation is
strictly to the east, the orient (i. e. the rising sun). E, therefore, is called ‘forward’
or ‘in front’, and the names for the other cardinal points are ‘left’ = N, ‘right’ = S,
and ‘back, behind’ = W. Compare (31). Note incidentally that the Sanskrit word
for ‘left’, uttara-, is a euphemism which in some ways tries to compensate for the
widespread prejudice against left-handers. Its original meaning is ‘upper’, hence
‘better’. Such euphemisms are not unusual for the notion ‘left’; compare Gk. aris-
terós (lit. ‘better’) and euṓnumos (lit. ‘well-named’), Lat. sinister (lit. ‘older’, hence
‘better’), or OEngl. winstre (lit. ‘friendlier’).

(31) a. Sanskrit:
E: prāñč-; pūrva- lit. ‘directed forward; first’
N: uttara- lit. ‘left’
S: dakṣiṇa- lit. ‘right’
W: pratīča-/paśčima- lit. ‘(directed to) behind’
b. Old Irish:
E: airther lit. ‘directed forward’
N: tuascert lit. ‘left direction’
S: descert lit. ‘right direction’
W: iarthar lit. ‘directed to behind’
c. Other languages:
N: Gmc. norþ- Compare Osc.-Umbr. nertro- ‘left’
Welsh gogledd Compare cledd ‘left’
S: Welsh deheu lit. ‘right (hand)’

What is significant for present purposes is that one early Indo-European language,
Avestan, systematically shifted the system in (31) clockwise by one point, so that
‘forward’ became S, ‘behind’ N, and ‘right’ W; see (32). The remaining term, the
one for E, should be ‘left’, but no unambiguous examples are attested, perhaps
by accident. (There is, to be sure, a vātō uparō ‘east wind’, whose uparō literally
means ‘upper’ and thus could be compared to Skt. uttara- ‘upper’ → ‘better’ →
‘left’ (see above); but a more literal interpretation ‘wind from the up-country’ has
also been proposed.)

(32) Avestan:
S: pauruua- lit. ‘directed forward; first’
N: apāxtara- lit. ‘directed to behind’
W: dašina- lit. ‘right’

The reasons for this shift in orientation may lie in the fact that the Zoroastrian reli-
gion of the Avestan texts presents a deliberate break with the earlier Indo-Iranian
218 Semantic change

tradition (as represented by Sanskrit). But this is mere speculation, for it is not at
all clear why the general break in religious tradition should have brought about
the specific break in terminology for the cardinal points.
A similar systematic shift in orientation is suggested by comparison of various
Afro-Asiatic languages; see the roots and forms in (33). In this case, perhaps, the
southern orientation of Egyptian can be attributed to the overwhelming signifi-
cance of the river Nile and the fact that it runs from south to north. Compare for
instance the root √ḫdy ‘go downstream, go north’; and note that √ḫntw ‘in front;
south’ can also mean ‘upstream’. But again, that is sheer speculation. Moreover,
in Hausa, which like Semitic, Berber, and Ancient Egyptian is a member of the
Afro-Asiatic family, the root corresponding to Sem./Egypt. √ymn means ‘west’,
just as in Egyptian.

(33) Semitic Berber Ancient Egyptian

S: √ḫntw lit. ‘in front’
N: √šml (Arab.) E: √yʔb lit. or alternative
meaning ‘left’
S: √ymn W: √ymn lit. ‘right’
W: ḍǝffǝr Compare Sem. √dbr ‘be
behind’

Interestingly, the religious significance of Mecca in Islam may have been respon-
sible for a similar shift. The root √qbl ‘facing, in front’ acquired the meaning ‘the
direction faced when praying’. A further reinterpretation as ‘south’ must have
taken place in areas where Mecca lies to the south.
Altaic furnishes further evidence for a pattern of naming the cardinal points
in terms of a southern orientation: Mongol bara-gun ‘right’, Kalmük ömnö ‘in
front’, and Mongol aru ‘back’ also mean ‘west’, ‘south’, and ‘north’, respectively.
The evidence of Altaic may be significant, since here we do not have an alternative
eastern orientation. This fact may suggest that southern-orientation systems need
not always be considered secondary realignments of original eastern-orientation
systems. In Indo-European, however, eastern orientation is pervasive and south-
ern orientation limited to Avestan. Under the circumstances it is more likely that,
eastern orientation is original in this language family, and southern orientation
just an areally restricted innovation.
The northern orientation of modern times reflects a later perspective, in
which reference to the magnetic north pole became the basis for navigation. This
northern orientation has given rise to completely new uses of ‘left’ and ‘right’ as
referring to ‘west’ and ‘east’, respectively. Moreover, speakers have introduced the
terms ‘up’ and ‘down’ as referring to ‘north’ and ‘south’, in reference to position
 The effects of semantic change 219

on a map or direction, as when a New Yorker says he is driving up to Boston but

down to Philadelphia; there is also the further change of north and south meaning
‘better’ and ‘worse’, respectively, the latter in, e. g., the expression go south for
‘become bad; go wrong’.

6.2.2. Syntax and systematic semantic shifts. In the discussion of ellipsis

(Chapter 5, § 3.2) we have already seen one example of how semantic change,
interacting with other changes, can have a systematic effect on linguistic struc-
ture, both on morphology and on syntax. That is the case of Engl. ne … nāwiht →
not, involving the hyperbolic use of nāwiht ‘nothing’ as reinforcement of the nega-
tion ne, fading of the hyperbolic connotations of nāwiht > not, and ellipsis of the
original negation ne. Here, the systematicity of the change is due to the fact that
negation is not just a semantic, or even morphological, phenomenon, but directly
interacts with syntax, which by its very nature is highly systematic.
Even more remarkably systematic effects can be found when a semantically
well-defined subsystem of the lexicon interacts with morphology and syntax.
This frequently is the case in the development of pronouns, especially when the
semantic trigger is politeness.
In ordinary usage, politeness is more or less the same as being considerate to
your fellow human beings, not hurting their feelings, saying nice things to them,
and so on. In linguistics, the term is often used in a more restricted sense, which
might be defined as showing appropriate deference or intimacy through highly
conventionalized means. In present-day English, politeness in this sense is just
about limited to the choice of address forms, such as sir or ma’am, the use or
non-use of titles, and the use of first or last names.
Earlier English, from about the thirteenth century till about 1700, had an addi-
tional means for expressing politeness, namely a choice between singular and
plural forms of the second person pronoun in reference to singular addressees.
The use of plural forms usually indicated a higher degree of politeness or defer-
ence, while singular forms were used either with subordinates or as signs of great
intimacy. Compare the examples in (34) from Chaucer’s Book of the Duchess. In
(34a) Juno, Queen of the Gods, addresses her messenger, a clear subordinate, and
uses the singular form of the pronoun, thou; in (34b) Morpheus, God of Sleep, uses
a form of the plural pronoun you to address his wife, Alcione, a Goddess herself.
For speakers of Modern English who are not familiar with such pronoun usage,
it should perhaps be added that second plural forms sometimes are used with
subordinates and second singular forms with persons who should normally be
addressed in the plural. Such violations of the norm convey special connotations
of irony, opprobrium, or insult, very similar to, say, a sudden switch from first
name to last name (or vice versa) in Modern English.
220 Semantic change

(34) a. “Go bet,” quod Juno, “to Morpheus

Thou knowest hym wel …”
‘“Go quickly,” said Juno, “to Morpheus – you know him well.”’
b. Ye shul me never on lyve yse
‘You shall never see me alive (again).’

Although this earlier English pronoun usage may have arisen under French influ-
ence, it ultimately reflects a semantic tendency found in many other languages,
namely to associate plurality with greater importance or “weight”. Modern
English preserves a trace of this in the so-called royal or editorial we.
If this were all, earlier English would differ from the modern language merely
by having an additional lexical category for indicating politeness. However, when
used as subjects, the singular and plural second person pronouns controlled dif-
ferent agreement markers on the verb. Compare (34’), where the singular of (34a)
is replaced by the plural, and the plural of (34b) by the singular. In (34’a), the verb
agreeing with plural ye appears without the second singular ending -st, in (34’b)
the singular form thou requires the verb to have the second singular ending -t. This
shows that the semantically determined choice of singular vs. plural has direct
repercussions in the syntax.

(34’) a. Ye know-Ø hym wel …

b. Thou shul-t me never on lyve yse

As time progressed, the members of the upper crust of English society increas-
ingly used only the plural pronoun, together with plural verb agreement – not so
much as a sign of deference or politeness in the usual sense, but as an indication
of their own refinement, as a sign of politeness to themselves and to their class, as
it were. Use of the old singular structures, by contrast, came to be reinterpreted as
a sign of lack of refinement, even boorishness. The behavior – and prejudices – of
the upper crust soon came to be imitated by the burgeoning bourgeoisie. And in
order not to be considered boors, the burghers did the same thing as the barons.
They increasingly gave up the use of thou in favor of you. Eventually, the use of
you instead of thou was adopted by the lower classes as well. The Quakers were
the only major source of resistance to these developments, and into the twentieth
century they insisted on retaining the use of the old singular pronoun, in the form
thee. Elsewhere, forms of the pronoun thou disappeared from the standard spoken
language, surviving only in fossilized form in religious contexts.
Interestingly, and perhaps ironically, because of its restriction to religious use,
thou has undergone a significant reversal in connotations. Originally the second
singular pronoun was used, as in most other languages with similar politeness
 The effects of semantic change 221

conventions, to signal the same kind of intimacy between God and worshiper as
the use of the word father in the Lord’s Prayer. Its modern restriction to the reli-
gious sphere invites a very different evaluation of thou, as a symbol of the very
special and deep reverence that human beings owe to the Lord. This re-evaluation
is no doubt one of the reasons that many forms of English-speaking Christianity
have begun changing from thou to the more familiar and intimate you in their
Bible translations and liturgical texts.
The semantically driven developments outlined above had their own lexical,
morphological, and syntactic effects. For one thing, the second person singular
disappeared from the lexicon of ordinary Standard English. Or rather, the dis-
tinction between second singular and plural pronouns disappeared. Further, the
second singular verb endings disappeared as well. Finally, as a consequence, the
syntax of English no longer requires a syntactic agreement distinction between
second singular and plural. In fact, English now has just one verbal agreement
marker, the third person singular ending -s (as in he know-s) which, moreover, is
limited to the present tense.
Some linguists have argued that the loss of the singular : plural distinction
in the second person was a great gain, not only because it simplified morphol-
ogy and syntax, but also because it made English speakers more democratic than
the speakers of most European languages who still use pronoun differences to
pay – or withhold – respect. But English speakers are able to do the same thing
by using different address forms. Not calling an officer sir or ma’am in the mili-
tary can have as disastrous consequences as using the “familiar” form of address,
the second singular, in earlier English or in most European languages. And con-
versely, calling your buddy sir or ma’am may be as inappropriate, ironic, or even
insulting, as the use of the “polite” second plural (or some other polite pronoun
form) in earlier English and many other languages. Moreover, the loss of the sin-
gular : plural distinction in the second person has clearly been felt to be a draw-
back by many English speakers. Otherwise, there wouldn’t be so many different
attempts at restoring the distinction by creating special pluralized forms such as
y’all [yɔl], you’uns [yɨnz] (< you ones), or yous(e) [yūz], not to mention the wide-
spread colloquial you guys.
Even so, English is not alone in having lost the singular : plural distinction in
the second person. Many varieties of Latin American Spanish have similarly gen-
eralized the old plural pronoun vos at the expense of the old singular tu, through
strikingly similar developments.
On the other hand, some languages expand the pronominal system to express
an even greater range of politeness distinctions. Thus, in Modern Hindi the old
second singular pronoun tū indicates either great intimacy or great rudeness,
depending on the social context; the old plural tum is the ordinary, unmarked
222 Semantic change

form of second-person address; and a new pronoun āp is used to indicate defer-

ence or social distance. And as in Middle and early Modern English, each of the
different pronouns controls a different kind of verb agreement.

7 Conclusion
Examples like the ones just discussed show that, given the right circumstances,
semantic change can have very sweeping and systematic effects, even on linguis-
tic structure. This should not, however, distract from the fact that in the majority
of cases semantic change is as fuzzy, self-contradictory, and difficult to predict as
lexical semantics itself.
At the same time, it must be admitted that semantic change can have pro-
found effects on the lexicon and is, in fact, so intimately tied to the lexicon that
some historical linguists subsume it under the heading of lexical change. In the
following two chapters we take a closer look at other processes that bring about
lexical change.
Chapter 8: Lexical borrowing
And I must borrow every changing shape
To find expression
(T. S. Eliot, Portrait of a Lady)

Neither a borrower, nor a lender be

…
This above all: To thine own self be true
(Shakespeare, Hamlet)

1 Introduction
Languages and dialects normally do not exist in a vacuum. They – or more accu-
rately, their speakers – always have some contact with other languages or dialects.
The degree of contact may vary considerably. It may involve the whole range of
language use, from informal, spoken to highly formal, written; or it may remain
confined to just one level of use, such as written discourse.
A very common result of linguistic contact is lexical borrowing, the adop-
tion of individual words or even of large sets of vocabulary items from another
language or dialect. Examples of such borrowings, or loans abound in English,
such as rouge (from French), macho (from Spanish), yen ‘craving’ (from Chinese),
or schwa (from Hebrew via German).
Generally, such “borrowed” items are not returned, nor is there any intent
to return them at the time of borrowing. In this regard, then, the terms theft or
embezzlement would be more appropriate, but they sound less genteel. Besides,
the donor language does not actually lose the borrowed word.
Such semantic quibbles aside, what is important is that like all other linguis-
tic terminology, terms such as borrowing, loan, and donor are used with special,
technical connotations. The connotations of such terms are bound to differ from
those found in the real world, no matter what terms we use.
Although in the act of borrowing there is no intention to return the borrowed
word, occasionally, by sheer coincidence, words do get returned, or are stolen
back. Consider for instance the English words redingote (a long, open, light-
weight coat without lining) and contredanse (a type of musical composition), or
the French sport ‘sport’. The first two words were taken from French which, in
turn, had borrowed them from English earlier; see (1a,b). Conversely, sport came
into French (and many other languages) from English which, in turn, had taken
the ancestor of the word from French; see (1c). We may even encounter something

https://doi.org/10.1515/9783110613285-008
224 Lexical borrowing

like mistaken returns, not to the original donor, but to a language closely related
to it. Compare the words in (2) which had been taken from Old Frankish, the lan-
guage of the Germanic overlords of Romance Gaul (see Chapter 2, § 3.3), and were
then passed on to English, another Germanic language, when England came to
be ruled by the French-speaking Normans. Interestingly, these words coexist in
English with native words which are still similar to the words that French had
borrowed; compare wise (as in in no wise) and ward.

(1) a. Engl. riding coat ⇒ Fr. redingote ⇒ Engl. redingote

b. Engl. country dance ⇒ Fr. contredanse (with folk etymology: contre
‘counter, opposite’ + danse ‘dance’) ⇒ Engl. contredanse
c. OFr. de(s)porter ‘divert; amuse’ ⇒ MEngl. disporten ‘divert, amuse
(oneself)’ → ‘play, frolic’ → NEngl. sport ⇒ Fr. sport

(2) a. OFrank. *wīsa ⇒ OFr. guise ⇒ Engl. guise

b. OFrank. *wardōn ⇒ OFr. guarder ⇒ Engl. guard

In some cases, words spread over vast territories through a chain of borrow-
ings. Such words are referred to as Wanderwörter (a borrowing from German,
meaning ‘migrating words’). Words for cultural items or concepts are especially
apt to become widely dispersed. Compare the examples in (3).

(3) a. Skt. śarkara- ‘sand, grit; sugar in granulated form’ ⇒ Pers. shakar ⇒
Arab. sukkar ⇒ (O)Ital. zucchero, OSpan. azúcar, OFr. sucre ⇒ Engl.
sugar; compare Germ. Zucker ⇐ Ital. zucchero, as well as Medieval Greek
sákkharon ‘sugar’, the source for Engl. saccharin.
b. Skt. khaṇḍa- ‘broken piece; sugar in large pieces, rock sugar’ ⇒ Pers.
qand ⇒ Arab. qandi ⇒ OIt. zucchero candi, OFr. sucre candi ⇒ Engl.
sugar candy (hence by further developments, candy); compare Germ.
Kandis(zucker).
c. Lat. centēnārius ‘a hundredweight’ ⇒ Gk. kentēnárion ⇒ Aramaic
qintinārā > qintārā ⇒ Arab. qintar ⇒ Medieval Lat. quintāle (a “returnee”
word) ⇒ OFr. quintal ⇒ Engl. quintal.
 The substance of borrowing 225

2 The substance of borrowing

The first thing that comes to mind when we think of borrowing is the adoption of
individual lexical items, such as rouge, macho, realpolitik, or one of our favorite
linguistic terms, umlaut. However, through vocabulary borrowing other linguistic
elements may be acquired.
For instance, extensive vocabulary borrowing can introduce new morphol-
ogy (see Chapter 5, § 4). Heavy borrowing from French, as well as Latin and Greek
(often via French), has introduced into English words like the ones in (4). Many
of these coexist with other borrowed words from which they can be considered
derived, see (5). They can therefore be analyzed as containing these basic words
plus the suffixes -able/ible, -ation/tion, -ance/ence. As a result, English has
acquired a considerable amount of originally foreign derivational morphology –
in addition to the words that it borrowed. Some of the new elements may combine
only with other borrowed elements, such as the -duct, -ceive of (6), words coined
in English from borrowed elements. Others can freely combine with non-bor-
rowed, native elements, such as the -able of (7). One of the most commonly used
derivational suffixes of English, the agentive suffix -er of words like singer, baker,
or transceiver, ultimately is a loan from (late) Latin -ārius ⇒ OE -ere (as in OE
leornere ‘learner; disciple’).

(4) equatable, legible, potable

derivation, deliberation, equation
deliverance, occurrence

(5) equate
derive
deliver

(6) trans-duct, trans-ceiv-er

(7) read-able, laugh-able, drink-able

Vocabulary borrowing also can introduce new sounds, or new contexts for old
sounds. The latter, more common development is observed in words like rouge,
prestige, garage with [ž] in word-final position. In more established English words,
[ž] is limited to medial position, as in measure, leisure. And the relative foreign-
ness of final [ž] is responsible for the common substitution of [ǰ], especially in less
prestigious words like garage. (In British English, the pronunciation [gǽriǰ] has
become more or less standard.)
226 Lexical borrowing

The introduction of a new sound is found for instance in the pronunciation of

the composer’s name Bach as [bax] by English-speaking aficionados of Baroque
music. The same sound has been introduced in the New York English expres-
sion yecch [yex] which seems to be of Yiddish origin. Speakers familiar with Scots
English, which has preserved the [-x] of earlier English, may also affect the sound
in the word loch (as in Loch Ness).
In addition to individual lexical items, languages may adopt combinations or
collocations of words, such as Engl. court martial, which comes from French.
Such collocations, in turn, have introduced into English a new manner of plural
formation. In native English collocations, the plural suffix -s normally attaches
to the last word, as in (8a). In French expressions of the type (8b), plural -s is
attached to the “head noun” of the construction, court, not the following adjec-
tive, martial. From the perspective of prescriptive grammar, this is the correct
pattern in English, too, even though many speakers may no longer understand
that martial is an adjective, and that court martial really means ‘martial (i. e., mil-
itary) court’. For these speakers, then, the structure in (8b) is anomalous, and they
tend to replace it with the regularized pattern in (8c), with the plural -s following
the last word of the constituent just as in (8a). For speakers who use the pattern
(8b), on the other hand, the borrowing has introduced a new syntactic pattern of
noun + adjective which is not generally found otherwise in English.

(8) a. parade marshal-s

b. court-s martial
c. court martial-s

In other cases the influence may be much greater. For instance, the French mode
of forming comparatives by means of plus ‘more’ plus the simple adjective (9a)
has given rise to the pattern (9b), with Engl. more substituting for Fr. plus. And
this pattern came to coexist with the native pattern (9c). The competition between
these two different modes of comparative formation eventually was resolved such
that monosyllabic adjectives (generally) take the inherited comparative in -er, as
do disyllabic ones in -y. Many speakers also have this pattern in disyllabic adjec-
tives in -er; but for some this is only optional. Adjectives which do not qualify for
taking -er make their comparatives by means of more. Compare example (10).

(9) a. beau : plus beau

b. beautiful : more beautiful
c. long : long-er
 The substance of borrowing 227

(10) Monosyllabic: long : long-er

Disyllabic: pretty : pretti-er
clever : clever-er/more clever
handsome : more handsome1
Polysyllabic: beautiful : more beautiful
(1 Some English speakers can say handsomer. There are further variations, especially in
nonstandard varieties of English; note also Lewis Carroll’s curiouser and curiouser. On
the other hand, some monosyllabic adjectives that are rarely used in the comparative
may prefer the pattern with more, such as more vague, more just.)

Examples like these may suggest that anything can be borrowed: lexical items,
roots and affixes, sounds, collocations, and grammatical processes. To some
extent this impression is well justified. Still, there are some differences.
From a purely linguistic perspective, the most important fact is that differ-
ent spheres of the vocabulary are borrowed more easily, others significantly less
easily. The most successful resistance to borrowing is offered by basic vocabu-
lary, words referring to the most essential human activities, needs, etc., such as
eat, sleep; moon, rain; do, have, be, or function words essential in syntax, such as
the demonstrative pronouns this and that, the definite article the, or conjunctions
like and, or, if, and when. In English, this is evident from the fact that in spite of
its pervasive and domineering influence, French contributed virtually nothing to
the most basic vocabulary. The only exception is the -cause of because. But note
that the initial be- of the word is solidly Anglo-Saxon, derived from earlier bi ‘by’.
The word because, thus, is not a borrowing, but was put together in English, by
combining the native prefix be- with the borrowed word cause ‘reason, etc.’ Words
of this type sometimes are referred to as hybrids, and are indeed, from an etymo-
logical standpoint, but not for the earlier English speakers who coined because
with lexical material available to them, whatever the source.
Although verbs may be borrowed, they are not as readily borrowed as nouns.
And if the need for borrowing does arise, many languages instead borrow a
nominal form of the verb and employ a native all-purpose verb such as do or make
as a means of turning that form into the equivalent of a verb. See the examples
in (11). The reason for this particular resistance probably lies in the fact that it is
easier to ask questions like “What do you call this (thing)?” than something like
“What is the verb you use to designate that somebody is doing this/acting in this
way?” Eventually, English expressions of the type (11) came to be used without the
verb ‘to do’, on the model of correspondences like (12).
228 Lexical borrowing

(11) Latin Early Modern English borrowing

Verb Verbal Adject.
imitāre imitātum do imitate
speculāre speculātum do speculate
corrigere correctum do correct

(12) they do go : they go

they do imitate : X

The relative resistance of verbs and especially of basic vocabulary does not mean
that they are totally impervious to borrowing. Under the right social circum-
stances (see § 5 below) both types of lexical items can be borrowed. For instance,
English borrowed the basic-vocabulary pronouns they, their, them from the lan-
guage of the so-called Danes. (On the identity of the Danes and their relationship
to the Anglo-Saxons see Chapter 2, § 3.3.) The same “Danish” language also was
the source for the fairly basic English verbs give and take. Moreover, English bor-
rowed a considerable number of not-so-basic verbs from French, such as perceive,
receive, and derive.
The most easily borrowed words belong to more specialized forms of dis-
course, often referring to technology or other phenomena that require a good
deal of mental and linguistic abstraction. Compare words like nation, inflation,
machine, engine, atom, finance, all of which are borrowings.
Other words, too, are commonly borrowed, especially the names for new arti-
facts and other cultural items which are subject to frequent change. Here belong
words such as telephone (made up of the borrowed components tele- ‘far’ and
phone ‘speech’, both from Greek) and lac/lacquer (ultimately from Hindi lākh or
its cognate in some other Modern Indo-Aryan language).
Borrowing of technological vocabulary is not just a modern phenomenon. For
instance, the ancient Gauls of what is now France had a highly developed tech-
nology in metallurgy; and the Germanic words for ‘iron’, such as OE isern (Mod.
Engl iron), OHG isarn, were borrowed from Gaul. isarno, a native Celtic word found
also in Irish iarn.

3 N
 ativization, or how do you deal with a word
once you have borrowed it?
The major difficulty with borrowing from a foreign language is that languages
may diverge considerably in their phonology. Thus, r is generally pronounced as
 Nativization, or how do you deal with a word once you have borrowed it? 229

a uvular fricative [ʁ] in Modern Standard French. Not having this sound in their
own language, English speakers find it difficult to articulate the sound in words
like rouge. Even those who have learned French and in the process have acquired
the pronunciation when speaking French usually have difficulties in maintaining
the pronunciation when speaking English. The problem basically is this. In order
to speak English, we have to “configure” our articulatory organs – and the neu-
rological processes that control them – for English, unless we want to speak with
a foreign accent. If, then, a word like rouge comes up in an English context, such
as She put on rouge, we can affect the French pronunciation only by reconfiguring
for French. That, however, is not only difficult and inconvenient, normally it also
brings about a noticeable and undesirable break in the utterance. Perhaps even
more important, our listeners may feel that we are putting on airs. To avoid all
these difficulties, we have to do what most English speakers do, we have to pro-
nounce the r as an English [r].
There are many other adjustments, beside changes in pronunciation, that
tend to accompany borrowing. What is common to all of them is that they nativ-
ize the borrowing by integrating it more firmly into the linguistic structure of the
borrowing language.
The most important nativization processes clearly involve phonology. Even
if we do nothing else, we have to make the borrowed word pronounceable in our
language.
When faced with a foreign sound that does not exist in our own language,
we think that the most natural thing to do is substitute the most similar native
sound. In principle, this usually is what happens. However in many cases it is
difficult to determine which sound is most similar.
The problem is that similarity or lack thereof comes in many different shades.
For instance, a voiced French sibilant (as in zéro) is very similar to a voiced English
sibilant (as in zero), even though the French sound may be more fully voiced than
its English counterpart. Under the circumstances, substituting anything but a
voiced sibilant would be perverse.
A slightly more complicated example is the English substitution of [k] for
foreign [x], as in the usual English pronunciation of Bach as [bak]. Here the pho-
netic difference between donor language and borrowing language is considerably
greater, and English (outside of Scots English) simply has no sound that would
closely match the foreign sound. Still, the substitution of [k] for [x] makes sense,
since both sounds share the fact that they are velar and voiceless. A substitution
of voiced velar [g] would make much less sense; and substitutions such as [p] and
[b] would be preposterous.
The situation often is much more complex. English is one of only a few Euro-
pean languages that have the voiceless dental fricative [θ]. When words with
230 Lexical borrowing

English [θ] are borrowed, such as the word thriller, there is a great amount of vari-
ation in the nativization of [θ]. It comes out as [s] in standard French and German,
but as [t] in many other European languages, including many nonstandard vari-
eties of French and German. These different choices cannot be fully explained by
the notion “most similar sound”. It is difficult to see how in standard German or
French [s] is more similar to [θ] than [t], while in other forms of speech, [t] is more
similar. Rather, it appears that [θ] is in some ways equally similar – and dissim-
ilar – to both [t] and [s]. Sibilants like [s] are super-fricatives which differ from
ordinary fricatives by having extra, “sibilant”, friction. The simple fricative [θ]
therefore can be considered to take an intermediate position between non-frica-
tive [t] and super-fricative [s]. Under the circumstances, the choice between [t] and
[s] is arbitrary; and the fact that different languages opt for one or the other sub-
stitution seems to result from something like conventionalization. In fact, some
German speakers use neither [s] nor [t], but [f] to nativize [θ], presumably because
it is acoustically closer to [θ] than either [s] or [t]. (Russian similarly substituted [f]
for Byzantine Greek [θ] in words like Fyodor ⇐ Gk. Theódōros [θ-].)
At times it is not just one sound which is substituted, but rather a combina-
tion of sounds which together can be said to be most similar to the foreign sound.
Thus, Fr. salon is borrowed as Engl. [sǝlɔn]. What motivates this development
is the following. The French word contains [ɔ̃ ] (written on), a single nasal vowel
that is absent in English. The nativization as corresponding oral vowel [ɔ] plus n
manages to “factor out” the vowel and nasal features of the French sound in terms
of permissible English sounds. Here again, there is some element of arbitrariness
in the selection of [n] to encode French nasality. It must be recognized too that
the spelling of the word could have played a role for English. Not so for German,
however: nonstandard German uses the velar nasal [ŋ] for the same purposes, as
in [zalɔŋ], whereas Standard German has adopted the nasal vowel from French,
as in [zalɔ̃ ].
Another example of such a process of factoring out the features of a non-native
single sound is the Middle English substitution of [iu] for Fr. [ü]; see example (13a).
Here, the frontness of Fr. [ü] is rendered by the front vowel [i], and its rounding by
the round vowel [u]. This substitution had important consequences for English.
In many varieties of English [iu] became [yū]; and the [y] of this new pronuncia-
tion triggered a subsequent process of palatalization. It is this development which
accounts for phonetic correspondences like Fr. mesure with [zü] : Engl. measure
with [žǝ] from earlier [zyū]. Similar examples can be found elsewhere, see (13b).

(13) a. Fr. pur [pür] ⇒ ME [piur] ‘pure’

b. Turk. göl ‘lake’ ⇒ Bulgar. gyol ‘puddle’
 Nativization, or how do you deal with a word once you have borrowed it? 231

In modern literate languages, nativization frequently takes place through spell-

ing. For instance, the common English pronunciation [rɔθ(s)čayld] for the name
Rothschild [rōtšild/t] is based on the fact that th and ch usually are pronounced
as Engl. [θ] and [č], respectively. Similarly the common English pronunciation of
word like Sanskrit [sǝnskrit] as [sænskrit], with [æ] rather than [ǝ], results from the
fact that the word is transcribed with the letter a, which in this context usually is
pronounced as English [æ]. To forestall such pronunciations, some Indian words
with [ǝ] are spelled with u, as in [pǝnǰāb] : Punjab. But interestingly, the spelling
Punjab has given rise to a pronunciation [punǰab], a hypercorrection based on
the fact that the letter u is commonly pronounced [u] in foreign words. (A similar
phenomenon is observed in the pronunciation [rāžā] for the South Asian word
raja(h) [rāǰā]. This hypercorrection arises from the belief that a foreign word must
have a “foreign” sound, so [ž] is substituted for [ǰ] to make the word sound like
other foreign words, such as prestige. More in § 4 below.)
Even more “exotic” nativization processes can be observed. One of these can
be termed etymological nativization. For instance, till recently literate speakers
of Russian nativized western foreign [h] as [g], as in gospital’ ‘hospital’ and Hans
Henrich Hock’s grandfather was addressed as Gerr Gock by Russian expatriates.
None of the nativization processes so far discussed would account for this sub-
stitution. But once we note that languages to the west such as Ukrainian, which
are closely related to Russian and in intensive contact with it, have changed
Proto-Slavic g to h, an explanation is possible: The relationship between native
Russian words and their Ukrainian (etc.) cognates provides a quasi-analogical
pattern which suggests that foreign h corresponds to native g; see (14). (As it turns
out, this nativization has now been given up, and to his dismay, Hans Henrich
Hock received a Russian visa for Xans Xenrix Xok.)

(14) Ukrainian hospod’ : Ru. gospod’ ‘God, Lord’

etc.
Germ. Hospital : X = Ru. gospital’ ‘hospital’
etc.

Phonological nativization may also be sensitive to phonological structure.

Many languages nativize foreign borrowings to make them conform to native
restrictions on word or syllable structure. For instance, in the northern varieties of
Standard German, all final stops have undergone final devoicing (see Chapter 4,
§ 5.1.1). Foreign words that do not conform to the pattern of final voicelessness,
such as Engl. trend with final voiced [d], are made to conform, through substitu-
tion of the nearest voiceless sounds, as in Trend [trent].
232 Lexical borrowing

Similarly, foreign borrowings in Japanese are consistently reshaped in order

to conform to the syllable structure of Japanese which tolerates only syllables
of the type CV, with just one initial consonant (if any), plus a vowel (or syllabic
nasal), and with no syllable-final consonants. This is commonly achieved through
vowel insertion, as in (15a). But it may be accomplished by other means, such as
the reduction of the initial consonant group and the dropping of the final conso-
nant in (15b), or by turning semivocalic [w] into the corresponding vowel [u], as in
(15c). (The final vowel in kuizu (15c), of course, reflects the same vowel insertion
as in (15a). Example (15a) additionally offers an instance of a nearest sound sub-
stitution, with [r] replacing English [l].)

(15) a. Engl. baseball ⇒ Jap. bēsubōru

crawl ⇒ kurōru (a swimming style)
b. sweater ⇒ sētā
c. quiz [kwiz] ⇒ kuizu

Phonological nativization, thus, can be accomplished by a great variety of devel-

opments, some fairly simple, others quite complex. In some cases, however, none
of these (or yet other) nativization processes are employed. Nativization, if it can
be called that, is accomplished by the simple adoption of a foreign sound or
by adoption of a sound in a context where it does not occur natively. The latter
development, for instance, accounts for the final [ž] of Engl. prestige, rouge, etc.
The former is found in the English pronunciation of Bach with [x], rather than
native English [k].
In many cases, nativization also takes place at the lexical level. And here
again we encounter the possibility of adoption, in this case without morpho-
logical modification. Examples are Engl. rouge, conceive, compassion, sympathy
which, respectively, are from Modern French, Old French, Latin (via French), and
Greek (via Latin and French). Such instances of lexical adoption, however, are
usually accompanied by phonetic or phonological nativization. The need to make
a foreign word pronounceable is much more basic than the need for lexical nativ-
ization.
Interestingly, in some cases of complete adoption, folk etymology secondar-
ily leads to greater nativization, in that it makes better sense of the often opaque
structure of borrowed words. (See Chapter 5, § 3.) English examples are the Algon-
quian animal name otček which by folk etymology has been made more trans-
parent by changing it to woodchuck, and Fr. chaise longue ‘sofa; lit. long chair’
has in many varieties of English been recast to chaise lounge or lounge chair. In
like manner, German has folk-etymologized a borrowing from Lat. arcuballista
‘crossbow; lit. bow-thrower’ as Armbrust lit. ‘arm-chest/breast’. In fact, borrow-
 Nativization, or how do you deal with a word once you have borrowed it? 233

ings are perhaps the most favored target of folk etymology, because their structure
frequently is opaque to the speakers of the borrowing language.
The polar opposite of lexical adoption is represented by loan shifts. These
involve changing the meaning of an existing native word so as to accommodate
the meaning of a foreign word. Put differently, a foreign concept is borrowed only
at the semantic level, without its linguistic form (which is supplied from native
sources) and consequently, no new lexical item is introduced into the borrowing
language.
Examples of this much more subtle and often undetectable process are found
in the semantic shifts which many older Germanic religious terms underwent
in response to the introduction of Christianity through the vehicle of Latin. For
instance, the words heofon ‘sky’, hel ‘underworld’, and god (non-Christian deity)
acquired new Christian meanings beside, or instead of, their earlier native conno-
tations. The semantic shifts were possible because the corresponding Latin terms
had a range of meanings that included both Christian and pre-Christian conno-
tations. The partial semantic agreement between Latin and Old English, then,
made it possible to extend the Old English meanings into new, Christian usages
covered by the Latin terms. As the formulation in (16) shows, developments of
this sort operate on something like a proportional model, similar to the one in the
analogical processes of four-part analogy, backformation, and hypercorrection
(see Chapter 5).

(16) Latin Old English

caelum ‘sky’ : heofon ‘sky’

‘abode of the gods’ : ‘abode of the gods and of
warriors fallen in battle’
‘Christian heaven’ : X

inferna ‘abode of the dead : hel ‘abode of the dead who

have not

(below the earth)’ fallen in battle

(below the earth)

‘Christian hell’ : Y

deus ‘deity’ : god ‘deity’

‘God’ : Z
234 Lexical borrowing

A process intermediate between adoption and loan shift consists of loan transla-
tions or calques. Morphologically complex foreign expressions are translated by
novel combinations of native elements that match the meanings and the structure
of the foreign expressions and their component parts. Compare for instance the
examples in (17). Like loan shifts, these words do not introduce foreign elements
into the language; but they do introduce new forms. Thus in (17b), the English
term world view owes its existence to Germ. Weltanschauung ‘view/outlook on
the world’, of which it is a loan translation. But unlike its occasional rival welt­
anschauung, it is composed entirely of native elements. Calquing was especially
common as an alternative to loan shifts when Christianity was introduced to the
early Germanic peoples; compare (17c).

(17) a. Engl. chain smoker : Germ. Kettenraucher

skyscraper : Wolkenkratzer
Fr. gratte-ciel
b. Germ. Weltanschauung : Engl. world view
c. Gk. ⇒ Lat. ev-angelium : OEngl. gōd-spell (both lit. ‘good
message’)
 > Mod.Engl. gospel

The examples in (17a) further show that the elements used to translate the com-
ponent parts of a foreign word are usually put together according to native mor-
phological patterns and processes. For instance, Engl. chain corresponds to Germ.
Kette and smoker to Raucher. But the German compound is Kette-n-raucher, not
Ketteraucher*, in accordance with a productive German process of compound
formation. Similarly, French translates Engl. skyscraper as gratte-ciel, lit. ‘scrape-
sky’, because that is the productive mode of making compounds that correspond
to the English pattern skyscraper (compare ouvre-porte ‘door opener’, lit. ‘open-
door’).
Germ. Wolkenkratzer, lit. ‘cloud scraper’ or ‘cloud scratcher’, further shows
that calques may occasionally be less than exact translations. In the present case,
the motivation for the inexact translation may be something like taboo. German
does not make a distinction between heaven and sky, but like Old English, uses a
single word, Himmel. A calque Himmelskratzer* thus might have been interpreted
as ‘heaven scraper’, invoking unfortunate associations with the Tower of Babel.
(Speakers of French evidently did not share this concern.)
Because calques look like perfectly “native” words, it is in most cases impossi-
ble to determine the direction of borrowing. Did English calque the German word
Kettenraucher or did German calque the English compound chain smoker? And
what about French fumeur à la chaine? Only where we have “outside” historical
 Nativization, or how do you deal with a word once you have borrowed it? 235

evidence can we tell the likely direction of borrowing. Thus, skyscraper probably
originated in American English when after the Great Fire of 1871, tall buildings,
“scraping the sky”, were erected in the prestige area of Chicago.
It might be added that calquing presupposes a certain familiarity with the
donor language and its grammatical structure. Otherwise, it would not be pos-
sible to recognize that a given item in the donor language is morphologically
complex, or to furnish a translation of the component parts. (This issue becomes
important in § 5.)
Calques are not necessarily limited to structures like (17), in which the compo-
nent parts of morphologically complex expressions are independent lexical items
in their own right. They can involve affixes. Consider for instance (18), where
Latin substitutes its native suffix -us for the -os of Greek, but leaves unchanged
the preceding root, Petr-.

(18) Gk. Pétr-os : Lat. Petr-us (name of the apostle Peter)

The suffix substitution in (18) probably has a special, morphological, motiva-

tion. The suffix -os is well established in Greek morphology, where it goes along
with a genitive singular -ou and other case forms. In Latin, however, the suffix
does not fit into the native inflectional system. At the same time, like Greek, Latin
requires nouns to be inflected for number and case. The substitution of native -us
introduces a suffix which can be inflected (e. g. genitive singular Petr-i) and thus
makes it possible to use the borrowed word in a grammatically correct manner.
Such structural nativization is possible only if the donor and borrowing lan-
guages have similar grammatical systems. Where the systems diverge, different
strategies need to be employed, at least in the case of languages which have fairly
complex morphological systems.
To illustrate the problem and how it is resolved, let us look at the gender
assignment of borrowings in languages which obligatorily have to mark every
noun for gender.
Gender assignment for borrowings seems to operate in terms of the following
parameters: (i) formal criteria; (ii) general semantic criteria; (iii) considerations
of the gender of semantically related native words; (iv) a default class to which
words are assigned if none of the other criteria provides a solution.
Consider the case of German. German has three genders – masculine, fem-
inine, and neuter. In this regard it looks similar to English, with our he, she, it
in the pronouns (though with nothing comparable in nouns). What complicates
matters is that gender assignment is “grammatical”, i. e., to a large extent seman-
tically arbitrary. Thus, ‘tables’ may be masculine, ‘doors’ feminine, and ‘books’
neuter, as in the examples of (19). At the same time, gender assignment is oblig-
236 Lexical borrowing

atory and has important syntactic consequences, since pronouns and adjectives
have to agree in gender with the nouns they refer to; see again the examples in
(19). To make things even worse, there is no consistent formal distinction between
the three genders of German. The best that can be said is that masculines and
neuters tend to inflect alike and in their nominative form tend to end in a conso-
nant, while feminines tend to end in a vowel, most commonly in -e [-ǝ].

(19) Masculine: Das ist ein alter Tisch; er kostet viel Geld.
‘That is an old table. It costs a lot of money.’
Feminine: Das ist eine starke Tür; sie kostet viel Geld.
‘That is a strong door. It costs a lot of money.’
Neuter: Das ist ein gutes Buch; es kostet viel Geld.
‘That is a good book. It costs a lot of money.’

Let us now look at how German assigns gender to words borrowed from other lan-
guages. Most examples are drawn from English which has “natural”, or sex-based
gender, but only for human or animate beings. One example comes from French
which has two genders, masculine and feminine.
The French system is more similar to the German one in that gender is seman-
tically largely arbitrary. However, in terms of its morphology, the grammatical
gender system of French is no more similar to that of German than the natural
gender system of English. German speakers, therefore, tend not to look to the
donor language for guidance in assigning gender but to rely on the criteria out-
lined above. And remarkably, in spite of the fact that the principles are quite
vague and heterogeneous in nature, German speakers show an amazing degree
of agreement in how they apply them. Consider for instance the examples in (20).

(20) a. Early NFr. garage [garažǝ] (m.) ‘garage’

b. Engl. computer
babysitter
trend
rush hour
panel

The French word in (20a) is masculine and should not cause any difficulties if
German gender assignment were based on the system of the donor language; it
should come out as masculine. However, formal criteria are potent enough to tilt
gender assignment in a different direction. French garage and all other French
words in -age come out as feminine in German, as in die Garage. The reason must
be sought in the tendency for German nouns in -e to be feminine.
 Nativization, or how do you deal with a word once you have borrowed it? 237

The fact that German has many borrowings from French ending in -age and
has consistently nativized them as feminines has an interesting consequence.
When Germans learn French, they naturally assume that French words in -age
are feminine. But when they use them as feminines they soon find out – much to
their annoyance – that the French “perversely” use them as masculines, not real-
izing that if there has been perversion it has taken place on the German side. Such
mismatches between apparently cognate native and foreign words are actually
quite common, not only as far as formal issues such as gender are concerned, but
also in meaning. For instance, the convenance of French mariage de convenance
means ‘agreement’, not ‘convenience’, in spite of the English calque marriage of
convenience, which is based on a misunderstanding of convenance. In language
teaching, mismatches of this sort are often referred to as “false friends”.
The problem of gender assignment is of course greatest in borrowings from
languages like English which have natural, not grammatical, gender. Here, too,
German may draw on formal criteria. For instance, the word computer fits the
native German class of instrument nouns in -er, such as Kratz-er ‘scraper’ (as in
the Wolkenkratzer of (17a) above). And since nouns of this type are masculine in
German, computer is assigned masculine gender.
By a similar reasoning we should expect babysitter to be nativized with mas-
culine gender, since German agent nouns in -er are masculine, such as Bäck-er
‘baker’. However, while masculine gender causes no difficulties if this word is
used in a generic sense, it becomes inappropriate if used in reference to a proto-
typical babysitter who, as in English, is female. In that case, German morphology
requires that the form be marked by the feminine suffix -in, as in Bäcker-in, the
female counterpart of Bäcker. But many Germans would balk at using Babysit-
ter-in, presumably because the word Babysitter has not been sufficiently nativized
to accept native derivational suffixes. Germans therefore are in something of a
quandary as to how to use Babysitter and tend to use the word only in the generic
sense, as in Sie arbeitet für uns als Babysitter ‘she works for us as babysitter’, while
avoiding expressions corresponding to Engl. The babysitter just called to say she
is running late.
Gender assignment for Trend is more complex, since semantic criteria gener-
ally work only in words for humans or animates. Moreover, formal considerations
provide only negative guidance. The word ends in a consonant, which suggests
masculine or neuter gender, rather than feminine; but the choice between mas-
culine and neuter is left undetermined. In this case, principle (iii) takes over,
namely consideration of the gender of semantically related native words. Native
near-synonyms which like Trend are monosyllabic and end in consonant are Zug
and Hang; and these are masculine. As a consequence, the word is nativized with
masculine gender.
238 Lexical borrowing

An example like rush hour, which comes out as feminine (die Rush-hour), sug-
gests that when formal and semantic criteria disagree, the latter may win out.
Formally, the word ends in consonant and should therefore get either masculine
or neuter gender. But semantically, Engl. hour corresponds to the German femi-
nine die Stunde ‘the hour’ and it is this gender that gets assigned to Rush-hour. (In
addition, the etymological German equivalent of hour is Uhr ‘watch, clock’, also
a feminine. Speakers familiar with the relationship between hour and Uhr may
draw on the relationship as another factor in favor of assigning feminine gender
to Rush-hour.)
Finally, in cases like Panel, none of the criteria examined so far will unambig-
uously assign a specific gender. Here the default provision takes over and turns
the word into a neuter, as in das Panel.
Similar considerations play a role in other languages. For instance, like other
Bantu languages, Swahili has a system of “noun classes”, with different “class
prefixes”, as in m-toto ‘child’, pl. wa-toto, or ki-swahili ‘Swahili language’. As in
German, adjectives and other words have to agree in class with the nouns they
refer to. Words borrowed from non-Bantu languages therefore have to be inte-
grated into the noun class system. In some cases, nativization takes place, as
in German, on the basis of formal criteria. Thus the Arabic word kitāb ‘book’ is
assigned to the “ki-class” as ki-tabu because of its initial ki-, even though that ki-
is not a prefix in Arabic. Accordingly, it makes its plural as vi-tabu. What helped
in this reassignment is the fact that ki- is the prefix for languages and other “lin-
guistic things” and thus is perfect for a word meaning ‘book’. Similarly, a traffic
pattern enjoined by signs with the verbal message keep left is now referred to by
the nativized expression ki-plefti whose plural, not surprisingly by now, is vi-plefti.
General semantic considerations decide the assignment of words like Engl.
settler to the m- (pl. wa-) class of human beings: m-setla, pl. wa-setla.
And again, a default class accommodates words that are not assignable by
other criteria. In Swahili this is the Ø-prefix class (with Ø-prefix also in the plural),
a class to which are assigned borrowings like Port. mesa [-z-] ‘table’, hence Swah.
Ø-mēza, pl. Ø-mēza.
The issue of nativization also arises in sign languages. For instance, Ameri-
can Sign Language (ASL) borrows many words from (oral) American English. The
mechanism for borrowing has been the use of finger-spelling, in which words that
do not have their own ASL sign can be spelled out in their English form, using the
manual alphabet where different hand-shapes correspond to letters of the English
alphabet. Borrowings from the oral language may be either short words (usually
two or three letters, and generally no more than five) or abbreviations, such as if
and OK (whatever its origin in English – see Chapter 9, § 1). Similar borrowings are
KO for ‘knockout’ (a boxing term) and NG, as an acronym for ‘No Good’, which,
 “Hyper-foreignization” – A further effect of borrowing 239

though not common today, had some currency in American English usage in the
1950s.
Finger-spelled loan words often undergo nativization, just like loan words
into oral languages. For example, the finger spelling for OK has undergone assim-
ilatory changes (see Chapter 4, § 5.1.1). Instead of the thumb contact and arcing
of the fingers characteristic of an independent finger-spelled O, we find that in
one version of the sign, the O has assimilated to K, so that the thumb is in contact
only with the first two fingers, the fingers used in the formation of the K. In many
instances, the changes to these loan words are so drastic that they lose all trace of
their origin as finger spellings. ASL users, for instance, are said to identify the sign
NG with the similarly formed sign for ‘eliminate’, the idea being that something no
good is to be thrown away. This identification constitutes a form of folk etymology
(see Chapter 5, § 3.2).

4 “Hyper-foreignization” – A further effect of

borrowing
As shown by the example of raja(h) in the previous section, the belief that foreign
words must have “foreign” sounds may lead to the introduction of hypercorrect,
or “hyper-foreign” substitutions, such as [ž] for the original [ǰ] of raja(h) – in spite
of the fact that the correct [ǰ] is a perfectly normal speech sound of English.
Hyper-foreignization is a relatively common phenomenon in English. One
that has been observed since at least the 1930s is found in pronunciations such as
[kū dǝ gra] for coup de grâce ‘final blow’ (lit. ‘blow of mercy [to relieve an animal
of incurable pain]’), whose French pronunciation is [ku dǝ gRas]. The motiva-
tion for this pronunciation is the belief that “French drops its final consonants”.
Now, it is perfectly correct that French commonly drops its final consonants, as in
gras ‘grease’ [gRa]; but this holds true only if the consonants are final in spelling.
Words with final orthographic -e follow a different rule which drops the -e; grâce
therefore is pronounced [gRas] in French. English speakers who pronounce coup
de grâce as [kū dǝ gra] evidently are not familiar with the intricacies of French
spelling and extend the drop rule to consonants that are final in pronunciation,
not in spelling. The resulting [kū dǝ gra] may be amusing to those more familiar
with French (including of course the French), because it corresponds to a spelling
coup de gras – which means a ‘blow of grease’ – which may be fatal, too, but for
a different reason.
Hyperforeignisms are not restricted to pronunciation but can also be based
on misperceptions of what constitutes foreign morphology. Thus the phrase No
240 Lexical borrowing

problemo, which has had some currency in relatively recent American usage,
being uttered, for instance, by Arnold Schwarzenegger’s character in the film
Terminator 2, is a morphological hyper-Spanish form. The actual Spanish form
for ‘problem’ is problema, with a final -a, but based on English borrowings from
Spanish such as taco, burrito, nacho, or macho, the perception has emerged that
final -o is a typically Spanish word-ending – whence problem-o. Other examples
include expressions such as el cheapo. Spanish speakers tend not to be amused
by this distortion of their language, referring to it as Mock Spanish.
A different phenomenon consists in the creation of “pseudo-foreignisms”,
as in the case of Germ. public viewing (mentioned at the beginning of Chapter 7),
or handy and beamer for mobile/cell phone and LCD-projector, words that are not
used in these meanings in English
At work in these cases is a phenomenon we have seen before and will see
again – most speakers are not linguists. What matters are ordinary speakers’ per-
ceptions of what makes a word or sound seem foreign, not what the actual facts of
the foreign language are. These facts are for linguists to worry about; speakers are
too busy using their own language to be concerned with such fine details.

5 W
 hy borrow? Motivations for borrowing
strategies
The motivation for borrowing which most readily comes to mind is need. If the
speakers of a given language take over new cultural items, new technical, reli-
gious concepts, or references to foreign locations, fauna, flora, there obviously is
a need for vocabulary to express these concepts or references. The easiest thing,
then, is to take over the foreign word together with the foreign article or idea.
Many of the examples that we have looked at so far are of this nature. Compare
especially (1a,b) and (3) above.
But need will not account for all borrowings. For one thing, languages have
internal resources to draw on, like componding or metaphor, for new concepts.
Moreover, it is fair to ask what need, for instance, English would have had for
borrowing words from French like the ones on the left side in (21) below? As the
inherited, Anglo-Saxon lexical items on the right show, there were perfectly work-
able indigenous words for these animals. The reason for the borrowing must be
sought in a different area, namely prestige. The words on the left side refer to
the animals as they were served at table, i. e., in a social sphere where French
culture and prestige dominated after the 1066 conquest of England. The terms on
the right side, by contrast, belong to the social spheres of raising and herding the
 Why borrow? Motivations for borrowing strategies 241

animals which after 1066 were relegated to the non-French-speaking community.

And even though we may no longer think of beef or veal as more prestigious, the
semantic differences between the terms on the left and those on the right are still
an echo of the original prestige difference. (Before the French words were bor-
rowed, of course, the Anglo-Saxon words were the general designations for the
animals, whether served at table or raised in the barnyard.)

(21) beef : cow, bull, ox

veal : calf
pork : pig/hog/swine

Prestige, rather than need, also accounts for borrowings like Germ. Trend (see
§ 3) or the loan shift by which Germ. Papier ‘(sheet of) paper’ came to include in
its range of meanings the notion ‘journal article, presentation at a professional
meeting’; see (22).

(22) English German

paper ‘piece, sheet etc. of paper’ : Papier ‘piece, sheet etc. of paper’
‘article etc.’ : X

Just as English had perfectly serviceable indigenous words for the items in (21)
prior to the Norman conquest, so German has perfectly adequate native words for
Trend and the new meaning of Papier – Zug, Hang, Anlage, Tendenz, etc. for the
first word, Aufsatz, Vortrag, etc. for the second.
If we look at the context in which these words entered the German language,
we can see the motivation for their getting borrowed. They were first used in post-
1945 West German sociology and related social sciences, in conscious imitation
of the corresponding English terms. The initial purpose was to indicate to the
world – or at least to one’s colleagues – familiarity with the most up-to-date and
prestigious literature in the field; and that literature happened to be written in
English. Even now, the terms Trend, and Papier in the meaning ‘article’, tend to
be limited to the somewhat trendy professional jargon of sociologists, pollsters,
and journalists.
It is of course possible to argue that the difference between need and pres-
tige is not really that great – if something is prestigious, we may feel a need to
imitate or borrow it. Nevertheless, the notion of prestige plays a significant role in
determining the extent of borrowing, as well as what kinds of words are likely to
be borrowed. Moreover, other, related social concepts affect the extent to which
foreign words are nativized, especially when that nativization is non-phonolog-
ical.
242 Lexical borrowing

5.1. Prestige relations and their effects. The varying effects of prestige on bor-
rowing can be illustrated by a brief look at the relationships of English with the
different languages it has come in contact with during the course of history.
Let us start with the Anglo-Saxons’ contact with the speakers of Celtic lan-
guages whom they encountered upon their arrival in England. In this particular
situation, the Anglo-Saxons clearly had the upper hand, militarily and politically.
As a consequence, they must have considered the Celts inferior. The low pres-
tige of the Celts, in turn, must be the reason that very few words of Celtic origin
were borrowed. The words that were borrowed were quite restricted in their deno-
tations and connotations. They were limited, in effect, to a few names for animals,
articles of clothing, and topology, such as brock ‘badger’ ⇐ Celt. brokko-, OE bratt
‘cloak’ ⇐ Gaelic (compare OIr. bratt), and crag ‘steep rock’ ⇐ Celtic (compare W
craig, Ir. carraig, Scots Gael. creag). In addition we find a fair number of place
names, including London, whose -don recurs in other place names of Celtic origin,
both in the British Isles and on the continent. Many of the place names appear
to have come to the Anglo-Saxons in Latinized form, reflecting the earlier Roman
domination of much of present-day England and the relatively greater prestige of
Roman culture.
The second important historical contact was with the Old Norse of the
so-called Danes who, after pillaging parts of England, eventually settled in the
so-called Danelaw, intermarrying and otherwise acting as equals with the indig-
enous English population. (See Chapter 2, § 3.3.) From this relationship between
equals resulted a very large number of borrowings, by some estimates more than
1,700. These borrowings affected everyday vocabulary, and included words such
as egg, guest, hit, husband, raise, skill, skin, skirt, sky. Even basic vocabulary was
borrowed, such as get, give, like, take, and the pronouns they, their, them. In some
cases, the borrowings continued to coexist with native, Anglo-Saxon words, such
as skin beside shin, skirt beside shirt, or the clitic form ’em (as in give ’em hell)
beside them. In others, the old Anglo-Saxon words were replaced by the borrow-
ings, as in guest and give, whose inherited counterparts would have been some-
thing like yest and yive.
Names, too, were affected, including the numerous English place names
ending in -by (ON bȳ ‘abode’) and the widespread pattern of family names ending
in -son, reflecting an Old Norse “patronymic” naming pattern preserved in Modern
Icelandic (as in Stefán Einarsson = Stefán, son of Einar).
In addition to the large amount of borrowing and the fact that even basic
vocabulary was affected, there are no special connotations (either positive or neg-
ative) attached to these loans. For instance, Scandinavian borrowings like skirt or
to raise do not differ significantly in social connotations from their Anglo-Saxon
counterparts, such as shirt or to rear.
 Why borrow? Motivations for borrowing strategies 243

The next important contact of English was with the French of the Norman con-
querors who in 1066 became overlords over the native English (and Anglo-Scan-
dinavian) population. This contact resulted in the largest number of borrowings.
Moreover, to the extent that special connotations are attached to the loans, they
almost invariably reflect the higher prestige enjoyed by the speakers of French.
Compare the examples in (21) above. At the same time, as in the case of the earlier
contact with Celtic but in contrast to the Danish contact, the most basic vocabu-
lary remained unaffected.
The last contact to be examined closes the circle in more than one sense. This
is the contact between English and the indigenous languages of North America.
From the perspective of the conquering Europeans, this was a contact of unequal
relationship very similar to the much earlier one between Anglo-Saxons and
Celts. The difference in prestige, again, is reflected in the types of borrowings that
were made from the Indigenous American languages. The most general sphere
of borrowings is that of place names – of the forty-eight contiguous states of the
United States, more than half bear names derived from indigenous languages.
Note for instance Illinois, Michigan, Ohio, Wisconsin. But even here we find a
strong tendency to use European-derived names, such as New York, Washington,
Virginia.
Beyond place names, borrowings most commonly are found in names for
fauna and flora, such as woodchuck and moose from Algonqu. otček and mōs.
Other borrowings of this sort include opossum, skunk, wapiti, hominy (grits),
tupelo (tree), persimmon, and succotash. However, here too the tendency is to
adapt European words to the new surroundings. Compare for instance the word
robin which in America refers to a bird quite different in size and taxonomic clas-
sification from the European bird of the same name.
Other borrowings are even more limited and tend to refer exclusively to Indige-
nous American life (compare moccasin, pow-wow, squaw, teepee, toboggan, totem,
wampum), very often with derogatory connotations, as in the case of squaw, or
with specifically American Indian connotations (such as teepee and wampum).
The contact situations just outlined and the nature of the borrowings asso-
ciated with them are fairly typical of linguistic contact in general. The different
types of relative social status of the participants in such contact situations can be
characterized by the terms adstrate, superstrate, and substrate. Languages
of roughly equal prestige, such as English and Norse in early England, are referred
to as adstrates. Where prestige is unequal, as between Normans and Anglo-Sax-
ons, between Anglo-Saxons and Celts, or between English-speaking Europeans
and Indigenous Americans, the terms superstrate and substrate are used, the
former referring to the language with higher prestige, the latter to the one with
lower prestige.
244 Lexical borrowing

Adstratal relationships, then, are most conducive for borrowings of every-

day-life vocabulary, even basic vocabulary. In contrast, there are much greater lim-
itations in contacts between languages of unequal prestige. Moreover, in such con-
tacts, the borrowings tend to reflect the social status of the donor language. If that
language is a superstrate, loans tend to come from the more prestigious sections
of the lexicon and their connotations likewise tend to be prestigious, at least at the
beginning. If the donor language is a substrate, loans tend to be limited to need
borrowings (such as new place names) and/or to have derogatory connotations.

5.2. Linguistic nationalism or the effect of social attitudes on nativization:

In § 3 we have noted a large variety of routines through which foreign words
are integrated into the borrowing language. In most of these, foreign words are
nativized in phonological or morphological structure. But occasionally they may
be adopted in unmodified form, without any significant degree of nativization.
Though such unmodified adoption is rare in the phonological structure of words,
it is quite common in morphology. In fact, all of the borrowings discussed in § 5.1
are of this type. But as we have seen, processes like calquing and the develop-
ment of loan shifts offer an alternative to morphological adoption – the creation
of words made up entirely from native sources. Let us contrast these two different
approaches to morphological integration as adoption vs. adaptation.
As seen in the preceding section, prestige plays a powerful role in determin-
ing the kinds of words that are borrowed and the quantity of borrowings. The
question of what determines the choice between adoption and adaptation like-
wise finds a satisfactory answer only when social factors are considered.
At first sight, one might suspect that the choice between adoption and adap-
tation is determined by linguistic factors. Adoption would be preferred where the
structures of donor and borrowing language are sufficiently similar to permit the
process to apply. Elsewhere, adaptation would be preferred. In support of this
view one might point to the difference between Chinese and English. Chinese
heavily favors adaptation. For instance the adaptation dian hua, lit. ‘lightning
speech’, is much more readily accepted as the rendition of Engl. telephone than
the rival adoption de lü feng, even though the latter sounds more similar to the
English original. English on the other hand quite readily adopts foreign words,
e. g. macho, rouge, umlaut.
This difference might be attributed to the fact that in its structure, Chinese
differs radically from the European languages, by having essentially monosyllabic
words (although compounding is possible), with severe restrictions on the occur-
rence of consonants, plus a rich system of tonal contrasts. When Chinese speak-
ers try to adapt by substituting similar-sounding syllables for the syllables of a
foreign word, they face the difficulty of having to match up not only the sounds
 Why borrow? Motivations for borrowing strategies 245

of the donor language, but also the meanings of these syllables so that they don’t
conflict with the meaning of the foreign word. For instance, a nativization of Engl.
telephone as de lü feng is quite good from the phonological perspective, but seman-
tically it causes difficulties, for de lü feng literally means ‘power-law-wind’, which
only vaguely fits the meaning of telephone. Chinese speakers therefore prefer to
adapt foreign words, by creating compounds of native words whose meanings are
more compatible with those of the foreign originals, such as dian hua, lit. ‘light-
ning speech’, for telephone. Even worse is the case of mai ke feng as a phonetic
approximation of Engl. microphone. Its literal meaning, ‘wheat-gram [unit of
weight]-wind’, cannot be considered even remotely related to the meaning of the
English word. But interestingly, in spite of these difficulties mai ke feng, generally
shortened to mai ke, has been accepted as the normal word for microphone.
Structure, however, cannot be solely responsible for preferring adaptation
to adoption. This is shown by the case of Modern Icelandic. Although the struc-
ture of Icelandic is much less different from that of English and other European
languages, it behaves like Chinese, generally preferring adaptation and limiting
adoption to foreign place names and terms for foreign fauna and flora. Contrast
the adoptions in (23a) with the adaptations in (23b). (English here is used as a
representative for the majority of European languages, which have a more tol-
erant, or at least mixed, attitude toward adoption.) In many cases, the adapta-
tions are more or less literal calques, as in (23b.i). Others are loan shifts; compare
(23b.ii), which actually involves the resurrection of an Old Icelandic word whose
meaning roughly was ‘wire’, as a translational equivalent of the English term wire
‘telegram’. In many other cases, the words are recreated from Icelandic elements,
without even attempting to provide a more or less precise translational equivalent
of the foreign words; see (23b.iii). On the other hand, adoptions like jeppi ‘jeep’
and berkill ‘tubercle’ are exceedingly rare. In fact, as examples like samríkismaður
‘Republican’ show, even foreign terms often are adapted, rather than adopted.

(23) a. Arabi Arab

Evrópa Europe
melóna melon
b. i. ljóshvolf photosphere ‘light concavity’
= ‘photo’ + ‘sphere’
fjarverða absence ‘far-being’
≈ ‘ab-’ + ‘sence’
samtal conversation ‘together-speak’
≈ ‘con-’ + ‘versation’
úrval selection ‘out-choice’
≈ ‘se-’ + ‘lection’
246 Lexical borrowing

ii. síma telegram, telephone,

telecommunication ‘wire’
talsíma telephone ‘speech wire’
ritsíma telegram ‘write wire’
iii. samríkismaður Republican ‘together-state-man’
hershöfðingi general ‘army-chieftain’
bókmentir literature ‘book-arts’

We know that this avoidance of adoption has not always been dominant in Ice-
landic. In texts of the sixteenth through early nineteenth centuries, innumerable
adoptions of foreign words can be found; compare (24). These came to Icelan-
dic through the Scandinavian languages, especially Danish. The Scandinavian
languages themselves had undergone extensive influence from the Low German
trade language of the Hanse, a commercial league of Northern German and Dutch
cities. As a consequence, many of the adopted words ultimately derive from Low
German. In addition, of course, there was the ever-present influence of Latin,
mediated through Danish and often also through Low German.

(24) borger ‘citizen, burgher’ (ultimately from Low German)

borgmeistari ‘mayor’ (ultimately from Low German)
dedicera ‘to dedicate’ (ultimately from Latin)
disputatía ‘dispute’ (ultimately from Latin)

The reason for this large-scale adoption of foreign words must be sought in two
factors: political domination by Denmark, and the introduction of Lutheran Chris-
tianity. Danish domination made Danish the prestige language. And the change
from Roman Catholicism to Lutheranism brought with it a large amount of new
Danish (ultimately German) terminology which was intimately linked with the
new form of religion and which, through this association, carried considerable
prestige.
Even at the height of their use in Icelandic, these adopted borrowings occurred
more frequently in informal writings than in more formal texts, suggesting that
they were trendy prestige borrowings, rather than need-based. Significantly, the
very writers of these informal texts inveighed against the use of foreign words.
They feared that the excessive use of such words would alienate Icelanders from
their own rich medieval literature, which is still highly revered by the Icelandic
people. As nationalist feelings increased markedly during the nineteenth century,
virtually all the foreign words adopted since the sixteenth century were elimi-
nated or replaced by adaptations, and with the exceptions noted above, new bor-
rowings were accepted only in adapted form.
 Why borrow? Motivations for borrowing strategies 247

The socially based motivating force behind these developments is now com-
monly referred to as linguistic nationalism or purism, the use of language to
assert the identity and prestige of one’s own people – in contrast to the prestige
that might be attached to foreign languages and their speakers.
Ironically, to the extent that it resorts to calquing, linguistic nationalism
requires a much fuller understanding of foreign linguistic structure than plain
adoption. When adopting a foreign word like photosphere, it is not strictly nec-
essary to understand that it is composed of the elements photo ‘light’ and sphere
‘sphere, concavity’. But such an understanding is essential for calquing pho-
tosphere as ljóshvolf. As a consequence, adaptations usually are introduced by
persons with a good understanding of the donor language’s morphology and
vocabulary.
In spite of its present-day aversion to adopting foreign words, Chinese, too,
has not always been resistant to adoption. When Buddhism came to China during
the Middle Chinese period, speakers of Chinese struggled valiantly to adopt
the Sanskrit terminology of Buddhism, in spite of the fact that, if anything, the
structural differences between Chinese and Sanskrit are even greater than those
between Chinese and English. One of the words that were borrowed in this contact
was later adopted in Japanese and is now known to us in its Japanese form. This
is the word zen whose ultimate source is Skt. dhyāna- ‘meditation’. The linguistic
nationalism of Modern Chinese, then, must be a more recent development – or a
rekindling of an earlier attitude after the influence of Buddhism had abated.
Linguistic nationalism is by no means limited to Icelandic and Chinese. It is
found in many other languages, although most of them show its effects only in a
very inconsistent, even erratic fashion.
Consider German. The erratic nature of linguistic nationalism is reflected
in two ways. First, in many cases, foreign words appear both in adapted and in
adopted form; see (25). Secondly, there is no consistency in the connotations
associated with adaptations vs. adoptions. As (25) shows, in some cases it is
the adopted borrowing, in others, the adapted word that is the more natural or
popular. The other word, then, often is used mainly in “officialese” or in special-
ized jargons. Thus, people might enter a phone booth which is marked Öffentli-
cher Fernsprecher ‘Public Far-Speaker’ = ‘Public Telephone’ (an officialese expres-
sion), but in the booth they use the Telefon (which is the normal word). And there
is at least one case (Auto : Wagen) where both words are commonly used, with
different speakers preferring one or the other term, but with no consensus as to
which one is more natural or popular. The degree of inconsistency becomes espe-
cially clear if we contrast the words for ‘telephone’ and ‘television’. In the case of
‘telephone’, normal use prefers the adoption, while the adaptation is officialese.
For ‘television’, the situation is just about the opposite.
248 Lexical borrowing

(25) Adoption Adaptation

Auto (general use) Wagen (general use)
Kraftwagen (officialese)
Radio (general use) Rundfunk (officialese)
Telefon (general use) Fernsprecher (officialese)
Television (TV industry jargon) Fernsehen ‘TV institution’
(general and officialese)
Fernseher ‘TV set’ (general)
Fernsehgerät ‘TV set’ (officialese)
Kopie ‘xerographic copy’ Ablichtung (small-town, officialese)
(large city, academia)

The situation is similar in most other languages of Europe (excepting Iceland).

Even in France, where linguistic nationalism has the powerful support of the pres-
tigious Académie Française and of the political and educational establishment,
attempts at purging the language of foreign adoptions have met with only limited
success. Words like redingote ‘riding coat’ (see (1a) above) and club (borrowed
from nineteenth-century British English) are fully integrated into the French
lexicon. And everybody knows that the Académie’s attempts to ban “Franglais”
have been rather ineffectual, at least in more trendy circles. Expressions like le
hot dog, le hamburger, le weekend, mainly of American English origin, have found
wide acceptance.
Even in Israeli Hebrew, linguistic nationalism has had less than spectacular
success. This is so in spite of the fact that the founders of the Hebrew revival in the
early part of the twentieth century made a deliberate, nationalistically inspired
attempt to revive the language in a truly Semitic guise, with total elimination of
the various European influences that had crept into the language over the centu-
ries. The founders of the movement expected the total elimination of European
influence to free Hebrew and its speakers from the yoke of 1,700 years of dias-
pora, second-class citizenship, and recurrent persecution. However, not even the
killing of six million Jews in the Nazi holocaust – an event that precipitated the
founding of the state of Israel – has been able to block the entrance of adopted
words like viza ‘visa’, student, or gymnazia (a type of secondary school).
Native speakers of English often find it difficult to understand why linguistic
nationalism should manifest itself in a preference for adaptations over adoptions.
However, linguistic nationalism is hardly alien to modern English. It is clearly
present in the area of pronunciation, which favors phonological adaptation over
the adoption of foreign sounds. Thus, most speakers would consider the pronunci-
ation of the composer’s name Bach with final [x] to be an affectation. (Contrast this
with the fact that educated German has adopted nasal vowels from French, as in
 Why borrow? Motivations for borrowing strategies 249

[zalɔ̃ ] ‘salon’, and [ǰ] from English, as in [meneǰǝr] ‘manager’.) At the lexical level,
however, adaptation is rare in Modern English and adoption just about the norm.
Even the preference for lexical adoption, however, is a fairly recent phenom-
enon. The introduction of Christianity in the Old English period, for instance,
was accompanied by a large number of loan shifts and calques; see (16) and (17c)
above. In this case, adaptation may have sprung not so much from linguistic
nationalism as from a conscious attempt by the missionaries to make the Chris-
tian religion less unfamiliar and therefore easier to accept by using terms that
the intended converts were familiar with. (This, in fact, is standard procedure in
Christian missionary efforts.)
Linguistic nationalism did, however, play a strong role in English during
the sixteenth and seventeenth centuries. As in many other parts of Europe, the
Renaissance had brought with it a rekindled interest and reverence for the lan-
guages of (western) European classical culture and civilization, and this interest
and reverence led many English writers to draw heavily on the classical languages
as sources for new vocabulary; compare (26). Many of the resulting borrowings,
such as the ones in (26a), have become an integral part of the English vocabulary.
Many others have not; see the examples in (26b).

(26) a. affirmation
negation
maturity
modesty
persist
b. adiuvate ‘help’
dominical ‘lordly’
ingent ‘enormous’
obtestate ‘beseech’

Words of the type (26b) failed to become generally accepted for several reasons.
Some of the words may have been considered excessively trendy by most contem-
poraries and thus would have passed out of usage anyway. But what may have
been even more important is that the large influx of unassimilated or poorly assim-
ilated foreign words met with a reaction very similar to the Icelandic response to
the influx of Danish words. English writers and scholars began to inveigh against
the excessive use of foreign words, often ridiculing them as “inkhornisms”. The
words in (26b) are all found in an ‘ynke-horne letter’ published by Thomas Wilson
as an illustration of usage that he condemned.
Attacks against inkhornisms came especially from two sides, the “Anglo-Sax-
onists” and the Puritans. Like the Icelandic critics of excessive Danish borrow-
250 Lexical borrowing

ings, the Anglo-Saxonists wanted to maintain the linguistic link with medieval
literature and tradition and regarded the flood of foreign words as a serious obsta-
cle. The Puritans equated “plain speech” with truth and saw the excessive use
of borrowings as a deviation from truth. One of the strongest advocates of plain
speech, John Cheke, used terms such as yeasay and naysay instead of the words
affirmation and negation (26a). Many other, similar adaptations were proposed,
such as unboundedness for infinity and gainrising for resurrection.
English differs from Icelandic in that most of these proposed adaptations have
met a similar fate as the inkhorn terms in (26b). Only a few, such as unbounded-
ness, have retained some degree of currency, but perhaps mostly only in technical
usage in mathematics or linguistics, and not necessarily in the precise meaning
of ‘infinity’. This difference no doubt results from the fact that the use of adopted
borrowings from Greek and Latin found strong, even vociferous, support among
many other English writers and scholars, who viewed the use of such terms highly
appropriate “for the necessary augmentation of our language” (Thomas Elyot).
Moreover, while medieval Icelandic was fairly free of foreign influence, the medi-
eval language of poets like Chaucer was too clearly influenced by the French of
the Norman conquerors to be describable as pure Anglo-Saxon. Finally, “plain
speech” may have come to be too closely associated with the sectarian activities
of the Puritans. Whatever the reasons, linguistic nationalism failed to become the
same powerful force as in Iceland.
This does not mean that linguistic nationalism faded away entirely. Occa-
sionally it was rekindled, especially when fueled by political nationalism. For
instance, antipathy to Germany during the First World War led to attempts to
replace adopted German borrowings such as weltanschauung by adaptations like
world view (see (17b) above), or even more daring replacements such as victory
cabbage for sauerkraut (⇐ Germ. Sauerkraut ‘sour cabbage’). But many of these
replacements did not succeed in the long run.
In general, the argument that adoptions enrich the English language has
carried the day. Linguistic nationalism survives mainly as an anti-intellectual
undercurrent, especially among vernacular speakers who abhor the “high-falu-
tin” sesquipedalianisms of the educated.
The unqualified success of linguistic nationalism in Modern Icelandic as well
as in Modern Chinese, then, is quite unusual and must be attributed to very special
circumstances. In both cases, the immediate reason for this success is the fact that
the attitude of linguistic nationalism is shared by virtually all layers of society.
In Icelandic, linguistic nationalism seems to have been supported by the
movement to achieve independence, as well as by a genuine fondness and rever-
ence in all layers of society for medieval literature whose written form remained
remarkably intelligible to the speakers of Modern Icelandic.
 Why borrow? Motivations for borrowing strategies 251

In the case of Chinese, linguistic nationalism is supported by a highly tradi-

tional, ethnocentric attitude. It has been claimed that an added factor lies in the
logographic nature of Chinese writing; see Chapter 3, § 5.1. The problem therefore
is not simply that phonetic adoption creates difficulties because all syllables are
meaningful; after all, there would be nothing to prevent Chinese from using for
nativization phonetic syllables that are not associated with any specific meaning.
Much more important is the fact that the writing system does not offer any
symbols for such meaningless syllables – all writing symbols are meaningful. Put
differently, Chinese has a triple identity between word, syllable, and logographic
symbol, and this identity places enormous obstacles in the way of the phonetic
adoption of foreign borrowings.
There is probably some truth to this argument; but as observed earlier,
Modern Chinese does offer instances of phonetic adoption, such as mai ke feng
for Engl. microphone, even though the literal meaning of the logographic symbols
employed, ‘wheat-gram [unit of weight]-wind’, clashes with the semantics of the
English word. Even more significant, Middle Chinese nativized a massive amount
of Sanskrit words through phonetic adaptation. As in the case of Icelandic, there-
fore, the real reason for preferring adoption over adaptation is more likely to lie
in the social attitude of linguistic nationalism.
In most other societies, the general populace seems to have a more ambivalent
attitude toward linguistic nationalism. However, even if its effects may be mixed,
there can be no doubt that in most societies, linguistic nationalism provides an
important counterbalance to the prestige of foreign culture and vocabulary.
The two socially based notions of prestige and linguistic nationalism, thus,
are the most important factors that determine the manner in which foreign words
are integrated into the lexicon of the borrowing language. In English, to be sure,
foreign prestige or domestic linguistic nationalism play a subordinate role, as
compared to the notion that adoptions enrich the language. But like prestige and
linguistic nationalism, the notion of enrichment is based on social attitude, not
on linguistic structure. (In fact, English has been characterized as ‘a language that
lurks in dark alleys, beats up other languages and rifles through their pockets for
spare vocabulary’.)
Before concluding this section, it may be appropriate to mention that linguis-
tic nationalism can take rather unexpected forms. For instance, during Japan’s
highly nationalistic phase, prior to its defeat in 1945, the speakers of Japanese
strongly resisted the adoption of foreign words. Instead, they preferred adapta-
tions. But these generally were made not in terms of native Japanese vocabulary,
but by means of words of Chinese origin. In fact, many of the Modern Chinese
adaptations of foreign words, such as dian hua, were created in Japan and spread
from there to China.
252 Lexical borrowing

The Chinese words had been adopted in the first millennium A.D., partly
because the Japanese adopted Buddhism from China. Chinese, thus, had acquired
in Japan a very similar role to that of the classical European languages, Latin and
Greek, in much of Europe. As a consequence, its vocabulary could be regarded
as indigenous and East Asian and thus more congenial to the Japanese language
than adoptions from western languages.
We find a similar situation in Modern Indonesian, which draws heavily on
Sanskrit lexical resources to adapt foreign words and concepts. Here Sanskrit had
been the source for a large number of earlier borrowings and thereby acquired the
role of an indigenous, Asian prestige language whose vocabulary can be drawn on
to indigenize foreign western words and concepts.
Understandably, Sanskrit, as the language of traditional Indian culture and
civilizations, plays a similar role in most of the modern languages of the Republic
of India, both Indo-Aryan and Dravidian. Compare for instance the Sanskrit-based
Hindi adaptations from English in (27). As in Icelandic, some of the adaptations
are straightforward calques (27a). Others are recreated from Sanskrit elements,
only partly influenced by the structure of the English model. Thus in (27b), the
first element of viśva-vidyālay(a) echoes the univers- of Engl. university, but the
rest combines Sanskrit elements into a compound ‘knowledge-abode’ = ‘school’.
Similarly, in (27c), the pro- of Engl. professor is calqued by its cognate pra-, to
which then is added one of the Sanskrit words for ‘teacher’, adhyāpaka.

(27) Hindi English Sanskrit elements

model
a. prag(a)ti progress pra- ‘forward’ + gati- ‘going’
(= Engl. -gress
from Lat. gradior ‘step, walk’)
prasār(a) promotion pra- ‘forward’ + sār(a) ‘causing to move’
lalit(a) kalā fine arts lalita- ‘lovely, charming’ = ‘fine’ + kalā
‘art’
b. viśvavidyālay(a) university viśva- ‘all, universal’ + vidyā- ‘knowledge’
+ ālaya- ‘abode, place’
bhāṣāviǰñān(a) linguistics bhāṣā- ‘language’ + viǰñāna- ‘knowledge,
science’
c. prādhyāpak(a) professor pra- ‘forward, in front’ + adhyāpaka
‘teacher’

Although such adaptations are Sanskrit in form, semantically they are eminently
English. For instance, Hindi uses the terms ārambh(a)- or samāroh(a)- in the
meaning ‘festive occasion’. In terms of the Sanskrit elements of which these words
 The effects of borrowing 253

are composed, one would expect meanings such as ‘beginning’; and these are in
fact attested for these words in traditional Sanskrit. The meaning ‘festive occa-
sion’ could only have arisen via the English semantics of commencement which
can mean both ‘beginning’ and ‘festive occasion (especially at a university)’. The
pervasive influence of English semantics can be noticed even in extended uses of
Sanskrit-based adaptations. For instance, the term pragati in (2ā) is beginning to
be used not only to designate the idea and ideology of ‘progress’ in an abstract
sense, but also the use of progress in expressions like work in progress, which is
calqued as kām pragati mẽ – much to the chagrin of purists who consider such
usage to be excessively influenced by English.
Two Indian languages resist adaptations by means of Sanskrit elements –
Urdu and Tamil. As an Islamic counterpart of Hindi (see Chapter 2, § 3.10.2),
intent on maintaining its distinctiveness vis-à-vis Hindi, Urdu draws on Arabic
and Persian sources to create counterparts for English terminology, such as lisani-
yat ‘linguistics’ (from Arab. lisan ‘tongue, language’ + a derivative suffix -iyat) and
funūn ē latīfā ‘liberal arts’ or ‘fine arts’ (where funūn = plural of Arab. fan ‘activity’,
ē = a Persian linking element, and latīf(ā) = ‘good, fine’).
Although Tamil has borrowed heavily from Sanskrit in the past (e. g. words
like āsiriyaṉ ‘teacher’ ⇐ Skt. ācarya), it now prefers to draw on its own resources
for adapting foreign terminology, including the Sanskrit-derived terminology of
other Indian languages. Thus, for viśvavidyālay(a) ‘university’ Tamil uses the
word palkalaikkar̤ akam, composed of pal ‘various’ ≈ viśva ‘all, universal’, kalai
‘art’ ≈ vidyā ‘knowledge, science’, and kar̤ akam ‘assembly’ ≈ ālay(a) ‘abode’. And
as a counterpart to prādhyāpak(a) ‘professor’ it offers pērāsiriyaṉ = pēr ‘great’ +
āsiriyaṉ ‘teacher’ = adhyāpak(a). These “Tamilizations” of Sanskrit-based termi-
nology are motivated by a more regionally defined form of linguistic national-
ism – the widespread feeling that Indo-Aryan political and cultural domination,
whether by Modern Hindi or by Sanskrit, has been excessive and has threatened
the separate, Dravidian identity of the Tamil people.

6 The effects of borrowing

As we have seen in the preceding section, English has a strong tendency to adopt
foreign vocabulary, usually restricting adaptation to phonology. In fact, since the
time of the Renaissance, English writers and scholars generally have claimed that
lexical adoptions lead to an enrichment of English vocabulary.
The extent of this enrichment may be gauged from the fact that between sixty-
five and seventy-five percent of present-day English vocabulary is of foreign origin.
254 Lexical borrowing

Much of that vocabulary comes from Romance. In most cases of Romance borrow-
ings, the source is French; see (28a). Less commonly, other Romance languages,
especially Spanish and Italian, are the donor languages, as in (28b). In addition,
of course, there is an abundance of borrowings from Graeco-Latin sources. But
the phonetic shape of these borrowings usually is closer to French than to either
Greek or Latin; see (28c). In some cases, this is because the word was borrowed via
French. This may be the case for nation. In others, the reason is that the English
word has been assembled from elements received via French, as in intercontinen-
tal. In yet others, the word may have been similarly assembled in French, such as
in the case of hydrogen. Note further that in some cases, English words may owe
their phonological shape to etymological nativization (see § 3) based on the pho-
netic correspondence between earlier borrowings from French and their English
equivalents, such as Fr. nation [nasyɔ̃ ] : Engl. nation [neyšn̥ ]; this is no doubt the
case for negation.

(28) a. place, receive, rouge, veal

b. plaza (from Spanish)
piazza (from Italian)
c. nation (Lat. natiō, -ōnis, Fr. nation)
intercontinental (Fr. inter- ‘between’, continental ‘continental’; compare
Lat. inter- and continentālis)
hydrogen (Fr. hydrogène; compare Gk. hudro- ‘water’ + genēs ‘engender-
ing’)
carbon (Lat. carbō, -ōnis, Fr. carbone)
negation (Lat. negatiō, -ōnis)

The effect of these borrowings is especially striking in technical prose, where

English and French (and other Romance languages) show a great degree of termi-
nological similarity, whereas German, with its adapted Wasserstoff ‘hydrogen’ (lit.
‘water matter’), Kohlenstoff ‘carbon’ (lit. ‘coal matter’), etc., appears to be quite
different and “Teutonic”. Many popular writers on English therefore claim that
English now really is a Romance, not a Germanic language.
Historical linguists have difficulties with such claims. First, although more
than sixty percent of the total vocabulary may be of foreign, largely Romance
origin (especially if Latin is included under the definition Romance), much of
that borrowed vocabulary is restricted to certain, highly technical or specialized,
spheres of usage. Compare the examples in (29).
 The effects of borrowing 255

(29) quadrant
quadrivium
questionnaire
quincunx
quodlibet

On the other hand, in basic vocabulary, cognates are much easier to find between
English and German than between, say, English and French; see (30a). The simi-
larities between English and German become especially striking if we look at the
morphology of basic vocabulary, such as the principal parts of irregular verbs,
as in (30b). Clearly, then, in this most basic, most indispensable, and most fre-
quently used part of its vocabulary, English looks very much like a “Teutonic”
language, not like Romance. Even where English and German do not agree (as
in sky : Himmel), English does not show any closer agreement with French (ciel).

(30) a. English French German

I je ich
you tu/vous du/ihr/Sie
he il er
she elle sie
it il es
do faire tun
be être sein
eat manger essen
drink boire trinken
earth terre Erde
sky ciel Himmel
and et und
to à zu
of de von
b. English drink (pres.), drank (past), drunk (past pple.)
German trinken (pres.), trank (past), getrunken (past pple.)
vs. French boire (pres.), buvais (imperf.), bus (passé défini), boirai
(fut.), bu (past pple.)

More than that, the so-called Romance component of English comes from differ-
ent, distinct Romance languages. For instance, the place of (29a) above, and the
plaza and piazza of (29b) all go back to Latin platēa ‘wide street’, which itself was
borrowed from Greek plateîa (hodós) ‘wide street’. But place shows developments
peculiar to French, plaza to Spanish, and piazza to Italian. There are many similar
256 Lexical borrowing

sets of multiple borrowings, made at very different periods and from different
sources, but coexisting in modern English; see (31).

(31) a. Modern English forms derived from Latin discus ‘quoit, disk’
(⇐ Gk. dískos)
Source
dais < ME deis ⇐ Fr. deis < Lat. discus
desk ⇐ Mediev. Lat. desca ⇐ Ital. desco < Lat. discus
dish < OE disc < West Germanic *diskaz ⇐ Lat. discus
disk/disc ⇐ Fr. disque ⇐ Lat. discus
discus ⇐ Lat. discus
b. Modern English forms derived from the word for ‘brother’ in various
languages
Source
fraternal, fraternity ⇐ Lat. frāter ‘brother’ (and derivatives)
Fra ⇐ Ital. fra ‘brother; designation of a friar’ < Lat.
frāter
friar ⇐ OFr. frere ‘brother’ < Lat. frāter
phratry ⇐ Gk. phratría/phrátra ‘a clan group’ (consisting of
‘brothers’ or ‘brethren’ in the extended sense)
pal ⇐ Romani p(h)al, phral ‘brother; buddy’ < Skt.
bhrā́tar-

Examples like (28a,b) or (31a) show that the “Romance” component of English is
a more or less accidental amalgam from different Romance languages.
Moreover, while the majority of Romance borrowings are French or Latin in
character, they have entered English at various times, with very different subse-
quent developments within English. In addition to the examples in (30a), compare
Engl. petty vs. petite, both borrowings from Fr. petit (m.), petite (f.) ‘small’, but
adopted at different times. While the more recently borrowed petite is phoneti-
cally quite close to its French counterpart, petty, borrowed in the medieval period,
has an accentuation which is more fully nativized. Consider also correspondences
like Engl. chant with [č-] vs. Fr. chant with [š-]. Here English preserves the initial
[č-] of Old French, while Modern French [š-] is the result of a later French sound
change. That is, English cannot be identified with any single chronological layer
of Romance.
Finally, with a few exceptions (such as rouge), the borrowings from Romance
(and other languages) have been completely nativized in their phonology and
thus have ceased to be French, Italian, or Spanish, but have become fully English.
As a consequence of having become nativized, they have undergone subsequent
 The effects of borrowing 257

changes that are peculiar to English and have been considerably altered in their
pronunciation, sometimes beyond recognition. Compare the borrowings from
French in (32).

(32) Modern English Modern French

measure [mɛžǝ(r)] mesure [mǝzür]
beef [bīf] bœuf [bɔ̈ f ]
curfew [kǝrfyū] couvre-feu [kuvr(ǝ)fö] (lit. ‘cover the fire’, i. e. “lights
out”)

In summary, then, it appears that lexical borrowing, even on a massive scale, is

highly unlikely to lead to a change in the genetic affiliation of a given language.
(For certain complications see Chapter 12, § 3.) On the other hand, the adoption
of borrowings may be conceived of as enriching the lexicon of a given language.
As we have seen, this belief is most widespread in English and probably accounts
for the fact that unlike most other languages, English almost voraciously adopts
foreign words.
Even in this respect, however, it is advisable to exercise caution. Languages
that adapt foreign words by means of calquing (as in Germ. Wasserstoff lit. ‘water
matter’ ≈ Fr. hydrogène ‘water engendering/matter’) expand their vocabulary just
as much as languages that adopt (such as Engl. hydrogen).
There is, however, one form of adaptation that does not enrich the lexicon,
namely adaptation through loan shift. By definition, loan shifts simply redefine
the meanings of existing lexical items, without adding new words. Moreover,
the result of loan shifting is an increase in polysemy, which potentially can lead
to ambiguities. Consider the case of Germ. Papier discussed in § 4 above. The
increase in the semantic range of this word from ‘(sheet of) paper’ to ‘(sheet of)
paper; journal article, presentation at a professional meeting’ increases the pos-
sibility of misunderstanding such sentences as Ich kann mein Papier nicht finden
‘I can’t find my paper’ (is it a sheet of paper or a journal article?), which can be
avoided by the old-fashioned distinction between Papier ‘(sheet of) paper’ and
Aufsatz ‘article’. The situation is even worse for the plural use, Ich kann meine
Papiere nicht finden, since Papiere (plur.) has the additional, specialized meaning
‘documents’.
As an alternative to loan shifts, therefore, borrowing by adoption can be justi-
fiably argued to enrich the language, by conveying the new meaning without any
ambiguity. In English, for instance, the German borrowing angst manages to more
clearly indicate the intended psychoanalytical connotations than anxiety or fear.
Even in other respects, adoption may enrich a language. For one thing, many
loans are not need-based, but prestige borrowings that introduce novel linguis-
258 Lexical borrowing

tic forms for already existing linguistic concepts and their corresponding native
forms. Thus when English borrowed from French the adjective royal, it already
had its own indigenous adjectival formation corresponding to king; namely kingly.
The new term royal then came to compete with the inherited form. And as noted
in § 5.5 of Chapter 7, such a competition usually is resolved through semantic spe-
cialization. In the present case, royal became the normal adjective corresponding
to king, while kingly survived in more specialized functions.
A less obvious enriching effect of adopted borrowings is that in many cases
they bring with them their own, novel morphological inventory and rules for the
combination of morphological elements. This is of obvious benefit in the area of
word coinage, the creation of new linguistic terms to express novel concepts. As
noted already in § 1, borrowed morphology provides an increase in the morpho-
logical elements and rules which can form a basis for coining new words. This
can be especially important for a language like English, in which the ability of
native derivational morphology to create complex new structures is fairly limited.
For instance, native English morphology rarely goes beyond structures like like-li-
hood or own-er-ship. The morphology abstracted from Latin and Greek sources, on
the other hand, permits the creation of complex derivations such as dis-establish
→ dis-establish-ment → dis-establish-ment-ary → dis-establish-ment-ari-an → dis-
establish-ment-ari-an-ism or dis-establish-ment-ari-an-ist → dis-establish-ment-
ari-an-ist-ic → dis-establish-ment-ari-an-ist-ic-al → dis-establish-ment-ari-an-ist-
ic-al-ly.
Additionally, the borrowed morphology frequently signals that the new word
is a technical term, not just an ordinary, everyday word. This is clearly the case
for such sesquipedalianisms as dis-establish-ment-ari-an-ist-ic-al-ly, or the word
sesquipedalianism itself, for that matter. But it affects many other spheres of the
vocabulary as well. Consider for instance the case of automobile. The word was
created from the elements auto- ‘self’ and mobile ‘moving’, extracted from bor-
rowings from Greek and Latin, respectively. The novel combination of these ele-
ments into automobile, then, signaled the technical nature of the resulting word
much more clearly than would have Engl. self-moving or Fr. mouvant par soi. Nev-
ertheless, some technical terms in English do not use borrowed morphology, but
are quite mundane Anglo-Saxon collocations; compare for instance black hole in
astrophysics. (The issue of coinage is discussed in fuller detail in the next chapter.)
In fact, perhaps the most important and overriding effect of large-scale adop-
tive borrowing on English is the creation of a clearly marked formal distinction
between an educated/technological variety and other, more everyday varieties of
the language. As can be seen from epithets like “sesquipedalian” or “high-falutin”
for the technological vocabulary, this distinction is very clear to native speakers,
no matter whether they are educated or not.
 The effects of borrowing 259

However, the special connotations just observed are limited to the vocabu-
lary that is more clearly of Latin and Greek origin. Contrast the difference in con-
notations between expressions like automotive, capability, antidisestablishmen-
tarianism on one hand and target practice, royal pain, enterprise on the other.
Technically, both sets of words are borrowings; but only the first set has special
technological or educated, if not pedantic, connotations.
There is reason to believe that the special connotations are directly attributa-
ble to the fact that the words are borrowed from Greek and Latin, scholarly pres-
tige languages in which many of the words had already been used with special
scholarly or technical connotations. The connotations of Graeco-Latin borrowings
therefore are exactly what one would expect.
In addition, recall that in languages like German, it is adaptations like Fernspre-
cher and Rundfunk which often have special technical or officialese connotations,
while adoptions like Telephon and Radio belong to the ordinary, non-technical
lexical layer of the language. Facts like these suggest that in many cases it is more
the sphere of usage than the origin of a particular lexical item which determines
its special connotations. Note in this regard that even in English, Anglo-Saxon
collocations like black hole have very special connotations if used in astrophysics,
and so do terms like borrowing, when employed in historical linguistics.
Finally, recall that many of the Graeco-Latin technical borrowings in English
are restricted to very specialized uses.
The common belief of English speakers that the adoption of vocabulary is
desirable, in that it “enriches the language”, thus, is difficult to justify on purely
linguistic grounds. But ultimately that may not be relevant. As in many other cases
we must remember that most speakers are not linguists. They are free to ignore
what makes sense to linguists and instead to act according to their own beliefs,
however naive these may appear to linguists. If, then, most English speakers are
persuaded that enlarging their vocabulary is a good thing, they can be expected
to behave accordingly and to adopt foreign words at a rate that far exceeds that
of other languages. Moreover, the linguistic nationalism of other languages is just
as much based on irrational beliefs, not on purely linguistic or structural facts.
Chapter 9: L exical change and etymology
The study of words
Good words are worth very much, and cost little.
(George Herbert, Jacula prudentum.)

What do you read, my lord?

Words, words, words.
(William Shakespeare, Hamlet II.191–2)

1 Introduction
The preceding chapters have emphasized the processes of change, detailing the
various forces that can bring about change in different domains of a language.
We have presented the types of change in terms of the different components of
grammar (phonology, morphology, syntax, semantics). Yet there is at least one
factor that goes beyond these different grammatical components and thereby
unifies these changes. All can have a profound effect on all the numerous and
varied elements that together make up a language’s lexicon.
A very basic, almost trivial, type of lexical change comes about through
regular sound change. When sound change affects a sound or class of sounds,
clearly the pronunciation of lexical items containing those sounds will undergo
a change. For instance, when the Latin word for ‘father’, pater [pater], became
French père [per] as the result of regular vowel changes including the regular loss
of intervocalic t, the lexical item for ‘father’ changed.
Analogical change can likewise bring about lexical change. For instance,
the leveling of the sibilant : [r] alternation in OE frēosan [frēozan] : frēas
[frēas] : fruron : (ge)froren in favor of -s- [z], ultimately yielding Mod. Engl.
freeze : froze : frozen (Chapter 5, § 2.1), may be said to have produced a change in
the phonological behavior of this lexical item.
More interesting are developments involving rhyming formation and related
processes discussed in Chapter 5, § 3.1, for these may introduce new words to the
language. Consider the English words in (1). All of these end in -ag, and all have
something to do with ‘slowness, fatigue, or tedium’. But as noted in Chapter 5, the
morphological composition of such forms is difficult to determine. If we were to
say that the meaning ‘slowness, fatigue, or tedium’ is associated with -ag, what
then would be the meaning of dr-, f-, fl-, l-, or s-?

https://doi.org/10.1515/9783110613285-009
 Introduction 261

(1) a. drag ‘lag behind’ < ME draggen < OE dragan or ON draga ‘drag, pull’
b. fag ‘exhaust, weary, grow weary’, presumably < ME fagge ‘droop’
c. flag ‘hang limply; droop’, probably of Scandinavian origin, from a word
akin to Old Norse flögra ‘flap about’
d. lag ‘fail to keep up; straggle’ < earlier English lag ‘last person’, ME lag-
‘last’, possibly from Scandinavian
e. sag ‘sink; droop’ < sixteenth century Engl. sacke, ultimately probably of
Scandinavian origin, compare Swed. sacka ‘(to) sink’.

The words in (1a–d) do in fact go back to earlier forms which already contained
-ag; so for these we might claim that the phonetic similarity does not result from
change, but is simply accidental. However, the matter is not so simple if we look
at the semantics. Only two items had earlier meanings compatible with ‘slowness,
fatigue, or tedium’, namely (1b) and (1d). Semantically, then, all the listed words
have undergone a change which has brought them closer together. And this fact
suggests that the relationship is not entirely accidental. The smoking gun (if we
can use that term for words with such meanings), which virtually proves that the
similarity is not due to chance, is seen in (1e), a word whose final consonant has
actually changed from [k] to [g]. Other words ending in [k], such as sack ‘bag’,
lack, crack have not undergone that change; the development in (1e) therefore
cannot be the result of regular sound change. The only explanation of the final
voiced -g of sag is that it came to be associated with the drag/fag/flag/lag “gang
of four” because of its meaning.
Thus an irregular, analogical change led to a change in a lexical item, and to
the strengthening in semantic coherence of a whole cluster of related lexical items.
These examples of lexical effects have started with well-understood items
whose history is easily documented. But there are vast numbers of other words
whose history we are uncertain of. What is the source of t-shirt, for instance? Is
it so named because it is shaped like the letter t, or like a golf tee, or is it an
abbreviation for t(ennis)-shirt? What about the incredibly common English lexical
item OK (also spelled O.K. and okay), which has spread practically all around the
world? Scholars have long been divided on the source of OK. Some consider it an
acronymic abbreviation for Old Kinderhook (a nickname of 1840 U.S. presidential
candidate Martin Van Buren); others see it as shortened from a jocular spelling
“oll korrect”, supposedly popular in the 1840 election; and still others treat it as
an Africanism that entered American English through contact with the usage of
African slaves in the South (see § 4.4 below for words that can be more clearly
derived in this manner).
The study of the origin of words is known as etymology. The first part of
this word comes from Greek étymon ‘true sense of a word’, so that etymology is
262 Lexical change and etymology. The study of words

the study of the true, i. e. original, forms of words. In a larger sense, etymology is
concerned with the history of words, how they arise, the factors that have affected
their ultimate shape and meaning, the semantic paths they have taken in their
development though time, and so on. Moreover, once we start exploring word
origins, the question arises, too, as to where various phrases and expressions –
idiomatic groups of words – come from. Why, for instance, do we say madder than
a wet hen, rarer than hen’s teeth, raining cats and dogs, or on a wild goose chase, to
take just a few animal-related expressions?
Etymology is fascinating and clearly has great popular appeal, as the large
number of books on word and phrase origins indicates. And many etymologies –
or facts showing the absence of an etymological connection among words – do
make for interesting trivia. For instance, although this may be hard to believe, the
word canary ultimately derives from Lat. canis ‘dog’. The birds known as canaries
bear their names because they originally came from the Canary Islands. These
islands, in turn, were named in Latin after the large canes (pl.) ‘dogs’ found there.
On the other hand, there is no etymological relationship between canary and the
French word canard ‘duck’, even though one might at first glance think of such
a connection; once you have canary and canard, can- in bird names seems like a
promising morphological division (“birds of a feather …”). Actually, though, the
can- of canard apparently is a syllable that onomatopoetically reflects the duck’s
quack, and thus is unrelated to canary – even though it is conceivable that for
speakers of French, in which the word for canary is canari, a folk etymological
connection (see Chapter 5, § 3.2) between canari and canard might seem right.
(English borrowed canard in the extended, metaphorical meaning ‘hoax’ ← ‘mali-
cious story’.)
Far from being just a matter for trivial pursuits, etymology is in a real sense
the basis of historical linguistics, for establishing the origin of a word is crucial to
understanding the changes it has undergone and the factors that have influenced
its development. Without a well-worked-out account of how bead could shift from
the meaning ‘prayer’ to ‘small roundish glass or ceramic object’ (see Chapter 3,
§ 2.2), we could not really establish its etymology, nor could we be sure about the
effects of sound changes such as Grimm’s Law in Germanic without first positing
etymologies for various lexical items that connect them with cognate words in
other languages (e. g. father as being from the same source as Latin pater, or ten
from the same source as Greek déka). Thus, once a good many well-established
cases are examined, working out the general principles that govern language
change can be undertaken. And it all starts with etymology.
Most of the changes discussed so far in a sense do nothing to alter the basic
inventory of lexical items. Whether ‘father’ is pronounced [pater] or [pɛr], whether
the past participle of ‘freeze’ has an -r- or a -z-, or whether we say sag or sack to
 Introduction 263

convey the meaning ‘sink, droop’, there is still a single form linked to a given
meaning, and thus no net gain in the number of lexical entries.
Other types of change, as we have seen in the chapters on semantic change
and borrowing, can significantly affect the lexical inventory. In this sense ety-
mology is also the study of the sources of words, of the word-formative resources
that languages have, and of how speakers use these resources. And the study of
etymology necessarily involves us in a study of lexical change, how words rise
and fall through time.
For instance, borrowing almost invariably adds to the vocabulary, even in
cases of calquing. The only exception is represented by loan shifts, which broaden
the meaning of existing words but ordinarily do not introduce new ones. Occa-
sionally, even loan shifts can enrich the lexicon. Consider the case of Modern
Icelandic síma ‘telecommunication’, used to translate Engl. wire, cable = telegram.
In this case, the loan shift was accomplished, not by expanding the meaning
of a word already in Modern Icelandic use, but by resurrecting an obsolete Old
Icelandic word of somewhat obscure signification which could be guessed at as
meaning something like ‘cable’ or ‘rope’.
Conversely, cultural and social changes may lead to obsolescence, the loss of
words, sometimes on an impressive scale. When Horatio tells Hamlet (Act I.230)
that he did not see the face of Hamlet’s father’s ghost because he wore his beaver
up, we need a textual note or a dictionary to tell us that a beaver is a term for
the visor on a helmet. If more of us wore suits of armor, this meaning would be
more commonly known. Consider also what happened to terminology such as
thill ‘shaft to attach an animal to a cart’ when the horse and buggy were replaced
by the automobile.
The social, cultural, and technological factors that lead to the obsolescence
of words at the same time may also necessitate the development or coinage of a
great deal of new vocabulary, not necessarily through borrowing. This phenom-
enon is evident in the lexical explosion occasioned by the introduction of the
internal combustion engine or, more recently, the advent of the computer.
In the sections that follow, we examine processes by which the lexical store
of a language can be enriched, by considering the etymology of new lexical items
and expressions.
264 Lexical change and etymology. The study of words

2 Coinage
In the preceding chapter we have seen many examples of borrowing of technical
terms, whether by adoption or by adaptation (calquing). Compare sets like the
one in (2) – perfect examples of the fact that speakers tend to borrow the words
which go along with the new artifacts or ideas that they adopt.

(2) English French Spanish Mod. Greek German

telephone téléphone teléfono tiléfono Telefon
Fernsprecher
Icelandic Hindi
tal-síma (ṭēli)fōn
dūrbhāš

What we have not examined is the question of how the language in which the
artifact or idea first originated acquired the new term to designate it. Clearly,
that language cannot resort to borrowing, but must create the term from its own
resources. In the case of telephone, this was accomplished by combining the ele-
ments tele- ‘far’ and phone ‘speech, voice; speak’ which had entered the language
through earlier borrowings from Greek (via Latin). In principle, this is not differ-
ent from the process that gives rise to the German alternative word, Fernsprecher
= fern ‘far’ + sprecher ‘speaker’, or the similar Hindi dūrbhāš = dūr ‘far’ + bhāš
‘speech, voice’. But while the German and Hindi words are re-creations, modeled
on the English form, the English word was an original creation, a new coinage.
There may have been a French model form predating telephone in English, though
not necessarily with the same meaning, but the account given here would be the
same for the French prototype.
In the case of telephone, the new word came about by combining the ele-
ments tele and phone in a productive morphological pattern through something
like four-part analogy; compare similar formations such as tele-scope, tele-graph,
or micro-phone. The same sort of process is responsible for many other neolo-
gisms. The word neologism itself is a neologistic coinage from the Greek elements
neo- ‘new’ and log- ‘word’ that had previously entered the language, and thus
literally means something like ‘new word’. Again, the French may have beaten the
English to it, with their earlier néologisme, but once again, the process described
here would be the same for either coinage.
Other neologisms include hard disk, dark matter, and black hole. Compared
to words like telephone, these expressions are much more mundane, in that they
do not significantly draw on the Graeco-Latin elements that are more typical of
technical terminology (see Chapter 8, § 5). But that does not diminish their being
 Coinage 265

new coinages, on a par with more technical-sounding ones like disk operation
system or magnetic resonance imaging.
In addition to four-part analogy, two other analogical processes frequently
are used for creating neologisms: blending (as in brunch) and backformation (as
in orientate). Compare the discussion in Chapter 5 and see also below.
Coinage may be accomplished by a large variety of other changes. In some
cases, something akin to loan shifting is involved, namely a simple semantic
extension. This has been the case for instance when horse-and-buggy terms like
wheel and tire came to be used in reference to automobile parts. Many slang terms
involve extension (see § 4 below), often utilizing the “part-for-the-whole” strategy
seen in Chapter 7, § 5.1, as with wheels as a term for car, skirt for woman, and suit
for businessman, lawyer, or administrator (i. e., someone forced to wear a suit to
work). And just as in some cases loan shifts may be accomplished by resurrecting
obsolete words of somewhat obscure meaning (see Icel. síma above), so neolo-
gisms sometimes are created by adopting a word of obscure significance.
This has been the case for the word quark, used in physics to designate a set of
elementary particles. The word was adopted from an enigmatic passage in James
Joyce’s Finnegan’s Wake: Three quarks for Muster Mark! (The whimsicality that led
to selecting this word is carried further with the attributes that distinguish differ-
ent kinds of quarks – top, bottom, strange, charmed, up, and down.)
The processes of coinage are seen quite vividly in the names given to new
products. Kleenex is clearly built on clean, Jell-O on the verb jell or the noun jelly
(alternatively, on gel and gelatin but with a distinctive spelling), and Xerox on the
Greek xero- for ‘dry’. But the source of the final parts (-ex, -O, -ox) is not entirely
clear. Perhaps they represent extensions from similar pieces in other words (such
as the -o suffix in a word like kiddo). Many of these well-known product names
have spread into general usage, as a generic term for the type of product they refer
to, e. g. kleenex for any type of facial tissue, jello for any type of flavored gelatin
dessert, xerox for any type of xerographic reproduction. Despite strenuous objec-
tions of the manufacturers, who have paid huge sums of money to some advertis-
ing team to think up the name for their product, such developments are difficult to
prevent; they are paralleled by many other, similar extensions in sphere of usage.
Coinage can also involve raw creation, sometimes for the sound or even visual
effect alone. For instance, the product name Kodak is said to have been created
because the letter < k > is somewhat rare in the spelling of English words, and
thus a word with an initial and a final < k > would make a striking – and thus pre-
sumably lasting – visual impression. Similarly Kleenex with an initial < K- > and a
medial < ee > has its own distinctive look visually, and the initial and final < x > in
Xerox is eye-catching too. Thus the final parts of these product names may simply
represent visually motivated creative word invention.
266 Lexical change and etymology. The study of words

More usually, perhaps, a striking sound impression is involved in such raw

creations. The word googol for ‘the number 10 raised to the power 100’ was created
out of the blue by Milton Sirotta, the nine-year-old nephew of the American math-
ematician Edward Kasner in the mid-twentieth century, apparently because it
sounded good! Note also that once googol was part of the language, it was easy to
create the word googolplex, for ‘the number 10 raised to the power googol’, as a
BLEND of googol and the -plex of words like duplex. In slightly different shape the
word morphed into the name of a popular internet search engine.
Many so-called NURSERY WORDS in virtually all languages involve creations
that are commonly believed to have a “baby-appropriate” sound structure, e. g.
Engl. Momma, Papa, Daddy, Nanny, even nicknames like Bubba or Sis(sie); Goth.
atta ‘father’ (from which the name Attila, lit. ‘daddy’, was derived), aiþei ‘mother’;
It. papà, mamma, bambino ‘child’; Turkish baba ‘father’, ana/anne ‘mother’;
Hindi bāp, mā̃, baččā ‘child’; Georgian dede ‘mommy’, mama ‘father’; and so on.
The structure, sounds, and often also the connotations of these words suggest that
they come from “baby talk”, i. e., the words and meanings which adults assign to
the early babbling of infants (see Chapter 14, § 1). Like other items that occur in
this kind of language, the words tend to have reduplicated syllables (as in pa-pa).
And the predominance of the vowel a and of the relatively unmarked consonants
p/b, t/d, m, and n is a common feature of early babbling. In a sense, then, these
words represent a type of borrowing from “baby talk” into regular adult usage
with babies and small children. Note further that despite modern American
and European sensibilities as to what is “appropriate” sound structure in these
words, there is nothing universal about the attachment of particular meanings
to particular combinations of sounds. Most of us might believe that words of the
type mama must refer to ‘mother’, and tata, dada to ‘father’; and in fact, one
famous linguist has claimed universal validity for this generalization. But Geor-
gian dede ‘mommy’, mama ‘father’ show that the generalization is not universally
true.
Many other processes can be drawn on to coin new words. Consider the word
car in the meaning ‘automobile’. At first glance we might be tempted to assume
that it originated by the same semantic transfer as other automotive terms, such
as wheel and tire above. But the common use of the term motor car in the early
period of the automobile suggests that car instead results from ellipsis. Once
motor cars had become the normal means of transportation, the element motor
could be considered redundant and therefore was deleted.
Many modern words have originated in this fashion. Consider the examples in
(3). Because of their derivations from place names and personal names, examples
of the type (3a) and (3b) are often singled out for special treatment, as toponyms
(3a) and eponyms (3b), respectively. In some cases, such developments can take
 Coinage 267

strange turns, as the earlier discussion of canary indicates. Moreover, as (3c) illus-
trates, words that have arisen via ellipsis, such as the jeans and canary of (3a),
may enter new compounds which may be subjected to another round of ellipsis
(recall how googol, once coined, could spawn googolplex).

(3) Modern English Source

a. damask Damask cloth (= cloth from Damascus)
(blue) jeans Gene (= Genoa) cloth
denim Fr. serge de Nîmes = cloth from Nimes
canary a bird from the Canary Islands
b. bowler Bowler hat (after a famous hat maker)
hansom Hansom car (after a famous carriage maker)
c. jeans blue jeans
canary canary yellow (after the coloring of canaries)
gum chewing gum

However, the terms toponym and eponym cover much wider territory, referring
not only to the results of ellipsis, but also to other secondary uses of place or per-
sonal names, mainly by way of metaphoric extension, as in (4). Thus, a maverick
is somebody who is like the rancher Maverick in not going along with prevailing
behavior. An even bolder extension of personal names can be seen in the nomen-
clature for measurements prevalent in physics, such as ohm, newton, and volt,
all named after famous physicists of the past. A very different secondary use of
names is found in the selected use of personal names, generally ones that are
or once were quite common, in various colloquial or even slang expressions, in
principle as a generic word for ‘human being’, ‘man, male’, ‘woman, female’, but
often with humorous or nasty overtones; see the examples in (4b) and (4c), where
maudlin reflects an older British pronunciation of Magdalen. The use of john as a
colloquial expression for ‘toilet’ probably owes its origin to such a development.

(4) a. maverick name of a nineteenth-century Texas rancher who unlike

other ranchers did not brand his cattle
dunce John Duns Scotus, a philosopher whose followers were
opposed to the humanism of the Renaissance and thus failed
to grasp the significance of the new movement
maudlin Mary Magdalene, who was iconographically portrayed as
weeping over the death of Jesus
b. jack-of-all-trades
jack-o’-lantern
Billy Goat
268 Lexical change and etymology. The study of words

John Doe
Jane Doe
c. Joe Blow
Johnny-come-lately
jack-ass
jenny(-ass)
Plain Jane
john (customer of a prostitute)
Dick, Peter (American English colloquial/slang/vulgar terms for the male
organ; note also Brit. Engl. Willy)

Moreover, while examples of the type (3) can be accounted for as reflecting ellip-
sis, there are many other cases of lexical shortening for which such an explana-
tion is much more difficult, or even out of the question. For instance, examples
like (5a) can with some stretch of the imagination be considered elliptical, elim-
inating the “redundant” elements show or filling. But what about examples like
(5b): Is the element ham here really redundant? If so, what is its meaning? Surely
it is not ‘ham’! Now, it is possible to argue that cheese burger is a blending of
cheese and hamburger. But that explanation does not account for the similarities
between (5a) and (5b). In both cases, a somewhat lengthy compound is made
shorter, more manageable, through some kind of reduction. It is this reduction
that seems to count most; and the method by which the reduction is accom-
plished is of much lesser significance. Reanalysis of words like cheeseburger, or
backformation, in turn has given rise to the word burger. (Note that hamburger in
origin is a toponym, going back to the German expression Hamburger Rundstück
‘round piece of Hamburg’, a bun (filled with a slice of chopped meat) prepared à
la Hamburg, i. e., as it would be prepared in the city of Hamburg.)

(5) a. TV host TV show host

gas station gas filling station
b. cheeseburger cheese hamburger
tacoburger taco hamburger

Some such shortenings can be quite extreme, and it is often difficult to figure out
how their parts end up going together to make the new whole. For example, the
Columbus Dispatch in the early 1980s reported on a dog owner who was repeat-
edly in violation of the law for some minor offense (e. g. not cleaning up after
the dog) but routinely ignored the tickets that were issued, and thus acted like a
scofflaw. The newspaper article referred to the dog as a scoffdog, apparently an
extreme shortening for something like scofflaw-owned dog. Obviously, the dog
 Coinage 269

was not the scoffer! Similarly, some grocery stores in Columbus have an aisle sign
for Baked Needs, where the intent is ‘items needed for (making) baked goods’,
and thus we see a shortening for (the admittedly somewhat clumsy) Baked Goods
Needs, even though in the resulting shortening, it would seem that the needs are
baked! Perhaps all that is needed is a vague associative reference to the words
underlying the shortened form.
The view that the basic motivation for words like gas station and cheeseburger
lies in a tendency toward abbreviation is supported by a wide variety of other
abbreviatory developments. Consider the examples in (6). Cases of the type (6a)
still bear a certain similarity to those in (5), in that the abbreviated version (such
as phone) is a meaningful component of the longer version (tele-phone). But the
shortened forms in (6b) cannot possibly be explained by some kind of morpho-
logical reduction. Reduction here operates entirely on phonological principles,
commonly by eliminating all but the accented syllable of the word, as in fridge,
though accented syllables can also be eliminated, as in the slang forms in (6c) and
the ordinary English word in (6d). The outcomes of this reduction of phonological
material are referred to as clippings.

(6) a. phone telephone

auto automobile
b. mike or mic microphone
lab laboratory
fridge refrigerator
frank frankfurter (⇐ Germ. Frankfurter Würstchen
‘little sausage à la Frankfurt’)
c. ’rents parents
’za pizza
d. bus omnibus (⇐ Lat. omnibus ‘for all’, where -bus is the dative
plural ending)

In some cases, one and the same form may be subjected to clipping in more than
one direction. For instance, taxicab (itself a shortening for taximeter cabriolet) has
yielded both taxi and cab as clipped forms with the meaning ‘taxicab’.
Even more daring reductions can be seen in the common use of acronyms,
such as US = United States or MRI = Magnetic Resonance Imaging. A special sub­
type of such acronyms arranges the basic words in such a way that the combi-
nation of their initial letters or sounds is a pronounceable word, as in laser =
Light Amplification by Stimulated Emission of Radiation or scuba = Self-Contained
Underwater Breathing Apparatus. Noteworthy, too, are acronyms based on suc-
cessive syllables of a single word, such as TV = TeleVision or PJs = PaJamas. Inter-
270 Lexical change and etymology. The study of words

estingly, abbreviations of the latter type are not significantly shorter in terms of
their pronunciation; they are shorter only in spelling. (Some scholars reserve the
name acronym for structures like laser, which are pronounced more like ordinary
words, and use the term initialism to distinguish abbreviations of the type US,
which are merely sequences of conventional letter names.)
In modern literate societies, acronyms generally operate in terms of the
written medium. Moreover, in fully alphabetic writing systems, they usually select
single initial letters. Examples of the type Germ. Flak = FLugzeug-Abwehr-Kanone
‘aircraft-defense-cannon’, operating with the initial consonant group of the first
element of the compound, are less common. In pre- or non-literate societies,
acronyms seem to be more naturally based on initial (or final) syllables. Thus, in
the oral tradition of indigenous Sanskrit grammar, going back to about the sixth
century B.C., finite verbs can be referred to as tiŋ, based on the initial (tip) and
final (mahiŋ) elements in an oral listing of finite-verb endings. And as the discus-
sion in Chapter 3 has shown, syllabic acrophony of this type played a major role
in the development of syllabaries.

3 Proper names: A case study in lexical origins

Proper names play a role in lexical change in providing a source for new ordinary
(or “common”) nouns, as in the cases of maverick or john discussed above. Yet
where do proper names come from in the first place? Many, it turns out, come from
a specialized use of common nouns, though there are other sources as well. As
an extended case study in the processes by which words arise, we examine here
the origins of various types of proper names and the occasional improper (so to
speak) uses to which proper names are put.

3.1. Names of peoples and places: A good place to start is with names of groups –
what are often termed “peoples”, and by extension the names of the countries
these peoples are located in.
All human peoples have names for themselves, as well as names for others,
and often consciously distinguish themselves from other groups in terms of their
names. Quite frequently, the group name is simply the word for ‘people’ or ‘human
being’, in the language of the group – making an implicit contrast between “us”,
the group members = human beings, and “others”, the nonmembers = nonhu-
mans or nonpeople. This ideology underlies names of peoples all over the world,
including American Indian tribal names such as the Illinois, the Lakota, and the
Kiowa; Uralic names, such as Mari (the Cheremis name for themselves), Nenets
 Proper names: A case study in lexical origins 271

(the Yurak self-designation), Komi (the name of the Ziryenes for themselves), Hanti
(the Ostyak self-label); the Munda ethnic name Kurku; and the Santal self-desig-
nation Hoṛ. The German self-label Deutsch looks like it belongs here, too, since its
original meaning is ‘of the people’. But in this case it is more likely that the term
originated in the medieval period when Latin was the dominant language of edu-
cation and when diutisk > diutsk > deutsch referred to the language of the people,
in the sense of “vernacular language” (as distinguished from the Latin prestige
language). A similar use is found for the Old English cognate þeodisc.
Related to the common self-definition as ‘the (real) people’ is a traditional
tendency to draw a distinction between one’s own group as speaking a real lan-
guage and others as incapable of doing so. A well-known example is the Ancient
Greek use of bárbaros (the source of Engl. barbarian) to refer to non-Greeks. To
Greek ears, other languages sounded like an inarticulate stammering, bar bar.
(The Latin adjective balbus, (roughly) cognate with Greek bárbaros, means ‘stam-
mering’.) A term barbara- with the same meaning is also found in Sanskrit, along-
side a word mlēččha- which no doubt, too, is intended to characterize foreigners
as only capable of producing ugly sounds (ml- being an unusual, and thus most
likely ugly-sounding, initial cluster in Sanskrit). Interestingly, when the Romans
conquered Greece they took over, along with Greek culture, the term bárbaros ⇒
barbarus, but because of their great respect for just about everything Greek, they
could not use the word to refer to the Greeks. So from that point on, the Romans
considered everyone a barbarian, except themselves and the Greeks.
Very commonly, names given to familiar groups by others bear negative con-
notations similar to bárbaros. For instance, Eskimo has been traced (via Spanish
and French) to Micmac eskameege ‘raw fish eaters’. Nemets, the name for Germans
among Slavic speakers, literally means ‘mute’; ‘unable to speak a real human
language’. (The name Slav, by contrast, may either be derived from slovo ‘glory;
word’ or result from a folk-etymological connection with slovo after a dissimila-
tory change had altered original *svobēn- ‘one’s own; our own [people, language]’
to *slobēn-.) Similarly, the word Apache derives from the Zuñi word for ‘enemy’,
and Comanche, from the Ute word kima ‘stranger’.
In some instances, the name of rulers has come to be used for the ruled
people. Such was the case with the names for French and France, named after the
Germanic Frankish peoples who conquered Gaul in the sixth century A.D. and
built a powerful empire there for some 400 years.
In many of the above examples, the group’s self-designation differs markedly
from our conventional name for the same group. (Consider for instance Nenets vs.
Yurak, or Deutsch vs. Nemets.) This conflict between self-designation and name
assigned by others is a very widespread phenomenon. For instance, the Greeks
nowadays refer to themselves as [élines], reflecting (via regular sound changes)
272 Lexical change and etymology. The study of words

an Ancient Greek name, Héllēnes, that was used originally (e. g. in the Iliad) for a
group from Thessaly (in central Greece) and later extended (e. g. by the Ancient
Greek historian Herodotos) to designate Greeks in general. Occasionally modern
Greeks may also use the term [rómii], lit. ‘Romans’, reflecting the Byzantine her-
itage of Greece and the fact that Byzantium once was the Eastern Roman Empire.
Nonetheless, speakers of English and most languages of Western Europe refer
to them as Greeks (or the equivalent thereof, e. g. French grec, Spanish griego),
reflecting the term Graeci used by the Romans for all Greeks, though the Ancient
Greek source, Graikoí, was originally applied just to one group, a tribe in the
northwest of Ancient Greece.
Similarly, Germans are referred to as aleman, lit. ‘Alemannic’, in French and
Spanish, and as Saxon in the Carpathian area of Romania. In addition, of course,
they are referred to as German in the English-speaking world and in countries
originally dominated by England, such as India, where we find Hindi ǰarman.
The original English word for ‘German’ was Dutch, based on Germ. deutsch, and
was used in reference to the Germanic-speaking inhabitants of Germany at a
time when that country included the Netherlands. But the Dutch that the English
had the greatest contact with came from the Netherlands. After the Netherlands
became independent in 1648, the English naturally used Dutch to refer to the
inhabitants of that country – which left them without a name for the inhabitants
of the remainder of Germany. In this situation they resorted to a word used by
the Romans to refer to their northern, Germanic neighbors. But while this choice
fixed the immediate problem, it turned out to cause difficulties for future stu-
dents of historical linguistics who may find the terms German and Germanic to
sound confusingly similar. (The Germans have no such problems, distinguishing
deutsch from germanisch.) A further complication results from the fact that the
English term Dutch has survived here and there as a designation of some groups of
German, not Dutch, origin – most notably in reference to the Pennsylvania Dutch.
Another case with similar complications is that of India and Indian. The
Indo-Aryans who centuries ago occupied what is now India referred to themselves
as ārya-, from the Indo-Iranian self-designation meaning ‘noble’ (see Chapter 2,
§ 3.10). The outside world has come to use a different name, of the type repre-
sented by English India (the country) and Indian (the inhabitants). This termi-
nology derives from Greek Indós, an adaptation of the Old Persian name hindu
for a river that is called sindhu in Sanskrit. The Indus River frequently formed the
boundary between Iranian and Indo-Aryan peoples, so that the Iranians could use
hindu to refer elliptically to the country and the people living ‘beyond the Indus’.
(The correspondence Skt. sindhu- : OPers. hindu- reflects regular sound changes.
The absence of initial h- in Greek reflects the fact that the Ionian Greeks living
closest to the Iranians had lost their aitches, just like Cockneys.)
 Proper names: A case study in lexical origins 273

The Iranian term hindu has become widely known with a different connota-
tion, referring to the most widespread indigenous religion of India. This connota-
tion developed when Persian-speaking Muslims conquered much of South Asia
and began to use the term hindu to refer to the majority population’s religion. At
the same time, they retained the more original meaning of hindu in the name for
the country, Hindustan, a term nowadays commonly used by Indians and other
South Asians to refer to India. (The official name of the country is Bhārat(a), ‘the
land of the descendants of Bharat(a)’, the mythological ancestor of an important
early royal dynasty.)
The use of the word Indian as a term for the indigenous peoples of the Ameri-
cas is an interesting misnomer resulting from Christopher Columbus’s belief that
he had reached India after crossing the Atlantic Ocean. In English, the “politi-
cally correct” term for these peoples now is Native or Indigenous Americans; and
given the possible confusion between Indian1 ‘citizen of India’ or, at an earlier
time, ‘inhabitant of South Asia’, and Indian2 ‘Indigenous American’, this choice
of terminology has a lot to recommend it, even if American Indians themselves
generally have not embraced it. (German ingeniously differentiates between the
two types of “Indians” by referring to the first group as Inder, and the second as
Indianer.) At the same time, the term Native American causes difficulties, since
members of virtually all human races may be born in the United States and in
that sense be native Americans (where America = US). Moreover, US-born whites
like to distinguish themselves as native Americans in contrast to the recent large
influx of non-white immigrants – an ironic twist of events, since at an earlier time,
whites used the term natives to refer to indigenous non-white peoples around the
world. A further wrinkle in this complex, even convoluted, situation is that, as
noted above, many “Indigenous Americans” prefer to be called Indians or Amer-
ican Indians.
As in the use of Indian as a term for the indigenous peoples of the Ameri-
cas, group names can be misapplied or perceived as misapplied, sometimes with
serious political consequences. A case in point are the terms Macedonia and Mac-
edonian. In ancient times, the Macedonians (whose name possibly derives from
Gk. makednós ‘tall’) spoke a language that may have been a sister-language of the
Hellenic (Greek) branch of Indo-European – or even a separate branch with close
affinities to Hellenic. Under Philip of Macedon and especially his illustrious son
Alexander the Great, Hellenistic culture and Greek language were introduced, and
Macedonia became part of Greece. Incursions of the Byzantine period brought
speakers of Southern Slavic who settled especially in the north of the area and
who have referred to themselves and their language as Macedonian for well over
a hundred years now. Politically, part of Ancient Macedonia now is a province of
Greece, with Greek as its official language. Another part was a state in the former
274 Lexical change and etymology. The study of words

Republic of Yugoslavia, but became independent in 1991; as a country it refers to

itself as the Republic of Macedonia and calls its Slavic majority population and
its language Macedonian.
There is thus no straight-line connection between the ancient Macedoni-
ans and the modern country or province of Macedonia or their speakers – other
than the name. Yet the terms Macedonia and Macedonian have become potent
symbols in a bitter political struggle between Greece and the somewhat newly
formed (Slavic) Republic of Macedonia, with Greece claiming that it alone is enti-
tled to use the name Macedonia. “What’s in a name?” one might ask. The answer
is “Plenty!”, especially if it is invested with the right sort of emotional and political
appeal.
Just as the proper name Maverick was the basis for a common noun maverick,
and names like John or Peter have taken on other meanings as common nouns,
so too the names of various groups may be used “improperly”, often with neg-
ative connotations. For instance, the word slave is ultimately (through a Greek
rendering of the name) related to Slav (see Chapter 2, § 3). The word gyp derives
from Gypsy, a misnomer related to Egyptian, based on the mistaken belief that the
Gypsies, who originated in India and call themselves Roma(ni) (< Prakrit ḍoma),
came instead from Egypt. The word cannibal comes from Christopher Columbus’s
rendering Caníbalis of the Arawak ethnic name caniba for the Carib peoples of
Cuba and Haiti, who, it was alleged, ate other human beings – an allegation not
substantiated by evidence. Note also such potentially offensive usages as to welsh
(or welch) on a bet or to jew someone. And the list goes on.
In western society today, with an increasing awareness of cultural differences
and a resurgence of ethnic pride, many of these negative terms are disappearing,
such as the verb to jew. For others terms, alternative words are being proposed,
such as Inuit, originally the name of a subgroup, for all Eskimos. As in other case
of taboo, the success of these attempts can be expected to be mixed. Yet some
names, such as Slavic Nemets for ‘German’, are so deeply ingrained and their
literal, etymological meaning so much lost in history that change is unlikely to
take place.
Moreover, those who find the misuse of ethnic names regrettable can take
some solace in the existence of relatively innocuous terms or phrases derived from
ethnonyms, such as Dutch treat, or It’s Greek to me. Interestingly, the Modern
Greek equivalent is mu íne kinézika, literally ‘It is Chinese to me’; and German has
Das kommt mir spanisch vor ‘That strikes me as Spanish’.

3.2. Names of persons: What is true for group names also holds for personal
names. Names of individuals show a variety of sources, in terms of both the type
of source and the language.
 Proper names: A case study in lexical origins 275

Most names were once meaningful words that presumably came to be applied
to some individual in recognition of some defining quality, and thus resemble
nicknames. For instance, the widespread name Paul (Span. Pablo, Ital. Paolo,
Russ. Pavel, Mod. Gk. Pávlos, etc.) is ultimately from Latin paulus ‘small’, the
name given to Saul of Tarsus when he converted to Christianity, and thus surely
a characterizing epithet at first. Similarly, Philip is ultimately from Greek phíl-ip-
pos ‘lover of horses’, again no doubt an epithet that originally characterized the
designee or the hopes that his parents had for him when he was born.
Biblical names commonly are of Hebrew origin. The name Adam is from the
Hebrew for ‘red’ (perhaps originally designating a characteristic skin color), David
is from a nursery word for ‘darling’ (later ‘friend’), Joseph originally means ‘may
Jehovah add’ (with an understood object ‘children’), and Mary is possibly from an
expression meaning ‘desired, longed-for [child]’.
From Germanic come names such as Edward (Old English Ead-weard ‘rich
guardian’) and Robert (Old English Hreod-beorht, Old High German Hrode-bert,
literally ‘bright of fame’). But as noted in Chapter 2, § 3, many names of Germanic
origin have come to English via French after the Norman conquest of 1066.
The Romans used ordinal numbers as the basis for some proper names, e. g.
Quintus (literally ‘the fifth [child or son]’) or Sextus (‘the sixth’). In rural England
and the Southern United States, names such as Easy or Early are attested, charac-
terizing the nature of their bearers’ birth. Christian virtues became names such as
Faith, Hope, Charity, in America at least, among settlers in Puritan New England.
The month-names April, May, and June are the basis for some women’s names,
originally perhaps motivated by the joyous, spring-like connotations associated
with these months.
In more recent years, people have found even more daring or, some would
say, outrageous, names for their children or for themselves. For instance, girls
have been named Georgia after the U.S. state of the same name; many English
girls are called Chelsea (originally after London’s artists’ quarter); and the British
actress Catherine Oxenberg called her daughter India. Note also the American
author Tennessee Williams, the fictional character Indiana Jones, and a New York
author named Gary Indiana (after the town Gary, Indiana). Some of these names
may be motivated by emotional attachment to the state or country; others simply
by a desire to be different.
A special case is the use, in the rural U.S. South of the 1930s, of such names
as Syphilis or Gonorrhea, which were chosen because they “sounded good” by
people not aware of their medical meanings. A similar, more widespread phe-
nomenon is the use of alternate, supposedly more refined or melodious spellings.
For instance, the former U.S. president Lyndon B. Johnson owed his first name to a
respelling of Linden because the use of y and o was considered more “euphonic”.
276 Lexical change and etymology. The study of words

A caricature of this convention is found in the jocular spelling of a very refined

dog Fido as Phideaux.
Just like common nouns, names can be borrowed, for their sound, their
meaning, or their symbolic value. Many, perhaps most, English names are bor-
rowed. This includes Germanic names like Henry or Robert, which came to English
via French; Celtic names such as Brian (from *bre ‘hill’); French names of Latin
and Greek origin, such as Renée, Pierre, and Philip; and Biblical names such as
Adam, David, Joseph, and Mary. Names of foreign origin that have not been nativ-
ized in English may be used as an indication of ethnicity. For instance, families
from Germany and Scandinavia may call their sons Hans and their daughters
Helga, African and Arabic names such as Abdul, Biko, Jamaal, and Kareem have
been especially popular among African Americans; and Hillel as a given name is
typically Jewish.
But with all names there is tendency toward bleaching of the original meaning
as the name takes on a life of its own, divorced from its etymological meaning and
ethnic associations. So you need not be of Anglo-Saxon ancestry to be named
Edward, or of Celtic background to be called Brian. And you don’t have to love
horses to carry the name Philip or to be religious to be called Faith; in fact, there
is nothing to prevent a Faith from being an atheist.
Just as names can be borrowed from other languages, so too can they be
borrowed internally, as it were, through extension from one lexical domain into
another. Thus, originally male names such as Billie or Jo have long been used
for females, and family names have been extended into use as first names, as
with Washington or Jefferson. In recent years, as more women have entered orig-
inally male-dominated professions, many women have requested that they be
addressed by more “male-sounding” variants or clippings of their given names,
such as Sam for Samantha or Chris for Christine. The motivation presumably lies
in the belief that male names make it easier for women to function in a male-dom-
inated environment.
Male names such as Sam and Chris are of course themselves innovations,
resulting from clipping and/or alterations in early child language. Sam and Chris
are clipped versions of Samuel and Christopher, as is Rob for Robert. But beside
Rob we also find Bob, with an assimilation of the initial consonant to the final
that is characteristic of young children’s speech. In addition, some people use
Bubba for Robert, although the word may also mean ‘brother’. Here again we find
features of early child language: A quasi-reduplicated form of Bob and/or assim-
ilation and simplification of the consonants in brother.
Hypocoristics of this sort often are quite old and reflect pronunciations or
phonetic variations that are no longer current. For instance, Dotty and Betty for
Dorothy and Elisabeth reflect an early pronunciation with [t], before [θ] was intro-
 Proper names: A case study in lexical origins 277

duced in the Renaissance on the Greek model (see Chapter 1, § 4). Ned most likely
reflects reanalysis of an original sequence mine Ed ‘my Ed’, at a time when the
possessive pronoun my showed an alternation between mine and my similar to
Mod. Engl. a : an. (A clear trace of this is found in the famous lines Mine eyes have
seen the glory of the coming of the Lord … from Julia Ward Howe’s “Battle Hymn
of the Republic”.)
Some names result from blends or coinages, and have a certain creative
quality to them. Popular names among African Americans such as Latisha or
Latrisha seem to derive from a blending of La- (from a source like LaVerne) and
t(r)isha (from a source like Tricia). The element La is found in many other names,
such as LaShawn and Latina.
Finally, names may carry significant social connotations. We have already
observed that certain names have a strong ethnic flavor. Social factors also play
a role in the popularity of certain names, with trends in naming often seeming
to catch on and spread like fads or slang expressions. The abundance of Jenni-
fers and Jasons among children born in middle-class America in the 1980s pre-
sumably reflects such a trend, a trend now decidedly untrendy, with the names
Jacob, Michael, Emma, and Madison being the most popular for babies born in
the 21st century. Often, common cultural elements can play a role. TV soap operas,
for instance, have been a source for the spread of many female names, such as
Crystal, with all its variant spellings (Kristal, Krystal, Cristal, etc.).
Family names also show varied origins and are moreover of relatively recent
origin. In fact, some cultures do not use them even today. They originally were
labels that further identified a person who was otherwise known just by given
name; they thus made it possible to specify which John or Jane was being referred
to.
Many are names of professions, such as Clark (a variant of clerk), Carpen-
ter, Goldsmith, Miller, Fletcher, Sawyer, and especially Smith – a popular source
for names in many other languages, such as Germ. Schmidt/Schmid/Schmitt, Fr.
Ferrier, Ital. Ferraro (and the derived form Ferrari), Sp. Herrero, Ru. Kuznetsov lit-
erally ‘Smith-son’, and Hung. Kovács. A very common alternative source consists
in place names, originally the place a person came from, such as Germ. Zumwalt/
Zumwald ‘at the forest’, (von) Hinüber ‘on the other side’, Engl. Milhouse ‘mill-
house’, Underhill, London, Hamburger, Welch/Walsh, Scott, and hundreds of
others.
Many family names are patronymic in origin, meaning ‘son of X’ or simply ‘of
X’, where ‘X’ is the father or grandfather. Compare English names with a final -s
(Adams, Richards, Roberts, etc.) representing the possessive suffix, or with a final
-son (Richardson, Robertson, Josephson, Adamson, Johnson, Davidson, etc.). In
some names, the -son is hidden by the spelling, such as Nixon or Dixon (originally
278 Lexical change and etymology. The study of words

‘Nick’s, Dick’s son’). English names in -son are usually of Scandinavian origin,
with -sen being the Danish version, as in Christiansen. Patronymics of Celtic origin
either have M(a)c- ‘son’ or O’, from OIr. ao < avi ‘grandson’, so that O’Henry liter-
ally would be ‘whose grandfather is Henry’. In addition note the rarer Fitz, as in
Fitzgerald, preserving the Old French ancestor of Mod. Fr. fils ‘son’.
Similar patterns are found in many other parts of the world, as in Hebrew
names with ben or bar (e. g. Ben Gurion, Bar Hillel) or Georgian names in -dze and
-shvili (such as Shevardnadze and Shalikashvili). The Modern Greek situation is
particularly interesting, for the form of patronymics differs regionally: -pulos for
families originally from the Peloponnesos, -iðis for those from Asia Minor, -akis
for those from Crete, and so on.
Metronymics are much rarer, but note the traditional Spanish pattern in names
such as Ramón Menéndez Pidal = Ramón whose father’s last name is Menéndez,
and whose mother’s last name is Pidal. (Traditionally, this causes difficulties for
illegitimate children, who would only have one last name, thus giving away their
origin.) In some areas of the world, metronymics are the norm. For instance, in
Gur languages of Burkina-Faso, Ghana, and Côte d’Ivoire we find names such
as Moses Kambou = Moses, son of Mrs. Kambou. (A father’s name is acquired at
initiation rites into adulthood, but that name remains secret.)
Finally, as with first names, some family names are originally epithets or nick-
names, originally reflecting some noteworthy, often physical, defining character-
istic, e. g. Germ. Schwarzkopf (literally, ‘whose head is black’); Engl. Whitehead,
Armstrong, Russell (from Fr. Rousell ‘red-haired’); It. Macchiavelli ‘son of the one
with dirty hair’; Lat. Cato ‘the sharp/clever one’, Caesar ‘the one with the mane’,
Cicero ‘the one who resembles a chickpea’, Naso (the family name of the poet
Ovid, literally ‘big-nose’).
While much more could be said about the origins of names, and examples
lined up from languages all over the world, this survey gives a fair sampling of the
range of sources for what some linguists – and many more non-linguists – have
felt is the most basic function of language, that of giving a name to an object or
individual.

4 Coinage in argots, jargons, and slang

As noted in Chapter 7, § 5.2, taboo may lead to a considerable turnover in the
lexicon, requiring the coining of many new words. Extensive and constant vocab-
ulary renewal is perhaps even more common in certain special forms of language
use which, like taboo, are socially motivated, namely in argots, jargons, and
 Coinage in argots, jargons, and slang 279

slang. Let us conclude the chapter by taking a closer look at coinage in these
forms of speech.
Although difficult to differentiate with absolute precision, these forms of lan-
guage use can be distinguished roughly as follows. Argots are secret languages,
intended for in-group communication that is to remain unintelligible to outsid-
ers. Argots commonly are employed by criminals; but they may also be used by
other groups, especially the suppressed or disadvantaged. The major purpose
of jargon is to serve in-group communication and social cohesion. Much of its
special vocabulary consists of technical terms, but there are also expressions,
often humorous, that serve as markers of solidarity. Slang, finally, is to ordinary
language what up-to-date, youthful, and somewhat outrageous fashion is to ordi-
nary dress wear. Because of their nature, argots and slang are in need of constant
lexical renewal. In the case of argots, the purpose is to maintain secrecy. If out-
siders hear argot words often enough, they can catch on to their meanings, and
the words are in danger of losing their secret nature. As for slang, the motivation
for constant lexical renewal is similar to the motivation for the constant change
in dress fashion. There is nothing more stale than outdated slang – or yesterday’s
fashion. Since the need for lexical renewal is strongest in slang and argots, most
of the examples given below come from these two forms of speech.
Note that the interrelation of slang, jargons, and argots with each other, as
well as with ordinary language, is complex. In many cases, the precise source for
a given word or the mechanism by which it acquired its meaning is shrouded in
mystery. Moreover, words often are borrowed from one sphere of language use to
the other. This is especially true for slang. Time and again we find that in order to
maintain its novelty, slang adopts words from argots and jargon. Finally, although
speakers tend to resist the intrusion of slang into ordinary language use, they are
far from successful in doing so; and slang (or jargon) words frequently become
part of ordinary vocabulary.
Consider for instance the case of fake, which entered English through argot
in the meaning of any illegal or criminal action, but especially that of stealing
or robbing. Lat. facere ‘do, make’ and Germ. fegen ‘wipe, swipe’ have been men-
tioned as possible sources for the word, in which case the semantic development
would be comparable to that in ‘make off with something’ or ‘swipe’ = ‘steal some-
thing’. But as in many other cases of argot and slang, the exact origin remains a
mystery. Further developments led to meanings like ‘deceive’. In these meanings
the word began to enter slang, as well as jazz musicians’ jargon, where it could be
used for improvising without prior preparation, as in If you don’t know it, just fake
it. From these contexts, the word has come to be increasingly accepted in ordinary
English, so that dictionaries like Webster’s New World Dictionary no longer con-
sider it necessary to label the word as slang or jargon.
280 Lexical change and etymology. The study of words

Other words of similar ancestry, but now in fairly common use, are kid (orig-
inally ‘young goat’), keister (originally from German or Yiddish kiste ‘box’?), ogle
(from Dutch oogelijn ‘little eye’), and pal (from Romani ph(r)al ‘brother’). Even
phrases can be liberated into common use, as in the widespread use of bottom
line to mean ‘the essential point’, originally a technical term in business jargon
referring to the final line in a financial statement.
Just as examples like daisy or, to a lesser degree, clear (in expressions like this
is not clear to me) have been referred to as faded metaphors, words like fake, then,
can be referred to as faded slang (or faded argot/jargon).

4.1. Coinage through semantic change: Semantic change is one of the major
vehicles for creating the vocabulary that distinguishes argots and slang from ordi-
nary language use or for maintaining the distance between these forms of speech
and ordinary language.
Consider recent argot words for ‘police’ in English, such as the heat, the fuzz,
and smokies. The expression the heat no doubt reflects the fact that the police
put the heat on criminals. And the expressions the fuzz (as in the expression Like
I was rappin’ to the fuzz at the beginning of Chapter 1) and smokies seem to derive
from the similarity between U.S. state troopers’ hats and the hat of Smokey the
Bear (who, of course, is fuzzy). Similar developments have given rise to a veritable
plethora of other argot words for ‘police’: bull, danger, signal, terror, elephant ears,
flatfoot, etc.
The act of informing the police about criminal activities has similarly been
expressed by many different words: bark, belch, bleat, chirp, rat (out), sing, squawk,
squeak, squeal, etc.
In the soldiers’ slang or jargon of late Roman antiquity and the early Middle
Ages, ‘battle’ was with a soupçon of gallows humor referred to at the ‘smashing
of pots (= heads) into shards or smithereens’. Hence heads could be referred to as
‘pots’ or even as ‘shards’. As these words began to penetrate ordinary language,
they became the ordinary words for ‘head’ in a number of European languages;
e. g. Germ. Kopf (related to Engl. cup), and Ital. testa, Fr. tête (from Lat. testa ‘pot,
potsherd’). Gallows humor can also be seen at work in argot words for the electric
chair such as hot seat, cinder seat, or barbecue stool.
As we saw earlier, jazz musician’s jargon contributed to the semantic devel-
opment of fake. In the U.S., this jargon has been a continuing source for slang
expressions. Two widely known products of jazz jargon are cool, originally refer-
ring to a mode of jazz performance that differed from earlier hot jazz by being
more smooth and intellectual, and groovy, first attested in the 1930s and com-
monly used in 1940s to 1960s jargon and slang, derived from earlier in the groove
= ‘going smoothly in the groove of a record’. The word groovy now is clearly dated,
 Coinage in argots, jargons, and slang 281

though occasionally enjoying a bit of a retro-revival; cool, on the other hand, has
shown amazing staying power, reappearing in ever-new slang forms and in ever-
new combinations. An oldish, somewhat dated, slang use of cool is found in the
combination cool dude, while a fairly recent slang use is That’s cool, meaning
‘that’s OK; nothing wrong about it’. But given the volatile nature of slang, this use
may already be beyond its prime. Note also the relatively recent pronunciation
of cool with a protracted vowel and a rising-falling intonation, as popularized by
television’s Bart Simpson, now possibly itself dated and lacking in cool.

4.2. Coinage through borrowing: In addition to semantic change, argots,

jargons, and slang commonly draw on borrowing as a major source for vocabulary
renewal. The donors for borrowing commonly are languages of groups that are
marginalized in society and thus often forced into illegal or criminal activities.
In English, these are especially Yiddish and Romani (traditionally referred to as
Gypsy; see § 3 above), as well as Shelta, a jargon spoken by tinkers and “vagrants”
in England and Ireland and deriving much of its vocabulary from Irish and Scots
Gaelic. But many other languages have contributed, including the speech of Dutch
and German criminals who played a significant role in the criminal underworld
of sixteenth-century England. The multicultural, multiethnic, and multilingual
nature of the sources of English argot can be gauged from the selected examples
in (7).

(7) cuffer ‘man’ Yid. kaffer ‘stupid fellow’ (?)

keister ‘behind; rear’ Yid. or Germ. kiste ‘box’
pal ‘companion’ Romani ph(r)al ‘brother’
stir ‘prison’ Romani staripan ‘prison’, stardo
‘imprisoned’ (?)
cove ‘man, rogue’ Romani kovo ‘that man’
tober ‘road’ Shelta tober (from Ir. bothar)
ogle ‘look at covetously’ Dutch oogelijn ‘eye (diminutive)’
mongee ‘food’ Fr. manger ‘eat; food’
feele ‘young woman’ Fr. fille or Ital. figlia ‘daughter’
kinchin ‘young’ Germ. kindchen ‘child (diminutive)’
ken ‘house’ Romani or Hindi khān(ā) ‘house’ or
Arab. xan ‘inn’
hoosegow ‘prison’ Span. juzgado ‘court of justice’
vamoose ‘run away, get lost’ Span. vamos ‘let’s go’
282 Lexical change and etymology. The study of words

African American Vernacular English has provided a similar source for much of
American English slang. Compare such expressions as rap ‘talk, chat, converse’
(in the introduction to Chapter 1) and bad or sick in the sense of ‘good’. See also
§ 4.4 below.

4.3. Other devices for coinage: Beside metaphorical extensions and borrow-
ings, argots and slang resort to a large variety of other means to coin new terms.
These include abbreviatory developments, as in (8), processes similar to taboo-in-
duced distortion (9), as well as a number of language games, such as “Pig Latin”
(with transposition of initial consonants to final position and addition of [ey],
(10), “rhyming slang” (11), and “back slang” (i. e. reverse pronunciation, based
on spelling) (12). While rhyming slang has at least some kind of counterpart in
ordinary linguistic change, in terms of the phenomenon of rhyming formation
(see Chapter 5, § 3.1), the other two phenomena normally are limited to argots and,
to a lesser degree, slang. (But note the physicists’ term mho which refers to the
basic unit of conductance and is derived by back slang from its reciprocal ohm,
the unit of resistance.) The fact that such processes are employed in these forms
of speech must be attributed to the unusually great need for vocabulary renewal
that is characteristic of these modes of communication.

(8) a. Slang:
def definitive (= ‘excellent’)
rad radical (= ‘excellent’)
triff terrific
abfab absolutely fabulous
b. Argot:
cutor prosecutor
davy affidavit
dan dynamite

(9) a. Slang:
grody grotty (?) (an older slang term)
b. Argot:
grift graft
glee see

(10) Slang and argot:

af-g-a(y) f-ag
am-scr-a(y) scr-am
ix-n-a(y) n-ix
 Coinage in argots, jargons, and slang 283

(11) Slang and especially argot:

April fools (burglary) tools
trouble and strife wife
bat and wicket ticket

(12) Argot:
enob bone
efink knife

4.4. Concluding notes: While argots usually are secret languages of the under-
world, they can arise under any other circumstances that call for secret commu-
nication, such as prisoner-of-war camps or slavery. Thus, it has been claimed that
the African American Vernacular English words on the left side of (13) below are
relics of an argot of early slavery, used to keep “the man” from understanding
important communications between the slaves. It has further been suggested that
these words can be traced to West African Wolof sources; see the forms on the
right side of (13). For the semantic development of honkey, compare the fact that
redneck is a common derogatory term for lower-class whites in the South of the
United States, presumably because they turn red in the sun.

(13) dig ‘understand’ dega ‘understand’

(hep/hip) cat ‘(smart) man’ hep kat ‘man in the know’
honkey ‘white (derogatory)’ hong/honk ‘red’

How well argots are capable of serving as secret languages can be gauged from
exchanges like the following which comes from a sixteenth-century book entitled
A caveat or warning, for commen cursetors. Could you figure them out without the
translation in (14’)?

(14) Argot:
a. Question: Why where is the kene that hath the bene bouse?
b. Answer: A bene mort here by at the signe of the prauncer

(14’) Translation:
a. ‘Now, where is the house that has the good drink?’
b. ‘A good wife close by at the sign of the horse.’
Chapter 10: Language, dialect, and standard
I speak a language, you speak a dialect, (s)he speaks like a barbarian
(Anonymous)

And the Gileadites took the passage of Jordan before the Ephraimites: and it was so, that
when those Ephraimites which were escaped said, Let me over; that the men of Gilead said
unto him, Art thou an Ephraimite? If he said, Nay;
Then they said unto him, Say now Shibboleth: and he said Sibboleth, for he could not frame
to pronounce it right. Then they took him, and slew him at the passages of Jordan; and there
fell at that time of the Ephraimites forty and two thousand.
(Judges 12: 5–6; translation from the Authorized Version)

1 Introduction
The preceding chapter has shown the effect that argots, jargon, and slang can
have on lexical change. And as illustrated by examples like pal, a Romani word
(ph(r)al) which entered English via a criminal argot, some of the effects may go
beyond these special varieties of speech and swim into the mainstream of the
ordinary language. In this chapter we take a closer and more general look at the
questions raised by such terms as varieties of speech and ordinary language,
including their relation to each other.
Relationships of this type are often said to involve a difference between
dialect and language or substandard and standard. (Linguists commonly employ
the term nonstandard or vernacular, instead of substandard, due to the judg-
mental connotations that “substandard” has.) And, of course, the terms dialect
and nonstandard cover much wider territory than argot, jargon, and slang. What
further complicates matters is that linguists would like to use the terms language
and dialect in a technical sense very different from their use by non-linguists. But,
as in many other situations, the maxim holds true, “Most speakers are not lin-
guists.” Because the opinions of non-linguists often do play a considerable role in
linguistic development, linguists cannot ignore them, no matter how much they
may disagree. As we will see, this is especially true for the relationship between
language and dialect.

https://doi.org/10.1515/9783110613285-010
 Language and dialect 285

2 Language and dialect

In ordinary, non-technical usage, the terms language and dialect tend to be asso-
ciated with strong value judgments. “Language” is prestigious, correct, stand-
ardized, and follows the rules of grammar. “Dialects” lack prestige, are incorrect,
substandard, and fail to obey the rules of grammar. They are, some might say, a
“depravation of what a language ought to be.” Contrast the English expressions
in (1)–(3), where those on the left are said to follow the rules of language, while
the ones on the right are considered “dialect”.

(1) ask aks

(2) I’ve taken I’ve took

(3) He doesn’t say anything to anybody at all He don’t say nothin’ to nobody
nohow

It is generally believed that the rules of grammar are sacrosanct, having been
established for all eternity. Dialectal deviations, under this view, are the result of
corruption brought about by carelessness and slovenly speech habits.
Linguists have difficulties with these views. They are keenly aware that even
standard languages undergo changes, and that yesterday’s “slovenly speech
habit” may become today’s standard – and conversely, the standard of yester-
day may be nonstandard today. For instance, in the Old English of Beowulf, the
Anglo-Saxon Chronicle, and many other fine texts, we find counterparts to both
of the forms in (1), not only with the sequence sk (spelled sc), now accepted as
correct, but also with ks (spelled cs), now considered “dialect”; see (1’). Construc-
tions with the participial form took, considered ungrammatical or dialectal today,
were perfectly normal in the language of Shakespeare, by all counts one of the
best English writers ever; see (2’). And structures with multiple negation compa-
rable to the right-side construction in (3) were perfectly acceptable in Chaucer;
see (3’a). Traces persist in the language of Shakespeare (3’b) and even in the eight-
eenth century (3’c), though under more restricted conditions. (See also Chapter 6,
§ 4.) It is only in fairly recent standard English that structures of this type have
become unacceptable.

(1’) ascian : acsian ‘to ask’

(2’) Betwixt mine eye and heart a league is took (Sonnets 47)
286 Language, dialect, and standard

(3’) a. Ther nas no man nowher so vertuous (Canterbury Tales A.251)

‘There was no man anywhere so virtuous.’
Lit. ‘There wasn’t no man nowhere so virtuous.’
b. … and live no more to shame nor me nor you (Sonnets 72)
c. No skill could obviate, nor no remedy dispel the terrible infection
(Oliver Goldsmith, History of Greece 1: 224)

Even today we find that different forms of Standard English disagree with each
other, not only in terms of such fairly well-known vocabulary differences as the
ones in (4) and the equally well-known pronunciation differences in (5), but also
on other points of pronunciation (6), and even of grammar (7). The difference
in (4c) is said to have caused genuine misunderstanding and, as a result, even
anger in the Allied High Command during the Second World War. Speakers of
(standard) British English tend to consider their r-less pronunciation in (5) to be
superior, in spite of the fact that the r-ful speech of (most) Americans preserves an
older stage of the language and thus is historically more correct. As for the word
herb, both British and American English speakers can find cause for finding fault
with one another; British speakers because of “that awful American r”, Ameri-
cans because of the Britishers’ incorrect sounding out of the initial h (“next thing
they’ll pronounce hour as [hawǝ] and hono(u)r as [hɔnǝ]!”). The situation is even
more complex in (7): Americans might take pride in being consistent in saying
both gotten and forgotten, while the British might feel superior for having the
same form, got, both in (7a) and in (7b). Historically, both got and gotten have
long-standing antecedents.

(4) British English American English

a. flat ‘apartment’ flat (tire)
b. knock up ‘call on’ knock up (slang) ‘make pregnant’
c. table ‘ put on the table for immediate table ‘put on the table =
consideration’ the “back burner”’

(5) car [kā] [kār]

bird [bǝ̄ d] [bǝ̄ rd]

(6) herb [hǝ̄ b] [ǝ̄ rb]

(7) a. I haven’t got your letter yetI haven’t gotten your letter yet
= ‘I haven’t received your letter yet.’
b. I haven’t got enough money I haven’t got enough money
= ‘I don’t have enough money.’
 Language and dialect 287

In the technical usage of linguistics, the terms language and dialect are used in a
very different manner. All the citations in (1)–(3) and (4)–(7) above are forms and
structures of the English language (as distinct from, say, the French language);
but they belong to different dialects. The left-hand citations in (1)–(3) belong to
the standard dialect, while the ones on the right belong to nonstandard or ver-
nacular dialects. And the differences in (4)–(7) characterize different standard
dialects of the English language.
This is not to say that the non-technical understanding of the terms language
and dialect is irrelevant. As we will see soon, the ideas associated with this dis-
tinction can play a considerable role in language change. But the technical usage
of the terms covers important aspects of linguistic reality as well.
The technical differentiation between language and dialect can be charac-
terized as follows. Varieties of speech that are relatively similar to each other,
whose divergences are relatively minor, are called different dialects of the same
language. A language, then, is the collection of such dialects – whether they are
standard or vernacular, urban or rural, regional or supraregional. Varieties which
differ from each other more noticeably, whose divergences are major, are called
different languages.
Ideally, the distinction between language and dialect is based on the notion
of mutual intelligibility. Dialects of the same language should be mutually
intelligible, while different languages should not be. This mutual intelligibility,
in turn, would be a reflection of the linguistic similarities between the different
varieties of speech.
Unfortunately, the mutual-intelligibility test does not always lead to clear
results. A person from rural Maine, speaking only the local dialect, will find it
difficult, if not impossible, to understand someone from rural Louisiana who
likewise can only speak the local dialect. Still, a person going on a slow, lei-
surely trip from Maine to Louisiana would find no language boundary along
the way, comparable to, say, the one between German and French. Rather, any
two neighboring local dialects along the way would be perfectly intelligible to
each other. The dialects of Maine and Louisiana, thus, are part of a dialect
continuum, linked to each other through a chain of mutual intelligibility, and
it is for this reason that they can be considered dialects of the same, English,
language – in perfect agreement with ordinary, non-technical perception and
usage.
The situation is more complex in cases like Scots English and American
English. For instance, confronted with a Scots English expression of the type (8a),
ordinary speakers of American English would have a hard time realizing that it
is English, and even greater difficulties in understanding the passage as being
simply a different pronunciation for the expression in (8b); in fact, Americans
288 Language, dialect, and standard

might well believe that (8a) is not English, but Scots Gaelic. How, then, can we
justify considering Scots English a form of English?

(8) a. [its a brādlext mōnlext next tonext]

b. It’s a broad-light moon-light night tonight

To justify the usual belief that Scots English is in fact a variety of English, just as
much as American English, we might point to the existence of a British dialect
chain that links Scots English with Standard British English, and to the general
mutual intelligibility of standard British and American English; American and
Scots English thus are ultimately linked with each other. But knowing that there
is such an indirect link does not really make it any easier for the Scots and Amer-
ican speakers to successfully communicate with each other in their own native
varieties of English.
An alternative justification for considering Scots and American English to be
different varieties of the same language might be that, given enough time (and
good will), speakers of the two varieties of English can achieve mutual intelligi-
bility. But this argument doesn’t get us very far, for with an even greater amount
of time (and of good will), a greater effort, and the right choice of words, French
and English might likewise become mutually intelligible.
In fact, there are numerous difficulties with the concept of mutual intelli-
gibility. For instance, Norwegian and Swedish are mutually quite intelligible,
and yet, most people – including linguists – would consider them to be different
languages. The reason for classifying them as different languages is that Norwe-
gian and Swedish have different standard dialects and literary traditions and,
even more important, are considered different from each other by their respec-
tive speakers. Here, then, cultural, social, and political considerations overrule
the mutual-intelligibility test. A language in this sense is, as one linguist put
it jokingly, “a dialect with an army and a navy”. One might add, “… and with
schools.”
But having different armies, navies, and schools does not necessarily mean
a difference in “language”. Consider the case of British vs. American English.
Britain and the United States certainly have different armies and navies; and
as we just saw, they have different standard dialects (although their differences
are smaller than those between Norwegian and Swedish). At least in the modern
period, they also have distinct literary traditions. But the overwhelming majority
of their speakers, on both sides of the Atlantic, continue to look upon their lan-
guage as one. Here as elsewhere, the attitudes of speakers matter greatly.
Consider further the case of Norwegian and Swedish vs. Danish. Norwegian
and Swedish are said to be quite readily intelligible to speakers of Danish. But the
 Language and dialect 289

intelligibility is not mutual, since speakers of Swedish and Norwegian find Danish
quite unintelligible.
A possible linguistic reason is the following. Danish had very extensive weak-
ening (see Chapter 4, § 5.1.2) of intervocalic consonants, while Norwegian and
Swedish preserved these consonants more faithfully. Compare Dan. flue : Swed.
fluga ‘fly’ or Dan. fjeder/fjer (both pronounced [fyeǝ]) : Swed. fjäder ‘feather’.
Danes would have learned to get by without the consonants; their presence in Nor-
wegian and Swedish would, for them, be redundant. But Norwegian and Swedish
speakers, used to the presence of intervocalic consonants in their own speech,
would depend on their presence for proper understanding and, not hearing them
in Danish, would fail to understand the Danes. So it seems, at least.
Linguistic factors of this type may in fact play a role in the lack of mutual
intelligibility among the Scandinavian languages. However, some – more honest –
speakers of Norwegian and Swedish admit that the lack of intelligibility is a con-
sequence more of attitude than of linguistic differences. As one Swede put it, “We
can understand the Danes; we just don’t want to.” Many Norwegian and Swedish
speakers consider the sound of Danish horrible – very guttural or throaty, in
part because of the use of a uvular [r] for the postdental [r] of Norwegian and
Swedish. Moreover, Danish makes extensive use of glottal stops, where Norwe-
gian and Swedish do not. Interestingly, this evaluation of Danish as guttural and
unpleasant is shared by many Danes. A great Danish linguist, for instance, is said
to have stated, “Danish is not a language, it is a throat disease.” Here, then, even
the notion of mutual intelligibility depends on speakers’ attitudes, not on purely
linguistic facts.
Finally consider cases like German and Dutch. A person traveling from
the southernmost areas of German speech (as far south as Northern Italy), via
Austria and Germany, into the Netherlands would find no boundary of mutual
non-intelligibility separating one local dialect from the other. The whole terri-
tory is a single dialect continuum, although dialects that are fairly removed from
each other geographically may be mutually quite unintelligible. Yet the standard
Dutch and German languages clearly lack mutual intelligibility and thus qualify
for being called different languages. Moreover, they meet the sociolinguistic cri-
teria for being considered distinct languages, since like Norwegian and Swedish,
they have different linguistic standards and literary traditions and are considered
different from each other by their respective speakers.
The situation is similar in the vast territory of the non-Balkan Romance lan-
guages, except that we have at least five distinct standard languages: Italian,
Romantsch (one of the officially recognized languages of Switzerland), French,
Spanish, and Portuguese. To these must now be added Catalan and Galician
which have recently been accorded official status (see also § 5 below).
290 Language, dialect, and standard

What these varying results and failures of the mutual-intelligibility test show
is that there is no clear line of demarcation between “different dialect” and “differ-
ent language”. Linguistic similarity or difference is a matter not of yes or no, but
of more or less. Even mutual intelligibility depends not only on linguistic factors,
but also on social ones.
Still, the terms dialect and language are useful because they define the extreme
points of a continuum. If in subsequent chapters certain developments are por-
trayed as characteristic of dialect contact on one hand, or language contact on the
other, we have to bear in mind that there may be relationships which show char-
acteristics intermediate between these two more extreme types of relationship.

3 Social dialects
In traditional historical linguistics, the notion dialect is almost exclusively
reserved for geographically defined local and regional speech varieties; this is
why Chapter 11, on dialectology, is devoted mainly to regional dialects. However,
dialect differences may also be correlated with social differences. Thus, the
Chicago Chain Shift, discussed in Chapter 4, § 5.4, was originally limited to certain
white working-class male – and macho – groups. Other speakers, within Chicago,
did not participate in the change. In fact, where the male-macho speakers pro-
nounce the name of their city as [šikægo], those who don’t want to be caught dead
being identified with these speakers affect a polarized pronunciation [šikɔgo].
The pronunciation [šikago] is relatively rare in Chicago. (See also Chapter 11, § 1.)
Social dialect differentiation of this type is very common and is by no means
confined to large urban areas like Chicago. We encountered the same phenome-
non in the case of fairly rural Martha’s Vineyard, with its socially polarized dia-
lects of [a]-centralization vs. non-centralization. (See Chapter 4, § 6.3.)
Similarly, in Central Illinois, a chain shift polarizes two social groups in two
small rural communities (Farmer City and Mansfield). The shift, involving the
fronting of all back vowels, is characteristic of one group, the so-called burnouts.
At the other extreme are the so-called rednecks, who refuse to participate in the
change, clinging instead to the standard language. The terms burnouts and red-
necks are used by local high school students as part of their school slang or argot,
and their meanings are rather different from those in ordinary American English.
Burnout no doubt is an extension of the more common term burn out, and probably
is influenced by “drug-scene” talk. The burnouts are students who use drugs or
alcohol and are not interested in going on to college. The use of redneck involves
somewhat more complicated developments. In origin, the word is a derogatory
 Discontinuous dialects – Professional jargons and related forms of speech 291

term for Southern U.S. rural whites, comparable to the Wolof-derived term honkey
(see Chapter 9, § 4). Because in the post-Civil Rights era, Southern poor whites
tended to make common cause with conservative politicians, the word redneck
appears to have locally acquired the extended meaning ‘conservative’. And by an
ironic twist, in this meaning the term could be used for upwardly mobile students
who were not at all poor, but conservative, in the sense that they did not “do drugs”
but instead went in for sports, as well as for a college education after high school.

4 D
 iscontinuous dialects – Professional jargons
and related forms of speech
In addition to continuous – social or regional – dialects, many languages have
at least some discontinuous, supraregional dialects, defined only in terms of
social factors, which extend across the boundaries of continuous dialects. Some
social dialects are by their very nature discontinuous and supraregional, because
they are defined in terms of social groups living in many, geographically discon-
tinuous locations.
Consider the English dialect of lawyers, with its heavy use of borrowings from
French and Latin, as in (9), as well as with certain peculiarities in grammar, such
as the past tense ple(a)d [pled] of plead [plīd], and especially the extremely long
and complex syntactic strings conjoined by multiple occurrences of whereas, fol-
lowed by a lonely therefore.

(9) feme couverte, feme sole

grand jury, petit jury
oyez, oyez, oyez (at the beginning of a court session)
habeas corpus
nolo contendere, nolle prosequi
venire (facias)

Social dialects of this type are commonly referred to as jargons or registers.

Of these two terms, “register” can be used in a more comprehensive manner,
as in “academic register” or “professional register”, while “jargon” tends to be
employed more specifically, as in “linguistic jargon” or “lawyers’ jargon”. More-
over, the two terms differ in connotations. “Register” focuses more on questions
of rhetoric and style, while “jargon” emphasizes the exclusiveness of dialects of
this type, which often makes it difficult for outsiders to understand them. In this
regard, jargons share features with argots and slang. (See Chapter 9, § 4.)
292 Language, dialect, and standard

While professional jargons and registers may share with argots and slangs the
fact that they are often difficult to understand for outsiders, their basic motivation
is not secrecy (or novelty, for that matter). Specialized fields of inquiry by their
very nature require specialized terminology, so as to express what is intended as
unambiguously and succinctly as possible. For instance, in historical linguistics
the term “sound change” is used, not to refer to just any change in sound struc-
ture, but to sound changes not conditioned by non-phonetic linguistic factors. If
each time we wanted to talk about sound change we had to use the lengthy cir-
cumlocution “sound changes not conditioned by non-phonetic linguistic factors”
we would never be able to get to the point. Similarly, the use of certain grammat-
ical constructions, such as the passive, is highly appropriate in many areas of the
sciences. The passive makes it possible to delete agents (as in The experiment was
conducted under the following conditions), and this in turn makes it possible to
state generally valid facts or claims, or observations that are believed to hold true
no matter who observes them.
Professional jargons or registers are not necessarily limited to lawyers, scien-
tists, and other professionals. They are also found in the trades and crafts.
One of the most widespread professional jargons of this type is that of sailors
and seafarers. But in some ways, this jargon is unusual. At least in the North Atlan-
tic, the vocabulary of the jargon is highly international, with liberal borrowings back
and forth among the languages of all the seafaring nations. And there is reason to
believe that medieval Mediterranean nautical jargons had similar characteristics.
Moreover, the vocabulary frequently comes not from the standard languages but
from coastal dialects. This is especially true for the German element, which comes
from the Low German (LG) dialects that differ considerably from southern-based
Standard German and are much closer to Dutch and Frisian. A consequence of
these special circumstances is that words may be borrowed back and forth several
times over, making it impossible in many cases to determine with certainty which
language was the ultimate source for a given term. Compare the English, German,
and French examples and their putative sources in (10). (Some of the terms in (10)
may be also used in non-nautical meanings, such as Engl. caboose, freight.)

(10) English German French Putative source

brig Brigg brig/brick Engl., from brigantine (⇐?)
boat Boot bateau (ME from) ON
buoy Boje bouée OFr., via Dutch (?)
cabin Kabine cabine E cabin ⇐ OFr. cabane
caboose Kombüse cambuse Dutch (⇐?)
fleet Flotte flotte Germ. ⇐ Fr. flotte ⇐ early
Gmc.
 Standard languages 293

flotilla Flotille flotille Fr. ⇐ Span. ⇐ Fr. etc.

freight Fracht fret Frisian, via Du./LG
gaff Gaffel OFr. ⇐ Span. or Port.
(larboard) Backbord bâbord LG/Du. /OE?
(First) Mate Matrose ‘sailor’ matelot Du/LG ⇐ Fr. ⇐ Du./LG or
Scand.
lee Lee Engl.(?), or ON (?)
luff Luv lof ON (?)
packboat Packboot paquebot Engl.
pilot Pilot pilote It. ⇐ Gk.
road Reede rade ON
schooner Schoner schooner Scots Engl. (?)
starboard Steuerbord tribord Fr. ⇐ early Gmc. (via?)

5 Standard languages
The most important supraregional dialects are standard dialects, more com-
monly referred to as standard languages. Historically, these can be of very differ-
ent origins. They can be regional or local dialects which for some reason acquired
sufficient prestige to be accepted as standard on a supraregional basis; compare
Standard French and English which developed out of the educated speech of Paris
and London, respectively. They may develop out of “koinés” (a type of contact lan-
guage to be discussed in Chapter 12 § 5), such as the Greek Koiné of Alexandrian
times or the Swahili of present-day Tanzania. They may result from deliberate
language planning or language engineering, as in the case of Nynorsk, one of the
standard languages of Norway.
To illustrate the role and effects of language planning in Norway, let us take
a closer look at the recent history of Norwegian. Centuries of Danish domination
had made Danish the language of the educated Norwegian elite, who were con-
centrated mainly in the large urban areas of the south. Over time, developments
similar to the ones that lead to regional accents in monolingual societies (see
further below) brought about a strongly Norwegianized form of Danish. The dif-
ference between this form of language and the Danish of Denmark was further
increased by a variety of sound changes, especially pervasive weakening, that
differentiated Danish from the rest of the Scandinavian languages. As a conse-
quence, the language of Norway came to differ so much from Danish that Danes
and Norwegians would consider it a different language altogether, namely Norwe-
gian. (Compare Norw. kjøbe [čöpǝ] or [čöbǝ] ‘buy’ vs. Dan. købe [köβǝ].)
294 Language, dialect, and standard

In the wake of the nationalism that swept nineteenth-century Europe, the

Danish ancestry of Norwegian became suspect in the eyes of nationalistically
inclined Norwegians. One of these was the school teacher Ivar Aasen. Inspired
by the puristic tendencies of Iceland (see Chapter 8, § 5.2), he set out to construct
a “truly Norwegian” language, drawing on the most conservative linguistic ele-
ments of conservative rural dialects. The grammar for this Landsmaal or Nynorsk
(i. e. ‘Country language’ or ‘New Norwegian’) was completed in 1864, and a com-
panion dictionary nine years later, in 1873.
Aasen’s heroic attempt at single-handedly creating a new, truly Norwegian
language was greeted with enthusiasm by many nationalists. But an even larger
number of Norwegians were not impressed. They did not feel any need to give up
their variety of Norwegian, which was widely spoken by the educated population
of southern Norway and even more widely understood, in favor of what to them
appeared to be a country-bumpkin language which could not claim any native
speakers or regional basis. (Recall that Nynorsk was a composite language, con-
sisting of elements from a variety of rural dialects.)
The result has been a long-standing, often uneasy, coexistence of two Norwe-
gian standard languages, the earlier Riksmaal or Bokmaal (i. e. ‘State language’ or
‘Book language’) and Aasen’s Nynorsk or Landsmaal. Both are recognized as offi-
cial languages, and both are used as media of instruction in the schools, with indi-
vidual localities opting for one or the other. Bokmaal still has the largest number
of speakers (about 80 %); but Nynorsk has acquired its own, loyal followers. In
recent years, the coexistence – and competition – of the two varieties has led to
the development of an intermediate variety which combines features from both
Nynorsk and Bokmaal. Perhaps this variety will some day resolve the competition
by replacing both of the currently recognized official languages.
Standard languages may also result from a combination of any or all of these
different sources. Thus, although Standard French and Standard English origi-
nated in the educated speech of the respective capital cities, Paris and London,
they did experience a certain amount of language planning, too. As noted in
Chapter 2, the Renaissance brought with it an increasing cultivation of the ver-
nacular, regional languages (at the expense of Latin). That cultivation consisted
not only of an expansion of vocabulary, commonly by borrowings from Latin, but
also in attempts to regulate grammar, usually on the model of Latin. In France this
was largely accomplished by the Académie Française. In England, which lacked
such an academy, it was the work of individual scholars and writers who often
disagreed with each other, especially in the area of vocabulary development. (See
Chapter 8, § 5.2; for syntax, see also Chapter 6, § 4.)
For an extended example of the complex ways in which standard languages
may originate, consider the case of Modern Standard German.
 Standard languages 295

The ultimate source was a regional written koiné of the chanceries (or admin-
istrative headquarters) of various East Central German principalities.
The reason that this regional variety acquired wider currency must be sought
in Lutheranism. Believing in the universal priesthood of all Christians, which
includes the ability to read the Bible for oneself, Luther felt it necessary to trans-
late the Bible into a language that could be understood by all Germans. And
because he was familiar with the East Central German chancery koiné, Luther
naturally used this as a starting point. But just as naturally, he had to expand
the vocabulary and diction of that bureaucratic language to make it suitable for
translating the Bible. To that end, he claimed, he “watched the people’s mouths”
to find words in common use that were widely understood, even by those who did
not use the words themselves.
The new Bible translation and its language, together with a catechism and a
large stock of church hymns, quickly spread to Luther’s Protestant followers who
were mainly located in the north of Germany. And both in church services and in
the schools that prepared pupils for reading the Bible, the new variety of German
increasingly came to be used as a spoken language. The religious divisions of
Germany, however, which led to a prolonged series of wars, made not only Luther-
anism, but also its language, suspect in the mainly southern and western parts of
Germany, which remained with Roman Catholicism.
As time progressed, the language came to be used for fine literature. At the
same time, it underwent a certain degree of deliberate archaizing at the hands of
people who were involved with attempts to purge the German language of foreign
influence and who wanted to maintain a connection with earlier, Middle High
German literature. In this latter respect, recall that the attempts to purify Icelandic
were in part inspired by a similar desire to maintain a literary connection with Old
Icelandic. However, in contrast to Iceland, the German attempts at purification
were only partly successful. The result is the very mixed effect of linguistic nation-
alism in Modern German that we noted in Chapter 8, § 5.2.
What turned this language from a regionally based and sectarian form of
German into a truly national medium was its use by the Romantic and Classical
literary writers, especially Goethe and Schiller. The sheer quality of their work was
such that even those who had so far resisted the German of Luther’s Bible trans-
lation could no longer hold back. Moreover, the topics of writers like Goethe and
Schiller were no longer tied to any particular form of Christianity. And perhaps,
too, the political and philosophical climate had changed enough to make the old
sectarian and regional differences appear less significant.
The written use by these writers and by contemporary scientists and philoso-
phers, then, increasingly became the model for correct usage, just as the written
(and oral) use of educated Paris and London speech had become the model for
296 Language, dialect, and standard

correct French and English. At roughly the same time, the language also became
the vehicle for anti-Napoleonic and anti-French sentiments, serving as the symbol
of an emerging nationalism. And slowly it came to be used as a spoken language,
especially in the northern areas of Germany and in the cities.
With increasing spoken use, it finally could become a native language, for at
least some speakers. Even today, however, it has remained a second language (or
dialect) for many German speakers. This is especially the case in Switzerland, but
even in southern Germany and Austria local and regional dialects are still the first
language for most speakers.
Whatever its origins, a standard language soon becomes an entity in its own
right, with a supraregional sociolinguistic basis. As a consequence, for instance,
Modern Standard English no longer is tied to (educated) London speech, not even
to Standard British usage. Rather, it is the language of educated speakers, no
matter where they may be located. And some of its linguistic innovations, such as
the “haw-haw” variety of the King’s (or Queen’s) English, with its [ew] for the [ow]
of other dialects (as in [ew, ay dewnt bilīv sew] ‘Oh, I don’t believe so’), seem to
have no regional basis but are limited to socially defined sub-varieties of Standard
British English.
As is shown by the case of English, which is used in Great Britain, Ireland, the
U.S., Canada, Australia, New Zealand, and many other countries (see Chapter 2,
§ 3.3), standard languages are not necessarily restricted to a single country. Simi-
larly, German is the standard language of Germany, as well as Austria, and parts
of Switzerland and Belgium; and French is standard not only in France but also
in parts of Belgium and Switzerland, as well as in West African countries like
Senegal and in the Canadian province of Quebec; in fact, French is recognized as
an official language in all of Canada, beside English.
Like Canada, other countries, too, have more than one standard or official lan-
guage: Nynorsk and Riksmaal/Bokmaal in Norway; Flemish, French, and German
in Belgium; French, German, Italian, and Romantsch in Switzerland. Within
France, Provençal is clamoring for recognition as a literary standard language, in
addition to Breton and Basque. And in Spain, at least three languages – Catalan,
Galician, and Basque – coexist with the Castilian Spanish standard language.
Examples like these show that standard language and national or officially
recognized language are not necessarily identical and that their relationship is
open to historical variation. Those who advocate special “language” rather than
vernacular status for, say, Provençal point to the fact that Provençal has a rich
literature of its own and thus is distinct from French. Those opposed to special
recognition argue that the “dialect” is not officially recognized as a national lan-
guage, may not be taught in the schools – because it is not officially recognized –
and so on.
 Standard languages 297

During the reign of the Franco regime in Spain, the latter argument prevailed,
and it is said that Catalunya abounded with signs that read “Don’t bark, speak
Spanish” (meaning “don’t talk Catalan, speak Castilian Spanish”). With the
demise of the regime and the introduction of democratic rule, the situation has
changed drastically, and Catalan, along with Galician and Basque, has been rec-
ognized as a co-official language. All three of these languages now are used freely
in schools, in publications, and in radio and television.
Finally, beside languages like German, French, English, or even Provençal
and Catalan, all of which are current in fairly large areas, there may be regionally
more restricted standard languages. And these may coexist with the supraregional
standard languages, as well as the local dialects, with different roles assigned to
each of these varieties of speech. For instance, in much of German-speaking Swit-
zerland, Standard German is used only for written communication. A regional
standard is said to be used for oral communication between speakers from differ-
ent areas, while elsewhere the local dialect is employed.
Standard languages often are written languages. In fact, in some societies
they exist only in written form (which of course may be read out or recited). This
was the case for early Standard German, and still is true for Standard German in
much of Switzerland. In such cases, standard language and local dialect or ver-
nacular coexist in a situation of “diglossia” (see below). However, in preliterate
societies and occasionally elsewhere, standard languages exist only in spoken
form, as is the case with the regional standard German of Switzerland.
An extreme case is that of Vedic Sanskrit, the language of the oldest sacred
texts of Hinduism. Until recently this form of Sanskrit was not put into written
form but handed down only through oral tradition – in spite of the fact that the
art of writing has been available in India since at least the third century BC. (See
Chapter 2, § 3.10.2, and Chapter 3, § 2.1.)
At the same time, the fact that many standard languages exist first and fore-
most in written form may have important repercussions for linguistic change. One
of the most common effects is that of spelling pronunciation, the replacement
of the historically justified pronunciation of a given word by one which is sug-
gested by the spelling.
It is this phenomenon which accounts for the fact that Engl. often frequently
is pronounced as [ɔftǝn], rather than as inherited [ɔf(ǝ)n]. Similarly, the initial
[h] of Engl. humble results from spelling pronunciation, the older pronunciation
surviving only in rural dialects, such as Southern Am. Engl. Be ’umble to the Lord.
The same explanation holds for the initial [h] of Brit. Engl. [hǝ̄ b] herb vs. Am. Engl.
[ǝrb]. (Once accepted by a speech community, of course, the result of spelling pro-
nunciations may become the norm; and the use of the older pronunciation may be
considered incorrect or rustic, as in the case of ’umble.)
298 Language, dialect, and standard

Standard languages, whether written or not, also can have a retarding effect
on linguistic change. After all, what characterizes standard languages is stand-
ardization. Standardization, then, becomes a measurement of correct speech,
to which people claiming to speak the standard language must adhere. Stand-
ard languages therefore tend to become fettered languages, which tend to retain
older patterns more tenaciously than vernaculars, especially when such patterns
become shibboleths. (For the origin of this expression see the second epigraph
of this chapter. The word shibboleth literally means ‘stream, river’ – a clever word
choice for testing the dialectal and ethnic identity of people at a river crossing.)
The preservation of archaic patterns can be observed in Standard German,
which still retains the option of the inflected genitive, as in (11a), or of the dative
singular ending -e [-ǝ], while most of the vernaculars have lost the [-ǝ] of the
dative and have replaced the inflected genitive construction with periphrastic
structures of the type (11b) and (11c). While structures like (11b) have come to be
accepted in the standard language, too, the type (11c) has acquired the status of
a shibboleth. Its use is not acceptable in written Standard German, and even in
the spoken language it is considered either substandard or, at the least, highly
colloquial.

(11) a. Das Haus des Vaters ‘the father’s house’

b. Das Haus vom Vater ‘the house of the father’
c. Dem Vater sein Haus (lit. ‘to the father his house’)

For possible further consequences of the conservative nature of standard lan-

guages, see § 6 below.
In addition to being conservative or “defensive”, standard languages may go
on the offensive, as it were. In many contemporary societies the standard lan-
guages are in the process of threatening the existence of the local, vernacular dia-
lects, or even of replacing them. This is often attributed to the influence of general
education and of the increased effectiveness of the mass media in penetrating all
layers of society. While these may be contributing factors, there are enough exam-
ples of similar developments in earlier times (when mass media played a minor
role, if any) to suggest that we are dealing with a tendency that holds true for all
languages, at all stages of their development.
The most celebrated and well-known instance of such a development, at
a time when there was nothing comparable to the modern mass media, is the
spread of the Greek Koiné during the Hellenistic period (roughly 300 BC to 300
AD), leading to the eventual replacement of virtually all of the older local dia-
lects. (For the nature and origin of the Greek Koiné, see Chapter 12, § 5.) Only one
present-day regional dialect can be clearly traced back to an ancient Greek local
 Standard languages 299

dialect, namely Tsakonian, a modern descendant of ancient Laconian (the dialect

of Sparta). And even Tsakonian has a very heavy admixture of Koiné elements. All
the other Modern Greek regional dialects are descended from the Koiné, although
they may preserve a few traces of older local or regional dialects.
Standard languages, thus, can in effect completely eliminate the results of
centuries of continuous interaction between neighboring dialects.
However, in many cases the local, vernacular dialects do not disappear
without any trace. As regional nonstandard speakers adopt the standard lan-
guage, they commonly retain some of the features of their original dialect and
wind up speaking the standard with a regional accent. For instance, speakers
from the U.S. Southeast adopting northern standard speech (which is generally
considered more prestigious) may quite successfully drop their “drawl” or their
“nasal twang” and, after some initial mistakes, learn where to put in a postvo-
calic [r], and where not. Even so, they may retain certain echoes of their original
speech, especially a pronunciation of words with northern [ay], such as mine and
ice, which to northern speakers sounds more like [æ] and therefore can lead to
misunderstandings, hilarity, or even embarrassment.
The spread of standard languages thus may not necessarily lead to completely
homogeneous and dialect-free results. In fact, the resulting regional accents of the
standard may acquire a life of their own and serve to newly define regional loyal-
ties – in opposition to the threat of homogenization, and to the political and social
powers behind that threat. Thus, in many European countries, local dialects have
been disappearing at a rapid rate. Their place, however, has commonly been
taken by regionalized forms of the standard. And partly because of the increasing
social significance of these varieties, partly also because of a greater tolerance
for diversity, regional accents are increasingly heard, even from newscasters on
national radio. Thus, where in the past, BBC news was invariably delivered in
“RP” (“received (standard) pronunciation” – received from whom, one wonders),
in more recent years it has become easy to identify announcers by their regional
accent as coming from England, Scotland, Wales, or Ireland – even though all
of them use otherwise impeccably standard British English. Similarly, U.S. pres-
idents like Jimmy Carter, who hail from the South, have been speaking Standard
American English, though with a Southern regional accent.
Yet a further – and paradoxical – complication is that the spread of a stand-
ard language by no means leads to everybody speaking the standard variety and
nobody speaking a vernacular any more. This seems to be due to two factors.
First, standard languages commonly have colloquial variants that are used in less
formal circumstances. Although these are variants of the standard, they are not
considered identical to it. It seems that these are the varieties that spread much
more successfully than the formal standard, which tends to remain limited to
300 Language, dialect, and standard

those who need to use very formal language, or believe they ought to. Secondly,
regional dialects, too, come in relatively formal, more colloquial, and downright
vernacular varieties. Among these, the vernacular varieties seem to be most suc-
cessful in preserving some of their linguistic features.
Regional accents thus can display a great degree of diversity. On one side of
the spectrum are the standard varieties, such as the different versions of Standard
British English nowadays heard on BBC. As mentioned, these differ mainly in fea-
tures of pronunciation. On the other extreme we find different vernaculars which,
though far removed from the original regional dialects, are perhaps equally far
removed from the standard. Not only is their pronunciation vastly different, they
also differ in lexicon, syntax, and style.
An interesting side effect of the development of regional accents is that not
all accents are necessarily considered equally acceptable. Frequently, accents
that are most different from the non-regionalized standard are considered more
interesting or pleasing than varieties that are closer to the standard. In part this
may involve the reverse stereotype of the “noble savage”, but in part it may also
result from the feeling that speakers of varieties relatively close to the standard
should “know better” and that their different accent, therefore, is simply due to
indolence or worse.

6 Diglossia
The conservative character of standard languages, if left unchecked, will bring
about an increasing gap between standard and vernacular, such that the standard
language ceases to be intelligible to vernacular speakers without special school-
ing. This is the case in many of the European languages, where the vernaculars
of the local dialects differ profoundly from the standard language; see the Scots
English in example (8). Even in the United States, where the differences between
standard and vernacular are smaller, some speech varieties find themselves in
this situation vis-à-vis the standard language, especially African American Ver-
nacular English and the speech of Appalachian rural whites.
In some societies, this situation has progressed to the point that the standard
is in effect a foreign or second language for all speakers, learned only in school,
but still considered to be the same language as that of the vernacular. Take for
instance the modern Arabic world and modern Greece, where in each case an
ancient, ancestral prestige language (or a derivative form of it) continued to be
in active use among the educated for centuries, side by side with its less pres-
tigious and increasingly different descendants. This special coexistence between
 Diglossia 301

an ancient prestige language and a related modern vernacular is referred to as

diglossia.
In some cases, such as in modern Greece and especially in the Arabic world,
the ancient prestige language has tended to be employed mainly in written form
or highly formal recitation, or for official business of various sorts, while the ver-
nacular is used elsewhere, and not normally in writing. The situation is similar
in the German-language area of modern Switzerland. In fact, it has been claimed
that nowadays some Swiss Germans are so uncomfortable with the spoken use of
Standard German that they prefer using English or French with native or fluent
speakers of Standard German.
In other traditions, the prestige language can – or could – also be employed
in freely spoken form. This was the case with Latin in medieval western Europe,
which was used as the common written and spoken language of the educated.
In India, Sanskrit has similarly continued in spoken form down to the present
time – more than three millennia after it first appeared on the scene. And in
Greece, a variety of Greek known as katharévousa (lit. ‘purifying [language]’)
which was based in its grammatical form on Ancient Greek and thus differed
markedly from the dimotikí (the ‘people’s [language]’), gained favor in the early
nineteenth century as a sign of Greek national identity and of independence from
the Ottoman Turks, who had ruled Greece for several centuries.
From the modern perspective, situations of this type may be hard to believe.
Why would people not use their own native language, instead of Latin, Sanskrit,
or some other non-native language? The situation becomes easier to understand
if we consider that medieval western Europe, just like modern Europe, was home
to a large variety of different languages, many of which clearly were not mutually
intelligible. Latin made it possible for students and scholars to go anywhere in
Europe and be understood by their peers, without the need of learning French
in France, Italian in Italy, German in Germany, English in England, Hungarian
in Hungary, etc. The case is very similar for Sanskrit in India. In both areas, inci-
dentally, the ancient prestige language was used not only to talk about scholarly
issues. It could be used for any other communication, such as asking about each
other’s background, gossiping, or even telling dirty jokes. In fact, English now
serves a very similar function in many parts of the world.
In principle, prestige languages in such diglossic situations are highly con-
servative, resisting the normal linguistic changes which affect the vernacular.
However, if they are freely used in spoken form, they often undergo what may be
called vernacularization, an intrusion of vernacular linguistic features. Thus,
spoken Sanskrit has in many areas of India acquired the vernacular pronuncia-
tion [ǰ] and [b] for initial [y] and [v]. Similarly, the increasing vernacularization of
medieval Latin led to the compilation of dictionaries which prescribed the correct,
302 Language, dialect, and standard

more classical pronunciation and proscribed the incorrect, popular pronuncia-

tion. And while highly conservative in its morphological structure, Katharevousa
Greek is thoroughly modern in its pronunciation.
At the same time, recall that under more ordinary circumstances the interac-
tion between dialect and standard may give rise to regional accents of the stand-
ard language which hold a position somewhere between the pure standard and
the pure dialect. In a similar manner, the interaction of vernaculars and the pres-
tige language in a diglossic relationship may lead to new varieties intermediate
between the pure vernacular and the pure prestige language.
Developments of this sort are found both in Greece and in many parts of the
Arab world (e. g., Egypt). In the Arab world, the prestige of the language of the
Qur’an is so great that intermediate varieties of this type have little chance to get
recognition. In Greece, the status of the prestige language has fluctuated consider-
ably, depending on political developments. For instance, after a period of decline
it was reinstated by the conservative military junta that ruled Greece from 1968 to
1974, but lost its official status with the downfall of the junta and the ascendancy
of a democratic and more progressive government. Subsequent political develop-
ments encouraged the use of an emerging intermediate language, which mixes
Katharevousa with Dimotiki elements, thus not only bridging the gap between the
two forms of language, but also between the generally more conservative support-
ers of Katharevousa and the more liberal proponents of Dimotiki.
In addition to vernacularization, the prestige language may undergo region-
alization, under the influence of the local languages or dialects. For instance, in
medieval times, different forms of Church Latin developed in this manner. Some
of these differed only in minor details, such as the pronunciation of c before i and
e. Church Latin of Italy had [č], that of France [s], and that of Spain [ts] (later > [θ]
or [s]). The most deviant of these varieties was that of England, especially after the
Great English Vowel Shift had changed the English vowel system beyond recogni-
tion (from the European perspective). Thus, exempli gratia (= e. g. ‘for example’),
pronounced as something like [eksempli gratsia] in Italian Church Latin and
similarly elsewhere on the continent, came out as [egzemplay grēšǝ] in England.
This highly idiosyncratic character of English Church Latin is probably one of
the reasons why speakers of English nowadays affect the Italian pronunciation
of Latin when singing compositions based on Latin texts, such as Bach’s B-Minor
Mass or Mozart’s Requiem, and when citing Latin expressions such as requiescat
in pace ‘may (s)he rest in peace’ (thus with [pače]).
Diglossic situations may have an enormous impact on people’s lives. In order
to become literate it is necessary in effect to learn a foreign language. But to make
things worse, the foreignness of that language is not even acknowledged. Stu-
dents are expected to learn it without great difficulty, since it is “their language”.
 Dialect borrowing 303

And if they do not succeed they will be considered dunces for “not knowing their
own language”.
Of course, such difficulties are not necessarily limited to diglossic situations.
In languages like English, too, we find that many students speaking vernacular
varieties have great difficulties learning the standard. Here, too, they run the risk
of being considered ignorant.
The difference between diglossic situations and languages like English with
their less extreme distinction between standard and vernacular is one of degree,
not an absolute one. The very notion of standard, with its insistence on immuta-
ble correctness, bears the seeds of diglossia. At the same time, the idea that
there should be a standard seems to permeate all human languages, presumably
because it is useful to have a form of speech that makes it possible to communi-
cate across different linguistic and social groups and even across time. And no
matter what the society may be, standard languages enjoy the highest prestige.
What is perhaps most interesting, and even puzzling at first sight, is the
paradox that, in spite of the well-recognized prestige of the standard, vernacular
forms of language have remarkable staying power. If standard languages are so
prestigious, why don’t vernacular speakers fall all over themselves and switch to
the standard?
Some speakers do in fact do so. But an amazingly large number of vernacular
speakers refuse. They may say things like I don’t talk so good. But if asked, Why
don’t you switch?, they will answer something like What, and sound like a sissy?!
It is no doubt this attitude that is responsible for the amazing stability of ver-
naculars. And one suspects that, ultimately, diglossia results not only from the
conservatism of standard dialects, but also from the tendency of vernaculars to
remain distinct and even to increase their distinctness by linguistic changes dif-
ferentiating them from the standard. As in the Labovian view of sound change,
group identification and group membership can be a powerful factor in linguistic
behavior and linguistic change.

7 Dialect borrowing
As observed in § 1 above, special speech varieties such as argots, jargons, and
slang may interact by borrowing from each other, as in the case of pal, mentioned
in § 1.
Examples like this show that the picture of borrowing painted in Chapter 8
is incomplete. Borrowing can take place, not only between different, distinct
languages, but also between dialects of the same language. Let us conclude this
304 Language, dialect, and standard

chapter by examining in greater detail the special consequences and difficulties

of dialect borrowing, including borrowing in diglossic situations.
Because the linguistic differences between dialects are smaller than those
between different languages, dialect borrowing does not require the same amount
of adjustment in phonological and other structures and thus is much easier to
accomplish. But precisely because of the greater similarities between donor and
borrowing language, dialect borrowings are much more difficult to detect. In
many cases, they differ from native words in just one or two sounds. Fortunately,
in many cases we have enough evidence to establish that given words are in fact
borrowings, and not native words with somewhat unusual phonological devel-
opments. But in some cases it is exceedingly difficult to establish a case for bor-
rowing.
Let us illustrate the problem with the example in (12). Normally, Old English
initial f and s remain unchanged in modern English; see (12a). But in a few words
we find a Modern v corresponding to OE f, and in one case, z corresponds to s; see
(12b). In addition, there are a few other words which entered English with initial f
after the Old English period, and which also have v in Modern English; see (12c).

(12) Old English Modern English

a. for for
fox fox
feccian fetch
full full
fyllan fill
b. fana vane
fæt vat
fyxin ‘female fox’ vixen
seax ‘sword’ zax ‘roofer’s knife-like tool’
c. pre-Mod. Engl. Modern English
fente vent ‘slit in pack or side of coat’
(⇐ Fr. fendre ‘to split’)
fiole vial
(⇐ Lat. phiala ‘drinking vessel’ ⇐ Gk. phiálē)
fanneer veneer
(⇐ Germ. fournier-)

On the face of it, the examples in (12b) and (12c) look like examples of irregular
sound change. And since no analogy or other special dialect-internal process can
be invoked to account for their initial voiced fricatives, they may be considered
counterevidence to the neogrammarian hypothesis that sound change is regular.
 Dialect borrowing 305

Even in the words of (12c), which were borrowed from other languages, the initial
voiced fricative cannot be explained as resulting from borrowing. The pre-Mod-
ern English forms clearly show that the words were borrowed into English with
initial f.
A solution to our problem is possible once we consider the dialectal situation
in England. While most modern dialects preserve initial voiceless f and s intact,
a group of southwestern dialects, including the dialect of Somerset [zǝmǝrzet],
regularly change all initial voiceless fricatives to voiced ones before vowel. That
is, in these dialects we get not only vat, vane, vixen, but also vor, vox, and vill. (For
instance, in Henry Fielding’s novel Tom Jones, situated in southwest England,
Squire Western regularly talks about going hunting for voxes.)
What we need to assume, then, is that the words in (12b/c) were borrowed
from “Somerset dialects” into the speech of London, the basis for the Modern
English standard language. In fact, closer examination of the words in (12b/c)
makes it possible even to venture a guess as to how and why the words may have
entered London speech. Except for vixen, all of the words – and these are all the
words with initial fricative voicing now used in Standard English – have techni-
cal meanings in pre-modern society: vats and veneers are terms associated with
woodworking, vanes and zaxes with roofers’ work, vents with tailoring, and vials
and vats with making containers for liquids or with filling containers with liquids.
This raises the distinct possibility that the words entered London speech through
craftsmen’s jargons (compare the nautical jargon in § 3 above), and that perhaps
even vixen found its way to London through the same vehicle.
Problems can arise, too, in diglossic situations where the standard language
is in effect the linguistic ancestor of the vernacular, and where at the same time it
remains in use for centuries, side by side with its vernacular descendants. Such
situations are found especially in the case of Latin and (early) Romance, and of
Sanskrit and its later, Middle or Modern Indo-Aryan descendants.
The continued coexistence of prestige language and vernacular may lead to
the same word being borrowed repeatedly, at different chronological stages of the
vernacular and the prestige language, and with very different results. Moreover,
some of the borrowings may come from vernacularized variants of the prestige lan-
guage. Early borrowings would be more similar to dialect borrowings (as in (12)),
since prestige language and vernacular would have had little time to diverge. Iron-
ically, however, such early borrowings would later come to diverge most, because
from the time of borrowing they have the longest time and therefore the greatest
number of opportunities for undergoing linguistic changes that differentiate them
from their sources. By contrast, very late borrowings may be more like foreign
language borrowings, easy to detect, but requiring a fair amount of nativization.
At the same time, they would usually be more similar to their sources than early
306 Language, dialect, and standard

borrowings, because they have much less time to become different through lin-
guistic change. The words borrowed most recently thus would be the most similar
to the source, while the words borrowed earliest would be the most different. As
if the situation were not complex enough, borrowings may come in somewhere
in the middle, at stages when the relationship between prestige language and
vernacular is intermediate between dialect and foreign language.
By way of illustration, let us look at a few concrete examples from Spanish.
In (13), it is possible to argue that only (13a) has a chance of being inherited from
Latin. It differs most strikingly from the Latin original, with loss of the medial
-i- and with a palatal nasal ñ reflecting the resulting -mn- (just as in Lat. somnu-
‘sleep, dream’ > Span. sueño). The form in (13b) is more likely to be an early bor-
rowing, presumably from a vernacularized form of Latin, which like the vernacu-
lar Spanish lost medial i, but retained the resulting consonant cluster mn, as well
as the vowel of the first syllable. (The r seems to be a kind of nativization of the
vernacularized Latin cluster mn, which no longer existed in vernacular Spanish,
and the preceding b can be explained as epenthetic between nasal m and oral r,
see Chapter 3, § 5.2.) Finally, the form in (13c) is virtually identical to the Latin orig-
inal. This suggests that it is a very recent borrowing and consequently had little or
no chance of undergoing changes indigenous to Spanish. In Spanish linguistics,
forms of the type (13c) and (13b) are commonly distinguished as cultismos and
semicultismos, where the semi- of the latter term indicates that such words are
only half-foreign or half-learnèd, by being more firmly integrated into the fabric
of the language than fully foreign or learnèd words.

(13) Latin Spanish

a. dominum dueño ‘lord’, Modern Span. ‘owner’
b. nominem nombre ‘name’
c. nomināre nominar ‘to name, nominate’

In (14), by contrast, the situation is less certain. As in (13b/c) we find phonetically

different reflexes of (virtually) the same Latin words. But the criteria that made it
possible to distinguish inherited from borrowed reflexes in (13) do not help us in
(14). The form in the middle column is phonetically closer to Latin, but the form
on the right is semantically closer. How, then, can we be certain which of these
forms – if any – is inherited, and which is borrowed?

(14) Latin Spanish

iunctum [y-] yunto junto
‘joined’ ‘close’ ‘joint, united’
 Dialect borrowing 307

Fortunately, examples of the type (14) are rare. However, they show most strik-
ingly the difficulties faced by linguists trying to establish historical developments
in situations of extended diglossia, such as that between Latin and its Romance
descendants.
Chapter 11: Dialect geography and
dialectology
“I knowed you wasn’t Oklahomy folks. You talk queer kinda – That ain’t no blame, you
understan’.”
“Everybody says words different,” said Ivy. “Arkansas folks says ’em different, and Okla-
homy folks says ’em different. And we seen a lady from Massachusetts, an’ she said ’em
differentest of all. Couldn’ hardly make out what she was sayin’.”
(John Steinbeck, The Grapes of Wrath.)

1 Introduction
The case of fox vs. vixen and other such examples in the preceding chapter shows
that different geographical dialects can interact with each other through lexical
borrowing. However, contact between speakers of neighboring dialects tends to
be pervasive, permeating all aspects of their daily lives. The effects of contact,
therefore, commonly extend beyond lexical borrowing and involve aspects of
general linguistic structure as well, including the extension of linguistic changes
from one dialect to another.
To illustrate this point let us return to a familiar phenomenon, the Chicago
sound shift discussed in Chapter 4, § 5.4, and exemplified in (1) below. Within
Chicago, the change is considered low in prestige and is confined mainly to
certain white, generally male and even macho, working-class circles. (See also
Chapter 10, § 3.)

(1) a. have [hæv] > [hæǝv], [heǝv]

Jan [ǰæn] > [ǰæǝn], [ǰeǝn]
taxi [tæksi] > [tæǝksi], [teǝksi]
b. Chicago [šikago] > [šikægo]
John [ǰan] > [ǰæn]

Significantly, the change has not remained limited to the city in which it origi-
nated. Especially its first phase, the change in (1a) has spread outside Chicago,
first in the “bedroom communities” of the “collar counties” around Chicago, and
later to other areas further downstate. If we examine the social evaluation of the
change in these areas, we can see why the change has come to be adopted. Outside
Chicago the change has been reinterpreted as a prestigious sign of urbanization.
It is therefore being affected by speakers who consider urbanization a good thing,
especially younger, upwardly mobile people, residing in more urbanized localities.

https://doi.org/10.1515/9783110613285-011
 Introduction 309

What is even more interesting is the response of women. In Chicago, where

the change has relatively low prestige, women have been reluctant to adopt it.
Outside Chicago, where the change is considered prestigious, women seem to be
in the vanguard of those who adopt it.
Given its relatively low prestige in Chicago, one might wonder how the change
came to be accepted by any group at all. The reason is that prestige is a relative
term. What may be prestigious in society at large may be considered unacceptable
in certain subgroups. In the case of Chicago, the sound shift in (1) seems to go
along with a certain macho, tough-guy image cultivated by a subsector of the male
white working class. And what counts as prestigious speech in other strata of
society would in this group be considered sissified. (See also Chapter 10, § 5.) This
social stereotyping of the change within Chicago accounts very well for the fact
that in the city, women are reluctant to accept it. Outside Chicago, on the other
hand, the change has a flavor of being fashionable. Women, being encouraged
to be more conscious of trends in fashion than men, therefore are more likely to
adopt it.
While in the present case prestige seems to be bringing about a steady and
fairly extensive spread of the change into downstate Illinois, it is worth remember-
ing that prestige can have very different effects. As may be recalled from Chapter 5,
§ 6, prestige differences between dialects frequently give rise to hypercorrections.
A case in point is illustrated in (2) below. In a group of dialects intermediate
between northern and southern U.S. speech, stretching from Maryland in the east
to Kansas in the west, final [-ǝ] at one time had changed to [-i] or [-ɨ]. Forms like
sody and opry (as in Grand Ole Opry) therefore were the regular counterparts of
northern and southern soda and op(e)ra. However, when the pronunciation with
final [-ǝ] came to be considered more prestigious, [-ǝ] tended to replace [-i], not
only in words with original [-ǝ], but also in words with original [-i], such as Mis-
souri.

(2) sody : soda

opry : opera
Missouri : X

The difference between the steady and extensive spread of the Chicago sound
shift and the much more sporadic phenomenon of the hypercorrect spread of [-ǝ]
may be explained as follows. The prestige of the Chicago shift in downstate Illi-
nois has come about more or less spontaneously, without outside interference. Its
acceptance generally is not a matter of conscious decision but takes place below
the level of conscious awareness. On the other hand, the replacement of dialectal
[-i] by [-ǝ] was in large measure a response to a very conscious process of school
310 Dialect geography and dialectology

instruction. It is therefore imposed from the outside and not spontaneous. More-
over, one suspects that school instruction tended to remain confined to correct-
ing the pronunciation of individual words, without providing a reliable, general
rule that would make it possible for pupils to know which words should be pro-
nounced with [-ǝ] and which with [-i].
In some form or another, then, prestige can be seen to play an important role
in the manner and degree to which features spread from one dialect to another.
Now, when looking at lexical borrowing, we saw that prestige finds a counter-
force in linguistic nationalism. In dialect contact, prestige similarly may be coun-
tered by sociolinguistic polarization, as illustrated by the following two cases.
(i) The fact that the changes in (1) are highly stigmatized as working-class
in Chicago has led to an interesting response in much of Chicago speech. First,
and not surprisingly, there is no diphthongization and raising of [æ]. Second, and
more surprisingly perhaps, the vowel [a] undergoes a development diametrically
opposed to the working-class fronting, becoming backed and (slightly) rounded,
at least in the word Chicago. Thus in Chicago, the name of the city tends to be
pronounced as something like [šikɔgo], rather than working-class [šikægo] – or
general Midwestern [šikago], for that matter.
(ii) On Martha’s Vineyard, centralization of [a] in the diphthongs [ay] and [aw]
toward the position of [ǝ] served to reassert linguistically the separate identity of
the islanders over against the mainlanders. In fact, as noted in Chapter 4, § 6.3,
as the change was being implemented, both the degree of centralization and the
number of words undergoing the process were highest for those who strongly
identified themselves as islanders, and lowest for those who had a positive atti-
tude toward the mainland.

2 Patterns of dialect interaction

2.1. The Chicago sound shift revisited. Consider again the spread of the
Chicago sound shift from working-class Chicago to downstate Illinois. A closer
look shows that the spread first of all is correlated with relative distance from
Chicago. The change applies more pervasively in the counties immediately sur-
rounding Chicago, the area in which it originated; its effects become more and
more attenuated with increasing distance from Chicago. In this sense, then, the
change radiates out from its source area and slowly peters out on the periphery.
Secondly, because the change is associated with urbanization or a positive
attitude to urbanization, it is implemented more thoroughly in urbanized than in
rural areas. Its effects, therefore, may be stronger in urban locations that are rel-
 Patterns of dialect interaction 311

atively distant from Chicago than in intervening rural areas. In this regard, then,
there is a certain discontinuity to the pattern in which the change is spreading.
These patterns of spread are by no means an isolated – or recent – phenom-
enon. In the following sections let us take a closer look at two celebrated cases
from the past which are a little more challenging because of their complexity, but
at the same time, also more instructive.

2.2. The fate of long *ū in the Low Countries. During the Middle Ages, old Ger-
manic [ū] began to shift to [ǖ] in the Flemish/Dutch dialect area of Belgium and
the Netherlands (the “Low Countries”). It is likely that the change originated in
the Flemish coastal area whose port cities held special prestige at that time. Evi-
dently it spread from that area to the south, north, and east. To the south, its
spread eventually was blocked by French, and to the north, by the sea. On the
eastern boundary, the situation was more complex. The prestige of the Flemish
area here was encroaching on another prestige area, dominated by the Low
German union of commercial cities known as the Hanseatic League, where [ū]
was retained unshifted.
In between these two areas we find something like a no-man’s land. Here
the change began to peter out, leaving in its wake speech islands where the shift
of [ū] to [ǖ] took place incompletely, affecting some words and leaving others
unchanged. Especially interesting are the differences in connotations between
shifted and unshifted lexical items. Shifted forms tend to have more prestigious
connotations. This is for instance the case for [hǖs] ‘house’, something that one
might brag about in talking to one’s neighbors (as in Come over and look at my
new [hǖs].) Unshifted forms occur in more “homey” vocabulary, such as [mūs]
‘mouse’, referring to objects that one would be less likely to bring to one’s neigh-
bors attention. Here the prestige of the innovating dialects evidently is reflected in
the choice of words permitted to adopt the innovating pronunciation.
Moreover, as in the case of the Chicago working-class vowel shift, this inter-
mediate area exhibits instances of discontinuous spread. The change may leap-
frog over territory that is only incompletely affected by the change, or not affected
at all. Compare the eastern periphery of Map 1 below, with its pockets of solid [ǖ]
territory within the larger [ū/ǖ] area, as well as the [ū/ǖ] enclave in the northeast,
in otherwise solidly [ū] territory.
What complicates matters is that in the sixteenth and seventeenth centuries,
a new development affected a smaller, relatively central part of the area: The [ǖ]
which had arisen from earlier [ū] now diphthongized to [öü] in the coastal cities
of the Netherlands; and the prestige of these cities led to the spread of this innova-
tion to most of the territory that had participated in the earlier change. The effects
of this change are ignored in Map 1.
312 Dialect geography and dialectology

Map 1: Outcomes of *ū in the Low Countries (early stage)

3 Focal, transition, and relic areas

Disregarding differences in detail, all the cases of dialectal spread that we have
examined so far share a general pattern.
There is a focal area within which the change originates and in which the
process is regular. This area contains the home dialect of the change, but may also
include neighboring territory into which the change spread early and thoroughly.
On the other end of the spectrum is a relic area, or several such areas, which
has (or have) not been affected by the spread.
Between these, there may be a transition area in which the spread loses in
generality or peters out. The extent of this degeneralization may be determined
by lexical factors, as in Dutch [hǖs] vs. [mūs], but they may be determined by
structural features as well (see § 5 below). The prestige factors responsible for the
incomplete spread remain transparent in some cases; see especially Dutch [hǖs]
vs. [mūs]. In many other cases the effect of prestige may no longer be discernible.
What is discernible, however, is the decreased regularity of the change.
 Dialectology as a diagnostic tool 313

Focal areas tend to be geographically or socially central areas, whereas relic

areas tend to be geographically or socially outlying or otherwise remote areas.
Thus in the case of the Low Countries, the focal area consisted of the socially
central, coastal regions. Relic areas are typically found on the periphery, relatively
far removed from the focal areas. However, other factors may play a role, too. For
instance, location in a relatively inaccessible mountain area may turn a dialect
into a relic area, even if it is geographically quite close to the focal area. On the
other hand, rivers ordinarily do not present obstacles to spread. In fact, they often
are conduits for accelerated spread because they tend to serve (or at least histori-
cally have served) as commercial links between the communities living on either
side. (Interstate highways nowadays play a similar role in the US.)

4 Dialectology as a diagnostic tool

As we have just seen, an examination of historically verifiable situations shows
that incomplete spread of a change may lead to irregularities in transition areas
(such as Dutch hǖs : mūs). This insight can be drawn on to account for similar pre-
historic situations, even if we may not have any direct evidence for the historical
sequence of events. This is especially important as regards sound change, for the
notion of transition area can be used to eliminate apparent counterexamples to
the neogrammarian hypothesis that sound change is completely regular. To illus-
trate the point let us look at a specific case in early Indo-European.
During the last century Indo-Europeanists were excited by the discovery
that at a very early time, a group of Indo-European languages changed the Pro-
to-Indo-European (PIE) palatal stops into sibilants. Compare PIE *ḱm̥tom ‘100’ >
Avestan satǝm. Other languages preserved the stop articulation, as in the Latin
outcome centum [k-] ‘100’. The two groups of languages are distinguished as
“satem” vs. “centum” languages (using the Avestan and Latin outcomes of the
word for ‘100’). In its most regular and general form, the change can be formu-
lated as in (3). Let us refer to the change as satem-assibilation.

(3) PIE *ḱ etc. > tš > š, s (etc.)

The reason for the excitement was that the satem-assibilation appeared to be a
very early development which boldly divided the Indo-European languages into
an eastern satem-branch, which among others included Indo-Iranian, Slavic, and
Baltic, and a western centum-branch, embracing Greek, Italic, Celtic, and Ger-
manic. The discovery that Hittite (to the south) and Tocharian (on the far eastern
314 Dialect geography and dialectology

periphery) do not exhibit satem-assibilation and thus are centum languages has
diminished the excitement somewhat, since this distribution calls into question
the earlier assumption of a clear east-west distinction. Nevertheless, there can be
no doubt that satem-assibilation is a very early change, one which possibly took
place within a Proto-Indo-European dialect continuum.
What is more important in the present context is that the generality and reg-
ularity of satem-assibilation is not evenly distributed. Indo-Iranian shows the
change in its most complete and regular form. Slavic and Baltic, by contrast,
exhibit the effect of the changes in some words, while others show unshifted
sounds. Even doublets can be found, in which the same original root shows
reflexes with and without shifted sounds. So it might appear as if sound change is
irregular in Slavic and Baltic. Compare the data in Chart 1. (Latin here represents
all of the centum languages; Lat. c = [k].)

PIE Sanskrit Avestan Lithuanian Old Ch.Slavic Latin

*ḱlew/ḱlu- śru- sru- klau-s-ī-ti slyšati in-clutus

‘hear, listen’ vs. šlōvē ‘fame’ slava ‘fame’ ‘famous’

*sweḱuros śvaśura- xvasura- šešuras svekŭrŭ socer

‘father-in-law’

*ḱm̥tom śatam satǝm šimtas sŭto centum

‘100’

Chart 1: Satem- vs. centum-forms

Keeping in mind the geographical distribution of the dialects (see Map 3), we are
justified in explaining the distribution of satem-assibilation outcomes as follows.
The change originated in a relatively centrally located focal area which included
Indo-Iranian. The centum languages on the periphery are relic areas in which
the change did not take place at all. Area II, taking in Baltic and Slavic, with its
irregular outcomes, is a transition area between the focal area and the western
relic areas; and as usual, within that transition area the change spread incom-
pletely. So the irregularities in Baltic and Slavic find a dialectological explanation
and need not be considered an exception to the hypothesis that sound change is
regular.
 Isoglosses and the problem of defining regional dialects 315

Map 2: Satem-assibilation in early Indo-European

5 I soglosses and the problem of defining regional

dialects
The cases we have examined so far have been relatively simple, with just one focal
area and with the transition area (more or less) neatly located between the focal
and relic areas. However, even in the case of a single change, the spread patterns
may be much more complex.
Consider the case of Old High German. Here, the voiceless stops that devel-
oped by Grimm’s Law from PIE voiced stops (Chapter 4, § 2), underwent a second
change that was similar to the part of Grimm’s Law that changed voiceless stops
to voiceless fricatives. In idealized form, the changes which implement the shift
can be formulated as in (4), disregarding certain minor complications.

(4) a. Voiceless stops become affricates initially and after consonant (except
after s).
Hence p > pf, t > tz, k > ch [(k)x].
b. They become “strong” (i. e. double) fricatives elsewhere (except, again,
after s).
Hence p > ff, t > zz, k > hh, where zz spells a long, dental sibilant [s̪ s̪] and
hh a long velar fricative [xx]

This shift, however, is completely regular only in Alemannic (Area I in Map 3

below), spoken in the southwest, along the upper Rhine valley. Outside the Ale-
mannic focal area, the results of the High German consonant shift decrease in
316 Dialect geography and dialectology

generality and regularity with increasing distance from that area. In some dia-
lects, such as Bavarian (Area II), the change arrives later and some relic forms
with unshifted p survive. Outside Alemannic and Bavarian, initial k remains
unchanged. The shift of initial p to pf is limited to only these two areas, plus East
Frankish (Area IV). Along the Rhine valley, the change of p to pf after consonants
gradually peters out, with Area Va showing the change after all consonants, Area
Vb showing it only after r and l, but not after m, and Area VI leaving p unchanged
after all consonants. Moreover, while all other areas shift non-initial t to zz in all
words, including the pronoun that > thazz ‘that’, Area VI retains that unchanged,
but changes t to zz in “normal” words, such as fat > fazz ‘vat, barrel’. Finally, Area
VI represents the relic area in which the sound shift failed to apply entirely.
Complexities of this sort are difficult to display on ordinary maps, such as
Map 3. Dialectologists employ a different approach to do so, namely the concept of
isoglosses, boundaries that mark the territory in which a given pronunciation or
other linguistic phenomenon is or isn’t found. Such an approach provides a much
more informative account of the complex spread pattern of the Old High German
sound shift; see Map 4 (which ignores Langobardian, a dialect that soon gave way
to varieties of Romance).

Map 3: Spread of Old High German sound shift

 Isoglosses and the problem of defining regional dialects 317

Map 4: Old High German sound shift: Isoglosses

Map 4, however, is quite incomplete, by focusing on just one phenomenon that

differentiates different dialects of Old High German. A more complete representa-
tion would look something like Map 5, whose added isoglosses reflect a variety of
developments and phenomena, partly phonological, partly morphological. For
instance, the isogloss just barely reaching into Area VI in the northwest indicates
a change of f to x in forms such as luft > lucht [luxt] ‘air’ that originated in the Low
Frankish of present-day Belgium and the Netherlands. And the east-west isogloss
cutting through Alemannic and Bavarian (indicated by “+” signs) separates a
northern area in which xs became ss (as in oxso > osso ‘ox’) from a southern relic
area where it did not. Most important, however, even this map is greatly simpli-
fied and offers only a glimpse of the full complexity on the ground.
The complex isogloss patterns of Map 5 are by no means unusual. In fact, as
time progresses, the dialect map of the High German dialects becomes even more
convoluted. If we include all the various lexical borrowings in such a map and
allow for enough time of continuous, uninterrupted contact, then the result may
well be that every word has its own history – the battle cry of early dialectol-
ogists who opposed the neogrammarian regularity doctrine.
While it probably is true that in the long run every word has its own history,
this does not justify the conclusion that the neogrammarian view of linguistic
318 Dialect geography and dialectology

Map 5: More complete isogloss map of Old High German

change is wrong. In spite of the amorphous nature of much of linguistic change,

there is also a great systematicity and regularity to language, without which it
would be difficult, if not impossible, to acquire – and understand – language
effectively. And that systematicity manifests itself most clearly in the regularity of
sound change – within its home territory. To ignore this regularity and systema-
ticity would severely limit our understanding of linguistic change.
At the same time, it is quite clear that extensive and prolonged contact, as it
is frequently found in areas long settled by speakers of the same language, leads
to isoglosses cutting across the territory in more or less crazy-quilt fashion. While
some of these isoglosses coincide with the major traditional dialectal divisions,
such as Alemannic, Bavarian, or various varieties of Frankish (see Map 3), many
others do not, but cut across these traditional dialect divisions.
This raises interesting questions. For instance, what is the basis for German
speakers continuing to think in terms of such dialectal divisions as Alemannic vs.
Bavarian, even to the present day, in spite of the fact there are isoglosses that link
Alemannic more closely with Rhine Frankish, or Bavarian with East Frankish?
What makes it possible for ordinary, linguistically untrained speakers to classify
the dialect of village X as Alemannic with strong influence of Bavarian, but the
dialect of Y, the next village to the east, as Bavarian with strong influence from
Alemannic? Are some isoglosses more salient, or more equal, than the others?
That is, do some isoglosses constitute some kind of shibboleth? (See Chapter 10,
§ 4.) And if so, which isoglosses are they and how did they come to be shibbo-
leths? There are no clear answers here, only intriguing, but real, questions.
 Migration and dialect leveling 319

6 Migration and dialect leveling

The dialect interactions examined so far basically involve stationary dialects,
whose speakers are fairly well settled. However, throughout the course of history
we find entire linguistic groups settling new territories through colonization or
conquest. A frequently observed result is that colonial speech, i. e., the speech of
groups that have left their homeland and settled in a new territory, has a much
smaller and more compact degree of linguistic diversity than the dialects that
have remained stationary. Thus, dialectal differences in general are much smaller
in America than in Great Britain. Moreover, within the United States dialectal dif-
ferences become smaller, and fairly homogeneous dialect areas become larger,
the farther west you go. In the Pacific west, the last-settled area of the United
States, dialect differences are the smallest, and only two major areas are gener-
ally distinguished, with relatively minor differences between them. See Map 6,
which ignores the special dialects of major cities such as Chicago, New York, or
Los Angeles, which often differ markedly. In England, by contrast, relatively small
geographic differences can be associated with striking dialect differences.
The situation is similar in the Spanish-speaking world. Within Spain, the Cas-
tilian core area has the greatest linguistic diversity. A smaller degree of diversity is
found in the South, which was settled by northern speakers after the destruction
of its Islamic (or Moorish) states, the expulsion of most of its Muslim or Jewish
population, and the forceful conversion of the rest to Christianity. In the same
year that the last of the Moorish kingdoms fell to the forces of Catholic Spain,
Columbus set sail and discovered the “New World”, much of which was settled by
Spanish speakers. And the dialectal diversity in the Spanish-speaking countries
of the Americas is even smaller than that of Southern Spain (though, again, more
recent developments have begun to introduce local differences)
The major factor responsible for this type of situation is dialect leveling.
It is a common occurrence in migration that speakers who in the homeland lived
far apart suddenly come into intensive, daily contact. Prior to migration, local
loyalties and patterns of communication prevailed, and communication with the
speakers of other dialects may have been minimal. As the result of migration,
however, the importance of original local loyalties becomes diminished, and the
job of communicating with the other settlers becomes paramount. Under these
circumstances, changes which eliminate excessive differences between different
speakers are highly favored, since they make communication and interaction
easier. The result is a dialectally more homogeneous language.
The effects of migration, thus, may in some ways be similar to the elimination
of dialect differences through the expansion of a supraregional, standard lan-
guage at the expense of the local dialects (see Chapter 10).
320 Dialect geography and dialectology

Map 6: A simplified map of major dialect areas in the contiguous continental United States

Migration, in addition, can lead to a decrease in contact between the speech

of the colonizers and that of the homeland. This often results in a relic area status
for colonial speech as compared to the homeland, for innovations which sweep
the homeland after the migration may reach colonial territory only incompletely,
or not at all. Thus the British English loss of r in syllable-final position affected
only the New England and Southern areas of what was to become the United
States, areas which during their early history stayed in closer contact with the
homeland. Other parts of the future United States remained unaffected. And as
it turns out, it was the speech of these other, r-ful areas that eventually became
more prestigious and now threatens the originally r-less dialects. (See Chapter 5,
§ 6.) Similarly the continued h-less pronunciation of herb in the United States is
an archaism compared to the British spelling pronunciation with h.
However, here as elsewhere, being a relic area does not offer immunity from
all linguistic change. Thus the strongly retroflex pronunciation of American r,
and changes like the Chicago shift of [æ] to [æǝ] etc. are testimony to the fact that
American English has undergone its own innovating changes.
Finally, where migration leads to a complete severance of communication
between colonial and homeland speech, different dialects become different lan-
guages. Without continued communication, the separated speech communities
are free to change in completely different directions by developments which over
 Migration and dialect leveling 321

time first reduce, and ultimately eliminate, any chance for mutual intelligibil-
ity – except through bilingualism (see the following chapters). This is no doubt
how Proto-Indo-European, once a single language with dialectal diversification,
turned into the different daughter languages briefly described in Chapter 2.
Chapter 12: Language spread, link
languages, and bilingualism
So keep mei Words in Mind bei Tag und Nacht:
A gute Noodlesupp’ tut Wunders workeh.
‘So keep my words in mind by day and night:
A good noodle soup works wonders.’
(Refrain of Mama’s Advice by Kurt M. Stein)

1 Introduction: Link languages and their sources

Just as in dialect continua there are supraregional forms of speech which facilitate
communication across regional differences, so in the case of distinct languages
there may arise supraregional means of communication. Whatever their origin,
supraregional languages may serve many different functions, ranging from that
of national language to auxiliary means of communication. A term that covers all
of this range is link language. (An alternative term is “lingua franca”, but this
may be confused with the early Mediterranean contact language called Lingua
Franca; see Chapter 14, § 3.3.)
Link languages and related phenomena are of interest to students of language
change because they often require the adoption of another language, generally
a second, but perhaps even a third. Adoption of another language, in turn, gen-
erally leads to changes in the adopted link language, or even in the native lan-
guage of the speaker, and thus can have profound consequences for linguistic
change.
The reasons for the supraregional use of languages are varied. They may range
from dominance by sheer power or conquest, over a combination of political and
cultural or technological/commercial predominance, to purely cultural preemi-
nence. Examples of “conquering languages” are Latin in the Roman Empire, and
English, French, and Spanish in more recent colonial empires. A combination of
political and cultural, technological, or commercial predominance is seen in the
case of the preeminence of French in eighteenth-century continental Europe, or
that of English in the present, largely post-colonial world. Cultural dominance
was responsible for the great prestige accorded to Greek as a link language in the
Roman Empire – in spite of the fact that the Romans had conquered Greece. Cul-
tural and religious reasons also were in part responsible for the fact that Sanskrit
became the predominant interregional language of medieval and much of early
modern India.

https://doi.org/10.1515/9783110613285-012
 Introduction: Link languages and their sources 323

But Sanskrit may have been helped by an almost diametrically opposed factor,
which can likewise lead to the selection of a particular language as a suprare-
gional language, namely its lack of association with any particular linguistic
group whose dominance might be perceived as a threat to the identity of other
groups. In the case of Sanskrit, it has been argued that just prior to its expansion
it was a language not tied to any particular region of India and that, in its classical
form, it was the most neutral language in a society where various forms of Prakrit
were the vehicles for Buddhism and Jainism, and where the sacred language of
Hinduism was pre-classical, Vedic Sanskrit. For further examples, see the later
discussion of English in modern India, or Koiné Greek in the empire of Alexander
the Great (covering Greece and most of the Middle East in the Hellenistic period
from about 300 BC to 300 AD).
And just as elsewhere, linguistic nationalism can act as a counterforce to
cultural or political dominance.
The relationship and interaction between these opposing forces can be seen
in many of the former European colonies, where debates rage over which lan-
guage should become the national language or the language of wider communi-
cation in the newly independent country.
On one side are those who argue for the retention of the former colonial lan-
guage as a supraregional means of communication, whether as the national lan-
guage or as an officially recognized auxiliary language. Proponents of this view
may in part be motivated by a desire to maintain the status which they derive
from the knowledge of the colonialist language, a language not accessible to the
uneducated and less privileged. Another important factor is that adoption of
the foreign language makes it possible to avoid the often violent consequences
of choosing one native linguistic variety over the others. And one should never
underestimate the advantages that the use of English offers for interaction with
the world at large.
On the other side are those who consider the former colonial language an
insult to their national identity. Moreover, they argue, only through the use of
indigenous languages will it be possible to compensate for the appalling lack of
general education that has been the legacy of colonial rule; for it is unrealistic to
expect people to become literate in an alien tongue when they do not even read
or write in their own language.
In actual fact, the arguments for both views are far from cogent. It is in many
cases possible to adopt an indigenous language which is just as “neutral” as the
former colonialists’ language and thus makes it possible to avoid the violent reac-
tions that result when a regionally, communally, or otherwise marked language
is selected. On the other hand, even if an indigenous language is chosen, that
language frequently is unfamiliar to the speakers of other indigenous languages.
324 Language spread, link languages, and bilingualism

These speakers, then, will still have to acquire literacy through a non-native lan-
guage. And even for native speakers of the chosen language, the trouble may not
be over, since almost invariably the variety of language chosen for administration
and schooling is far removed from the vernacular speech of the ordinary people
in vocabulary, syntax, and style. In fact, the relationship between the official lan-
guage and the vernacular in many cases comes very close to being diglossic (see
Chapter 10, § 6).
The case of modern India is especially illuminating. When the British ruled
India, Hindi, in its politically neutral, “Hindustani” variety (see Chapter 2,
§ 3.10.2), increasingly came to be the symbol of national unity against English, the
language of the foreign oppressor. And Hindustani was learned widely through-
out India, even in Bengal in the east and in the Dravidian south, areas whose
speakers took great pride in their own linguistic and cultural heritage. But after
independence there arose a great deal of opposition to attempts at making Hindi
the national language, resulting in fierce political debate and even riots. Those
opposed to Hindi argued for the retention of English as national language.
There are several reasons for this reaction. Perhaps the most important one
was the fear of political hegemony. Since Hindi was the mother tongue of the
largest single group of Indians, speakers of other languages considered the impo-
sition of Hindi as a national language to be a threat to their own linguistic com-
munities. This fear was especially great in Bengal and in the south, where the
regional languages were considered to have a far greater literary tradition and
prestige than Hindi.
Whatever the motivation, the opponents of Hindi came to see English as an
ideal alternative. After the departure of the British, it ceased to be a threat within
India. Moreover, it was spoken as a native language by only very small and politi-
cally insignificant groups. Unlike Hindi, therefore, it did not bestow unfair advan-
tages to large masses of native speakers. Rather, virtually all Indians were at the
same advantage – or disadvantage – of having to learn English as a second or
additional language.
As it turns out, the forces in favor of Hindi and those opposed to it have settled
into something like a permanent stalemate. The issue of whether Hindi or English
should serve as the national link language of India remains unresolved to the
present day, and both Hindi and English continue to be used.
In the modern period, there have been several attempts to avoid the difficulties
connected with having to select an existing language as national or international
link language through the creation of artificial languages such as Volapük
(invented in 1879 by the German linguist Johann Schleyer) or Esperanto (created
in 1887 by the Polish physician L. L. Zamenhof). Like Nynorsk (Chapter 10, § 5),
these are “constructed” languages, requiring a fair amount of linguistic engineer-
 Introduction: Link languages and their sources 325

ing. But while Nynorsk was constructed in the service of linguistic nationalism,
languages like Volapük and Esperanto are created in the hope of bridging the
narrow boundaries of nationalism, as languages that are truly international in
scope.
The creators of these languages therefore tried to develop languages that are
relatively free of the idiosyncrasies of regionally based languages. However, in
this regard, their success has been limited. For instance, Volapük, intended as
the pük ‘speech’ of the whole vol ‘world’, professes to derive the components of
its name from a specific language, English (vol = world and pük = speak – hard
to believe but true); and in the element pük it uses a somewhat unusual speech
sound [ü] that is absent even from many European languages and therefore poten-
tially difficult to acquire for their speakers. Esperanto insists on using a feminin-
izing suffix -ino for words referring to females, even where not necessary (as in
fraulino ‘Miss’, evidently from Germ. Fräulein, which does not have a male-refer-
ence counterpart) and in spite of the fact that many languages (such as Chinese)
get by very well without using such suffixes.
Esperanto has acquired a small, but very dedicated number of adherents
around the world who speak the language regularly and are said to use it effec-
tively for communication across linguistic boundaries. There are even said to be
some native speakers. But the attempt to make it a world language, superseding
the traditional languages with all their nationalist baggage, has failed.
Probably the greatest problem faced by artificial languages like Esperanto is
that they are artificial. There seems to be a widespread prejudice against artificial
languages, and that prejudice tends to be buttressed by arguments such as the
following. “Esperanto is not a native language and therefore cannot adequately
be used in many of the contexts that natural languages are able to serve, such
as children’s play, joking, or making love,” or “Esperanto lacks an indigenous
literature and therefore is in no position to compete with languages like English
which can boast of a long and rich literary history.” From an objective perspective,
arguments such as these are not very strong. As the case of Nynorsk demonstrates,
it is possible for a constructed language to become a native language, used in all
social contexts and even boasting its own literature. But recall that the success of
Nynorsk, too, has been less than spectacular. Only one sixth of the overall rela-
tively small Norwegian population has embraced it. Subjective – and clearly neg-
ative – value judgments (as well as inertia) are responsible for the fact that the
majority of the population continues to use Bokmaal.
The language which in today’s world comes closest to functioning as a truly
international means of communication, English, owes its status to a combination
of different factors. Its initial spread around the world was clearly driven by a
nationalist, colonialist expansion by Britain which, by the middle of the twentieth
326 Language spread, link languages, and bilingualism

century, had spread the language to every continent. But by that time English had
already lost some of its cohesion, through the development of a second stand-
ard variety, American English, in the aftermath of the American Revolution. This
variety itself had begun its own colonial expansion, especially in the Philippines.
And even though colonial empires were crumbling, first slowly and then in a
virtual avalanche, the political, cultural, and technological power of the United
States became ascendant, giving a further boost to the spread of English.
In the meantime, the form and function of colonial English were profoundly
influenced by the bilingual context in which it took root. Post-colonial English
reflects the impact of the indigenous languages of those who came to use it for
their own purposes. Through the incorporation of structural and lexical features
from the indigenous languages, new varieties of English have arisen which have
been called indigenized, such as Indian or West African English. Such indigeni-
zation has increased the development toward a pluricentric English language,
with regional – standard and nonstandard – varieties not only in England and the
United States, and not only in the other countries with large populations of native
English speakers, but also in large parts of the world where English is not used as
a native language. (See Chapter 2, § 3.3.)
In many cases, the use of English in non-native contexts was motivated by
social considerations, such as prestige or regional neutrality, as well as by the
need for a link language in a multilingual, multicultural society (as in present-day
India). The development no doubt was helped by the indigenization of English,
which made it a more adequate means of communication in its new contexts. But
conversely, the increasing use of English as a link language must have reinforced
the process of indigenization.
And just as success breeds success in many other spheres, so the increasingly
larger domain of English use, both in native contexts and as an indigenized link
language in non-native contexts, has further increased the status of English as a
truly international means of communication. Even in areas of the world such as
continental Europe, Latin America, or East Asia, where it traditionally competed
with other languages – or was not even considered a serious candidate as an inter-
national link language – English now has become the most widely learned second
language, to be used not only in order to communicate with native speakers of
English, or with speakers of indigenized varieties of English (such as Indian or
West African English), but with speakers of all the world’s languages.
Before concluding this section, it may be worth mentioning that link lan-
guages may come to coexist with regional languages in something very similar
to a diglossic relationship. In the case of Sanskrit and the Middle Indo-Aryan
Prakrits or Latin and the medieval Romance languages, we have of course clas-
sical examples of diglossia as defined in Chapter 10. But as noted in the same
 Interference and interlanguage 327

chapter, Sanskrit and Latin held a very similar position vis-à-vis the Dravidian
languages and the non-Romance languages of western Europe, respectively. Thus,
for a long time, the major language for written communication in western Europe
was Latin, while the regional languages, whether Romance or non-Romance,
were employed mainly as vernaculars.
Similarly, the prestige of Sanskrit in India, combined with the fact that most
of the Dravidian languages engaged in heavy borrowing from Sanskrit, has led
not only Indo-Aryans, but also Dravidian speakers to believe that the Dravidian
languages are descended from Sanskrit.
This raises the possibility that English may assume a similar role vis-à-vis
the languages with which it has come to coexist. However, in contradistinction to
earlier times, our modern period may have a sense of historicity too consciously
developed to permit us to lose sight of the historical antecedents and relation-
ships between languages. It is therefore unlikely that speakers of languages like
Hindi or Yoruba will come to think of their language as descended from English.

2 Interference and interlanguage

As time progresses, link languages – no matter what their origin – may be mod-
ified, often to a considerable degree. This phenomenon, similar to the effect of
dialects on standard languages examined in Chapter 10, § 5, or of the vernacular
on the prestige language in diglossic situations (see Chapter 10, § 6), is responsible
for the indigenization of English mentioned in the preceding section.
Let us take a closer look at the phenomenon by examining some aspects of
Indian English, a language which for most speakers differs considerably from the
British model.
The linguistic component most subject to indigenous influence is the vocabu-
lary, including fixed collocations. The reasons for adding to the lexicon often are
quite straightforward, such as the need to refer to objects, concepts, customs for
which there are no ready-made terms in the language chosen as a link language.
Lexical borrowing, with its various routines of nativization, may serve to bridge
this gap as in the Indian English examples of (1a). Some of these borrowings even
spread to the English used outside of India; see (1b). However, not all lexical inno-
vations are motivated by need. For instance, the Indian English forms in the left
column of (1c) replace native English expressions like the ones on the right.
328 Language spread, link languages, and bilingualism

(1) a. saree (a garment worn by South Asian women)

tahsil (an administrative unit)
lathi (a long bamboo stick used by Indian police for crowd control)
twice-born (calque of Skt. dviǰa- ‘twice-born; member of the three upper
castes’)
b. khaki
yoga
c. key bunch : bunch of keys
God-love : love of God/God’s love

Lexical innovations of this type should not be surprising. All languages have to
adjust to the needs of those who use them, whether they are native languages
or link languages largely used by non-native speakers. (Compare recent English
innovations such as supercomputer, prioritize, or dweeb.) What is more impor-
tant is that grammar may be affected, too. This is most strikingly the case in the
phonology of Indian English, which is characterized by large-scale substitutions,
similar to those which are used in the nativization of English vocabulary. As a
consequence, expressions like I am going to the station may be pronounced as
in (2), with post-dental retroflex for the English “dentals” t, d (which actually
are themselves post-dental, too, though alveolar, not retroflex), with dental d for
the dental fricative ð of English, and with various other changes. (As noted in the
Appendix to Chapter 1, retroflex consonants are marked by subscript dots, as in ṭ.)

(2) [āi ēm gōiŋ ṭū da (i)sṭēšan]

Morphology and syntax may likewise be affected. Thus the expression in (2) is
more likely to be said without the article – see (2’) – just as in indigenous lan-
guages of India. A further deviation from the native varieties of English can be
observed in (2”), found in many vernacular varieties of Indian English.

(2’) [āi ēm gōiŋ ṭū (i)sṭēšan]

(2”) [āi ēm ǰasṭ nāu gōiŋ ṭū (i)sṭēšan]

The variant (2”) reflects a general tendency of Indian English to exhibit system-
atic differences in verb formation, compared with British English, the original,
colonial source of Indian English. See the correspondences in (3). (Differences in
pronunciation are ignored in this example.) Though the use of just now in exam-
ples like (2”) and (3c) at this point is not obligatory, we find here the makings of a
complete and systematic shift in the formation of the present-tense system.
 Interference and interlanguage 329

(3) Indian English British English

a. I am knowing this I know this
b. I am going to school I go to school
c. I am just now going home I am going home

While examples like those in (1a/b) can be motivated by the need for vocabulary
adequate to the new context in which the language is used, it is difficult to moti-
vate developments like those in (1c), (2), or (3) by need. What is at work here is a
principle that can be observed in all second-language acquisition.
This principle has often been called interference or transfer, the influ-
ence of one’s native language on the structure of the acquired, second language.
Thus the Indian English example God-love in (1c) above may be considered to
have been formed on the native model of dēva-bhakti-, a compound of dēva-
‘God’ and bhakti- ‘devotion’. And the phonological substitutions in (2) impose
on English the phonological structure of Hindi and other South Asian languages,
in which post-dental retroflex consonants contrast with pure dentals, and where
the post-dental stops of English therefore are perceived as retroflex. Similarly, the
slightly aspirated voiceless stops of English words like to are replaced by unaspi-
rated stops (as in [ṭū]), rather than the much more heavily aspirated voiceless
stops found in the indigenous languages.
However, the concept of interference or transfer is not sufficient to account
for keybunch in (1c) or for the correspondences in (3). In contrast to God-love, key-
bunch cannot be motivated in terms of an existing indigenous compound. Rather,
the word must result from an overextension of the English process of compound-
ing or, possibly, of its indigenous counterpart. In either case, the resulting struc-
ture is the product of a creative process, not simple (or simple-minded) transfer
or interference.
Let us return now to the changes reflected in example (3). The substitution in
(3b) can be explained by transfer or interference, as a morphological and syntac-
tic calque of the Hindi expression in (4a) or of similar structures in other South
Asian languages. Here the Hindi participle ending -tā is translated by the English
participle (pple.) ending -ing, and the auxiliary (AUX) hū̃ (lit. ‘(I) am’) is matched
by its English counterpart am. The elements then are combined according to the
syntactic rules of English. The explanation is similar for (3a). However, there is no
pattern which would directly motivate the type (3c), whose Hindi counterpart is
given in (4b). (A literal translation of the auxiliary rahā of this construction would
be ‘remained’ or ‘remaining’.)
330 Language spread, link languages, and bilingualism

(4) a. maĩ (i)skūl ǰā- tā hū̃

I school go pple. AUX
‘I go to school’
b. maĩ ghar ǰā rahā hū̃
I home go AUX AUX
‘I am going home’

Rather, the type (3c) seems to reflect an attempt to retain the English distinction
between (3b) and (3c) and – even more important – the corresponding South
Asian distinction between (4a) and (4b), within a novel, “transfer” grammar
which encodes (4a) as I am going to school. The just now which may optionally be
used even in the British English version of (3c), then, seems to have been recruited
in order to achieve that goal.
Modifications of the “target” language in second-language acquisition thus
are not always explainable as resulting exclusively from interference or trans-
fer. They can be more satisfactorily accounted for as arising from the fact that
language learners must formulate for themselves a grammatical rule system
which will account for the target language. The formulation of that rule system is
influenced not only by the speakers’ native language but also by their – correct
or incorrect – assumptions about the nature of the target language. And in the
process, novel structures may arise which are unprecedented in both the native
and the target language.
To account for this different conceptualization of the second-language learn-
ing process and to differentiate it from the older conceptualization as interference
or transfer, the term interlanguage has been introduced. This term is used in the
remainder of this chapter, as well as in subsequent chapters.
Interlanguage is not limited to “exotic” areas of the “Third World”. It plays a
role in all second-language acquisition, being responsible for the “accent” (pho-
netic or otherwise) with which foreigners (or in many cases, their descendants)
speak our language. For instance, when Pennsylvania Dutch speakers use the
English expression Outen the light in the meaning ‘turn off/extinguish the light’,
the verb outen has come into existence as the result of interlanguage. In their
native Pennsylvania Dutch, a variety of German, the speakers would express the
idea of extinguishing or turning off a light by using the verb aus-machen, lit. ‘to
make or do or out’). Many verbs of similar structure, such as rot machen ‘make
red’, have English equivalents of the type redden. The verb outen, then, results
from extending the pattern of red : redden to out : X, because of the German par-
allelism of rot machen and aus-machen.
A more complex example of the effect of interlanguage is found in the passage
in (5a), uttered by a German graduate student when one of his American friends
 Interference and interlanguage 331

treated everybody to drinks. (The example has been slightly altered to simplify
the presentation.) Some of the peculiarities of this utterance, which was enor-
mously difficult to process for virtually all who were present, can be explained
as transfer, such as the final devoicing in [hes] = [hæz] has, [dait] = died, and [of]
= [ǝv] of, or the [v-] for [w-] in [wǝn] one. And the pronunciation [aunts] for [ants]
or [ænts] aunts was clearly influenced by the spelling. But these were not the
major obstacles to comprehension. More problematic was the initial [hes dait] =
has died, which clearly did not sound very much like English. Interestingly, it is
not motivated by German grammar either. German instead would say something
like (5b). Slowly it became clear that the student had meant to say something
like (5c).
But what had gone wrong to produce (5a)? Evidently the student had learned
that English has a similar strategy for forming questions as German, namely to
front the finite verb. He also had learned that English is different from German,
by normally requiring the finite auxiliary to be placed next to the non-finite “main
verb”, as in (5d) vs. (5e). (See also Chapter 6, § 5.) Where he went wrong was in
overextending the English pattern in (5d) to produce the question in (5a). As in the
earlier example of Indian English, interlanguage has produced a structure that is
unprecedented in either the native language or the target language.

(5) a. Hes dait van of yur aunts?

b. Ist eine deiner Tanten gestorben?
is/has one of your aunts died
c. Has one of your aunts died?
d. One of your aunts has died recently
e. Eine deiner Tanten ist kürzlich gestorben

Ordinarily the effects of interlanguage are relatively short-lived or limited to indi-

vidual learners. Moreover, different learners can have different interlanguages,
even if their native language is the same.
However, in cases where the second language serves as a link language and
is used in that capacity over an extended period, there is a greater chance that
interlanguage phenomena may become a permanent feature. This is especially
true if the target language is used as in South Asia – as a means of communication
primarily between speakers of mutually unintelligible indigenous languages, and
not with native speakers of the language. In situations of this sort, interlanguage
phenomena are less likely to be eliminated under the corrective influence of the
native form of the language. As a consequence they can become cumulative and
the results tend to become institutionalized as the linguistic norm of an entire
speech community.
332 Language spread, link languages, and bilingualism

The extent to which the effects of interlanguage have become institutionalized

in Indian English can be gauged from the fact that speakers employing varieties
closer to native-speaker English may be subjected to correction or even ridicule,
much as someone who speaks English with a foreign accent. This is what often
happens when speakers of Indian English return to India from an extended stay
in a country whose native language is English. During their stay in, say, the United
States, they may have gotten used to aspirating their voiceless stops, simply in
order to make it easier for native speakers of English to understand them cor-
rectly. But if they aspirate these stops after their return to India, they meet with
responses such as “What’s this [ṭhū ṭhū]? Do you think you’ve become an Ameri-
can? Why don’t you say [ṭū] like an Indian?”
Language pedagogues tend to use the term interlanguage in a negative sense,
as something that should be overcome by additional and better language instruc-
tion. And given their perspective, this evaluation is understandable. However,
as examples such as Indian English show, given the right circumstances, inter-
languages do not simply remain a somewhat annoying intermediate stage on the
way to complete language acquisition. Rather, they can become institutionalized,
and interlanguage may lead to the indigenization of the language, which makes
it adequate not only on a need basis but on an emotional basis, as the unique
property of its new speakers, in a manner not substantially different from the
emotional functions of, say, English for monolingual native speakers in the United
States or Britain.
In this way, then, new varieties of English are born, which come to coexist
with the older native varieties (such as British and American English) and thereby
increase the pluricentrism of the English language. Indeed, many linguists now
use the term “World Englishes”, with a plural form of the language name, to
reflect the fact that there are so many indigenized varieties of English around the
globe.

3 Code switching and code mixing

Language contact may lead to several other responses. But most of these seem to
have a lesser effect on linguistic change than interlanguage.
One common response, found among many fluent bilinguals, consists of
switching back and forth between the coexisting languages, such that portions of
a given sentence or utterance are in one language, other parts in another language.
This process is referred to as code switching. The examples in (6) illustrate how
code switching works. Example (6a) gives the pure Hindi version, while (6b–f)
 Code switching and code mixing 333

present different code-switched variants, with the English portions in small caps
and cited in ordinary orthography for easier recognition.

(6) a. kahtē haĩ ki āǰkal bahut lōgō̃ kō yah nahĩ ̃ pasand

saying are that nowadays many people to this not pleasing
hai ki rāǰ nārāin pāgal kē samān aikṭ kar rahā hai
is that Raj Narain fool like act AUX is
‘They say that nowadays many people don’t like it that Raj Narain is
acting like a fool.’
b. kahtē haĩ ki āǰkal bahut lōgō̃ kō yah nahĩ ̃ pasand hai ki Raj Narain acts
like a fool
c. kahtē haĩ ki nowadays many people don’t like it ki rāǰ nārāin pāgal kē
samān aikṭ kar rahā hai
d. kahtē haĩ ki āǰkal many people don’t like it that/ki Raj Narain acts
like a fool
e. kahtē haĩ ki āǰkal bahut lōgō̃ kō yah nahĩ ̃ pasand hai ki Raj Narain acts
pāgal kē samān
f. kahtē haĩ ki āǰkal bahut people don’t like it …

Switching occurs most easily at major syntactic boundaries (between clauses),

as in (b) and (c). On the other hand, the type (f), with a break in the middle of a
constituent (between the modifying adjective bahut and its head noun people),
borders on being ungrammatical. In addition, observe that switching tends to be
limited to syntax and morphology, without a comparable switch in phonology.
The phonology which is employed throughout such code-switched utterances
usually is the property of the speaker’s native or most dominant language. (The
English portions of (6) thus are pronounced with unaspirated retroflex stops, etc.,
much like the passages in (2) above.)
Many linguists recognize a parallel phenomenon of code mixing. While code
switching takes place on the syntactic level, code mixing is a lexical phenomenon.
Consider for instance the aikṭ kar rahā hai of (6a), where aikṭ is Engl. act, while
kar rahā hai ‘is doing/making’ is Hindi and serves to make the verb aikṭ usable in
a Hindi sentence. A more extended example, actually a caricature of code mixing,
is found in the motto of this chapter. Examples like this show that code mixing
consists of the insertion of content words from one language into the grammatical
structure of another.
It is difficult to distinguish this process from lexical borrowing. But the term
“code mixing” may perhaps be useful to refer to varieties of language use in which
the admixture of foreign elements is much heavier than in normal borrowing sit-
uations – so heavy that in extreme cases virtually all content words are of foreign
334 Language spread, link languages, and bilingualism

manufacture, while the native language furnishes only the phonology, morphol-
ogy, and basic vocabulary. Such varieties of language use are currently common
in South Asia and many other parts of the world, including varieties of Spanish
spoken in the United States. In most cases, they are limited to individuals and do
not appear to have lasting consequences.
However, in areas of South America, a heavily code-mixed language use
appears to have become institutionalized as the norm of a particular linguistic
community. In the border area between certain Spanish- and Quechua-speaking
territories, a new, mixed language, called “Media Lengua” (‘Half Language’), is
said to have arisen, whose vocabulary by and large is Spanish, while grammar
and basic vocabulary are Quechua. A similar form of speech, Michif, arose in the
Dakotas and adjoining areas of Canada in bilingual contact between early French
discoverers and speakers of Algonquian languages (especially Cree). Its nouns
almost exclusively come from French, while its pronouns, verbs, and basic struc-
ture are Algonquian. Languages like these are difficult to classify in terms of their
genetic affiliation. Should we base our classification on the vast majority of the
vocabulary? In that case, Media Lengua is Spanish. Or should we place more trust
in basic grammar and vocabulary? In that case, it is Quechua. Most historical
linguists would opt for the latter classification. But the issue is controversial; and
some linguists would consider such languages to be genetically unclassifiable or
even to constitute a new “re-rooting” of a language family tree.
The situation is at first sight similar for written languages like “High Urdu”,
“High Hindi”, Classical Modern Persian, or Osmanli Turkish. Here, too, we find
a pervasive admixture of foreign words (Arabic and Persian in Urdu, Sanskrit in
Hindi, Arabic in Persian, and Arabic and Persian in Turkish). And in certain texts,
these can reach proportions similar to Media Lengua, with virtually everything
but the morphology, the pronouns, and the function words in foreign garb. But
in these cases we are only dealing with certain written varieties of the language,
while the ordinary spoken language keeps the number of foreign words in normal
limits. Similarly, it has been observed that in the South Indian language Kannada,
the jargon of professional wrestlers is heavily code-mixed with English so that,
again, all content words are in English and only the grammatical structure, pro-
nouns, and function words are in Kannada. What distinguishes varieties like
Media Lengua and Michif is that they constitute ordinary, every-day spoken lan-
guage. Language mixture in such cases, therefore, has affected the totality of the
language.
 Substratum 335

4 Substratum
As discussed above, in contrast to code mixing and especially to code switch-
ing, the effects of interlanguage are institutionalized quite frequently, leading
to distinctively new language varieties. Examples in recent, observable history
include Indian English and West African English, both used as link languages
in their respective areas and both showing phonological, syntactic, and lexical
characteristics which markedly differentiate them from native varieties of
English. A less radical effect of interlanguage (involving Yiddish and English) can
be seen in the special variety of English that arose in New York among Jewish
immigrants.
One of the features of this variety is a much higher incidence of syntactic
structures with fronting of constituents other than subject to sentence-initial
position, such as This movie I really could do without. Note that such fronting is
widespread in Yiddish and other continental European languages. In this case,
it could be argued that the effect of interlanguage is relatively minor. While topic
fronting is increasingly falling out of favor in many varieties of English, it still
is marginally possible in the language as a whole. Yiddish speakers, thus, have
simply exploited a marginal construction of native-speakers’ English and used it
to encode a mode of discourse organization favored in their own native speech.
Other features of “Yiddish English” differ more markedly, such as expressions
like You want I should give you a ride? These, too, reflect syntactic patterns of
Yiddish (and many other continental European languages), while traditional
native-speakers’ English prefers structures like Do you want me to give you a
ride? (The institutionalization of Yiddish English seems to result from the fact
that for several generations, communication with speakers outside the Jewish
ghetto was fairly limited. Many interlanguage phenomena therefore remained
unchecked.)
Interestingly, Yiddish English, once institutionalized, took on a life of its own.
Many of its current users may know little if any Yiddish and certainly cannot be
identified as native speakers of the language. Similarly, South Asian and West
African English have become established varieties of English, learned as such by
new generations of speakers, rather than being created anew.
Extrapolating from such known cases, many linguists have postulated similar
developments in earlier, often prehistoric, contact situations. The scenario most
commonly envisioned is language shift, a situation in which contact results
from invasion and where one language (usually that of the invaders) eventu-
ally replaces one or more indigenous languages. In such situations, it has been
claimed, the substratum of the indigenous languages can have as systematic
and far-reaching an effect on the language of the conquerors as, say, the South
336 Language spread, link languages, and bilingualism

Asian languages had on Indian English. (Note that this use of the term “substra-
tum” is different from the use of the similar term “substrate” in Chapter 8, § 5.1,
where a sociolinguistic distinction between substrate, superstrate, and adstrate
is made. Both uses of the term are too well established in linguistics to permit
replacing one of them with a less confusing term.)
It has for instance been claimed that the far-reaching Western Romance weak-
ening, as in (7) below, is to be attributed to a Celtic substratum. For, it is said, the
area in which lenition is found is coterminous with the territory settled by the
Celts before the Roman expansion. Moreover, similar weakenings are found in
attested Celtic languages, such as Old Irish and Middle Welsh; see (8).

(7) Latin Spanish French

amicus amigo [-γ-] amiØ ‘friend’
vidēre veØer > ver veØoir > voir ‘see’

(8) PIE Old Irish Middle Welsh

tewtā tūath [θ] tud ‘people’

Similarly the change of Lat. ū to Fr. [ü] has been attributed to a Celtic substratum.
For, again, the Celts held Gaul before the Roman invasion, and a fronting change
of *u is found in Welsh, a Celtic language; see (9). Other linguists, especially native
speakers of German, have attributed the fronting of ū to ü to the influence of the
Germanic Franks, who gave France its first dynasty of rulers, as well as its name.
This alternative account illustrates how even an objective field like linguistics is
not always immune to political belief or bias.

(9) PIE Middle Welsh

*uḱsen- ych [ɨx] ‘ox’
*tū ti [ti] ‘you (sg.)’

The Castilian Spanish and Southern French change of f to h (> Ø) in examples like
(10) has been explained in terms of a Basque substratum. For, it is said, Basque
had no f when this change occurred.

(10) Latin Spanish

filius hijo ‘son’
farina harina ‘flour’

These examples are typical of traditional “substratist” explanations. However,

they also suffer from the typical weaknesses of most of these explanations.
 Substratum 337

For instance, it is not at all clear that the weakening of the relatively late-at-
tested Insular Celtic languages Old Irish and Middle Welsh was also a feature
of the Continental Celtic dialects of Gaul and Iberia (which died out very early).
Moreover, the process is found in many Italian dialects that are spoken in areas
not originally settled by the Celts.
As for the French change ū > ü, the Celtic fronting of u-vowels was restricted to
Welsh. There is no evidence that it occurred elsewhere in Celtic. In fact, languages
like Old Irish provide positive evidence against the assumption that u-fronting
was a general Celtic phenomenon. And here again the change in question has
parallels in other Romance dialects where Celtic influence is less certain. Mutatis
mutandis, the same arguments apply to the claim that French fronting reflects
Frankish influence.
The substratist case is best for the change of f to h. The change is not limited
to Spanish dialects that are close neighbors of Basque; it is also found in southern
French dialects (Gascon) that border on Basque; see (10’a). What lends further cre-
dence to the substratum explanation is that Gascon and the Spanish f > h dialects
(which include Castilian Spanish) are not direct neighbors of each other; rather,
they are separated by Basque territory. Still, even within Romance the change is
not limited to Gascon and Spanish. It is also found in southern Italian dialects,
such as Calabrian, which are far removed from Basque; see (10’b).

(10’) a. Latin Gascon

faber hàure ‘smith’
festus hèsto ‘festive; festival’
b. Latin Calabrian
farina harina ‘flour’
filum hilu ‘string’

Perhaps even more significant is the fact that the above changes are by no means
unusual and do not require an unusual or special substratum explanation. Inter-
vocalic weakening is so widespread that it would be more noteworthy to find a
language that did not undergo it at some point of its history than to find a language
that did. The fronting of u-vowels likewise is common and recurs in a large variety
of other languages and dialects, such as the Attic-Ionic of Ancient Greek, Slavic,
Dutch, and most varieties of the modern Scandinavian languages. The fact that
these languages were able to front their u’s without the aid of the Celts suggests
that the same may be true for French. Finally, special weakening developments
in labials, though not as common as medial weakening, are found in a number
of other languages, such as early Celtic, Japanese, Armenian, and the Dravidian
language Kannada. The specific change of f to h has a parallel in the history of
338 Language spread, link languages, and bilingualism

Japanese, as well as in Hawaiian, where f > h goes back to earlier Malayo-Polyne-

sian *p (as in *pa- ‘four’ > Samoan, Tongan fa, Haw. ha), and according to some
scholars even in ancient dialectal Latin.
Perhaps the most striking example of the way in which substratist accounts
fail to provide a meaningful explanation is that of “retroflexion”, a phenomenon
most consistently attributed to substratal influence, wherever it is found. This
approach has been carried to an extreme in the case of the Sicilian dialectal devel-
opment of tr to ṭ(r) and similar developments elsewhere in Romance. After finding
that no known language could be held responsible for this change, one linguist
claimed that it must be attributed to an unknown “substratum X”!
Even Grimm’s Law has been attributed to some unknown substratum. Presum-
ably the change was motivated by the fact that the substratum language had a dif-
ferent phonological system which required substitutions along the lines of the top
part of Figure 1. But you have to ask yourself why the substitutions were not made
in a more straightforward manner, along the lines of the lower part of Figure 1.

PIE Substratum X / PGmc.

p, t, k f, þ, x

b, d, g p, t, k

bh, dh, gh b, d, g

p, t, k f, þ, x

b, d, g p, t, k

bh, dh, gh b, d, g

Fig. 1: Substratum and Grimm’s Law

A more plausible substratist explanation of Grimm’s Law would assume that the
initial stage of the change postulated in Chapter 4, § 5.4, the aspiration of orig-
inal voiceless stops, was introduced by substratum speakers who pronounced
voiceless stops with aspiration. But similar changes are found elsewhere, as in
Xhosa and other Southern Bantu languages, as well as in Old High German (see
Chapter 11, § 2.3). Even ardent substratists would shrink from claiming that the
same substratum speakers are responsible both for Grimm’s Law and for the
Southern Bantu shift. The only way that a substratist explanation could be moti-
vated for all such cases is if it could be shown that every single observable case of
 Koinés 339

such a change is unambiguously the result of substratum influence. Otherwise,

the fact that we have no direct evidence for the alleged substrata – not to mention
the nature of their phonological systems – turns substratist explanations into
nothing better than speculations.
Substratist explanations are not limited to phonological phenomena, but
have also been proposed for syntax and other grammatical components. An
example of a substratist account in syntax is found in the case of the Greek infini-
tive. Ancient Greek had a highly developed system of infinitives, used in functions
similar to their English counterparts (such as Engl. to err in to err is human or to
go in I want to go home). In Modern Greek, the infinitives have been eliminated
(see Chapter 13, § 3). It has been suggested that a substratum language without
infinitives was responsible for the Greek loss of the infinitive, and Thracian and
Illyrian have sometimes been mentioned as likely “suspects”. However, the sub-
stratum account runs afoul of the problem of circularity, for very little is known
about either Thracian or Illyrian (see Chapter 2, §§ 3.6 and 3.12), and the claim that
they lacked an infinitive is based entirely on the effect they supposedly had on
Greek! There are also problems with the chronology, for we would have to believe
that Thracian and Illyrian had an effect on Greek syntax about a thousand years
after they were last attested.
The conclusion we draw must be that although examples like Indian English
demonstrate that interlanguage may result in far-reaching structural changes,
many of the commonly alleged prehistoric instances of substratal changes are
dubious. Not that the proposed solutions are completely impossible. In fact, in the
case of Gascon and Spanish f > h, a substratist explanation looks attractive. But in
most cases, substratist accounts simply are either unnecessary and unenlighten-
ing, or difficult to establish beyond a reasonable doubt, or both.
It is in the more specific areas of koiné formation, “convergence”, and the
development of pidgins and similar varieties of language that interlanguage plays
a more clearly relevant role. One of these, koiné formation, is discussed in the
next section. Convergence and pidginization are sufficiently different in their
characteristics to be treated in separate chapters.

5 Koinés
A special type of link language, commonly referred to as koiné, tends to arise
under very special linguistic conditions, characterized by the following features:
– The varieties of speech that are in contact with each other are closely related
languages or even mutually intelligible dialects.
340 Language spread, link languages, and bilingualism

– For cultural or political reasons, these linguistic varieties are considered by

their speakers to be of about equal prestige, each being the proper linguistic
vehicle for a group with its own cherished identity. Speakers therefore are not
willing to give up their speech in favor of one of the other speech varieties.
– No outside language suggests itself as a link language.

Put simply, koinés may be defined as deregionalized languages or dialects which

because of their deregionalization become potential vehicles as link languages
in areas meeting the above description. The mechanism for deregionalization, in
turn, no doubt lies in interlanguage.
The classical example of koiné, which gave its name to the type of link lan-
guage we are concerned with, is the Hellenistic Koiné, the koinḗ glôssa ‘common
language’ of the Greece of Alexander the Great and subsequent times. This lan-
guage transcended the local languages (or rather, dialects) of the various Greek
city states and confederations of cities, with their jealously guarded separate
political, cultural, and linguistic identities. In fact, as noted in Chapter 10, § 5, the
Koiné eventually replaced virtually all of the dialects of ancient Greece.
The Koiné was essentially a (partially) de-Atticized variant of the Attic Greek
of Athens, a city which had become one of the most important, perhaps the most
important state in Greece, culturally (in terms of its arts, literature, and learning),
politically (in terms of heading the Attic League, the main line of defense against
the Persians), and economically. Even so, without de-Atticization, its dialect
might not have been acceptable as a link language, since accepting it would have
given the undesirable impression of accepting the political hegemony of Athens.
Fortunately for posterity, a de-Atticized version arose.
This form of speech probably first came about in the Athens harbor, where
Attic came into daily contact with virtually all the other Greek dialects, espe-
cially with Ionic. Under these circumstances an interlanguage variety of Attic
could readily develop. The language received further impetus from the contact
between different seafaring Greek city states in the Attic League. Whatever its
early origins, however, this nonstandard de-Atticized Attic was ideally suited as a
link language, since it could be looked at as unaffiliated with any of the compet-
ing standard local dialects.
Some examples may illustrate the way in which de-Atticization was accom-
plished. The majority of Greek dialects have -ss- as the outcome of earlier *-ky-,
*-khy-, and other consonant clusters, while Attic has -tt-, a development shared
by only a few other dialects. The Koiné generally offers the -ss- of the majority of
dialects, as in (11a). Similarly, older -rs- was assimilated to -rr- in Attic and a few
other dialects. Elsewhere, the unchanged -rs- prevailed. Again, the Koiné went
with the majority pattern, against the idiosyncratic Attic usage; see (11b). Attic,
 Outlook 341

together with parts of Ionic, had changed earlier *-ayw- into -ā- before vowel,
whereas the majority of dialects had -ai-. And yet again, the Koiné sided with the
majority; see (11c).

(11) Attic Koiné

a. glôtta glôssa ‘tongue, language’
phuláttō phulássō ‘guard, watch’
téttares téssares ‘four’
b. árrēn ársēn ‘male’
c. elā́ā elaíā ‘olive’

The developments that brought about the Koiné did not, of course, arise from
speakers going into a huddle and deciding to subvert Attic. Rather they must have
resulted from a slow process of – semi-conscious or even subconscious – selection
of the non-Attic features which had fortuitously arisen through the Attic interlan-
guage in Athens harbor and which differentiated this form of speech sufficiently
from standard Attic to make it acceptable as a general link language.
We can see similar developments in various African koinés, especially in
the Bantu area. What is interesting is that in many cases the deregionalization is
brought about by selective simplification, the reduction or elimination of just
those grammatical features which differ most widely in the various languages and
dialects involved.
For instance, there is good reason to believe that the Bantu languages orig-
inally had either an accent which was not bound to any particular syllable, or
something more like the tonal system of languages like Chinese. However, as the
result of linguistic change, the accent or tone systems differ considerably from
one Bantu language to the other. Languages like Swahili, which appear to have
originated as koinés, instead show an accent that is fixed on the next-to-last syl-
lable of all words. The reason for this simplification may have been that by drop-
ping a feature which is idiosyncratically different from language to language, and
by substituting in its stead a completely predictable feature, Swahili achieved a
degree of deregionalization which made it more suitable as a koiné.

6 Outlook
The development of koinés brings us back to the two themes established in the
first two sections of this chapter, the development of link languages and the role
of interlanguage in linguistic contact. But note that ordinary link languages can
342 Language spread, link languages, and bilingualism

have many sources other than koinés; and the phenomenon of interlanguage is
not restricted to link languages but operates in all situations of second-language
learning. Moreover, in other link languages, interlanguage is a response phenom-
enon, conditioned by the fact that a language has come to be used as a means of
inter-language communication. In koiné-formation, by contrast, interlanguage is
the very foundation for the development of the link language; without interlan-
guage there would be no koiné.
In the next chapter we look at another area of language contact in which
interlanguage plays a fundamental role. But whereas the effects examined in
this chapter are essentially unidirectional, from substrate languages to link lan-
guages, the phenomena examined in the next chapter involve bidirectional effects
of interlanguage, with results that are perhaps even more profound than the ones
that we encountered in this chapter.
Chapter 13: Convergence: Dialectology
beyond language boundaries
He spoke more than ten different languages fluently – all in Russian.
(A common, half-joking claim about the famous Russian-born linguist Roman Jakobson,
co-founder of the modern approach to convergence studies)
The division between them, in their leading character, blends away.
(Charles Darwin, Fertilisation of Orchids, V: 159)

1 Introduction: Convergence defined

Until recently, the ideal of modern western societies has been that of the nation-
state with a single national language. There are a few European countries like
Switzerland and Belgium with a clear and long-established policy of bi- or mul-
tilingualism; and in North America, Canada has tried its hand at English-French
bilingualism. However, the linguistically inspired strife or even violence which
keeps flaring up in Belgium and the threat of Canada’s splitting in two along lin-
guistic lines demonstrates how precarious the notion of bilingualism tends to
be in modern western society. In fact, until very recently, the norm in modern
western European countries has been one of discouraging, even suppressing the
use of minority languages. As a consequence, many languages have died out (see
Chapter 15), such as Cornish, the Celtic language indigenous to Cornwall (the
southwestern tip of England), and Dalmatian, a variety of Romance once spoken
along the east coast of the Adriatic Sea. Cornish was replaced by English and
Dalmatian, as seen in Chapter 2, § 3.4, by Italian and varieties of Slavic. Other lan-
guages have undergone severe attrition in use, in the number of speakers, even in
linguistic structure, such as Upper and Lower Sorbian in Germany (see Chapter 2,
§ 3.4), or Welsh, Irish, and Scots Gaelic in the British Isles. And as observed in
Chapter 10, § 5, the repression of Catalan, Galician, and Basque in the Spain of
Franco’s fascist regime has only recently given way to official toleration, recogni-
tion, and even cultivation.
In many ways, however, this aversion to bi- or multilingualism is anomalous.
Large areas of the world are habitually bilingual or even multilingual. (Henceforth
the term bilingual and its derivatives are used to cover both bi- and multilin-
gual situations.) These include not only the Balkans, South Asia, much of Africa
and the Americas, but even premodern, medieval western Europe. In addition,
of course, bilingualism is widespread at national/linguistic boundaries, such as
between France and Germany.

https://doi.org/10.1515/9783110613285-013
344 Convergence: Dialectology beyond language boundaries

Such habitual bilingualism may be most likely to occur if the languages in

contact are more or less equal in strength, or if they are more or less equal in
prestige. However, there may be other social reasons for maintaining a pattern of
bilingualism.
In South Asia, for instance, one’s position within society, especially within
the complex caste system, has traditionally been intimately tied up with language.
Bengalis, for instance, would marry Bengalis (from the appropriate caste); and to
establish yourself as a Bengali you had better speak the Bengali language – even
if you live in Hindi-speaking territory. Religious affiliation, too, is often linked to
language, especially among Muslims, most of whom speak Urdu, no matter where
they live. In either case, in traditional Indian society you can’t simply decide not
to be Bengali, Muslim, or whatever your affiliation may be, and to give up the
language that is appropriate for your affiliation – unless you want to become an
outcast from society, unable to marry and have a family. At the same time, over
the centuries there has been considerable group mobility across political and lin-
guistic lines, generally for cultural and economic reasons. So, even if you have
to speak Bengali or Urdu at home (to establish your credentials as a Bengali or a
Muslim and therefore your place in society), you may need to know one or more
other languages to interact with the people on the outside.
Although the details differ, the situation is remarkably similar in the Balkan
peninsula of southeastern Europe. Repeated conquests and reconquests and other
migrations have in many areas introduced settlement patterns where speakers of
different languages are thrown together in the same locality or where their vil-
lages are intermingled in a crazy-quilt fashion, situations which guarantee daily
bilingual contact. At the same time, as recent developments in former Yugoslavia
remind us, the old ethnic identities – and differences – remain a potent force,
precluding linguistic amalgamation or the victory of any of the rival languages.
In short, it is not necessary to love your neighbors in order to engage in daily
bilingual contact with them.
Whatever the social motivations, in societies with long-standing bilingual-
ism, the use of different languages is in many ways comparable to the use of dif-
ferent social or regional dialects in monolingual societies. Different varieties of
speech are appropriate under different social conditions. One language might be
used at home, another on the job, a third in religious contexts. But to a large
extent their spheres of usage are mutually exclusive; each has its own appropriate
niche in society.
Conditions of this sort commonly give rise to a phenomenon that is referred
to as convergence, the increasing agreement of languages not only in terms of
vocabulary (which may in fact remain quite distinct), but especially in aspects of
their overall structure.
 Introduction: Convergence defined 345

If bilingual contact is prolonged, over several centuries, the interlanguages

that naturally arise in the process of speaking a second language can have a much
more profound effect on the structure of languages than the fairly short-lived sub-
stratum situations examined in the preceding chapter.
This is not simply because interlanguage has a longer time span within which
it can operate. Rather, with the passage of time, a syndrome of the following sort
will arise. Let the interaction begin with two languages, A and B, producing the
interlanguages AB, based on native knowledge of language A and acquired knowl-
edge of B, and its counterpart BA, based on native language knowledge of B and
acquired knowledge of A. These interlanguages, in turn, will interact with each
other, as well as with relatively unchanged A and B. The result will be a build-up
of increasingly complex and mixed interlanguages, with increasingly longer – and
more complex – series of “superscripts”. Compare the illustration in (1). In the
long run it may become impossible to determine which of the shared features
originated where.

(1) A B
B A
A B
A B A B
B B A A
A A B B
A B A B A B A B
A A B B A A B B
B B B B A A A A
A A A A B B B B

In contrast to the phenomena discussed in the preceding chapter, then, conver-

gence saliently involves not just a unidirectional effect of interlanguage, but a bi-
or multidirectional one. In fact, some evidence suggests that interlanguage itself
may have a bidirectional side to it. True, under ordinary circumstances we tend to
notice the effects of interlanguage mainly in the “accent” with which a foreigner
speaks our language. But when that foreigner returns home after many years
abroad, those who remained behind will likewise notice a certain “accent” in his
or her native speech. This is often considered an affectation, and to some extent it
may be. But to a certain degree this “foreign accent” in one’s native language is as
natural and normal as the “foreign accent” in one’s second, non-native language.
An important mechanism that can bring about this “backflow” is accommo-
dation, the tendency to adjust one’s speech to that of the persons one is speak-
ing to. Recent empirical research, both in phonetics/phonology and in syntax,
shows that accommodation is a common feature of conversational interactions,
even among speakers of the same language. Since accommodation often works
both ways, it is likely that it accounts not only for “backflow” but also for fea-
tures of interlanguage. This bidirectionality of interlanguage, then, may play an
important role in convergence, in addition to the complex interaction of inter-
346 Convergence: Dialectology beyond language boundaries

languages modeled in (1) above. Recent research has further established that the
internalized grammars of native bilinguals are different from the grammars of
persons who speak only one of the two languages, no matter which of these two it
may be.
In the process of accommodation, one suspects, a certain degree of selec-
tion plays a role, too, accelerating convergent developments and eliminating
more divergent ones. As we saw in Chapter 8, when borrowing foreign words we
generally nativize them, especially phonetically. If we fail to do so, we need to
“reconfigure” our articulation in mid-sentence. This is no doubt the reason, too,
for the fact noted in the preceding chapter (§ 3) that in code switching between dif-
ferent languages, the phonology tends to be solidly from one of these languages
(usually one’s native language). Naturally, then, somebody who has to function as
a bilingual on a daily basis likewise has to deal with the problem of reconfiguring
from one of the languages to the other. Under the circumstances, those varieties
of interlanguage will be favored which are most conducive to making the job of
switching back and forth easier. In fact, the ideal outcome would be varieties of
language that are virtually identical in linguistic structure, so that speakers only
have to plug in different words or morphological elements in order to satisfy the
requirement of maintaining one’s linguistic identity.
Note finally that for convergence to take place, it is not necessary for all speak-
ers of the involved languages to be bilingual (or even equally proficient as bilin-
guals), or for all dialectal areas of these languages to be bilingual. It is perfectly
possible for convergence to start in a relatively small area of intense bilingualism,
such as the border between two different linguistic groups. From this area, then,
the results of convergence can spread to new speakers and to new dialect areas by
the usual processes of dialectal spread (see Chapter 11). The loss of past tense dis-
tinctions in modern French, Romantsch, northern Italian dialects, and Southern
German discussed in § 5 below is a good example of such a development.

2 Convergence illustrated – Kupwar

In the most famous, well-studied, and large-scale convergence areas, the Balkans
and South Asia, the exact manner in which convergence arose is partly shrouded
in the mystery of the unobservable past and, therefore, also partly controversial.
Fortunately, there are cases where convergence has happened quite recently
and where, therefore, we have a good idea of how it came about. Perhaps the most
celebrated of these is the case of Kupwar, a small locality of about 3,000 inhabit-
ants at the border of the modern Indian states of Maharashtra and Karnataka. Its
 Convergence illustrated – Kupwar 347

population includes three major groups, speaking the following languages which
outside Kupwar are clearly distinct and whose history is well understood:

Urdu (Indo-Aryan): used by Muslim landholders, a socially prestigious or pow-

erful group.
Kannada (Dravidian): spoken by Jain landholders and Hindu craftspeople who
likewise hold a fairly high prestige.
Marathi (Indo-Aryan): used by Hindu “untouchables” and landless laborers, i. e.,
persons on the lower end of the prestige spectrum. In addition, it is the state
language and the primary means of education.

In spite of the obvious prestige differences, the languages until recently coex-
isted without any appreciable threat of one replacing the others. Within its own
communal setting, each of the groups or communities stuck to its own language
as a mark of its separate identity. In intergroup relations, however, there was a
great amount of bilingualism and multilingualism, especially among the men and
largely, though not exclusively, in favor of Marathi. Most speakers were at least
passively competent in all the languages of the locality.
This complex and intensive bilingualism is known to have extended over more
than 300 years. And during that period it brought about such a remarkable degree
of convergence that the phonology and syntax of the individual languages have
been claimed to be virtually identical. Only the vocabularies and grammatical
elements remained clearly distinct, with borrowing limited to a few lexical items.
Presumably it is this lexical distinction which makes it possible for people
to have nearly identical structures and still feel that they are speaking different,
communally appropriate languages. This should actually not be surprising, given
that non-linguists, when trying to characterize differences between dialects or
languages, find it much easier to talk about different lexical choices than about
structural differences.
The structural parallelism of the examples in (2) below may provide an initial
glimpse of the nature of this convergence. While there are some differences in
detail, such as the fact that the “absolutive marker” is an independent (but clitic)
word in Urdu and Marathi, but a suffix in Kannada, there is an exact, word-by-
word, suffix-by-suffix parallelism in the linear arrangement of the sentences and
in the meanings and functions of the morphological elements and words that are
used. Put differently, the sentences are exact calques of each other. (The non-ital-
icized words represent one of the rare examples of recent lexical borrowing; the
source language is Urdu. – Abs. = marker of the so-called absolutive, a non-finite
verbal form which functions as something like a verbal adverb and whose literal
translation is ‘having VERBed’; TA = tense/agreement marker.)
348 Convergence: Dialectology beyond language boundaries

(2) Ur. pālā ǰarā kāṭ kē lē kē ā- yā

Ma. pālā ǰarā kāp - un ghē - un ā- lō
Ka. tāplā ǰarā khōd – i tagōnd – i ba- yn
greens some cut Abs. take Abs. come TA
‘Having cut some greens, having taken (them), I came.’
= ‘I cut some greens and brought them’

What complicates matters is that much of the structural agreement in this example
is shared by (virtually) all the South Asian languages, as a result of historically
earlier and geographically more widespread convergence. But the Kupwar varie-
ties of Urdu, Marathi, and Kannada have converged far beyond the ordinary con-
vergence of the South Asian languages, specifically of the ordinary varieties of
Urdu, Marathi, and Kannada that are spoken outside Kupwar.
Thus, Urdu and Marathi both have arbitrary or grammatical gender, just like
German (see Chapter 8, § 3): Nouns not referring to human beings are arbitrarily
assigned masculine (see Urdu pālā in (2)), or feminine gender (see Urdu kitāb
‘book’). And Marathi has yet a third, neuter gender. Like Marathi, Kannada has
a three-gender system, but with a clear semantic basis for gender assignment,
like English: Nouns referring to male humans are masculine; to female humans,
feminine; and all others are neuter. The similarities and differences of these three
systems are summarized in (3).

(3) masculine feminine neuter

Ur. ± human ± human
Ma. ± human ± human ± human
Ka. + human + human – human

Just as in German, nominal gender assignment has important grammatical con-

sequences, for it controls agreement on adjectives and verbs, as in (4). The dif-
ferences between the Urdu, Marathi, and Kannada systems therefore were not
resolved simply by giving up gender marking altogether. Rather, the semantically
more transparent system of Kannada was extended to Kupwar Marathi and Kupwar
Urdu. As a consequence, Kupwar Marathi has neuter agreement for non-human
nouns that are masculine or feminine in Standard Marathi. And Kupwar Urdu,
lacking a separate neuter gender, employs the “unmarked” masculine gender for
the same purpose. Compare the examples in (5). (Here and elsewhere below, St. =
Standard, Ku. = Kupwar variety; m. = masculine, f. = feminine, n. = neuter; Adj. =
adjective, N = noun, V = verb.)
 Convergence illustrated – Kupwar 349

(4) Ur. aččhī kitāb paṛhī

Adj. f. N f. V f.
good book read
‘(I) read a good book.’

(5) St.Ur. vahā̃ nadīā-ī

f. f.
St.Ma. tith nadīā-lī
f. f.
Ku.Ur. hvā nadīā-yā
m.
Ku.Ma. tith nadī ā-lō
n.
Ka. yalli hwaḷi ba-ttu
n.
there flood came
‘A flood came there.’

In other areas of structural disagreement, the pattern of Urdu and/or Marathi

has won out. Thus, Standard Kannada does not have a verb ‘to be’ after predi-
cate adjectives. On the model of Urdu and Marathi, Kupwar Kannada has come to
employ a form of ‘be’ in this context; see (6).

(6) St.Ka. ii mane nim-da Ø i-du nim mane Ø

Ku.Ka. id mani nimd eti id nimd mani eti
Ku.Ma. hē ghar tumčā hāi hē tumčā ghar hāi
Ku.Ur. yē ghar tumhārā hai yē tumhārā ghar hai
this house your is this your house is
‘This house is yours.’ ‘This is your house.’

Example (6) further illustrates that as in the development of koinés, convergence

may bring about selective simplification. In Standard Kannada, adjectives and
demonstratives appear in two morphologically distinct forms. One occurs when
the form is used as a predicate or as a noun (as in nim-da, i-du), the other is used
elsewhere (as in ii, nim). Kupwar Kannada has given up the distinction in favor of
the predicate/nominal forms (as in id and nimd).
Significantly, then, convergence in Kupwar has not been unidirectional, with
one language doing all the giving and the others doing all the taking. Rather, the
common features of the Kupwar varieties of Kannada, Marathi, and Urdu reflect
influence from all of the three languages.
350 Convergence: Dialectology beyond language boundaries

As it turns out, the situation just described has recently come to an end. In
Maharashtra, as in most of the other states of India, there is a tremendous pres-
sure to learn the official state language in order to get a job. Kannada-speaking
Jains are encouraging their children to speak “proper” Marathi and Urdu-speak-
ing Muslims tend to send their sons to Marathi-medium schools. The older local
varieties of Kannada, Urdu, and even Marathi are fading out. Convergence, thus,
does not thrive when there is strong pressure to shift one’s language loyalty.

3 The Balkans
One of the classical examples of convergence is found in the area of the Balkans
(Map 1), the mountainous peninsula in southeastern Europe that is home to lan-
guages which belong to five distinct subgroups of Indo-European: (i) Bulgarian,
Macedonian, and the national languages that take the place of the former “Ser-
bo-Croatian”: Bosnian, Croatian, Montenegrin, and Serbian, abbreviated as BCMS
(all Slavic); (ii) Romanian and the related Aromanian and Meglano-Romanian
(all Romance); (iii) Albanian; (iv) Modern Greek; and (v) Romani (the Indo-Ar-
yan language of the so-called Gypsies). In addition, the non-Indo-European lan-
guage Turkish has figured prominently in the Balkans. Other languages, too, are
found in this region, but they tend to be less well represented than those men-
tioned above, and in any case do not show the same extent of convergence. This
holds as well for various dialects of the convergent languages. In particular, in
the southeastern dialects of Serbian, called Torlak, show the greatest extent of
convergence; and in Albanian, the northern (Geg) dialects show less convergence
and fewer convergence features than the southern (Tosk) dialects. Significantly,
the entire area is characterized by a high degree of long-standing bilingualism
and multilingualism.
It should therefore not be surprising that the languages of the area have over
the centuries come to share remarkable similarities in structure, especially in the
past 800 years or so. Many of the shared features, or “Balkanisms”, are clearly
common innovations, since they were absent from Proto-Indo-European or from
earlier forms of the Balkan languages (in particular, Ancient Greek, in the case of
Modern Greek; Latin, in the case of Romanian; Old Church Slavic, in the case of
Bulgarian, Macedonian, and Torlak Serbian; and Sanskrit, in the case of Romani).
What makes the languages of the Balkan so interesting, then, is not just their
mutual convergence, but also their individual and collective divergence from their
historical antecedents, as well as from their non-Balkan relatives. For instance,
the innovated features which Bulgarian, Macedonian, and Torlak Serbian share
 The Balkans 351

Map 1: Political map of the Balkans

with the other Balkan languages are absent from the rest of Slavic; and the case
is similar for Romanian compared to the rest of Romance. Moreover, as becomes
clear from the following presentation, even though we know a great deal about
the earlier history of most of the Balkan languages, there is still ample room for
uncertainty or scholarly disagreement as to how the Balkan convergence area
arose.
In fact, it was the study of this area that gave rise to one of the terms used to
refer to such areas, sprachbund, lit. ‘language league’ (in German), a term which
has been adopted in many English-language publications on convergence. Other
terms used in English are convergence area and linguistic area. Of these,
“convergence area” is probably the most transparent in English and therefore is
used in the remainder of this chapter.
The features shared by all or most Balkan languages range over phonology,
morphology, and syntax. Unlike the Kupwar situation, there are also many shared
352 Convergence: Dialectology beyond language boundaries

loan words, many of them having spread from one Balkan language or another,
but also some that have diffused from Turkish, which was very influential despite
being a relative late-comer to the Balkans. Compare examples like Mod. Gk.
ðrómos, Alb. dhrom, Bulg., BCMS Rmn. drum ‘way, road’, Rmi. drom (from Greek),
or Mod. Gk. [boyá] (spelled μπογιά = mpogiá!), Bulg., Mac., BCMS boja, Rmn. boia,
Alb. bojë ‘paint, color’ (from Turk. boya).
The important structural features shared by many or most of the Balkan lan-
guages include the following:
– The absence of long vowels. Even though earlier stages of the languages
had a contrast between long and short vowels, most of the members of the
Balkan convergence area do not have contrastive vowel length. As in the case
of other Balkan features, however, there are exceptions, in some areas on the
periphery. Thus, vowel length is found in BCMS, except for Torlak Serbian
(geographically closest to Bulgarian and Macedonian); and some Albanian
dialects have long vowels as the result of more recent, secondary develop-
ments.
– A postposed definite article, as in (7). This feature is found in Albanian, Roma-
nian, Bulgarian, Macedonian, and Torlak Serbian. The rest of BCMS does not
use a definite article at all; and Greek and Romani have a preposed article, as
in Gk. o fílos ‘the friend’, Rmi. o phral ‘the brother’. The articles of the Slavic
and Romance Balkan languages have outside relations, such as the demon-
strative pronoun (*)tŭ ‘that’ in Slavic or the article of It. il duomo ‘the dome,
Fr. le chat ‘the cat’, Sp. el lobo ‘the wolf’. However, outside the Balkans, the
articles and demonstratives are preposed; see Sp. el lobo vs. postposed Rom.
lupu-l. Note further that though the Balkan languages (other than Greek and
Romani) agree on the postpositive placement of the article, they do not agree
on its form. Rather, each language employs an indigenous form.

(7) plain noun noun + article

Mac.
Bulg. voda voda-ta ‘water’/‘the water’
Torlak
Rom. lup lupu-l ‘wolf’/‘the wolf’
Alb. shok shok-u ‘comrade’/‘the comrade’

– Use of dependent clauses in place of infinitival structures; see for example

(8a). The development has been carried through most consistently in Greek,
Macedonian, Torlak Serbian, and Bulgarian, and is found robustly as well in
Romanian and Tosk Albanian. It peters out in the rest of BCMS, with Serbian
generally using the dependent-clause structure and the more northerly
 The Balkans 353

and westerly Croatian preferring the older infinitive construction; see (8b).
Related non-Balkan languages and earlier stages of the Balkan languages use
infinitives, as in Fr. je veux écrire or Ancient Greek thélō gráphein ‘I want to
write’.

(8) a. Mod. Greek θélo na γráfo

Rom. vreau să scriu
Alb. dua të shkruaj
Mac. sakam da pišam
Bulg. iskam da piša
Rmi. kamav te hramonav
I want that I write
‘I want to write.’
b. Serbian hoću da pišem
I want that I write
Croatian hoću pisati
I want to write
‘I want to write.’

– The marker for the future tense is based on the verb ‘want, wish’ used as an
auxiliary, except in certain, mostly northern (Geg) Albanian dialects which
use the verb kam ‘have’. Compare the examples in (9). This future marker
generally is an invariant, uninflected particle and is followed by the depend-
ent-clause construction illustrated in (8); the only exception is BCMS which,
outside the Torlak dialects, uses an auxiliary that is inflected for person and
number and can combine with an infinitive. Historically, the invariant future
markers derive from full verbs, generally third person singular forms, which
usually have undergone clitic reduction (as with Engl. is to ’s; see Chapter 4,
§ 5.5). For instance, Gk. θa derives from θéli na (lit. ‘it wants that’; see (8a))
and BCMS ću from the hoću of (8b). Again, the outside languages and earlier
stages of the Balkan languages have very different constructions; see (10).

(9) Gk. θa γráfo ‘I will write’ (*‘it wants that I write’)

Rom. o să scriu
Alb. do të shkruaj
Mac. k’e (da) pišam
Bulg. šte da piša
Rmi. ka hramonav
vs. BCMS pisa-ću (*‘to-write I-want’)
354 Convergence: Dialectology beyond language boundaries

(10) Span. escribir-é < escribir hé ‘I will write’ (originally ‘I have to write’,
i. e., infinitive of ‘write’ + ‘have’)
Anc.Gk. gráp-s-ō ‘I will write’ (= ‘write’ + future suffix + first singular
ending)
Russ. budu pisat’ ‘I will be writing’ (= ‘I am’ + infinitive of ‘write’)

While the present-day facts are well established, the historical developments
responsible for these facts are much less certain.
As far as the elimination of long vowels goes, the best that can be said at this
point is that it seems to have been a communal effort, so to speak. It could well
be the result of selective simplification, stripping away more complex features, a
move toward a lowest common denominator system, as it were.
For the remaining, syntactic features, a number of different theories have
been proposed or are at least conceivable. Most of them are problematic. In fact, it
may be inappropriate to expect all of the features to have developed from a single
source, rather than through the same kind of give-and-take observed in Kupwar.
An older theory tried to attribute all of the syntactic Balkan features (as well
as others) to Greek. While Greek origin could be argued for some of the features,
the postposed article can hardly be explained this way, since the definite article
of Greek is preposed, not postposed, and has been preposed since Ancient Greek
times.
Since postposed articles, the replacement of the infinitive by dependent-clause
constructions, and the use of ‘want to’ as future auxiliary cannot be traced to the
ancestors of Greek, Romanian, or the Balkan Slavic languages, it would be tempt-
ing to attribute them to Albanian. The “advantage” of this assumption is that the
earlier history of Albanian is unknown and attempts to trace it back to other early
Indo-European languages like Illyrian and Thracian are problematic. But that is
also the disadvantage – we would simply be trying to explain the unknown by the
even less known. Moreover, the fact that northern Albanian (Geg) uses the verb
‘have’ as future auxiliary makes Albanian an unlikely source for the more general
Balkan use of ‘want’.
A more viable hypothesis is the claim that the loss of the infinitive is of Greek
provenience, since the development is found attested earliest in that language.
This view finds support in the fact that the replacement has taken place most com-
pletely in Greek and in the Slavic languages neighboring Greek – Macedonian and
Bulgarian. By contrast, Romanian, for instance, shows some productive (though
limited) uses of its old infinitive.
A further complication arises from the fact that some scholars have argued
that both the loss of the infinitive and the use of ‘want to’ as future auxiliary may
have arisen in a convergence area of the Late Roman Empire that included both
 The Balkans 355

Greek and Latin. According to this hypothesis, it was Latin that was indirectly
responsible for the infinitive loss; for even in Old Latin, infinitive and depend-
ent-clause structures coexisted as alternatives, while early Greek preferred
infinitives. Even if this hypothesis is correct – and it is by no means generally
accepted by Balkanists – the subsequent elimination of the infinitive in favor of
the dependent-clause structure could have been a Greek contribution. Latin and
its non-Balkan Romance descendants did not participate in this development.
The case is similar for the future auxiliary, but the evidence is a bit more
robust. Again, it is claimed that the late Roman Imperial convergence area used
both ‘want to’ and ‘have to’, as well as ‘begin to’ as auxiliaries. In fact, this triple
choice is also found in Old Church Slavonic, the earliest attestation of South
Slavic, as well as in Gothic, the language of the earliest coherent Germanic
texts, which happens to have been spoken at the periphery of the late Roman
Empire including in the Balkans. This suggests that the use of ‘want’ in most of
the Balkans (as in (9)) and ‘have’ in (most of) non-Balkan Romance (e. g. Span.
escribir-é, as in (10)) could have been the result of different attempts to resolve the
competition between the two auxiliaries. Further evidence for this view may be
seen in the fact that there are exceptions on both sides of the divide: Geg Albanian
uses ‘have’, and certain Italian dialects use ‘want’, and in some of the Balkan
languages, ‘want’ vs. ‘have’ futures are grammatically differentiated, with ‘have’
in negative futures and ‘want’ otherwise. What is puzzling is why none of the
languages selected ‘begin’ as their future auxiliary.
Unfortunately, in this case it is difficult to tell whether the use of future aux-
iliaries originated in Greek or in Latin – or possibly even in South Slavic. Earlier
stages of both Greek and Latin had future tenses that did not employ auxiliaries;
and for Slavic we simply have no earlier records. Moreover, the use of formations
meaning ‘want to’ or ‘have to/be obliged to’ to indicate future tense is by no means
unusual. Consider English, which uses both will (originally ‘want’) and shall (orig-
inally ‘be obliged’). So some linguists believe that the choice of one or another of
these constructions in the Balkans and in non-Balkan Romance could have come
about independently. Many other linguists, however, doubt that the use of ‘want’
in most of the Balkan languages can be attributed to chance, especially given the
existence of other salient features that are shared by all or most of the Balkan
languages. (The question of how to explain that not all Balkan languages share
all the features is addressed in Section 6 below.)
What remains, then, is that a significant number of convergent features can
be observed in the present-day Balkans and that the geographical concentration
of these features is not likely to be due to accident. It may be disconcerting that
there is controversy and uncertainty as to how the convergence came about, but
this is by no means unusual. The very fact that there is controversy over the his-
356 Convergence: Dialectology beyond language boundaries

torical origin of Balkan convergence demonstrates the significance and intrinsic

interest of the phenomenon.

4 South Asia
Another famous convergence area is that of South Asia: Beside Burushaski in the
Northwest, for which we have no known outside relations, there are at least four
major linguistic families which over the course of millennia have come to show an
increasing agreement in a large number of overall structural features. These are:
– Indo-Aryan and some of the neighboring Eastern Iranian languages (such as
Pashto), belonging to the Indo-European language family
– the Dravidian languages, which may perhaps be distantly related to the Uralic
or Finno-Ugric family
– the Munda languages, related to Southeast Asian (“Austro-Asiatic”) languages
such as Mon and Khmer
– Tibeto-Burman languages on the northern periphery of South Asia, which
share many features of the convergence area

For the approximate location of these language groups in modern South Asia, see
Map 2.
All of these languages tend to share certain features. Exceptions do occur,
such as some Munda languages, which lack absolutives, or Kashmiri, which has
innovated in the area of word order; but such exceptions are rare. The common
features are as follows:
– A contrast between dental and retroflex consonants, as in (11)
– An unmarked SOV order, as in (12)
– The tendency to use “absolutives”, something like verbal adverbs, where
European languages would employ dependent or coordinate clauses; see (13)

(11) Skt. pāta- ‘flight’ : pāṭa- ‘portion’

(12) Hindi maĩ kitāb paṛh rahā hū̃

I (S) book (O) read AUX AUX (V)
‘I am reading a book.’

(13) Hindi a. kitāb paṛh kē baiṭh lō

book read Abs. sit take
‘Having read the book/after you have read the book, sit down.’
or: ‘Read the book and sit down.’
 South Asia 357

b. baiṭh kē kitāb paṛh lō

sit Abs. book read take
‘Having sat down/after you have sat down, read the book.’
or: ‘Sit down and read the book.’

Map 2: Approximate distribution of modern South Asian languages

Controversy continues as to when the convergence of South Asian languages

began and which language group is responsible for the features listed above.
Many scholars argue that the source is Dravidian, since all three features can be
reconstructed for Proto-Dravidian. And, these scholars maintain, the ancestors of
the other languages lacked some, or even all of these features. Moreover, since all
of the features are found in the earliest attested stage of Indo-Aryan, it is claimed
that convergence between Indo-Aryan and Dravidian must have begun in the
second millennium BC. Under this view, then, the similarities between Dravidian
and the other South Asian languages are due, not to bidirectional convergence,
but to unidirectional substratum influence.
358 Convergence: Dialectology beyond language boundaries

Other linguists have pointed to evidence which supports the view that all
the syntactic features also are indigenous in Indo-Aryan and that the basic prin-
ciples of syntactic organization underlying these structures are inherited from
Proto-Indo-European. SOV plus absolutive structures also appear to be inherited
in Tibeto-Burman, as well as in many languages to the north and west of South
Asia, including Altaic, Uralic, and several ancient Near Eastern languages, such
as Elamite, Akkadian, and Sumerian. If these features are not inherited in the
respective languages or language families, they may have arisen in an earlier,
much larger Eurasian convergence area which extended far beyond South
Asia.
Interestingly, both Indo-Aryan and Dravidian, throughout their history, offer
an alternative to subordination by means of absolutives (or other non-finite forma-
tions), namely a special type of relative construction in which the relative clause is
not embedded into the main clause, but is juxtaposed before (or after) that clause.
In this type of construction the main clause commonly contains a “correlative”
pronoun (CP) which answers to the relative pronoun (RP) of the relative clause.
The pairing of relative and correlative pronouns, then, accomplishes the same
purpose as the English placement of relative clauses after the constituents that
they modify. Compare the examples in (13’).

(13’) a. Skt. yaḥ puruṣaḥ paṭhati saḥ (puruṣaḥ) vidvān bhavati

RP human studies CP human wise becomes
lit. ‘Which human (being) studies, that human (being) becomes
wise.’
‘A human being who studies becomes wise.’
b. Tamil evaṉ naṉṟāka ur̤ aikkiṟāṉ -ō avaṉ vār̤ kkaiyil
RP hard works Clitic Particle CP in life
muṉṉēṟuvāṉ
will succeed
lit. ‘Which (being) works hard, that (being) will succeed in life.’
‘He/She who works hard, will succeed in life.’

Unaware of their wide distribution in Dravidian and their appearance even in

the earliest Dravidian texts, some linguists have attributed these structures to
Indo-Aryan influence. Ironically, one linguist has made the converse claim, that
the relative-correlative construction of Indo-Aryan indirectly results from Dravid-
ian influence. Recent research suggests that these structures, too, are inherited in
both language families.
If the arguments that the overall syntactic agreements reflect common inher-
itance are correct, then the substratum hypothesis loses much of its cogency. It
 South Asia 359

should be noted, however, that many adherents of the substratum hypothesis

remain unconvinced of these alternative arguments.
There is controversy even over whether Dravidians and Indo-Aryans were in
contact early enough for the allegedly substratum-induced features to be present
in the earliest Indo-Aryan. One piece of evidence which seems to favor early
contact is the appearance of one modern Dravidian language, Brahui, in today’s
Pakistan, i. e., in the extreme northwest which is commonly assumed to have been
the place of earliest Indo-Aryan settlement in South Asia. But the present-day
location of Brahui could be the result of migration from further south. True, the
general direction of migrations has been from the northwest toward the south and
east of South Asia. But other Dravidian-speaking groups have moved northward.
And one Indo-Aryan group, the Roma (formerly called Gypsies), has moved even
beyond northwestern South Asia, into Central Asia, the Middle East, and thence
into Europe.
Similar uncertainties exist over another issue which might support the
hypothesis that Indo-Aryan and Dravidian were in contact early enough for
Dravidian to influence the structure of Indo-Aryan. This is the question of early
Indo-Aryan borrowings from Dravidian. The reasoning here goes as follows: If
there is evidence for such borrowings then it may be safely assumed that there
must have been contact.
Some linguists find massive evidence for borrowings. But even substratists
now consider many of these uncertain. Linguists on the other side of the argument
have claimed that none of the proposed early borrowings from Dravidian stands
up under scrutiny. But even if they are right, an absence of lexical borrowing does
not preclude the possibility of contact. As noted earlier, the languages of Kupwar
have undergone extensive structural convergence, while lexical borrowing has
been minimal.
This leaves the retroflex/dental contrast, as in (11) above. Opponents of the
substratist hypothesis have pointed out that such a contrast has arisen in other
parts of the world (see Chapter 12, § 4) and that it is therefore unnecessary to
assume Dravidian influence on Indo-Aryan. They also have observed that the
sound systems of early Dravidian and Indo-Aryan are much more divergent than
should be the case if Indo-Aryan retroflexion resulted from Dravidian influence:
As Figure 1 shows, early Dravidian actually has a triple contrast between dental
(unmarked), alveolar (post-dental; marked by underlining), and retroflex. In San-
skrit, the earliest attested form of Indo-Aryan, only [r] is phonetically alveolar;
for the rest the contrast is one between dental and retroflex. Dravidian has a ret-
roflex sound, r̤ , very similar to Modern American English [r], which is absent in
Indo-Aryan. On the other hand, Sanskrit has a retroflex sibilant ṣ that is absent in
Dravidian. And so on.
360 Convergence: Dialectology beyond language boundaries

Sanskrit Dravidian
dent. alv. retr. dent. alv. retr.
stop t ṭ t ṯ ṭ
th ṭh
d ḍ
dh ḍh
sib. s ṣ
nas. n ṇ n ṉ ṇ
liqu. l ṟ ḻ ḷ
ṟ r̤
Fig. 1: Early Indo-Aryan and Dravidian systems

Contrast this situation with the modern one, especially as it occurs in the central
area of South Asia, where Dravidian and Indo-Aryan languages are in closest
contact. Except in the extreme south and northwest, the idiosyncratically Dra-
vidian r̤ and the equally idiosyncratic retroflex sibilant ṣ of Sanskrit have been
eliminated; and so has the alveolar series of early Dravidian. Moreover, secondary
developments have given rise to a retroflex flapped ṛ, and in some of the lan-
guages to a retroflex liquid ḷ. And in both groups of languages, dental nasals are
conditioned variants of more basic alveolar nasals. Compare Figure 2. Here, then,
we have genuine convergence by way of mutual accommodation, while the early
situation in Figure 1 looks more like divergence.

Indo-Aryan Dravidian
dent. alv. retr. dent. alv. retr.
stop t ṭ t ṭ
th ṭh
d ḍ
dh ḍh
sib. s
nas. n ṉ ṇ (n) ṉ ṇ
liqu. ḻ ḷ ḻ ḷ
ṟ ṛ ṟ ṛ
Fig. 2: Modern Indo-Aryan and Dravidian systems
 South Asia 361

Finally, there is even evidence that the Dravidian contrast between dental, alveo-
lar, and retroflex stops may be secondary, the result of assimilation of dentals to
preceding alveolar and retroflex nasals and liquids, as in *ceṉ-t-ēṉ > Tamil ceṉṯēṉ
‘I went’, *āḷ-t-ēṉ > āṇṭēṉ ‘I ruled’. The claim that retroflexion is inherited in Dra-
vidian is therefore open to question.
Nevertheless, the fact that both early Dravidian and early Indo-Aryan have a
retroflex : dental contrast is difficult to attribute to pure chance, even if it looks
like an innovation in both groups. The very fact that it looks like an innovation
in both groups of languages, at roughly the same time and in roughly the same
area, makes the assumption of chance similarity even more difficult to accept. The
fact that both groups seem to have innovated, however, makes unilateral substra-
tum influence from Dravidian on Indo-Aryan (or vice versa, for that matter) just
as unlikely. Perhaps, then, we should entertain the idea that the contrast arose
from sound changes which were convergent even though they yielded different
outcomes, simply because they operated on different inputs – a retroflex : dental
contrast in Sanskrit, and a retroflex : alveolar : dental contrast in Dravidian. These
differences subsequently would have been eliminated by convergent accommo-
dating developments.
While this alternative view of retroflexion is speculative and has not been
generally accepted, it has the virtue of overcoming some of the objections to the
substratist view. Perhaps even more important, it replaces that view with a con-
vergence analysis much more in keeping with later South Asian historical devel-
opments that in most cases are convergent, rather than reflecting unilateral sub-
stratum influence.
Only for the Munda languages do we need to assume extensive unidirec-
tional influence. SOV order and the retroflex : dental contrast must be the result
of contact; for the Austro-Asiatic languages of Southeast Asia have basic SVO and
lack the retroflex : dental contrast. Note however that there is independent evi-
dence that the speakers of Munda languages have a very different social status
from that of (most) Dravidians and Indo-Aryans. Munda speakers live only in
so-called tribal societies, in relatively isolated and economically disadvantaged
areas. Their languages and customs, and they themselves are the subjects of
widespread discrimination. Further, many areas now inhabited by Indo-Aryans
and Dravidians bear place names suggesting that they originally were settled by
Mundas who must have been displaced by Indo-Aryans and Dravidians. Given
these circumstances, it should not be surprising if the Munda speakers were also
linguistically on the receiving end.
Perhaps most important, all these attempts at explanation address only
part of the South Asian picture. What ultimately remains a mystery is that the
dental : retroflex contrast is found not only in Indo-Aryan, Dravidian, and Munda,
362 Convergence: Dialectology beyond language boundaries

Map 3: Approximate distribution of the dental : retroflex contrast in South Asia

but also in Burushaski, in Iranian languages, and even in Tocharian (once spoken
in present-day Xinjiang), i. e. far to the north of South Asia. Remarkably, it is also
found in the languages of the Andaman Islands, far to the southeast (closer to
Southeast Asia than to South Asia). Further, a sizable number of Tibeto-Burman
languages have the feature as well. On the other hand, Tibeto-Burman languages
at the northeastern periphery have no such contrast but have alveolars instead,
and so do some neighboring Indo-Aryan languages, especially Assamese. (See
Map 3 for the approximate distribution in South Asia.) It is hard to believe that the
widespread distribution of the retroflex : dental contrast results from pure chance.
However, the presence of the contrast even in the Andamanese languages, which
apparently have not been in contact with the other South Asian languages for
 The dialectology of convergence areas 363

millennia, raises questions about accounts that focus only on well-documented

languages such as Indo-Aryan and Dravidian and suggests that whatever devel-
opments led to the present-day distribution must have begun far into prehistory.

5 Europe
Convergence has not always been limited to “exotic” areas. There is good reason
for believing that prior to the development of the notion of the monolingual
nation-state, much of medieval and early modern Europe was a convergence area.
In fact, convergence can be observed even in a more recent development
that affected colloquial French, Romantsch, northern Italian dialects, southern
German, and parts of Dutch/Flemish. This development consists of the replace-
ment of the simple past by the present perfect; see example (14). The phenomenon
is most widespread in French, whereas in Italian, German, and Dutch/Flemish,
it is limited to dialects that are geographically close to French. This suggests that
the change originated in French and spread from there into the neighboring lan-
guages. Within Germany it is now spreading into more northern dialects, pre-
sumably through ordinary dialect diffusion. Given the lateness of the spread into
non-French territory, it is possible that it originated in border-area bilingualism.

(14) a. Traditional Innovative

German southern German
ich bin gegangen ich bin gegangen ‘I have gone’
ich ging ich bin gegangen ‘I went’
b. Traditional Innovative
French French
je suis allé je suis allé ‘I have gone’
j’allais je suis allé ‘I went, used to go’
j’allai je suis allé ‘I went’

6 The dialectology of convergence areas

The case of the replacement of the simple past by the present perfect raises inter-
esting questions. The spread of the phenomenon from colloquial French to the
colloquial varieties of neighboring languages looks very much like the spread of
innovations in dialect areas (see Chapter 11). Moreover, the spread then continues
within German through ordinary dialect diffusion. In this regard, then, conver-
364 Convergence: Dialectology beyond language boundaries

gence areas are similar to dialect areas, and there may be relatively smooth tran-
sitions from one to the other.
But the similarities go farther. Just as in dialect areas certain features may
spread over larger territory, while others are more limited in their distribution, so
in convergence areas we may find that certain features cover virtually the entire
area, while others are more restricted in their occurrence. Consider the occurrence
of a ‘want’-future, which covers virtually the entire Balkan area (except for north-
ern Albanian), vs. the postposed definite article (which does not occur in Greek
or in most of BCMS). In fact, just as in dialect areas, it is possible to draw isogloss
maps for convergence areas; see Map 4.

Map 4: An approximate isogloss map of the Balkan convergence area

 The dialectology of convergence areas 365

This similarity between convergence areas and dialect areas should actually not
be surprising, for as noted at the beginning of this chapter, the use of different
languages in bilingual societies is in many ways comparable to the use of different
dialects in monolingual societies. The isogloss evidence now permits us to state
this insight even more boldly: Languages spoken in bi- or multilingual societies
are the functional equivalent of dialects in monolingual societies, not only in their
social function, but also in their interaction, including the spread of linguistic
features and innovations.
Convergence and convergence areas, however, have even broader signifi-
cance. As we have seen, where the evidence is readily available (as in Kupwar),
convergence commonly is not a one-way street, but rather a situation where all of
the languages involved tend to make their contributions. Even in cases such as the
general South Asian convergence area or that of the Balkans, we have seen that
attempts to attribute the shared features to just one language tend to be problem-
atic. The concept of convergence, thus, is an important alternative to the tradi-
tional notion of – one-way – substratum influence.
Chapter 14: Pidgins, creoles, and related
forms of language

For Better or For Worse© 1992 Lynn Johnston Prod. Inc. Reprinted with permission of
­Universal Press Syndicate. All rights reserved.

1 Introduction: Foreigner Talk, “Tarzanian”, and

other simplified forms of speech
In the last two chapters we examined various effects of ordinary bilingual contact
between speakers of different languages. Let us now turn our attention to more
extreme results of contact which include the development of pidgins. To set the
stage for our discussion, let us begin with a look at a rather ordinary form of
language contact which, in spite of its ordinariness, bears the seeds for the more
extraordinary developments discussed in the rest of this chapter.
In response to the enormous and ever-growing international demand for
instruction in English, the profession of teaching English as a Second Language
(ESL) has become a highly attractive and lucrative business. The ranks of ESL
teachers have been filled by many people who are highly trained and qualified
professionals. But there are also individuals whose only qualification is that they
speak English. Teachers of the latter type often do not know the native languages
of their students and the grammars of those languages. They may be quite unfa-
miliar with the formal grammar of their own language, English. And they may
even lack the most rudimentary acquaintance with language pedagogy.
Now, imagine what happens when such teachers are confronted with stu-
dents who have no previous knowledge of English – not an unlikely event, since
the students have come to them to become familiar with English. Clearly, such
situations can be highly frustrating for both teacher and students. What interests

https://doi.org/10.1515/9783110613285-014
 Introduction: Foreigner Talk, “Tarzanian”, and other simplified forms of speech 367

us in the present context is the manner in which such teachers often respond to
the situation. This response can be best illustrated by the following anecdote.
An ESL teacher, let us call him or her Jan Johnson, was confronted with a group
of foreign students who had no prior knowledge of English but were clearly eager
to learn. At the same time, the teacher had no knowledge of the students’ native
language. After spending most of the first class session explaining in English the
purposes of the course, the requirements, and so on, and feeling reassured by the
students’ polite smiles that they understood, Jan concluded the session, saying
For the next class, read chapter one.
The blank gazes in the students’ faces made Jan realize that the students did
not understand the instruction. At best, they had come to realize that the initial
pleasantries were over, and that their teacher now was telling them to do some-
thing. But what that something was clearly eluded them.
Jan’s first response was to repeat the same message, only at a much slower
pace and somewhat more loudly. The results were no better than after the first
try. One or more additional tries, at an even slower pace and even greater volume,
met with the same results.
At this point, feeling highly frustrated, Jan held up the book and said the
following – at an even slower pace, even greater volume, and in a heavy, chunky
rhythm, with each word intoned as if it were a complete sentence:

(1) Here – book. See – book? Open – page – one.

And to make sure that the students would understand their assignment, Jan
pointed to the chapter title on the first page and added in the same voice:

(2) Chapter – one. Read. Next – time.

The story does not tell us how successful this last effort was. But one suspects that
the students figured out that they were supposed to read Chapter One (or Page
One?), even if they could not actually do so – since they did not know English as
yet. At any rate, their eagerness to acquire English would sooner, rather than later,
enable them to learn enough of the language to complete the course and to employ
it in the contexts of their choice – with, of course, varying degrees of success.
Jan’s case is not an isolated one, and utterances such as the ones he or she
resorted to in sheer desperation are not limited to certain ESL contexts. They are
extremely common when any two or more people not knowing each other’s lan-
guage try to communicate. In fact, they are such a common phenomenon that
linguists have introduced a special term to refer to this type of communication,
namely Foreigner Talk.
368 Pidgins, creoles, and related forms of language

Most of us have encountered Foreigner Talk more than once in our lives; and
while our examples are drawn from English, virtually all languages – or rather,
their speakers – have a form of Foreigner Talk. Modern tourism is a prime context
for utterances such as (3), uttered to a taxi driver by a tourist afraid she might
miss her plane.

(3) No – speak – Spanish. Go – airport … … airplanes – zoom-zoom-zoom.

Quick. You – get – me – airport – me – pay – big bucks. Kapeesh?

Another common context is warfare. Thus, in one of the episodes of the American
television series M*A*S*H, Colonel Potter finds himself saying something like (4)
to one of the Koreans he is trying to communicate with, only to stop himself by
uttering (5).

(4) Me – no – want – watch

(5) My God, I’m beginning to talk pidgin!

In fact, not only have most of us come across Foreigner Talk; many of us have used
it ourselves, under similar circumstances. There is indeed only one common alter-
native, a practice which anthropologists refer to as silent barter. In many parts
of the world, people not knowing each other’s language engage in trade without
any serious attempt at using language, simply by displaying their wares, selecting,
rejecting, and finally trading objects. The whole procedure is conducted in silence.
If the trade is not mutually satisfactory, the whole cycle begins again. While evi-
dently quite effective for simple trading purposes, silent barter is highly limited in
its application. It could conceivably have been used in the situation that gave rise
to utterance (4), but not in the situations addressed by the utterances in (1)–(3).
Even those of us who might have a personal prejudice against the use of For-
eigner Talk are quite familiar with it and its linguistic peculiarities, and we are
able to judge whether a particular utterance is a proper example of Foreigner Talk
or not.
We know that in addition to increase in volume, decrease in speed, and a
chunky, word-by-word delivery, Foreigner Talk exhibits a number of peculiarities
in its lexicon, syntax, and morphology, most of them consisting of attrition and
simplification.
In the lexicon we find most noticeably an attrition in terms of the omission of
function words such as a, the, to, and. There is also a tendency to use onomato-
poetic expressions such as (airplanes –) zoom-zoom-zoom, colloquial expressions
such as big bucks, and words that sound vaguely international such as kapeesh.
 Introduction: Foreigner Talk, “Tarzanian”, and other simplified forms of speech 369

In the morphology we find a tendency to simplify by eliminating morpholog-

ical complexities. As a consequence, where ordinary English distinguishes I vs.
me, Foreigner Talk tends to use only me.
Syntactic simplification consists of the absence of complex syntax (such as
the use of relative and other dependent clauses). The latter feature, combined
with a number of lexical and morphological ones, can best be illustrated by com-
paring (3) above with (3’) below. Even a cursory glance will convince us, whether
we use Foreigner Talk or not, that (3) is a much more appropriate example of
Foreigner Talk than (3’).

(3’) I – do – not – speak – Spanish. Go – to – the – airport. Quickly. If – you –

get – me – to – the – airport – I – will – pay – more – money. Understand?

Part of our familiarity with the structural peculiarities of Foreigner Talk and our
ability to judge its quality may stem from our familiarity with another widespread
form of speech that is structurally similar, namely Baby Talk. This is a form of
speech commonly employed by adults, or even older children, in talking with
babies.
The term Baby Talk reflects the common assumption that “this is the way
babies talk”. But linguists who study the early stages of children’s language acqui-
sition know that this assumption is not founded in fact. Rather, Baby Talk is a
response by adults to a situation remarkably similar to the contexts in which For-
eigner Talk tends to arise – the desire or need to communicate with somebody
whose language we don’t understand and who apparently does not understand
our language either. Scholars working on early child language acquisition there-
fore prefer to use terms such as nursery talk or care-giver talk to refer to this
form of speech. But the term Baby Talk has a certain usefulness, in that it more
accurately reflects what ordinary, linguistically naive adults believe; and as noted
on several earlier occasions, such a belief often is more important for linguistic
change than the more objective accounts of trained linguists.
Given the communicative similarities between the contexts in which Baby
Talk and Foreigner Talk arise, it is not surprising that, like Foreigner Talk, Baby
Talk is characterized by extensive lexical attrition and morphological and syn-
tactic simplification, as in (6). At the same time, the differences between the two
types of situation also have consequences in linguistic structure and in “delivery”.
Baby Talk tends to exhibit a great degree of phonological simplification, as in
seep for sleep; and while Foreigner Talk tends to be characterized by a chunky
and rather loud delivery, Baby Talk features a more “lilting” or “sweet” delivery.
370 Pidgins, creoles, and related forms of language

(6) Baby wan(t) seep?

‘Does the baby/do you want to sleep?’

Many people use Baby Talk when talking with their lovers or even their pets, i. e.,
in other situations in which a lilting, sweet delivery seems appropriate. Dogs,
however, may instead be subjected to a form of simplified speech that is anything
but lilting and sweet. This is the special form of language used in dog obedience
training, especially in expressions like No sniff ‘do not sniff around’ or No scratch
‘don’t scratch yourself’ – or even No lick paw, to discourage a dog from licking her
paws till they get sore.
In addition to these forms of simplified speech we can draw on yet another,
similar form of speech as a model for producing Foreigner Talk. This is the literary
caricature of something like Foreigner Talk that we find exemplified in the famous
Me Tarzan – You Jane cited at the beginning of this chapter. This form of language
is found in numerous other literary contexts, many of them much earlier than the
Tarzan stories. See for instance the passage in (7). Even so, linguists have begun
to use the term Tarzanian to refer to this form of language use.

(7) “Kill-e,” cried Queequeg, twisting his tattoed face into an unearthly
expression of disdain, “ah! him bevy small-e fish-e; Queequeg no kill-e so
small-e fish-e: Queequeg kill-e big whale!” (Herman Melville, Moby Dick)

Interestingly, then, by at least partly drawing on the model of Tarzanian, Foreigner

Talk may be drawing on a caricature of itself – in a kind of circular development.
But while Tarzanian no doubt draws partly on Foreigner Talk and our quasi-intu-
itive understanding of its peculiarities, it also tends to caricature another form of
speech which, in spite of great structural similarities, differs considerably in its
use and social setting from Foreigner Talk. This form of speech is called pidgin
(note Colonel Potter’s use of the term in example (5) above).
The pidgin influence is especially clear in the passage cited in (7). Both the
general setting of Melville’s novel and the specific use of expressions like tattoed
face places the utterance in the prototypical context for pidgins to arise – the
interaction between Europeans and non-Europeans in the period of European
colonial expansion. At the same time, there is clear empirical evidence that, in
spite of their similarities, pidgin and Foreigner Talk are very distinct forms of
speech. These matters are discussed more fully in the next two sections.
 Pidgins defined 371

2 Pidgins defined
In the wake of the European colonialization of much of today’s “Third World”,
there arose all over the globe a series of speech varieties that are commonly
referred to as pidgins. These languages in many ways differ radically from any
of the other types of language resulting from linguistic contact. True, here too we
find evidence of interlanguage influence from the various indigenous languages.
And one of the major characteristics of pidgins, structural simplification, may
be found in other types of language contact, such as koiné formation or conver-
gence. But in other situations, structural simplification is selective and merely
serves to eliminate (excessive) linguistic differences. In pidgins, by contrast, there
is a radical simplification of linguistic structure, plus a radical reduction
or attrition of vocabulary. Thus, all morphological variation, all phonological
alternation, major syntactic phenomena such as the passive, and all syntactic
embedding tend to be eliminated. And the lexicon tends to be limited to 1,000 or
2,000 words. Most significant, by all appearances, simplification and reduction
take place rapidly, within one or two generations.
Before trying to determine the precise linguistic developments and the special
social conditions that give rise to pidgins, let us take a brief look at an often-cited
example from an English-based pidgin spoken in New Guinea and Melanesia.
This is a structure supposedly used to express the notion ‘piano’:

(8) Big pela bokas yu pait-im i krai

Literally: big fellow box you fight-him he cry (= cries)
Actually: big * box you fight * * cry
large (suit)case one hit shout
great wooden … touch … make sounds
… container …
…
= ‘A big type of wooden structure which, if you touch it (or its
keys), emits (musical) sounds’

Even a cursory glance will show that this is a rather lengthy expression for the
notion ‘piano’. The length of the expression is a consequence of the extremely
limited lexicon of pidgins. If you have only 1,000 to 2,000 words, it does not pay
to have special terms for such notions as ‘piano’, or ‘philosophy’ for that matter.
Especially terms for things and ideas like ‘piano’ and ‘philosophy’ that are far
removed from the social context in which pidgins must function are likely to be
expressed by circumlocution, as in (8).
372 Pidgins, creoles, and related forms of language

But even some very basic notions may be expressed by circumlocution. For
instance, the New Guinean/Melanesian pidgin expression for hair is gras (bi)lɔŋ
hed, lit. ‘grass belong head’ ≈ ‘grass-like growth on the head’.
Moreover, the words actually used in circumlocutions of this type exhibit
characteristics attributable to the extremely limited range of the lexicon. As the
“Actually” glosses in (8) for big, bokas, pait, and krai illustrate, to cover enough
semantic territory each word has to have a wide range of polysemy, that is, mul-
tiple meanings for a given word (see Chapter 7, § 2). Thus, big covers ‘big’, ‘large’,
‘great’, and many other related meanings.
At the same time, all of the vocabulary used in (8) is of European origin. This
is the normal pattern in the classical pidgins that arose in the context of Euro-
pean colonialization. Non-European words are rare, except for names of places,
flora, and fauna which tend to come from the indigenous languages. A few other
non-European words may likewise come from the indigenous languages, such as
kanæka ‘native’ in Melanesian/New Guinean pidgin. But they may also stem from
other languages, such as Melanesian/New Guinean pidgin kau-kau ‘food’, which
has been traced to Hawaiian sources.
Further, note the evidence for extensive structural simplification which man-
ifests itself in the absence of inflectional morphology in krai, rather than crie-s;
in the absence of a relative pronoun (or other type of relative marker); and in the
absence of the conditional marker ‘if’. Again, these characteristics are a general
feature of classical pidgins.
Other features include the use of just one universal preposition. New Guinean/
Melanesian pidgin, for instance, employs the all-purpose preposition (bi)lɔŋ to
express the notions covered by ordinary English of, to, for, at, in, with, and all the
other prepositions, as with gras (bi)lɔŋ hed ‘hair’, discussed above.
Similarly, our pidgin uses mi and em not only for the object me and him, but
also for the subject I and he. In the third person, em indicates not only the singular
masculine ‘he’, but also feminine ‘she’ and neuter ‘it’. And so on.
The last statement seems to be contradicted by the use of im and i in example
(8), whose literal translation is ‘him’ and ‘he’. However, here as elsewhere, the
“Literally” gloss at best gives us an indication of the English source, not of the
actual meaning or function of a given word. In the case of im and i, the “Actually”
gloss supplies an asterisk, just as it does for pela, lit. ‘fellow’. This is because in
both cases, English words have been recruited to express grammatical features of
the indigenous languages: im indicates that the preceding verb is transitive and
i signals aspects of discourse continuity. Both of these elements, thus, serve as
something like function words or affixes, calquing grammatical peculiarities of
the indigenous languages.
 Pidgins defined 373

The use of pela generally is motivated by the fact that the indigenous lan-
guages employ “classifiers”. Classifiers are marginally present in English, too,
where they are used to “individualize” mass nouns, as in rice : one grain of rice/
two grains of rice, bread : a loaf of bread, a slice of bread. In the indigenous lan-
guages of New Guinea/Melanesia (and in many other languages around the world)
classifiers are obligatory with all nouns that are preceded by modifiers. The use
of pela, then, is a calque of this important grammatical pattern. At the same time,
like all other structural and lexical pidgin features, it exhibits extreme attrition,
being expressed by a single, all-purpose form, and like most of the vocabulary
it comes from English. (A second use of pela, not illustrated in (8), is to turn the
pronouns mi and yu into plurals, yielding mi-pela ‘we (all)’ (including persons
other than ‘you’ and ‘I’) and yu-pela ‘you all’, beside yu-mi ‘we’ = ‘you and I only’.
Here again, the use of pela, in mi-pela vs. yu-mi, serves to encode an important
grammatical distinction of the indigenous languages.)
Finally, yet another consequence of the extreme lexical and grammatical
reduction has occasionally been commented on, namely the need to use gestures,
facial expressions, and changes in voice quality to make up, as it were, for the
limited linguistic means that speakers can use to convey their ideas. This phe-
nomenon has perhaps been most strikingly described in the following late-nine-
teenth-century account of a pidgin-like form of language, Chinook Jargon, spoken
in the northwestern United States, British Columbia, and Alaska. The account
betrays its date by its use of expressions such as “a party of the natives”. (See § 4
below for more on Chinook Jargon.)

The Indians in general are very sparing of their gesticulations. No languages, probably,
require less assistance from this source than theirs … We frequently had occasion to observe
the sudden change produced when a party of the natives, who had been conversing in their
own tongue, were joined by a foreigner, with whom it was necessary to speak in the Jargon.
The countenances, which had before been grave, stolid, and inexpressive, were instantly
lighted up with animation; the low, monotonous tone became lively and modulated; every
feature was active; the head, the arms, and the whole body were in motion, and every look
and gesture became instinct with meaning. (Hale 1890)
374 Pidgins, creoles, and related forms of language

3 Pidgin origins
The question of how pidgins originated, and why so many of them arose in the
wake of European colonialist expansion, has elicited many different responses.

3.1. “Imperfect learning”. One explanation, coming in different variants, boils

down to the assertion that pidgins resulted from the inability of the non-Europe-
ans to learn European languages correctly, i. e., pidgins result from “imperfect
learning”.
However, one must ask why pidgins arose only among non-Europeans. And
why should the non-Europeans have exhibited this peculiar inability only in
learning European languages, while having no difficulties with learning other,
non-European languages? Note in this regard that the indigenous populations of
most non-European areas of the world have a long tradition of bilingualism (and
multilingualism). That bilingualism, however, did not typically result in pidgins
or pidgin-like languages, but in convergence, koiné-formation, and similar devel-
opments. (See Chapters 12 and 13.) Moreover, in more recent times large numbers
of non-Europeans have learned European languages very well – evidently because
they were both permitted and encouraged to do so.
The idea that pidgins and creoles result from the incomplete acquisition of
European target languages has recently been revived, but without racist over-
tones. Scholars arguing for this position claim that many of the “nonstandard”
features of pidgins and creoles can be accounted for by assuming that the target
of acquisition was not the standard European language, but rather vernacular
varieties.
The latter claim is, of course, quite sensible, given the social circumstances
in which pidgins and creoles were formed (see also § 5 below). However, any
assumption that pidgins arose from imperfect learning runs into a serious diffi-
culty. This difficulty is connected with the very process of structural simplifica-
tion. If we only had English-based pidgins, it would be easy to attribute the lack
of inflection in forms such as the krai of (8) above to the fact that the non-Euro-
pean learners, not understanding the function of inflectional endings such as -s,
-ing, -ed, ignored them altogether. The result is the simple root krai, whose choice
would be further supported by the fact that most forms of the present tense, as
well as the imperative and infinitive, consist of the simple root, without inflec-
tional ending. But in Romance-based pidgins the situation is much more complex.
Consider for instance the following (incomplete) Portuguese paradigm of the verb
‘to speak’.
 Pidgin origins 375

(9) Present singular plural

1 falo falamos
2 falas falais1
3 fala falam
Imperative 2 fala falai1
Infinitive falar
Gerund falando
Past Participle falado
(1The second plural form has become obsolescent in present-day Portuguese.)

How should this complex set of forms be simplified, reduced to a single, invariant
form? Should we use the simplest form, without any inflectional ending, i. e., the
second person singular imperative fala? But this form is just one of many. And
in fact, it is not the form that was used as the invariant, uninflected form of the
verb. Perhaps the third person singular should have been used, since this is the
form that most frequently occurs in speech? Or the first person singular, since we
like to talk about ourselves? Again, these forms were not used. Or should we just
randomly choose the first verbal form we come across? That would mean that the
invariant pidgin verb forms should randomly reflect any verbal form of Portu-
guese. And again, that is not what we find.
Instead, the normal invariant pidgin form of the verb is that of the infinitive,
falar in our case.
Since the late nineteenth century, many linguists have argued that this con-
sistent choice is unexplainable under the assumption that pidgins arose from
imperfect learning. To be able to consistently choose one grammatical structure,
the infinitive, as the invariant shape of the verb requires a degree of grammatical
knowledge of the European language which is far from imperfect. Moreover, one
must ask, if the non-European learners had such a high degree of grammatical
knowledge of the European language that they could recognize the infinitive, why
did they not use the full range of grammatical forms of that language – or at least
a much larger range than the mere infinitive?

3.2. The “racial-inferiority” argument. The claim that pidgins result from the
imperfect learning of European languages by non-Europeans has often – espe-
cially in colonialist times – been supported by the racist allegation that the
non-European “natives” are genetically inferior to the European colonizers and
that this is the reason they are unable to learn the European languages. This view
has often been supported by asserting that the highly reduced structure and
vocabulary of pidgins are prima facie evidence for the intellectual inferiority of
376 Pidgins, creoles, and related forms of language

the “natives”. The similarity between pidgins and Baby Talk “confirms” that the
“natives” only have the mental capacity of infants.
The strongest linguistic arguments against this racist view can be based on
the following facts. Many of the native languages of these “natives” can rival any
European language in structural complexity. Moreover, as just noted, the “natives”
had no difficulties with being bilingual in non-European languages. And, when
given the opportunity and encouragement to do so, they were perfectly able to
learn the European languages. Beyond this linguistic evidence, it is sufficient to
note that there is simply no credible evidence to support the view that non-Euro-
peans are mentally inferior to Europeans.
Nevertheless, several elements in the traditional racist arguments are signif-
icant, since they provide us with important information on the social attitudes of
the European colonizers and since, as we see shortly below, the mere belief that
the “natives” are mentally inferior may have indirectly contributed to the institu-
tionalization of pidgins.

3.3. The Portuguese Proto-Pidgin hypothesis. It has also been alleged that all –
or at least most – pidgins are descended from a single source, a “Proto-Pidgin”.
This claim, if correct, would be highly attractive, since it would automatically
explain the similarities in structural and lexical attrition found in all of the pidgins.
Some scholars claim that the Proto-Pidgin consisted in the original Lingua
Franca or Sabir, a Romance-based contact language employed in the Mediterra-
nean area from the time of the Crusades into the 19th century, and sharing with
the pidgins of the colonialist period a high degree of lexical and structural reduc-
tion. According to these scholars, Sabir was taken as a contact language from
the Mediterranean to the world at large by the Portuguese, who were the first to
engage in the explorations that ultimately led to the domination and exploitation
of most of the non-white world by a tiny minority of European, white nations.
Other scholars, instead, believe that the Portuguese developed their own Pro-
to-Pidgin contact language, without influence from Sabir.
As other European nations, with different languages, entered the colonialist
scene, it is claimed, they took over the ready-made pidgin of the Portuguese –
together with their nautical and other relevant non-linguistic expertise. But
instead of taking it over intact, these nations adjusted the Portuguese pidgin to
their languages by substituting words from their own lexica for the Portuguese
lexical items. This process, which has been called relexification, would a priori
seem to be relatively easy, given that we are talking about a very limited vocabu-
lary of at most 2,000 words.
The relexification hypothesis receives apparent empirical support from
the fact that many, if not most, of the non-Portuguese European-based pidgins
 Pidgin origins 377

have certain lexical items which are most likely to be of Portuguese origin. Most
notable among these are the words in (10) below. Significantly, the earlier English
form sabby, clearly Romance in origin, agrees best with Port. saber [-b-] ‘know’,
whereas Fr. savoir and Sp. saber have [v] and [β] respectively. Similarly, pickaninny
and its relatives are most easily derived from Port. pequeno, diminutive pequen-
ino ‘small’, while Spanish has pequeño with palatal nasal. More than that, a few
pidgins, such as Saramaccan (spoken in Surinam), seem to have stopped relex-
ifying in midstream. About 27 percent of Saramaccan vocabulary is traceable to
Portuguese; the rest is mainly of English origin.

(10) Engl. pidgin savvy

Earlier Engl. pidgin sabby
Fr. pidgin sabé
Engl. pidgin pickaninny
Du. pidgin pikien
Span. pidgin piquinini

This theory likewise is open to several doubts and reservations. First, even if it
were established beyond a reasonable doubt that all pidgins are descended from
a single Proto-Pidgin, we must still explain how that Proto-Pidgin came about. Re-
lexification does not solve the problem; it merely pushes it back farther in history.
Second, some pidgins or pidgin-like languages (see § 4 below) clearly arose
independently, in areas and social situations without any possible access to
Sabir, the hypothetical Portuguese Proto-Pidgin, or any of the pidgins supposedly
descended from it. Even some of the alleged descendants of the Portuguese Pro-
to-Pidgin exhibit features suggesting that they arose independently. Thus, early
reports show that in the French-based pidgins of the West Indies the issue as to
which structure should be used as the uninflected, general verb form was not
yet fully resolved. Both the infinitive (as in savoir ‘know’) and, less commonly, a
regularized form of the past participle (savé ‘known’) still were in competition.
Had there been simple relexification of an already established Proto-Pidgin, we
would not expect such fluctuations. It is only later, as the pidgins come to be more
established, that the infinitive form is used across the board, just as it is in any
other pidgins based on languages with relevant verb morphology. (English-based
pidgins, of course, are not helpful in this regard, since their invariable verb form
is identical not just to the English infinitive, but to the verbal root.)
More than that, the heterogeneous vocabulary of Saramaccan and a few
other similar cases can be explained by a different scenario, which has the added
advantage that it is much less hypothetical than the relexification hypothesis:
Let us assume that Saramaccan started as a Portuguese-based pidgin, a likely
378 Pidgins, creoles, and related forms of language

assumption since we know that the Portuguese had a significant South American
presence (which still survives in Brazil) and since, moreover, the Portuguese were
heavily involved in the slave trade to South America. Under the circumstances,
we would expect a Portuguese-based pidgin to have arisen. Now, while in many
colonial situations the speakers of the source language for the pidgin remained in
power, Surinam, formerly known as Dutch Guyana, experienced a rather check-
ered colonial history. At an early period the colony came under British control,
which continued in British Guyana but gave way to Dutch control in Surinam
by the late seventeenth century. It is the early change from Portuguese to British
influence which can be held responsible for the lexically mixed character of Sara­
maccan. As we see in § 5, pidgins – if they survive for an extended period – can
undergo a process of creolization or depidginization which most prominently
manifests itself in vocabulary expansion. For the majority of pidgins, the source
for the original pidgin lexicon and the source for the expanded creole lexicon were
the same European language. But in Saramaccan, the situation must have been
different. While Portuguese furnished the source for the pidgin lexicon, lexical
expansion must have taken place largely during the period of British control and
therefore would draw on English vocabulary. (Dutch vocabulary in Saramaccan
reflects the later Dutch control of the colony.)
In the majority of pidgins the words of Portuguese origin are much more
limited. The most widespread are the ones in (10) above. The wide diffusion of
this limited lexical set, however, can be explained without the assumption of a
Portuguese Proto-Pidgin. The fact is undeniable that there was a great amount
of contact between Portuguese navigators, sailors, and (slave) traders and their
counterparts from other European nations as they entered the colonialist and
slave-trading “enterprise”. In the process, a fair amount of vocabulary connected
with the enterprise must have been passed on from the Portuguese to the other
Europeans, as part of a special conquistador/slave trader jargon. (A possible par-
allel is the North Atlantic nautical jargon of Chapter 10, § 4.)
Probable traces of this jargon, which are not limited to pidgins (and creoles,
discussed in § 5 below), can be found in a fair amount of the terminology of the
slave trade, including Engl. Negro, mulatto, quadroon, and their counterparts in
other European languages, as well as the term creole, whose original meaning is
said to have been ‘child of a non-European mother and a European father, born in
the house of the father’. There is nothing to prevent us from assuming that savvy
and pickaninny were likewise diffused from Portuguese through the medium of
the conquistador/slave trader jargon, rather than through relexification from a
Portuguese Proto-Pidgin.
That words of this sort, together with their social connotations, can be picked
up by people who do not speak a pidgin – or are in the process of relexifying a
 Pidgin origins 379

pidgin – is shown by the fact that savvy and pickaninny, as well as Negro, mulatto,
quadroon, and creole have entered the general vocabulary of English and are used
by people who have no firsthand acquaintance with pidgins. Similarly, the word
kapeesh of example (3) above was no doubt first picked up by American soldiers
in World War II who fought in Italy and learned a smattering of Italian, including
capisci ‘do you understand?’, regionally pronounced more like [kapiš]. Having
done so, they transferred the word to similar contexts, i. e., when talking with
speakers of other foreign languages that they did not understand. Subsequently,
many Americans who have never been to Italy have adopted the word and use it
in similar situations, without of course knowing its origin. Note similarly the word
kau-kau ‘food’ in Melanesian/New Guinea pidgin, a word which has been traced
to Hawaiian origin and which, significantly, is believed to have come to the area
through South Sea sailors’ jargon.

3.4. Foreigner Talk and the origin of pidgins: So far, the most plausible hypoth-
esis is that (beside interlanguage), the most potent force in the development of
pidgins is foreigner Talk, the form of speech discussed in detail in § 1 above.
As we have seen, there are great formal similarities between Foreigner Talk
and pidgins. Both exhibit a great amount of structural and lexical reduction.
Moreover, there is good reason to believe that pidgins originated in contexts very
similar to those that give rise to Foreigner Talk: Speakers find themselves in a
situation – in this case the context of colonial expansion and the slave trade –
where they are forced to communicate with others whose language they do not
understand and who do not understand their language.
Given these great similarities between Foreigner Talk and pidgins, it is tempt-
ing to view pidgins simply as institutionalized forms of Foreigner Talk. However,
the link cannot be quite so direct. We know that Foreigner Talk is a very common
tendency in the context of first linguistic contact. We must therefore ask ourselves
why it is not institutionalized more commonly.
A plausible answer to this question can be given if we consider sociolinguistic
factors. Under normal circumstances, the expectation is that a foreign language
(or even several languages in contact) will be learned to the point of complete or
at least adequate mastery. Foreigner Talk therefore generally is only a transitory,
first-generation or first-contact phase.
The expansion of European colonialism brought with it a very different expec-
tation on the part of (most of) the Europeans: The “natives” were held to be infe-
rior and thus proper objects of colonialist and racist exploitation, even of slavery.
As we have seen in § 3.2 above, they were also commonly believed to be incapable
of correctly learning the European languages. If, then, they began to imitate the
Foreigner Talk of the Europeans, the similarity of their production to Baby Talk
380 Pidgins, creoles, and related forms of language

only strengthened the colonialists’ and slave-traders’ mistaken belief that these
“natives” had the mentality of infants and that Foreigner Talk therefore was the
only proper way of speaking to them. Under the circumstances, the use of For-
eigner Talk did not just remain a transitory phenomenon, but became institution-
alized as the proper vehicle for communication with the “natives”.
The extent to which attitudes of this type permeated society can be gauged
from the fact that even in the early part of the twentieth century, the Encyclopedia
Britannica characterized Pidgin English as an “unruly bastard jargon, filled with
nursery imbecilities, vulgarities, and corruptions.” Except for the function words,
virtually every lexical item in this passage expresses prejudice. Note further the
use of the word nursery, a clear echo of the belief that Pidgin = Baby Talk.
Another important factor may have been that by providing Foreigner Talk
as the only model which the “natives” could imitate, the Europeans were able to
keep them “in their place”. Foreigner Talk, then, became a marker of the social
distance between European masters – who spoke the “real, proper” form of Euro-
pean language – and their non-European subjects or slaves – who were confined
to an “inferior, bastardized” form of the language. Being relegated to Foreigner
Talk excluded the subject peoples from the European language of power, while
the reduced structure of the European-based Foreigner-Talk made it both easy
to learn and perfectly adequate for the limited communication – mainly giving
orders – that the Europeans wanted to engage in.
Recent research suggests that Portuguese pidgins arose from a deliberate deci-
sion by the Portuguese to use Foreigner Talk, rather than their normal language.
In the early phase of Portuguese expansion down the western coast of Africa,
attempts were made to communicate with the local population through interpret-
ers familiar with Portuguese and with Arabic, which at that time served as the
major link language in all of northern Africa. As the Portuguese moved farther
south, this approach no longer was feasible. At first, the Portuguese tried to teach
their language to members of the local community who would then serve as inter-
preters. After a while, realizing that this was a very time-consuming process, they
switched to teaching a simplified, Foreigner Talk variety of Portuguese. It is this
variety which seems to have been the basis for the Portuguese pidgins.
In addition to having sociolinguistic plausibility on its side, as well as the
general similarities between Foreigner Talk and pidgins, the Foreigner Talk hypoth-
esis has the advantage of explaining an important linguistic feature of pidgins
which other theories find very difficult to explain. This is the fact, noted in § 3.1,
that the invariable verb form of Romance-based pidgins generally is the infinitive
of the European source language, rather than another specific form – or, randomly,
any form – of the verbal paradigm. As it turns out, the infinitive also is the normal
invariable verb form in the Foreigner Talk varieties of the Romance languages.
 Trade Jargons and other pidgin-like languages 381

The Foreigner Talk hypothesis further explains why many pidgins have picked
up highly colloquial, even vulgar, expressions from the European languages, such
as Melanesian Pidgin Engl. bagerap ‘destroy, ruin …’ from vulgar Engl. bugger up.
As observed in § 1 above, such expressions are quite common in Foreigner Talk.
While the Foreigner Talk hypothesis thus is the most fruitful account of pidgin
origins, there is clear evidence that pidgins, once established, differ markedly
from Foreigner Talk. After the First World War, the former German colony of New
Guinea was placed under Australian trusteeship. Australian officials who took
over the administration believed that they could talk to the “natives” simply by
using their own version of Foreigner Talk, with a few elements (im, i, and pela)
thrown in randomly to capture the most striking features of New Guinea Pidgin.
But it is reported that the “natives” were not impressed. At least when amongst
each other, they laughed derisively at what to them was an incompetent imitation
of their pidgin.

4 Trade Jargons and other pidgin-like languages

The importance of sociolinguistic factors can also be seen in a more recent and
therefore more observable development. In Germany (and similarly in a number
of other European countries), the industrial boom of the 1960s led to a heavy
influx of foreign laborers or “guest workers” who were brought in from economi-
cally less advantaged countries to do work which Germans considered undesira-
ble. Most guest workers had no prior knowledge of German, and their languages
were equally unfamiliar to most of their German supervisors. Out of the Foreigner
Talk employed by the socially more powerful Germans there developed a special
variety of language called Gastarbeiterdeutsch (‘guest worker German’, abbrevi-
ated GAD). GAD has undergone a certain degree of institutionalization and may
for instance be used by guest workers of different linguistic backgrounds when
talking to each other. Moreover, although in the early stages, native Germans
apparently used Foreigner Talk quite freely with guest workers, recent studies
suggest that Foreigner Talk now is used rather sparingly. That is, foreign laborers
no longer seem to need the input of native speakers’ Foreigner Talk for acquiring
GAD.
At the same time, the German sociolinguistic conditions are not conducive
for the development of a fully institutionalized pidgin. In spite of considerable
difficulties with a bureaucracy that tries to prevent their acquiring citizenship and
in spite of widespread social segregation and recurring episodes of xenophobic
excesses, most foreign laborers and their families have tried to stay in Germany
382 Pidgins, creoles, and related forms of language

and have therefore made considerable efforts to learn German more fully. Most
native speakers of German, on their part, tend to switch to normal, non-simplified
German as soon as they feel that a particular foreign laborer has begun to acquire
more than cursory control of the language. It is only among those guest workers
who develop a very negative attitude to Germany and to German society (because
of cultural disillusionment or because they have been victims of xenophobia) that
GAD becomes relatively “fixed”. But workers with this attitude tend to return to
their home countries. Their version of GAD therefore has no chance of becoming
institutionalized.
The relatively unsettled sociolinguistic nature of GAD is mirrored by relatively
unsettled linguistic characteristics. For instance, instead of generalizing a single
morphological form as the all-purpose, uninflected form of the verb, GAD has at
least three different formations: an uninflected form of the verb, formally similar
or identical to the imperative, a form identical to the infinitive, and the past parti-
ciple; see (11). Interestingly, each of these can be used in ordinary German to give
orders; see (11’). It has been observed that the generalization of formations which
can be used as imperatives is a common feature of pidgins. Thus the infinitive,
commonly used in Romance-based pidgins, can also be used as an imperative
in the Romance languages. Presumably, this choice reflects the social context
in which the more privileged or powerful give orders to the less privileged. In
German, the use of infinitives and past participles is especially common in more
impersonal commands, as for instance in the military. These forms are therefore
especially appropriate for giving orders to people with whom one does not want
to be in close, personal contact. The choice of the imperative may, however, also
be motivated by the fact that it is virtually identical to the root and thus morpho-
logically the simplest verb form.

(11) GAD Ordinary German

a. mach arbeit ich mache/tue die Arbeit ‘I do the work’
wir machen /… die Arbeit ‘we do the work’
ich habe die Arbeit getan ‘I did the work’
b. nix nach haus gehn ich gehe nicht nach hause ‘I don’t go home’
c. tag geschlafen ich schlafe am Tag /… ‘I sleep during the day’
(said by a guest worker who was working night shifts)

(11’) a. iss deine Suppe! (sing. 2 imperative) ‘eat your soup’

b. Essen fassen! (infinitive; distancing) ‘get (your) chow’
c. stillgestanden! (past pple.; military) ‘stand still; (stand at) attention’
 Trade Jargons and other pidgin-like languages 383

While pidgins and pidgin-like languages such as GAD arise in an environment

of rather extreme inequality, similar types of language can come about under
conditions of equality. Interestingly, the different social conditions are reflected
in the fact that such Trade Jargons also differ from pidgins in their linguistic
characteristics.
One such language may be the so-called Chinook Jargon referred to in § 2
above, which used to be employed by trappers and traders in the northwest of
the United States, British Columbia, and Alaska. Extant descriptions show that
when the Europeans arrived, the jargon was used for relatively short-term trading
relations in a social setting of equality. Here, too, we find a great degree of struc-
tural and vocabulary simplification, exceeding what one encounters in normal
language contact. However, there is an important difference: The vocabulary and
the linguistic structures employed, as well as their simplification, are much more
variable than those of the classical pidgins. Thus, the vocabulary did not predom-
inantly come from a single source; but a variety of languages contributed to it.
Moreover, the extent to which different groups contributed to the vocabulary was
subject to considerable fluctuation. (Some linguists believe that Chinook Jargon
began as a pidgin, because there are some indications that it was used with slaves;
but for all we know, slavery in the Native American context of the northwest was
very different in nature from the chattel slavery of European colonialism and its
successors in countries like the United States.)
A similar language, called Russenorsk, arose in the north of Norway, as a
means of communication between Russian traders and Norwegian fishermen
during the relatively short fishing season. Just like Chinook Jargon, the language
was employed for relatively short-term trading relations between social equals.
Here, too, we find considerable reduction in structure and vocabulary. And again,
the vocabulary is heterogeneous, with Russian and Norwegian elements mixed in
roughly equal proportions, in addition to a few lexical items which seem to have
come from the nautical jargon of the northern Atlantic. Moreover, some areas of
grammatical structure exhibit the effect of something like selective simplifi-
cation. For instance, the all-purpose preposition po is phonetically more or less
identical to the Norwegian preposition på and the Russian po (although the mean-
ings of these two prepositions are not identical).
Trade jargons like Russenorsk and Chinook Jargon give the impression of
being semi-institutionalized forms of Foreigner Talk, without the homogeneity of
fully institutionalized classical pidgins. However, in spite of their linguistic and
social differences, they share one feature with pidgins: In both types of contact,
there is no expectation of full acquisition of the other’s language. What differs
is the specific reason for not expecting full acquisition. In the case of pidgins,
that reason is the colonialists’ notion of extreme social inequality. In the case
384 Pidgins, creoles, and related forms of language

of trade jargons, by contrast, the reason is that two or more groups engage in
contact which (by design or necessity) is restricted to just a few activities. What is
interesting in this regard is that when the Russian merchants decided to engage
in less limited trading relations with Norway, they sent their sons to Oslo (or, as it
was called then, Christiania) to learn “proper” Norwegian; and conversely, Nor-
wegians went to the city of Arkhangelsk to learn Russian.

5 Creoles

Many pidgin-like forms of language may have developed in the extended history
of human language, only to disappear later – usually without any distinct
trace. But under certain conditions they came to be employed in a manner that
ensured them a more lasting place in history, as link languages or even as native
languages.
Given their severe limitations in grammar and especially in vocabulary,
pidgins and similar varieties of language may be very useful, even appropriate
for the very restricted social conditions in which they arose. However, the limi-
tations are considerable obstacles when languages of this type are to be used in
a broader range of social and linguistic contexts. At a minimum, an expansion
of context requires a vastly expanded vocabulary which more unambiguously
accommodates the large range of meanings ordinarily expressed through lan-
guage. A certain expansion of grammar is no doubt required as well.
This process of expansion, called creolization (or depidginization), is
commonly believed to take place only when a pidgin “acquires native speakers”.
According to this view, the starting point is a linguistically highly diversified com-
munity in which parents begin using pidgin with each other as their only common
means of communication. The pidgin therefore becomes the sole basis for a new
generation of speakers to acquire as a native language. And, it is argued, while
the pidgin may have been sufficient as an auxiliary language for the parents, it is
clearly inadequate as a native language and therefore must undergo expansion
and elaboration. Linguists subscribing to this view will reserve the term creole
for languages which arose in this manner.
The American linguist Derek Bickerton, in fact, has based an elaborate theory
of creolization on this view. According to him, the need to create a native language
makes it necessary for children to draw on a “bioprogram”, part of the innate
endowment of human beings, which determines the structure of the creole. At
the same time, it explains idiosyncratic features that supposedly are shared by
 Creoles 385

all creoles and cannot be attributed to the influence either of the European or of
the non-European languages. One of these is “double negation” and “negative
spread” as in (12).

(12) a. English-based creoles:

Cameroon Pidg. i neva giv no nɔting fɔ papa
‘He never gave anything to the old man.’
b. Portuguese-based creoles:
i. São Tomé: i’nẽ na ka ’tlaba na’i fa ‘They do not
negative negative work here.’
Angolares: ɛ’nɛ na ka ’taba ngɛ wa ‘They do not
negative negative work here.’
ii. Papiamentu: Mi no tin nada ‘I have nothing.’
no(t) nothing

There are several reasons why Bickerton’s theory is highly controversial. Most
important, the supposedly idiosyncratic features of creoles are not as unusual as
Bickerton claims, and many, perhaps all, can be explained as reflecting influence
from relevant European or non-European languages. For instance, while the neg-
ative spread in (12a) does not look like Standard English, it is perfectly natural
in vernacular English, as in (12’a). And let us not forget that the majority of early
slave traders and colonialists were not highly educated and were therefore more
likely to speak the vernacular than the standard. In the Romance languages, neg-
ative spread is found even in the standard languages, see (12’b.ii). And one study
of early Portuguese Foreigner Talk provides examples of double negation in that
form of speech; see (12’b.i).

(12’) a. Vernacular English: He never gave nothin’ to the old man

b. Portuguese:
i. Foreigner Talk: nunca a mi cadella nam
never be(?) me dog not
‘Never am I a dog.’
ii. Standard: não dou nada nunca a ninguém
not nothing never nobody
‘I do not ever give anything to anybody.’

While Bickerton’s theory is considered dubious by most pidgin and creole special-
ists, the general belief that creoles arise when pidgins acquire native speakers has
remained remarkably unshaken.
386 Pidgins, creoles, and related forms of language

Perhaps it is true that in some cases slave owners, fearful of African slave
revolts (especially after the successful revolution in Haiti), may have attempted
to prohibit the use of African languages and to force slaves to resort to pidgin.
However, if such attempts were made, they were not very successful. For instance,
reports that early North American fugitive slave patrols frequently had Wolof
interpreters suggest that instead of pidgin, Wolof and perhaps other African lan-
guages were used as link languages among the slaves. (Recall the presence of
Wolof-based words in African American English Vernacular; Chapter 9, § 4.4.)
More important, it must be seriously doubted whether a language as restricted
as a pidgin would have been picked up as a native language by large groups of
children, or whether it would have been used as the only means of communica-
tion in the parental generation. In order for that to happen, the pidgin would have
to have undergone considerable prior expansion and elaboration. Note further
that studies of plantation populations have shown that there were not always
large numbers of children who would have been in a position to acquire and
expand the pidgin.
In fact, creolization or depidginization can take place without a pidgin’s
acquiring native speakers. This suggests that creolization ordinarily is a slow,
continuous process of depidginization, rather than an overnight, “catastrophic”
phenomenon.
Especially illustrative is the case of the varieties of Pidgin English used in
Papua New Guinea and the Solomon Islands. (These are now commonly referred
to as Tok Pisin or Neo-Melanesian, and Neo-Solomonic, respectively.) These
languages came to be employed as administrative auxiliary languages by the
European colonial administrations in communicating with a linguistically highly
diversified indigenous population, as increasingly popular link languages
between the various local communities, and as vehicles for missionary activities.
Each of these expanded uses brought with it an elaboration in vocabulary and
structure so as to enable the language to be employed in its new social contexts.
Tok Pisin has now become a language of parliamentary debates and of the news
media, requiring yet further expansion and elaboration.
Acquisition of native speakers, on the other hand, has proceeded at a much
slower pace. Even recently, only about five percent of all Tok Pisin users were
native speakers. Moreover, while native speakers were reported to use a more
“advanced” form of language in their early years, during their teens they were
said to adjust to the norm of the majority population of non-native (but fluent)
speakers.
Although the exact earlier history of other creoles is to a large extent shrouded
in mystery, circumstantial evidence suggests that similar developments took
place here, too. Thus, our early information about Caribbean pidgins comes from
 Creoles 387

missionaries’ reports or, even more significant, from grammars and translations
of the Bible and the catechism, which they produced for the purpose of convert-
ing slaves to Christianity. Clearly, such activities required considerable expansion
of the pidgin, especially of the lexicon – a vocabulary of 1,000 to 2,000 words
would hardly have sufficed to translate the Bible. The form of language that we
can discern from these sources, therefore, is no longer the simple, highly reduced
pidgin, although it may not be the full creole either.
More than that, though the colonialist establishment strongly disapproved of
the practice, we have numerous reports of Europeans having “gone native”, living
with, or even marrying, non-European women, begetting children and accepting
the children as their own legal offspring, or even altogether adopting non-Eu-
ropean ways. From all we can tell, these practices were much more widespread
than the extant – generally highly disapproving – reports let on. In fact, it is in
this context that the word creole is believed to have arisen, to refer to the children
of European/non-European matches. (The source word, Port. crioulo, is said to be
derived from criar ‘to create, beget’.)
This “domestic” context, too, can be expected to have encouraged an expan-
sion in vocabulary and structure, to make the language usable for the more
expanded communicative demands within the family or household. As in Tok
Pisin, it is possible that some children acquired the resulting form of speech as
their native language; but there is nothing to guarantee that their form of speech
immediately became dominant.
The clear evidence of Tok Pisin and the more circumstantial evidence of other
pidgins/creoles, then, suggest that the distinction between pidgins and creoles is
gradient, rather than absolute. The distinction pidgin vs. creole may be useful for
linguistic classification, but just like distinctions such as Old English vs. Middle
English, it seems to be an idealization. And just as in reality, speakers of Old
English did not wake up one fine morning finding themselves speaking Middle
English, so pidgin-speaking societies probably did not switch to creole in a short,
cataclysmic upheaval.
What is more important is that, once the process of depidginization has run
its full course and the language thereby has acquired the lexicon and grammar
necessary for full communication, the resulting creoles will be functionally indis-
tinguishable from any other form of “full” language. It is only their history which
makes them different.
In the majority of cases, the resulting language is a vernacular which is used
only for ordinary everyday communication, while another language (usually a
European standard language) serves as a means of more intellectual and written
communication. This result, then, is something very similar to diglossia (see
Chapter 10, § 6). In fact, the Haitian relationship between the speech of the edu-
388 Pidgins, creoles, and related forms of language

cated elite (modeled on Parisian French) and the French-based creole of the
majority population has been cited as a paradigm case of diglossia.
However, creoles are not “condemned” to forever remain vernaculars. The
case of Tok Pisin shows that creoles are just as much usable as intellectual and
written languages as any other form of speech, if there is the need.

6 D
 ecreolization and African American Vernacular
English
Where creoles are used as a vernacular, their relationship to the coexisting Euro-
pean prestige language may be of two types. On one side is the diglossic rela-
tionship between Haitian Creole and French. On the other side, where society
is less rigidly stratified, as in the post-slavery English-speaking Caribbean, the
result may be quite different. In this environment, an ever-increasing section of
the population has found it possible, convenient, or necessary to become actively
bilingual in the creole and the European standard language. Through interlan-
guage, then, varieties of language have arisen which are intermediate between the
European standard and the creole. Note the similar development of intermediate
varieties in the Modern Greek diglossic relationship between Katharevousa and
Dimotiki (Chapter 10, § 6) and the Norwegian competition between Nynorsk and
Bokmaal (Chapter 10, § 5).
In the Caribbean, the process probably was helped by the fact that both creole
and (more or less) Standard English speakers consider the creole a dialect, i. e., a
vernacular variety of English. Creole speakers trying to approximate the standard
therefore do not see this as learning a different language; and standard speakers
expect such approximation, with the justification that “They really should know
their own Language!”
Interestingly, here again, speakers behave in accordance with their own social
attitudes and prejudices and not according to the linguists’ view. Linguists would
argue that the creole is really a separate language, not just a dialect of English,
because of its special historical origins and its formidable structural differences
compared to the standard. But such distinctions evidently are of no great signifi-
cance to most ordinary speakers.
In fact, the common assumption among professional linguists that the lin-
guistic approximation necessarily involves a full-fledged creole and a European
standard language is open to some question. Nothing prevents speakers from
beginning to approximate the European standard language even at the pidgin
stage, if the need should arise – or to approximate the European vernacular, for
 Decreolization and African American Vernacular English 389

that matter! As a matter of fact, the very process of creolization or depidginization

can be considered as simply a first step in this process of linguistic approximation.
Here as elsewhere linguists need to constantly reexamine their views, so as not to
fall victims to their own idealizations.
Whatever the truth may be in a particular situation, the ultimate result of the
creole’s (or pidgin’s) approximation to the European prestige language is note-
worthy enough to have received a special name, namely creole continuum. The
continuum is characterized by a range of speech varieties which extend from the
more or less pure European language to the more or less pure creole (or pidgin),
with all kinds of intermediate varieties which, depending on education, motiva-
tion, or attitude, exhibit a varying admixture of standard and creole elements. The
different layers within this creole continuum then function much like the social
dialects of monolingual societies, such that different speakers will be proficient in
a number of different varieties (but not necessarily in all). That is, for all practical
purposes, the two languages have merged into a single entity.
The ultimate outcome is complete decreolization. The grammar of the orig-
inal creole becomes more or less fully integrated into the grammar of the Euro-
pean language, the pure creole disappears, and its vestiges acquire characteristics
which even linguists would term social dialects of the European language.
A probable example is found in African American Vernacular English (AAVE)
or, as it became known in important research of the 1970s and 1980s, Black Ver-
nacular English. In its more conservative or archaic variants, AAVE still preserves
traces of the original pidgin or creole grammar, such as the “completive” marker
done in (13a) or the lack of gender distinction in (13b). But these are definitely on
the wane.

(13) a. He done tole me

b. He a nice girl

Other creole features have proved much more vigorous, such as the absence of the
past tense marker -(e)d in tole of (13a) or the absence of the verb ‘to be’ in (13b).
However, they have done so in a curious fashion.
In the case of the past tense, there is evidence that AAVE now generally has
acquired the ending -(e)d: Forms like lied, teed (off) have practically invariant
final [-d]. Where the addition of -(e)d results in a final consonant group, the ending
is variably absent or present, as in clean(ed), walk(ed). Its absence is especially
common in forms like tole, where the vowel change in the verb root is sufficient
to mark the form as the past tense of tell, even without any affix. What seems to
have happened here is that the original absence of the past-tense marker has been
“salvaged”, by having been reinterpreted as the result of word-final simplification
390 Pidgins, creoles, and related forms of language

of consonant groups, a phenomenon also widely present in the fast speech of

white English vernacular speakers.
Similarly, the absence of the “copula” verb ‘to be’ in structures like (13b) has
been integrated into the system of non-creole English grammar: Unlike prototyp-
ical creoles, AAVE does in fact have a copula. But that copula may be “deleted”,
and deletion takes place under the same conditions under which standard English
“contracts”; see (14a). Where standard English cannot contract, AAVE does not
delete (14b). Put differently, AAVE differs from standard English by having dele-
tion where the latter has contraction.

(14) Standard English Black Vernacular English

a. She’s a nice girl She a nice girl
b. She ís1 She ís1
She isn’t nice, is she? She ain’t nice, is she?
The girl I saw yesterday is nice The girl I saw yesterday is nice
(1 response to the claim that she is not)

Through the integration of its creole features into the grammar of “ordinary”
English AAVE has become a decreolized dialect of English. But note that much
of the decreolization took place in the American South, based on Vernacular
(Southern) White English, and not on the standard language. This factor proba-
bly accounts for the fact that AAVE has been rather slow to adopt the third-per-
son singular present ending -s. The absence of this ending (or its generalization
throughout the present, as in we goes) appears to be an old feature of nonstandard
white Southern speech, carried over from regional dialects on the British Isles,
especially the so-called Midland dialects. In this case, then, the structure of the
European-based speech that was available as a model for decreolization rein-
forced the pidgin/creole feature of not having inflectional endings.
The fact that some of the features of AAVE can be traced to European sources
has given rise to theories that AAVE can be exhaustively explained as a regional
dialect, just like any other dialect of American English. However, features like
the variable presence of the copula, peculiarities of the past tense formation, and
relics like the ones in (13) persuasively argue that it did start out as a creole. Further
evidence comes from African American speech in the so-called Tidewater Area,
islands off the southern East Coast of the United States. This variety of English is
considerably more creole-like than AAVE. Its conservatism is explained by the fact
that when the northern troops retreated after the liberation of the slaves during
the Civil War, the former white landholders of this area did not return: unlike the
rest of the South, the islands no longer provided an economically viable oppor-
tunity for plantation farming. Until the islands were “opened up” again to the
 Decreolization and African American Vernacular English 391

outside world during the 1930s, the population had relatively little contact with
white speech and thus maintained a form of language much closer to the original
creole.
While decreolization thus is a possible final development in what sometimes
is called the “life cycle” of pidgins, it is not a necessary event. The major devel-
opmental step lies in the process of creolization (or depidginization), which turns
a radically simplified and socially highly restricted form of communication into
a full-blown language, with the complexity and social versatility of ordinary lan-
guages. From this perspective, decreolization is simply a step “sideways”, from
one form of fully developed language to another.
Chapter 15: Language death
Gaelic’s no use to you through the world.
(Said by a Gaelic speaker justifying why she is teaching her children English, not Gaelic.
Reported by Nancy Dorian in The loss of language skills, ed. by R. D. Lambert & B. F. Freed,
1982.)

But today, by reason of the immense augmentation of the American population …, the
Indian races are more seriously threatened with a speedy extermination than ever before
in the history of the country.
(Donehogawa, first American Indian to be Commissioner of Indian Affairs, Report of the US
Department of the Interior, 1870)

To continue in the “life-cycle” metaphor introduced at the end of the preceding

chapter, we turn our attention now to developments at the end of the life cycle of
languages when, in essence, a language loses its native speakers and thus under-
goes language death. Like the birth of languages through pidginization and cre-
olization, language death results from language contact.
Throughout history, speakers have given up their native language in favor
of another, more prestigious – or economically or socially powerful – form of
speech, whether the latter be the language of a foreign conqueror, a link language,
a koiné, or a creole. While this much has been known for a long time, the manner
in which communities switch language loyalties and the effects of this switch on
linguistic structure were only poorly understood. Even less well examined is the
related issue of dialect death; for dialects, too, can lose their speakers and die out.
In fact, dialect death is even more widespread than language death. Although we
continue in this chapter to speak of the death of languages, our observations also
apply to dialects.
Traditional linguistic literature may mention the date at which “the last
speaker” of a given language died. But it does not examine questions such as
“Whom did that last speaker speak to?” or “What was his or her language like?”
Only recently has systematic research begun on the issue of language death.
Findings still are limited in number and spotty in terms of the languages and the
aspects of grammar that have been studied. Nevertheless, certain general patterns
are beginning to emerge.
First, given the metaphor of language death, it is useful to distinguish differ-
ent manners in which languages may be said to die – “language suicide”, “lan-
guage murder”, and other language death. Second, we must distinguish between
individual loss of language skills and the disappearance of entire speech com-
munities. While both are interesting, only the latter development has permanent
effects on language history.

https://doi.org/10.1515/9783110613285-015
 Language death 393

Language suicide generally is a matter of individual decision. As such it is

very common among immigrants to areas with a linguistically homogeneous pop-
ulation and without any tradition of bi- or multilingualism. Thus, many immigrant
families in the United States decide soon after entering the country that their chil-
dren should be fluent in English so as to be able to compete with native speakers
of English. To this end the parents decide not to use their native language any
more, but only English. Sometimes the decision to abandon one’s native language
is motivated by special social or political considerations. For instance, refugees
from Nazi Germany, both Jews and non-Jews, were so strongly opposed to what
was happening in Germany and in the name of Germany that they decided never
to speak German again. Similarly, immigrants to Israel generally have decided to
switch from their native languages to Modern Israeli Hebrew, the language of the
country that promised to give them a clear identity and freedom from persecution.
Although widespread as an individual phenomenon, language suicide does
not seem to be common for whole speech communities. Still, its effects should
not be underestimated. Yiddish, the language of most Eastern European Jews,
has undergone severe attrition as the result of language suicide – either in Israel
(in favor of Modern Hebrew) or in the United States (in favor of English). (The
number of Yiddish speakers has of course also been severely diminished by the
mass murder of Jews under the Nazis.)
Language murder can be said to take place if individuals or whole speech
communities are forced to abandon their native language in favor of a language
favored by a politically more powerful group. Again, this is not an uncommon phe-
nomenon, at least on an individual basis. For instance, after World War I, United
States schools, with support from national teachers’ organizations and various
forms of government, strongly discouraged or prohibited the use of any language
other than English by the children of immigrants. And in the early twentieth
century, the United States collected American Indian children into government
schools and forced them to abandon their native languages and cultures in favor
of the language and customs of the English-speaking population. Most teachers’
organizations and governments have since then taken a different position, and
the repression of languages other than English has to a large extent given way to
greater tolerance. Bilingual education programs have been set up, generally con-
ceived of as transitions to full acquisition of English. Many educators have argued
in favor of bilingualism as a goal in itself. On the other hand, new organizations,
and an entire new movement (“English Only”), have been formed with the goal
of making English the official national language and of doing away with more
permissive bilingual programs – and with bilingualism in general. (Closer exam-
ination reveals that much of the English Only movement is motivated not just by
a rejection of bilingualism but by xenophobia.)
394 Language death

Similarly, until very recently non-private schools in Wales, in Great Britain,

and Brittany, in France, have, as a matter of policy, prohibited the use of Welsh
and Breton, respectively, by school children, even during recess, requiring them to
use English and French (respectively) at all times. Other minority languages have
generally received the same treatment throughout early modern Europe and into
the twentieth century.
Although such repressive policies are directed at entire speech communities,
they do not necessarily meet their goals. In many cases, enough individuals have
braved social ostracism or even severe punishment to maintain their language,
or to prevent the immediate death of the language. At the same time, there is no
denying that many languages have disappeared and are still dying around the
globe. One recent estimate is that more than half of the world’s approximately
6,000 distinct languages are facing extinction within this century. Significantly,
although being distinct languages, almost all of them lack the “army and navy”
which one linguist jokingly claimed are needed to define a form of speech as a
language, not a dialect; see Chapter 10, § 2.
On the level of the speech community, language death commonly does not
take place suddenly, within just one generation. Rather, it is a slow process which
may extend over more than three generations. Death, moreover, is not necessar-
ily caused by overt repression. It often arises in situations where a non-native
language is considered more prestigious or useful (compare the first quotation at
the beginning of the chapter). Or use of the non-native language may be required
in certain contexts (such as school or dealing with governmental authorities),
but not necessarily in others. As a result, the native language is employed less
frequently and in fewer social contexts.
Even for individuals, language attrition tends to be a fairly slow process.
Speakers who have been abroad for only a short period and have regularly used
a foreign language during that time, may return with a noticeable “accent”. They
also may have missed out on some of the most recent innovations in their home
country, especially in vocabulary and colloquialisms or slang. However, they
generally have no difficulty expressing themselves. Extended stays abroad can
have more serious results. Thus, when H. H. Hock returned to his native Germany
after thirteen years in the United States, he experienced an excruciatingly frus-
trating first week. Having used English not just for speaking but for teaching,
publishing, even thinking, he found even the most simple sentences difficult
to produce without at least some errors in usage. (Eventually, fluency returned,
and he was complimented on how good his German was after so many years of
absence.)
Losing much of one’s native language skills, however, does not imply for all
practical purposes becoming a native speaker of the “other” language. Depending
 Language death 395

on motivation and opportunity, the outcome may be a lack of native competence

in both the native and the second language.
Language attrition of this sort becomes most relevant if for some reason it
becomes a feature of whole speech communities. In that case, reduction in use
and competence limits the input on which new speakers of the language can draw
in order to learn their language. In many cases, the most common constructions
in the language may remain unaffected. Less frequently used structures, however,
may come to be heard so rarely that the next generation of learners finds it diffi-
cult, if not impossible, to determine how to use them correctly.
The new generation of speakers may therefore avoid using such construc-
tions, thus further reducing the input for the next generation of speakers, and so
on. In this manner the rule system of the language undergoes a slow process of
atrophy. Studies on language death in very different locations, focusing on very
different grammatical phenomena, agree that grammatical attrition is not across
the board or random, but takes place in terms of general rules of the language
fading out, and in this process, rules which are from some perspective more dif-
ficult are lost first.
Interestingly, grammatical atrophy is not matched by a similar decrease in
vocabulary or in the younger generations’ ability to understand older speakers.
Passive and active command of the language may differ considerably.
At a certain point, the atrophy reaches a point when a new generation of
speakers is no longer able to learn the grammatical rule system of the language.
The members of this generation, often referred to as semi-speakers, fluently
understand even their grandparents’ speech, but generally admit that they are
unable to speak the language themselves. At this point, transmission of the lan-
guage has come to an end, and the language has effectively died. One suspects
that many of the last speakers mentioned in traditional accounts were in fact the
last members of such a generation of semi-speakers.
Even when language death has happened, the language is not necessarily
dead forever. As long as there are sufficient records of the language as it once was
used, it is possible to resurrect it. Similarly, it is possible for a speech community
that has been experiencing language attrition to reverse the process and to revive
the language to its earlier status, by turning to the more fully intact speech of the
older generation.
The most celebrated example of the resurrection of a language that had died
out as a normal spoken medium is that of Modern Israeli Hebrew. While Hebrew
continued to be used as a language of scripture, ritual, and to a lesser degree
scholarship, it had died out as a spoken language among most Jews. The first
steps toward its revival were undertaken by the nineteenth and twentieth century
Zionist movement in its attempt to assert the Semitic identity of the Jewish people.
396 Language death

But its call to use Hebrew as an overt sign of this identity was challenged by other
groups who advocated Yiddish as the language of Jewish identity. The revival of
Hebrew as a spoken language was therefore limited to relatively small groups. The
Nazi holocaust and the founding of Israel as a country where Jews would be able
to live without fear of further persecution radically changed the situation, and
within a very short time there was an almost complete switch from Yiddish and
many other European languages to Modern Hebrew.
Many other peoples have attempted to revive their language, but generally
with more mixed success. For instance, there have been recent attempts in the
British Isles to revive Manx and Cornish, ancient, (near-)extinct Celtic languages
spoken on the Isle of Man and in Cornwall, respectively. The European continent
is witnessing attempts to revive Latin as a spoken language. Some even hope
to make Latin the common language of the European Union, thus avoiding the
problem of having to decide which link language to use. Similarly, groups in India
are attempting to revive Sanskrit, which since the 1970s has been dying rapidly in
its spoken use among traditionally educated scholars. Here, too, it can be argued
that the use of Sanskrit as a common link language avoids the problem of having
to decide between Hindi and English (see Chapter 12, § 1). It combines within
itself the advantages of both Hindi and English. Like Hindi it is an indigenous
language, and like English it is not a native language for any sizable community
within India and therefore does not bestow special privileges on one community
while discriminating against the others.
In the United States, there have been repeated attempts by American Indians
to revive or resurrect, or as some say, to “reawaken” their languages. In part,
these attempts are motivated by the fact that the United States accords special
legal status to peoples who can demonstrate cultural and linguistic continuity
with their indigenous roots. To succeed in these attempts, many communities
have asked linguists and anthropologists to make available to them recordings
of indigenous languages made in the nineteenth and early twentieth century,
when the languages were still spoken, or were spoken in fuller, less diminished
form.
Language death thus presents interesting and important challenges to speak-
ers, to be sure, but also to linguists. The most important is the question whether
we, as linguists, have a special responsibility to ensure that languages do not die.
Some linguists take a “Darwinian” position, arguing that languages always have
died and always will die when they no longer are useful to their speakers, and
that linguists have no business to interfere with this natural development. Others
feel that we cannot force speakers to maintain their languages if they do not want
to. A third group takes a more interventionist position, arguing that any loss of
language diminishes the world, just as the death of any animal or plant species
 Language death 397

threatens our ecosystem. These linguists actively support groups that are trying to
preserve or revive their language, or even encourage them to do so.
The truth probably is on the side of those who argue that one cannot force
speakers to maintain their language. In fact, it is probably just as imperialist or
paternalist to tell speakers that they must do so as it is to try to suppress their
language.
Linguists, however, need not remain on the sidelines. They can help groups
interested in reviving or preserving their language by providing relevant infor-
mation on grammar, vocabulary, and usage, or by preparing teaching materi-
als which enable the members of these groups to provide formal instruction in
their language. Through such efforts, these groups can counter or overcome the
common prejudice that a form of speech is a language only if it is taught in school
and has a formal grammar, while all other forms of speech are “dialects” and
therefore do not merit preservation. Linguists can also help in developing the
vocabulary necessary to permit the language to branch out from its traditional
setting and to become fully functional in the modern world.
Linguists can serve the cause of language preservation in another, more indi-
rect way, by collecting the greatest possible amount of grammatical and lexical
information, as well as entire texts, of languages that are in danger of becoming
extinct. These materials can then be made available to speakers if they decide
at some future point to reverse the course of language death or to revive the lan-
guage. Without such materials, even the most successful attempt at language res-
urrection, that of Modern Hebrew, would have come to naught.
Ultimately, of course, we must accept the fact that no linguist can stem the
course of language death. Only the speakers of the language can do so – if they
have the motivation, the opportunity, and the wherewithal.
Chapter 16: Comparative method:
Establishing language
relationship
The Sanscrit language, whatever be its antiquity, is of a wonderful structure; more perfect than the Greek, more
copious than the Latin, and more exquisitely refined than either, yet bearing to both of them a stronger affinity, both
in the roots of verbs and in the forms of grammar, than could possibly have been produced by accident; so strong
indeed, that no philologer could examine them all three, without believing them to have sprung from some common
source, which, perhaps, no longer exists: there is a similar reason, though not quite so forcible, for supposing
that both the Gothick and the Celtick, though blended with a very different idiom, had the same origin with the
Sanscrit; and the old Persian might be added to the same family, if this were the place for discussing any question
concerning the antiquities of Persia
(Sir William Jones, Third Anniversary Discourse, on the Hindus, Royal Asiatic Society, 1786)

1 Introduction
The epigraph above, which readers will remember from Chapter 2, has had a
double significance for the history of linguistics. On one hand, it provided an
important stimulus for research in comparative Indo-European linguistics, a
field which soon became the most thoroughly investigated area of historical and
comparative linguistics and which to the present has remained the most impor-
tant source for our understanding of linguistic change. This is the issue that we
pursued in Chapter 2.
More important yet, Jones’s statement is significant because it offers a suc-
cinct and explicit summary of what have turned out to be the basic assumptions
and motivations of comparative linguistics: accounting for similarities which
cannot be attributed to chance, by the assumption that they are the result of
descent from a common ancestor.
To establish this kind of account we must naturally look for languages that
seem to share enough similarities to suggest that there may be a genetic relation-
ship. In many cases, this is not all too difficult, once we accept the basic notion
that languages may be genetically related to each other. To illustrate the point,
consider Table 1 below.
As the table shows, even seven lexical items – if selected with care – can
furnish strong evidence that the Indo-European languages of Europe (Breton–
Latvian) are related to each other. The case is similar for the Uralic languages
(Finnish, Estonian, and Hungarian), although the case for Hungarian may be
less obvious. Moreover, the table permits us to distinguish subgroups within,

https://doi.org/10.1515/9783110613285-016
 Introduction 399

say, Indo-European: Celtic (Breton, Welsh, Irish), Germanic (Icelandic–

German), Romance (French–Romanian), Slavic (Bulgarian–Russian), and Baltic
(Lithuanian and Latvian); Greek and Albanian constitute subgroups of one
each.

Tab. 1: Vocabulary correspondences in the major European languages

‘one’ ‘two’ ‘three’ ‘head’ ‘ear’ ‘mouth’ ‘nose’

Breton ünan dau tri penn skuarn genu fri

Welsh ɨn dai tri pen klɨst keg truɨn
Irish ȫn dɔ tri kyan kluǝs byal srōn
Icelandic eidn tveir þrīr hȫfüð eira münnür nēf
Danish en tōʔ trēʔ hōðǝ ōrǝ monʔ nǣsǝ
Norwegian ēn tō trē hōvǝd ȫrǝ mund nǣsǝ
Swedish ēn tvō trē hȫvud ȫra mun nǣsa
Dutch ēn tvē drī hōft ōr mont nȫs
English wǝn tū θrī hed īr mawθ nowz
German ʔains tsvai drai kɔpf ʔōr munt nāzǝ
French œ̃/ün dö trwa tɛ̄ t orɛ̄ y buš ne
Spanish uno dos tres kaβeθa orexa boka nariθ
Portuguese ũ doyš treyš kǝbesǝ orela bokǝ nariž
Italian un(o) due tre testa orɛkkyo bokka nazo
Romanian un doy trey kap ureke gurǝ nas
Albanian ñǝ dü tre kok(ǝ) veš goy(ǝ) hundǝ
Greek éna ðyó trís kefáli aftí stóma míti
Bulgarian yedan dva tri glava uxo usta nos
Serbian yedan dva tri glava uho usta nos
Czech yeden dva tři hlava uxo usta nos
Polish yeden dva tši gwova uxo usta nos
Russian adyin dva tryi galavá úxo rot nos
Lithuanian vyíenas dú tryīs̃ galvá ausyís burná nṓsyis
Latvian viens divi trīs galva auss mute deguns
Finnish üksi kaksi kolme pǟ korva sū nenä
Estonian üks kaks kolm pea kõrv sū nina
Hungarian eǰ kēt hārom fȫ/fey fül sāy orr
Turkish bir iki üč baš kulak āız burun
Basque bat bi hirür bürü belari aho südür

(Note: Except for French ‘one’, the numerals are cited without gender variation. Finnish and Esto-
nian ä = [æ].))
400 Comparative method: Establishing language relationship

Tab. 2: Further data from selected Asian languages

Hindi ēk dō tīn sir/sar kān mũh nāk

Marathi ēk dōn tīn ḍōī, muṇḍ kān tō̃ī nāk
Kashmiri akh zɨʔ trɨʔ kalɨ kan ɨs nas
Persian yek do se sær guš dæhɔn bini
Ossetic iw dɨwwǝ ǝrtǝ sǝr qus dzɨx fɨnd
Armenian mi erku erekh glux unkn beran r̄ǝngunkh
Tocharian B ṣe, sana wi trai, tarya āśce klautso koyṁ meli
Tamil ondrɨ iraṇḍɨ mūndrɨ talɛi kāðɨ, sevi vāy mūkkɨ
Kannada ondu eraṛu mūru tale kivi bāy, mūti mūgu

(Note: Unlike the other forms, the Classical Armenian and Tocharian B forms are not given in
phonetic transcription but in a transliteration of their original spelling. The Tocharian words for
‘one’ and ‘two’ distinguish between masculine and feminine forms.)

Given the evidence of just the seven words in Table 1, there might appear to be
a somewhat weaker case for a Turkish-Basque relationship. Compare the simi-
larities in the words for ‘one’, ‘head’, and possibly also ‘three’. As far as we can
tell, however, the similarities are misleading. Once the basis for comparison is
enlarged to, say, a hundred lexical items, it turns out that the Turkish-Basque
similarities are most likely the result of chance. In fact, up to this point it has not
been possible to successfully establish a genetic relationship between Basque and
any other language or language group. The fact that there can be such accidental
similarities raises important questions about our ability to establish genetic rela-
tionship by sheer inspection of vocabulary. (This matter is pursued further in § 2
below and in Chapter 17.)
The situation gets more complex once we introduce selected Asian languages,
as in Table 2, a continuation of Table 1.
On one hand, the sets Hindi dō : Marathi dōn : Persian do : Osset. dɨwwǝ
‘two’, Hindi and Marathi tīn : Kashmiri trɨʔ : Tocharian trai/tarya ‘three’, and Kash-
miri nas ‘nose’ (perhaps also Hindi nāk ‘nose’) may suggest relationship to the
Indo-European languages of Europe because of the phonetic similarities between
the words. Note also Hindi mũh ‘mouth’ and especially Marathi muṇḍ ‘head’ on
one side and the Germanic words for ‘mouth’ on the other.
On the other hand, Hindi/Marathi ēk, Kashm. akh, Pers. yek ‘one’ look more
similar to Finn. üksi, Est. üks, Hung. eǰ, and so do Hindi kān, Kashm. kan ‘eye’
to Finn. korva, Est. kõrv – as well as to Kannada kivi. In fact, Hindi nāk ‘nose’,
perhaps also Kashm. nas, could just as well be considered related to Finn. nenä,
Est. nina as to the words for ‘nose’ in the European members of the Indo-Euro-
pean language family. And while Hindi mũh and Marathi muṇḍ bear strong resem-
 Introduction 401

blances to the Germanic words for ‘mouth’, they are similar, too, to Kannada
mūti.
But there are more problems. First, the evidence accumulated by more than
a century of comparative linguistics suggests an especially close relationship
between Indo-Aryan Hindi, Marathi, and Kashmiri on one hand and Iranian
Persian and Ossetic (Iron dialect) on the other. This relationship, however, does
not come out well in Table 2, except perhaps for the word for ‘head’.
Further, the evidence for considering Tocharian an Indo-European language
appears to be limited to one word, the numeral ‘three’, and there seems to be no
evidence for considering Armenian an Indo-European language. Again, this con-
flicts with what we know as the result of extensive work in comparative Indo-Eu-
ropean linguistics.
Finally, the evidence does correctly indicate that the Dravidian languages
Tamil and Kannada (in the south of India) are not particularly closely related
to any of the other language families. However, it fails to indicate that there are
recurrent similarities between Dravidian and Uralic which may suggest a possible
relationship (see Chapter 17).
As it turns out, a number of the similarities that we just noted are accidental,
just like those between Turkish bir and Basque bat ‘one’, and therefore do not
reflect genetic relationship.
This is the case for the Hindi/Marathi and Germanic words for ‘mouth’. Hindi
mũh derives from Sanskrit mukha- which, if inherited, would reflect an earlier
*mukho-; Marathi muṇḍ goes back to Skt. mūrdhan-, a reflex of PIE *melǝdh/
ml̥ǝ̯ dh-; while the Germanic words reflect PGmc. *munþa- which must go back to
PIE *mn̥ to-. Note further that Kannada mūti is related to Tam. mūñči ‘face’, which
appears to be older both in its phonetic shape and in its semantics. That is, as we
trace these similar forms back in history we find that they become less similar. The
modern similarities, thus, must be due to chance.
The similarities between the Hindi/Kashmiri/Persian and Finnish/Estonian/
Hungarian words for ‘one’ likewise are accidental. The Indo-Aryan words go back
to Skt. ēka-, derived from PIE *oy-ko-’one, single’, which in turn represents an
extension by a suffix -ko- of the same root which, extended by the suffix -no-,
underlies the numeral ‘one’ that is found in the majority of the Indo-European
languages of Europe.
The similarities between Hindi and Persian/Ossetic in the word for ‘head’
reflect the fact that the Hindi word has been borrowed from Persian. Here, too,
then, the similarities do not reflect inheritance from a common ancestor.
On the other hand, a number of genuine cognates are difficult to detect
without extensive comparative research. For instance, the Tocharian and Arme-
nian words for ‘one’ are ultimately related to the one found in Greek; all three of
402 Comparative method: Establishing language relationship

these derive – believe it or not – from PIE *sem- ‘same, similar, identical’. The
Armenian and Ossetic words for ‘three’ are perfect cognates of the words found
in the rest of the Indo-European languages. Again, given the evidence in Tables 1
and 2, this may be hard to believe; but more than a century of research has shown
that the Armenian and Ossetic words are related. The Armenian form results from
a change of *tr- > θr-, loss of the θ, and prefixation of a vowel before the resulting
initial r; the Ossetic form involves metathesis of initial *tr- to *rt-, an areal phe-
nomenon in the Caucasus, plus prefixation of a vowel before initial *rt- and other
changes. The Armenian word for ‘two’ can likewise be related to its counterparts
elsewhere in Indo-European, through a sequence of even more complex develop-
ments.
The upshot is that not all similarities – or dissimilarities – between languages
in their vocabulary are indicative of genetic relationship, and that in order to
establish genetic relationship we have to go significantly beyond comparing just
seven vocabulary items.

2 Chance similarities, onomatopoeia, and

“nursery words”
Probably any given pair of languages will offer at least some formally and seman-
tically similar linguistic items whose similarities are simply due to chance. We
have seen a few examples of this type in the preceding section, such as the Hindi,
Marathi, Kannada, and Germanic words for ‘mouth’ (or ‘head’). Even among the
Indo-European languages of Table 1, some resemblances are accidental. This is
certainly true for the similarity between the words for ‘one’ found in most of the
languages and Modern Greek éna ‘one’. The Modern Greek form reflects Ancient
Greek heîs, mía, hén ‘one (m., f., n.)’. This form, in turn, derives from earlier *sem-s,
*sm-ia, *sem, with a root *sem- related to Engl. same. The remainder of the words
for ‘one’ derive from a different ancestor, *oy-no-.
Modern Greek has a word mati ‘eye’ whose phonetic and semantic resem-
blance to Malay mata is remarkable. But again, from what we know about the
earlier history of these languages, the similarity between the words is due to
chance. Mod. Gk máti goes back to earlier Gk. ommátion, a diminutive form of
ómma ‘eye’, which in turn derives from an earlier *okw-m(e)n-, in which *-m(e)
n- was a derivational suffix, and only the first element *okw- originally meant
‘eye’. Ironically, this element, found also in ṓps ‘eye, face’, as well as in Kúklōps
‘Cyclops, lit. having a (single) circular eye’ and related to Engl. eye, has disap-
 Chance similarities, onomatopoeia, and “nursery words” 403

peared from Mod. Gk. máti. What is left consists only of suffixal material, histor-
ically speaking.
There are similar problems for Mal. mata. While most Malayo-Polynesian
languages have cognates of mata ‘eye’, many offer evidence for a related form
kita ‘see’ (e. g. Tagalog kita), and others testify to a form buta ‘be blind’ (e. g. Fiji
buto). These variant forms suggest that mata is morphologically composite, con-
sisting of a root -ta meaning something like ‘sight’ and a prefix ma- which fixes
the meaning of -ta to ‘eye’, while ki- and bu- alter the meaning to ‘see’ and ‘be
blind’ respectively.
Similarly, Modern English and Modern Persian have phonetically and seman-
tically virtually identical forms for ‘bad’: bad [bæd] and bad [bæd] (with a vowel
slightly more retracted than that of the English word). The origin of the English
word is somewhat controversial. Two derivations have been proposed, one from
OE bæddel ‘hermaphrodite, effeminate man’, the other from OE (ge)bæded ‘cap-
tured’. (Both of these involve extensive semantic shifts, generally involving pejo-
ration.) The Persian word, on the other hand, derives from earlier Pahlavi wad,
whose initial w- cannot possibly be related to the b- of the English form, either of
its putative Old English ancestors, or any other imaginable ancestral form.
We can avoid being misled by chance similarities if we insist that our com-
parison be based on a very large data base. For if we find striking similarities in
pronunciation and meaning in, say, a thousand words, the possibility that these
similarities are due to chance becomes rather remote. Note that the data base
must be very large, for as (1) below shows, it is not at all difficult to find a fairly
large number of chance similarities between any given pair of languages. (San-
skrit and English are of course related to each other. However, we know their
linguistic histories sufficiently well to be certain that the similarities in (1) do not
reflect genetic relationship. On this matter see also Chapter 17.)

(1) Sanskrit Mod. Engl.

kōṇa- corner
ǰhampa- jump
taru- tree (correct cognate: Skt. dāru-)
tōraṇa- door (more accurate cognate: Skt. dvāra-)
krōśati cries
gati- gait
lōkati looks
marīčikā mirage
rāga- rag (musical terms)
vāmā woman
(and others, the total being at least 26 items)
404 Comparative method: Establishing language relationship

In addition, certain types of vocabulary are notoriously unreliable for establishing

genetic relationship. One of these is onomatopoeia. Although details may differ,
onomatopoetic expressions come out remarkably similar in different languages.
See for instance the rooster calls in examples (17) and (17’) of Chapter 7. In spite
of their differences, Engl. cockadoodledoo, Germ. kickericki, Fr. cocorico share a
repetition of velar stops, an inserted liquid (generally an r-sound), as well as,
generally, accent on the last syllable.
Another area in which caution is advisable consists in vocabulary such as
Engl. dad(dy), mom(my), baby, It. papà, mamma, bambino, Hindi bāp, mā̃, baččā.
The structure, sounds, and often also the connotations of these words suggest
that they are nursery words, i. e., words and meanings which adults assign to
the early babbling syllables of infants. Like other items that occur in this kind of
language, the words tend to have reduplicated consonant-vowel syllables (as in
pa-pa). And the predominance of the vowel a and of the consonants p/b, t/d, m,
and n in these words is a common feature of early babbling. (See Chapter 9, § 2.)

3 Similarities due to linguistic contact

Even if we eliminate chance similarities, onomatopoeia, and nursery words, we
are not necessarily home free. We may be confronted with situations of the type
(2). Here it appears as if English is simultaneously and equidistantly related to
two distinct languages, French and German, with no evidence for genetic rela-
tionship between these two languages if we limit ourselves to the evidence in
(2). In biology, such a dual relationship might not be entirely unexpected, since
there is such a thing as cross-breeding. But in genetic/comparative linguistics,
this type of relationship is considered suspect. The suspicion always arises that
such a relationship is attributable to borrowing. And in fact, in the present case
we know from the history of English that the correspondences between English
and French result from the secondary contact between the two languages after the
Norman conquest of England.

(2) English French German

calf Kalb
veal veau
cow Kuh
beef bœuf
swine Schwein
pork porc
 Similarities due to linguistic contact 405

Even if we did not have this direct historical knowledge, we would be able to make
a good case for a borrowing relation between English and French by looking at
other items, such as the ones in (3).

(3) English French German

to à zu
too trop zu
two deux zwei
twenty vingt zwanzig
eat manger essen
bite mordre beissen
father père Vater [f-]
mother mère Mutter
three trois drei
thou tu du

In these correspondences (which could be multiplied many times over), it is

easy to see that there is a close relationship between English and German, while
French generally offers very different forms. True, closer examination now reveals
some recurrent similarities which also involve French (see the last four words
in (3)); but it is also clear that French does not exhibit any closer affinities with
English than it does with German (or vice versa). In fact, in the words for ‘father’
and ‘mother’, and in many others like them, the similarities between English
and German are much more striking than those of either of the two languages
to French. The special affiliation of English with French that might have been
suggested by the correspondences in (2) thus turns out to be contradicted by the
evidence of additional data.
What is even more significant, a comparison of the data in (2) with those in
(3) shows that the English/French similarities are restricted to certain, limited,
spheres of the vocabulary. On the other hand, the German/English similarities
pervade the whole lexicon, including basic vocabulary. As noted in Chapter 8,
borrowing tends to be limited to certain spheres of the lexicon. In addition, it is
often, though not always, restricted to technical vocabulary. And it has the least
effect on basic vocabulary.
Cases like the English/French/German relationship are important because
they provide insights that make it possible to detect borrowings in other cases,
where we do not have direct historical evidence. If the similarities between two
given languages are limited to certain spheres of the lexicon and if they cover
little, if any, basic vocabulary, then there is a strong reason to suspect that they
result from borrowing, not from genetic relationship.
406 Comparative method: Establishing language relationship

Contact-induced similarities can also be found in overall structure, as the

result of convergence. Recall for instance the case of the Balkans, of South Asia,
or – even more strikingly – of Kupwar (Chapter 13, § 2). On the other hand, diver-
gence in overall structure does not necessarily argue against genetic relation-
ship. For instance, the modern Indo-European languages, though clearly related,
exhibit the following basic word orders: Indo-Aryan, Iranian, and Armenian have
SOV (Subject : Object : Verb); Celtic offers VSO; German, Dutch, and Frisian have
a mixture of SOV and SVO characteristics; and most of the others exhibit SVO.
Similarities and differences in overall structure, thus, are not a reliable guide to
establishing relationship. (But see § 5 below on similarities in specific aspects of
structure.)
Comparative linguists therefore usually concentrate on vocabulary and on
correspondences that emerge from an examination of vocabulary. Moreover, since
basic vocabulary is less likely to be borrowed, the evidence of such vocabulary
receives the highest priority.

4 Systematic, recurrent correspondences

We can strengthen our argument for genetic relationship between given languages
by showing that their similarities are not helter-skelter or sporadic, but that they
are systematic and recur in large sets of words. In fact, given that sound change
is overwhelmingly regular, we must expect a great degree of systematicity and
recurrence in the phonetic similarities between putatively related languages.

Consider again the case of English and German. If we add the data in (4) to those
in (2) and (3), we note some important phonetic differences between English
words with t and their German counterparts. However, within these differences,
we can establish a great systematicity; see the summary in (5). Moreover, even
though there may be differences, the German counterparts of English t are pho-
netically similar, in that like t, they are dental. And if we expand our horizon to
include words with English p and k, we find a very similar situation, allowing
for some minor differences; see (6). Given these facts, the conclusion becomes
almost inescapable that these words, and many others like them, go back to a
common ancestor and have become different through the operation of regular
sound change. The ability to find such regular and systematic correspondences
between languages is the cornerstone of establishing genetic relationship. In fact,
systematic correspondences are to be expected, given the overwhelming regular-
ity of sound change.
 Shared idiosyncrasies 407

(4) English German

frost Frost
chest Kiste

(5) English German

t z [ts] (initially and after consonant)
t ss [s] (intervocalically)
t t (after s)

(6) English German

pound Pfund
p- : pf-
penny Pfennig
ape Affe
-p- : -f-
hope hoffen
aspen Espe
-sp- : -sp-
wasp Wespe
cool kühl
k- : k-
card Karte
make machen
-k- : -x-
cook Koch

5 Shared idiosyncrasies
We can yet further improve our case if we can find shared idiosyncrasies in mor-
phology.
Consider the English and German comparatives of Engl. good and its German
counterpart gut, which are formed from what looks like a completely different
lexical item – better, best and besser, best-; see (7a). Morphological relation-
ships of this type are commonly referred to as suppletion. Contrast the supple-
tion in (7a) with the normal pattern in Engl. warm : warmer : warmest, Germ.
warm : wärmer : wärmst-. Now, (7a) demonstrates that French likewise has supple-
tion; but significantly, the English and German data exhibit systematic and recur-
rent similarities with each other, while the French forms are radically different. If
we had to choose which of these patterns of suppletion must result from genetic
relationship, we would surely opt for the patterns found in English and German.
To select English and French would border on the perverse.
Similarly, the early Indo-European languages and even some modern ones
exhibit striking similarities in the third person singular and plural forms of the
408 Comparative method: Establishing language relationship

verb ‘to be’, including a remarkable paradigmatic alternation between vowel +

s- in the singular and simple s- in the plural. Compare (7b).

(7) a. English German French

good gut bon
better besser meilleur
best best- le meilleur
(For the correspondences -t- : -s-, -st- : -st-, see (4) above. The other correspondences
are similarly supported by “outside” evidence; compare the g- : g- in Engl. great,
give : Germ. gross, geben.)

b. Sanskrit Latin Mod. Germ. Old Church Slavic

as-ti ‘is’ es-t ‘is’ is-t ‘is’ es-tŭ ‘is’
s-anti ‘are’ s-unt ‘are’ s-ind ‘are’ s-ǫtŭ ‘are’

Highly idiosyncratic phonological alternations such as those in (7b) do not nor-

mally get borrowed, nor do suppletive patterns like the ones in (7a). The morpho-
logical idiosyncrasies exhibited by these patterns, therefore, combined with the
fact that they involve systematic phonological correspondences, would be diffi-
cult to explain except as reflecting common heritage.

6 Reconstruction
Most historical linguists believe that the ultimate proof of genetic relationship lies
in reconstruction, i. e., in reversing linguistic history, as it were, by postulating
linguistic forms in an ancestral or proto-language from which the attested forms
can be derived by plausible linguistic changes. Note that “proof” here is to be
understood more or less as in a court of justice, as establishing a case beyond a
reasonable doubt. Moreover, to be probative, the reconstruction must be based on
a large amount of lexical items and at the same time conform to a set of evaluative
principles that are presented below.
For an illustration, consider the data in (8) as they bear on the reconstruction
of Proto-Indo-European vowels.
 Reconstruction 409

(8) Sanskrit Greek Latin Germanic Reconstruction

a. idam id(em) Go. ita ‘it, that’ *i
likta- é-lip-on (re-)lic-tus ‘left’
b. yugam zugón iugum Go. juk, ‘yoke’ *u
OE geoc
budh- puth- OE budon, ‘(a)bide,
geboden awake’
c. asti estì est Go. ist, OE is ‘is’ *e
atti édomai edō Go. itan,
OE etan, itiþ ‘eat’
d. aṣṭau oktṓ octō Go. ahtau ‘eight’ *o
e. aǰati ágō agō ON aka ‘drive’ *a
f. pitar- patḗr pater Go. fadar ‘father’ *ǝ
(Note: The data above are representative of larger sets. Given the overall available evidence, it is
not possible to assume that (some of) the above sets can be “collapsed” in the process of recon-
struction and that the observed differences might be attributed to special developments in the
individual languages. That is, each set must be reconstructed separately for the proto-language.)

In reconstructing, we must keep in mind the following principles which deter-

mine the acceptability and plausibility of our reconstruction:
– Reconstructed items and systems and postulated linguistic changes should
be natural. A corollary of this principle is that postulated sound changes
must be regular, in conformity with the regularity principle of Chapter 4.
A further corollary is that there must be a phonetic value attached to a recon-
structed sound. For instance, it would be unacceptable to reconstruct set (8 f)
as *#. What would be the phonetic value of such a symbol? We must assume
that a reconstructed (proto-)language is essentially like any language observ-
able today – since known languages do not have sounds like #, presumably
neither did Proto-Indo-European. A natural reconstruction would instead be
*[ǝ]; for as (9) shows, [ǝ] can naturally change to either [a] or [i].

(9) i u

e ǝ o

Moreover, it would be dubious to reconstruct *[a] for both sets (8e) and (8 f) –
or for sets (8d) and (8e), for that matter. To do so would require the assump-
tion that contrary to normal expectations, sound change operates in a spo-
410 Comparative method: Establishing language relationship

radic, irregular fashion, such that *[a] can change randomly to i or to a in

Indo-Iranian, without any motivation for the different developments.
– The reconstruction must not violate occam’s razor. According to this maxim,
attributed to the medieval English philosopher William of Occam, Entia (or
essentia) non sunt multiplicanda praeter necessitatem ‘Entities [in an argu-
ment] should not be multiplied beyond necessity.’ This maxim, incidentally,
is fundamental to all scientific inquiry.
In comparative reconstruction, such “entities” are (i) reconstructed items,
and (ii) changes required to convert these items into the forms attested in the
descendant languages.
For instance, it would be a violation of Occam’s Razor if we reconstructed
*[æ] for set (8a). This reconstruction would require the entirely unnecessary
assumption either that all the languages innovated by changing *[æ] to [i] or
that there was such a change in the ancestor language.
On the other hand, while the reconstruction *[ǝ] for set (8 f) introduces an
additional reconstructed sound, which is not attested as such in any of the
daughter languages, it does so by necessity. To do otherwise would result in a
violation of our first principle, since it would entail the unnecessary assump-
tion of irregular sound change.
Finally, Occam’s Razor argues that set (8d) should be reconstructed as *[o].
This reconstruction makes it possible to distinguish the behavior of this
set from set (8e), in accordance with the expectation that sound change is
regular. In addition, the reconstruction *[o] makes it possible to account for
the attested forms with a minimum of changes. All we need to assume is that
outside of Greek and Latin, the distinction between *[o] and *[a] was lost in
favor of [a]. The Greek and Latin forms, then, represent unchanged outcomes.
If however we reconstructed, say, *[ɔ], then we would have to unnecessarily
assume that the original sound was changed in all the languages, including
Greek and Latin.
– Wherever we can, we use the oldest available stages of languages. This
makes reconstruction simpler, since less time has passed from the time of the
proto-language, and thus there has been less chance for linguistic changes to
obscure the relationship between the languages.
Consider the data in (10). Most striking are the examples in (10a), where the
similarities – and correspondences – between Sanskrit and Old English are
quite clear, while their modern counterparts have come to differ greatly as the
result of linguistic change (mainly lexical replacements). As examples such
as (10b) illustrate, in some cases the relationship remains transparent, even
in the modern languages. But compared to the older stages, such modern
correspondences are much rarer. In fact, (10c) shows that in some cases even
 Reconstruction 411

the oldest stages of the languages have undergone enough changes that the
relationship of words that we know to be inherited from the Indo-European
ancestor has become greatly obscured. Their relationship can be established
only after extensive research in comparative reconstruction.

(10) Sanskrit Old English Hindi Modern English

a. asti is hai is
sa sē vō he
vayam wē ham we
svasar sweostor bahan sister
śvaśrū sweogor sās mother-in-law
vēda wāt ǰāntā hai (he) knows
b. dvā(u) twā dō two
trayas þrī tīn three
pād- fōt pā̃v- foot
c. čakra- hweogol čakkā wheel (< *kwekwlo-)
śr̥ ŋga- horn sī̃g horn (< *ḱer/ḱr̥ -)
bhavati bēon hōnā be (< *bhū-)
ṣaṭ seox chah six (< *s(w)eḱs)
pluśi- flēah pissū flea (< *pl(o)uḱ-)

We are able to assert that the words in (10c) are in fact related to each other
partly because of the evidence of cognates in the early stages of other related
languages, such as Gk. kéras ‘horn’. In other cases, such as Skt. čakra-,
OE hweogol, the relationship can be demonstrated only because we have
reconstructed the Proto-Indo-European ancestral language, established the
sound changes from PIE to languages such as Sanskrit and Old English, and
are therefore able to show that both forms are derivable from a PIE form
*kwekwlo-, which is also reflected in Gk. kúklos ‘wheel, circle’. Moreover,
because we have reconstructed not just the sound system and the lexicon
of Proto-Indo-European, but also its morphology, we are able to explain
*kwekwlo- as a morphological derivative of the independently reconstructed
PIE root *kwel- ‘move, turn’ with an original meaning along the lines of ‘the
thing that keeps turning around’. (The morphological processes are complex,
involving among other things a productive phonological alternation between
el and l [hence *kwel- beside *-kwl-] and a process of reduplication, which
copies the initial consonant and vowel of the root [hence the initial *kwe- of
*kwe-kwlo-].)
412 Comparative method: Establishing language relationship

Evidence of the type (10a), combined with that of (10c), suggests that there
may be an optimal closeness of related languages that is necessary to success-
fully establish genetic relationship. If too many centuries of linguistic changes
have increased divergence beyond that optimal stage, the evidence may become
too limited, and establishing genetic relationship may become difficult or even
impossible. (See also Chapter 17.)

7 W
 hat can we reconstruct and how confident are
we of our reconstructions?
As seen earlier, comparative linguistics places the greatest amount of confidence
in sound correspondences found in lexical comparisons, especially basic vocab-
ulary. Nevertheless, we can reconstruct other aspects of the ancestral language,
beside the lexicon. Based on the methods and assumptions illustrated in the
preceding section, we can reconstruct a fair amount of the phonology of the pro-
to-language; and using different and more sophisticated methods, we can gain a
pretty good picture of the morphology of the proto-language and of aspects of its
syntax.
Ironically, although we base our reconstructions on lexical evidence, lexical
reconstruction in many cases is done with less confidence than the reconstruc-
tion of phonology, morphology, and syntax. Consider the case of Algonquian
‘fire-water’ in example (11). There is no doubt that the words for ‘fire’ and ‘water’
are inherited. Given the evidence in (11), we might feel similarly confident about
reconstructing a word ‘fire-water’. But appearances are deceiving. We know that
“fire-water”, i. e., alcohol, was introduced with the arrival of Europeans, long
after Proto-Algonquian was spoken. We must therefore conclude that the words
for ‘fire-water’ were assembled secondarily, from indigenous roots and according
to inherited processes of compound formation. Moreover, we can assume that the
words were not created independently, but that they were diffused through the
Algonquian languages by calquing (for which see Chapter 8.)

(11) Fox Cree Menomini Ojibwa

‘fire’ askotēwi iskotēw eskōtēw iškotē
‘water, liquid’ -āpō- -āpo- -āpō- -āpō-
‘fire-water’ iskotēw-āp-oy eskōtēw-āp-oh i-skotēw-āp-ō

Examples like this show strikingly that in some cases we are more successful
in reconstructing basic morphological elements, such as the roots for ‘fire’ and
 What can we reconstruct and how confident are we of our reconstructions? 413

‘water’, and the morphological patterns according to which they can combine,
than complete, complex words. The best we can do is to establish that Proto-Al-
gonquian had the morphological elements and machinery to assemble a word like
‘fire-water’ – if the occasion had arisen. The problem is that the occasion arose
only much later.
Such problems are not limited to lexical reconstruction. In syntax, too, we are
able to reconstruct syntactic patterns, but reconstructing specific sentences runs
into even greater difficulties than reconstructing complex words. True, we may be
quite certain that a speaker of Proto-Indo-European must have been able to utter
a simple sentence like *pǝtēr (e)gwemt ‘the father came/arrived/went’. But even
for a simple sentence like this there are problems, such as the fact that Indo-Eu-
ropeanists are not in full agreement as to whether we should reconstruct gwemt or
egwemt for the form meaning ‘came/arrived/went’. Moreover, there is the problem
that the same idea may be expressed in more than one way. And we cannot know
that any speaker of Proto-Indo-European ever actually uttered such a simple sen-
tence. For complex sentences, the problems are obviously even greater.
The problem runs even deeper, for as the example of gwemt vs. egwemt illus-
trates, comparative linguists often disagree with each other. Their disagreement
may concern matters of relatively minor detail, such as whether past-tense forms
of the type gwemt should be reconstructed with the “augment” *e- for all of Pro-
to-Indo-European or for only some dialects of the proto-language, or whether the
prefix was introduced in the early stages of some of the daughter languages.
The reason for this disagreement, briefly, is this. Among the ancient Indo-Eu-
ropean languages, the augment is limited to Indo-Iranian, Armenian, Greek, and
a few other, less well attested languages. These languages were close geograph-
ical neighbors. On the other hand, Hittite, Latin, and the other early Indo-Eu-
ropean languages show no clear traces of the augment. What is especially
embarrassing is that Hittite lacks it, since Hittite is attested earlier than either
Sanskrit or Greek (or Latin). Some linguists therefore consider the augment a
regional innovation, either in dialectal Proto-Indo-European (comparable to
the centum : satem phenomena discussed in Chapter 11, § 4) or even later (pre-
sumably as the result of convergent developments; see Chapter 13). Other lin-
guists argue that Hittite, Latin, and other early languages that lack the augment
exhibit other innovations in verbal morphology and that the absence of the
augment in these languages can therefore be considered a similar morphological
innovation.
Even the reconstruction of the Indo-European sound system has been a
matter of controversy and/or change of opinion. In the nineteenth century the
stop system was reconstructed as in (12a), with a neat four-way contrast between
voiceless, voiceless aspirated, voiced, and voiced aspirated, just as it is found
414 Comparative method: Establishing language relationship

in Sanskrit. (See also Chapter 2.) More recently, scholars have argued that the
voiceless aspirated series of Sanskrit (indirectly attested also in Iranian) can be
explained as the result of secondary developments. Occam’s Razor, therefore,
should prevent us from postulating it as a feature of the proto-language. As a
consequence, the system in (12b) was postulated.
More recently yet it has been claimed that the system in (12b) is unnatural.
The most important argument is the claim that no known languages have voiced
aspirates without also having voiceless aspirates. Scholars adhering to this view
reconstruct the system in (12c), with voiceless stops (± aspiration), “glottalized”
stops (accompanied by a glottal-stop element), and voiced stops (± aspiration)
corresponding, respectively, to the voiceless, voiced, and voiced aspirated stops
of (12b).

(12) a. Nineteenth-century reconstruction

Labial Dental Palatal Velar Labiovelar
voiceless p t ḱ k kw
voiceless aspirated ph th ḱh kh kwh
voiced b d ǵ g gw
voiced aspirated bh dh ǵh gh gwh
b. Standard twentieth-century reconstruction
voiceless p t ḱ k kw
voiced b d ǵ g gw
voiced aspirated bh dh ǵh gh gwh
c. “Glottalic” reconstruction
voiceless (± asp.) p(h) t(h) ḱ(h) k(h) kw(h)
glottalized p’ t’ ḱ’ k’ kw’
voiced (± asp.) b(h) d(h) ǵ(h) g(h) gw(h)

The so-called glottalic system in (12c) differs markedly from the ones in (12a)
and (12b) and, if correct, would have enormous consequences for comparative
Indo-European linguistics. The system is virtually identical to the one found in
certain modern Armenian dialects and postulated for early Armenian by the advo-
cates of the “glottalic theory”. This has the virtue that the system is precedented
and therefore can be considered natural. But another consequence is that the
sound shift traditionally postulated for Armenian (see Chapter 4) can no longer be
maintained. Instead, we must assume that Armenian essentially retained the stop
system of Proto-Indo-European. An extension of this argument is that Grimm’s
Law, which, as noted in Chapter 4, is remarkably similar to the traditionally pos-
tulated Armenian sound shift, must likewise be rejected. The Germanic sound
system, then, is claimed to be nearly as archaic as that of Armenian.
 What can we reconstruct and how confident are we of our reconstructions? 415

But there are further consequences. The striking differences between Arme-
nian and Germanic on one hand and the rest of Indo-European on the other must
now be attributed, not to innovations on the part of Armenian and Germanic,
but to sound shifts in the other Indo-European languages. These shifts must be
of similar proportions to the ones traditionally postulated for Armenian and Ger-
manic. Moreover, these shifts would have to be considered independent of each
other. If this assumption is correct, we would have to postulate some ten or twelve
major sound shifts, instead of the two traditionally assumed for Armenian and
Germanic. Such a proliferation of shifts, in turn, could be considered an argument
against the reconstruction in (12c), since it would violate Occam’s Razor.
Moreover, as noted in Chapter 2, the glottalic system found in some of the
modern Armenian dialects may be attributed to convergence with the neighboring
Caucasic languages. In this regard, note that Ossetic, an Iranian language which
likewise is spoken in this region, has a similar glottalic system. But in this case,
the evidence of the other Iranian languages makes it clear that the glottalic system
is an innovation, no doubt the result of convergence with the other languages
of the Caucasus. These facts weaken the arguments for considering the glottalic
system of Armenian to be an archaism.
Finally, it has been observed that some languages do in fact have voiced aspi-
rates without contrasting voiceless aspirates. One area in which such languages
are found is part of the Indonesian archipelago. Members of the West African
group of Kwa languages likewise offer such supposedly impossible sound systems.
The evidence of these languages shows that one of the most important foun-
dations of the glottalic theory cannot be maintained, namely the claim that lan-
guages with voiced aspirates but no contrasting voiceless aspirates are unnatural.
There are thus a number of arguments that weaken the cogency of the glot-
talic theory. Most Indo-Europeanists, therefore, prefer reconstruction (12b) to
(12c); but proponents of the glottalic theory remain convinced that (12c) is a supe-
rior reconstruction.
Such disagreements must appear disconcerting to the non-linguist, and even
to linguists working in other areas of specialization, who are unfamiliar with the
often arcane arguments of comparative linguists. In principle the disagreement
should come as no surprise. All reconstructions basically are hypotheses about
the nature of the proto-language, and by their very nature hypotheses are – well,
hypothetical. They are meant to be tested, not to be taken as truths just by being
posited. True, we try to exclude questionable hypotheses by appealing to such
principles as Occam’s Razor and naturalness. But these are only very general
guidelines. They are not simple algorithms which, if properly applied, will auto-
matically yield correct solutions. They require judgments on the part of compara-
tive linguists. And that is where disagreements can arise.
416 Comparative method: Establishing language relationship

At the same time, we don’t really have any choice; we have to develop
hypotheses, even if they are “hypothetical” and sometimes controversial. If we
really knew what the proto-language was like, we wouldn’t have to do recon-
struction.

8 Language families other than Indo-European

The present chapter, just like much of the rest of this book, so far has concentrated
on Indo-European languages. This is because since their “discovery” in the late
eighteenth century, these languages have received the attention of more compar-
ative and historical linguists than any other language family. In part this reflects
the fact that until relatively recently, most linguists were speakers of Indo-Euro-
pean languages.
But this is not a sufficient explanation. Much of the work on the Semitic lan-
guages and the larger Afro-Asiatic family of which Semitic is a member has also
been done by native speakers of Indo-European languages, and similarly, pio-
neering fundamental research on language families such as Bantu, Malayo-Poly-
nesian, or the languages of the Americas has been conducted by speakers of
Indo-European languages.
What is more important is that most Indo-Europeanists begin with a good
foundation in the classical languages of Greek and Latin. Since these are clearly
Indo-European, it is natural for such scholars to expand their horizon (if they
choose to do so) to other, related, Indo-European languages.
It may, however, also be true that the early Indo-European languages present
something close to the optimal stage for comparison mentioned in § 6 above. This
makes the initial task of establishing genetic relationship, as well as the job of
reconstruction, relatively easy. True, even the Indo-European family includes
members whose earliest attestations come from less optimal times (such as Alba-
nian and Tocharian), or whose written attestations present other difficulties (such
as Hittite). But just as in the case of examples like Skt. čakra-, OE hweogol ‘wheel’
in (10c) above, it is possible to at least begin to unravel the mysteries presented by
such languages, because – in spite of the difficulties outlined in § 7 – we do have
a fairly firm understanding of reconstructed Proto-Indo-European and therefore
are able to draw on that understanding to make hypotheses as to how recalcitrant
forms like čakra- and hweogol or recalcitrant languages like Albanian, Tocharian,
and Hittite may be derived from Proto-Indo-European.
The following presents a brief look at the often more mixed successes of com-
parative linguistics as regards other language families. For selected languages
 Language families other than Indo-European 417

the relationships are illustrated with examples of lexical correspondences. In

some cases, the phonetic similarities in the correspondences are strong enough
to strike even the non-specialist. In others, the similarities are more remote, but
extensive comparative work makes it certain that the forms are cognate. In a few
cases (especially Altaic), sets of highly dissimilar correspondences are included
specifically to illustrate the problems that lead linguists to disagree on whether
the languages in question are genetically related.
The highly controversial issue of whether it is possible to establish longer-
range genetic relationships (as between Indo-European and Semitic or Afro-Asi-
atic) or even a genetic relationship between all of the world’s languages is taken
up in the next chapter.
As we saw in the introduction to this chapter, beside Indo-European there
are members of at least two other language families in Europe. One of these fam-
ilies is the Finno-Ugric group which includes Finnish, Estonian, and Hungar-
ian, as well as a number of other less well-known languages such as Lapp (now
called Saami, found in the northern parts of Norway, Sweden, and Finland), and
Ostyak (also called Hanty, in western Siberia). Finno-Ugric, in turn, is part of a
larger group, called Uralic, which includes Samoyed (in the northern part of the
Russian Republic, east of the Ural mountains). The relationship between these
languages can be illustrated by the sample correspondences below, given here
in traditional transcription. Some of the forms are obviously similar, such as the
words for ‘winter’ in all three languages, or the words for ‘fish’ in Finnish and
Hungarian. For others, the relationship is less obvious, such as the words for ‘one’
and ‘two’; but a closer look yields better results. For instance, the words for ‘one’
all begin in vowel, and the Finnish -k- of this word can be related to the -gy = [ǰ]
in Hungarian; and so forth. In fact, more than a century of comparative work on
Finno-Ugric/Uralic makes it certain that all the correspondences involve genuine
cognates, descended from a common ancestor.

Finnish Hungarian Ostyak

‘one’ yksi egy it, ij
‘two’ kaksi kettő/két katǝn, kăt
‘three’ kolme három xutǝm
‘fish’ kala hal xut
‘heart’ sydän szív sam
‘winter’ talvi tél tatǝ, tał
418 Comparative method: Establishing language relationship

The other non-Indo-European family represented in Table 1 above is Altaic, of

which only one language with a literary tradition is found in Europe, namely
Turkish. Turkish is part of a closely-related subgroup, called Turkic, members of
which are found as far east as Central Asia and Siberia. Other members of the
Altaic group include Mongol, Manchu, and Tunguz. Recent research suggests that
Korean, and perhaps also Japanese, may be related to the Altaic languages. But
both claims, especially the claim that Japanese is related, are controversial.
Even the Altaic group as more traditionally defined is controversial. Some
scholars deny the validity of the “Altaic hypothesis” altogether and claim that
Manchu and Tunguz are related to Korean and possibly to Japanese, but that
there is no relationship between this group and the rest of what traditionally has
been called Altaic. (Perhaps the number of well-established correspondences is
at a level that some scholars are satisfied that it is high enough to establish rela-
tionship, while others feel that it is not.) A comparison of the data below with
those for Uralic readily illustrates how much more remote the Altaic languages
are from each other. (Some of the Tunguz forms are missing, due to insufficient
information.) Similarities are limited to Mong. ǰirin : Tung. ǰū(r) ‘two’, and Tu.
yürek : Mong. dzürx ‘heart’. However, semantically less exact correspondences
such as Turk. balık ‘fish’ : Mong. balgu ‘carp’, and Turk. kıš ‘winter’ : Mong. kul-de,
Tung. kēl-di ‘cold’ can easily be added.

Turkish Mongol Tunguz

‘one’ bir negen umun
‘two’ iki qoyor/ǰirin ǰū(r)
‘three’ üč gurban ilan
‘fish’ balık dzagas
‘heart’ yürek dzürx mēwan
‘winter’ kıš öböl

Some scholars have argued for genetic relationship not just between the Altaic
languages, but even between Uralic and Altaic, pointing to lexical similarities
such as those below. (The glosses on the left in many cases are only approximate.
For instance, the range of meanings for Alt. *al- includes ‘underside’, ‘frontside’,
‘lower part, backside, rump’, and so on.) Some of the correspondences are indeed
quite striking; others, such as *ñele- : *dalag- ‘lick’ are less impressive. Whatever
the merits of such similarities, the Ural-Altaic hypothesis is considered even less
well established than the Altaic one, and therefore even more controversial.
 Language families other than Indo-European 419

Uralic/Finno-Ugric Altaic
‘under, below’ *al- *al-
‘tongue, language’ *kelä *kele
‘we’ *me- *min-
‘what’ *mǝ *mu
‘lick’ *ñele- *dalag-
‘three’ *kolme Mong. gurban

As noted in the introduction to this chapter, in addition to Indo-European, Uralic,

and Altaic, Europe also is host to Basque, which does not seem to belong to any of
the other well-established language families. It may well be that Basque had rela-
tives in prehistoric times that died out with the coming of the Indo-Europeans to
Europe in the second millennium BC, but there are no records of such languages.
Many other language isolates like Basque are found around the world, such as
Sumerian in ancient Mesopotamia and Burushaski in the extreme north of South
Asia. Languages like these present even greater challenges to comparative linguis-
tics than controversial groupings like Altaic.
Language families spoken in the Asian part of Eurasia are discussed below.
Illustrative correspondences are given only for selected language families. In most
cases, these are the words for the numerals ‘one’, ‘two’, and ‘three’, but for some
groups other words are cited.
The Caucasus area is home to a number of Indo-European and Altaic (Turkic)
languages, including Armenian and Ossetic (an Iranian language), as well as
Azeri (Azerbaijani, a Turkic language). In addition, there are three groups of Cau-
casic languages which are commonly considered “unrelated” (i. e., unrelatable)
to any outside languages or language families: the Northwest Caucasic languages
(e. g. Abkhaz); the Northeast Caucasic languages (e. g. Chechen-Ingush and Dag-
estani); and Kartvelian. The best known Kartvelian language, and the one with
the longest literary attestation (since the fifth century AD) is Georgian.
The Caucasic languages are notorious for very rich consonant systems. One
feature, shared by the Indo-European languages Armenian and Ossetic, is the
existence of a series of glottalized stops. (See § 7 above.) Some of the languages,
especially those of Northwest Caucasic, have consonant systems unexcelled in
any other attested human language. Ubykh, for instance, is said to have close to
80 consonants. And some of the same languages have been claimed to have the
lowest vowel inventories. Abkhaz, for example, probably only has two vowels, a
low vowel a and a central vowel ǝ.
In the eastern area we find Sino-Tibetan, a family that includes the Chinese
language family, as well as Tibeto-Burman, of which Tibetan and Burmese are
major members. Chinese has been attested since probably the seventeenth century
420 Comparative method: Establishing language relationship

BC, Tibetan since the eighth century AD, and Burmese from the twelfth century
AD. Although the Sino-Tibetan family is generally considered well established,
reconstructive work has not progressed very far as yet, and many aspects of the
internal subgrouping of Sino-Tibetan are still uncertain. There have been pro-
posals in the past that Thai belongs to Sino-Tibetan, but recent research suggests
that it may rather be distantly related to Austronesian. The following correspond-
ences may illustrate the relationship between Chinese, Tibetan, and Burmese. As
in many other cases, some word sets exhibit much more transparent similarities
than others; compare the words for ‘three’ and ‘I’ vs. the words for ‘two’; but all
forms can be considered cognates. (Chinese forms are from the Middle Chinese
period; the Tibetan and Burmese forms come from the written forms of these lan-
guages.)

Chinese Tibetan Burmese

‘two’ ñžyi- gnyis hnac
‘three’ sam gsum sûm
‘I’ nguo nga ŋa
‘name’ myäng ming ǝ-mañ
‘tree, wood’ syen shing sac

South Asia is home to Dravidian, a family of languages spoken mainly in the

south of India and parts of Sri Lanka. But one member, Brahui, is spoken much
farther north, in present-day Pakistan. The major literary languages are Tamil,
Malayalam, Kannada, and Telugu. The following correspondences, in traditional
transcription, suggests the degree to which the Dravidian languages are related
to each other.

Tamil Malayalam Kannada Telugu Brahui

‘one’ oṉṟu onnu ondu okaṭi asi(ṭ)
‘two’ iraṇḍu raṇḍu eraṛu reṇḍu ira(ṭ)
‘three’ muṉṟu mūnnu mūru mūṛu musi(ṭ)

The Austro-Asiatic language family includes Mon and Khmer in present-day

Kampuchea, the Munda languages in Central and East-Central India, and Khasi
in the northeast of South Asia.
The Malayo-Polynesian or Austronesian languages form a far-flung, but
linguistically close-knit family. They include Malay, Indonesian, Javanese, Tagalog
(in the Philippines), Maori, Hawaiian, Samoan, as well as Malagasy, the language
of Madagascar, just east of Africa. Compare the following correspondences.
 Language families other than Indo-European 421

Indonesian Javanese Tagalog Samoan Malagasy

‘one’ satu siǰi isa tasi isa
‘two’ dua loro dalawa lua rua
‘three’ tiga telung tatlo tolu telu

Afro-Asiatic, as the name suggests, extends from Africa into Asia. The group
includes the Semitic languages (Hebrew, Arabic, as well as Assyrian and Baby-
lonian of ancient Mesopotamia), Ancient Egyptian (and its descendant, Coptic),
as well as Berber (in North Africa), Cushitic (including Somali), and Chadic
(including Hausa). Compare the following correspondences. (Only putatively
related words are given; hence some of the blanks. The hieroglyphic script of
Ancient Egyptian indicates only the consonants, not the vowels.)

Hebrew Arabic Egyptian Berber Cushitic Chadic

‘to beat’ dɔqaq daqqaqa dḳw dǝgdǝg daku dōka
‘bone’ qaṣṣ qs ixs, iγs k’ašī
‘ear, hear’ šmaʕ samiʕa sǰm asim māsuw sim
‘heart’ lɛ̄ v lubb yb ul lēb, nibbo nǝfu
‘mouth’ pe fam emi (y)af po
‘nose, smell’ snsn san sunsunā

In Africa, the following families are recognized, in addition of course to Afro-Asi-

atic.
The most widespread language family is Bantu, ranging from Swahili in Kenya
and Tanzania to Setswana and Zulu in southern Africa. The Bantu languages are
generally considered part of a larger family, Niger-Congo, which includes West
African and sub-Saharan languages like Wolof, Fula, and Yoruba. An even larger
putative language family is Niger-Kordofanian, which adds several languages
of the Sudan to Niger-Congo. Of these different genetic classifications, Bantu is by
far the best established; Niger-Congo is more uncertain; and Niger-Kordofanian
is controversial. The correspondences below are from selected members of the
Niger-Congo family.

Bantu Other

Swahili Lingala Setswana Ahlõ Efik

‘one’ m-oja m-ɔkɔ ŋ̊ ŋwe ili kiet

‘two’ m-bili/wili mi-bale pedi iwa iba
‘three’ tatu mi-sato tharo ita ita
422 Comparative method: Establishing language relationship

In the extreme south of Africa are located the Khoisan languages, famous for their
click sounds, which are indicated by such arcane symbols as ≠k, ≠g, and !k(x).
Formerly these languages were called Bushman and Hottentot; but the names
have been given up because of negative connotations. Two languages of Tanzania,
Sandawe and Hadza, have been claimed to be distant relatives of the Khoisan
languages. The following correspondences have been claimed to establish the
claimed relationship. However, recent research shows that the Khoisan lan-
guages do not constitute a single language family; and the relationship, if any, of
Sandawe and Hadza becomes even more tenuous.

Sandawe Khoisan
Naron Khoi
‘ear, hear’ keke ≠kē ≠gai
‘four’ haka haga haka
‘valley’ Goʔa !xubi !kxowi

It has been argued that the majority of the remaining African languages (includ-
ing Nubian, Sudanic, and Songhai) form a single language family, called Nilo-Sa-
haran. But like many others, this genetic classification is controversial.
The Americas are home to a large variety of indigenous languages. Accord-
ing to some scholars, most of these are related to each other, and there are only
three “super-families” in the Americas. But this view remains highly controver-
sial. A more conservative approach would recognize, among others, the following
groups, but would consider the genetic affiliation of many languages to be still
unsettled.
Eskimo-Aleut is a group of languages extending from Alaska and Northern
Canada to Greenland, of which Eskimo, now often referred to as Inuit, is the best-
known member. As noted in Chapter 9, the term Eskimo originally is a derog-
atory word, apparently derived from Micmac eskameege ‘raw fish eaters’. The
term, however, is still used in some technical writing and by some Indigenous
Alaskans.
The Athabaskan family is named after Athabaskan, spoken in Alaska
and Northwest Canada, but includes many other languages, known for their
rich consonant systems, a large number of glottalized consonants, and highly
complex consonant groups. Navajo, with the largest number of speakers of any
American Indian language in the United States (some 150,000), and Apache, are
also members of the Athabaskan family, though spoken much farther south (in
present-day Arizona and adjacent areas). The Athabaskan family is considered
related to two other groups, the nearly extinct Eyak (Alaska), and the Tlingit
group (Alaska and Northwest Canada). Some linguists argue for a larger family,
 Language families other than Indo-European 423

“Na-Dene”, which also includes Haida (Alaska and British Columbia); but that
affiliation is controversial.
Algonquian is a widespread family of closely related languages, extending
from the Canadian prairie provinces across the Great Lakes area to northeastern
North America, and originally along the eastern seaboard as far south as Virginia.
Well-known members include Blackfoot, Cheyenne, Cree, Chippewa or Ojibwa,
Fox, Menomini, Ottawa, the Illinois Confederation, and Shawnee. The corre-
spondences below may illustrate the relative closeness of the members of this
family; see also the words for ‘fire’, ‘water’, and ‘firewater’ in (11) above. Recon-
struction of the linguistic ancestor, Proto-Algonquian, has made considerable
progress during the twentieth century.

Fox Cree Menomini Ojibwa

‘one’ nekoti nikot- nekot ninkot-
‘two’ nīšwi nīso nīs nīš
‘three’ neswi nisto neʔniw nisswi

Two languages spoken in California, Wiyot and Yurok, have been shown to be
related to Algonquian, but at a much greater distance. The fact that Algonquian
thus has relatives in California raises interesting questions about the earlier distri-
bution of the language family, or about prehistoric migrations in North America.
Iroquoian is a family of languages in the eastern United States and Canada
with members that bear some particularly familiar names from American history,
for it comprises the members of the “Five Nations” confederacy (also known as
the “Iroquois League”): Cayuga, Mohawk, Oneida, Onondaga, and Seneca, along
with Tuscarora, which as a later addition, turned the confederacy into the “Six
Nations”. Other Iroquoian languages are Cherokee (see Chapter 3, § 5.3 for its
writing system), Erie, Huron, and Wyandot. Iroquoian is sometimes classified as
related to Siouan.
Siouan is a very far-flung family, embracing the languages of the Sioux or
Dakotas, as well as Crow, Iowa, Omaha, Osage, Winnebago, and many others. The
family at one time extended as far north as the Dakotas and Central Canada, as far
east as Virginia and the Carolinas, and as far south as the Gulf coast. There have
been attempts to relate Siouan to Hokan, languages spoken in the Southwest of
the United States, which include Mojave, Chumash, and Yuman. But that classifi-
cation is generally doubted; and there are even doubts as to whether all the Hokan
languages are really related to each other or whether their similarities are mainly
attributable to centuries or even millennia of mutual borrowing.
Uto-Aztecan is a large family in the western United States, Mexico, and
Central America, including Nahuatl (the language of the ancient Aztec empire),
424 Comparative method: Establishing language relationship

Hopi (in Arizona), and Ute (in Utah and Colorado). The following correspond-
ences may illustrate the relationship.

Comanche Tübatulabal Luiseño Hopi Papago Nahuatl

‘one’ sǝmǝʔ čīč supúl sǝ̄́ ka hǝmako seem-
‘two’ waha(h)- wō wḗx, wéʔ lȫ́yö-m gōk oomi
‘three’ pahi- pāi- pā́hi pā́yo-m vaik eeyi

Mayan, in Mexico and Central America, is a group of fairly closely related lan-
guages, named after their most well-known member, the language of the ancient
Maya civilization. As observed in Chapter 3, the Mayan civilization developed a
writing system of its own, long before the arrival of the Europeans. The decipher-
ment of the writing system has been increasingly successful in recent years.
Arawakan now is found mainly in northeastern South America, but once
extended into the Caribbean as well.
Quechua, a far-flung family with members in Peru, Ecuador, Bolivia, as well
as in border areas of Argentina, Chile, and Colombia, was the language of the
ancient Inca empire. The modern varieties of Quechua are very closely related to
each other, as can be seen from the following correspondences. Some scholars
have grouped Quechua and Aymara into a larger, “Andean” or “Quechumara”
family. But like most other attempts at establishing larger genetic families in the
Americas, this proposal has remained controversial.

Ancash Junín Cajamarca Amazonas Ecuador

‘two’ iškē iškay iškay iškē iškay
‘three’ kimsa kimsa kimsa kimsa kimsa
‘six’ hoxta suʔta soxta suχta suχta
‘language’ qaλu aλu qažu kadžu kažu

Ayacucho Cuzco
iskay iskay
kimsa kinsa
soxta soxta
kaλu qaλu

Australia, too, is home of a large number of languages. By some estimates, at

least 200 languages were spoken in Australia at the time of the European arrival.
The languages have suffered an enormous degree of language death. About fifty
percent of the original languages are now extinct. Many others are dying. Some
scholars claim that all the indigenous languages of Australia are related. Others
 Language families other than Indo-European 425

class the large majority into a Pama-Nyungan family, distributed over most of
Australia, and assume a certain number of smaller genetic groups for the remain-
ing languages, many of which are found in the northwest. But many details of
these and other proposed genetic classifications still need to be worked out. In
the meantime, Australian languages continue to die at a rapid rate; and with the
languages the evidence dies out that they might contribute to a more complete
understanding of Australian linguistic relationships.
In addition, we can mention various sign(ed) languages and the question
of their genetic affiliation. The number of such manually based languages is gen-
erally assumed to be very large – at least in the hundreds, but possibly in the
thousands. In their natural state (i. e., leaving aside codes such as finger-spelled
versions of spoken languages), true signed languages are unrelated to their “co-ter-
ritorial” oral languages. For instance, American Sign Language (ASL) has nothing
to do with American English, either historically, or structurally, or lexically. The
same holds true for the relationship between British Sign Language (BSL) and
British English, French Sign Language (FSL) and French, and so on. Interestingly,
however, ASL and FSL are related to each other historically, and neither is related
to BSL. FSL originated around 1760 through the efforts of a French teacher to the
deaf, Abbé de l’Épée, and later spread to America (where it became the basis for
ASL), to Russia, to Ireland (from where it spread to Australia), and to several other
European countries, whose sign languages thus are related and form a language
family. Through similar developments, Japanese and Korean Sign Language are
related to each other. BSL and Chinese Sign Language, by contrast, constitute
something like signed counterparts to oral language isolates such as Basque.
Our knowledge of relatedness among signed languages is partly based on
what is known about their historical spread, but also on applying the standard
methods of comparative linguistics – by comparing systematic similarities and
differences in hand shapes, hand orientation, and hand movements for particular
signs, in the meanings associated with these signs, and in the morphology and
syntax of signed languages. Thus, just as examples in the earlier chapters have
shown that sign languages are affected by the same kinds of linguistic change
that are observable in oral languages, so also it is true that the principles of com-
parative linguistics apply equally well to signed languages as to oral languages.
Chapter 17: Proto-World?
The question of long-distance
genetic relationships
And the Lord said, Behold, the people is one, and they have all one language; and this they
begin to do: and now nothing will be restrained from them, which they have imagined to
do. Go to, let us go down, and there confound their language, that they may not understand
one another’s speech. So the Lord scattered them abroad from thence upon the face of all
the earth: and they left off to build the city. (Genesis 11: 6–8)

1 Introduction
The question whether all of the world’s languages are related, i. e., whether we
can establish a “Proto-World” from which all human languages are descended,
has intrigued humankind for centuries, even millennia. Perhaps the earliest,
and certainly the most famous testimony to this interest in the western world is
the story about the tower of Babel (cited above, from the “Authorized Version” of
the English Bible translation). Similar stories are told in other parts of the world,
including in many indigenous languages of the western United States. Consider
the following two examples.

Mouse was sitting on top of the assembly house, playing his flute and dropping pieces of
coal through the smokehole, when Coyote interrupted him. Those who sat closest to the
smokehole received fire and therefore cook their food and speak correctly. Those farther
removed did not receive fire and remained in the cold; that is why their teeth chatter when
they talk. If Coyote had not interrupted Mouse, all people would have received the fire and
would have spoken in one language. (Adapted from a tale by the Maidu of California.)

When the people emerged from the lower world to the upper world through the sipapuni,
Mocking Bird stood beside Old Spider Woman and assigned them to different groups. “You
will be Hopi and speak Hopi,” he said to one group. “You will be Navajo and speak Navajo,”
he said to another group. In this way he assigned everyone to a tribe and language – the
Hopis, the Navajos, the Apache, the Paiutes, the Zunis, and so on, down to the whites.
(Adapted from a myth of the Hopi of Arizona.)

Historical linguists have long considered it impossible at the present state of our
knowledge to establish that all the world’s languages are genetically related.
More than that, many doubt whether it is possible to establish relationships more
distant than, say, Indo-European, Uralic, Bantu, or Algonquian. As we have seen
in the preceding chapter, other, relatively well-established, language families
such as Altaic still remain controversial.

https://doi.org/10.1515/9783110613285-017
 Introduction 427

At the same time, since nearly the beginning of comparative linguistics,

there have been attempts to establish longer-distance relationships. And time
and again, traditional historical linguists have claimed that these attempts are
premature or poorly established. For instance, the Ural-Altaic hypothesis (see
Chapter 16, § 8) has been doubted, since it was proposed mainly on the basis of
general structural similarities, features which could easily be attributed to con-
tact-induced convergence. Vocabulary correspondences, on the other hand, are
quite limited, too limited to convince scholars to exclude the possibility that they
are borrowings, onomatopoeia, nursery words, or even due to accident. Adopting
an expression used in second-language teaching, let us refer to such non-cognate
similarities as false friends.
The situation is similar for the nineteenth-century proposal of an Indo-
Semitic family, consisting of Indo-European and Semitic. An additional problem
in this case is that Semitic belongs to a larger family, now referred to as Afro-Asi-
atic (see Chapter 16, § 8). The comparison, therefore should be of Indo-European
with Afro-Asiatic, not just with Semitic. More recent work meets this objection
by comparing Indo-European with Afro-Asiatic and proposing a larger family
which one scholar refers to as Lislakh (lit., ‘language of the people’, from lisan,
the Arabic word for ‘tongue, language’, and *laHHuwa- ‘people’, a form recon-
structed by one scholar for the ancestor of Hittite and the other Indo-European
languages).
Probably the largest language family proposed by scholars attempting to
establish relationship by traditional methods is Nostratic, a family consisting
of Indo-European plus a range of other language families which differs from
researcher to researcher and tends to include such language groups as Semitic
or Afro-Asiatic, Uralic, Altaic, Caucasic languages, and/or languages like ancient
Sumerian or Basque which most traditional scholars consider to be without
known relatives (“isolates”, see Chapter 16, § 8). The term “Nostratic” is somewhat
ethnocentric, meaning ‘belonging to us/to our part of the world’, i. e., to Eurasia
plus Northern Africa. Depending on one’s views, this may be considered exciting
or deplorable. What is more significant is that even more than such shorter-dis-
tance groupings as Ural-Altaic or Lislakh, the Nostratic hypothesis has failed to
convince most comparative linguists because of concerns that the similarities in
vocabulary and structure may be of the false-friend type.
Most recently, the American linguist Joseph Greenberg and some of his asso-
ciates have claimed that long-distance relationships can be established more
effectively – and more easily – by employing an approach totally different from
the traditional methods. This is an approach called lexical “mass comparison” or
“multilateral comparison” which establishes relationship through comparison of
a large number of languages, on a fairly limited lexical basis (including function
428 Proto-World? The question of long-distance genetic relationships

words and affixes), by means of scoring cognates based simply on similarities in

sound and meaning.
Greenberg used this method to posit a genetic classification of the languages
of Africa and, more recently, of the indigenous languages of the Americas. He
argued that most of the latter languages can be grouped together into a single
family, called Amerind.
Greenberg further claimed that his approach makes it possible to establish
a relationship between Amerind on one hand and, on the other hand, Eski-
mo-Aleut, and four linguistic families of the “Old World”: Afro-Asiatic, Indo-Eu-
ropean, Uralic, and Dravidian. In principle, he claimed, his method makes it pos-
sible to show that all of the world’s languages are related. Some “Nostraticists”
have made similar claims recently.
Not unexpectedly, Greenberg’s claims and similar claims of certain Nostrat-
icists have created a lot of excitement among non-linguists, since they promise,
at least in principle, to provide a positive answer to the question raised at the
beginning of this section: “Are all the world’s languages related?” By contrast,
traditional comparative linguists have subjected Greenberg’s methodology and
claims to close scrutiny – and severe criticism.
The debate that ensued between Greenberg and his critics has at times been
highly acrimonious. Greenberg has been accused of having predetermined the
outcome of his work on Amerind before actually applying his methodology. A
close associate of Greenberg has suggested that opposition to Greenberg’s meth-
odology is based either on indolence or, worse, on a Eurocentric refusal to even
entertain the idea that the languages of Europe might be related to non-Euro-
pean languages. Despairing of finding acceptance among traditional historical
linguists, advocates of such long-distance genetic comparisons have taken to the
semi-popular press, especially Scientific American. Even popular journals, such
as US World and News Report and Der Spiegel, have carried features on their work.
In the following let us take a closer look at the kinds of arguments and evi-
dence that are relevant in discussions of this type.
Our focus is on oral languages, since they are the topic of the controversy
between Greenberg (and his followers), the Nostraticists, and what may be called
the Traditionalists. But as noted at the end of Chapter 16, there are numerous
signed languages throughout the world, some related to each other but, crucially,
none connected to or derived from any oral language. Whatever may emerge from
the “Proto-World” controversy, therefore, affects the genetic relationship of oral
languages – sign languages seem to require the assumption of different origins.
(See also § 5 below.)
 Longer-distance comparison 429

2 Longer-distance comparison
For language families like Indo-European and Uralic, the evidence that the
members of each respective family are indeed related amongst each other is so
strong and overwhelming that to doubt the relationship would be bordering on the
perverse. Most important, in both of these families it has been possible to success-
fully reconstruct the phonology, much of the morphology, and the basic outlines
of the syntax. Controversies (such as over the Indo-European “glottalic theory”)
are concerned with details of the reconstruction or with the precise nature of the
reconstruction (phonetic or otherwise), not with the overall results of comparative
reconstruction or the question of whether it is possible to reconstruct at all, and
certainly not with the question of whether the languages are related.
For many other putative language groups, the evidence is not sufficient to
conduct comparative reconstruction and thus to establish genetic relationship. In
many cases the evidence is so limited that most linguists would consider hypoth-
eses of genetic relationship extremely dubious.
In some cases, the evidence is at least strong enough to establish a shared
idiosyncrasy, or the number of recurrent lexical similarities is massive enough to
make genetic relationship very likely, even if it cannot be established beyond a
reasonable doubt. A handy term for a group of this sort is phylum. The term was
originally introduced in comparative work on Indigenous American languages
to refer to “super-families” consisting of putatively related languages. But it is
useful for designating any group of languages for which there is tantalizing – but
insufficient – evidence of relationship.
A case in point is the question of the possible relationship between Uralic and
Dravidian. Earlier comparisons of the modern members of these language fami-
lies suggested relationship to some scholars, while others remained skeptical. The
completion of etymological dictionaries and the working out of (approximate)
reconstructions for each of the two families have pushed back our knowledge
of each family by several thousand years – by essentially undoing some of the
effects of linguistic change. Comparison of the resulting reconstructed forms of
Uralic and Dravidian has made it easier to establish similarities; and many sys-
tematically recurring correspondences in non-technical, basic vocabulary have
been uncovered. Compare the selected examples in (1).

(1) Proto-Uralic Proto-Dravidian

*tag- *ta(-r)- ‘give’
*täm- *tev- ‘fill’
*tug- *tur̤ - ‘river’
*käte- *kay- ‘hand’
430 Proto-World? The question of long-distance genetic relationships

*kele- *kēḷ- ‘speech’

*sükese- *čuk(k)- ‘autumn’
*pekse- *pak- ‘arrow’

In addition we find systematically recurrent phonetic correspondences in the

shared syntactic idiosyncrasy of expressing negation by a finite verb; see (2). For
an illustration of how such negative verbs operate, compare the Finnish example
in (3), in which it is the negation which is inflected for person and number, while
the accompanying main verb appears in an uninflected form.

(2) Ural. *äl- : Drav. *al(l)-

(3) sg. 1 juo-n ‘drink’ e-n juo ‘do not drink’

2 juo-t e-t juo
3 juo ei juo
pl. 1 juo-mme e-mme juo
2 juo-tte e-tte juo
3 juo-vat ei-vät juo

Now, we do find “negative verbs” elsewhere, such as in a number of Australian

languages; and the modern Tsakonian dialect of Greek developed one through
regular sound changes affecting combinations of a negative marker and an obliga-
tory auxiliary verb. Marathi, an Indo-Aryan neighbor of Dravidian, has developed
something like a negative verb, too, presumably by convergence with Dravidian.
However, as usual in convergence situations, the actual form of this negative
verb consists of inherited, Indo-Aryan elements (compare the present stem nāhī
≈ Hindi nahī̃ ‘not’). It is thus quite different and exhibits no systematic recurrent
correspondences with either Dravidian or Uralic. The correspondence in (2), by
contrast, does.
Still, there must be some reservations concerning the value of the shared idi-
osyncrasy in (2). First, the elements involved are very short. This raises the possi-
bility that we might be dealing with a case of chance similarity. The longer, more
complex, and more idiosyncratic putative cognates are, the less the likelihood
that their similarities are due to chance; but in short elements, consisting of just
one or two sounds, the chance of accidental similarity is much greater. Second,
Dravidian al(l)- originates as a negative form of the verb ā(k)- ‘to be’ and its use
as negative auxiliary may well be secondary. Third, there is an apparently older
way of forming negative verbs in Dravidian, partly by suffixation and partly by
vowel change in the root. As in many other cases, then, closer examination raises
important questions about claims of linguistic relationship.
 Longer-distance comparison 431

In the case of many other languages, the evidence is even more limited, such
that the encountered similarities could well be the result of borrowing or conver-
gence, or might even reflect simple chance.
For instance, there are a number of intriguing correspondences between
Uralic and Proto-Indo-European, including pronominal forms, as well as verb
and noun endings. These similarities include some twenty lexical items such as
the words for ‘name’ and ‘water’ (4a), demonstrative and personal pronouns (4b),
and even inflectional endings (4c).

(4) Uralic Proto-Indo-European

a. *nime- *Hnomen- ‘name’
*wete *wodor/weden/uden- ‘water’
b. *tä-
*te/o- ‘this, that’
*to-
*se- *se/o- ‘this, that’
*ke-, ku- *kwe/o- ‘who, what’
c. *-m *-m (Accusative singular)
*-n *-om (Genitive plural)

However, the number of correspondences, proposed by one of the most ardent

proponents of the Indo-Uralic hypothesis, is highly limited. Some of them are
controversial. For instance, one Uralicist has doubted the reconstruction of the
accusative singular and genitive plural endings in (4c). No shared idiosynrasies
involving systematically recurring correspondences seem to exist. And given the
extremely limited number of correspondences, the possibility cannot be excluded
that the similarities are false friends.
As we saw in example (1) of Chapter 16, chance similarities are by no means
rare. And example (10) of the same chapter, repeated as (5) below, shows that
given enough time of separation and independent development, clearly related
languages can come to be different enough that many of their genuine cognates
are difficult or even impossible to recognize.

(5) Sanskrit Old English Hindi Modern English

a. asti is hai is
sa sē vō he
vayam wē ham we
svasar sweostor bahan sister
śvaśrū sweogor sās mother-in-law
vēda wāt ǰāntā hai (he) knows
432 Proto-World? The question of long-distance genetic relationships

b. dvā(u) twā dō two

trayas †rī tīn three
pād- fōt pā̃v- foot
c. čakra- hweogol čakkā wheel
śṛŋga- horn sī̃g horn
bhavati bēon hōnā be
ṣaṭ seox chah six
pluśi- flēah pissū flea

Let us pursue this issue a little further by taking a closer look at the relationship
between Modern Hindi and English – pretending that we do not know that they
are related, and trying to establish their relationship by vocabulary comparison.
This is actually more difficult than it appears. It is all too easy to be influenced by
one’s knowledge of the historical relationship between the languages and there-
fore to notice the genuine cognates, or even to underestimate the effects of lin-
guistic change on the recognizability of genuine cognates.
An open-ended search of Modern Hindi and English dictionaries yields some
55 genuine cognates which are still close enough phonetically and semantically
to look like they are related; see the selected examples in (6a). There are some 30
further genuine cognates which probably would be unrecognizable without one’s
knowing their historical antecedents; see (6b) for examples. In fact, without that
knowledge, one might well feel that if forms like these can be related, then any-
thing can. In addition, there are at least 45 Hindi borrowings from Sanskrit which
have English cognates but which – since they are borrowings – are not strictly
speaking cognates between Hindi and English. See (6c) for a few examples. Some
of these, such as the first four items under (6c) are difficult to recognize, just like
the words in (6b). Words of the type (6c) probably should be disregarded since on
one hand they are borrowed from Sanskrit but, on the other hand, their Sanskrit
sources are cognates of their English counterparts. These words, thus, are simul-
taneously cognate and non-cognate. Some 5 correspondences involve borrowings
from Persian into Hindi (6d), and another 10 or more other borrowings not directly
involving Hindi and English; see (6e) for selected examples. Finally, and most
significantly, there are some 60 correspondences which, given our knowledge of
the history of these languages, clearly are accidental similarities; see the selected
examples in (6 f).
 Longer-distance comparison 433

(6) Hindi English Hindi English

a. bā̃dh-nā bind bhāī brother
bhaũ brow āṭh eight
pā̃č five pā̃v, pā̃y foot
čār four māī mother
dā̃t tooth dō two
kaun who
b. hō-nā be ǰān-nā can, know
gayā ‘went’ come harā green
sī̃g horn sun-nā listen
ǰī-nā ‘live’ quick sūkhā ‘dry’ sear
thā ‘was’ stand ǰuā yoke
c. hanu chin švān ‘dog’ hound
hiraṇya gold vāñčhā wish
pitā father ant end
mānav man nābhī navel
tanu thin vidhavā widow
d. band ‘string’ bond sar(a)d cold
dar(vāzā) door nāf navel
gar(a)m warm
e. Hindi English
kēndra (from Greek via Sanskrit) center (from Greek via Latin)
mistrī ‘craftsman’ (from Portuguese) master (from Latin)
path path (from Iranian)
vāk (from Sanskrit) voice (from French)
f. Hindi English Hindi English
gannā cane gāṛī car, cart
čakhnā ‘taste’ to check čabānā chew
kōylā coal kōnā corner
gin-nā to count kāṭ-nā cut
dūn ‘valley’ downs gati gait
ghās grass kēš hair
xudā/khudā God kuttā ‘dog’ hound
lōčan ‘eye’ look tōṛ-nā ‘break’ tear
434 Proto-World? The question of long-distance genetic relationships

Disregarding Sanskrit borrowings of the type (6c) which, as noted, are simultane-
ously cognate and non-cognate, we find that the ratio of cognates that are both
genuine and recognizable (6a) to false friends (6d–f) is about 55 to 75. Even if
we add the difficult type (6b) to the genuine cognates we wind up with a ratio of
about 85 to 75. That is, no matter what we do, there is only about a 50 : 50 chance
that correspondences are genuine cognates.
The situation is very much the same when we look at similarities in pronouns,
function words, and grammatical suffixes. Compare the data in (7). Here again
we find that in languages that are clearly related, the ratio of genuine cognates to
false friends is not much better than 50 : 50.

(7) Hindi English

hū-ṁ a-m (directly or indirectly from PIE *-mi)
maiṁ me (directly or indirectly from PIE *me)
kaun/kyā who/what (related, from PIE *kwe/o/i-, but hard to
­recognize as cognates)
aur [ɔ̄ r] or (false friends; Hindi aur < Skt. aparam ‘other;
‘and’ moreover’; Engl. or < OE oþðe)
ǰā-nā go-ing (false friends; the cognate of the Hindi suffix
(infinitive) (gerund) has been replaced by -ing in English)

Given these fairly dismal results for languages that we know to be related, the
small number of correspondences between Indo-European and Uralic must be
considered too limited to be persuasive. Much more massive evidence would be
required for traditional comparativists even to entertain the possibility of genetic
relationship.
The question, then, must be “How massive is ‘massive’?” Clearly, one corre-
spondence is not enough; nor are twenty. And just as clearly, a thousand corre-
spondences with systematic recurrences of phonetic similarities and differences
would be fairly persuasive. Are 500 enough, then? And if not, are 501 sufficient?
Nobody can give a satisfactory answer to these questions. And this is no doubt
the reason that linguists may disagree over whether a particular proposed genetic
relationship is sufficiently supported or not.
Traditional historical linguists would believe that ultimately the question is
irrelevant, since genetic relationship is safely established only through recon-
struction, not just through simple vocabulary comparison. After all, they would
argue, only reconstruction gives us the ability to state with confidence that the
similarities in (6c-e) are false friends.
Moreover, traditional comparativists feel that in order to be successful,
reconstruction has to be based on a certain minimal amount of lexical evidence.
 Longer-distance comparison 435

Without such evidence it becomes difficult, if not impossible, to decide on a par-

ticular reconstruction, rather than another, a priori equally possible one.
For example, there was a recent attempt to show that Dravidian is related to
Elamite, the language of the rulers who preceded the Old Persian king-emperors
(6th–4th c. BC). The corpus of Elamite texts is very small, providing an extremely
limited basis for comparison. Still, the Elamo-Dravidian hypothesis at first
looked promising, since it went beyond vocabulary comparison and proposed a
reconstruction. Many traditional comparativists, however, felt that the evidence
of Elamite is too limited to permit reconstructing a sufficiently large number of
lexical items to establish the case beyond a reasonable doubt. Under the circum-
stances, it would be difficult to be certain that the reconstructions – and the pos-
tulated changes relating the reconstructed forms to their attested descendants –
are not ad hoc. They might work just for the limited number of lexical items on
which the reconstruction is based; but there is no guarantee that they would work
if new data were to be added. Only a very large amount of data would make it
possible to test the basic validity of the reconstruction.
Critics soon felt that their skepticism was confirmed, because the author of
the hypothesis proposed another, radically different reconstruction, on the basis
of the same limited data and without being able to convince other linguists that
this reconstruction was any better than the first. If two radically different recon-
structions can be proposed without any persuasive arguments for choosing one
over the other, then there must be something wrong with the entire approach.
Traditional historical linguists are equally skeptical about recent attempts
at establishing a Nostratic family, even though these attempts, too, go beyond
lexical comparison and propose reconstructions. Again, they feel that the lexical
evidence is too limited to be certain that the reconstructions are on the right
track.
The traditionalists’ skepticism is confirmed by the fact that Nostraticists dis-
agree amongst each other over whether particular words should be considered
genetically related, or rather ancient borrowings.
This is the case for relationships such as Altaic *poku- (or the like) ‘ox, bovine
animal’, inferrable from Monguor fuγuor, Mong. hüker, Turk. öküz, and possibly
related to Ainu peko, Jap. beko vs. PIE *peḱu- ‘cattle, property’. Some Nostrati-
cists believe that the words are related, descended from Proto-Nostratic, others
consider the words of the Altaic and other non-Indo-European languages to be
borrowings from Indo-European. Yet others claim that the Indo-European word is
borrowed from Semitic. (Note incidentally that Turk. öküz : Engl. ox are a perfect
example of chance similarity.)
Skeptics find further grounds for doubt in reconstructions such as *kuyon or
*küyna ‘dog, wolf’, *mad/med- ‘honey, mead’, or *ʔam(m)/ʔem(m) ‘mother’.
436 Proto-World? The question of long-distance genetic relationships

Cases like *ʔam(m)/ʔem(m) ‘mother’ may be nursery words, comparable to

mama, amma ‘mother’ in many modern languages. (See Chapter 16, § 2.)
The type *mad/med- ‘honey, mead’ can be suspected to belong to the cat-
egory of words for cultural artifacts or techniques which are notoriously easily
diffused through borrowing. Compare Chapter 16, § 7 on Algonquian ‘fire-water’,
and Chapter 18, § 4.2 on the word for ‘hemp’. See also Chapter 8, § 1, for the wide
diffusion of words for ‘sugar’ and ‘candy’ from India via Iran and the Arab world
to the west.
Some linguists have also questioned the value of the reconstruction *kuyon or
*küyna ‘dog, wolf’ since, they argue, dogs are not descended from wolves but are
related to jackals. Proponents of the Nostratic hypothesis can counter by pointing
to Southern Arabic dialects which use the same word for ‘dog’ and ‘wolf’. (Actu-
ally, dogs are now considered to be descended from wolves.)
What seems to have escaped both sides of this last issue is that the phonetic
similarities in words such as PIE *ḱuwon- which form the basis for the recon-
struction can be attributed to onomatopoeia. As noted in Chapter 7, § 5.3, in a
number of Indo-European languages, the original word for ‘dog’ was replaced
by words with initial ku-, such as Sanskrit kurkura-, literally no doubt ‘the one
that snarls, growls, or barks, i. e., makes the sound [kurkur]’, Mod. Hindi kuttā,
Modern English cur. Similar developments are found outside Indo-European, as
in Tamil kurai ‘to bark’ : kūraṇ ‘dog’.
Strictly speaking this does not prove that the reconstruction *kuyon or
*küyna is impossible. However, the fact that ku-initial words for ‘dog’ can arise
independently, presumably as onomatopoeia, casts doubt on the cogency of the
reconstruction.
It is possible that as linguists keep adding to the evidence in favor of longer-
range groupings such as Indo-Uralic or Nostratic, they will eventually reach the
point where the majority of comparativists will accept the relationship between
these languages. At this point the evidence still is too limited. If the languages in
question are related, their relationship may be too remote, too far removed from
that optimal stage for comparison mentioned in Chapter 16, § 6, to be establisha-
ble beyond a reasonable doubt.

3 Are there any unrelated languages?

In the case of many other languages, traditional comparative linguists feel that
there is no probative evidence whatever for outside genetic relationships. Such
languages include Basque along the western part of the French/Spanish border,
 Lexical mass comparison: Can it establish “Proto-World”? 437

Sumerian in ancient Mesopotamia, and Burushaski in Northwest India. As seen

in Chapter 16, § 8, languages of this sort are often called isolates. Even more
commonly they are referred to as “unrelated” to any other languages.
Strictly speaking, the term “unrelated” is not accurate – the word “unrelata-
ble” would be a much better description. Except perhaps in the case of trying to
relate a signed language with an oral language, we can never prove that two given
languages are not related. It is always conceivable that they are related, but that
the relationship is of such an ancient date that it has been completely obscured
by millennia of divergent linguistic changes.
Ultimately, this issue is tied up with the question of whether there was a
single or a multiple origin of Language, writ large (see § 5 below). According to
traditional comparativists, this question can be answered only in terms of unver-
ifiable speculations, given the fact that even with the added time depth provided
by reconstruction, our knowledge of the history of human languages does not
extend much beyond ca. 5000 BC, a small “slice” indeed out of the long prehis-
tory of language. And as we have seen in (5) above, even shorter time-spans can
obscure the evidence for genetic relationship.

4 L exical mass comparison: Can it establish

“Proto-World”?
As noted in § 1 above, the American linguist Greenberg claims that the difficulties
that beset traditional comparative linguistics in trying to establish longer-range
genetic relationships can be overcome by the method of lexical “mass compari-
son”. Under this approach, genetic relationship can be established through com-
parison of a large number of languages, on a fairly limited lexical basis (about
200 to 250 lexical notions), simply by means of scoring cognates based on simi-
larities in sound and meaning. While his work has been mainly concerned with
establishing long-distance genetic groupings such as a proposed Amerind family,
Greenberg claims that his method in principle makes it possible to show that all
(or most) of the world’s languages are related.
Critics have argued that the ease with which chance similarities of the type
(6 f) can be established casts doubt on Greenberg’s method of determining cog-
nation merely on the basis of phonetic and semantic similarities. They argue that
false friends can only be eliminated by traditional comparative methodology
which, as noted earlier, makes it possible to distinguish between true cognates
and false friends.
438 Proto-World? The question of long-distance genetic relationships

Greenberg has replied to such criticism by claiming that his method yields
correct insights in spite of these difficulties because it operates as mass or multi-
lateral comparison, comparing a large number of languages at the same time:
“The method of multilateral comparison is so powerful that it will give reliable
results even with the poorest of materials. Incorrect material should have merely a
randomizing effect.” Put differently, under mass comparison, errors, he claimed,
will cancel each other out.
Greenberg’s claim has recently been subjected to a rigorous statistical test on
the basis of randomly generated lists of artificial vocabulary items. The test sug-
gests that rather than reducing the possibility of chance similarities, an increase
in the number of compared languages actually leads to an increase in the chance
of accidental resemblances.
Similar conclusions have been reached in an empirical test of Greenberg’s
methodology, using a limited word list, when applied to Hindi, English, and
Finnish – languages whose earlier history is known to us. As in the open-ended
comparison of Hindi and English presented in § 2 above, the method produced not
only genuine cognates, but also false friends. In fact, the ratio between genuine
cognates and false friends is nearly 1 : 2 (no better than the results of the open-
ended Hindi/English comparison reported in § 2 above). That is, even for clearly
related languages, the method has less than a 50 : 50 chance of yielding correct
results. Application of the method also suggests relationship between Hindi and
English on one hand, and Finnish on the other. However, except for one corre-
spondence, all of these similarities are false friends. Finally, expanding the basis
by including data from German and Marathi fails to reduce the ratio of false friends.
Given this evidence, Greenberg’s methodology of mass comparison must be
considered to be of dubious reliability.
Greenberg and his associates, especially Merritt Ruhlen, find further support
for his approach by proposing individual etymologies which they believe show
that many, perhaps most, languages of the world are related. One of these etymol-
ogies is *tik ‘finger’, with alleged reflexes in fifteen language families. However,
numerous empirical and methodological difficulties have been pointed out for this
etymology, suggesting again that the similarities are simply due to chance. Given
that the root consists of only three sounds, the possibility of chance similarity
should not be surprising. Traditional historical linguists have always argued that
chance similarities are more likely to occur in short words than in longer ones.
Most of Greenberg and Ruhlen’s other long-range etymologies suffer from the
same difficulty, of being overly short. There is, however, one exception. This is
the etymology *maliq’a ‘throat, swallow’ which Greenberg finds attested in his
postulated Amerind family, in Eskimo-Aleut, and in four linguistic families of the
“Old World” – Afro-Asiatic, Indo-European, Uralic, and Dravidian. See Table 1.
 Lexical mass comparison: Can it establish “Proto-World”? 439

Tab. 1: Greenberg & Ruhlen’s *maliq’a ‘swallow, throat’

Language family Language Form Meaning

a. Afro-Asiatic PAfr.-As. *mlg ‘to suck, breast, udder’

b. Arabic mlǰ ‘to suck the breast’
c. Old Egyptian mndy ‘woman’s breast, udder’
d. Indo-European PIE *melg- ‘to milk’
e. English milk ‘to milk, milk’
f. Latin mulg-ēre ‘to milk’
g. Uralic P-Finno-Ugric *mälke ‘breast’
h. Saami mielga ‘breast’
i. Hungarian mell ‘breast’
j. Dravidian Tamil melku ‘to chew’
k. Malayalam melluka ‘to chew’
l. Kurux melkhā ‘throat’
m. Eskimo-Aleut Central Yupik melug- ‘to suck’
n. Amerind Proto-Amerind *maliq’a ‘to swallow, throat’
o. Almosan Halkomelem mǝlqw ‘throat’
p. Kwakwala m’lXw-’id ‘chew food for the baby’
q. Kutenai u’mqolh ‘to swallow’
r. Penutian Chinook mlqw-tan ‘cheek’
s. Takelma mülk’ ‘to swallow’
t. Tfaltik milq ‘to swallow’
u. Mixe amu’ul ‘to suck’
v. Hokan Mohave malyaqé ‘throat’
w. Walapai malqi’ ‘throat, neck’
x. Akwa’ala milqi ‘neck’
y. Chibchan Cuna murki- ‘to swallow’
z. Andean Quechua malq’a ‘throat’
α. Aymara malyq’a ‘to swallow, throat’
β. Macro-Tucan. Iranshe moke’i ‘neck’
γ. Equatorial Guamo mirko ‘to drink’
δ. Macro-Carib Surinam e’mōkï ‘to swallow’
ε. Faai mekeli ‘nape of the neck’
ζ. Kaliana imukulali ‘throat’

(The organization of Table 1 closely follows that of Greenberg and Ruhlen; the
major difference consists of the addition of identifying letters in the left margin
for easier cross-reference.)
Unlike etymologies such as *tik ‘finger’, this is a “robust” etymology, con-
sisting of three syllables, and including three consonants and three vowels. Intu-
itively, this robustness seems to support Greenberg and Ruhlen’s assertion that
the “probability for a random similarity among [the] six families [examined]” is
440 Proto-World? The question of long-distance genetic relationships

“about one chance in 10 billion”, followed by the remark, “So much for accidental
resemblances.”
As it turns out, the etymology is not at all as robust as it appears on first sight.
First, Greenberg permits a fair amount of latitude in the phonetic correspond-
ences, including metathesis (in items u., q., ε., and ζ. of Table 1) and loss of one or
another root consonant (e. g. β., δ.). Moreover, vowel correspondences are ignored
altogether. (In this context note the epigraph at the beginning of Chapter 4.) Sim-
ilarly there are considerable variations in the semantics, including ‘swallow,
throat’, ‘suck’, ‘chew’, ‘milk’, ‘breast’, and ‘neck’.
Now, most of the developments that might be responsible for these varia-
tions are quite natural, or at least not unusual. In fact, given enough time, such
variations in form and meaning are not only possible; they are to be expected.
Compare example (5c) above for the phonetic divergences between Modern Hindi
and English as compared to the earlier, Sanskrit and Old English stages.
The real problem lies first of all in the fact that some of Greenberg’s data are
suspect. For instance, the Dravidian (Tamil) melku most likely consists of a root
mel-, actually attested in the same meaning, plus a suffix -ku-. The etymology of
the Indo-European words for ‘milk’ is controversial; but most Indo-Europeanists
prefer derivation from a root *melǵ- ‘stroke, wipe’, attested in this meaning in San-
skrit mṛǰ-, with a semantic development comparable to that found in Latv. slaukt
‘to milk’ vs. Lith. šliaukti ‘sweep’. The Finno-Ugric words for ‘breast’ ordinarily
refer to the chest or forepart of animals, not to women’s breasts.
Even more important, given enough phonetic and semantic leeway – which
we should expect, given the great time-depth at which the languages must be
related (if they are related) – it is amazingly easy to find alternative candidates as
descendants of *maliq’a ‘throat, swallow’. See the examples in (8), with in each
case an indication (i) of the phonetic developments and (ii) the semantic changes
that would putatively relate a given word to *maliq’a ‘throat, swallow’.

(8) a. Afro-Asiatic
Egypt. ʕm ‘swallow’, Kushitic am ‘eat, devour’, Somali ʕon/ʕun ‘eat’
(i) l > Ø; metathesis of stop and nasal; q’ > ʕ
(ii) No significant changes
Arab. qmm ‘devour’, Kushitic qam (etc.)
(i) l > Ø; metathesis of stop and nasal; Arabic dou-
bling of m (?)
(ii) No significant changes
Sem. lqq ‘lick’, Egypt. Demotic lkh ‘lick’, etc., Kushitic lanqi (etc.) ‘tongue’
(i) m > Ø, in a form like *mliq’a; stop doubling
(ii) ‘swallow’ → ‘suck’ → ‘lick’ → ‘tongue’
 Lexical mass comparison: Can it establish “Proto-World”? 441

b. Indo-European:
*leiǵh- ‘lick’ (i) m > Ø, in a form like *mliq’a
(ii) ‘swallow’ → ‘suck’ → ‘lick’
*melH- ‘grind’ (i) q’ > “laryngeal” H
(ii) ‘swallow’ → ‘chew’ → ‘grind’
*gel- ‘throat, (i) m > Ø, in a form like *mliq’a; metathesis of l and
swallow’ stop
(ii) No change
*gwer- ‘throat, (i) Similar to preceding; but velar > labiovelar (as in
swallow’ items o. and r. of Table 1; and l > r as in item y. of Table 1)
(ii) No change
c. Uralic:
*ñele ‘swallow’ (i) Palatalization of m; loss of stop
(ii) No change
*ñole ‘lick’ (i) Palatalization of m; loss of stop
(ii) ‘swallow’ → ‘suck’ → ‘lick’
*ñɤkkœ ‘neck’ (i) Palatalization of m; assimilation of l
(ii) see item β. in Table 1.
*kelä ‘tongue, (i) m > Ø; metathesis of stop and l
language’ (ii) ‘swallow’ → ‘suck’ → ‘lick’ → ‘tongue’ etc.

d. Dravidian
mir̤ uŋku ‘swallow’ (i) l > r̤ ; “prenasalization” of k, common in Dravid.
(ii) No change
mār ‘breast’ (i) l > r, q’ > Ø
(ii) See items a.-c., g.-i. of Table 1
mulai ‘breast’ (i) k > Ø
(ii) See items a.-c., g.-i. of Table 1
mukku ‘gobble’ (i) Assimilation of l to k
(ii) ‘swallow’ → ‘gobble (up)’

Alternatives like these are by no means limited to the languages in example (8).
Similar alternatives can be found in numerous Indigenous American languages,
as well as in such families as Altaic, Bantu, and Austronesian.
The fact that it is so easy to find alternatives of this type raises important
questions about Greenberg’s wide-ranging etymology. What are the criteria for
choosing among different alternatives without being arbitrary? Which are the
genuine cognates, if any? And which are the false friends? Important here is the
fact that barring special circumstances, at most one form per language or lan-
guage family can be a true cognate; the others must be false friends.
442 Proto-World? The question of long-distance genetic relationships

Greenberg does not address this issue. Implicitly he seems to be operating

with the idea that the forms should be maximally similar both phonetically and
semantically. But this is hardly a realistic approach. As examples like (6b) above
illustrate, given enough time, genuine cognates often become differentiated
beyond recognition.
There may in fact be no answer to this question, short of comparative recon-
struction. But to do reconstruction requires much richer evidence and a method-
ology markedly different from Greenberg’s mass comparison.
Further, the very ease with which alternatives of the type (8) can be found
calls into question Greenberg’s claim that the similarities in his data cannot be
due to chance. Whatever may be the mathematical foundation of his calculation
that the chance of finding the similarities in Table 1 is one in ten billion, the empir-
ical evidence suggests that the conclusion is incorrect.
Far from corroborating Greenberg’s mass comparison, the long-range etymol-
ogy *maliq’a ‘throat, swallow’ actually casts further doubt on any approach that
involves setting up “cognates” merely on the basis of similarities in sound and
meaning. What it suggests instead is that only the traditional approach, however
cumbersome and time-consuming it may be, makes it possible to distinguish true
cognates from false friends.
At this point, then, there is no credible alternative to the cumbrous and
time-consuming traditional method of comparative linguistics. And, as noted
earlier, this method is at this point incapable of successfully establishing even
longer-range relationships, not to mention settling the question of whether all the
world’s (spoken) languages are related.
This may be considered regrettable, for it would indeed be marvelous if we
could establish longer-distance relationships or even prove that all the world’s
languages are descended from a common ancestor. But perhaps it is better to
acknowledge our ignorance, our inability to come up with definite conclusions,
rather than to believe that we have found the answers.
This way the challenge remains to look farther and deeper. And even if we may
never reach the goal of establishing relationship for all the world’s languages, we
are bound to gather new insights in our quest for that goal and over time we may
even succeed in more firmly establishing longer-range relationships such as those
of Ural-Altaic, Indo-Uralic, Lislakh, or even Nostratic.
 The origin of Language 443

5 The origin of Language

The controversy over Proto-World is really about the origin of all known (oral) lan-
guages; and indeed, under ideal, but most likely unattainable conditions, we might
perhaps catch a glimpse of the nature of the “mother tongue” from which all spoken
languages are derived – either through application of the traditional method of com-
parative linguistics (if sufficient new evidence is found), or through Greenberg’s mass
comparison (if it can be made to work), or some other as-yet undeveloped methodol-
ogy. Even so, catching a glimpse of such a mother tongue is a far cry from receiving an
answer to what might be thought of as the ultimate question in historical linguistic
research: Where did this mother tongue itself come from, and how did it originate?
In other words, how did Language (writ large) come about in the first place?
Over the centuries, some peoples have sought the solution by looking to the
heavens or to supernatural forces, as in the myths of various Indigenous American
peoples (see § 1 above), and as also supposed by the ancient Egyptians, and early
Christian writers such as Origen (3rd c. AD). Others saw language as a human
invention and worried about what the inspiration was for the form of the invented
words – did they arise as imitations of sounds in nature (the so-called bow-wow
theory), or as derivatives from human interjections (the so-called pooh-pooh or
yo-he-ho theory), or in yet other ways? In 1866, the Linguistic Society of Paris
declared that any theories on the origin of language were by their very nature so
speculative that they should no longer be discussed in its meetings. But the issue
of language origin is too interesting a question, too basic to our understanding of
ourselves as humans, and therefore too important to human beings for its discus-
sion to be forever banished from scholarly or even popular consideration.
Since at this point, comparative linguistics (whether Greenbergian, Nostrat-
icist, or Traditionalist) evidently cannot provide answers to the question of lan-
guage origin, scholars interested in this issue have drawn on a wide range of other
resources. The evidence that researchers have looked at includes the following:
– observable examples of language genesis, including normal language devel-
opment in children; language development in hearing or deaf children with
no exposure to language input of any sort; the origination of pidgin lan-
guages; and the development of creoles
– physiological evidence that might reveal aspects of the evolutionary develop-
ment of the human capacity for language, such as comparative anatomy and
fossil records
– the language – and other – abilities of creatures evolutionarily related to
humans, as these abilities manifest themselves both under natural conditions
and under training by human instructors or intimate exposure to human lin-
guistic behavior
444 Proto-World? The question of long-distance genetic relationships

We briefly examine below some of the insights that might emerge from these
sources.
The study of language development in children seems quite promising at first.
The stages of normal development have been studied and are fairly well under-
stood. It is tempting to extrapolate from these stages and to hypothesize that they
recapitulate the evolutionary stages of human language development, in accord-
ance with the Biogenetic Law, “Ontogeny recapitulates phylogeny”, first formu-
lated by Ernst Haeckel in 1866. However, one finding of child language develop-
ment research is that, to be successful, the acquisition of language depends on
the stimulus of other language users (generally the adult care-givers at first, but
later the child’s peer group). This finding is reinforced by the differences between
normal child language acquisition and abnormal – and fortunately, quite rare –
cases, in which children have to develop a language without human stimulus. (In
fact, well-documented cases of abnormal language development always involve
some kind of human stimulus, even though the stimulus may come very late, or
in spoken form when signed language input is needed.) The study of child lan-
guage development, therefore, cannot provide definite answers to the question of
how human language may first have arise, since by definition there was no prior
human input at that point, whenever it may have been.
Similar concerns arise when one looks to the development of pidgins and
creoles; these, too, depend on human language input, in the form of already-exist-
ing languages. Bickerton’s “bioprogram” hypothesis (Chapter 14, § 5) might prove
more useful, in that it postulates an innate form of grammar that manifests itself
exclusively in creolization and therefore differs substantially from the grammar
of ordinary languages. It might be hypothesized that this innate grammar is closer
to the grammar of the early stages of human linguistic evolution. But as noted in
Chapter 14, Bickerton’s hypothesis is controversial; it therefore does not provide
a solid foundation for theories on the origin of language. Moreover, in a sense it
would at best push the origin of language back one stage, to the question of how
this innate grammar came to be part of human capabilities in the first place.
Some researchers have looked for clues to the origin of language in evidence
from physical anthropology, including the study of human fossils, comparative
anatomy, or structures in non-humans – or even non-primates – that are anal-
ogous to the human organs of speech (such as the vocal tracts of apes or even
the air passages of frogs and lungfish). Some important findings have emerged,
such as the observation that the speech organs did not evolve for the purpose
of speech, but rather, that speech is an “overlay” function, superimposed on a
vocal tract originally developed for other purposes. Measurements of the vocal
tracts of apes show them to be quite different from those of adult humans, and
actually somewhat similar to those of human newborns. The study of fossil skulls
 The origin of Language 445

of Neanderthals, an evolutionary cousin to modern humans, has suggested that

Neanderthals may not have been able to articulate the full range of speech sounds
found in known human languages, and thus may also have lacked the perceptual
apparatus to process the full range of human speech sounds. These observations,
taken together, suggest that human language as we know it may be a fairly recent
evolutionary innovation, perhaps on the order of 50,000 to 100,000 years old,
although less complex and less highly structured forms of communication in
pre-human ancestors could be much older. These claims, however, remain con-
troversial. For instance, the claim that Neanderthals were incapable of producing
and understanding language as we know it is based on examination of a single
skull that may have been damaged by arthritis and physical trauma. Moreover,
the evolutionary situation is very complex. As many as 100 developmental factors
have been identified by some scholars as contributing, directly or indirectly, to
the ultimate emergence of human language; and the evolutionary history of many
of these factors is still poorly understood. Nevertheless, the comparative physio-
logical approach has improved our understanding of the physical developments
associated with the emergence of language.
Other researchers have turned to the linguistic abilities of primates, such as
chimpanzees and gorillas, as they can be inferred from both their natural call
systems in the wild and their latent abilities, drawn out through instruction in a
sign language or in the use of representational symbols such as plastic tokens.
The idea here is that the study of primate abilities and behavior can help us
understand the linguistic abilities and behavior of human beings just before the
development of language as we know it. The natural communication systems of
apes consist of a closed set of calls, about twenty in number, and certainly show
nothing resembling the complexity of human language. However, some observers
have seen in these systems the potential for developing the open-endedness and
complexity of human language. The attempts to tap the latent linguistic abilities
of apes have resulted in some dramatic claims of language learning by these apes;
but many linguists consider these claims controversial. More important, it is not
at all obvious that the linguistic training of apes by human instructors can tell
us anything about how humans may have first acquired language at a time when
there were no other living beings that could have instructed them.
What may be more significant is that the successes in training apes in some
kind of language (however far removed that language may have been from human
speech) were accomplished by employing non-oral means of communication –
modified forms of signed language or systems based on plastic tokens. Moreover, as
we have seen in Chapter 16, § 8, numerous signed language systems have developed
spontaneously around the world. One theory of language origin, therefore, becomes
quite attractive, namely the so-called gestural theory, according to which human
446 Proto-World? The question of long-distance genetic relationships

language first originated in a gestural “channel”, and the shift to the oral channel
was a secondary development. This switch in channel, it is suggested, may have
started with movements in the vocal organs, especially the tongue, that mimicked
manual gestures. An alternative hypothesis, not necessarily in conflict with this
view, is that vocal sounds at first were emphasizing accompaniments to meaningful
manual gestures, just as facial expressions and gesticulations now can accompany
speech as a form of non-verbal “paralanguage”. Some researchers further claim that
the complete switch to the oral channel was motivated by the fact that it enabled
human beings to use their hands for purposes other than communication and still
to effectively communicate. Nevertheless, many researchers remain skeptical.
These various lines of investigation, as important and as interesting as they
are, have largely focused on aspects of the form of the various (literally) “moving
parts” that go into the generation of speech and of human language. The question
of why these pieces and structures should have been created in the first place, let
alone put into use in the way they have been, is not addressed. Another promising
line of research, accordingly, involves looking at behavioral aspects of interactions
among humans, among nonhuman primates, and among other species of animals.
That is, humans show cooperative behavior, a cornerstone of language use,
far more than any other animal species, including other primates. Cooperative
communicative interaction occurs among nonhumans, just to a lesser extent and
not involving speech; a baby ape, for instance, can develop a symbolic way, say,
by poking, to show its mother that it wants to be nursed. This is social negotiation
that leads to the exchange of information via symbolic means, thus something
like a linguistic signal. Two further insights from the animal world help here.
First, studies of finches, foxes, and bonobos suggest that the cooperative nature
may have emerged in an environment with a lower stress level, perhaps due to
more plentiful sources of food. Second, a comparison of humans with other pri-
mates shows that only humans imitate processes that lead to goals; by contrast,
chimpanzees imitate only goals. Process-oriented imitation can transmit complex
patterns of behavior, allowing culture in a broad sense to emerge. Since language
is a cultural phenomenon, there may be a link here in evolutionary terms.
As fascinating and provocative as all these threads of evidence are, they
do not offer a definitive answer for the origin of language. In fact, all of these
notions and constructs, i. e. a model of the co-evolution of different systems,
may be needed. Thus, the controversy over the origin of language will likely go
on, perhaps forever. But if language really is one of the defining characteristics of
human beings – a view dear to most of us, even if open to debate – it may well be a
good thing that the answer to what started us on the road to humanity lies beyond
our grasp, for this will encourage us to continue to examine ourselves, our place
in the world, and the role language plays for us as human beings.
Chapter 18: Linguistic palaeontology:
Historical linguistics, history,
and prehistory
We have found a strange footprint on the shores of the unknown.
We have devised profound theories, one after another,
to account for its origin. At last we have succeeded in
reconstructing the creature that made the footprint.
And lo! it is our own.
(Arthur Stanley Eddington, Space, Time, and Gravitation, Chapter 12)

The past is a foreign country; they do things differently there.

(Lesley Poles Hartley, The Go-Between, prologue.)

1 Introduction
One of the most exciting aspects of doing historical work of any kind is the thrill of
getting a glimpse of events that may have happened eons before our time. While
the results of historical research are sometimes speculative, they are always inter-
esting, challenging, or even highly controversial. In this chapter we see a number
of ways in which historical-comparative linguistics, often combined with the tes-
timony of ancient texts, contributes to the challenges and controversies of history,
especially prehistory, and to archaeology.
The idea that historical-comparative linguistics might be of any relevance in
this regard may strike many as odd. However, the fact that we can reconstruct
vocabulary of the proto-language has implications, for the words that we recon-
struct must have referred to real objects, animals, plants, ideas, and the like and
thus can be expected to open a window on the world and the world view of the
people that spoke the language. As archaeologists often put it, “Pots don’t speak”.
Supplementing the archaeological findings with the insights of linguistics can be
expected to make the pots “speak”.
For instance, the fact that we can reconstruct PIE words for ‘horse’ (*eḱwos),
‘cow, bovine animal’ (*gwōws), and ‘dog’ (*ḱuwōn) tells us a great deal about the
degree to which the speakers of PIE had succeeded in domesticating animals; for
clearly, it makes no sense to have the words without also having the animals that
the words refer to.

https://doi.org/10.1515/9783110613285-018
448 Linguistic palaeontology: Historical linguistics, history, and prehistory

Drawing on linguistic evidence of this type to make inferences about the

culture, society, and ecology of prehistoric peoples is known as linguistic pa-
laeontology.
Although linguistic palaeontology can in principle be applied to the ances-
tor of any established language family, the following examples are drawn from
Indo-European, since research both on reconstruction and on linguistic palaeon-
tology has been conducted most extensively for this family.
Even so, many issues of Indo-European linguistic palaeontology are contro-
versial – especially the ones that would be of greatest interest to the student of
general prehistory. Moreover, there are numerous cases in which linguistic palae­
ontology has been drawn on to support ideological movements or theories, such
as the “Aryan Myth” that prepared the ground for the Nazis. There are also cases in
which ideological movements reject the insights of comparative-historical linguis­
tics and of linguistic palaeontology because they conflict with beliefs cherished by
the movements. For instance, radical Hungarian nationalists reject the a ­ ffiliation
of Hungarian with the other Uralic/Finno-Ugric languages, preferring a more “glo-
rious” affiliation with the Huns (and/or the Sumerians). Similarly, Indian nation-
alists tend to reject the theory that the “Aryans”, i. e. the speakers of Indo-Aryan,
came to India from the outside (from the Eurasian steppes or Anatolia), because
it conflicts with their view that the “Aryans” are indigenous to India. A different
perspective on what is often called the “Aryan Invasion Theory” (AIT) was that
of British “Orientalists” who looked at the theory as establishing a deep affinity
between India and the West. British “Anglicists”, by contrast, viewed the theory
as proof of the superiority of westerners; and this perspective was shared by the
Nazis and their ideological forerunners (see § 9.1 below) and is still espoused by
neo-Nazis. The AIT was accepted for yet a different ideological reason by Tamil/
Dravidian nationalists who viewed the northern, “Aryan” populations of India as
intruders who suppressed the indigenous Dravidian people.
With the exception of the textual evidence that seemed to support the theory
that the “Aryans”, i. e. the speakers of PIE, were white (§ 9.1), we will focus
mainly on hypotheses and theories proposed by serious scholars of linguistics
and archaeology – these turn out to offer enough controversy and differences of
opinion. For instance, a recent proposal by scholars trained in genomic research
located PIE in Anatolia at about 6500 BC (see § 8.5), rejecting as irrelevant lin-
guistic palaeontological evidence that contradicts the proposal and insisting on
the superiority of their mathematically based methodology. The journal Science
rejected submissions that tried to refute the proposal. And in a 2015 publication
Pereltsvaig (a linguist) and Lewis (a specialist in historical and regional geogra-
phy) characterized the proposal as “garbage in/garbage out”.
 Preliminary I: The term “Aryan” 449

2 Preliminary I: The term “Aryan”

In the early stages of Indo-European linguistics, before the sound changes had
been worked out, the phonetic similarity between the term ārya used by speakers
of early Indo-Iranian as a self-designation and the name Ériu, Mod. Ir. Éire for
Ireland was interpreted as indicating that the speakers of Proto-Indo-European
used the term ārya to refer to themselves, and “Aryan” came to be used by some
linguists instead of the more clumsy term Indo-European or one of its hyphen-
ated variants. Once the sound laws of Indo-European had been worked out, it
became clear that there is no etymological relation between ārya and Ériu; and the
use of the term “Aryan” in the meaning Indo-European was given up in linguistic
circles.
However, as is well known, racist ideologues, especially the Nazis, appropri-
ated the term “Aryan” for theories of western or Germanic “racial” superiority.
As a result, many Indo-Europeanists have been reluctant to use the term even
in reference to the Indo-European languages of northern India, preferring to use
the term “Indic”. But Indians use the latter term to refer to all languages of India,
whether of Indo-European descent or not; and they prefer the term “Aryan” or
“Indo-Aryan” for the Indo-European languages of South Asia. After some hesita-
tion (because of the misuse of the term) many scholars have now come to accept
the term “Indo-Aryan” as the designation for these languages.
To further complicate matters, in India the term “Aryan” is widely used to
refer both to Indo-Aryan languages and to their speakers, with special focus on
Sanskrit, the earliest form of Indo-Aryan.

3 Preliminary II: How many PIEs?

The relationship between the Indo-European languages is often modeled in terms
of a “tree”, with PIE sitting at the top, subfamilies such as Anatolian or Indo-Ira-
nian in the middle, and individual languages at the bottom; see Figure 1.
Ever since the discovery that Hittite and other Anatolian languages (see
Chapter 2, § 3.8) are Indo-European a large amount of linguistic evidence has been
unearthed that shows that the Anatolian languages must have broken off from the
undivided Proto-Indo-European ancestor earlier than any of the other languages.
More recent scholarship converges on considering Tocharian the next group
of languages to have broken off from Proto-Indo-European. Especially interesting
in this regard are semantic changes shared by the Indo-European languages other
than Tocharian (and Anatolian); see (1). As can be readily seen, the meaning in
450 Linguistic palaeontology: Historical linguistics, history, and prehistory

Proto-Indo-European

Anatolian Greek Italic Celtic Germanic Slavic Indo- Tocharian

Iranian

Hittite etc. Latin etc. Sanskrit etc. A B

Fig. 1: Conventional tree diagram of the Indo-European languages

Tocharian must be the original one and that of the other languages secondary –
the change can only work in one direction.

(1) PIE *yebh/yobh-

Tocharian A, B yäp- ‘enter’
vs. Sanskrit yabhati ‘copulates’
Greek oíphō ‘copulate’
Russian ebu [yebu] ‘copulate’

As a consequence, the diagram in Figure 2 represents a more accurate portrayal

of the chronological relationships between the Indo-European languages. The
importance of this for the present chapter lies in the fact that we have to be careful
about how we use the term “PIE”. Is it the “Proto-Indo-European” ancestor on top,
before the departure of Anatolian? Is it the stage after the Anatolian departure and
before Tocharian splits off? Or is it the stage after the Tocharian split?

To distinguish these different stages, it will sometimes be useful to use different

labels – “PIE I” for the ancestor of all the languages, “PIE II” for the stage after the
Anatolian departure, and “PIE III” for what remains after the Tocharian split. (The
latter stage is commonly referred to as Core Indo-European, but the term sounds
a bit ethnocentric.)
Some linguists postulate yet a further split, a migration of the ancestors of
the Italic, Celtic, and Germanic speakers away from the other varieties of PIE III.
 Religion, society, and law 451

Proto-Indo-European

Anatolian (Hittite)
Tocharian

Other IE languages

Fig. 2: The relationship of Anatolian and Tocharian to the rest of Indo-European

4 Religion, society, and law

Information on Indo-European religious beliefs, society, and aspects of law comes
largely from textual evidence, especially of the languages that are attested ear-
liest – Hittite, Sanskrit (and to a lesser extent, Avestan), Greek, and Latin. But
linguistic reconstruction plays a considerable role as well.

4.1 Religion, mythology, and poetic tradition

Indo-European religion is often portrayed as a strongly male-oriented one, whose

deities are associated with the sky. Support for this view comes in the fact that
*dyēw-s, lit. ‘sky’, is the only divine name that we can reconstruct with confidence
for Proto-Indo-European. The name refers to a male God who plays a prominent
role in Greek and Latin, as Zeús or Juppiter, chief of the Gods. (Note also the Ger-
manic deity hidden in Engl. Tuesday; see Chapter 4, § 3.3.) Moreover, words for
‘God’ that are in widespread use in early Indo-European languages, such as Skt.
dēva-, Lat. deus, OIr. día, and Lith. dievas, are derivable from an adjective *deyw-
os, whose literal meaning is ‘belonging the sky’.
Note further that Lat. Juppiter reflects a collocation *dyew pǝter which liter-
ally means ‘O father sky’. The collocation evidently is old, since it has counter-
parts in Greek Zeús patḗr and Vedic Sanskrit dyaús (…) pitā́.
Significantly, we have no evidence for reconstructing the name of a female
deity, comparable to the image of the Mother Goddess found throughout much
of prehistoric Eurasia. When Mother Goddesses do occur in early Indo-European
texts, their presence is said to reflect the substratum of conquered, pre-Indo-Eu-
ropean indigenous populations who worshiped the Goddess.
So, at least, the story goes.
452 Linguistic palaeontology: Historical linguistics, history, and prehistory

But there are difficulties with this traditional view of early Indo-European
religion. And these difficulties concern not only our picture of the Indo-Europe-
ans, but also of the pre-Indo-European indigenous peoples. Perhaps the major
problem is that for the Indo-Europeans we have to rely on traditions conveyed
through language; for the pre-Indo-European populations of Europe and Asia
we depend on artifacts. Now, there can be no doubt that the textual traditions
of the Indo-Europeans – just as those of the ancient Near East, of Egypt, or of
Meso-America – were handed down by males and that this male dominance may
well have skewed the information that was preserved for posterity. Artifacts, in
this regard, may be more “gender-neutral”. But here, too, the evidence may have
been skewed, even if unintentionally. Archaeologists have been impressed with
the widespread presence of Mother Goddess images in early prehistoric sites and
have inferred from this presence a female-dominated religion. But fine-grained
research on the Mediterranean island of Malta suggests that in at least one society
of this type, the female-dominated religious sphere coexisted with a male-domi-
nated one. The former was found in subterranean burial complexes, the latter, in
above-ground sanctuaries. The heavy presence of female images in earlier archae-
ological digs, then, might reflect the fact that such digs often unearth burial sites
which, being placed underground, had a better chance of being preserved than
above-ground sanctuaries.
The new archaeological evidence invites reexamination of early Indo-Euro-
pean tradition. True, we find strong evidence for male deities associated with
the sky; but we also find recurrent myths and other references to female deities
associated with the earth and the fertility that lies in the earth – and even to the
impregnation of ‘earth’ by ‘sky’ as the act through which the world was created.
(Impregnation, of course, may still be interpreted as an act of male domination;
but if we take this interpretation to its strict conclusion, then all human soci-
eties are male-dominated, and the distinction between male-dominated and
female-dominated societies becomes meaningless.)
The primordial embrace of father sky – above – and mother earth – below – is
no doubt the ideology underlying the Vedic Sanskrit passage in (2), which jux-
taposes ‘father sky’ and ‘mother earth’. Note further that the Mother Goddesses
of the Greek “Mysteries” generally were associated with the earth or with under-
ground caves.

(2) dyaúr me pitā́ ǰanitā́ nā́bhir átra

bándhur me mātā́ pṛthivī́ mahī́yám
‘Father sky is my creator, the navel here;
Mother earth is my kin, this great one.’
 Religion, society, and law 453

As the case of ‘father sky’ and ‘mother earth’ shows, the comparative method
permits us to reconstruct a fair amount of Indo-European non-material culture,
especially if it draws not only on individual words, but also on collocations of
words, as well as on the literary traditions of the early Indo-European peoples.
Especially in the area of poetic language we find numerous fixed collocations
that have been handed down as a part of Indo-European poetic heritage. Compare
for instance Gk. áphthiton kléos : Skt. ákṣiti śravas, both meaning ‘imperishable
fame’ and, in spite of their phonetic differences, both derivable from a common
source, PIE *ń̥ -gwhdhi-tom or *ń̥ -gwhdhi-ti ḱléwos. Scholars have also begun to
reconstruct elements of Indo-European myths, especially of one in which a great
hero or God slays a large snake or dragon. In fact, in a publication titled How to kill
a dragon: Aspects of Indo-European poetics, the American linguist Calvert Watkins
proposed to reconstruct a sentence which must have formed part of the traditional
telling of the myth and which states quite simply and appropriately that “He [the
hero] killed the dragon”.

4.2 Society

From similar evidence it can be gathered that the structure of Indo-European

society was patriarchal, with the woman moving into the husband’s father’s
extended family; that it was organized around the clan as its smallest unit beyond
the extended family; and that clans could form tribes, the basic units of early
Indo-European ethnicity.
The early traditions of the Indo-Europeans tell us that they were warlike
people whose leaders were heroes: They could even slay dragons and, under more
mundane conditions, could lead their people to victory over others. In historical
times we find one Indo-European group after the other invading other peoples’
territory. Note, for instance, the ransacking of Rome and Delphi by the Celts, the
imperialist expansions of the Romans, the raids on the Roman Empire by Ger-
manic and other tribes, or the Slavic incursions into the Byzantine empire and
much of present-day Russia.
This evidence is generally taken to indicate that the Indo-Europeans spread
over most of their later territory by military expansion. An eminent prehistorian,
Colin Renfrew, has painted a very different picture: The Indo-European languages
spread peaceably, along with agricultural innovations, from a much more south-
ern heartland in Anatolia (see § 8.5 below).
The facts, as they are preserved in early Indo-European tradition and in the
historical records, favor the traditional interpretation. However, it is quite likely
that the military exploits of the early Indo-European peoples are only part of the
454 Linguistic palaeontology: Historical linguistics, history, and prehistory

total story and that there was another, more peaceful dimension in which cul-
tural innovations could spread from tribe to tribe. In fact, we believe that in many
cases the invaders did not replace the local populations, but merely constituted
a thin – but powerful – overlay. This makes it at least possible that preexisting
networks of communication between different local populations remained rel-
atively undisturbed by the comings – and goings – of conquerors. It may have
been by this route that the agricultural words of § 6.3 spread into much of Western
Indo-European. Put differently, the conflict between Renfrew’s archaeologically
founded scenario and the traditional view of linguists and philologists can be
resolved if we envisage the world of early Indo-European expansion as consisting
of at least two levels: one being the “heroic” one of military conquest, the other
being a more peaceful one that might include agricultural and general cultural
diffusion. In this regard note the similar division between father sky and mother
earth discussed in the preceding section.
The image of the division of labor between father sky and mother earth can be
extended in an even more appropriate manner. The archaeologist E. J. W. Barber
shows us that far from being relegated to being the helpmates of their heroic
husbands, fathers, or sons, Indo-European women dominated an important and
much more peaceful sphere of their own – the production of textiles; and in that
sphere they either developed important new technological innovations, or passed
them on from one prehistoric society to the other. (This, of course, does not mean
that Indo-European women were treated as true equals. But none of the societies
for which we have early records accorded equal status to women. In this regard,
too, the Indo-Europeans may have differed little from other contemporary cul-
tures – whether we approve or not.)
In fact, it is highly unlikely that all (male) Indo-Europeans were militaristic
conquerors who, in the manner of traditional warfare, raped, pillaged, and plun-
dered peaceful non-Indo-European populations and their territories. True, early
historical records emphasize the military exploits; but we also get scenes of love,
tenderness, and compassion, as well as great thinkers like Socrates and compas-
sionate visionaries like the Buddha. We probably make a mistake in imagining
the early Indo-Europeans as exclusively great heroes – or great brutes, depending
on our perspective. In the area of human nature it is most likely that the “strange
footsteps” that we discover “on the shores of the unknown” are “our own”.
Although social structure probably was not as rigid in early times as in later
Greek, Roman, medieval European, and especially Indo-Aryan society, there was
no doubt some differentiation. As noted before, the early traditions tell of heroic
leaders, who could even fight dragons. There must also have been priests who
took care of the people’s spiritual needs and watched over the rituals to father
sky and perhaps also mother earth. In addition, there must have been people who
 Religion, society, and law 455

tended the cattle, cultivated the crops, and otherwise saw to it that everybody
could eat. Such a threefold division of society into warriors, priests, and common
people is, in a way, so mundane as to be unremarkable. But under the French
scholar Georges Dumézil, the division was turned into an “ideology”, which sup-
posedly governed all of Indo-European culture.
Dumézil’s views have given rise to an extended debate. Arguments have
been met by counterarguments, and so on. What has largely been ignored in the
context is the fact that early Indo-European society had at least one additional,
fourth, component – the slaves. As in all ancient societies, prisoners of war – if
they were permitted to live at all – were enslaved (see Chapters 2, § 3.4, and 9, § 3.1
on Slav : slave), and even members of the dominant group could become slaves
if, for instance, they committed certain crimes. True, unlike the chattel slavery
of the Caribbean and the Americas, traditional slavery was not hereditary, and
“manumission” often offered a release from serfdom. But the life of slaves was
a difficult one; and many societies depended on the work of slaves to maintain
them. (This is most strikingly true for Athenian democracy, which was built on the
backs of slaves who, of course, were barred from participating in the democracy
that made Athens famous.)
In addition to slaves, another group may have formed a special layer of early
Indo-European society, beyond the warriors, priests, and commoners, namely the
traders and artisans who provided tools and luxuries of foreign origin or intro-
duced the production of such items. A recent article argues that the presence of
traders and artisans accounts for striking similarities between prehistoric arti-
facts from northern India to the Balkans and even Scandinavia; and it points to
parallels in historical times of artisans and other specialists who migrate from
society to society, including the so-called Gypsies (or Roms). Like the slaves, such
migrant specialists would have been an important component of society and yet,
would not have been given full legal standing within society. In this sense, then,
they were outside society or, at best, on its margins.
It is interesting that a late Avestan text divides society into four, not just three,
strata: Warriors, priests, cattle-herders (i. e., commoners), and finally artisans.
Even more interesting, a similar explicit division of society, underlying the later
caste system, is found in early Sanskrit texts. But here the fourth stratum consists
of both artisans and slaves. Here, then, the fourth estate included all persons who
were outside society, even though they formed an important – and indispensa-
ble – component of that society.
456 Linguistic palaeontology: Historical linguistics, history, and prehistory

4.3 Comparative linguistics and comparative law

The comparison of textual evidence can be extended to many other spheres,

including law.
To take one example: In the ancient Roman, Greek, and Hittite legal tradi-
tions, one finds a parallel treatment of offenses committed by someone who is not
considered a legal person – a slave, a child, a cow, or the like. In principle, several
outcomes are possible; for example, declaring the offense to be a nonoffense (as
is done with some juvenile offenses in the United States); providing restitution;
turning the offender over to the person who suffered from the offense; and so
on. What is interesting is that all three ancient traditions offer the same resolu-
tion – a choice is allowed between restitution and turning the offender over to
the plaintiff. Thus, the structures of the legal codes are completely parallel, with
not just one outcome being prescribed but a choice between two outcomes and,
moreover, the same choices (out of several conceivable ones) are specified in all
three traditions. Further, cognate vocabulary is used: Latin uses the verb sarcire
‘to make amends for’ and Hittite the related verb sark- ‘make restitution for’. For
the alternative action, of ‘giving over’, Greek uses the verb para-dídōmi (root: dō-
‘to give’) and Latin the verb dēdere (also from the root dō-).
The structural parallels combined with the linguistic parallels thus permit us
to infer a common origin for the legal practices being compared and even provide
us with an idea of the technical legal language covering such situations in Pro-
to-Indo-European.

5 Material culture
From the perspective of the prehistorian it is a significant question whether the
Indo-Europeans are to be assigned to the Stone Age or to the (early) Metal Age.
The available evidence suggests a weak “yes” on both counts.
We can only reconstruct one word for metal that is usable for tools. This is
*ayos/ayes, the ancestor of Engl. ore, Goth. aiz ‘copper’, Lat. aes ‘copper (ore),
bronze’, Skt. ayas- ‘metal, iron’. The original meaning most likely included both
‘copper’ and its principal alloy, ‘bronze’. The evidence of this word, then, would
place the Indo-Europeans in the Bronze Age.
But there is much more evidence pointing in the direction of the Stone Age,
presumably the Neolithic. Many of the tools bear names suggesting that they
were originally made of stone or rock. Thus Germanic sahs, a short sword that
proved very handy in battle against the Romans and provided the base for the
 Economy 457

name Saxon, is related to Lat. saxum ‘rock, stone’. A modern reflex of the word is
Engl. zax, the name of a cutting tool used by roofers (with “Somerset” voicing of s
> z; see Chapter 10, § 7). The English word saw and its cognates in other Germanic
languages have been considered derived from the same root. Similarly, hammer
is a cognate of Slav. kamy ‘rock’; and Greek ákmōn ‘anvil’ and Lithuanian ašmens
‘cutting edge’ are related to Sanskrit aśman- ‘rock, stone’.
Further support for the view that the Indo-Europeans had a strong neolithic
background comes from the fact that wooden and stone instruments, rather than
metal ones, are used in many early Indo-European rituals. Rituals tend to be the
most conservative aspects of religion and therefore to preserve both linguistic and
material archaisms.
We may conclude, then, that the Indo-Europeans essentially had a neolithic
background but had recently, in a relatively late stage of Proto-Indo-European,
entered the Bronze Age.

6 Economy
Another issue that is of keen interest to prehistorians is the question whether the
Indo-Europeans were pastoralists and hence presumably nomadic, or agrarian
and hence presumably sedentary. Here again we find contradictory evidence; and
there is greater variation in the interpretation of that evidence.
Partly the variation reflects the fact that the terms “pastoral” and “agrarian”
and the economic patterns they refer to are not as clearly defined as suggested by
the assumption that “pastoral” means “nomadic” and “agrarian”, “sedentary”.
First, it is well known that agrarian societies often include a significant compo-
nent of cattle raising. Second, in many parts of the world (including present-day
South Asia) an essentially agrarian society coexists with an ethnically different
pastoralist society, in a system of mutual dependence. Most important, however,
is the fact that pastoral societies often are not “nomadic” in the sense that the
term is commonly understood but engage in a system referred to as “transhu-
mance”. Tribes move with their cattle in a cyclical fashion, putting up camp in
different areas with the change of the seasons. As they set up camp for a season,
societies like this can be called semi-sedentary. Moreover, research shows that if
seasonal stays last long enough, the transhumance system allows for seasonal,
but relatively low-level agriculture.
458 Linguistic palaeontology: Historical linguistics, history, and prehistory

6.1 Early perspectives

As noted earlier, we can reconstruct PIE words for ‘horse’ (*eḱwos), ‘cow, bovine
animal’ (*gwōws), and ‘dog’ (*ḱuwōn). We can therefore infer that Indo-Europeans
had made advances in the domestication of animals.
From the economic perspective it is especially interesting that the word
*peḱus could be used to refer both to ‘cattle’ (compare Skt. paśu- ‘cattle, esp.
bovine cattle’, Lat. pecus ‘(herd of) cattle’, Germ. Vieh ‘cattle, beast’) and to ‘prop-
erty, wealth’ in general (Lat. pecunia ‘property, wealth; money’, Engl. fee). Opin-
ions are divided on whether we should reconstruct ‘cattle’ or ‘property, wealth’
as the original meaning. But one thing we can be certain of – cattle and pastoral-
ism formed a significant basis, perhaps the major one, for measuring wealth in
Indo-European society.
The strong evidence in favor of cattle raising contrasts with the much weaker –
and controversial – evidence for farming or agrarianism. We find cognate words
for ‘field’, ‘plough’, and ‘sow, seed’ in the western Indo-European languages.
But cognates for these words, in these meanings, are absent in Indo-Iranian. To
account for this fact, 19th-century scholars proposed three different explanations;
and some scholars still find one or another of these attractive.
The first explanation considers the western Indo-European words and their
agricultural meanings to be innovations. For instance, the words for ‘field’ and
‘sow’ which could be reconstructed as *aǵros and *sē- respectively, can be derived
from roots with more general, less specifically agricultural meanings: *aǵ- ‘drive
(especially of cattle)’ and *sē- ‘throw, cast’. In the case of the word for ‘field’,
therefore, the original meaning may have been ‘area where the cattle are driven’
= ‘pasture’; the meaning ‘field (for agriculture)’, then, would be a later seman-
tic extension. As for ‘sow’ it is interesting that words for ‘throw, cast’ or ‘put,
place’ are used to refer to sowing in several Indo-European languages, such as
Gk. speírō, Mod. Ir. cuirim, and Skt. vap-. The specialization of *sē- ‘throw, cast’ to
refer to ‘sow’, therefore, could likewise be an innovation of the western members
of the family. According to this view, then, the Indo-Europeans may have been
pastoralists who were just at the threshold of agriculture.
A different explanation is also possible. The absence of cognate words for
‘field’, ‘plough’, and ‘sow, seed’ in Indo-Iranian could be explained by assum-
ing that the Indo-Iranians migrated away from an Indo-European homeland with
mixed pastoral/agricultural society. And because they migrated over vast dis-
tances and for many generations, they might have given up agricultural life and
the words that went along with it and instead reverted to pastoralism, which was
easier to maintain during their long treks. This account, of course, presupposes
 Economy 459

the assumption that the original home was somewhere in Europe – an assump-
tion that would require independent justification.
There was even a third hypothesis: The absence of cognate agricultural termi-
nology in Indo-Iranian may reflect a division within Proto-Indo-European society.
The more western tribes had begun to develop some form of agriculture, presum-
ably because they lived in areas with soil and climate conditions that were con-
ducive to farming. The more eastern tribes, by contrast, lived in an area with less
fertile soil and harsher climate and therefore stuck to pastoralism.
Given the evidence available to 19th-century scholars, it would be difficult to
make a non-arbitrary choice between these different proposals.

6.2 The evidence of Tocharian and Anatolian

The recognition of Tocharian and Hittite (Anatolian) as belonging to the Indo-Eu-

ropean family, and the testimony of these languages complicates matters even
further.
First, Hittite has a cognate of the words for sē- ‘sow’, namely say/siya-, and
interestingly the word has both the meaning ‘sow’ and the meaning ‘cast, throw’.
Should we take this to indicate that the meaning ‘sow’ is inherited from PIE after
all? Or should we assume that Hittite innovated independently from the western,
European languages by adding the meaning ‘sow’ to a verb whose basic meaning
was ‘cast, throw’? And how do we decide? Note that like Indo-Iranian, Tocharian
uses a different word for ‘sow’, Toch. A kat-, B kät-, which also means ‘strew,
scatter’. How should we assess the relevance of this fact for the reconstruction –
or nonreconstruction – of a word for ‘sow’?
Second, while words for ‘plough’, derived from a root *arǝ- (or more exactly
*h2erh3-), are widely attested in western Indo-European languages, including Gk.
aróō and Lat. arō ‘to plough’ but not in Indo-Iranian, there also seems to be a
cognate in Toch. B āre ‘plough (noun)’, on the far eastern edge of the Indo-Euro-
pean languages; and this might indicate that the word, in this meaning, goes back
to PIE. Cognates in Hittite, however, have different meanings, as in GIShah(ha)r(a)
‘rake’, hahharie- ‘to rake’, as well as harra- ‘break, bruise, grind’. Should we there-
fore reconstruct a meaning ‘break (up)’ and consider the meaning ‘to plough’ to
be secondary, possibly innovated independently in the western languages and in
Tocharian? Or should we reconstruct the meaning ‘plough’ for PIE II? What, then,
should we do about the absence of the word in Indo-Iranian?
460 Linguistic palaeontology: Historical linguistics, history, and prehistory

6.3 Evidence for grain farming?

As it turns out, we do have evidence for at least one Proto-Indo-European word

for a cereal grain, *yewo-, attested in Skt. yava- ‘grain’, later ‘barley’, Avest. yava-
‘grain’, Lith. javai ‘grain’, and with semantic specialization comparable to later
Skt. ‘barley’, Gk. zeaí ‘spelt’, Hitt. e(u)wa(n)- ‘special type of barley’(?). There
is even a Tocharian word (Toch. B yap ‘millet’) that has been compared; but the
historical phonology creates difficulties. The presence of the word in Anatolian,
Indo-Iranian, and possibly Tocharian strongly suggests inheritance from PIE and
could therefore be taken as evidence for agriculture. However, this is not the only
possible conclusion.
First, the Indo-European word for ‘grain’, in turn, was borrowed into Fin-
no-Ugric, compare Finn. jyvä ‘grain’, Votyak ju ‘threshed grain’. The fact that the
word could later diffuse into Finno-Ugric suggests that it could have done the
same thing in earlier Indo-European – as a “wanderwort” comparable to many
other agricultural items that are borrowed from language to language, often
through a long chain of intermediate languages. (See Chapter 8, § 1 on the words
for ‘sugar’ and ‘candy’.)
Second, and probably more important, the fact that only one word can be
reconstructed and that its meanings vary across different languages and across
time should raise concerns. Was *yewo- a single, cultivated kind of cereal grain? If
so, what was it? If not, was it one or another wild variety of cereal plant?
Wild varieties could have been harvested by both hunter-gatherers and pas-
toralists. But even cultivated varieties are known to have been grown by pasto-
ralists practicing transhumance. For instance, a recent report shows that Central
Asian pastoralists of the third to second millennia BC grew at least three varieties
(wheat, barley, and millet).

6.4 Conclusions

While it is fairly certain, then, that the Indo-Europeans practiced pastoralism, it is

less certain that they also practiced agriculture. What is perhaps most significant
is that we can reconstruct only one word for grain. Even if, against the evidence of
Hittite, we were to reconstruct a word for ‘plough’, this does not necessarily indi-
cate full-fledged agriculture; for we don’t really know what kind of implement the
word referred to. Was it a “digging stick” (the earliest phase in the developments
leading to the full plough), an “ard plough” (an intermediate phase), or some-
thing closer to what we now understand by the word plough?
 Horses, wheels, and more 461

Further, although many prehistorians take the evidence to indicate Indo-Eu-

ropean agrarianism, we have seen that some kind of agricultural activity is per-
fectly compatible with transhumant pastoralism. And unlike full-fledged agri-
culture, transhumant pastoralism is not fully sedentary but shifts camps in a
seasonal cycle.
Further support for the early Indo-Europeans having been (transhumant) pas-
toralists and not (sedentary) agrarians comes from the fact that we cannot recon-
struct a PIE word for ‘village’ (not to speak of ‘town’ or ‘city’). Sanskrit grāma, for
instance, which later means ‘village’, still refers to a troop of people who may be
on the move, and its Slavic counterpart gramada/gromada means ‘heap’ or ‘pile’.
Similarly, OE þorp, OHG dorf ‘village’ may be related to Lat. turba ‘crowd; tumult’,
Gk. túrbē ‘noise, bustle’. Other connections have been proposed, but ON þyrpast
‘to crowd, throng’ seems to support the etymology, and so does the fact that the
Germanic word was borrowed as Old French troupe ‘troop’ (with metathesis).
There is evidence for reconstructing a word meaning ‘housing, abode’, namely
PIE *dem- from a root meaning ‘construct’; but in a transhumant pastoral society
this could have referred to a set of prefabricated walls (etc.). This possibility
receives support from the fact that an early Vedic word for ‘house, abode’ is gṛhāḥ
(plural!; the singular gṛham becomes the normal word in post-Vedic); similarly,
in Lithuanian the plural namai is much more common than the singular namas.
These conclusions may come as a disappointment to prehistorians. Pastoral-
ism is a widespread phenomenon at the relevant period in prehistory. If the com-
parative evidence permitted us to be certain that the Indo-Europeans practiced
sedentary agriculture, this would provide much more specific information for
identifying the Indo-Europeans with physical artifacts that have been unearthed
by archaeologists, and with the particular cultural sphere in which those artifacts
were used, including built-up settlements.

7 Horses, wheels, and more

Most practicing Indo-Europeanists agree that three cultural/technological events
are the most relevant for trying to locate PIE in space and time. These are the
domestication of horses, the invention of the wheel, and the development of two-
wheeled, horse-drawn (battle) chariots. In this section we start the discussion by
looking at the linguistic and archaeological facts, as they are known at present.
We will take a closer look at the implications of these findings in § 8.
462 Linguistic palaeontology: Historical linguistics, history, and prehistory

7.1 The linguistic evidence

Phonologically related words for ‘horse’ are found in all the branches of Indo-Eu-
ropean, and a common ancestral from can be reconstructed; see (3). The forms in
(3c) have only recently been shown to be related. Note that the phonological var-
iation between forms with e and without is a pervasive feature of PIE phonology
and morphology; for other examples see Chapter 16, § 6 and also further below
regarding the root *kwel/kwl-.

(3) a. Reconstruction: *h1(e)ḱwos

b. Reflexes of *h1eḱwos
Skt. aśva
Avest. aspa
Arm. ēš (with changed meaning ‘donkey’)
Lat. equus
OEngl. eoh
OIr. ech
Lith. ašva ‘mare’
Toch. A yakwe
Luwian (Anatolian) atsuwa
c. Reflexes of *h1ḱwos
Gk. híppos
Czech sve-řep(ec) ‘breeding stallion’, Old Pol. świe-rzepa ‘mare’

While there can thus be no doubt that the word for ‘horse’ must be reconstructed
for all of Indo-European, including Anatolian, the same cannot be said for
‘donkey’. Here we find one set of forms in the western, European languages, as
in (4), and another set of forms in the eastern languages, as in (5). Not only is
there no etymological relation between the two sets; even within the sets there are
discrepancies such as in the Greek and Latin words. Further, among the western
languages, only the Greek and Latin words are old; the words in Germanic, Slavic,
and Celtic are all borrowings from Latin or borrowings of borrowings from Latin.
The fact that there may be a Luwian form asna hidden in tark-asna and that there
is a Sumerian anšu strongly suggests that the Greek and Latin words are also bor-
rowings, ultimately from Sumerian or some other Near Eastern language.
 Horses, wheels, and more 463

(4) a. Greek ónos

b. Latin asinus
⇒ pre-Old High German asilaz, Gothic asilus
⇒ Old Church Slavic osĭlŭ, Lithuanian asilas, etc.
⇒ Early Irish a(s)san
⇒ Old English assa

(5) a. Sanskrit gardabha

Tocharian B kercapo
Tamil kar̤ utai
b. Sanskrit khara
Younger Avestan xara
Saka khara

The eastern forms in (5) are of two major subtypes and, in the case of (5a) even
may include a non-Indo-European language (unless the Tamil word is a chance
similarity). A great variety of different etymologies has been proposed. For (5a)
these range from onomatopoeia for Skt. gardabha and Toch. kercapo, to possible
inheritance for the pair Skt. gardabha : Toch. kercapo (which could go back to
*gordhebho), to very early borrowing from Indo-Iranian to Tocharian, to borrow-
ing of Sanskrit gardabha from Dravidian, to possibly borrowing of the Sanskrit
and Tocharian words from a Central Asian substrate language. Regarding (5b),
the existence of a Sanskrit adjective khara ‘harsh, etc.’ and the fact that khara can
also refer to crows makes the assumption of onomatopoeia attractive (‘the animal
with the harsh shout’); but a borrowing connection with Akkadian ḫârum ‘male
donkey’ has also been considered.
At any rate, there is no evidence that would permit reconstruction of a PIE
word for ‘donkey’.
Reconstruction of a word *kwekwlo (or kwekwl(h1)o) ‘wheel’ for PIE II is required
by the evidence of Old English/Old Norse, Greek, Indo-Iranian, and Tocharian;
see (6). Evidence for an alternative word, *rotHo, is limited to PIE III; see (7). The
first of these two words is linguistically remarkable. It is a “reduplicated” forma-
tion, with a prefixed copy of the initial consonant plus the vowel e and a reduced
form of the root *kwel ‘turn’ (i. e. *kwe-kwl-o); see Chapter 16, § 6. While reduplica-
tion is common in verb morphology, it is rare in nouns.

(6) a. Germanic *hwehla/hwegla (OE hweol/hweogol, ON hvél/hjól)

b. Greek kúklos
c. Indo-Iranian *čakra (Skt. čakra, Avest. čaxra)
d. Tocharian A kukäl, B kokale (→ ‘chariot, wagon’)
464 Linguistic palaeontology: Historical linguistics, history, and prehistory

(7) a. OIr. roth ‘wheel, chariot’, MW hrod

b. Lat. rota
c. OHG rad, OS rath, OFris. reth
d. Lith. rãtas (pl. ‘wagon, wain’), Latv. rats
e. Indo-Iranian *ratha (Skt. ratha, Avest. raθa → ‘chariot’)

Although Hittite has a word for ‘wheel’ (hurkis), that word is not a cognate of the
words in the other languages and looks like an independent creation.
Evidence for chariots comes from early Indo-European texts, but unlike the
words for ‘horse’ and ‘wheel’, there is no evidence that there was a PIE word for
‘chariot’ (whether in PIE I, II, or III). In some languages, older words for ‘wheel’
were used to refer to chariots – Indo-Iranian ratha, OIr. roth, Toch. kukäl, kokale,
with a semantic shift similar to Engl. nice wheels. Others created new terms, e. g.
Gk. hárma ‘chariot; chariot and horses’ from the root in arariskō ‘join together’.

7.2 The archaeological and archaeozoological facts

Evidence for horse domestication appears from the beginning of the 4th millen-
nium BC, in the Eurasian steppe area stretching from eastern Ukraine to Kazakh-
stan. Early evidence involved abrasion of back teeth, suggesting the use of bitting.
Doubts about these findings should be laid to rest by a recent study that provides
evidence for the milking of mares around 3,500 in Kazakhstan.
The hypothesis that horse domestication started in the Eurasian steppes is
supported by recent genomic work. In Anatolia and the neighboring Near East,
wild horses were extinct by the end of the palaeolithic. The case is similar for
India/South Asia and Greece; and the relative genetic/genomic homogeneity of
domesticated horses in Central Europe suggests outside origin, rather than local
domestication. The only area with genomic diversity comparable to that of the
Eurasian steppes is the Iberian peninsula; but this area has no evidence for the
use of bitting or the milking of mares.
It is instructive to compare the case of the horse with that of the other equid
that was domesticated at roughly the same time – the donkey. Current research
shows that donkeys were domesticated from (northern) African stock. The ear-
liest appearance of domesticated donkeys dates to between 4000 and 3000 BC
in Egypt, with Southwest Asian (including Anatolian) dates being later, between
2,400 and 2,200 BC.
Wheels and wheeled vehicles turn up in the archaeological record more or
less simultaneously in a large area of western Eurasia and South Asia around 3500
BC, without clear evidence as to which area(s) may have developed wheels earliest.
 Horses, wheels, and more 465

Two-wheeled, horse-drawn chariots appear first in the Sintashta Culture,

southeast of the Urals near the present border of Russia and Kazakhstan (ca.
2000 BC). In the same area we also find the first occurrences of horse and/or
chariot burials, commonly in connection with the burial sites of what appear to
be chieftains. The archaeological record shows that chariots and horse burials
spread from this area west, south, and east. In some cases, there is clear evi-
dence that the spread coincided with the spread of Indo-European speakers, e. g.
in Mycenaean Greece where Greek language, images of horse-drawn chariots, and
burials of paired horses appear during the 16th century BC. In Anatolia, chariots
are first mentioned in a 17th-century Hittite text, but a later text from about 1500
BC suggests that the art of horse-and-chariot use was brought from the outside, by
speakers who used Indo-Iranian (or Indo-Aryan?) terms such as aikawartana ‘one
turn (around the race course)’. In South Asia we find evidence for horse burials
around 1600 BC, in the Swat Valley, close to the Khyber Pass, the traditional point
of entry from Central Asia.

7.3 Putting it all together

As we have seen, the word *h1(e)ḱwos ‘horse’ must be reconstructed for PIE. In
virtually all of the languages it must have referred to domesticated horses, since
wild horses had become extinct in palaeolithic times in Anatolia, Greece, and
South Asia; and Central Europe likewise does not appear a likely area for horse
domestication. The only plausible way we can explain this situation is by assum-
ing that PIE was spoken at a time and in an area close to the time and place of
horse domestication, i. e., the early 4th millennium BC in the Eurasian Steppes. By
contrast, the fact that we cannot reconstruct a word for donkey suggests that PIE
was not spoken close to the area of donkey domestication, i. e. close to Egypt and
the neighboring Near East.
The fact that words for ‘wheel’ are shared by the PIE II (or PIE III) languages,
but not by Anatolian, might indicate that Anatolian split from the other languages
before the invention of the wheel; but other explanations are conceivable. Inter-
estingly, one scholar has proposed that the wheel was invented by the Indo-Eu-
ropeans and that the word *kwekwlos was borrowed into the languages of the Cau-
casus such as Kartvelian and beyond (e. g. Sumerian). In his view this hypothesis
is supported by the fact that Kartvelian and Sumerian have similar reduplicated
forms (8) and that the variation between labial and velar in these languages pre-
sents different nativizations of the PIE labiovelars. The hypothesis, however, is
weakened by the fact that the languages have a similar variation in verbs meaning
‘turn, roll’ from which the reduplicated noun forms can be derived; see the right
466 Linguistic palaeontology: Historical linguistics, history, and prehistory

side of (8). It is highly unlikely that these mundane verbs also were borrowed
from PIE. This leaves the possibility of calquing, the recreation of morphologically
complex forms through native morphology. But as we have seen in Chapter 8,
§ 3, without some “outside” historical evidence, the direction of calquing cannot
be determined, and there is no independent archaeological evidence that would
establish where wheels originated. Assuming Indo-European origin therefore is
arbitrary.

(8) a. PIE *kwe-kwl-os ‘wheel’ *kwel/kwl ‘turn’

b. Kartvel. *gr-gar- / *br-bar- ‘wheel, wagon’ *bar- / *gar- ‘roll’
c. Sumer. GIŠgi-gir ‘wagon’, gúr / bùr ‘circle’ gir- ‘roll’

Strong circumstantial evidence supports the view that the speakers of Anatolian
split off before the introduction of the two-wheeled, horse-drawn chariot. First,
there is the possibility that the Hittites learned the use of horse-drawn chariots
from outside people who used Indo-Iranian technical terms. Second, we find evi-
dence for the presence of Hittite speakers in Anatolia as early as about 1900 BC,
too close in time – and too far in space – to the introduction of horse-drawn two-
wheeled chariots in the Sintashta area around 2000 BC.
For the non-Anatolian languages, it is likely that their ancestor must have
been spoken near the Sintashta area, where chariots seem to originate. However,
the fact that different languages coined different words for ‘chariot’ makes it
unlikely that the speakers of PIE II were identical with the Sintashta people –
whatever their language may have been – and that they directly or indirectly took
over the new technology from Sintashta.

8 Time and original home (“Urheimat”) of PIE

The findings which we just summarized are of great relevance to the larger issue
of when and where Proto-Indo-European was spoken. Especially the issue of
“where”, frequently referred to as the “Urheimat” (‘original home’), has been of
keen interest right from the beginnings of Indo-European studies. It has also been
highly contentious and controversial.
 Time and original home (“Urheimat”) of PIE 467

8.1 Early views

Still under the spell of biblical thinking, William Jones (1776) tried to classify lan-
guages in terms of three major families – Semitic, Hamitic, and “Yafetic”, after
the three sons of Noah. Since “Semitic” and “Hamitic” were already taken for
other groupings, he used the term “Yafetic” for Sanskrit, Greek, Latin, “Gothick”,
and “Celtick”, but also included other languages such as Quechua. Regarding the
original home, he opted for the Iranian plateau, perhaps influenced by the belief
that this is the area where Noah’s ark had landed.
Schlegel, who transmitted Jones’s proposal in modified form and inspired
further work on comparative Indo-European linguistics (Chapter 2, § 3), believed
Sanskrit to be the ancestor of the Indo-European languages and hence claimed
India as the homeland. In his view, the language was superior to other languages
(such as Chinese or American Indian languages) and was brought out of India by
a superior people.
Schlegel’s Indian-origin hypothesis became problematic when it became
clear that Sanskrit is not the ancestor of Greek, Latin, and the other Indo-Euro-
pean languages, but rather a sister language. This insight tended to favor hypoth-
eses that placed the Urheimat farther to the west, even in Central Europe.

8.2 The Beech Tree Hypothesis

One famous – or infamous – hypothesis that placed the Urheimat in or near

Central Europe centers on the existence of Latin and Germanic words for ‘beech
tree’ – a tree with edible nuts – and an Ancient Greek word for an oak tree with
edible acorns; see (10a). The phonological correspondences suggest that the
words can be reconstructed for the ancestral Proto-Indo-European (10b). If the
meanings of Latin and Germanic are taken to be the original ones, then it is pos-
sible to argue that Proto-Indo-European must have been spoken in an area with
beech trees. Significantly, the evidence of prehistoric pollen samples suggested
that this area was limited to the Central European area west of Lithuania – close
to Germany. And scholars, especially from Germany, concluded that therefore
Germany was closest to the Indo-European Urheimat, and Germans closest to the
Indo-European ancestors.
468 Linguistic palaeontology: Historical linguistics, history, and prehistory

(10) a. Lat. fāgus, OHG buohha, OE bēc (> Mod. Engl. beech) ‘beech tree’
Gk. phēgós ‘oak with edible acorns’
b. PIE *bhāgos? – meaning???
c. PIE *bhag- ‘share, partake of → eat’
Skt. bhaj- ‘partake (of), eat’
Gk. phageîn ‘to eat’

Other scholars, most of them also Germans, quickly pointed out that the entire
argumentation suffers from a variety of flaws. Only two can be mentioned here.
First, the assumption that the original meaning was ‘beech tree’ is arbitrary.
It would be just as possible to reconstruct a meaning ‘tree with edible acorns’.
Second, all the attested forms could be derived from the root *bhag- ‘partake of,
eat’ (10c), in which case the words could have been created independently to des-
ignate whatever tree with edible acorns or nuts the Romans, Greeks, or Germanic
tribes encountered.
As a matter of fact, the beech-tree hypothesis engages in circular reasoning.
It assumes that the meaning found in Latin and Germanic is the original one
and that the word is inherited. Having done so, it reconstructs an ancestral form
*bhāgos with the meaning ‘beech tree’. The reconstruction then is “confirmed” by
the Latin and Germanic forms together with their meanings, while the different
meaning in Greek and the absence of the word in other languages are attributed
to special developments.
Another hypothesis that favors a (northern) Central European origin is based
on the evidence of river names. Similar sounding names, such as Germ. Saale,
Lith. Salótas are limited to Europe and are said to cluster especially in the area
around present-day Lithuania. Assuming that these names reflect Proto-Indo-Eu-
ropean heritage, then, leads to the further assumption that PIE must have been
spoken in (northern) Central Europe.
Like the beech-tree hypothesis, the “river hypothesis” is problematic. The
assumption that the river names reflect Proto-Indo-European heritage is gratui-
tous, but without this assumption the claim that the area characterized by these
names is the original home of Proto-Indo-European falls apart. Some scholars
propose the alternative hypothesis that the names come from a pre-Indo-Euro-
pean substrate language, possibly related to Basque or even to Semitic. Consider
the many American river names taken from Native American sources, such as
Mississippi, Missouri, or Ohio.
 Time and original home (“Urheimat”) of PIE 469

8.3 The Balkan Hypothesis

An alternative perspective places the Urheimat in the Balkans, based on the fact
that in the early historical period the linguistic diversity among Indo-European
languages appears to be greatest in this area and in closely adjacent territory. In
addition to Greek on the southern periphery of the Balkans, we find evidence for
languages such as modern Albanian, ancient Macedonian, Thracian, and Illyrian,
and closely adjacent, Phrygian. (Most of the ancient languages are attested only
in fragmentary form; see Chapter 2, § 3.12.)
Now, as we saw in Chapter 11, § 6, where the historical record is clear, linguis-
tic diversity usually is greatest in the homeland, and smallest in colonial territory.
By extrapolation, then, the area of the Balkans could be considered the original
home of the Indo-Europeans.
Even this argument has its problems. An alternative hypothesis is possible,
namely that the Balkans were a kind of bottleneck, where different linguistic
groups “got stuck” in their migrations. This view can be buttressed by evidence
from known history. The Balkans have witnessed an enormous amount of migra-
tion – of the Romans (who left their trace in the Romanians), of the Huns and their
linguistically highly diversified allies, the Goths, the Vandals, and other Germanic
(and non-Germanic) tribes; of the Bulgars (a Turkic tribe that gave Bulgaria its
name); of the Magyars (Hungarians); of the Slavs; of the Turks; of the Roms. And
the result is the present-day highly diverse linguistic map of the Balkans that we
saw in Chapter 13, § 3.

8.4 The Caucasus Hypothesis

In recent years, some linguists have tried to locate the original home of the
Indo-Europeans in an area near the southern Caucasus. The arguments are partly
based on claims of early borrowings from or into Caucasic languages, Semitic
or Afro-Asiatic, and other, less well known ancient languages once spoken in
the area. These include words such as the Germanic words for ‘goat’ (including
Engl. goat) and their Latin cognate haedus ‘kid’ which bear strong resemblance
to Semitic words like Hebr. gǝðī́ ‘kid’. Similarly, the Greek, Latin, Armenian, and
Hittite words for ‘wine’ (oînos < *woyno-; vinum, gini < *weyno-; wiyana-) have
Semitic counterparts, such as Arab. wayn- ‘black grapes’, Akkadian īnu- ‘wine’.
Further, based on correspondences such as Gk. léōn, Lat. leō, OHG leo, lio, lewo,
Lith. levas, Pol. lew, it has been proposed that a word for ‘lion’ must be recon-
structed for PIE and that PIE therefore must have been spoken south of the Cau-
casus, in an area with lions.
470 Linguistic palaeontology: Historical linguistics, history, and prehistory

An additional argument is based on the “glottalic” reconstruction of the Pro-

to-Indo-European stop system (see Chapter 16, § 7): In its overall structure, the
glottalic system is remarkably similar to the system reconstructed for the ances-
tor(s) of modern Caucasic languages. This similarity, it is claimed, must be due to
convergence under close bilingual contact.
As we have seen in Chapter 16, § 7, the “glottalic” reconstruction of PIE is
controversial and may violate Occam’s Razor. Words like the ones for ‘goat’ and
‘wine’ or ‘grape’ are commonly considered borrowings from Semitic and other
Near Eastern languages and are limited to the more western Indo-European lan-
guages. As for words like ‘lion’, the Old High German, Lithuanian, and Polish
words can hardly be inherited, since the animal they designate is not found in
the areas in which the languages are spoken – and have been spoken through-
out their attested history. When they appear in the early literature of these
languages they refer to lions found in other parts of the world – or to lions in
the metaphorical sense, as symbols of power and heroism. In either case, they
must be borrowings from languages spoken in areas with lions. Moreover, South
Asia, where Sanskrit was (and still is) spoken, includes areas where lions roam;
but Sanskrit uses an entirely different word, siṁha-, of somewhat uncertain
origin. (This word, too, is used to indicate valor and power and underlies the
common Indian name Singh, as well the name of the city-state Singapore, lit.
‘the lion city’.) On balance, then, similar-sounding words for ‘lion’ are limited
to Greek and Latin and to languages – Germanic, Baltic, and Slavic – which
are likely to have borrowed the words from Greek or Latin. Indo-European-
ists tend to be leery of words that are geographically limited in this way since
such words may well be regional innovations (as in the case of the “beech tree”
argument).

8.5 Anatolia vs. the Steppes of Eurasia

As we have seen in § 6, a large amount of evidence, both linguistic and archaeo-

logical/archaeozoological, favors the hypothesis that Proto-Indo-European was
spoken in an area and at a time close to the area and time of horse domestication
in the Eurasian Steppes, and that the same holds true for post-Anatolian Proto-In-
do-European as regards the development of two-wheeled, horse-drawn (battle)
chariots. Scholars accepting this perspective generally believe that it was the
possession of domesticated horses and, somewhat later, of two-wheeled, horse-
drawn chariots that made it possible for speakers of PIE or early Indo-European
to spread out over the large territory in which they are found at the beginnings of
recorded history.
 Time and original home (“Urheimat”) of PIE 471

There may be some disagreement as to whether PIE was pastoralist or agrar-

ian (see § 6.4). Pastoralism, whether fully nomadic or of the transhumance variety,
presumably would make it easier for people to spread over long distances, and to
do so in relatively short time – the time-depth from the period of horse domes-
tication to the earliest appearance of Anatolian speakers in the Middle East is
only about 1500 years; and the time-depth of about 500 years between the first
appearance of two-wheeled, horse-drawn chariots and the earliest appearance of
Indo-European speakers in Greece is even shorter. But with enough ingenuity it
is possible to construct scenarios by which sedentary agrarian populations can
accomplish the same spread, within the same time frames.
The most significant rival of the “Steppe Hypothesis” is the “Anatolian
Hypothesis”. The first serious modern proposal of the latter hypothesis comes
from Colin Renfrew, an eminent archaeologist. More recently, the hypothesis has
been supported by a team of scholars around Quentin D. Atkinson who apply sta-
tistical models derived from genomic research in order to determine the original
home and timeframe of Proto-Indo-European, based on the evidence of lexical
comparisons.
Renfrew proposes that the spread of Indo-European languages was connected
with a largely peaceful spread of agriculture and agriculturalists from the Fertile
Crescent, starting in the 8th millennium BC. The fact that recent genomic research
establishes a population movement out of that area, into western Eurasia (see
§ 9.2), might be considered to support Renfrew’s hypothesis and thus to counter
some of the objections by Indo-Europeanists.
One objection to Renfrew’s view of a peaceful, agricultural spread is that the
earliest Indo-European texts are full of references to conflict and war; see also
§ 4.2. But the most important counterargument is that an Anatolian origin fails
to explain the existence of words for ‘horse’ in all the major Indo-European lan-
guages and the fact that the word must refer to domesticated horses (§ 7.2). Ana-
tolia, where wild horses became extinct in the Palaeolithic, would hardly be an
area in which horses could be domesticated; and there is no evidence for horse
domestication anywhere before the early to middle 4th millennium BC, long after
the spread of agriculturalists from the Fertile Crescent.
Adherents of the Anatolian hypothesis counter this problem in various
ways. One of these is to claim that the words for ‘horse’ and ‘wheel’ may have
diffused through borrowing long after the time of Proto-Indo-European; another
one accepts the possibility that the words are inherited but suggests that they
originally had different meanings and that the meanings ‘horse’ and ‘wheel’ are
secondary. Attempts like these are highly problematic. For instance, if the word
for ‘horse’ had been borrowed after the descendants of PIE had become distinctly
different languages, we would expect the borrowed words to look more similar
472 Linguistic palaeontology: Historical linguistics, history, and prehistory

to the word in the donor languages. If, say, Greek had been the source, we would
expect correspondences as in (11a) with some morphological adjustments, rather
than the actual ones in (11b). Further, if ‘horse’ is assumed not to have been the
original meaning of *h1(e)ḱwo- it is legitimate to ask that an original meaning
be provided. Surely, it could not have been the other domesticate equid, the
donkey. As we have seen, the early Indo-European words for ‘donkey’ are dif-
ferent from *h1(e)ḱwo- and look like late borrowings from non-Indo-European
languages.

(11) a. Hypothetical borrowings from Greek hippos

Skt. hippa
Lat. hippus
Luwian hippa
b. Skt. aśva
Lat. equus
Luwian atsuwa

And yet, Atkinson and his group argue that statistical models derived from
genomic research and applied to Indo-European basic vocabulary establish
about 6,500 BC as the date, and Anatolia as the place of Proto-Indo-European.
The model appears to be predicated on the assumption that the language spoken
in the original home must be the most archaic and on the view that the rate of
lexical replacement in basic vocabulary can be gauged accurately.
The first assumption is problematic on purely linguistic grounds. The example
of American and British English shows that transplanted varieties in many cases
preserve archaisms lost in the homeland, as in Am. Engl. [rǣðǝr] (rather), with
older [æ] and retention of final [-r], vs. Brit. Engl. [rāðǝ].
The second assumption suffers from an even greater amount of problems.
The statistical methodology has been challenged. The reliance of a methodol-
ogy based solely on lexical evidence and ignoring the evidence of morphological
change has been questioned.
Most important, like the earlier “glottochronological” approach based on
lexical (basic-vocabulary) evidence, it has been shown to lead to results that are
empirically indefensible. In the case of “glottochronology”, it was shown that the
assumption of a constant rate of lexical change, the foundation of the method, is
disconfirmed by the evidence of cases such as Icelandic vs. Norwegian (Bokmaal):
Where Icelandic shows a replacement rate of 4 % per millennium, Norwegian has
20 %, neither of which agrees with the postulated rate of 14 %. A probable expla-
nation is that Norwegian has borrowed heavily from Low German, even in the area
of basic vocabulary, while Icelandic has been relatively isolated from contact and,
 Time and original home (“Urheimat”) of PIE 473

moreover, as a result of linguistic nationalism has eliminated borrowings that

were found in early Modern Icelandic; see Chapter 8, § 5.2. Borrowing/language
contact (and its counterforce of linguistic nationalism) thus can have a profound
effect on the rate of lexical change.
A similar problem has been pointed out regarding the new statistical approach,
according to which Romani is claimed to have split from the rest of Indo-Aryan
about 1500 BC or, according to a more recent version, 500 BC. Linguistic facts,
such as the close relationship of Romani to the Modern Indo-Aryan languages,
rather than to the Sanskrit of 1500 BC or the Prakrit of 500 BC, show that Romani
cannot have split away from the rest of Modern Indo-Aryan any earlier than about
1000 AD. Here again, the statistical approach is misled by a heavy amount of
borrowing. (The effects of language contact also have led to erroneous results in
the application of the method to the Austronesian languages.) Proponents of the
lexical statistical approach dismiss such problems as “minor” and claim that their
method gets things right in most cases.
A modified form of this methodology has more recently been applied to the
lexical and phonological and morphological evidence of the Indo-European lan-
guages. Significantly, the results of this study support the Steppe Hypothesis and
a date of 4500 to 3500 BC for PIE.
Recent palaeogenomic research by two teams working independently of
each other (and published in the same volume of the same journal) concludes
that in addition to an original hunter-gatherer population and the agricultural-
ists coming out of the Fertile Crescent, a more recent migration, around 2500 BC,
brought pastoralists to Eastern and Central Europe. One of the studies directly
associates the migration with the spread of Indo-European and Indo-Europeans;
the other is a bit more cautious.
There are indeed reasons for being cautious. First, the studies only cover the
Steppes and eastern/central Europe (as well as, possibly, the area of Tocharian);
they do not cover southern Europe, including Greece, or the areas settled by
Indo-Iranians. (Palaeogenomic evidence is not readily available in these areas.)
Second, the time of the population spread precedes the introduction of horse-
drawn, two-wheeled chariots; the existence of chariots in the earliest forms of
Greek would therefore be difficult to explain in terms of such a population spread.
Third, since the late 19th century it has been known that there is no direct link
between genetic or genomic type and language. Finally, the common view that
language spread implies massive population movement is problematic; we now
know that relatively small groups can spread a language, if social and political
conditions are right. For instance, Turkish is in origin a Central Asian language,
but a recent genomic study shows that only about 3.4 % of the Turkish population
can be genetically traced to Central Asia.
474 Linguistic palaeontology: Historical linguistics, history, and prehistory

Even so, an origin in the Eurasian Steppes is the best hypothesis, given the
linguistic and archaeological/archaeozoological evidence currently available.

9 Genetics, genomics, and “race”

During the 19th and early 20th centuries there was a keen interest in the issue of
what the Indo-Europeans looked like – the question of their “race”. The second
half of the 20th century saw a strong tendency to avoid talking about this topic.
Partly this was a reaction to Nazi racism; but in part it was also based on the
insight that the concept “race” is a social construct and that, contrary to earlier
and still popular beliefs, there are no “pure races”. The advent of genomic studies
has led to a renewed interest in issues of genetic and genomic differences and
similarities between different population groups. Especially relevant for our
purposes is work in palaeogenomics. Advances in DNA analysis make it possi-
ble to sequence evidence extracted from prehistoric human remains, especially
in areas with relatively cool climates in which tissues have a better chance of
being preserved. We have already seen some references to the way in which this
kind of research contributes to issues in linguistic palaeontology. § 9.2 deals
with additional remarkable insights. We will start the discussion with 19th-cen-
tury views on “race”, their apparent justification in early Vedic Sanskrit texts,
and the way these views seemed to justify the “Nordic Race” ideology of the
Nazis.

9.1 N
 ineteenth-century views on “Race” and the issue of
racism and Nazism

As is well known, the Nazis subscribed to the view that the Indo-Europeans were
“Nordic”, i. e. light-skinned, blond, and blue-eyed “Aryans”, that they were supe-
rior to any other group of people, especially the Jewish people, and that these
“facts” justified the Nazis’ attempts to suppress and exterminate such “inferior”
groups as the Jews and the Roms (“Gypsies”).
It is therefore appropriate to ask several questions: Who was responsible
for introducing this “racial” definition of the Indo-Europeans as light-skinned,
blond, and blue-eyed? What, if any, is the evidence on which this definition
was based? To what extent were Indo-Europeanists and philologists working on
ancient Indo-European languages responsible for the atrocities committed by the
Nazis? To what extent could they have prevented these atrocities by taking seri-
 Genetics, genomics, and “race” 475

ously the implications of their work and by challenging ideologically motivated

misinterpretations and misapplications?
As it turns out, answers to these questions do not come easy, especially since
the racism fueling Nazi ideology pervaded much of 19th- and early 20th-century
western civilization and most fields of scientific inquiry, including anthropology,
genetics, and the study of “exotic” and “oriental” cultures. It is therefore entirely
possible that Nazism would have developed even in the absence of any textual
support for defining the Indo-Europeans as “Nordic”. Moreover, the “Nordic
hypothesis” turns out to have been much more popular among anthropologists
and historians than among Indo-Europeanists. At the same time, as we will see,
Indo-Europeanists failed to take seriously their responsibility of subjecting the
“Nordic” hypothesis to the same degree of scientific and scholarly scrutiny as
other, more purely “linguistic” theories. Had they done so, they might well have
been able to debunk the “Nordic” hypothesis and thus to call into question the
“intellectual” foundation of this hypothesis and of the racist ideology that it was
believed to justify. While it is, of course, too late now to change the course of
history, we believe that Indo-Europeanists and comparative linguists in general
must learn from the past and try to prevent similar misinterpretations and misap-
plications of their findings in the future.
It is for this reason that we present a sketch of the history of ideas relevant
to the issue of Indo-European “race” and conclude with a brief outline of what
Indo-Europeanists can do to counter current ideologically motivated attempts to
use linguistic and other prehistory as a means for establishing the superiority of
one group and justifying the exclusion or suppression of others.
At the outset, it is probably useful to understand that the concept “race”,
supposedly defined in terms of robust physical and mental differences, has been
called into serious question by modern anthropologists. Wherever we look, pure
“races”, as commonly postulated in the 19th and early 20th centuries, simply do
not exist. Rather, different areas show varying degrees of admixture of different
types of physical appearance (what are technically called phenotypes), one area
typically blending into the next one which merely exhibits a slightly different
admixture. Archaeological research, based largely on skeletal evidence, shows
that this is not just a recent phenomenon. As it is used today, the term “race” has
no biological validity but is simply a cultural construct (although, as the race-
based classification of US citizens shows, a very powerful one). It is for this reason
that we enclose the term in quotation marks throughout this discussion.
The first tentative step in the direction of associating Indo-Europeans or their
languages with some kind of “racial” differences seems to have been made in
1808 by the German scholar Friedrich Schlegel. Although his work has many other
facets, what is most important in the present context is his claim that the Indo-Eu-
476 Linguistic palaeontology: Historical linguistics, history, and prehistory

ropean languages are unique in their structure, and more nearly approach per-
fection than any other human languages, including the Semitic ones – languages
that Schlegel considered to have an “animalistic” origin. From the perspective
of modern linguistics, Schlegel’s claims about the structural superiority of the
Indo-European languages are naive at best, unsupportable by empirical evidence,
and exceedingly ethnocentric; and Schlegel later changed his view on the Semitic
languages. But a potential seed had been planted for the claim that Indo-Euro-
pean languages and their speakers are in their very essence superior to other lan-
guages and their speakers, including the Semitic ones.
It remained to Christian Lassen to propose in 1847/1867 the hypothesis that
the Indo-Europeans were white, that their homeland was outside of India, and
that the “Aryans”, i. e. the speakers of Indo-Aryan, migrated into India from
the northwest. Lassen was no Indo-Europeanist but a philologist, and he drew
support for his claims, not from language, but from the “racial” observations and
classifications of the peoples of India by western, mainly British anthropologists.
Like many of his contemporaries, Lassen was virtually obsessed with “race”, skin
color, and ethnocentric judgments about the relative beauty of different Indian
“races” (considering those groups most beautiful that come closest to Europeans
in their appearance). He further echoed Schlegel’s claims about the structural
superiority of Indo-European over any other languages, including the Semitic
ones. But beyond that, he asserted – without being able to furnish even a shred of
substantiation – that the Indo-Europeans were (and are) of superior moral char-
acter compared to the Semites, whom he characterized as egocentric, exclusivist,
and intolerant. It is Lassen’s perspective, not that of Schlegel, that is referred to
by later “cultural historians” preaching “Aryan” superiority, including Gobineau
and Chamberlain, people whom Hitler referred to as part of his intellectual herit-
age. (Gobineau’s racist perspective was refuted by August Friedrich Pott – a rare
example of an Indo-Europeanist taking an explicit stand against the racist misap-
propriation of the theory of Indo-European relationship.)
The final step, which appeared to provide conclusive textual evidence that
the Indo-Europeans, or at least the (Indo-)Aryans, were white-skinned and blond,
was made in a book by the German Indologist Heinrich Zimmer, published in
1879. Zimmer’s major focus was on skin color; his comments on blondness were
more in the nature of off-hand remarks and were developed more fully by later
authors.
Zimmer found evidence in the oldest Sanskrit text, the Rig Veda, for a black-
white division between the (Indo-)Aryan invaders of India and the indigenous
population, the dāsas or dasyus. Consider for instance the following verse, which
seems to make a clear distinction between the ‘dark skin’ of the dāsas or dasyus
and that of the āryas.
 Genetics, genomics, and “race” 477

(12) ā́ryaṁ prā́vad … svàrmiḷheṣv … |

… tvácaṁ kṛṣṇā́m arandhayat
‘He (Indra) helped the … ārya … in the battles for the sun(light). … he
made … the black skin (of the dāsa/dasyu) subject [to Aryan control].’

In addition, Zimmer observed that several Rig-Vedic deities are occasionally

referred to as hari-keśa and hari-śmaśāru, lit. ‘yellow-haired’ and ‘yellow-bearded’,
terms which he interpreted as meaning ‘blond-haired’ and ‘blond-bearded’ and
which were subsequently interpreted as indicating that at least some of the āryas
were blond.
Even if you have not grown up in the 19th or early 20th century, when “race”
was highly prominent in the minds of most westerners (and many others, too),
you may find the passage above to be quite persuasive, especially in light of its
reference to the tvácaṁ kr̥ ṣṇā́m ‘black skin’ of the dāsas/dasyus. And because of
Zimmer’s scholarly reputation as a specialist on the Vedic texts, Indo-European-
ists accepted his interpretations, and concluded that since even in the farthest
eastern realms there was evidence for light skin color and blond hair, the Indo-Eu-
ropeans – or at least some of them – must have been white, blond, and hence
also blue-eyed. And this conclusion became part of the standard fare of Indo-Eu-
ropean linguistics, both in Germany and elsewhere, without being subjected to
the critical scrutiny that other theories about the Indo-Europeans were routinely
exposed to. Even an earlier edition of our book reflects this interpretation.
True, like most Indo-Europeanists, that edition raised important questions
about whether it is legitimate to conclude that all Indo-Europeans were white
(etc.) or that they differed significantly in phenotype from their non-Indo-Euro-
pean counterparts. In fact, even during the height of the Nazi regime, in 1942,
the German Indo-Europeanist Hans Krahe raised similar questions; and in a set
of scholarly volumes published in 1936, well after the take-over of the German
university system by the Nazis and their henchmen, Indo-Europeanist con-
tributors expressed strong reservations about the equation of language with
race. (By contrast, a large number of articles in the same volumes, by anthro-
pologists, prehistorians, and scholars focusing on various aspects of “culture”,
contain numerous and extensive claims about the “Nordic race” of the Indo-
Europeans.)
Unfortunately, Indo-Europeanists did not go beyond voicing such reserva-
tions and they failed to examine critically the validity of Zimmer’s interpretations.
Had they done so, they might not have been able to change the views of those
who were ideologically firmly committed to the idea of “racial superiority” of
the “Nordic” Indo-Europeans and their most faithful descendants, the Germanic
peoples, but they could have deprived such claims of their apparent scholarly
478 Linguistic palaeontology: Historical linguistics, history, and prehistory

foundation and could, therefore, have been able to dissuade those who were
uncommitted from giving credence to the racist claims of the ideologues.
In fact, closer examination of the above passage and of some nine additional
passages (which are not always as explicit) casts doubt on the validity of Zimmer’s
interpretation. In every single passage that provides enough context, the ‘black/
dark’ color of the dāsas/dasyus contrasts, not with a white skin of the āryas, but
with the sun or the light that they possess or seek to possess. Consider in this
regard the svàr- of the first line of (12), a word unambiguously referring to the
sun (and in fact cognate, in a complicated way, with English sun). This finding
suggests that the term “black/dark” here does not refer to skin color, but rather to
the perhaps universal tendency to equate black or dark, the color of the dangerous
night, with evil persons or forces, and white or light, the color of daylight, with
good ones. (In modern times, compare e. g. the white hats of the “good guys” and
the black hats of the “bad guys” in Western movies.) As a matter of fact, given that
the struggle between dark/evil and light/good forces is a major theme of the Rig
Veda, this appears to be the better explanation. Even the term tvácaṁ ‘skin’ does
not prove that the passage in (12) refers to differences in skin color, for the term
is used elsewhere in the Rig Veda for any covering, including the surface of the
earth; and in a telling metaphor, the plants are referred to as the ‘body-hair’ (on
the skin) of the earth.
The case is even weaker for “blondness”. The same fire and sun-related
Gods that are referred to as hari-keśa and hari-śmaśāru ‘yellow-haired’ and ‘yel-
low-bearded’ are also designated as hiraṇya-keśa and hiraṇya-śmaśāru ‘gold-
haired’ and ‘gold-bearded’, and some of them are further characterized as hiraṇya-
bāhu ‘gold-armed’. Unless we are ready to accept blond-armed beings, we must
conclude that the terms ‘yellow’ and ‘gold’ in both sets of epithets refer to the
fiery, bright golden color and nature of these deities.
Significantly, the Vedic passages in question are the only textual evidence
suggesting that the Indo-Europeans may have conceived of themselves as “white”
or as lighter-skinned than their non-Indo-European opponents. But as we have
seen, that evidence fails to be cogent, and this ipso facto deprives the “racial”
and racist intellectual precursors of Nazi ideology of their apparent scholarly
foundation. Unfortunately, traditional Indo-Europeanists failed to engage in
the necessary critical reexamination of the evidence. This suggests one impor-
tant lesson for Indo-Europeanists and comparative linguists in general: Do
not restrict your approach of strict scrutiny of the evidence to “purely linguis-
tic” issues, but extend the approach also to other, “softer” issues – especially
those that have the potential for misuse by ideologues who try to establish their
own superiority over others by appealing to linguistic and textual history and
prehistory.
 Genetics, genomics, and “race” 479

There is yet another important lesson to be learned. As noted earlier, Indo-Eu-

ropeanists tended to be rather cautious about purely “racial” or even racist inter-
pretations of the Rig-Vedic evidence. This caution at least in part reflected the
increasing realization among some of the contemporary anthropologists that
there are no “pure races” and that, even if we accept the concept of “race” as
having some kind of physiological validity, not even the early Germanic people
were “racially pure”. Interestingly, some of the Nazi “race scientists” were aware
of this problem, too, and they tried to deal with it by resorting to the nebulous and
empirically unverifiable – and hence unchallengeable – concept of the folk spirit
or folk soul (“Volksgeist” or “Volksseele”) which was considered to define the
Indo-European and especially the Germanic “race” as superior to other “races”.
(In this regard these people were, of course, close to Schlegel’s and especially
Lassen’s earlier views.) We may, from our present-day perspective, find this neb-
ulous, “spiritual” perspective utterly ridiculous; but in the early 20th century it
found great support in a strong anti-scientist scholarly movement which rejected
as “positivist” the scientific approach and methodology of the mainstream of
Indo-European linguistics, with its emphasis on testing the validity of hypotheses
against empirical evidence. Not that all of these scholars would have subscribed
to the Nazi ideology; in fact, many did not and had to flee, or wound up in the con-
centration camps. However, by rejecting scientific, empirical methodology and
by accepting an unverifiable, non-empirical approach, they deprived themselves
of the means to demonstrate that their interpretations were preferable to those of
their racist opponents, they overestimated their ability as scholars to persuade
ordinary people of the correctness of their views, and they vastly underestimated
the ability of racist ideologues to reach the masses. It may be worth mentioning
that the lesson that we need to rely on empirically verifiable or challengeable
evidence is relevant not only for Indo-Europeanists and other comparative lin-
guists, but for any other scholarly endeavor, especially in cases where nebulous,
loose hypothesizing can be exploited by ideologues to support their own racist or
exclusionist agenda.

9.2 Recent findings in genomics

Because of the way that notions like phenotype and “race” had been misused
in the past, many linguists and other scholars do not always feel comfortable
with recent genomic research, which tends to make broad generalizations regard-
ing the spread of different human groups out of Africa into Eurasia and beyond.
What is important to note, however, is that genomic research is not driven by
any preconceptions and biases regarding the supposed superiority of one group
480 Linguistic palaeontology: Historical linguistics, history, and prehistory

over another. Moreover, genomics is a fast-moving field where the findings of one
group of scholars are being challenged by other researchers almost immediately
after publication. As in other fields, including linguistics, this lively pattern of
what might be called academic combat serves both to advance the field and to
keep scholars honest. And as in other academic fields, findings are always con-
tingent, to be challenged or overturned based on new evidence or interpretation
of the evidence.
What is especially important are the advances in palaeogenomics which in a
number of cases have overturned accounts based solely on the genomic evidence
of modern populations.
As we have seen, palaeogenomic research has led to the finding that in addi-
tion to the earliest, hunter-gatherer population and the later advent of agricultur-
alists from the Fertile Crescent, at least one other movement must be recognized,
that of pastoralists from the Eurasian Steppes.
Certain specific findings are especially remarkable, since they challenge
wide-spread beliefs about the “racial” characteristics of Europeans. Only the first
layer of human settlement, that of the hunter-gatherers, had blue eyes – but also
dark skin. Lighter skin – with brown eyes – is characteristic of the later new-
comers, whether agriculturalists from the Fertile Crescent or pastoralists from the
Eurasian Steppes. Moreover, there is no genetic link between blue eyes or light
skin and blondness. In short, the “Nordic” ideal of blue eyes, blond hair, and light
skin may be a very late development, specific to Europe, and in no way a feature of
the Indo-European Urheimat – whether in the Eurasian Steppes or, less likely, in
Anatolia. This, of course, should be cold comfort for neo-Nazis and others believ-
ing in “Aryan”, white European supremacy.

10 I deology vs. comparative linguistics and

­linguistic palaeontology
As we have seen in § 8.1, the comparative-historical demonstration that the Indo-Eu-
ropean languages are related, combined with claims by 19th-century philologists
regarding the “racial” identification (or superiority) of the “Aryans”/Indo-Europe-
ans, has served as an alleged scientific foundation for European racism, including
the extreme antisemitic racism of the Nazis. Some of the attempts to locate the
Indo-European Urheimat in Central Europe (see § 8.2) may have been driven by
Central European/German ethnocentrism; and Jones’s locating the original home
on the Iranian Plateau (§ 8.1) appears to have been informed by biblical considera-
tions. There is thus ample evidence that the significance of comparative-historical
 Ideology vs. comparative linguistics and ­linguistic palaeontology 481

Indo-European linguistics and linguistic palaeontology goes far beyond the small
circle of Indo-Europeanists and general historical linguistics and is of acute inter-
est to various ideologically motivated groups and individuals.
Let us now look at some other, recent cases in which comparative-historical
linguistics “meets” ideology.
The fact that chariots seem to have been first developed in Sintashta, in
southern Russia close to the border with Kazakhstan, has been taken by some
linguists and archaeologists to indicate that the Indo-Iranians and maybe all
Indo-Europeans originate in the area of Sintashta. The Indo-Iranian theory has
been especially advocated by the eminent Russian archaeologist, Elena Efimovna
Kuz’mina. Russian nationalists of various stripes eagerly seized upon this identifi-
cation as an indication that the original home of Indo-Europeans or “Aryans” was
in Russia and that the Slavs are the major inheritors of this “Aryan” ancestry. The
large, impressive urban site of Arkaim (https://en.wikipedia.org/wiki/Arkaim)
has been taken to be especially significant in establishing Russia’s “glorious
past”, and in 2005 the site was graced by a visit of President Putin. However, the
Sintashta culture area is characterized by built-up town settlements, a fact that
creates problems for identifying it directly with the Indo-Europeans, especially
with the Indo-Iranians, because of the strong evidence (especially in the case of
the Indo-Iranians) that they practiced transhumance and hence had no perma-
nent built-up settlements; see § 6 above. Moreover, the lack of a common word
for ‘chariot’ in PIE II makes identity of the Indo-Europeans and the people of Sin-
tashta unlikely; see § 7.1 above. (It is for these reasons that in our earlier discussion
we hedged our bets by saying that the evidence of horse domestication and the
development of two-wheeled horse-drawn chariots favors an Urheimat at or near
the area(s) where these developments took place.)
Still, the best evidence points to an area in or near the Eurasian steppes and,
at the time of the development of chariots, near the Sintashta cultural area. This,
in turn, requires the assumption that Indo-European speakers had migrated to
all the other areas in which they are found at the beginning of recorded history.
The notion that the “Aryans”, in the sense of speakers of Indo-Aryan, migrated
into India/South Asia has especially given rise to a great variety of ideological
responses. During British rule, the migration was commonly characterized as an
invasion, and the “Aryans” were portrayed as white, European conquerors who
brought “civilization to the natives” – the “Aryan Invasion Theory” (AIT). This per-
ception naturally did not please the “natives”, and there were counterproposals
that the “Aryans” had always been in India and that other Indo-Europeans must
have migrated out of India – the “Out of India Theory” (OIT). The fact that William
Jones had been in the service of the British East India Company was even taken to
indicate that the relationship between Sanskrit and the Indo-European languages
482 Linguistic palaeontology: Historical linguistics, history, and prehistory

outside India was developed by the colonialists as a means to somehow denigrate

and divide the people of India.
At the same time, advocates of a separate cultural and political existence of
the Dravidians, i. e. the speakers of Dravidian languages, heartily accepted the
AIT as evidence that the “Aryans” came from the outside and suppressed the
Dravidians, and that Dravidians must now rise up against “Aryan” rule. While
the Dravidian anti-“Aryan” perspective has abated in recent times, proponents
of “Aryan” indigenism continue to push the OIT. Publications advancing “Aryan”
indigenism sometimes are content with rejecting the AIT as an ideologically moti-
vated hoax, but there have also been attempts to provide linguistic arguments.
These arguments, however, are problematic for various reasons, the most impor-
tant of which is the rejection (or ignorance) of the regularity of sound change
principle, for which see Chapter 4, § 4 and especially § 7.
Other proponents of “Aryan” indigenism simply reject the entire methodology
of comparative-historical linguistics, and one such proponent claims that recon-
structed proto-languages (specifically Proto-Dravidian) are just a theory and that
nobody ever has seen such a proto-language. The claim, made by somebody char-
acterized as a former NASA engineer, ignores that other scientific insights are also
“mere theories”, such as the theory of Black Holes; but that as in comparative-his-
torical linguistics, rejecting such theories makes it impossible to account for the
broad range of phenomena that the theories explain.
The approach of rejecting comparative-historical linguistic methodology is
not limited to India but is found, for instance, in Hungary, where nationalists
reject the classification of Hungarian as a Uralic (or Finno-Ugric) language as an
invention of scholars in the service of the Habsburgs’ attempt to belittle the Hun-
garians. Hungarian nationalists instead prefer a more “glorious” affiliation with
the Huns, or even with the Sumerians.
To some extent such skepticism seems to find support in the fact that even
some linguists have questioned various aspects of comparative-historical linguis-
tics.
The Italian scholar Angela Marcantonio, for instance, has questioned the
validity of both the Indo-European and the Uralic language families. One of her
main arguments is that the methodology suffers from circularity: Reconstructions
require the assumption of an inordinate amount of linguistic changes to derive the
daughter languages, and without these changes the reconstructions would fall
apart. This ignores the fact that even to derive Modern English from Old English
requires an “inordinate” amount of linguistic changes; and similarly for all other
languages with attested earlier stages. Moreover, if we reject the comparative
method, we are left without an explanation for the extensive similarities (and
differences) between the Indo-European languages, and the lack of such exten-
 Ideology vs. comparative linguistics and ­linguistic palaeontology 483

sive similarities between Indo-European and other language families. Even so,
Marcantonio’s claims have been embraced by Hungarian nationalists and “Aryan”
indigenists.
Some historical linguists, too, have expressed doubts about whether linguis-
tic reconstructions can make any claim to being realistic. This perspective has
obvious implications for linguistic palaeontology – if the reconstructions are not
realistic, how can we use them as indicators of prehistoric culture, society, or even
migrations?
These skeptics argue that we can never hope to fully reconstruct the ancestral
language, since lexical items and grammatical forms tend to become obsolete.
The impact is so pervasive that many forms may be lost from many of the related
languages, leaving no evidence that might be used for reconstruction. (Consider
the rapid and extensive loss of horse-and-buggy terminology in recent times, after
the introduction of the internal combustion engine.)
Even in phonological reconstruction, the greatest success story of the com-
parative method, it is argued that we can never be certain about the precise pro-
nunciation of what we reconstruct as, say, voiced, or aspirated, or palatal. In fact,
the controversy over the glottalic theory shows how little we can be certain about
certain specific phonetic details of Proto-Indo-European sounds.
Finally, the skeptics argue, the comparative method by its very design has
to reduce all the variation found in the daughter languages to invariance. Real,
natural languages always have some variation: This may range from morphologi-
cally conditioned phonological alternations like Engl. sing : sang : sung to dialec-
tal variation such as [šikægo] : [šikɔgo] : [šikago] for Chicago. Languages without
variation, especially without different dialects, are unnatural. If the comparative
method forces us to reconstruct such languages, then our reconstructions by
necessity are unrealistic.
The skeptics conclude that the best we can say for reconstructions is that
they serve as convenient cover formulae, summarizing our understanding of the
linguistic relationship between given languages.
Now, it is perfectly true that our reconstructions are hypotheses and thus nec-
essarily somewhat hypothetical. It is also true that there is disagreement over
issues such as the glottalic theory. And as we have seen in some of the preceding
sections, there are also controversies over the reconstruction of particular lexical
items. But these disagreements and controversies concern matters of detail.
Lexical items whose reconstruction is controversial constitute a minute fraction of
the hundreds, even thousands of lexical items, roots, and morphemes that we can
reconstruct without difficulties – items such as the words for ‘father’, ‘mother’,
‘brother’, ‘sister’; or ‘eat’, ‘drink’, ‘sleep’; or even personal pronouns and the func-
tion word ‘and’. It is just that many of these well-assured items are not particularly
484 Linguistic palaeontology: Historical linguistics, history, and prehistory

interesting for linguistic palaeontology: What language does not have words for
basic concepts like these?
Even disagreements over issues such as the glottalic theory are not as serious
as they might appear. True, we may be arguing over whether particular sounds
were voiced unaspirated or whether they were glottalized. But we do agree on
the need to reconstruct three series, rather than just one or two. Further, where
most of the languages present a labial stop or a plausible outcome of such a stop,
none of us would reconstruct a glottal stop instead. There are many more aspects
of phonological reconstruction that we agree on – even the skeptics – than those
that we do battle over.
The claim that the comparative method by necessity eliminates all vestiges of
morphologically conditioned variation likewise is exaggerated. It is only through
the comparative method that we can postulate for Proto-Indo-European alter-
nations of the type *sengwh- / *songwh- / *sn̥ gwh- which are the source for Engl.
sing : sang : sung; note also the variation in the word for ‘horse’ (*h1eḱwos/h1ḱwos)
and in the root underlying *kwekwlos ‘wheel’ (*kwel/kwl). The comparative method
even furnishes evidence of dialectal variation in Proto-Indo-European, such as
the centum : satem division mentioned in Chapter 2, § 3.6 and Chapter 11, § 3.
Here again, specialists may differ on matters of detail. And it is certainly pos-
sible, even likely, that some aspects of Proto-Indo-European variation escape us –
because the evidence for the variation has been irretrievably lost. But this does
not detract from the fact that we are able to reconstruct aspects of the phonologi-
cal and dialectal variation characteristic of natural languages.
More than that, the American linguist Robert A. Hall, Jr. has put the compar-
ative method to the test by doing reconstruction in Romance and comparing the
results with Latin. Making allowances for the fact that the Classical Latin of Caesar
and Cicero is not identical with the vernacular Latin from which the Romance lan-
guages descended, the results are impressive. Most striking is the fact that Hall
was able to reconstruct something approximating the Latin length distinction in
vowels – even though all Latin long vowels had become short in Romance! True,
the evidence of the Romance languages does not permit us to define the feature
distinguishing the two vowel sets as length; it could conceivably be characterized
by other terms such as “tense”. In this sense, then, we are in a situation similar
to the glottalic controversy. But the fact remains that the reconstructed contrast,
however defined, closely corresponds to the length contrast of Latin and thus is a
realistic one, not just a figment of one scholar’s imagination.
There are also cases where Indo-Europeanists have reconstructed phonolog-
ical elements and subsequent discoveries have confirmed these reconstructions.
For instance, dialectal differences in the Greek outcomes of the PIE labiovelars as
p, t, etc. led scholars to postulate an early stage of Greek in which the PIE labiove-
 Conclusions and outlook 485

lars had been preserved. When the Mycenaean inscriptions were deciphered as a
very early form of Greek (Chapter 3, § 3.3), it turned out that there was a separate
series of symbols corresponding to the Indo-European labiovelars, confirming
what had previously been “only a theory” – the retention of labiovelars in an
early stage of Greek.
On balance, a proper assessment of the value of our reconstructions would be
that they approximate some prehistoric reality, even if some aspects of that reality
may escape us. Where different scholars disagree with each other regarding such
issues as the glottalic theory and linguistic palaeontology, their disagreement is
not so much over the goal of approximating reality, but over which road leads
more effectively to that goal. In fact, if they were resigned to considering recon-
structions mere “convenient cover formulae”, there would be no point to such
disagreements: Any cover formula would be as good as the next, as long as it
manages to “sum up the facts”.

11 Conclusions and outlook

In this chapter we have examined a variety of ways in which the methods and
results of historical and comparative linguistics can be applied to issues outside
linguistics proper. By looking at Proto-Indo-European, we have seen that the
comparative method permits us to go beyond the mere reconstruction of sounds,
morphology, syntax, and words, and to reconstruct aspects of prehistoric culture
and civilization. As a consequence we have realized that the “footprints on the
unknown shore” are, to some extent, ours: Human beings tend to be the same
no matter where or when they live. Some aspects of Indo-European culture, such
as the limited use of agriculture (if any) or of metals, will remain to us “a foreign
country”. But even if the Indo-Europeans “did many things differently”, we may
have become a little more familiar with them – through the fruits of the compar-
ative method.
At the same time, we have come to know that there are remaining controver-
sies, especially regarding the time and place at which Indo-European was spoken,
with two major contending theories, the Anatolian and the Eurasian Steppe
hypotheses. But perhaps there is some benefit in having traveled the road to this
realization. Examining the different proposals and hypotheses in some detail has
given us a better understanding of the arcane world of the Indo-European special-
ists and should prepare us for a better appreciation of new claims that we might
come across in the future, whether they involve Indo-European and the Indo-Eu-
ropeans or other proto-languages and their speakers.
Chapter notes and suggested readings
We have deliberately written this book with no footnotes in the text and without
a large number of references worked into the presentation, for we felt that they
might be more of a distraction than a help. However, such notes and references can
serve a very useful purpose, by pointing you in the direction of further readings
if you want to explore certain issues more fully; by giving you information on the
standard sources of the ideas we present and the findings we discuss; by generally
acknowledging contributions that others have made to our understanding of the
processes of language change over the years; or by making additional comments
of interest that do not quite fit into the main text. In these chapter notes, we give
relevant additional information, mainly keyed to publications listed in the Refer-
ences, with reference to specific chapter sections where possible or appropriate.
Many of these publications are at an advanced level, representing the find-
ings of professional linguists investigating language change. Our expectation is
that seeing them listed here will give you a feel for the full range of scholarship
on language change that has provided the basis for what we have distilled in this
textbook.
At the same time, many topics we have covered lend themselves particularly
well to popular and thus highly accessible treatments (e. g., writing, etymology,
the origin of language, linguistic palaeontology). Indeed, some of the works cited
below come from sources such as Scientific American or the publications of the
British Museum, and are designed for the interested, intelligent, and intellectu-
ally curious non-specialist reader. Thus, we specifically note those works which
are appropriate for beginners in the field.
As you can see, historical linguistics is a vast field and there is a great amount
of information available; these notes will give you an idea of where to start looking
to go beyond what we have given you.
One type of resource we do not generally mention here, because advances
in technology have made specific listing unnecessary, are websites. As opposed
to the situation in 1995 when we were putting together the notes for the first
edition, there is now, some 23 years later, a proliferation of websites and elec-
tronic resources on individual languages and language families, on etymology,
on writing systems, on language relationships--you name it; and they can all be
found rather easily with the powerful search engines available these days. As with
any resource, one has to approach internet sites with a critical eye, since they are
of uneven quality, and may present idiosyncratic or ideologically biased points
of view. But there are plenty of excellent materials to be found and we encourage
you to explore on your own, or with some guidance from your instructors. Still,
we mention occasional websites where they have something particular to offer.

https://doi.org/10.1515/9783110613285-019
 Chapter notes and suggested readings 487

Chapter 1.
Introduction
Since we survey briefly in this chapter what is in store in the rest of the book, we
start these notes with reference to some general works on historical linguistics.
Among these are classic works that are basically manifestos on the nature of lan-
guage and so treat language change as an essential part of understanding lan-
guage in general. Others are textbooks more in the familiar modern sense, though
often aimed at student populations with backgrounds in linguistics quite different
from that assumed in this textbook.
Among the classic works are Bloomfield 1933 ([1965]), Jespersen 1921, Paul
1920, Sapir 1921, de Saussure 1916.
Textbooks at the beginning level include Aitchison 2001, Arlotto 1972, Crowley
& Bowern 2010, Trask 1994, 1996, and Campbell 2013. Trask 2000 and Campbell
& Mixco 2007 are useful glossaries of terms and concepts in historical linguistics,
and Luraghi & Bubenik 2010 has brief and readable encyclopedia-like articles on
a wide range of key historical topics.
More advanced introductions, presupposing greater familiarity with linguis-
tics, include Fox 1995 (focus on reconstruction), Goyvaerts 1975, Jeffers & Lehiste
1979, Lehmann 1992, Sihler 2000, and Sturtevant 1917.
Other even more advanced introductions and handbooks are Anttila 1988,
Bynon 1977, Hale 2007, Hock 1991, Hoenigswald 1965, Labov 1994, 2001, 2010,
McMahon 1994, and Ringe & Eska 2013. Somewhat advanced historical linguis-
tics texts in German include Boretzky 1977, Sternemann & Gutschmidt 1989, and
Szemerényi 1989a.
Additionally, though they are aimed at quite advanced students and profes-
sional practitioners of historical linguistics, we mention a few anthologies that
offer surveys of the field: Polomé 1990a, Jones 1993, Bowern & Evans 2014, Joseph
& Janda 2003, and Janda, Joseph, & Vance 2019; also, Dawson & Joseph 2013 con-
tains some one hundred of the most important articles in historical linguistics,
mostly from the 20th century. All of these works listed here offer many more exam-
ples of the types of changes given throughout this book.
In this book, many of our examples come from English. We note here just a
few of the extremely large number of useful sources on English and its history:
Barber 1976, Bloomfield & Newmark 1963, Crystal 1988, Crystal 1995 (organized in
a highly readable and innovative way), and Pyles & Algeo 1982 (somewhat tradi-
tional in style but solid).
Similarly, since our references to change in sign languages and to their history
are spread here and there across the chapters, we note several relevant, and gen-
488 Chapter notes and suggested readings

erally highly accessible works on sign languages and their diachrony: Battison
1978, Frishberg 1975, Frishberg 1976, Lane 1980, Lucas & Valli 1990, Lucas, Bayley,
& Valli 2003, Perlmutter 1986, Stokoe 1974, Supalla et al. 2019.
As for phonetics, presented in the Appendix to this chapter, virtually any
introductory textbook in linguistics – and there are dozens – gives essentially the
same information and symbols. A standard text on phonetics itself, which goes
into far greater detail than is presented here, is Ladefoged 1993.
Finally, an excellent resource, aimed at a generalist audience, in which a
brief overview of linguistic concepts and terms given throughout this book can be
found, is Bright 1992, updated and revised as Frawley 2003; see also Brown 2006.

§ 1. There is other evidence that English is not standing still besides the somewhat
“archaic” feel that Arlo Guthrie’s words have or innovations like free-standing
ish (see https://slate.com/human-interest/2014/06/ish-how-a-suffix-became-
an-independent-word-even-though-it-s-not-in-all-the-dictionaries-yet.html for
an interesting perspective on this usage with citations from the 1980s and early
2000s). Relatively recently, the word uber, a borrowing from German, has become
a handy prefix in English meaning ‘over and above, very’ (one-upping mega,
as it were) as in That guy is uber hot (giving rise, with the founding of the ride
company Uber, to puns about being uber-careful/Uber careful when using that
service). Similarly, because, which for centuries was a composite preposition and
required the use of of with it, as in Because of linguistics, I now enjoy my studies
(a result of its etymology, from by cause of), has in recent years begun to be used
as a preposition proper on its own, as in Why do I enjoy my studies now, you ask?
Because linguistics.

§ 3: We are using the spelling Romanian here and throughout, as opposed to

Rumanian. The spelling with o is preferred by many native speakers of the lan-
guage, as it emphasizes its Romance-language origin and connection with ancient
Rome; it is thus a matter of language attitudes where, as we note throughout this
book, speakers’ sensibilities rule.

§ 5: There is a t-shirt one can buy that says “English is weird. It can be understood
through tough, thorough, thought though.”
 Chapter notes and suggested readings 489

Chapter 2.
The discovery of Indo-European
§ 1. Pedersen 1959 gives an intriguing and readable history of the development
of historical linguistic research, and especially of Indo-European studies (and
includes pictures of the scholars known otherwise just by name).
The term “vernacular” is an important one that occurs frequently in this book
(e. g. in Chapters 1, § 2; 10, § 2; 12, § 1). It refers to a form of speech that is spoken by
“ordinary people” and thus has lower status than a “prestige” (or more properly,
“high prestige”) variety with which it coexists. Depending on the circumstances,
a vernacular may be a different language from the prestige variety (such as early
forms of French, Spanish, English, or German in medieval Europe, which were
vernaculars compared to the prestige language, Latin); or it may be something
more like a nonstandard dialect (such as nonstandard forms of English compared
to Standard English).

§ 2. The field of IE studies has such a vast literature that we cannot do it justice
here. The following all provide some general information on the individual lan-
guages and the proto-language: Meillet 1937, Lockwood 1969, 1972 (these two
being designed particularly for interested nonlinguists), Krahe 1970, Lehmann
1993, Baldi 1983, Meier-Brügger 2003, Mallory & Adams 2006, Clackson 2007 (a
more advanced publication), Fortson 2010, and Beekes & DeVaan 2011.
Classic treatments of Proto-Indo-European grammar, in whole or in part,
include Brugmann 1897–1916 and 1904, more recently Watkins 1969 (on the IE
verb) and Mayrhofer 1986 (on phonology); see also Sihler 1995 (focus on Greek
and Latin). The 1879 publication by de Saussure, published when the author was
just twenty-one years old, eventually revolutionized the view of the Proto-In-
do-European vowel system (this is translated in part in Lehmann 1967). Gam-
krelidze & Ivanov 1984, 1994 give a still controversial reassessment of much of
the classical view of Proto-Indo-European grammar (see Chapter 16, § 7 on their
view of Proto-Indo-European phonology).
Buck 1949 is an interesting source on the Proto-Indo-European lexicon,
organized in terms of words for related concepts (like a thesaurus); Gvozdano-
vić 1992 is an advanced compendium on one sector of the Proto-Indo-European
lexicon, the numerals (see also Winter 1992 and Justus 1988). Finally, Cardona,
Hoenigswald, & Senn 1970 and Polomé 1982 offer important collections of high-
level papers on various aspects of Indo-European. Mention can be made here too
of Mallory & Adams 1997 and 2006, encyclopedic works with article-length entries
on hundreds of topics relevant to Indo-European studies, broadly conceived.
490 Chapter notes and suggested readings

§ 3. Beyond the individual chapters in Ramat & Ramat 1997 and now Klein,
Joseph, & Fritz 2017–2018, there are also useful reference works on each of the
branches of Indo-European, often with surveys of the important languages and/
or bibliography, including the following:
Celtic: MacAuley 1992, Ball & Fife 1993, and Gregor 1980
Latin (Italic): Palmer 1954 and Baldi 1999 (Latin), Wallace 2007 (non-Latin ancient
Italic), and Posner 1966, Hall 1976, and Ledgeway & Maiden 2016 (Romance)
Germanic: König & van der Auwera 1994 and Robinson 1992; see also Haugen
1982 for the Scandinavian languages and Antonsen 1975 for early runic, and
Stearns 1978 on Crimean Gothic.
Slavic: Comrie & Corbett 1993
Lithuanian (Baltic): Fraenkel 1950, Senn 1942
Baltic-Slavic or Balto-Slavic: Szemerényi 1957
Albanian: Hamp 1972 (a primarily bibliographic essay), Lloshi 1999, Orel 2000,
Demiraj 2006
Greek: Palmer 1980, Horrocks 2010, Christidis 2007
Iranian: Schmitt 1989a, Korn 2016
Indo-Aryan: Burrow 1973 (Sanskrit, with information on Indo-Iranian in general),
Hock 2016, Cardona & Jain 2003 (modern Indo-Aryan), see also Masica 1991.
Anatolian (Hittite): Ceram 1973 and Gurney 1954; unfortunately, no generally
accessible treatment of the Anatolian languages exists; Melchert 1994 is a
fairly technical, though comprehensive treatment of Anatolian historical pho-
nology.
For Armenian and Tocharian there are no generally accessible surveys.
On the dialectally highly diverse foundations of Classical Armenian see Winter
1966.
On Etruscan, see Wellard 1973, intended for a general audience; Wallace 2008 is
an excellent manual for learning Etruscan, together with inscriptions to read.

Chapter 3.
Writing: Its history and its decipherment
General references: Diringer 1962, Friedrich 1957, Gelb 1963, and Nakanishi 1992,
all designed for a general audience; somewhat more scholarly in orientation are
Carter & Schoville 1984, Coulmas 1989, 1996, 2002, Daniels 1990, Daniels & Bright
1995, Günther & Ludwig 1994–1995, Jensen 1958, 1970, Trager 1974.
 Chapter notes and suggested readings 491

§ 2.1. On orality in early India see Falk 1990 and 1993 (the latter is an advanced
scholarly survey of the issue). Roots was published in 1976 by Doubleday & Co. It
is a powerful testimony to the vitality and accuracy of oral traditions, since Haley
reports that when he finally visited the West African village that his ancestors
came from, he heard a griot recite the family history. He describes the moment
thus: “This man [the griot] whose lifetime had been in this back-country African
village … had just echoed what I had heard all through my boyhood years on my
grandma’s front porch in Henning, Tennessee”.

§ 2.2. Denise Schmandt-Bessarat has initiated the important research into the role
of tokens in the prehistory of writing, as described, for instance, in her 1992 book.

§§ 2.3–6 The works by Diringer, Gelb, Friedrich, Carter & Schoville, and Jensen
cited in the General References above all treat the ancient Near Eastern develop-
ments in considerable detail and discuss the development of the alphabet.
On cuneiform writing, see Walker 1987; Chiera 1938 describes what has been
learned from the cuneiform tablets about life in this part of the world in ancient
times. On the Egyptian hieroglyphs, see Davies 1987. The writing system of the
Mayans, as well as its decipherment, is described in Coe 1992, Houston 1989,
Johnson 2013, and Stuart & Houston 1989. All of these publications are aimed at
a general audience.
On the development of the Persian syllabary, see Schmitt 1989b. Daniels &
Bright 1995 deals with Semitic “abjads”. Senner 1989 is an interesting collection
of papers, dealing in part with the spread of writing and the alphabet. Jeffery 1961
offers an excellent account of the development of the Greek writing system out of
Semitic originals. A very accessible discussion of the Germanic runes is found in
Page 1987. The issue of the Germanic “feather runes” and their possible relation
to the Old Irish Ogham script is discussed in Pedersen 1959.

§ 3. Cleator 1962, Friedrich 1957, Gelb 1963, Gordon 1982, and Pedersen 1959 give
interesting details on the decipherments of cuneiform and the Egyptian hiero-
glyphs, and of other decipherments. See also Walker 1987 on cuneiform, Chad-
wick 1958, 1987 on the decipherment of Linear B, Andrews 1981 on Egyptian hier-
oglyphs, and Coe 1992 and Johnson 2013 on Mayan writing, all quite accessible
works.

§ 4. On the interpretation of written records (philology) in general, see the many

articles in Fisiak 1990, some of which also address the specific question of pho-
netic interpretation. See Sturtevant 1940 for a survey of evidence that helps us in
determining the pronunciation of ancient languages known only through writing.
492 Chapter notes and suggested readings

The problem of the Old English “digraphs” is discussed in Antonsen 1967, Kuhn
1961, Stockwell & Barritt 1961.

§ 5. Sampson 1985 is an excellent source on East Asian writing systems, and

writing typology in general, and Unger 2004 debunks some common myths
about Chinese/East Asian writing. Salomon 1998 is an excellent source on Brahmi
writing. Many of the works mentioned above, but especially Diringer 1962 and
Gelb 1963, as well as Daniels 1990, 2018, Daniels & Bright 1995, and Trager 1974,
give typologies of writing systems with discussions of their historical develop-
ment. Nakanishi 1992, a popular treatment, gives samples of the various writing
systems of the world. Fuller discussion of the different types is found in Trager
1974, Daniels & Bright 1995, and Günther & Ludwig 1994–95. It is worth noting
that there is a museum devoted to writing, and specifically the alphabet – the
Museum of the Alphabet, in Waxhaw, Tennessee. The website www.omniglot.com
bills itself as “the online encyclopedia of writing systems & languages” and has
samples of a wide range of alphabets, abjads, syllabaries, and such.

Chapter 4.
Sound change
§§ 2–3. Lehmann 1967 provides, in English translation with annotations, many
of the nineteenth century works that reported the important breakthroughs in
Indo-European linguistics mentioned here, including Grimm 1819–34, 1893,
Lottner 1862, Rask 1818, and Verner 1877 (Sir William Jones 1786, the source of
the quotation in Chapter 2, is also included). Bopp 1816 and Pott 1833–36 were
among the early explorations of IE grammar. Vennemann 1984 gives a novel inter-
pretation of the Germanic sound shifts (partly in accordance with the “Glottalic
Theory” of PIE consonants – see Chapter 16 § 7). On the Southern Bantu sound
shift, see Doke 1954. For other parallels to Grimm’s Law, see Sapir 1931 and Labov
1981.

§ 4. On taboo deformations, see Havers 1946. For a fairly advanced discussion

of the neogrammarian hypothesis, see Jankowsky 1990, and Labov 1981, 1994.
Osthoff & Brugmann 1878, the source of the epigram, is the classic statement of
the regularity hypothesis. Schuchardt 1885 is the most outspoken representative
of those who were opposed to the neogrammarian regularity hypothesis. On the
regularity principle as applied to American Sign Language, see Moser 1990.
 Chapter notes and suggested readings 493

§ 5. Additional examples of the various types of sound change can be found in

virtually any of the textbooks mentioned above in the notes to Chapter 1 (though
in some instances the terminology may be slightly different). An especially com-
prehensive treatment is found in Hock 1991, Chapters 3–8. For the phenomenon
of loss with compensatory lengthening see Hock 1986. Martinet 1964 is the stand-
ard source concerning chain shifts, and Labov’s highly influential work draws on
some of these insights and elaborates on them (see especially Labov 1994).
The Great Vowel Shift in English, along with Grimm’s Law, is one of the most
celebrated sound changes that have been studied, and the literature on this
subject is enormous; see Bloomfield & Newmark 1963, Pyles & Algeo 1982 or any
other work on the history of English (see the Notes to Chapter 1, above) for addi-
tional discussion of this change.
Hoenigswald 1964 discusses the issue of sporadic changes. The connection
of sporadic changes with speech errors goes back to Paul 1920. The aspiration
dissimilation now known as Grassmann’s Law was discovered by Hermann Grass-
mann, a noted mathematician whose work is still important in modern mathe-
matics; Grassmann 1863 is translated in Lehmann 1967.

§ 6.1–2. Jespersen 1941 and Whitney 1877 discuss the notion of change as improve-
ment not decay, while Zipf 1929, Mańczak 1987, and Phillips 2006 try to establish
a link between word frequency and sound change. The various neogrammarian
attempts to deal with the motivation of sound change are discussed in Paul 1920.
Arguments against the claim that linguistic change originates in early child lan-
guage learning are found in Bybee & Slobin 1982 and in Vihman 1980.

§ 6.3. Forerunners of Labov’s work are Gauchat 1905, Hermann 1929 (a follow-up
study on Gauchat), and Sturtevant 1917. Labov 1963 is the ground-breaking study
that led to the recognition that sound change is observable and is directed by
social factors. See also Labov 1965, Weinreich, Labov, & Herzog 1968 (an impor-
tant early position paper), and especially Labov 1994, 2001 for a full discussion,
at a very advanced level, of the nature of sound change.

§ 7. The attack on the regularity hypothesis reported here is that of Berlant 2008. –
The (unattested) individual-language forms in example (19) were generated out
of the reconstructed PIE forms by applying a large variety of different linguistic
changes that are possible in human language, avoiding any repetition of changes
in the same language.
494 Chapter notes and suggested readings

Chapter 5.
Analogy and change in word structure
Analogy has a directly psychological basis, inasmuch as it involves drawing a
relationship between two (or more) linguistic entities, whether separate but
semantically related words, forms linked together in a grammatical paradigm,
sound-alike words, or whatever. Anttila & Brewer 1977 has references to hundreds
of works on analogy, both linguistic and psychological. Joseph 1998 (updated in
Joseph 2019) provides a compendium of types of morphological change, with
extensive examples, though the discussion is at a somewhat high level. The three
studies on analogy (Anttila 2003, Dressler 2003, and Hock 2003) in Joseph & Janda
2003 address the nature of analogy from different theoretical perspectives – at a
rather advanced level, though with numerous relevant and illuminating exam-
ples; Fischer 2019 discusses iconicity and analogy, with attention to language
learning and general cognitive strategies.

§ 2.1. See Lipton 1991 for an entertaining look at collective terms of the type a pride
of lions. Winter 1989 presents some parallels in Welsh, Armenian, and Russian to
English zero-plurals.

§ 3.1. See for instance Winter 1989. Baron 1989 (especially Part III) is a highly
readable work with several examples of blends and related formations, especially
in advertising. The Yiddish rhyming pattern with shm- has parallels in Tamil,
Marathi, and other South Asian languages, Turkish, Bulgarian, and Greek, and
many other languages from India across central Asia and the middle East into
Europe (see Southern 2005). The Yiddish pattern may well be the result of a long-
range geographical diffusion (see also Chapter 8, § 1 for examples of the long-dis-
tance spread of words).

§ 3.2. Probably any parents you ask can give you examples of their children’s
reinterpretations and folk etymologies (and such examples provide material for
many popular comic strips), though these processes clearly are not restricted
just to children. Parker 1883 is a huge collection of forms created by these spo-
radic processes (though some may now be dated). Instances like the Gladly …
example have come to be known as mondegreens, based on 20th century writer
Sylvia Wright’s understanding, while a child, of the lyrics of a Scottish ballad,
“The Bonny Earl of Murray”. The song says that when the earl died, “they
laid him on the green”, i. e. buried him in the village green, but Wright under-
stood that as referring to a partner of his, “Lady Mondegreen”. An interesting
 Chapter notes and suggested readings 495

collection of such examples can be found at http://www.uh.edu/~mbarber/

mondegreens.html.

§ 4. A more detailed discussion is found in Hock 2003.; see also Menner 1937 for
examples from American English that, interestingly, are still current, and Crowley
& Bowern (2010: 204–205) and Campbell (2013: Chp. 4) for examples from lan-
guages other than English.

§ 6. For recent discussions of morphological change see Dressler 2003, Hock

2014, Joseph 2019. Many recent publications on morphological change focus on
“grammaticalization”, the origination of morphological elements and the extent
to which their emergence is a unidirectional development; see e. g. Hopper &
Traugott 2003, Traugott & Heine 1991, and Fischer, Norde, & Perridon 2004, and
for a comprehensive treatment of grammaticalization from every conceivable
angle see Narrog & Heine 2011.

Chapter 6.
Syntactic change
Much of this chapter deals with aspects of usage that are commented on routinely
by prescriptivists, who are interested more in what they feel speakers ought to
be saying than in what speakers actually do say. As indicated, there is a long
history of prescriptivism in English; see Crystal 1995 for a brief overview, as well
as any source on the history of English (see notes to Chapter 1 above). Pinker 1994,
Chapter 12 has a highly readable discussion of prescriptivism in contemporary
America. Many examples of syntactic change due to language contact are given
in Chapter 13 below.

§ 5. Hock 1982 is the source for the discussion of the role of auxiliary clitics in the
change from SOV to SVO word order.

Chapter 7.
Semantic change
Material for this chapter is drawn from Algeo 1990a, Ogden & Richards 1923, Stern
1931, Trier 1931, 1973, Ullman 1957, 1962, as well as personal research by H. H.
496 Chapter notes and suggested readings

Hock. Traugott & Dasher 2001 is a recent theoretical presentation of semantic

change from the perspective of “grammaticalization” theory (for which see the
notes to Chapter 5, § 4), with many revealing examples. Steinmetz 2008 is a highly
readable popular treatment of changes in meaning, aimed at a general audience.

§ 4. The notion of the arbitrary relation between linguistic form and linguistic
meaning was first emphasized by Ferdinand de Saussure (see de Saussure 1916).

§ 5.1. Crystal 1995 and Baron 1989 (Chapter 11) discuss “doublespeak” (euphe-
mism in officialese and in daily use). Baron 1989 (Chapter 18) has numerous
examples of punning in business names (you should be able to find examples in
your own area). See also Lutz 1989 and Cutts & Maher 1984, similarly aimed at the
general reader.

There are technical terms for some of the different sorts of metaphor illustrated
here; many people (ourselves included!) often have trouble distinguishing them,
but we mention them in this note in case you might have run into such “figures
of speech” in an English or literary analysis class: synecdoche is the designa-
tion of a thing or person by means of its most salient part (as in (19c) wheels for
‘car’); metonymy is the designation of a group of things or persons by means of
a word referring to something with which that group is habitually associated (as
in (20a) the pulpit for ‘clergy’); hyperbole is exaggeration or overstatement for
effect (as in (21a) terribly sorry, where there is nothing terrible about the apology);
and litotes is understatement, often with a negative involved (as in (22b) not
inconsiderable, meaning ‘a lot’), thus a case where two negatives linguistically
can make a positive (see Chapter 6, § 4 on double negation in Latin and in both
standard and nonstandard English).

§ 5.2. Pisani 1937 and Winter 1982 discuss ‘tongue’ in Indo-European; see Havers
1946 on the effects of taboo in general.

§ 5.4. Gilliéron 1915 and 1918 are the classical sources on “intolerable homon-
ymy”.

§ 6.2.1. Buck 1949 organizes the Indo-European lexicon by meanings, and thus
provides numerous opportunities to view semantic changes, including the
extended Indo-European example given here. The non-Indo-European data are
based on research by H. H. Hock; sources include Cohen 1947 and information
provided by Iwona Kraska-Szlenk.
 Chapter notes and suggested readings 497

Chapter 8.
Lexical borrowing
For general references, see Algeo 1990b and Haugen 1950, as well as Hock 1991,
Chapter 14; Thomason & Kaufman 1988 is a provocative work that kickstarted
much of the recent work on language contact; they define “borrowing” in a much
broader sense, including developments discussed in Chapters 12–14. Winford
2003, a general survey of contact-induced change, has extensive discussion of
borrowing. On aspects of borrowing (and contact more generally) involving sign
languages, see Lucas & Valli 1989, and Lucas 1990b, 1992.

§ 2. Just like the noun + adjective pattern from French in (8c) in which the bor-
rowing makes something possible in the language that was not there previously,
borrowed Greek and Latin roots have made certain “additive” compounds without
conjunctions possible in English that are not possible with native material;
compare oto-rhino-laryng-ologist with ear, nose, and throat doctor (impossible
without and: *ear, nose, throat doctor).

§ 3. See Hock 1991, § 14.3. Arndt 1973 deals with the issue of gender assignment of
German loanwords.

§ 4. See Janda, Joseph, & Jacobs 1994 for more examples of hyperforeignisms and
discussion of the significance of this phenomenon. On “Mock Spanish” see e. g.
Hill 1998 and Callahan 2014.

§ 5.2. For general discussion, see Hock 1991, § 14.5.3. The information on Icelandic
is based on research by H. H. Hock (conducted in 1962–1963); see also Haugen
1982: 204–205. See Sampson 1985: 166–7 for some discussion of nativization of
foreign words in Chinese. Sampson argues that the nature of the writing system
makes adaptation a difficult strategy. He also notes that as speakers of the lan-
guage of high culture in East Asia, the Chinese have felt little motivation through-
out history for the Chinese to borrow from neighboring languages. At the time
of the adoption of Buddhism, however, Chinese made numerous adoptions from
Sanskrit (with phonetic nativization); see e. g. Chen 2000.
498 Chapter notes and suggested readings

Chapter 9.
Lexical change and etymology: The study of words
General reference: A recent survey is found in Zgusta 1980.

§ 1. For numerous examples of phonesthematic attraction in English, see Samuels

1972; Bergen 2004 is a recent experimental study of this and related phenomena.
Popular books on English etymology and the origins of phrases, all of which make
for fascinating reading, include Black 1988, Feldman 1989, Flexner 1982, Funk
1985, 1986, Hendrickson 1983, Partridge 1961. Works on English word origins that
are on a more scholarly level include: The American Heritage Dictionary (espe-
cially the 1992 and 2000 editions, and now usable on-line at https://ahdictionary.
com/), Onions 1966, and especially the Oxford English Dictionary (first or second
edition), which traces the history of English words from their earliest attestation to
the present, providing extensive examples of usage and changes in usage. Davies
1981 is a popularized collection of Indo-European roots, whereas Watkins 1985
is a highly readable but more scholarly collection of all the Indo-European roots
that have a reflex in English, whether by inheritance or by borrowing, together
with a listing of all the English words that are derived from these roots. Buck 1949,
organized in terms of semantic classes, makes it possible to trace lexical change in
the Indo-European languages. Works dealing more with the practice of etymology
include Malkiel 1993 and Schmitt 1977.

§ 2. On words that derive from names (eponyms), see Partridge 1950a and Hen-
drickson 1988. Feldman & Feldman 1994 provides a popularized account of new
acronyms in English. The loss of the middle part of three-element compounds
(such as cheese hamburger → cheese burger) was already noted by the indigenous
Sanskrit grammarians and is discussed briefly in Wackernagel & Debrunner 1942
(especially pp. 164–165).

§ 3. For the etymological sources of names, there are literally dozens of popular
books that provide information, vignettes, derivations, etc. A few of the more
useful ones for English given names and surnames, both in the United States and
in England, are: Dunkling 1977, Ewen 1931, Hanks & Hodges 1988, 1990, Harrison
1918, Hook 1982, Lambert & Pei 1961, McKinley 1990, Stewart 1979, and Withy-
combe 1977. It is possible to find similar books on names in other languages and
for various ethnic traditions, such as Woods 1984 on Hispanic names, Kolatch
1989 on Hebrew names, Puckett 1974 on African American names or Guggen-
heimer & Guggenheimer 1992 and Kaganoff 1977 on Jewish names (ask your refer-
 Chapter notes and suggested readings 499

ence librarian for help regarding other languages and/or ethnic backgrounds, or
try your favorite internet search engine!).

§ 4. On slang, see Beale 1989, Dillard 1976, Grose 1796, Partridge 1950b, 1967. An
interesting exercise is to compare your local college slang temporally (e. g. with
what your parents or professors recall of their college slang) or geographically
(e. g. with what your friends at other schools report), for some insights into varia-
tion and lexical replacement in slang. On argot, see Kluge 1901 and Partridge 1967.
On African elements in African American Vernacular English, see Turner 1949 and
Dalby 1972 (the latter deals with the Wolof elements discussed in this section).

Chapter 10.
Language, dialect, and standard
General references: Chambers & Trudgill 1983; Francis 1983, and Trudgill 1983,
1986, 1990a, 1990b, 1994.

§ 2. The exact source of this (half-)joking statement about the difference between
dialects and languages is disputed; but whatever its source, it is particularly apt.
Browne 2002 has expanded on and updated this sentiment by mentioning as well
“a flag and a national anthem, and lately an airline …, a seat in the UN, and a
soccer team with the national colors”.

§ 3. For the Central Illinois chain shift see Habick 1980.

§ 4. The nautical jargon example comes from Hock 1991, Chapter 15. For Mediter-
ranean nautical jargon see Kahane, Kahane, and Tietze 1958.

§ 5. A good discussion of the development of Nynorsk and its relation to Bokmaal

is found in Haugen 1982. On the Hellenistic Greek Koiné, see Thumb 1901 and
Bubenik 1989; for briefer and highly readable discussions, see Browning 1982
and Horrocks 2010. Joseph 1987 has extensive discussion of the standardization
process in various European countries. On regional accents, see Trudgill 1983.

§ 6. The landmark study of diglossia is Ferguson 1959. The emotional side of the
“language question” in Greece has been so strong at times that there have lit-
erally been riots over the use of katharevousa versus dimotiki. Stylistic distinc-
tions somewhat like diglossia are found in virtually all languages; interestingly,
500 Chapter notes and suggested readings

this is true even for nonliterate societies, as documented by Bloomfield 1927 for
Menominee (an Algonquian language of the north central US, currently spoken
in Wisconsin).

Chapter 11.
Dialect geography and dialectology
General references: Chambers & Trudgill 1983, Francis 1983, Hock 1991 (Chapter 15),
Jaberg 1908, Mattheier 1983, and Trudgill 1983, 1986, 1990a, 1990b, 1994.

§ 2.1. On the geographical spread of the Chicago Sound Shift, see Callary 1975.

§ 2.2. See Kloeke 1927.

§ 4. For attempts at substratist explanations of Grimm’s Law see e. g. Bréal 1907

and Wiik 2002.

§ 5. For fuller discussion of the Old High German sound shift and the dialectol-
ogy of Old High German, see Hock 1991. Wolfram & Schilling-Estes 2003 surveys
several different models of diffusion and addresses how to make sense of situa-
tions with numerous, cross-cutting isoglosses.

Wolfram & Schilling-Estes 1998 gives a full account of American English and its
dialects, while Labov, Ash & Boberg 2006 offers a comprehensive look, in a mul-
timedia manner (e. g., via sound files and map displays on an associated website
and CD), at contemporary North American English dialectology.

§ 6. A catchy and highly quotable statement about American and British English
that has been attributed to George Bernard Shaw, to Oscar Wilde, and to Winston
Churchill – the truth as to the actual source is unclear – is that Americans and
British are “one people separated by a common language.”
 Chapter notes and suggested readings 501

Chapter 12.
Language spread, link languages, and
bilingualism
General reference: Hock 1991, § 16.1, Lehiste 1988, Thomason & Kaufman 1988,
Thomason 2001, Winford 2003.

§ 1. Esperanto and Volapük are two of the better-known artificial languages but
there have been others; see Large 1985 for more information. See also Libert 2000,
2003 on other sorts of artificial languages. Note the existence in Vienna of the
Esperanto Museum and Collection of Planned Languages.

§ 2. On Indian English (and more generally, world-wide varieties of English), see

Kachru 1965, 1986, 1992; on interlanguage, see Selinker 1972 and the recent reap-
praisal in Selinker 1992; for a different view on interlanguage and a discussion
of indigenization see Sridhar & Sridhar 1986. It is telling that the use of a plural,
“Englishes”, is quite common now in talking about postcolonial varieties of
English, and note the existence of a journal entitled World Englishes. See Schnei­
der 2003 and Trudgill 2004 on the creation of “new Englishes”, and Trudgill and
Hannah 2002 on international varieties of standard English.

§ 3. Muysken 1981 and 1997 discuss Media Lengua. See also Thomason & Kaufman
1988. Information on Michif is in part based on 1980s dissertation research at the
University of Illinois by James Kapper; see Bakker 1997 for a more recent, fuller
account. Douaud 1985 discusses the Canadian Métis, which is somewhat similar
to Michif, from an ethnolinguistic perspective. Media Lengua, Michif, and several
similar languages have been argued to be sufficiently different from other out-
comes of language contact to be recognized as a special category called “Bilingual
Mixed Languages”; for discussion see Matras & Bakker 2008.

§ 4. See Millardet 1933 for “substratum X”. The Balkan loss of the infinitive is
discussed in Joseph 1983; see also Hock 1988. See the notes on Chapter 10, § 5
concerning the Koiné.

§ 5. See the notes on Chapter 10, § 5 concerning the Koiné.

502 Chapter notes and suggested readings

Chapter 13.
Convergence: Dialectology beyond language
boundaries
General references: Hock 1991 (§ 16.3), Lehiste 1988, Southworth 1990, Thomason
& Kaufman 1988, Thomason 2001, Ureland 1990, Weinreich 1968 (a true classic in
the field), and Winford 2003.

§ 1. Recent studies of the effect of a second language on one’s first language are
Cook 2003 and Aysan 2012. Regarding the principle of accommodation see Bran-
igan et al. 2000, 2007, Giles et al. 1991, Pardo 2006, Pardo et al. 2013, Shepard et
al. 2001.

§ 2. The Kupwar convergence is described in Gumperz & Wilson 1971. Two recent
studies by Kulkarni-Joshi (2008, 2016) show that the traditional pattern of multi-
lingualism in Kupwar has come to an end and that, in order to get jobs, families
are shifting to the official State Language, Marathi. The changes in Kupwar situ-
ation show that the social context can keep on changing and affecting language
change (in keeping with the theme in Chapter 1 that “language is always chang-
ing”).

§ 3. The classic work on the Balkans is Sandfeld 1930; see also Schaller 1975, Solta
1980, Banfi 1985, Feuillet 1986, 2012, Asenova 2002, Demiraj 2004. There are no
book-length publications in English yet, though Friedman & Joseph 2020 will
remedy that (in the meantime, see Hock 1988, as well as Joseph 1986, 1992, 2003,
and Friedman 2006 for brief but readable presentations). The shared loan words
include calques (loan translations) not just of words and expressions but also pro-
verbial sayings, indicating a long-standing intimate and intense contact situation.

§ 4. The standard descriptive source on the South Asian convergence area is

Masica 1976. On the historical development of convergence, see Emeneau 1956
and the collection of papers in Emeneau 1980; a different view is advocated in
Hock 1975, 1984, 1996.

§ 6. See Hock 1988 on the dialectology of convergence areas. (This publication

also provides a detailed account of the various features and isoglosses in Map 4.)
 Chapter notes and suggested readings 503

Chapter 14.
Pidgins, creoles, and related forms of language
General references: Hall 1966, Hancock 1990, Holm 1989, Hymes 1971, Michaelis et
al. 2013ab, Schuchardt 1883–1888, 1978, 1980, Singler 1988, Thomason & Kaufman
1988, Thomason 2001, Winford 2003; see also Baron 1977 and Hock 1991 (§ 16.4).

§ 1. As far as H. H. Hock can recall, the anecdotal characterization of the unpre-

pared ESL teacher is adapted from Hatch 1978. Ferguson 1971 is the standard ref-
erence on Foreigner Talk and Baby Talk and their relation to other types of sim-
plification. See also the collection of articles in Clyne 1981.

§ 2. On Melanesian Pidgin English/Tok Pisin, see Hall 1943, Mühlhäusler 1970,

1982, 1983. The passage on Chinook Jargon at the end of this section comes from
Hale 1890.

§ 3. The idea that Foreigner Talk is the most important source for the development
of pidgins goes back to Schuchardt (1883–1888). Schuchardt, too, appears to be
the first one to have discussed the significance of the choice of the infinitive as
an all-purpose, uninflected form of the verb. On the deliberate use of Foreigner
Talk by the Portuguese, see Naro 1978. On the unsuccessful attempts of Australian
officials to “fake” Tok Pisin, see Mühlhäusler 1981.

§ 4. GAD is discussed in Clyne 1968, and Rost-Roth 1995 has some observations
about the extent of integration into German society. Regarding specific trade
jargons, see Broch 1927 on Russenorsk, and Thomason 1983 on Chinook Jargon.

§ 5. See Bickerton 1981 concerning the “bioprogram” hypothesis. Muysken &

Smith 1986 provides a good sampling of views critical of Bickerton’s claims; see
also Roberts 2000 for discussion of the demographics of Hawaiian Creole English
use in the formative period.

§ 6. On the AAVE copula, see Labov 1969. Dillard 1972 is a survey of African Amer-
ican Vernacular English; Schneider 1989 offers information on the early stages of
AAVE, and more recently Winford 1997, 1998 have extensive discussion. See also
Rickford 1999 (his website, http://www.stanford.edu/~rickford/ebonics, provides
a fair and informative discussion of recent controversies concerning AAVE), as
well as Wolfram and Thomas 2002 (who provide evidence against the view that
AAVE started out simply as a dialect of Southern U.S. speech).
504 Chapter notes and suggested readings

Chapter 15.
Language death
Ground-breaking studies on language death are Dressler 1972, Dressler & Wodak
1977, and Dorian 1981. Dorian 1989 is an important anthology, with case-studies of
a number of different language-death situations; Robins & Uhlenbeck 1991 offers
reports on the endangerment situation for languages in all parts of the world.
See also Schmidt 1985. Thomason 2015 is a recent and highly readable textbook
on the subject. Hock 1983 and 1992 deals with language attrition in an ancient
prestige language, Sanskrit, which is now dying out in its spoken use. Lambert
& Freed 1982 treats language loss in individuals. On the “English Only” or “Offi-
cial English” movement, see Baron 1990, Adams & Brink 1990. Fishman 1991 dis-
cusses language maintenance and language revival.
A fascinating debate on the role of the linguist in dealing with endangered
languages is provided by the exchange involving Hale et al. 1992, Ladefoged 1992,
and Dorian 1993.
Many additional book-length studies and anthologies have come out in
recent years dealing with different aspects of language endangerment, language
death, and language maintenance, including Abley 2003, Austin & Sallabank
2011, Campbell & Belew 2018, Janse & Tol 2003, Brenzinger 2007, Crystal 2000,
Grenoble & Whaley 2005, Harrison 2007, Hinton & Hale 2001, Rehg & Campbell
2018, and Tsunoda 2005. See also UNESCO’s Atlas of the world’s languages in
danger (http://www.unesco.org/languages-atlas/). A particularly richly popu-
lated website is that of the Endangered Languages Project (part of the Alliance for
Linguistic Diversity), at http://www.endangeredlanguages.com/.
Various organizations are supporting work on language documentation
and language preservation. These include Documenting Endangered Languages
(http://www.nsf.gov/funding/pgm_summ.jsp?pims_id=12816), Dokumenta-
tion bedrohter Sprachen (http://dobes.mpi.nl), the Endangered Language Fund
(http://www.endangeredlanguagefund.org), and the Hans Rausing Endangered
Languages Documentation Programme (http://www.hrelp.org/).
 Chapter notes and suggested readings 505

Chapter 16.
Comparative method: Establishing language
relationship
General references: See Anttila 1988, Hock 1991, Fox 1995, Winter 1990, Campbell
& Poser 2008.

§§ 1–6. For a good summary see Winter 1970. The most cogent statement of the
principles of the Comparative Method is Meillet 1925 (available in an English
translation), but see also Campbell 1988 and Joseph 2016. Baldi 1990 contains a
number of important articles on the results and methods of comparative linguis-
tics applied to a variety of language families (e. g. Campbell & Goddard 1990).
Anttila 1988 illustrates comparative reconstruction with Uralic data; Hock 1991,
with data from Germanic. The articles in Durie & Ross 1996 argue for a counter-
vailing perspective on the potential success and applicability of the Comparative
Method.

§ 7. For an application of the comparative method to syntax, see Hall 1968 and
Hock 1985, as well as Harris & Campbell 1995. The issue of Proto-Indo-European
syntactic reconstruction is discussed at length in Hock 1991, Chapter 19 and the
references cited there.
On the “Glottalic Theory”, developed more or less at the same time by Gam-
krelidze (working with Ivanov), and by Hopper; see Gamkrelidze & Ivanov 1973,
1984, 1994, Hopper 1973, Gamkrelidze 1988, all rather advanced works. Relevant
also are Vennemann 1989, Haider 1985, Hock 1991 (§ 19.5.2), Stewart 1989 (on
voiced aspirates in West African Kwa languages), Stevens 1992; see also Vogt 1958
(on Armenian glottalized stops). Salmons 1993 provides a good overview of the
theory, pro and con, while Szemerényi 1989b gives a critical appraisal (see also his
1967 pre-glottalic appraisal of Proto-Indo-European phonology).
The validity of the Comparative Method has been tested against controls pro-
vided by the Romance languages by Hall 1950 and 1976.

§ 8. Sources on some of the language families mentioned include the following:

Uralic: Collinder 1965a, 1965b, Décsy 1990, Rédei 1986–1988; Sajnovics 1770
Altaic: Poppe 1960, Ramstedt 1957–1966, Miller 1971, 1991 (Japanese included),
Unger 1990 (doubts about Altaic); the Altaic hypothesis has been revived by
Robbeets (e. g. 2013) under the label Transeurasian.
Basque: Tovar 1957, Trask 1996, Hualde, Lakarra & Trask 1996
Caucasic languages: Hewitt 1989
506 Chapter notes and suggested readings

Chinese and Sino-Tibetan: Haudricourt 1954, Karlgren 1949, Norman 1988; Thur-
good & LaPolla 2003, Baxter & Sagart 2014
Dravidian: Andronov 1970, Caldwell 1974, Steever 1998, Krishnamurti 2003, Koli-
chala 2016
Austro-Asiatic: Anderson 2008, Jenny & Sidwell 2015
Malayo-Polynesian/Austronesian: Blust 1990, Dempwolff 1938, Kahlo 1941
Afro-Asiatic: Cohen 1947, Lieberman 1990 (the Cushitic and Chadic sub-groups
comprise a variety of related languages; the forms that we cite come from
different languages within each group)
Bantu: Doke 1954, Meinhof 1899 (and see also Greenberg 1966 on the relatives of
Bantu within Africa)
Indigenous languages of the Americas: Campbell & Mithun 1979, Campbell 1997,
Mithun 1999. For a sympathetic and intriguing account of the diversity of
indigenous languages in California, written for a general audience, see
Hinton 1994.
Algonquian: Campbell & Goddard 1990, Goddard 1990; the classic work on the
family is Bloomfield 1946 (available on-line at https://home.cc.umanitoba.
ca/~oxfordwr/bloomfield1946/)
Uto-Aztecan: Miller 1967
Siouan: Chafe 1976
Hokan-Siouan: Langdon 1974
Mayan: Campbell 1990ab
Quechua: Cerrón-Palomino 1987
Australian languages: Dixon 1980, 1990, 2002, Evans 2003, Bowern & Koch 2004.

Surveys of languages of the world, intended for a general audience include Camp-
bell 1991, 1995, Katzner 1995, Lyovin et al. 2017, Pereltsvaig 2017, Wendt 1961; these
books typically provide a thumbnail sketch (at best) of a number of languages,
including quite obscure ones. A collection of more scholarly accounts (of a smaller
number of languages, some 40 in all) is to be found in Comrie 2018. On sign lan-
guages, see Lane 1980, Perlmutter 1986, and Stokoe 1974.
 Chapter notes and suggested readings 507

Chapter 17.
Proto-World? The question of long-distance
genetic relationships
General references: See the papers in Lamb & Mitchell 1991 and Campbell & Poser
2008. Statistical methods are becoming more and more prevalent for testing
claims of relatedness; see Ringe 1992 for an early example, and the papers in
Forster & Renfrew 2006 for more discussion (with further references). See also
the notes on Chapter 18.

§ 1. The quoted passages from Indigenous American languages are based on

Hinton 1994 and Courlander 1971. On Nostratic, which comes in different “vari-
eties” depending on the researcher(s) involved, see Bomhard 1984, 1990, 2008,
Bomhard & Kerns 1994, Collinder 1965b, 1974, Illič-Svityč 1964, 1971, 1976, and the
papers in Salmons & Joseph 1998.

§ 2. On Uralic and Dravidian, see Tyler 1968. For attempts to link Indo-European
and Uralic, see Collinder 1965b, 1966, 1974, and Ringe 1995; Décsy 1990 expresses
doubts about reconstructing the Uralic case endings. On the proposed connec-
tion between Dravidian and Elamite, see McAlpin 1974, 1975, 1981. Other connec-
tions involving Indo-European have been proposed over the years, most notably
Indo-European and Hamito-Semitic or Afro-Asiatic; see, e. g., Cuny 1946, Hodge
1990, Møller 1907, 1917. On the question of the meaning ‘wolf’ for Nostratic *kuyon
or *küyna, see Manaster-Ramer 1992.

§ 4. Greenberg’s controversial claims about lexical mass comparison, leading

to his classification of Indigenous American languages, are presented in Green-
berg 1987, and condensed for a general audience in Greenberg & Ruhlen 1992;
Ruhlen 1994 presents the mass comparison evidence for a Proto-World. (Earlier
statements and applications of Greenberg’s views are found in Greenberg 1957
and 1966.) See also Ross 1991 and Shevoroshkin 1990, both aimed at the general
reader.
Greenberg’s method of mass comparison is critically examined by Peter 1991
(an empirical test of the methodology) and Ringe 1992 (an evaluation of the sta-
tistical foundation of the methodology). Hock 1994 is the source of the critical
discussion of the putative reconstruction *maliq’a, while Salmons 1992 critically
examines the *tik reconstruction; historical linguists are generally skeptical of
Greenberg’s methods, see e. g. Hamp 1992.
508 Chapter notes and suggested readings

An interesting scholarly debate on methodology and results in this area is

represented in a series of publications by Campbell 1988, Greenberg 1989, and
Matisoff 1990.

§ 5. The injunction of the Linguistic Society of Paris is to be found in its by-laws,

Article 2: “La Société n’admet aucune communication concernant, soit l’origine
du langage, soit la création d’une langue universelle” (“The Society will accept
no contribution dealing with either the origin of language or the creation of a
universal language”).
Valuable resources include Jespersen 1921 and Lieberman 1975, 1984, 1998,
and the anthologies of de Grolier, Lock, Peters, & Wind 1983; Lock 1978; von Raf-
fler-Engel, Wind, & Jonker 1991; Wind, Chiarelli, Bichakjian, Nocentini, & Jonker
1992; Wind, Jonker, Allott, & Rolfe 1994; and Wind, Pulleyblank, de Grolier, &
Bichakjian 1989; Armstrong, Stokoe, & Wilcox 1995 present a case for the role of
sign language in language origin, while many of the papers in Gibson & Ingold 1993
explore the relationship between tool use in early hominids and the emergence
of language. Danesi 1993 surveys several centuries of speculation on the origin of
language in Western European philosophy, and ultimately sees the development
of representational art as a relevant factor in the development human language.
Attempts to learn about language origins from teaching language to animals are
surveyed in Wallman 1992; there are many popular treatments of this topic, some
focusing more on the larger (and compelling) perspective of human-animal inter-
action in general, e. g. Hess 2008. Lightfoot 1999 discusses insights into language
evolution that the study of generative syntax might offer, and Hauser, Chomsky,
& Fitch 2002 is similarly oriented toward what a generative approach can contrib-
ute. Carstairs-McCarthy 2000 speculates that sounds and syllables predated the
emergence of coherent meanings, looking at the origin of language from a fresh
viewpoint compared with most other discussions. On the more anthropological
front, Hockett & Ascher 1964 is a classic work of relevance. See also Pinker 1994,
Chapter 11, for a relatively brief but quite readable discussion of the origin of lan-
guage. Other recent works with a popular orientation include Everett 2017, Fitch
2010, Hurford 2007, 2012, 2014, and Kenneally 2007. Callaghan 1997 surveys four
books on the topic, citing several others that came out in the 1990s, and Tamariz
2019 is a fine overview of current thinking; see also Erard & Matacic 2018. As of
2015 (internet publication; first print issue: 2016), there is now a Journal of Lan-
guage Evolution, published by Oxford University Press.
We use the spelling “Neanderthal” here, pronounced with [θ] by many speak-
ers, and [t] by others, and note that one also sees the spelling “Neandertal”. There
is no real consensus in the current literature, and both have some justification;
 Chapter notes and suggested readings 509

science fiction writer Robert J. Sawyer has an interesting discussion of both sides
of this point at https://www.sfwriter.com/hotal.htm.

Chapter 18.
Historical linguistics, history, and prehistory:
Linguistic palaeontology and other applications
of our methods
§ 1. General references on linguistic palaeontology: Winter & Polomé 1992 (an
important anthology), Polomé 1990b. Early work on Indo-European cultural
reconstruction includes Schrader 1886, 1890, 1906–1907. General works on lin-
guistic palaeontology as applied to various aspects of Indo-European prehistory
include Benveniste 1973, Polomé 1990b, Scherer 1956, Schmidt 1992, and Skomal
& Polomé 1987.

§ 2. The identification ārya = Eire (Old Irish Ériu) appears to go back to Max Müller
1864. Later research suggested rather that Eire goes back to *piHweryōn ‘land of
substance’, which via *fīweryōn > (*)hiweryōn (compare Latinized Hibernia),
changed into Old Ir. Ériu (MacBain 1911).

§ 3. On the position of Anatolian in Indo-European see Melchert 1998; the position

of Tocharian is established on lexical grounds in Winter 1997b. See Ringe et al.
2002 for fuller discussion.

§ 4.1. On Indo-European poetics and mythology, see Watkins 1982, 1987, 1989,
1995. Lehmann & Zgusta 1979 presents an attempt at reconstructing a Proto-In-
do-European connected text. See Malone et al. 1993 on recent archaeological finds
in Malta that bear on the issue of male vs. female deities.

§ 4.2. For an analysis of Proto-Indo-European institutions, based on lexical

items and their use in the individual languages, including some observations on
Indo-European law, see Benveniste 1973. Dumézil (1958) argues for a tripartite
social stratification of Indo-European society. The role of traders in early societies
has been dealt with in Taylor 1992. On Indo-European women and textiles, see
Barber 1991. On Renfrew’s “Anatolian hypothesis,” see § 8.5 below.
510 Chapter notes and suggested readings

§ 4.3. The example of Indo-European comparative jurisprudence is adapted from

Calvert Watkins’s class lectures (Harvard University, 1973).

§ 5–6. Benveniste 1973 is an excellent source on Proto-Indo-European material

culture and economy, and Mallory & Adams 1997, 2006 are recent compendia on
Indo-European culture based on linguistic palaeontology and archaeology. On
transhumance see e. g. Suttie & Reynolds 2003; on the use of grains by “mobile
pastoralists” in early 3rd-millennium Central Eurasia, see Spengler et al. 2014.

§ 6.1. A good survey of early views is found in Schrader 1906–1907.

§ 6.3. See also Kümmel 2017 on Indo-Iranian. For a different perspective see
Joseph 2017. Weiss 2018 raises serious questions about the inclusion of Hitt. e(u)
wa(n) as well as the original meaning of *yewo. Recent archaeological research
shows that grains were used for cooking as early as 15,000 years BP, long before
the introduction of agriculture; see Dunne et al. 2016.

§ 7.1. On the Greek word for ‘horse’ see Bozzone 2013; for Slavic see Blažek 2009;
for Anatolian see Melchert 2012. On the non-European words for ‘donkey’, see e. g.
Parpola 2010, Pinault 2003, 2006, 2008. Winter 1997a provides evidence that the
Armenian meaning ‘donkey’ is innovated.

§ 7.2. On horse domestication in the Eurasian steppes in the early 4th millen-
nium BC see Anthony et al. 1991, Anthony 2007, and Outram et al. 2009. Bendrey
2012 and Warmuth et al. 2012 present archaeological and genomic arguments
in support of this geographic and chronological setting. (Bendry considers the
Iberian peninsula a possible alternative; but there is no positive evidence for
horse domestication at that time.) Regarding the domestication of donkeys see
Rossel et al. 2008, and see Vilà et al. 2006 on horse and donkey domestication.
The almost simultaneous appearance of wheels and wheeled vehicles in mid-4th
millennium Eurasia is discussed in Fansa & Burmeister 2004.

§ 7.3. On the supposed Indo-European invention of the wheel, see Parpola 2008.

§ 8.2. For the problematic “beech tree” hypothesis, see Hoops 1905, Krogmann
1955–1956 vs. Feist 1913, Bynon 1977. Thieme 1954 added PIE *lokso ‘salmon’ to
the supposed beech tree evidence; but that argument was questioned by Diebold
1985; see also Mayrhofer 1955.

§ 8.3. See Djakonov 1985.

 Chapter notes and suggested readings 511

§ 8.4. The most outspoken advocates for a homeland in the Caucasus are Gam-
krelidze & Ivanov (1984, 1985, 1990). See also Dolgopolsky 1989 and the papers in
Markey & Greppin 1990.

§ 8.5. An early advocate of the Steppe Hypothesis is Gimbutas 1970, 1985. More
recent publications include Anthony 2007 and Anthony & Ringe 2015. Renfrew’s
publications in support of the Anatolian Hypothesis include Renfrew 1987, 1989,
2004. (Interestingly, Renfrew 1973 expressed serious concerns about diffusionist
theories.) Recent publications in favor of the Anatolian Hypothesis include Atkin-
son & Gray 2006, Bouckaert et al. 2012, and Gray & Atkinson 2003. Responses to
the Atkinson et al. perspective include Ringe 2009 (linguistic issues), Holm 2007
(statistical models and methodology), and Pereltsvaig & Lewis 2015 (extensive
general critique, including the issue of Romani/Indo-Aryan). On the problems of
glottochronology see Bergsland & Vogt 1962. Chang et al. 2015 argue that a more
sophisticated application of Atkinson et al.’s methodology supports the Steppe
Hypothesis. For the suggestion that the meanings ‘horse’ and ‘wheel’ are second-
ary, see Heggarty 2008.

Palaeogenomic arguments for a migration from the steppes to western Eurasia

ca. 4500 BP are Allentoft et al. 2015 and Haak et al. 2015 (the latter publication
directly links the migration to the spread of Indo-European languages). The
genomic study of Turkey is Cinnioğlu et al. 2003.

§ 9.1. General references: Schlegel 1808; Lassen 1847, Latham 1851, Gobineau
1853 (opposed by Pott 1856), Chamberlain 1899; Zimmer 1879 (vs. Schetelich 1991,
Hock 1999); Günther 1934 (vs. Krahe 1942). The set of scholarly volumes of 1936 is
Arntz 1936. On the influence of Gobineau and Chamberlain on Hitler’s thinking
see Spielvogel & Redles 1986.

§ 9.2. Recent publications include Allentoft et al. 2015, Haak et al. 2015, Jones et
al. 2015, Lazaridis et al. 2014, Mathieson 2015 et al. On prehistoric blue-eyed, dark-
skinned European hunter-gatherers see Olalde et al. 2014; on blond hair as a late
phenomenon see Guenther et al. 2014.

§ 10. On Sintashta archaeology and identifying its culture as Indo-Iranian see

Kuz’mina 2007 (vs. Lamberg-Karlovsky 2002); for the appropriation of Arkaim
by nationalists and neo-nazis see Shnirelman 2007. For an early British view see
Max Müller 1847; the acceptance of the AIT by Dravidian and other movements
is dicussed in Bergunder 2002. A leading opponent of the AIT is Talageri (e. g.
2008); publications questioning comparative-historical linguistics include Jha &
512 Chapter notes and suggested readings

Rajaram 2000. For the Hungarian nationalist rejection of the classification of Hun-
garian as Uralic, see Klaniczay 2011. Marcantonio’s 2002 attack on Uralic linguis-
tics and comparative-historical linguistic methodology has met with strong refu-
tation by Laakso (2004); her similar attack on Indo-European linguistics (2009)
suffers from similar shortcomings but is still awaiting a comprehensive response.
The question of the reality or realism of reconstructions is debated by Pulgram
1959 and Hall 1960; see also Hall 1968 on the overall reliability of the comparative
method for Romance.
References
Abley, Mark. 2003. Spoken here: Travels among threatened languages. New York: Random
House.
Adams, Karen, and Daniel T. Brink (eds.) 1990. Perspectives on official English: The campaign
for English as the official language of the USA. Berlin/New York: Mouton de Gruyter.
Aitchison, Jean. 2001. Language change: Progress or decay? 3rd edn. Cambridge: Cambridge
University Press.
Algeo, John. 1990a. Semantic change. In Polomé 1990a: 399–408.
Algeo, John. 1990b. Borrowing. In Polomé 1990a: 409–413.
Allentoft, Morten E., and 65 co-authors. 2015. Population genomics of Bronze Age Eurasia.
Nature 522: 167–174.
American Heritage Dictionary of the English Language. 2000. 4th edn. Boston: Houghton Mifflin
Co.
Anderson, Gregory D. S. (ed.) 2008. The Munda languages. Oxford/New York: Routledge.
Andrews, Carol. 1981. The British Museum book of the Rosetta Stone. New York: Dorsett.
Andronov, Mikhail Sergeevich. 1970. Dravidian languages [transl. from the Russian by
D. M. Segal]. Moscow: Nauka.
Anthony, David W. 2007. The horse, the wheel, and language: How Bronze-Age riders from the
Eurasian steppes shaped the modern world. Princeton: Princeton University Press.
Anthony, David W., and Don Ringe. 2015. The Indo-European homeland from linguistic
and archaeological perspectives. Annual Review of Linguistics 2015 (1): 199–219.
http://www.annualreviews.org/journal/linguistics.
Anthony, David, Dimitri Y. Telegin, and Dorcas Brown. 1991. The origin of horseback riding.
Scientific American, December 1991: 94–100.
Antonsen, Elmer H. 1967. On the origin of the Old English digraph spellings. Studies in
Linguistics 19: 5–17.
Antonsen, Elmer H. 1975. A concise grammar of the older runic inscriptions. Tübingen:
Niemeyer.
Anttila, Raimo, and Warren A. Brewer. 1977. Analogy: A basic bibliography. Amsterdam/
Philadelphia: Benjamins.
Anttila, Raimo. 1988. Historical and comparative linguistics, revised edition. Amsterdam/
Philadelphia: Benjamins.
Anttila, Raimo. 2003. Analogy: The warp and woof of cognition. In Joseph and Janda 2003:
425–440.
Arlotto, Anthony. 1972. Introduction to historical linguistics. Boston: Houghton-Mifflin. (Repr.
1981, Washington, DC: University Press of America.)
Armstrong, David F., William F. Stokoe, and Sherman E. Wilcox. 1995. Gesture and the nature of
language. Cambridge: Cambridge University Press.
Arndt, Walter W. 1973. Nonrandom assignment of loan words: German noun gender. Word 26:
244–253 (1970–1972).
Arntz, Helmut (ed.). 1936. Germanen und Indogermanen: Volkstum, Sprache, Heimat, Kultur:
Festschrift für Herman Hirt, 2 vols. Heidelberg: Winter.
Asenova, Petja. 2002. Balkansko Ezikoznanie: Osnovni Problemi na Balkanskija Ezikov Sûjuz.
2nd edn. Sofia: Faber [1st edn., 1989].
Atkinson, Quentin D., and Russell D. Gray. 2006. How old is the Indo-European language family?
Illumination or more moths to the flame? In Peter Forster and Colin Renfrew, 91–109.

https://doi.org/10.1515/9783110613285-020
514 References

Austin, Peter K. and Julia Sallabank (eds.) 2011. The Cambridge handbook of endangered
languages. Cambridge: Cambridge University Press.
Aysan, Zeynep. 2012. Reverse interlanguage transfer: The effects of L3 Italian & L3 French on
L2 English pronoun use. Bilkent University MA thesis. http://www.thesis.bilkent.edu.
tr/0006019.pdf
Bakker, Pieter. 1997. A language of our own: The genesis of Michif, the mixed Cree-French
language of the Canadian Métis. (Oxford Studies in Anthropological Linguistics, No 10.)
Oxford/New York: Oxford University Press.
Baldi, Philip (ed.). 1990. Linguistic change and reconstruction methodology. Berlin/New York:
Mouton de Gruyter.
Baldi, Philip. 1983. An introduction to the Indo-European languages. Carbondale/ Edwardsville:
Southern Illinois University Press.
Baldi, Philip. 1999. The foundations of Latin. Berlin/New York: Mouton de Gruyter.
Ball, Martin J., and James Fife (eds.) 1993. The Celtic languages. London/New York: Routledge.
Banfi, Emmanuel. 1985. Linguistica balcanica. Bologna: Zanichelli.
Barber, Charles. 1976. Early Modern English. London: Andre Deutsch.
Barber, Elizabeth J. W. 1991. Prehistoric textiles. Princeton: Princeton University Press.
Baron, Dennis. 1989. Declining grammar and other essays on the English vocabulary. Urbana,
IL: National Council of Teachers of English.
Baron, Dennis. 1990. The English-only question: An official language for Americans? New
Haven: Yale University Press.
Baron, Naomi S. 1977. Trade jargons and pidgins: A functionalist approach. Journal of Creole
Studies 1977: 5–28.
Battison, Robin. 1978. Lexical borrowing in American Sign Language. Silver Spring, MD: Linstok
Press.
Baxter, William H., and Laurent Sagart. 2014. Old Chinese: a new reconstruction. New York:
Oxford University Press.
Beale, Paul (ed.) 1989. Partridge’s concise dictionary of slang and unconventional English. New
York: Macmillan.
Beekes, Robert S. P. and Michiel DeVaan. 2011. Comparative Indo-European Linguistics: An
Introduction. Second Edition. Amsterdam/Philadelphia: Benjamins.
Bendrey, Robin. 2012. From wild horses to domesticated horses: A European perspective. World
Archaeology 44 (1): 135–157. https://www.academia.edu/1785218/From_wild_horses_to_
domestic_horses_a_European_perspective
Benveniste, Emile. 1973. Indo-European language and society. (Transl. from the French (Le
vocabulaire des institutions indo-européennes. Paris: Les Éditions de Minuit [1969: 2
volumes]) by E. Palmer.) Coral Gables, FL: University of Miami Press.
Bergen, Benjamin K. 2004. The psychological reality of phonaesthemes. Language 80:
290–311.
Bergsland, Knut, and Hans Vogt. 1962. On the validity of glottochronology. Current
Anthropology 3: 115–153.
Bergunder, Michael. 2002. Umkämpfte Vergangenheit: Anti-brahmanische und hindu-na-
tionalistische Rekonstruktionen der frühen indischen Religionsgeschichte. “Arier” und
“Draviden”, ed. by Michael Bergunder and Rahul Peter Das, 135–180. Halle: Verlag der
Franckeschen Stiftungen.
Berlant, Stephen R. 2008. Deconstructing Grimm’s Law reveals the unrecognized foot and leg
symbolism in Indo-European lexicons. Semiotica 171: 265–290.
 References 515

Bickerton, Derek. 1981. Roots of language. Ann Arbor, MI: Karoma.

Black, Donald C. 1988. Spoonerisms, sycophants, and sops. New York: Harper & Row.
Blažek, Václav. 2009. On hypothetical Slavic continuants of IE *H1eku ̂ ̯ o- ‘horse’. http://www.
pies.ucla.edu/WeCIEC/blazek_v_2009 f.pdf
Bloomfield, Leonard. 1927. Literate and illiterate speech. American Speech 2: 10: 432–439.
Bloomfield, Leonard. 1933. Language. New York: Holt, Rinehart & Winston. (Repr. 1984,
Chicago University Press; chapters on historical linguistics issued separately as
“Language history”, New York: Holt, Rinehart & Winston, 1965.)
Bloomfield, Leonard. 1946. Algonquian. Linguistic structures of Native America, ed. by Harry
Hoijer et al., 85–129. (Viking Fund Publications in Anthropology 6.) New York: Viking Fund.
Bloomfield, Morton W., and Leonard Newmark. 1963. A linguistic introduction to the history of
English. New York: Alfred A. Knopf.
Blust, Robert. 1990. Summary report: Linguistic change and reconstruction methodology in the
Austronesian language family. In Baldi 1990: 133–153.
Bomhard, Allan R. 1984. Toward Proto-Nostratic: A new beginning in the reconstruction of
Proto-Indo-European and Proto-Afro-Asiatic. Amsterdam/Philadelphia: Benjamins.
Bomhard, Allan R. 1990. A survey of the comparative phonology of the so-called “Nostratic”
languages. In Baldi 1990: 331–358.
Bomhard, Allan R. 2008. Reconstructing Proto-Nostratic. Comparative phonology, morphology,
and vocabulary (2 vols). Leiden: Brill.
Bomhard, Allan R., and John C. Kerns. 1994. The Nostratic macrofamily: A study in distant
linguistic relationship. Berlin/New York: Mouton de Gruyter.
Bopp, Franz. 1816. Ueber das Conjugationssystem der Sanscritsprache in Vergleichung mit
jenem der griechischen, lateinischen, persischen und germanischen Sprache. Frankfurt:
Andreä.
Boretzky, Norbert. 1977. Einführung in die historische Linguistik. Reinbek: Rowohlt.
Bouckaert, Remco, Philippe Lemey, Michael Dunn, Simon J. Greenhill, Alexander V.
Alekseyenko, Alexei J. Drummond, Russell D. Gray, Marc A. Suchard, and Quentin D.
Atkinson. 2012. Mapping the origins and expansion of the Indo-European language family.
Science 337: 957–960.
Bowern, Claire, and Bethwyn Evans (eds.) 2014. Routledge handbook of historical linguistics.
New York: Routledge.
Bowern, Claire, and Harold Koch (eds.) 2004. Australian languages: Classification and the
comparative method. Amsterdam/Philadelphia: Benjamins.
Bozzone, Chiara. 2013. Initial “yod” in Greek and the etymology of Greek ἵππος ‘horse’.
Proceedings of the 24th Annual UCLA Indo-European Conference, ed. by Stephanie W.
Jamison, H. Craig Melchert, and Brent Vine, 1–26. Bremen: Hempen.
Branigan, H. P., M. J. Pickering, and A. A. Cleland. 2000. Syntactic co-ordination in dialogue.
Cognition 75: 13–25.
Branigan, H. P., M. J. Pickering, J. F. McLean, and A. A. Cleland. 2007. Syntactic alignment and
participant role in dialogue. Cognition 104: 163–197.
Bréal, Michel. 1907. La loi de Grimm. La revue de Paris 14: 6. 52–96.
Brenzinger, Matthias (ed.) 2007. Language diversity endangered. Berlin/New York: Mouton de
Gruyter.
Bright, William (ed.) 1992. Oxford International Encyclopedia of Linguistics. London/New York:
Oxford University Press.
Broch, Olaf. 1927. Russenorsk. Archiv für slavische Philologie 41: 209–262.
516 References

Brown, Keith (ed.) 2006. Encyclopedia of language and linguistics. 2nd edn. Oxford: Elsevier.
Browning, Robert. 1982. Medieval and Modern Greek (2nd edn.). Cambridge: Cambridge
University Press.
Browne, E. Wayles. 2002. What is a standard language good for, and who gets to have one?
and Open and closed accent types in nouns in Serbo-Croatian. (Naylor Memorial Lecture
Series, 3.) Columbus: Ohio State University Department of Slavic and East European
Languages and Literatures.
Brugmann, Karl. 1897–1916. Vergleichende Laut-, Stammbildungs- und Flexionslehre der
indogermanischen Sprachen. Straßburg: Trübner. [Revised edition of K. Brugmann and
B. Delbrück 1886–1900, Grundriß der vergleichenden Grammatik der indogermanischen
Sprachen. Straßburg: Trübner.]
Brugmann, Karl. 1904. Kurze vergleichende Grammatik der indogermanischen Sprachen.
Straßburg: Trübner.
Bubenik, Vit. 1989. Hellenistic and Roman Greece as a Sociolinguistic Area. Amsterdam/
Philadelphia: Benjamins.
Buck, Carl Darling. 1949. A dictionary of selected synonyms in the principal Indo-European
languages. Chicago/London: University of Chicago Press. Paperback edition 1988,
Chicago/London: University of Chicago Press.
Burrow, Thomas. 1973. The Sanskrit language, new and revised edition. London: Faber.
Bybee, Joan L., and Dan I. Slobin. 1982. Why small children cannot change language on
their own: Suggestions from the English past tense. Papers from the 5th International
Conference on Historical Linguistics, ed. by A. Ahlqvist, 29–37. Amsterdam/Philadelphia:
Benjamins.
Bynon, Theodora. 1977. Historical linguistics. Cambridge: Cambridge University Press.
Caldwell, Robert. 1974. A comparative grammar of the Dravidian or South-Indian family of
languages, revised and edited by J. L. Wyatt and T. Ramakrishna Pillai. New Delhi: Oriental
Books Reprint Corp. (Reprint of the 1913 edition, London: Kegan Paul, Trench, Trubner.)
Callaghan, Catherine A. 1997. Progress on the origin of language. Diachronica 14: 2. 345–362.
Callahan, Laura. 2014. The importance of being earnest: Mock Spanish, mass media, and the
implications for language learners. Spanish in Context 11:2. 202–220.
Callary, Robert E. 1975. Phonological change and the development of an urban dialect in
Illinois. Language in Society 4: 155–169.
Campbell, George L. 1991. Compendium of the world’s languages, two volumes. London/New
York: Routledge.
Campbell, George L. 1995. Concise compendium of the world’s languages. London/New York:
Routledge.
Campbell, Lyle. 1988. Review article on Greenberg 1987. Language 64: 591–615.
Campbell, Lyle. 1990a. Mayan languages and linguistic change. In Baldi 1990: 115–129.
Campbell, Lyle. 1990b. Philological studies in Mayan languages. In Fisiak 1990: 87–105.
Campbell, Lyle. 1997. American Indian languages: The historical linguistics of Native America.
Oxford: Oxford University Press.
Campbell, Lyle. 2013. Historical Linguistics. An Introduction. 3rd edn. Edinburgh: Edinburgh
University Press.
Campbell, Lyle, and Anna Belew (eds.) 2018. Cataloguing the world’s endangered languages.
New York: Routledge.
Campbell, Lyle, and Ives Goddard. 1990. Summary report: American Indian languages and
principles of language change. In Baldi 1990: 17–32.
 References 517

Campbell, Lyle, and Marianne Mithun (eds.) 1979. The languages of native America. Austin:
University of Texas Press.
Campbell, Lyle, and Mauricio Mixco. 2007. Glossary of historical linguistics. Edinburgh:
Edinburgh University Press / Salt Lake City: University of Utah Press.
Campbell, Lyle, and William Poser. 2008. Language classification: History and method.
Cambridge: Cambridge University Press.
Cardona, George, and Dhanesh Jain (eds.) 2003. The Indo-Aryan languages. London/New York:
Routledge.
Cardona, George, Henry Hoenigswald, and Alfred Senn (eds.) 1970. Indo-European and
Indo-Europeans. Philadelphia: University of Pennsylvania Press.
Carstairs-McCarthy, Andrew. 2000. The origins of complex language: An inquiry into the
evolutionary beginnings of sentences, syllables, and truth. New York: Oxford University
Press.
Carter, Martha L., and Keith N. Schoville (eds.) 1984. Sign, symbol, script: [Guide to] An
exhibition on the origins of writing and alphabet. Madison: University of Wisconsin,
Department of Hebrew and Semitic Studies.
Ceram, C. W. (Pseudonym for Karl W. Marek). 1973. The secret of the Hittites: The discovery of
an ancient empire. Transl. by R. and C. Winston from the 1955 German original. New York:
Schocken Books.
Cerrón-Palomino, Rodolfo. 1987. Lingüística quechua. Cuzco: Centro de los studios rurales
andinos.
Chadwick, John. 1958. The decipherment of Linear B. New York: Vintage Books.
Chadwick, John. 1987. Reading the past: Linear B and related scripts. Berkeley/Los Angeles:
University of California Press and British Museum.
Chafe, Wallace L. 1976. The Caddoan, Iroquoian, and Siouan languages. The Hague:
Mouton.
Chamberlain, Houston Stewart. 1899. Grundlagen des neunzehnten Jahrhunderts. München:
Bruckmann.
Chambers, John K., and Peter Trudgill. 1983. Dialectology. Cambridge: Cambridge University
Press.
Chang, Will, Chundra Cathcart, David Hall, and Andrew Garrett. 2015. Ancestry-constrained
phylogenetic analysis supports the Indo-European steppe hypothesis. Language 91:
194–244.
Chen, Shu-Fen. 2000. Rendition techniques in the Chinese translation of three Sanskrit
Buddhist scriptures. University of Illinois at Urbana-Champaign PhD dissertation.
Chiera, Edward. 1938. They wrote on clay. (Reprinted 1966). Chicago: University of Chicago
Press.
Christidis, Anastasios-Ph. (ed.) 2007. A history of Ancient Greek: From the beginnings to late
antiquity. Cambridge: Cambridge University Press.
Cinnioğlu, Cengiz, Roy King, Toomas Kivisild, Ersi Kalfoğlu, Sevil Atasoy, Gianpiero L. Cavalleri,
Anita S. Lillie, Charles C. Roseman, Alice A. Lin, Kristina Prince, Peter J. Oefner, Peidong
Shen, Ornella Semino, L. Luca Cavalli-Sforza, Peter A. Underhill. 2003. Excavating
Y-chromosome haplotype strata in Anatolia. Human Genetics 114: 127–148. http://
evolutsioon.ut.ee/publications/Cinnioglu2004.pdf.
Clackson, James. 2007. Indo-European linguistics: An introduction. Cambridge: Cambridge
University Press.
Cleator, Philip E. 1962. Lost languages. New York: New American Library.
518 References

Clyne, Michael. 1968. Zum Pidgin-Deutsch der Gastarbeiter. Zeitschrift für Mundartforschung
35: 130–139.
Clyne, Michael (ed.) 1981. Foreigner talk. (International Journal of the Sociology of Language, 28.)
Coe, Michael D. 1992. Breaking the Maya code. New York: Thames & Hudson.
Cohen, Marcel. 1947. Essai comparatif sur le vocabulaire et la phonétique du chamito-
sémitique. Paris: Librairie Ancienne Honoré Champion.
Collinder, Björn. 1965a. An introduction to the Uralic languages. Berkeley/Los Angeles:
University of California Press.
Collinder, Björn. 1965b. Hat das Uralische Verwandte? Eine sprachvergleichende Untersuchung.
(Acta Universitatis Upsaliensis: Acta Societatis Linguisticae Upsaliensis, n.s., 1: 4.)
Uppsala: Almqvist & Wiksell.
Collinder, Björn. 1966. Distant linguistic affinity. In: Ancient Indo-European dialects, ed. by
H. Birnbaum and J. Puhvel, 199–200. Berkeley: University of California Press.
Collinder, Björn. 1974. Indo-Uralisch – oder gar Nostratisch? Vierzig Jahre auf rauhen Pfaden.
In: Antiquitates Indogermanicae: Gedenkschrift für Hermann Güntert, ed. by M. Mayrhofer
et al., 363–375. (Innsbrucker Beiträge zur Sprachwissenschaft, 12.) Innsbruck: Institut für
Sprachwissenschaften.
Comrie, Bernard, and Greville Corbett (eds.) 1993. The Slavonic languages. London: Routledge.
Comrie, Bernard (ed.) 2018. The world’s major languages, 3rd edn. Abingdon: Routledge.
Cook, V. 2003. Effects of the second language on the first. Clevedon: Multilingual Matters.
Coulmas, Florian. 1989. Writing systems of the world. Oxford: Blackwell. (2nd edn, 1991).
Coulmas, Florian. 1996. The Blackwell Encyclopedia of Writing Systems. Oxford: Blackwell.
Coulmas, Florian. 2002. Writing systems. An introduction to their linguistic analysis.
Cambridge: Cambridge University Press.
Courlander, Harold. 1971. The fourth world of the Hopis: The epic story of the Hopi Indians as
preserved in their legends and traditions. 5th printing 1992. Albuquerque: University of
New Mexico Press.
Crowley, Terry, and Claire Bowern. 2010. An introduction to historical linguistics, 4th edn.
Oxford: Oxford University Press.
Crystal, David. 1988. The English language. London: Penguin.
Crystal, David. 1995. The Cambridge encyclopedia of the English language. Cambridge:
Cambridge University Press.
Crystal, David. 2000. Language death. Cambridge: Cambridge University Press.
Cuny, Albert. 1946. Invitation à l’étude comparative des langues indo-européennes et des
langues chamito-sémitiques. Bordeaux: Brière.
Cutts, Martin, and Chrissie Maher. 1984. Gobbledygook. London: George Allen & Unwin.
Dalby, David. 1972. The African element in Black American English. Rappin’ and stylin’ out, ed.
by T. Kochman, 170–186. Urbana: University of Illinois Press.
Danesi, Marcel. 1993. Vico, metaphor, and the origin of language. Bloomington: Indiana
University Press.
Daniels, Peter T. 1990. Fundamentals of grammatology. Journal of the American Oriental Society
110: 727–731.
Daniels, Peter T., and William Bright (eds.) 1995. The world’s writing systems. Oxford: Oxford
University Press.
Daniels, Peter. 2018. An exploration of writing. Sheffield: Equinox Publishing.
Davies, Peter. 1981. Roots: Family histories of familiar words. New York: McGraw-Hill.
Davies, William Vivian. 1987. Reading the past: Egyptian hieroglyphs. London: British Museum.
 References 519

Dawson, Hope C., and Brian D. Joseph (eds.) 2013. Historical linguistics (Critical concepts in
linguistics, 6 volumes). London: Routledge.
Décsy, Gyula. 1990. The Uralic protolanguage: A comprehensive reconstruction. Bloomington,
IN: Eurolingua.
Demiraj, Shaban. 2004. Gjuhësi ballkanike. 2nd edn. Tiranë: Akademia e Shkencave e
Republikës së Shqipërise (Instituti i Gjuhësisë dhe i Letërisë) [1st edn., 1994]
Demiraj, Shaban. 2006. The origin of the Albanians (linguistically investigated). Tirana:
Academy of Sciences of Albania.
Dempwolff, Otto. 1938. Vergleichende Lautlehre des austronesischen Wortschatzes, vol. 3:
Austronesisches Wörterverzeichnis. (Beihefte zur Zeitschrift für Eingeborenen-Sprachen,
19.) Berlin: Reimer.
Diebold, A. Richard, Jr. 1985. The evolution of Indo-European nomenclature for salmonid fish.
(Journal of Indo-European Studies, Monograph 5). McLean (VA): Institute for the Study of
Man.
Dillard, Joey L. 1972. Black English: Its history and usage in the United States. New York:
Random House.
Dillard, Joey L. 1976. American talk: Slang and American usage. New York: Random House.
Diringer, David. 1962. Writing. London: Thames and Hudson.
Dixon, R. M. W. 1980. The languages of Australia. Cambridge: Cambridge University Press.
Dixon, Robert M. W. 1990. Summary report: Linguistic change and reconstruction in the
Australian language family. In Baldi 1990: 393–401.
Dixon, R. M. W. 2002. Australian languages: Their nature and development. Cambridge:
Cambridge University Press.
Djakonov, Igor M. 1985. On the original home of the speakers of Indo-European. Journal of
Indo-European Studies 13: 92–174.
Doke, Clement Martyn. 1954. The Southern Bantu languages. London/New York/Capetown:
Oxford University Press/International African Institute.
Dolgopolsky, Aron. 1989. Cultural contacts of Proto-Indo-European and Proto-Indo-Iranian with
neighbouring languages. Folia Linguistica Historica 8: 3–36.
Dorian, Nancy D. 1981. Language death: The life cycle of a Scottish Gaelic dialect. Philadelphia:
University of Pennsylvania Press.
Dorian, Nancy D. (ed.) 1989. Investigating obsolescence: Studies in language contraction and
death. Cambridge: Cambridge University Press.
Dorian, Nancy D. 1993. A response to Ladefoged’s other view on endangered languages.
Language 69: 575–579.
Douaud, Patrick. 1985. Ethnolinguistic profile of the Canadian Metis. Canadian Ethnology
Service Paper No. 99. (National Museum of Man Mercury Series.)
Dressler, Wolfgang U. 1972. On the phonology of language death. Papers from the 8th Regional
Meeting of the Chicago Linguistic Society, 448–457. Chicago: University of Chicago
Department of Linguistics.
Dressler, Wolfgang U. 2003. Naturalness and Morphological Change. In Joseph and Janda 2003:
461–471.
Dressler, Wolfgang U., and Ruth Wodak (eds.) 1977. Language death. (International Journal of
the Sociology of Language 12; published also as vol. 191 of the journal Linguistics, 1977.)
Dumézil, Georges. 1958. L’idéologie tripartite des Indo-Européens. (Collection LATOMUS, 31.)
Brussels: LATOMUS, Revue d’Études Latines.
Dunkling, Leslie. 1977. First names first. New York: Universe Books.
520 References

Dunne, Julie, Anna Maria Mercuri, Richard P. Evershed, Silvia Bruni, and Savino di Lernia. 2016.
Earliest direct evidence of plant processing in prehistoric Saharan pottery. Nature Plants 3,
article 16194.
Durie, Mark, and Malcolm Ross (eds.) 1996. The comparative method reviewed: Regularity and
irregularity in language change. Oxford: Oxford University Press.
Emeneau, Murray B. 1956. India as a linguistic area. Language 32: 3–16.
Emeneau, Murray B. 1980. Language and linguistic area: Essays selected by A. S. Dil. Stanford:
Stanford University Press.
Erard, Michael, and Catherine Matacic. 2018. Did kindness prime our species for language?
Science 361 (6401): 436–437 (DOI: 10.1126/science.361.6401.436).
Evans, Nicholas (ed.) 2003. The Non-Pama-Nyungan languages of Northern Australia:
Comparative studies of the continent’s most linguistically complex region. Canberra:
Pacific Linguistics, Research School of Pacific and Asian Studies, Australian National
University.
Everett, Daniel L. 2017. How language began: The story of humanity’s greatest invention.
London: Liveright Publishing Co.
Ewen, Cecil L. 1931. A history of surnames of the British Isles: A concise account of their origin,
evolution, etymology, and legal status. London: K. Paul, Trench, Trubner. (Reprinted 1968,
Detroit: Gale Research Co.)
Falk, Harry. 1990. Goodies for India: Literacy, orality, and Vedic culture. Erscheinungsformen
kultureller Prozesse, ed. by W. Raible, 103–120. (ScriptOralia, 13.) Tübingen: Gunter Narr.
Falk, Harry. 1993. Schrift im alten Indien: Ein Forschungsbericht mit Anmerkungen. Tübingen:
Gunter Narr.
Fansa, Mamoun, and Stefan Burmeister (eds.) 2004. Rad und Wagen: Der Ursprung einer
Innovation: Wagen im Vorderen Orient und Europa. (Beiheft der Archäologischen
Mitteilungen aus Nordwestdeutschland 40.) Mainz: Verlag Philipp von Zabern.
Feist, Sigmund. 1913. Kultur, Ausbreitung und Herkunft der Indogermanen. Berlin: Weidmann.
Feldman, David. 1989. Who put the butter in butterfly? New York: Harper & Row.
Feldman, Gilda, and Phil Feldman. 1994. Acronym soup: A stirring guide to our newest word
form. New York: William Morrow and Co.
Ferguson, Charles A. 1959. Diglossia. Word 15: 325–340.
Ferguson, Charles A. 1971. Absence of copula and the notion of simplicity: A study of normal
speech, baby talk, foreigner talk, and pidgins. In Hymes 1971: 141–150.
Feuillet, Jack. 1986. La linguistique balkanique. (Cahiers Balkaniques No. 10). Paris:
INALCO.
Feuillet, Jack. 2012. Linguistique comparée des langues balkaniques. Paris: Institut d’études
Slaves.
Fischer, Olga, Muriel Norde, and Harry Perridon (eds.) 2004. Up and down the cline: The nature
of grammaticalization. Amsterdam/Philadelphia: Benjamins.
Fischer, Olga. 2019. What role do iconicity and analogy play in grammaticalization? In Janda et
al., Chapter 15.
Fishman, Joshua A. 1991. Reversing language shift. Clevedon: Multilingual Matters.
Fisiak, Jacek (ed.) 1990. Historical linguistics and philology. Berlin/New York: Mouton de
Gruyter.
Fitch, W. Tecumseh. 2010. The evolution of language. Cambridge: Cambridge University Press.
Flexner, Stuart B. 1982. Listening to America: An illustrated history of words and phrases from
our lively and splendid past. New York: Simon and Schuster.
 References 521

Forster, Peter, and Colin Renfrew. 2006. Phylogenetic methods and the prehistory of languages.
Cambridge: McDonald Institute.
Fortson, Benjamin. 2004. Indo-European language and culture: An introduction. Oxford:
Blackwell.
Fox, Anthony. 1995. Linguistic reconstruction: An introduction to theory and method. Oxford:
Oxford University Press.
Fraenkel, Ernst. 1950. Die baltischen Sprachen. Heidelberg: Winter.
Francis, W. Nelson. 1983. Dialectology: An introduction. New York: Longman.
Frawley, William (ed.) 2003. International encyclopedia of linguistics, 2nd edn. Oxford: Oxford
University Press.
Friedman, Victor A. 2006. The Balkan languages. In Brown 2006: 1: 657–672.
Friedman, Victor A. and Brian D. Joseph. 2020. The Balkan languages. Cambridge: Cambridge
University Press.
Friedrich, Johannes. 1957. Extinct languages. Philosophical Library.
Frishberg, Nancy Jo. 1975. Arbitrariness and iconicity: Historical change in American Sign
Language. Language 51: 696–719.
Frishberg, Nancy Jo. 1976. Some aspects of the historical development of signs in American
Sign Language. PhD dissertation, University of California, San Diego.
Funk, Charles E. 1985. A hog on ice and other curious expressions. New York: Harper & Row.
Funk, Charles E. 1986. Heavens to Betsy! and other curious sayings. New York: Harper & Row.
Gamkrelidze, Thomas Valerianovich. 1988. The Indo-European glottalic theory in the light of
recent critique. Folia Linguistica Historica 9: 3–12.
Gamkrelidze, Thomas Valerianovich, and Vjacheslav V. Ivanov. 1973. Sprachtypologie und die
Rekonstruktion der gemeinindogermanischen Verschlüsse. Phonetica 27: 150–156.
Gamkrelidze, Thomas Valerianovich, and Vjacheslav V. Ivanov. 1984. Indoevropejskij jazyk
i indoevropejcy: Rekonstrukcija i istoriko-tipologičkij analiz prajazyka i protokul’tury
[Indo-European and the Indo-Europeans: A reconstruction and historical typological
analysis of a protolanguage and a proto-culture.] Tbilisi: Publishing House of the Tbilisi
State University.
Gamkrelidze, Thomas Valerianovich, and Vjacheslav V. Ivanov. 1985. The ancient Near East and
the Indo-European question and the migration of tribes speaking Indo-European dialects.
Journal of Indo-European Studies 13: 2–91.
Gamkrelidze, Thomas Valerianovich, and Vjacheslav V. Ivanov. 1990. The early history of
Indo-European languages. Scientific American, March 1990: 110–116.
Gamkrelidze, Thomas Valerianovich, and Vjacheslav V. Ivanov. 1994. Indo-European and the
Indo-Europeans. Berlin/New York: Mouton de Gruyter.
Gauchat, Louis. 1905. L’unité phonétique dans le patois d’une commune. Aus romanischen
Sprachen und Literaturen: Festschrift Heinrich Morf, 175–232. Halle: Niemeyer.
Gelb, Ignace J. 1963. A study of writing. Revised edition. Chicago: University of Chicago Press.
Gibson, Kathleen, and Tim Ingold (eds.) 1993. Tools, language, and cognition in human
evolution. Cambridge, MA: Harvard University Press.
Giles, H., Coupland, J., and N. Coupland. 1991. Contexts of accommodation: Developments in
applied sociolinguistics. Cambridge: Cambridge University Press.
Gilliéron, Jules. 1915. Pathologie et thérapeutique verbales. Paris: Champion.
Gilliéron, Jules. 1918. Généalogie des mots qui designent l’abeille. Paris: Champion.
Gimbutas, Marija. 1970. Proto-Indo-European culture: The Kurgan culture during the fifth,
fourth, and third millennia BC. In Cardona et al. 1970: 155–197.
522 References

Gimbutas, Marija. 1985. Primary and secondary homeland of the Indo-Europeans. Journal of
Indo-European Studies 13: 185–202.
Gobineau, Joseph Arthur, Comte de. 1853. Essai sur l’inégalité des races humaines, 1. Paris/
Hanover: Firmin Didot Frères/Rumpler.
Goddard, Ives. 1990. Algonquian linguistic change and reconstruction. In Baldi 1990: 99–114.
Gordon, Cyrus H. 1982. Forgotten scripts: Their ongoing discovery and decipherment. Repr.
1993, New York: Barnes & Noble.
Goyvaerts, Didier L. 1975. Present-day historical and comparative linguistics: Introductory
guide to theory and method. Ghent: E. Story-Scientia P.V.B.A.
Grassmann, Hermann. 1863. Ueber die Aspiraten und ihr gleichzeitiges Vorhandensein im An-
und Auslaute der Wurzeln. Zeitschrift für vergleichende Sprachforschung auf dem Gebiete
des Deutschen, Griechischen und Lateinischen, 12: 2. 81–138.
Gray, Russell D., and Quentin D. Atkinson. 2003. Language-tree divergence times support the
Anatolian theory of Indo-European origin. Nature 426: 435–439.
Greenberg, Joseph H. 1957. Essays in linguistics. Repr. as Phoenix Book P 119, 1963.
Greenberg, Joseph H. 1966. The languages of Africa. Bloomington, IN/The Hague: Indiana
University / Mouton.
Greenberg, Joseph H. 1987. Language in the Americas. Stanford: Stanford University Press.
Greenberg, Joseph H. 1989. Classification of American Indian languages: A reply to Campbell.
Language 65: 107–114.
Greenberg, Joseph H., and Merritt Ruhlen. 1992. Linguistic origins of Native Americans.
Scientific American, November 1992: 94–99.
Gregor, Douglass Bartlett. 1980. Celtic: A comparative study of the six Celtic languages: Irish,
Gaelic, Manx, Welsh, Cornish, Breton seen against the background of their history,
literature, and destiny. Cambridge/New York: The Oleander Press.
Grenoble, Lenore, and Lindsay Whaley. 2006. Saving languages: An introduction to language
revitalization. Cambridge: Cambridge University Press.
Grimm, Jacob. 1819–1834. Deutsche Grammatik, 4 vols. Göttingen: Dieterich.
Grimm, Jacob. 1893. Deutsche Grammatik, 2nd edn., v. 1. Gütersloh: Bertelsmann. (Engl. transl.
of pp. 580–592 in Lehmann 1967: 46–60.)
de Grolier, Eric, Andrew Lock, Charles R. Peters, and Jan Wind (eds.) 1983. Glossogenetics:
The origin and evolution of language. (Models of scientific thought, 1). Chur and London:
Harwood Academic Publishers.
Grose, Francis. 1796. A classical dictionary of the vulgar tongue, ed. by E. Partridge. Repr. 1992,
New York: Dorset Press.
Guenther, C. A., B. Tasic, L. Luo, M. A. Bedell, and D. M. Kingsley. 2014. A molecular basis for
classic blond hair color in Europeans. Nature Genetics 46: 748–752.
Guggenheimer, Heinrich W., and Eva H. Guggenheimer. 1992. Jewish family names and their
origins: An etymological dictionary. Hoboken, NJ: Ktav Publishing House.
Gumperz, John J., and Robert Wilson. 1971. Convergence and creolization: A case from the
Indo-Aryan/Dravidian border. In Hymes 1971: 151–168.
Günther, Hans F. K. 1934. Die Nordische Rasse bei den Indogermanen Asiens: Zugleich ein
Beitrag zur Frage nach der Urheimat und Rassenherkunft der Indogermanen. München:
Lehmann.
Günther, Hartmut, and Otto Ludwig (eds.) 1994–1995. Schrift und Schriftlichkeit/Writing and
its use: Ein interdisziplinäres Handbuch internationaler Forschung/An interdisciplinary
handbook of international research, 2 vols. Berlin: de Gruyter.
 References 523

Gurney, Oliver R. 1954. The Hittites, 2nd edn. Baltimore: Penguin Books.
Gvozdanović, Jadranka (ed.) 1992. Indo-European numerals. Berlin/New York: Mouton de
Gruyter.
Haak, Wolfgang, and 38 co-authors. 2015. Massive migration from the steppe is a source for
Indo-European languages in Europe. Nature 522: 207–211.
Habick, Timothy. 1980. Sound change in Farmer City: A sociolinguistic study based on acoustic
data. Urbana: University of Illinois PhD dissertation.
Haider, Hubert. 1985. The fallacy of typology: Remarks on the PIE stop-system. Lingua 65:
1–27.
Hale, Horatio. 1890. An international idiom: A manual of the Oregon Trade Language or Chinook
Jargon. London: Whittaker & Co.
Hale, Kenneth, et al. 1992. Endangered languages. Language 68: 1–42.
Hale, Mark. 2007. Historical linguistics: Theory and method. Oxford: Blackwell.
Hall, Robert A., Jr. 1943. Melanesian Pidgin English. Baltimore: Linguistic Society of America.
Hall, Robert A., Jr. 1950. The reconstruction of Proto-Romance. Language 26: 6–27.
Hall, Robert A., Jr. 1960. On realism in reconstruction. Language 36: 203–206.
Hall, Robert A., Jr. 1966. Pidgin and creole languages. Ithaca, NY: Cornell University Press.
Hall, Robert A., Jr. 1968. Comparative reconstruction in Romance syntax. Acta Linguistica
Hafniensia 11: 81–88.
Hall, Robert A., Jr. 1976. Proto-Romance phonology. (Comparative Romance grammar, 2.) New
York: Elsevier.
Hamp, Eric P. 1972. Albanian. Current trends in linguistics, vol. 9: Linguistics in western Europe,
ed. by T. A. Sebeok, 1626–1692. The Hague: Mouton.
Hamp, Eric P. 1992. On misusing similarity. Explanation in historical linguistics, ed. by
G. W. Davis and G. K. Iverson, 95–103. Amsterdam/Philadelphia: Benjamins.
Hancock, Ian. 1990. Creolization and language change. In Polomé 1990a: 507–525.
Hanks, Patrick, and Flavia Hodges. 1988. A dictionary of surnames. New York: Oxford University
Press.
Hanks, Patrick, and Flavia Hodges. 1990. A dictionary of first names. New York: Oxford
University Press.
Harris, Alice C., and Lyle Campbell. 1995. Historical syntax in cross-linguistic perspective.
Cambridge: Cambridge University Press.
Harrison, Henry. 1918. Surnames of the United Kingdom: A concise etymological dictionary.
London. Repr., Baltimore: Clearfield Co.
Harrison, K. David. 2007. When languages die: The extinction of the world’s languages and the
erosion of human knowledge. Oxford: Oxford University Press.
Hatch, Evelyn M. 1978. Discourse analysis and second language acquisition. Second language
acquisition: A book of readings, ed. by E. M. Hatch. Newbury House.
Haudricourt, André-Georges. 1954. Comment reconstruire le chinois archaïque. Word 10:
351–364.
Haugen, Einar. 1950. The analysis of linguistic borrowing. Language 26: 210–231.
Haugen, Einar. 1982. The Scandinavian languages: A comparative historical survey.
Minneapolis: University of Minnesota Press.
Hauser, Marc D., Noam Chomsky, Tecumseh W. Fitch. 2002. The faculty of language: what is it,
who has it, and how did it evolve? Science 298: 1569–1579.
Havers, Wilhelm. 1946. Neuere Literatur zum Sprachtabu. (Sitzungsberichte der Akademie der
Wissenschaften Wien, phil.-hist. Klasse, 223: 5.)
524 References

Heggarty, Paul. 2008. Calling the bluff on linguistic palaeontology: The horse, the wheel and …
the king? (Abstract of paper at Workshop “New Directions in Historical Linguistics”, Lyon
2008. http://www.ddl.ish-lyon.cnrs.fr/colloques/NDHL2008/Powerpoint/13-Heggarty.pdf.
Hendrickson, Robert. 1983. Animal crackers: A bestial lexicon. New York: Viking Press.
Hendrickson, Robert. 1988. The dictionary of eponyms: Names that became words. New York:
Dorset Press. (Reprint of Human words, 1972, Philadelphia: Chilton Book Co.)
Hermann, Eduard. 1929. Lautveränderungen in der Individualsprache einer Mundart.
Nachrichten der Gesellschaft der Wissenschaften zu Göttingen, phil.-hist. Klasse, 9:
195–214.
Hess, Elizabeth. 2008. Nim Chimpsky: The chimp who would be human. New York: Random
House.
Hewitt, B. George (ed.) 1989. The indigenous languages of the Caucasus. Delmar, NY: Caravan
Books.
Hill, Jane. 1998. Language, race, and white public space. American Anthropologist 100:
3.680–689.
Hinton, Leanne, and Kenneth Hale (eds.) 2001. The green book of language revitalization in
practice. San Diego: Academic Press.
Hinton, Leanne. 1994. Flutes of fire: Essays on California Indian languages. Berkeley, CA:
Heyday Books.
Hock, Hans Henrich. 1975. Substratum influence on (Rig-Vedic) Sanskrit? Studies in the
Linguistic Sciences 5: 2. 76–125.
Hock, Hans Henrich. 1982. AUX-cliticization as a motivation for word order change. Studies in
the Linguistic Sciences 12: 1. 91–101.
Hock, Hans Henrich. 1983. Language-death phenomena in Sanskrit: Grammatical evidence
for attrition in contemporary spoken Sanskrit. Studies in the Linguistic Sciences 13: 2.
21–35.
Hock, Hans Henrich. 1984. (Pre-)Rig-Vedic convergence of Indo-Aryan with Dravidian? Another
look at the evidence. Studies in the Linguistic Sciences 14: 1. 89–107.
Hock, Hans Henrich. 1985. Yes, Virginia, syntactic reconstruction is possible. Studies in the
Linguistic Sciences 15: 1. 49–60.
Hock, Hans Henrich. 1986. Compensatory lengthening: In defense of the concept ‘mora’. Folia
Linguistica 20: 431–460.
Hock, Hans Henrich. 1988. Historical implications of a dialectological approach to convergence.
Historical dialectology, ed. by J. Fisiak, 283–328. Berlin/New York: Mouton de Gruyter.
Hock, Hans Henrich. 1991. Principles of historical linguistics, 2nd edn. Berlin/New York: Mouton
de Gruyter.
Hock, Hans Henrich. 1992. Spoken Sanskrit in Uttar Pradesh: Profile of a dying prestige
language. Dimensions of sociolinguistics in South Asia: Papers in memory of Gerald
Kelley, ed. by E. C. Dimmock, B. B. Kachru, and Bh. Krishnamurti, 247–260. New Delhi:
Oxford University Press.
Hock, Hans Henrich. 1994. Swallow tales: Chance and the “world etymology” Maliq’a ‘swallow,
throat’. Papers from the 29th Regional Meeting of the Chicago Linguistic Society, 1:
215–238. Chicago: University of Chicago Department of Linguistics.
Hock, Hans Henrich. 1996. Pre-Ṛgvedic convergence between Indo-Aryan (Sanskrit) and
Dravidian? A survey of the issues and controversies. Ideology and status of Sanskrit:
Contributions to the history of the Sanskrit language, ed. by J. E. M. Houben, 17–58.
Leiden: Brill.
 References 525

Hock, Hans Henrich. 1999. Through a glass darkly: Modern “racial” interpretations vs. textual
and general prehistoric evidence on ārya and dāsa/dasyu in Vedic society. Aryan and
Non-Aryan in South Asia: Evidence, interpretation, and ideology, Proceedings of the
International Seminar on Aryan and Non-Aryan in South Asia, University of Michigan, Ann
Arbor, 25–27 October, 1996, 145–174, ed. by Johannes Bronkhorst and Madhav Deshpande.
Harvard Oriental Series, Opera Minora, 3. Repr. 2012, Delhi: Manohar.
Hock, Hans Henrich. 2003. Analogical Change. In Joseph and Janda 2003: 441–460.
Hock, Hans Henrich. 2014. A morphosyntactic chain shift in the Hindi-Panjabi area: Explications
and implications. Journal of South Asian Languages and Linguistics 1: 5–30.
Hock, Hans Henrich. 2016. Old and Middle Indo-Aryan. In Hock & Bashir 2016: 18–35.
Hock, Hans Henrich, and Elena Bashir (eds.). 2016. The languages and linguistics of South
Asia: A comprehensive guide. Berlin/Boston: de Gruyter Mouton.
Hockett, Charles, and Robert Ascher. 1964. The human revolution. Current Anthropology
5.135–168.
Hodge, Carleton T. 1990. The role of Egyptian within Afroasiatic(/Lislakh). In Baldi 1990:
639–659.
Hoenigswald, Henry M. 1964. Graduality, sporadicity, and the minor sound change processes.
Phonetica 11: 202–215.
Hoenigswald, Henry M. 1965. Language change and linguistic reconstruction. Chicago:
University of Chicago Press.
Holm, Hans J. 2007. The new arboretum of Indo-European “trees”: Can new algorithms reveal
the phylogeny and even prehistory of IE? Journal of Quantitative Linguistics 14:2–3.
167–214.
Holm, John. 1989. Pidgins and creoles, 2 vols. Cambridge: Cambridge University Press.
Hook, Julius N. 1982. Family names: How our surnames came to America. New York: Macmillan.
Hoops, Johannes. 1905. Waldbäume und Kulturpflanzen im germanischen Altertum. Straßburg:
Trübner.
Hopper, Paul J. 1973. Glottalized and murmured occlusives in Indo-European. Glossa 7:
141–166.
Hopper, Paul, and Elizabeth Traugott. 2003. Grammaticalization, 2nd edn. Cambridge:
Cambridge University Press.
Horrocks, Geoffrey. 2010. Greek: A history of the language and its speakers, 2nd edn.
Chichester: Wiley Blackwell.
Houston, Stephen D. 1989. Reading the past: Maya glyphs. London: British Museum.
Hualde, José Ignacio, Joseba A. Lakarra, and R. L. Trask. 1996. Towards a history of the Basque
language. Amsterdam/Philadelphia: Benjamins.
Hurford, James R. 2007. The origins of meaning. Oxford: Oxford University Press.
Hurford, James R. 2012. The origins of grammar: Language in the light of evolution II. Oxford:
Oxford University Press.
Hurford, James. R. 2014. The origins of language. A slim guide. Oxford: Oxford University Press.
Hymes, Dell (ed.) 1971. Pidginization and creolization of language. Cambridge: Cambridge
University Press.
Illič-Svityč, Vladislav Markovič. 1964. Drevnejšie indoevropejsko-semitskie jazykovye kontakty
[The oldest contacts between Indo-European and Semitic languages]. Problemy indoev-
ropejskogo jazyknoznanija, ed. by V. Toporov, 3–12. Moscow: Nauka.
Illič-Svityč, Vladislav Markovič. 1971, 1976. Opyt sravnenija nostratičeskix jazykov [An attempt at
reconstructing Nostratic]. 2 vols. Moscow: Nauka.
526 References

Jaberg, Karl. 1908. Sprachgeographie. Aarau: Sauerländer.

Janda, Richard D., Brian D. Joseph, and Neil G. Jacobs. 1994. Systematic hyperforeignisms
as maximally external evidence for linguistic rules. The reality of linguistic rules, ed. by
S. D. Lima et al., 67–92. Amsterdam/Philadelphia: Benjamins.
Janda, Richard D., Brian D. Joseph, and Barbara Vance (eds.) 2019. Handbook of historical
linguistics, vol. 2. Oxford: Wiley.
Jankowsky, Kurt R. 1990. The Neogrammarian hypothesis. In Polomé 1990a: 223–239.
Janse, Mark, and Sijmen Tol (eds.) 2003. Language death and language maintenance:
Theoretical, practical, and descriptive approaches. Amsterdam/Philadelphia: Benjamins.
Jeffers, Robert J., and Ilse Lehiste. 1979. Principles and methods for historical linguistics.
Cambridge, MA: MIT Press.
Jeffery, Lilian Hamilton. 1961. The local scripts of archaic Greece. Oxford: Clarendon Press.
Jenny, Mathias, and Paul Sidwell (eds.) 2015. The handbook of Austroasiatic languages. Leiden:
Brill.
Jensen, Hans. 1958. Die Schrift in Vergangenheit und Gegenwart. Berlin: VEB Deutscher Verlag
der Wissenschaften.
Jensen, Hans. 1970. Sign, symbol, and script, transl. of 3rd edition of Jensen 1958. London: Allen
& Unwin.
Jespersen, Otto. 1921. Language: Its nature, development, and origin. Repr. 1964, New York:
W. W. Norton & Company.
Jespersen, Otto. 1941. Efficiency in linguistic change. (Det Kgl. Danske Videnskabernes Selskab,
27: 4.)
Jha, N., and Navaratna Srinivasa Rajaram. 2000. The deciphered Indus script: Methodology,
readings, interpretations. New Delhi: Aditya Prakashan.
Johnson, Scott A. J. 2013. Translating Maya hieroglyphs. Norman: University of Oklahoma
Press.
Jones, Charles (ed.) 1993. Historical linguistics: Problems and perspectives. London/New York:
Longman.
Jones, Eppie R., and 22 co-authors. 2015. Upper Palaeolithic genomes reveal deep roots of
modern Eurasians. Nature Communications 6, Article 8912.
Jones, Sir William. 1786. The third anniversary discourse, on the Hindus. Asiatick Researches 1:
422. (Published 1788; repr. in Lehmann 1967.)
Joseph, Brian D. 1983. The synchrony and diachrony of the Balkan infinitive: A study in areal,
general, and historical linguistics. Cambridge: Cambridge University Press.
Joseph, Brian D. 1986. A fresh look at the Balkan Sprachbund: Some observations on H. W.
Schaller’s Die Balkansprachen. Mediterranean Language Review 3: 105–114.
Joseph, Brian D. 1992. The Balkan languages. In Bright 1992: 1: 153–155.
Joseph, Brian D. 1998. Diachronic morphology. Handbook of morphology, ed. by A. Zwicky and
A. Spencer, 351–373. Oxford: Blackwell Publishers.
Joseph, Brian D. 2003. The Balkan languages. In Frawley 2003: 194–196.
Joseph, Brian D. 2016. The comparative method: Simplicity + power = results. Veleia 33: 39–48
(Special issue for the Bopp Bicentennial).
Joseph, Brian D. 2017. Expanding the methodology of lexical examination in the investigation of
the intersection of early agriculture and language dispersal. In Robbeets 2017: 259–274.
Joseph, Brian D. 2019. Historical morphology: An overview and update. In Janda et al.,
Chapter 4.
 References 527

Joseph, Brian D., and Richard D. Janda (eds.) 2003. Handbook of historical linguistics. Oxford:
Blackwell.
Joseph, John. 1987. Eloquence and power: The rise of language standards and standard
languages. London: Pinter.
Justus, Carol F. 1988. Indo-European numerals and numeral systems. A linguistic happening in
memory of Ben Schwartz, ed. by Y. L. Arbeitman, 521–541. Louvain: Peeters.
Kachru, Braj B. 1965. The Indianness in Indian English. Word 21: 391–410.
Kachru, Braj B. 1986. The alchemy of English. Oxford: Pergamon Press. (Repr. 1990, Urbana:
University of Illinois Press.)
Kachru, Braj B. (ed.) 1992. The other tongue: English across cultures, 2nd edn. Urbana:
University of Illinois Press.
Kaganoff, Benzion G. 1977. A dictionary of Jewish names and their history. New York: Schocken
Books.
Kahane, Henry, Renée Kahane, and Andreas Tietze. 1958. The Lingua Franca in the
Mediterranean: Turkish nautical terms of Italian and Greek origin. Urbana: University of
Illinois Press. (Repr. 1988, Istanbul: ABC Kitabevi.)
Kahlo, Gerhard. 1941. Kleines vergleichendes malayo-polynesisches Wörterbuch. Leipzig:
Harrassowitz.
Karlgren, Bernhard. 1949. The Chinese language: An essay on its nature and history. New York:
Ronal Press.
Katzner, Kenneth. 1995. The languages of the world. London and New York: Routledge.
Kenneally, Christine. 2007. The first word: The search for the origins of language. New York:
Penguin.
Klaniczay, Gábor. 2011. The myth of Scythian origin and the cult of Attila in the nineteenth
century. Multiple antiquities – multiple modernities: Ancient histories in nineteenth
century European cultures, ed. by Gábor Klaniczay, Michael Werner, and Ottó Gecser,
185–212. Frankfurt: Campus Verlag.
Klein, Jared, Brian D. Joseph, and Matthias Fritz (eds). 2017–2018. Handbook of comparative
and historical Indo-European linguistics (3 volumes). Berlin: de Gruyter Mouton.
Kloeke, Gesinus Gerhardus. 1927. De Hollandsche expansie in de zestiende en zeventiende
eeuw. ’s Gravenhage: Nijhoff.
Kluge, Friedrich. 1901. Rotwelsch: Quellen und Wortschatz der Gaunersprache und der
verwandten Geheimsprachen. Repr. 1987, Berlin: de Gruyter.
Kolatch, Alfred J. 1989. The new name dictionary: Modern English and Hebrew names. Middle
Village, NY: J. David Publishers.
Kolichala, Suresh. 2016. Dravidian languages. In Hock & Bashir 2016: 73–107.
König, Ekkehard, and Johan van der Auwera (eds.) 1994. The Germanic languages. London/New
York: Routledge.
Korn, Agnes. 2016. Iranian. In Hock & Bashir 2016: 51–66.
Krahe, Hans. 1942. Review of Hans F. K. Günther, Herkunft und Rassengeschichte der
Germanen. Indogermanische Forschungen 58: 190–191.
Krahe, Hans. 1970. Einleitung in das vergleichende Sprachstudium, ed. by W. Meid.
(Innsbrucker Beiträge zur Sprachwissenschaft, 1.) Innsbruck.
Krishnamurti, Bhadriraju. 2003. The Dravidian languages. Cambridge: Cambridge University
Press.
Krogmann, Willy. 1955–1956. Das Buchenargument. Zeitschrift für vergleichende Sprach-
forschung 72: 1–29, 73: 1–25.
528 References

Kuhn, Sherman M. 1961. On the syllabic phonemes of Old English. Language 37: 522–538.
Kulkarni-Joshi, Sonal. 2008. Deconvergence in Kupwad? Indian Linguistics 69: 153–162.
Kulkarni-Joshi, Sonal. 2016. Forty years of language contact and change in Kupwar: A critical
assessment of the intertranslatability model. Journal of South Asian Languages and
Linguistics 3(2): 147–174.
Kümmel, Martin Joachim. 2017. Agricultural terms in Indo-Iranian. In Robbeets 2017: 275–290.
Kuz’mina, Elena Efimovna. 2007. The origins of the Indo-Iranians, ed. by J. P. Mallory. Leiden/
Boston: Brill.
Laakso, Johanna. 2004. Sprachwissenschaftliche Spiegelfechterei (Angela Marcantonio: The
Uralic language family. Facts, myths and statistics). Finnisch-Ugrische Forschungen 58:
296–307. English version at https://homepage.univie.ac.at/Johanna.Laakso/am_rev.html
Labov, William. 1963. The social motivation of a sound change. Word 19: 273–309.
Labov, William. 1965. On the mechanism of linguistic change. Georgetown University
Monographs on Languages and Linguistics 18: 91–114.
Labov, William. 1969. Contraction, deletion, and inherent variability of the English copula.
Language 45: 715–762.
Labov, William. 1981. Resolving the neogrammarian controversy. Language 57: 267–308.
Labov, William. 1994. Principles of linguistic change, vol. I: Internal factors. Cambridge, MA /
Oxford, England: Blackwell.
Labov, William. 2001. Principles of linguistic change, vol. II: Social factors. Oxford: Blackwell.
Labov, William. 2010. Principles of linguistic change, vol. III: Cognitive and cultural factors.
Oxford: Blackwell.
Labov, William, Sharon Ash, and Charles Boberg. 2006. The atlas of North American English:
Phonetics, phonology, and sound change. Berlin/New York: Mouton de Gruyter.
Ladefoged, Peter. 1992. Another view of endangered languages. Language 68: 809–811.
Ladefoged, Peter. 1993. A course in phonetics, 2nd edn. Fort Worth: Harcourt Brace Jovanovich
College Publishers.
Lamb, Sydney, and E. Douglas Mitchell (eds.) 1991. Sprung from some common source: Investi-
gations into the prehistory of languages. Stanford: Stanford University Press.
Lamberg-Karlovsky. 2002. Archaeology and language: The Indo-Iranians. Current Anthropology
43: 63–88.
Lambert, Eloise, and Mario Pei. 1961. Our names, where they come from and what they mean.
New York: Lothrop, Lee & Shepard Co.
Lambert, Richard D., and Barbara F. Freed (eds.) 1982. The loss of language skills. Rowley, MA:
Newbury House.
Lane, Harlan. 1980. Historical: A chronology of the oppression of sign language in France and
the United States. Recent perspectives on American Sign Language, ed. by H. Lane and F.
Grosjean, 119–161. Hillsdale, NJ: Lawrence Erlbaum Associates.
Langdon, Margaret. 1974. Comparative Hokan-Coahuiltecan studies. The Hague: Mouton.
Large, Andrew. 1985. The artificial language movement. Oxford: Basil Blackwell.
Lassen, Christian. 1847. Indische Alterthumskunde, v. 1. Bonn: Koenig.
Latham, Robert Gordon. 1851. The Germania of Tacitus, with ethnological dissertations and
notes. London: Taylor, Walton & Maberly.
Lazaridis, Iosif, and 119 co-authors. 2014. Ancient human genomes suggest three ancestral
populations for present-day Europeans. Nature 513: 409–416.
Ledgeway, Adam, and Martin Maiden (eds.) 2016. The Oxford guide to the Romance languages.
Oxford: Oxford University Press.
 References 529

Lehiste, Ilse. 1988. Lectures on language contact. Cambridge, MA: MIT Press.
Lehmann, Winfred P. (ed.) 1967. A reader in 19th century historical Indo-European linguistics.
Bloomington: Indiana University Press.
Lehmann, Winfred P. 1992. Historical linguistics: An introduction, 3rd edn. London and New York:
Routledge
Lehmann, Winfred P. 1993. Theoretical bases of Indo-European linguistics. London: Routledge.
Lehmann, Winfred P., and Ladislav Zgusta. 1979. Schleicher’s tale after a century. Festschrift
for Oswald Szemerényi, ed. by Bela Brogyanyi, 1: 455–466. Amsterdam/Philadelphia:
Benjamins.
Libert, Alan. 2000. A priori artificial languages. Munich: Lincom.
Libert, Alan. 2003. Mixed artificial languages. Munich: Lincom.
Lieberman, Philip. 1975. On the origins of language: An introduction to the evolution of human
speech. New York: Macmillan.
Lieberman, Philip. 1984. The biology and evolution of language. Cambridge, MA: Harvard
University Press.
Lieberman, Philip. 1998. Eve spoke: Human language and human evolution. New York: W. W.
Norton & Co.
Lieberman, Stephen J. 1990. Summary report: Linguistic change and reconstruction in the
Afro-Asiatic languages. In Baldi 1990: 565–575.
Lightfoot, David. 1999. The Development of language: Acquisition, change, and evolution.
Oxford: Blackwell.
Lipton, James. 1991. An exaltation of larks: “The ultimate edition.” New York: Viking Books.
Lloshi, Xhevat. 1999. Albanian. Handbuch der Südosteuropa-Linguistik, ed. by Uwe Hinrichs,
277–299. Wiesbaden: Harrassowitz.
Lock, Andrew (ed.) 1978. Action, gesture, and symbol: The emergence of language. London/
New York: Academic Press.
Lockwood, William B. 1969. Indo-European philology: Historical and comparative. London:
Hutchinson University Library.
Lockwood, William B. 1972. A panorama of Indo-European languages. London: Hutchinson
University Library.
Lottner, Carl. 1862. Ausnahmen der ersten Lautverschiebung. Zeitschrift für vergleichende
Sprachwissenschaft 11: 161–205. (Transl. in Lehmann 1967.)
Lucas, Ceil (ed.) 1990a. Sign language research: Theoretical issues. Washington, DC: Gallaudet
University Press.
Lucas, Ceil. 1990b. ASL, English, and contact signing. In Lucas 1990a: 288–307.
Lucas, Ceil. 1992. Language contact in the American deaf community. San Diego: Academic
Press.
Lucas, Ceil and Clayton Valli. 1989. Language contact in the American deaf community. The
sociolinguistics of the deaf community, ed. by Ceil Lucas, 11–40. San Diego: Academic
Press.
Lucas, Ceil, Robert Bayley, and Clayton Valli. 2003. What’s your sign for pizza? An introduction
to variation in American Sign Language. Washington, DC: Gallaudet University Press.
Luraghi, Silvia and Vit Bubenik (eds.) 2010. Bloomsbury companion to historical linguistics.
London: Bloomsbury (originally published as Continuum companion to historical
linguistics, 2010).
Lutz, William (ed.) 1989. Beyond nineteen eighty-four: Doublespeak in a post-Orwellian age.
Urbana, IL: National Council of Teachers of English.
530 References

Lyovin, Anatole V., Brett Kessler, and William R. Leben. 2017. An introduction to languages of
the world. Oxford: Oxford University Press.
McAlpin, David W. 1974. Toward Proto-Elamo-Dravidian. Language 50: 89–101.
McAlpin, David W. 1975. Elamite and Dravidian: Further evidence of relationship (with
discussion by M. B. Emeneau, W. H. Jacobsen, F. B. J. Kuiper, H. H. Paper, E. Reiner,
R. Stopa, F. Vallat, R. W. Wescott, and a reply by D. W. McAlpin.) Current Anthropology 16:
105–115.
McAlpin, David W. 1981. Proto-Elamo-Dravidian: The evidence and its implications.
(Transactions of the American Philosophical Society, 71: 3.) Philadelphia.
MacAulay, Donald (ed.) 1992. The Celtic languages. (Cambridge Language Surveys.) Cambridge:
Cambridge University Press.
MacBain, Alexander. 1911. An etymological dictionary of the Gaelic language, 2nd revised
edition. Repr. 1982, Glasgow: Gairm Publications.
Malkiel, Yakov. 1993. Etymology. Cambridge: Cambridge University Press.
Mallory, J.P., and D. Q[uincy] Adams (eds.) 1997. Encyclopedia of Indo-European Culture.
London: Fitzroy Dearborn Publishers.
Mallory, J. P., and D. Q[uincy] Adams. 2006. The Oxford introduction to Proto-Indo-European
and the Proto-Indo-European world. Oxford/New York: Oxford University Press.
Malone, Caroline, et al. 1993. The death cults of prehistoric Malta. Scientific American,
December 1993: 110–117.
Manaster-Ramer, Alexis. 1992. On anecdotal universals in historical linguistics. Diachronica 9:
135–137.
Mańczak, Witold. 1987. Frequenzbedingter unregelmässiger Lautwandel in den germanischen
Sprachen. Wrocław: Polska Akademia Nauk.
Marcantonio, Angela. 2002. The Uralic language family: Facts, myths and statistics. Oxford/
Boston: Philological Society.
Marcantonio, Angela. 2009. Evidence that most Indo-European lexical reconstructions are
artefacts of the linguistic method of analysis. The Indo-European language family:
Questions about its status, ed. by A. Marcantonio, 10: 1–46. Washington, DC: Institute for
the Study of Man.
Markey, Thomas, and John Greppin (eds.) 1990. When worlds collide: Indo-Europeans and
pre-Indo-Europeans. Ann Arbor, MI: Karoma.
Martinet, André. 1964. Économie des changements phonétiques, 2nd edn. Bern: Francke.
Masica, Colin P. 1976. Defining a linguistic area: South Asia. Chicago: University of Chicago
Press.
Masica, Colin P. 1991. The Indo-Aryan languages. Cambridge: Cambridge University
Press.
Mathieson, Iain, and 16 co-authors. 2015. Eight thousand years of natural selection in Europe.
bioRxiv preprint doi: http://dx.doi.org/10.1101/016477.
Matisoff, James A. 1990. On megalocomparison. Language 66: 106–120.
Matras, Yaron, and Peter Bakker (eds.) 2008. The Mixed-Language debate. Berlin/New York:
Mouton de Gruyter.
Mattheier, Klaus J. 1983. Aspekte der Dialekttheorie. Tübingen: Niemeyer.
Mayrhofer, Manfred. 1955. Altindisch lakṣā́: Die Methoden einer Etymologie. Zeitschrift der
Deutschen Morgenländischen Gesellschaft 105: 175–183.
Mayrhofer, Manfred. 1986. Indogermanische Grammatik, 1. Heidelberg: Winter.
McKinley, Richard A. 1990. A history of British surnames. London/New York: Longman.
 References 531

McMahon, April M. S. 1994. Understanding language change. Cambridge: Cambridge

University Press.
Meier-Brügger, Michael. 2003. Indo-European Linguistics (in cooperation with Matthias
Fritz and Manfred Mayrhofer, translated by Charles Gertmenian). Berlin/New York: De
Gruyter.
Meillet, Antoine. 1925. La méthode comparative en linguistique historique. Paris: Librairie
Honore Champion. (“The comparative method in historical linguistics”, transl. by Gordon
B. Ford, Jr., 1970.)
Meillet, Antoine. 1937. Introduction à l’étude comparative des langues indo-européennes,
8th edn. Paris: Hachette. (Reprinted 1964, University of Alabama Press.)
Meinhof, Carl. 1899. Grundriss einer Lautlehre der Bantusprachen. (Abhandlungen für die
Kunde des Morgenlandes, 11: 2.) Repr. 1966, Kraus Reprint Limited.
Melchert, H. Craig. 1994. Anatolian historical phonology. Amsterdam and Atlanta: Rodopi.
Melchert, H. Craig. 1998. The dialectal position of Anatolian within Indo-European. Proceedings
of the Twenty-Fourth Meeting of the Berkeley Linguistics Society: Special session on
Indo-European subgrouping and internal relations, 24–31. Berkeley: Berkeley Linguistic
Society.
Melchert, H. Craig. 2012. Luvo-Lycian dorsal stops revisited. The sound of Indo-European 2, ed.
by Roman Sukač and Ondřej Šefčik, 206–218. München: Lincom.
Menner, Robert J. 1937. Hypercorrect forms in American English. American Speech
12.3:167–178.
Michaelis, Susanne Maria, Philippe Maurer, Martin Haspelmath, and Magnus Huber (eds.)
2013a. The survey of pidgin and creole languages, Volume 1. English-based and
Dutch-based languages. Oxford: Oxford University Press.
Michaelis, Susanne Maria, Philippe Maurer, Martin Haspelmath, and Magnus Huber (eds.)
2013b. The Atlas of pidgin and creole language structures. Oxford: Oxford University Press.
Millardet, Georges. 1933. Sur un ancien substrat commun à la Sicile, le Corse, et la Sardaigne.
Revue de linguistique romane 9: 346–369.
Miller, Roy Andrew. 1971. Japanese and the other Altaic languages. Chicago: University of
Chicago Press.
Miller, Roy Andrew. 1991. How many Verner’s Laws does an Altaicist need? Studies in the
historical phonology of Asian languages, ed. by W. G. Boltz and M. C. Shapiro, 176–204.
Amsterdam/Philadelphia: Benjamins.
Miller, Wick R. 1967. Uto-Aztecan cognate sets. (University of California Publications in
Linguistics, 48.) Berkeley/Los Angeles: University of California Press.
Mithun, Marianne. 1999. The languages of Native North America. Cambridge: Cambridge
University Press.
Møller, Hermann. 1907. Semitisch und Indogermanisch, vol. 1. Copenhagen.
Møller, Hermann. 1917. Die semitisch-vorindogermanischen laryngalen Konsonanten.
(Mémoires de l’Académie Royale des Sciences et Lettres de Danemark, 7me série, 4: 1.)
Copenhagen.
Moser, Margaret G. 1990. The regularity hypothesis applied to ASL. In Lucas 1990: 50–56.
Mühlhäusler, Peter. 1970. Growth and structure of the lexicon of New Guinea Pidgin. Canberra:
Pacific Linguistics, C-52.
Mühlhäusler, Peter. 1981. Foreigner Talk, Tok Masta in New Guinea. In Clyne 1981: 93–113.
Mühlhäusler, Peter. 1982. Tok Pisin in Papua New Guinea. English as a world language, ed. by
R. W. Bailey and M. Görlach, 439–466. Ann Arbor, MI: The University of Michigan Press.
532 References

Mühlhäusler, Peter. 1983. The development of word formation in Tok Pisin. Folia Linguistica 17:
463–487.
Müller, Friedrich Max. 1847. On the relation of the Bengali and the Arian and aboriginal
languages in India. Report of the British Association for the Advancement of Science 17:
319–350.
Müller, Friedrich Max. 1864. Lectures on the science of language, 2 vols, 4th ed. London:
Longman, Green, Longman, Roberts, & Green.
Muysken, Pieter. 1981. Halfway between Quechua and Spanish: The case for relexification.
Historicity and variation in creole studies, ed. by A. R. Highfield and A. Valdman, 52–78.
Ann Arbor, MI: Karoma.
Muysken, Pieter. 1997. Media Lengua. Contact languages: A wider perspective, ed. by Sarah G.
Thomason, 365–426. Amsterdam/Philadelphia: Benjamins.
Muysken, Pieter, and Norval Smith (eds.) 1986. Substrata versus universals in creole genesis.
Amsterdam/Philadelphia: Benjamins.
Nakanishi, Akira. 1992. Writing systems of the world: Alphabets, syllabaries, pictograms.
Rutland, VT: Charles E. Tuttle Co.
Naro, Anthony Julius. 1978. A study on the origins of pidginization. Language 54: 314–347.
Narrog, Heiko, and Bernd Heine (eds). 2011. The Oxford handbook of grammaticalization.
Oxford: Oxford University Press.
Norman, Jerry. 1988. Chinese (Cambridge Language Surveys.) Cambridge: Cambridge University
Press.
Ogden, Charles Kay, and Ivor Armstrong Richards. 1923. The meaning of meaning. New York:
Harcourt.
Olalde, Iñigo, and 23 co-authors. 2014. Derived immune and ancestral pigmentation alleles in a
7,000-year-old Mesolithic European. Nature 507: 225–228.
Onions, C. T. 1966. The Oxford dictionary of English etymology. Oxford: Clarendon Press.
Orel, Vladimir. 2000. A concise historical grammar of the Albanian language: Reconstruction of
Proto-Albanian. Leiden: Brill.
Osthoff, Hermann, and Karl Brugmann. 1878. (Preface to) Morphologische Untersuchungen
auf dem Gebiete der indogermanischen Sprachen, vol. 1. (Engl. transl. in Lehmann
1967.)
Outram, Alan K., Natalie A. Stear, Robin Bendrey, Sandra Olsen, Alexei Kasparov, Victor Zaibert,
Nick Thorpe, Richard P. Evershed. 2009. The earliest horse harnessing and milking.
Science 323: 1332–1335.
Oxford English Dictionary. 1884–1928. Oxford University Press. (Repr. 1933 with supplements.)
Oxford English Dictionary. 1989. 2nd edition, Oxford University Press.
Page, Raymond Ian. 1987. Reading the past: Runes. London: British Museum.
Palmer, Leonard R. 1954. The Latin language. London: Faber & Faber.
Palmer, Leonard R. 1980. The Greek language. Atlantic Highlands, NJ: Humanities Press, Inc.
Pardo, Jennifer S. 2006. On phonetic convergence during conversational interaction. Journal of
the Acoustical Society of America 119 (4): 2382–2393.
Pardo, Jennifer S., Isabel Cajori Jay, Risa Hoshino, Sara Maria Hasbun, Chantal Sowemi-
mo-Coker, and Robert M. Krauss. 2013. Influence of role switching on phonetic
convergence in conversation. Discourse Processes 50 (4): 276–300.
Parker, A. Smythe. 1883. Folk-etymology: A dictionary of verbal corruptions or words perverted
in form or meaning, by false derivation or mistaken analogy. Repr. 1969, New York:
Greenwood Publishers.
 References 533

Parpola, Asko. 2008. Proto-Indo-European speakers of the Late Tripolye culture as the inventors
of wheeled vehicles: Linguistic and archaeological considerations of the PIE homeland
problem. Proceedings of the 19th Indo-European Conference, UCLA, ed. by Karlene
Bley-Jones et al., 1–59. Washington, DC: Institute for the Study of Man.
Parpola, Asko. 2010. The Indus Script and the wild ass. The Hindu, 23 June 2010.
http://www.thehindu.com/todays-paper/tp-opinion/the-indus-scriptand-the-wild-ass/
article481377.ece.
Partridge, Eric. 1950a. Name into word: Proper names that have become common property: A
discursive dictionary, 2nd edn. London: Secker & Warburg.
Partridge, Eric. 1950b. A dictionary of the underworld. London: Routledge & Kegan Paul.
Partridge, Eric. 1961. Adventuring among words. Oxford: Oxford University Press.
Partridge, Eric. 1967. Dictionary of slang and unconventional English, 6th edn. New York:
Macmillan.
Paul, Hermann. 1920. Prinzipien der Sprachgeschichte, 5th edn. Halle: Niemeyer.
(Engl. translation of 2nd edn.: Principles of language history, 1889, New York: Macmillan.)
Pedersen, Holger. 1959. The discovery of language. (Transl. by J. W. Spargo from the 1931
Danish original.) Repr. 1959, Bloomington: Indiana University Press.
Pereltsvaig, Asya. 2017. Languages of the world: An introduction. Cambridge: Cambridge
University Press.
Pereltsvaig, Asya, and Martin W. Lewis. 2015. The Indo-European controversy: Facts and
fallacies in historical linguistics. Cambridge: Cambridge University Press.
Perlmutter, David M. 1986. No nearer to the soul. Natural Language and Linguistic Theory 4:
515–523.
Peter, Steven Joseph. 1991. Barking up the wrong family tree? Greenberg’s method of mass
comparison and the genetic classification of languages. Urbana: University of Illinois BA
honors thesis.
Phillips, Betty. 2006. Word frequency and lexical diffusion. New York: Palgrave MacMillan.
Pinault, Georges-Jean. 2003. Une nouvelle connexion entre le substrat indo-iranien et le
tocharien commun. Historische Sprachforschung 116: 175–189.
Pinault, Georges-Jean. 2006. Further links between the Indo-Iranian substratum and the BMAC
language. Themes and tasks in Old and Middle Indo-Aryan linguistics, ed. by Heinrich
Hettrich and Bertil Tikkanen, 167–196. Delhi: Motilal Banarsidass.
Pinault, Georges-Jean. 2008. Chrestomathie tokharienne: Textes et grammaire. Leuven/Paris:
Peeters.
Pinker, Steven. 1994. The language instinct: How the mind creates language. New York: William
Morrow and Co.
Pinnow, Heinz-Jürgen. 1959. Versuch einer historischen Lautlehre der Kharia-Sprache.
Wiesbaden: Harrassowitz.
Pisani, Vittore. 1937. Toch. A käntu und das idg. Wort für ‘Zunge’. Zeitschrift für vergleichende
Sprachwissenschaft 64: 100–103.
Polomé, Edgar C. (ed.) 1982. The Indo-Europeans in the fourth and third millennia. Ann Arbor,
MI: Karoma.
Polomé, Edgar C. (ed.) 1990a. Research guide on language change. Berlin/New York: Mouton
de Gruyter.
Polomé, Edgar C. 1990b. Linguistic paleontology: Migration theory, prehistory, and archeology
correlated with linguistic data. In Polomé 1990a: 137–159.
534 References

Poppe, Nikolaus. 1960. Vergleichende Grammatik der altaischen Sprachen. Wiesbaden:

Harrassowitz.
Posner, Rebecca. 1966. The Romance languages: A linguistic introduction. Anchor Books. Repr.
1970, Gloucester, MA: Peter Smith.
Pott, August Friedrich. 1833–1836. Etymologische Forschungen auf dem Gebiete der indoger-
manischen Sprachen, 2 vols. Lemgo: Meyer.
Pott, August Friedrich. 1856. Die Ungleichheit menschlicher Rassen: Hauptsächlich vom sprach-
wissenschaftlichen Standpunkte, unter besonderer Berücksichtigung von des Grafen von
Gobineau gleichnamigem Werke. Lemgo & Detmold: Meyer’sche Hofbuchhandlung.
Puckett, Newbell N. 1974. Black names in America: Origins and usage. Boston: G. K. Hall.
Pulgram, Ernst. 1959. Proto-Indo-European reality and reconstruction. Language 35:
421–426.
Pyles, Thomas, and John Algeo. 1982. The origins and development of the English language,
3rd edn. New York: Harcourt Brace Jovanovich.
Ramat, Anna Giacolone, and Paolo Ramat (eds.) 1997. The Indo-European languages. London:
Routledge.
Ramstedt, Gustaf John. 1957–1966. Einführung in die altaische Sprachwissenschaft, 3 vols.
(Suomalais-Ugrilaisen Seuran Toimituksia, 104: 1–3) Helsinki: Suomalais-Ugrilainen
Seura.
Rask, Rasmus. 1818. Undersögelse om det gamle Norske eller Islandske Sprogs Oprindelse.
København: Gyldendal.
Rédei, Károly (ed.) 1986–1988. Uralisches etymologisches Wörterbuch. Wiesbaden:
Harrassowitz.
Rehg, Kenneth L., and Lyle Campbell (eds.) 2018. The Oxford handbook of endangered
languages. Oxford: Oxford University Press.
Renfrew, Colin. 1973. Before civilization: The radiocarbon revolution and prehistoric Europe.
New York: Knopf.
Renfrew, Colin. 1987. Archaeology and language: The puzzle of Indo-European origins.
Cambridge: Cambridge University Press.
Renfrew, Colin. 1989. The origins of the Indo-European languages. Scientific American, October
1989: 106–114.
Renfrew, Colin. 2004. Time depth, convergence theory, and innovation in Proto-Indo-European.
Languages in prehistoric Europe, ed. by Alfred Bammesberger and Theo Vennemann.
Heidelberg: Winter.
Rickford, John. 1999. African American Vernacular English. Oxford: Blackwell.
Ringe, Jr., Donald A. 1990. Evidence for the position of Tocharian in the Indo-European family?
Die Sprache 34.59–123.
Ringe, Jr., Donald A. 1992. On calculating the factor of chance in language comparison.
(Transactions of the American Philosophical Society, 82: 1.). Philadelphia: American
Philosophical Society.
Ringe, Jr., Donald A. 1995. “Nostratic” and the factor of chance. Diachronica 12: 1: 55–74.
Ringe, Don. 2009. Inheritance vs. borrowing of reconstructed vocabulary. Language Log
January 10, 2009. http://languagelog.ldc.upenn.edu/nll/?p=994.
Ringe, Don, and Joseph Eska. 2013. Historical linguistics: Toward a twenty-first century reinte-
gration. Cambridge: Cambridge University Press.
Ringe, Don, Tandy Warnow, and Ann Taylor. 2002. Indo-European and computational cladistics.
Transactions of the Philological Society 100: 1. 59–129.
 References 535

Robbeets, Martine. 2013. Transeurasian: A linguistic continuum between Japan and Europe.
From contact linguistics to Eurolinguistics: A linguistic odyssey across Europe and beyond,
ed. by Sture Ureland, 151–166. Berlin: Logos Verlag.
Robbeets, Martine (ed.) 2017. Language dispersal beyond farming. Amsterdam/Philadelphia:
Benjamins.
Roberts, Sarah Julianne. 2000. Nativization and the genesis of Hawaiian Creole. Language
change and language contact in pidgins and creoles, ed. by John McWhorter, 257–300.
Amsterdam/Philadelphia: Benjamins.
Robins, R. H., and Eugenius Uhlenbeck. 1991. Endangered languages. Oxford: Berg.
Robinson, Orrin W. 1992. Old English and its closest relatives: A survey of the earliest Germanic
languages. Stanford: Stanford University Press.
Ross, Philip E. 1991. Hard words: How deeply can language be traced? Radical linguists
look back to the Stone Age, traditionalists disagree. Scientific American, April 1991:
138–147.
Rossel, Stine, Fiona Marshall, Joris Peters, Tom Pilgram, Matthew D. Adams, and David
O’Connor. 2008. Domestication of the donkey: Timing, processes, and indicators.
Proceedings of the National Academy of Sciences 105 (10): 3715–3720.
Rost-Roth, Martina. 1995. Language in intercultural communication. The German language
and the real world: Sociolinguistic, cultural, and pragmatic perspectives on contemporary
German, ed. by Patrick Stevenson. Oxford: Oxford University Press.
Ruhlen, Merritt. 1994. On the origin of languages: Studies in linguistic typology. Stanford:
Stanford University Press.
Sajnovics, Joannis. 1770. Demonstratio idioma ungarorum et lapponum idem esse. Tyrnavia.
(Repr. 1968, Indiana University Publications, Uralic and Altaic Series, 91.)
Salmons, Joe. 1992. A look at the data for a global etymology: *tik ‘finger’. Explanation
in historical linguistics, ed. by G. W. Davis and G. K. Iverson, 207–228. Amsterdam/
Philadelphia: Benjamins.
Salmons, Joseph C. 1993. The glottalic theory: Survey and synthesis. (Journal of Indo-European
Studies, Monograph 10.) McLean, VA: Institute for the Study of Man.
Salmons, Joseph C., and Brian D. Joseph (eds.) 1998. Nostratic: Sifting the evidence.
Amsterdam/Philadelphia: Benjamins.
Salomon, Richard. 1998. Indian epigraphy: A guide to the study of inscriptions in the
Indo-Aryan languages. New York: Oxford University Press.
Sampson, Geoffrey. 1985. Writing systems: A linguistic introduction. London: Hutchinson & Co.
Samuels, Michael Louis. 1972. Linguistic evolution, with special reference to English.
Cambridge: Cambridge University Press.
Sandfeld, Kristian. 1930. Linguistique balkanique: Problèmes et résultats. Paris: Librairie
Ancienne Honoré Champion.
Sapir, Edward. 1921. Language. New York: Harcourt.
Sapir, Edward. 1931. The concept of phonetic law as tested in primitive languages by Leonard
Bloomfield. Methods in social science: A case book, ed. by S. A. Rice, 197–306. Chicago:
University of Chicago Press.
de Saussure, Ferdinand. 1879. Mémoire sur le système primitif des voyelles dans les langues
indo-européennes. Repr. 1968, Hildesheim: Olms. (Excerpts in Engl. transl. in Lehmann
1967.)
de Saussure, Ferdinand. 1916. Cours de linguistique générale, ed. by Charles Bally, Albert
Sechehaye, and Albert Reidlinger. Paris: Payot.
536 References

Schaller, Helmut W. 1975. Die Balkansprachen: Eine Einführung in die Balkanphilologie.

Heidelberg: Winter.
Scherer, Anton. 1956. Hauptprobleme der idg. Altertumskunde. Kratylos 1: 3–21.
Schetelich, Maria. 1991. Die “schwarzen” Feinde der Arya im Ṛgveda. Altorientalische
Forschungen 18 (1): 151–162.
Schlegel, Friedrich von. 1808. Ueber die Sprache und Weisheit der Indier: Ein Beitrag zur
Begründung der Alterthumskunde. Heidelberg: Mohr und Zimmer.
Schmandt-Besserat, Denise. 1992. Before writing, vol. 1: From counting to cuneiform. Austin:
University of Texas Press.
Schmidt, Annette. 1985. Young people’s Dyirbal: An example of language death from Australia.
Cambridge: Cambridge University Press.
Schmidt, Karl Horst. 1992. Celtic movements in the first millennium B.C. Journal of
Indo-European Studies 20: 145–178.
Schmitt, Rüdiger (ed.) 1977. Etymologie. Darmstadt: Wissenschaftliche Buchgesellschaft.
Schmitt, Rüdiger (ed.) 1989a. Compendium linguarum iranicarum. Wiesbaden: Reichert.
Schmitt, Rüdiger. 1989b. Altpersisch. In Schmitt 1989a.
Schneider, Edgar W. 1989. American Earlier Black English: Morphological and syntactic
variables. Tuscaloosa/London: The University of Alabama Press.
Schneider, Edgar W. 2003. The dynamics of New Englishes: From identity construction to dialect
birth. Language 79: 233–281.
Schrader, Otto. 1886. Linguistisch-historische Forschungen zur Handelsgeschichte und
Warenkunde, vol. 1. Jena.
Schrader, Otto. 1890. Prehistoric antiquities of the Aryan peoples: A manual of comparative
philology and the earliest cultures. (Transl. of first edition of Schrader 1906–1907.)
London: Griffin.
Schrader, Otto. 1906–1907. Sprachvergleichung und Urgeschichte: Linguistisch-historische
Beiträge zur Erforschung des indogermanischen Altertums, 3rd edn, 2 vols. Jena:
Kostenoble.
Schuchardt, Hugo. 1883–1888. Kreolische Studien, 8 vols. Wien: Gerold.
Schuchardt, Hugo. 1885. Über die Lautgesetze: Gegen die Junggrammatiker. Berlin:
Oppenheim.
Schuchardt, Hugo. 1978. The ethnography of variation: Selected writings on pidgins and
creoles, transl. and selected by T. L. Markey. Ann Arbor, MI: Karoma.
Schuchardt, Hugo. 1980. Pidgin and creole languages, selected and transl. by G. G. Gilbert.
Cambridge: Cambridge University Press.
Selinker, Larry. 1972. Interlanguage. International Review of Applied Linguistics 10: 209–231.
Selinker, Larry. 1992. Rediscovering interlanguage. London: Longman.
Senn, Alfred. 1942. The Lithuanian language: A characterization. Chicago: Lithuanian Cultural
Institute.
Senner, Wayne M. (ed.) 1989. The origins of writing. Lincoln: University of Nebraska Press.
Shepard, C. A., H. Giles, and B. A. Le Poire. 2001. Communication accommodation theory. The
new handbook of language and social psychology, ed. by W. P. Robinson and H. Giles,
33–56. New York: Wiley.
Shevoroshkin, Vitaly. 1990. The mother tongue: How linguists have reconstructed the ancestor
of all living languages. The Sciences, May/June 1990: 20–27.
Shnirelman, Victor A. 2007. Archaeology, Russian nationalism, and the “Arctic Homeland”.
Selective remembrances: Archaeology in the construction, commemoration, and
 References 537

consecration of national pasts, ed. by Philip L. Kohl, Mara Kozelsky, and Ben-Yehuda
Nachman, 31–70. Chicago/London: University of Chicago Press.
Sihler, Andrew L. 1995. New comparative grammar of Greek and Latin. New York/Oxford: Oxford
University Press.
Sihler, Andrew L. 2000. Language history: An introduction. Amsterdam/Philadelphia:
Benjamins.
Singler, John Victor. 1988. The homogeneity of the substrate as a factor in pidgin/creole
genesis. Language 64: 27–51.
Skomal, Susan N., and Edgar C. Polomé. 1987. Proto-Indo-European: The archeology of a
linguistic problem. Washington, DC: Institute for the Study of Man.
Solta, Georg Renatus. 1980. Einführung in die Balkanlinguistik mit besonderer Berück-
sichtigung des Substrats und des Balkanlateinischen. Darmstadt: Wissenschaftliche
Buchgesellschaft.
Southern, Mark. 2005. Contagious couplings: Yiddish shm- and the contact-driven
transmission of expressives. Westport, CT: Praeger.
Southworth, Franklin C. 1990. Contact and interference. In Polomé 1990a: 281–294.
Spengler, Robert, Michael Frachetti, Paula Doumani, Lynne Rouse, Barbara Cerasetti, Elissa
Bullion, and Alexei Mar’yashev. 2014. Early agriculture and crop transmission among
Bronze Age mobile pastoralists of Central Eurasia. Proceedings of the Royal Society, B 281:
20133382. http://dx.doi.org/10.1098/rspb.2013.3382
Spielvogel, Jackson J, and David Redles. 1986. Hitler’s racial ideology: Content and occult
sources. Simon Wiesenthal Center Annual 3: 227–246.
Sridhar, Kamal, and S. N. Sridhar. 1986. Bridging the gap: Second language acquisition and
indigenized varieties of English. World Englishes 5: 1. 3–14.
Stearns, MacDonald. 1978. Crimean Gothic: Analysis and etymology of the corpus. Berkeley:
Anma Libri.
Steever, Sanford B. (ed.) 1998. The Dravidian languages. London/New York: Routledge.
Steinmetz, Sol. 2008. Semantic antics: How and why words change meaning. New York:
Random House.
Stern, Gustaf. 1931. Meaning and change of meaning. Bloomington: Indiana University
Press.
Sternemann, Reinhard, and Karl Gutschmidt. 1989. Einführung in die vergleichende Sprachwis-
senschaft. Berlin: Akademie-Verlag.
Stevens, Christopher M. 1992. The use and abuse of typology in comparative linguistics: An
update of the controversy. Journal of Indo-European Studies 20: 45–58.
Stewart, George R. 1979. American given names: Their origin and history in the context of the
English language. New York: Oxford University Press.
Stewart, John M. 1989. Kwa. The Niger-Congo languages, ed. by J. Bendor-Samuel, 217–245.
Lanham/New York/London: University Press of America.
Stockwell, Robert P., and C. Westbrook Barritt. 1961. Scribal practice: Some assumptions.
Language 37: 75–82.
Stokoe, William C., Jr. 1974. Classification and description of sign languages. Current trends
in linguistics, vol. 12: Linguistics and adjacent arts and sciences, ed. by T. A. Sebeok,
345–371. The Hague: Mouton.
Stuart, David, and Stephen D. Houston. 1989. Maya writing. Scientific American, August 1989:
82–89.
Sturtevant, Edgar H. 1917. Linguistic change. Chicago: University Press.
538 References

Sturtevant, Edgar H. 1940. The pronunciation of Greek and Latin. Baltimore: Linguistic Society
of America. (2nd edn. 1967.)
Supalla, Ted, Fanny Limousin, and Betsy Hicks McDonald. 2019. Historical change in American
sign language. In Janda et al., Chapter 20.
Suttie, J. M., and S. G. Reynolds (eds.) 2003. Transhumant grazing systems in temperate Asia.
Rome: Food and Agriculture Organization of the United Nations.
Szemerényi, Oswald. 1957. The problem of Balto-Slav unity: A critical survey. Kratylos 2:
97–123.
Szemerényi, Oswald. 1967. The new look of Indo-European: Reconstruction and typology.
Phonetica 17: 65–99.
Szemerényi, Oswald. 1989a. Einführung in die vergleichende Sprachwissenschaft, 3rd edn.
Darmstadt: Wissenschaftliche Buchgesellschaft.
Szemerényi, Oswald. 1989b. The new sound of Indo-European. Diachronica 6: 237–269.
Talageri, S. G. 2008. The Rigveda and the Avesta: The final evidence. New Delhi: Aditya
Prakashan.
Tamariz, Monica. 2019. A comparative evolutionary approach to the origins and evolution of
cognition and of language. In Janda et al., Chapter 23.
Taylor, Timothy. 1992. The Gundestrup cauldron. Scientific American, March 1992:
84–89.
Thieme, Paul. 1954. Die Heimat der indogermanischen Gemeinsprache. (Akademie der Wissen-
schaften und der Literatur, Mainz.) Wiesbaden: Steiner.
Thomason, Sarah Grey. 1983. Chinook Jargon in areal and historical context. Language 59:
820–870.
Thomason, Sarah Grey. 2001. Language contact: An introduction. Washington, DC: Georgetown
University Press.
Thomason, Sarah Grey. 2015. Endangered languages: An introduction. Cambridge: Cambridge
University Press.
Thomason, Sarah Grey, and Terrence Kaufman. 1988. Language contact, creolization, and
genetic linguistics. Berkeley/Los Angeles: University of California Press.
Thumb, Albert. 1901. Die griechische Sprache im Zeitalter des Hellenismus: Beiträge zur
Geschichte und Beurteilung der Koinē. Straßburg: Trübner.
Thurgood, Graham, and Randy J. LaPolla (eds.) 2003. The Sino-Tibetan languages. New York:
Routledge.
Tovar, Antonio. 1957. The Basque language. Transl. by Herbert Pierrepont Houghton.
Philadelphia: University of Pennsylvania Press.
Trager, George L. 1974. Writing and writing systems. Current trends in linguistics, vol. 12:
Linguistics and adjacent arts and sciences, ed. by Thomas A. Sebeok, 373–496. The
Hague: Mouton.
Trask, R[obert] L[awrence]. 2000. Dictionary of historical and comparative linguistics. London:
Routledge.
Trask, Robert Lawrence. 1994. Language change. London and New York: Routledge.
Trask, R[obert] L[awrence]. 1996. Historical linguistics. London: Routledge.
Trask, Robert Lawrence. 1997. The history of Basque. London: Routledge.
Traugott, Elizabeth Closs, and Bernd Heine (eds.) 1991. Approaches to grammaticalization,
2 vols. Amsterdam/Philadelphia: Benjamins.
Traugott, Elizabeth, and Richard Dasher. 2001. Regularity in semantic change. Cambridge:
Cambridge University Press.
 References 539

Trier, Jost. 1931. Der deutsche Wortschatz im Sinnbezirk des Verstandes: Die Geschichte eines
sprachlichen Feldes, vol. l: Von den Anfängen bis zum Beginn des 13. Jahrhunderts.
Heidelberg: Winter.
Trier, Jost. 1973. Aufsätze und Vorträge zur Wortfeldtheorie, ed. by A. van der Lee and O.
Reichmann. The Hague: Mouton.
Trudgill, Peter. 1983. On dialect: Social and geographical perspectives. New York: New York
University Press.
Trudgill, Peter. 1986. Dialects in contact. Oxford: Blackwell.
Trudgill, Peter. 1990a. The dialects of England. Oxford: Blackwell.
Trudgill, Peter. 1990b. Dialect geography. In Polomé 1990a: 257–271.
Trudgill, Peter. 1994. Dialects. London and New York: Routledge.
Trudgill, Peter. 2004. New-dialect formation: The inevitability of Colonial Englishes. Oxford:
Oxford University Press.
Trudgill, Peter, and Jean Hannah. 2002. International English: A guide to the varieties of
Standard English, 4th edn. London: Arnold.
Tsunoda, Tasaku. 2005. Language endangerment and language revitalization. Berlin/New York:
Mouton de Gruyter.
Turner, Lorenzo Dow. 1949. Africanisms in the Gullah dialect. Chicago: University of Chicago
Press.
Tyler, Stephen A. 1968. Dravidian and Uralic: The lexical evidence. Language 44: 798–812.
Ullmann, Stephen. 1957. The principles of semantics, 2nd edn. Glasgow: University Publications.
Ullmann, Stephen. 1962. Semantics: An introduction to the science of meaning. Oxford:
Blackwell.
Unger, J. Marshall. 1990. Japanese and what other Altaic languages? In Baldi 1990: 547–561.
Unger, J. Marshall. 2004. Ideogram: Chinese characters and the myth of disembodied meaning.
Honolulu: University of Hawai’i Press.
Ureland, P. Sture. 1990. Contact linguistics: Research on linguistic areas, strata, and
interference in Europe. In Polomé 1990a: 471–506.
Vennemann, Theo. 1984. Hochgermanisch und Niedergermanisch: Die Verzweigungstheorie
der germanisch-deutschen Lautverschiebungen. Beiträge zur Geschichte der deutschen
Sprache und Literatur 106: 1–45.
Vennemann, Theo (ed.) 1989. The new sound of Indo-European: Essays in phonological
reconstruction. Berlin/New York: Mouton de Gruyter.
Verner, Karl. 1877. Eine Ausnahme der ersten Lautverschiebung. Zeitschrift für vergleichende
Sprachforschung 23: 97–130. (Engl. transl. in Lehmann 1967.)
Vihman, Marilyn May. 1980. Sound change and child language. Papers from the 4th
International Conference on Historical Linguistics, ed. by E. C. Traugott et al., 303–320.
Amsterdam/Philadelphia: Benjamins.
Vilà, Charles, Jennifer A. Leonard, and Albano Bejo Perreira. 2006. Genetic documentation
of Horse and Donkey domestication. Documenting domestication: New genetic and
archaeological paradigms, ed. by Melinda A. Zeder. Berkeley & Los Angeles: University of
California Press.
Vogt, Hans. 1958. Les occlusives de l’arménien. Norsk Tidskrift for Sprogvidenskap 18: 143–161.
von Raffler-Engel, Walburga, Jan Wind, and Abraham Jonker (eds.) 1991. Studies in language
origins, vol. 2. Amsterdam/Philadelphia: Benjamins.
Wackernagel, Jacob, and Albert Debrunner. 1942. Indo-Iranica. Zeitschrift für vergleichende
Sprachforschung 67: 154–182.
540 References

Walker, C. B. F. 1987. Reading the past: Cuneiform. London: British Museum.

Wallace, Rex E. 2007. The Sabellic languages of ancient Italy. Munich: Lincom
Wallace, Rex E. 2008. Zikh Rasna: A manual of the Etruscan language and inscriptions. Ann
Arbor: Beech Stave Press.
Wallman, Joel. 1992. Aping language. New York: Cambridge University Press.
Warmuth, Vera, Anders Eriksson, Mim Ann Bower, Graeme Barker, Elizabeth Barrett, Bryan
Kent Hanks, Shuicheng Li, David Lomitashvili, Maria Ochir-Goryaeva, Grigory V. Sizonov,
Vasiliy Soyonov, and Andrea Manica. 2012. Reconstructing the origin and spread of horse
domestication in the Eurasian steppe. Proceedings of the National Academy of Sciences of
the United States 109, no. 21. http://www.pnas.org/content/109/21/8202.full.pdf+html.
Watkins, Calvert. 1969. Indogermanische Grammatik, 3: 1: Formenlehre: Geschichte der
indogermanischen Verbalflexion. Heidelberg: Winter.
Watkins, Calvert. 1982. Aspects of Indo-European poetics. In Polomé 1982: 104–120.
Watkins, Calvert (ed.) 1985. The American heritage dictionary of Indo-European roots. Boston:
Houghton Mifflin Co.
Watkins, Calvert. 1987. How to kill a dragon in Indo-European. Studies in Memory of Warren
Cowgill (1929–1985): Papers of the Fourth East Coast Indo-European Conference, ed. by
C. Watkins, 270–299. Berlin: de Gruyter.
Watkins, Calvert. 1989. New parameters in historical linguistics, philology, and culture history.
Language 65: 783–799.
Watkins, Calvert. 1995. How to kill a dragon: Aspects of Indo-European poetics. New York and
Oxford: Oxford University Press.
Weinreich, Uriel, William Labov, and Marvin I. Herzog. 1968. Empirical foundations for a
theory of language change. Directions for historical linguistics, ed. by W. P. Lehmann and
Y. Malkiel, 95–195. Austin: University of Texas Press.
Weinreich, Uriel. 1968. Languages in contact: Findings and problems. The Hague:
Mouton.
Weiss, Michael. 2018. Barley in Anatolian, Tocharian, and elsewhere: A fine-grained analysis.
228th Meeting of the American Oriental Society, Pittsburgh, PA.
Wellard, James. 1973. The Search for the Etruscans. New York: Saturday Review Press.
Wendt, Heinz F. 1961. Sprachen. Frankfurt: Fischer Bücherei.
Whitney, William D. 1877. The principle of economy as a phonetic force. Transactions of the
American Philological Association 8: 121–134.
Wiik, Kalevi. 2002. Eurooppalaisten juuret [‘The roots of the Europeans’]. Jyväskylä:
Atena.
Wind, Jan, Abraham Jonker, Robin Allott, and Leonard Rolfe (eds.) 1994. Studies in language
origins, vol. 3. Amsterdam/Philadelphia: Benjamins.
Wind, Jan, Brunetto Chiarelli, Bernard Bichakjian, Alberto Nocentini, and Abraham Jonker (eds.)
1992. Language origins: A multidisciplinary approach. (NATO ASI Series D: Behavioural
and Social Sciences, 61.) Dordrecht: Kluwer Academic Publishers.
Wind, Jan, Edward G. Pulleyblank, Eric de Grolier, and Bernard H. Bichakjian (eds.) 1989.
Studies in language origins, vol. 1. Amsterdam/Philadelphia: Benjamins.
Winford, Donald. 1997. On the origins of African American Vernacular English: A creolist
perspective. Part 1: Sociohistorical background. Diachronica 14: 2. 305–344.
Winford, Donald. 1998. On the origins of African American Vernacular English: A creolist
perspective. Part 2: Linguistic features. Diachronica 15: 1: 99–154.
Winford, Donald. 2003. An introduction to contact linguistics. Oxford: Blackwell.
 Conclusions and outlook 541

Winter, Werner. 1966. Traces of early dialectal diversity in Old Armenian. Ancient Indo-European
dialects, ed. by H. Birnbaum and J. Puhvel, 201–211. Berkeley and Los Angeles: University
of California Press.
Winter, Werner. 1970. Basic principles of the comparative method. Method and theory in
linguistics, ed. by P. J. Garvin, 147–156. The Hague: Mouton.
Winter, Werner. 1982. IE for ‘tongue’ and ‘fish’. Journal of Indo-European Studies 10: 167–186.
Winter, Werner. 1989. Thoughts about markedness and normalcy/naturalness. Markedness
in synchrony and diachrony, ed. by O. M. Tomić, 103–109. Berlin/New York: Mouton de
Gruyter.
Winter, Werner. 1990. Linguistic reconstruction: The scope of historical and comparative
linguistics. In Polomé 1990a: 11–21.
Winter, Werner. 1992. Some thoughts about Indo-European numerals. In Gvozdanović 1992:
11–28.
Winter, Werner. 1997a. A loan word and its implications. Language and its ecology: Essays in
memory of Einar Haugen, ed. by Stig Eliasson and Ernst Håkon Jahr, 435–440. Berlin/New
York: Mouton de Gruyter.
Winter, Werner. 1997b. Lexical archaisms in the Tocharian languages. Historical,
Indo-European, and lexicographical studies: A festschrift for Ladislav Zgusta on the
occasion of his 70th birthday, ed. by H. H. Hock, 183–193. Berlin: Mouton de Gruyter.
Winter, Werner, and Edgar C. Polomé (eds.) 1992. Reconstructing languages and cultures.
Berlin/New York: Mouton de Gruyter.
Withycombe, Elizabeth G. 1977. The Oxford dictionary of English Christian names, 3rd edn. New
York: Clarendon Press.
Wolfram, Walt, and Natalie Schilling-Estes. 1998. American English: Dialects and variation.
Oxford: Blackwell.
Wolfram, Walt, and Natalie Schilling-Estes. 2003. Dialectology and linguistic diffusion. In
Joseph and Janda 2003: 713–735.
Wolfram, Walt, and Erik Thomas. 2002. The development of African American English. Oxford:
Blackwell.
Woods, Richard D. 1984. Hispanic first names: A comprehensive dictionary of 250 years of
Mexican-American usage. Westport, CT: Greenwood Press.
Zgusta, Ladislav. 1990. Onomasiological change: Sachen-change reflected by Wörter. In Polomé
1990a: 389–398.
Zimmer, Heinrich. 1879. Altindisches Leben. Berlin: Weidmann.
Zipf, George Kingsley. 1929. Relative frequency as a determinant of phonetic change. Harvard
Studies in Classical Philology 40: 1–95.
Language index
Abkhaz 419 Babylonian 71, 73, 86, 421
Afrikaans 43 Baltic 39, 43, 44, 45, 46, 47, 49, 54, 313, 314,
Afro-Asiatic 218, 416, 417, 421, 427, 428, 399, 470, 490
438, 439, 440, 469, 506, 507 Balto-Slavic 46, 55, 490
Ahlõ 421 Bamum 103
Ainu 435 Bantu 107, 122, 238, 338, 341, 416, 421, 426,
Akkadian 358, 463, 469 441, 492, 506
Akwa’ala 439 Basque 89, 296, 297, 336, 337, 343, 399,
Albanian 47, 55, 97, 198, 350, 352, 353, 354, 400, 401, 419, 425, 427, 436, 468, 505
355, 364, 399, 416, 469, 490 BCMS, see Bosnian-Croatian-Montenegrin-
Algonquian 67, 102, 232, 334, 412, 413, 423, Serbian
426, 436, 500, 506 Belorussian 45
Almosan 439 Bengali 30, 54, 344
Altaic 43, 218, 358, 417, 418, 419, 426, 427, Berber 218, 421
435, 441, 442, 505 Blackfoot 423
American Sign Language 119, 121, 156, 238, Bokmaal 294, 296, 325, 388, 472, 499; see
425, 492 also Norwegian
Amerind 428, 437, 438, 439 Bosnian 45, 55, 350; see also Bosnian-
Anatolian 48, 49, 449, 450, 451, 459, 460, Croatian-­Montenegrin-Serbian, BCMS,
462, 464, 465, 466, 470, 471, 485, 490, Serbo-Croatian
509, 510, 511 Bosnian-Croatian-Montenegrin-Serbian
Andean 424, 439 (BCMS) 55, 350–354, 364; see also
Angolares 385 Serbo-Croatian
Apache 271, 422, 426 Brahmi 95, 98, 99, 100, 101, 102, 492
Arabic 36, 53, 70, 73, 76, 85, 198, 238, 253, Brahui 359, 420
276, 300, 301, 334, 380, 421, 427, 436, Breton 38, 296, 394, 398, 399
439, 440 British English 12, 13, 15, 16, 20, 108, 119,
Aramaic 224 125, 153, 159, 162, 163, 165, 175, 185,
Arawakan 424 197, 203, 225, 248, 286, 288, 296, 299,
Armenian 49, 50, 52, 54, 55, 107, 108, 115, 300, 320, 328, 329, 330, 425, 472, 500
337, 400, 401, 402, 406, 413, 414, 415, British Sign Language 425
419, 469, 490, 494, 505, 510 Brythonic 38
Assyrian 71, 73, 86, 421 Bulgarian 45, 55, 350, 352, 354, 399, 494
Athabaskan 102, 422 Burmese 419, 420
Australian languages 425, 430, 506 Burushaski 356, 362, 419, 437
Austro-Asiatic 53, 356, 361, 420, 506 Bushman 422
Austronesian 420, 441, 473, 506
Avestan 51, 52, 55, 85, 86, 214, 217, 218, 313, Caribbean pidgins 386
314, 451, 455, 463; see also Younger Catalan 39, 289, 296, 297, 343
Avestan Caucasic 49, 415, 419, 427, 469, 470, 505
Aymara 424, 439 Cayuga 423
Azerbaijani 419 Celtic (also “Celtick”) 29, 36, 37, 38, 41, 43,
Azeri 419 55, 83, 228, 242, 243, 276, 278, 313,
336, 337, 343, 396, 398, 399, 406, 450,
462, 467, 490

https://doi.org/10.1515/9783110613285–021
544 Language index

Chadic 421, 506 Efik 421

Chechen-Ingush 419 Egyptian 42, 48, 50, 62, 63, 64, 70, 73, 77,
Cherokee 73, 102, 423 85, 87, 88, 90, 218, 274, 302, 440, 452,
Cheyenne 423 464, 465
Chibchan 439 Elamite 88, 358, 435, 507
Chinese 11, 14, 29, 54, 63, 67, 70, 95, 96, 97, English 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13,
99, 150, 223, 244, 245, 247, 250, 251, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24,
252, 274, 325, 341, 419, 420, 425, 467, 25, 26, 27, 33, 34, 36, 38, 41, 42, 43, 46,
492, 497, 506 53, 55, 57, 58, 59, 62, 65, 67, 70, 73, 76,
Chinese Sign Language 425 79, 80, 84, 87, 90, 91, 92, 93, 94, 97,
Chinook 373, 383, 439, 503 102, 104, 105, 106, 108, 109, 110, 111,
Chipewyan 107 112, 114, 115, 117, 118, 119, 120, 121, 123,
Chippewa 423 125, 126, 127, 128, 129, 130, 131, 132,
Chumash 423 134, 138, 139, 140, 141, 142, 143, 144,
Cockney 16, 18 145, 146, 147, 148, 149, 150, 151, 153,
Comanche 271, 424 154, 155, 156, 157, 158, 159, 160, 161,
Coptic 88, 421 162, 163, 164, 165, 166, 167, 168, 169,
Cornish 38, 343, 396 170, 171, 172, 173, 175, 176, 177, 178,
Cree 102, 198, 334, 412, 423 179, 180, 181, 182, 183, 184, 185, 186,
creole 378, 379, 384, 385, 387, 388, 389, 187, 188, 189, 190, 191, 193, 197, 198,
390, 391, 392 199, 201, 202, 203, 204, 205, 206, 207,
Croatian 45, 53, 55, 56, 350, 353 209, 210, 211, 212, 213, 214, 215, 216,
Crow 423 219, 220, 221, 222, 223, 224, 225, 226,
Cuna 439 227, 228, 229, 230, 231, 232, 233, 234,
Cushitic 421, 506 235, 236, 237, 238, 239, 240, 241, 242,
Cypriot 48 243, 244, 245, 247, 248, 249, 250, 251,
Cyril 44, 81 252, 253, 254, 255, 256, 257, 258, 259,
Cyrillic 44, 45, 73, 80, 81, 82 260, 261, 262, 263, 264, 265, 267, 268,
Czech 44, 45, 87, 399, 462 269, 271, 272, 273, 275, 276, 277, 278,
279, 280, 281, 282, 283, 284, 285, 286,
Dagestani 419 287, 288, 290, 291, 292, 293, 294, 295,
Dakotas 334, 423 296, 297, 299, 300, 301, 302, 303, 304,
Dalmatian 39, 343 305, 320, 322, 323, 324, 325, 326, 327,
Danish 13, 33, 42, 43, 85, 104, 111, 228, 243, 328, 329, 330, 331, 332, 333, 334, 335,
246, 249, 278, 288, 289, 293, 294, 399 336, 339, 343, 348, 351, 353, 355, 358,
Dimotiki 302, 388 359, 366, 367, 368, 369, 371, 372, 373,
Dravidian 27, 30, 52, 53, 125, 252, 253, 324, 374, 377, 378, 379, 380, 381, 385, 386,
327, 337, 347, 356, 357, 358, 359, 360, 387, 388, 389, 390, 392, 393, 394, 396,
361, 363, 401, 420, 428, 429, 430, 435, 399, 402, 403, 404, 405, 406, 407, 408,
438, 439, 440, 441, 448, 463, 482, 506, 410, 411, 425, 426, 431, 432, 433, 434,
507, 511 435, 436, 438, 439, 440, 451, 456, 457,
Dutch 13, 43, 160, 186, 214, 246, 272, 274, 463, 464, 468, 469, 472, 478, 482, 483,
280, 281, 289, 292, 311, 312, 313, 330, 484, 487, 488, 489, 492, 493, 494, 495,
337, 363, 378, 399, 406 496, 497, 498, 499, 500, 501, 502, 503,
504, 505
East Slavic 45 Equatorial 439
Eastern Arctic Eskimo 102 Erie 423
 Language index 545

Eskimo 102, 271, 422, 428, 438, 439 238, 241, 246, 247, 248, 250, 254, 255,
Eskimo-Aleut 422, 428, 438, 439 257, 259, 264, 268, 271, 272, 273, 274,
Estonian 46, 398, 399, 400, 401, 417 275, 280, 281, 287, 289, 292, 295, 296,
Etruscan 40, 80, 89, 490 297, 298, 301, 311, 315, 316, 317, 318,
Eyak 422 324, 330, 331, 336, 338, 346, 348, 351,
363, 381, 382, 393, 394, 399, 404, 405,
Faai 439 406, 407, 408, 438, 463, 470, 472, 475,
Faroese 43 476, 477, 480, 487, 488, 489, 497, 500,
Farsi 52 503
Fiji 403 Germanic 13, 14, 21, 26, 33, 36, 37, 40, 41,
Finn(ish) 32, 97, 132, 198, 398, 399, 400, 42, 43, 44, 55, 73, 79, 80, 81, 83, 104,
401, 417, 430, 438 105, 106, 108, 109, 110, 111, 112, 115, 117,
Finno-Ugric 356, 417, 419, 439, 440, 448, 125, 126, 132, 133, 134, 155, 181, 183,
460, 482 209, 210, 216, 224, 228, 233, 234, 254,
Flemish 43, 296, 311, 363 256, 262, 271, 272, 275, 276, 311, 313,
Fox 412, 423, 487, 505 336, 355, 399, 400, 401, 402, 409, 414,
Frankish 40, 43, 45, 209, 213, 224, 271, 316, 415, 449, 450, 451, 453, 456, 457, 461,
317, 318, 337 462, 463, 467, 468, 469, 470, 477, 479,
French 1, 4, 8, 9, 10, 11, 13, 15, 22, 23, 24, 490, 491, 492, 505
30, 31, 34, 38, 39, 41, 42, 45, 46, 51, 55, Goidelic 38
94, 118, 121, 123, 131, 135, 140, 150, 153, Gothic (also ”Gothick”) 29, 36, 42, 43, 44,
155, 157, 160, 162, 167, 175, 179, 198, 45, 50, 55, 104, 105, 106, 109, 355, 398,
200, 203, 205, 207, 208, 209, 211, 213, 463, 467, 490
215, 220, 223, 224, 225, 226, 227, 228, Greek (ancient) 4, 8, 11, 15, 16, 29, 31, 32,
229, 230, 232, 234, 236, 237, 239, 240, 33, 35, 36, 38, 40, 43, 44, 45, 46, 47, 48,
241, 243, 248, 250, 254, 255, 256, 257, 50, 51, 52, 55, 77, 78, 79, 80, 81, 82, 84,
258, 260, 262, 264, 271, 272, 275, 276, 85, 87, 88, 89, 90, 94, 105, 106, 107, 111,
278, 287, 288, 289, 291, 292, 293, 294, 128, 132, 140, 141, 160, 167, 171, 206,
296, 297, 301, 311, 322, 334, 336, 337, 224, 225, 228, 230, 232, 235, 250, 252,
343, 346, 363, 377, 388, 394, 399, 404, 254, 255, 258, 259, 261, 262, 264, 265,
405, 407, 408, 425, 433, 436, 455, 461, 271, 272, 273, 274, 275, 276, 277, 278,
489, 497 293, 298, 299, 301, 302, 313, 322, 323,
French Sign Language 425 337, 339, 340, 350, 352, 353, 354, 355,
Frisian 13, 43, 186, 292, 293, 406 364, 388, 398, 399, 401, 402, 409, 410,
Fula 421 413, 416, 430, 433, 450, 451, 452, 454,
456, 457, 462, 463, 465, 467, 468, 469,
Gaelic 38, 73, 84, 242, 281, 288, 343, 470, 472, 473, 484, 485, 489, 490, 491,
392 494, 497, 499, 510; see also Mycenaean
Gaulish 55 Greek, West Greek
Georgian 49, 266, 278, 419 Greek (modern) 141, 206, 274, 278, 299,
German (also “NHG”) 2, 4, 5, 6, 10, 11, 12, 339, 350, 388, 402; see also Tsakonian
13, 14, 18, 21, 22, 23, 24, 33, 34, 36, 39, (Greek)
43, 44, 45, 46, 51, 55, 85, 86, 87, 94, Guamo 439
104, 105, 113, 119, 127, 132, 134, 135, Gujarati 30, 54
139, 144, 145, 146, 162, 186, 187, 188,
189, 198, 200, 206, 210, 211, 214, 223, Hadza 422
224, 230, 231, 232, 234, 235, 236, 237, Haida 423
546 Language index

Haitian 387, 388 314, 315, 321, 350, 354, 356, 358, 398,
Haitian Creole 388 399, 400, 401, 402, 406, 407, 408, 409,
Halkomelem 439 411, 413, 414, 415, 416, 417, 419, 426,
Hamitic 467 427, 428, 429, 431, 434, 435, 436, 438,
Han’gŭl 97, 98, 99, 100, 103 439, 440, 448, 449, 450, 451, 452, 453,
Hanty 417 454, 455, 456, 457, 458, 459, 460, 461,
Hausa 218, 421 462, 463, 464, 465, 466, 467, 468, 469,
Hawaiian 338, 372, 379, 420, 503 470, 471, 472, 473, 474, 475, 476, 477,
Hebrew 24, 30, 73, 76, 85, 223, 248, 275, 478, 479, 480, 481, 482, 483, 484, 485,
278, 393, 395, 396, 397, 421, 469, 498 489, 490, 492, 496, 498, 505, 507, 509,
Hindi 24, 30, 53, 54, 98, 198, 200, 206, 208, 510, 511, 512
214, 221, 228, 252, 253, 264, 266, 272, Indo-Iranian 50, 54, 217, 272, 313, 314, 410,
281, 324, 327, 329, 332, 333, 334, 344, 413, 449, 458, 459, 460, 463, 464, 465,
356, 396, 400, 401, 402, 404, 411, 430, 466, 481, 490, 510, 511
431, 432, 433, 434, 436, 438, 440 Indonesian 252, 415, 420, 421
Hindi-Urdu 53 Indo-Semitic 427
Hindustani 53, 54, 324 Indo-Uralic 431, 436, 442
Hittite 48, 49, 57, 73, 86, 87, 88, 313, 413, Inuit 274, 422
416, 427, 449, 450, 451, 456, 459, 460, Iowa 423
464, 465, 466, 469, 490 Iranian 32, 40, 44, 49, 50, 51, 52, 54, 55, 63,
Hokan 423, 439, 506 85, 108, 217, 272, 273, 313, 314, 356,
Hopi 424, 426 362, 401, 406, 410, 413, 414, 415, 419,
Hottentot 422 433, 449, 450, 458, 459, 460, 463, 464,
Hung(arian) 32, 125, 277, 301, 398, 399, 400, 465, 466, 467, 480, 481, 490, 510, 511
401, 417, 439, 448, 482, 483, 512 Iranshe 439
Huron 423 Irish 38, 40, 50, 55, 82, 83, 216, 217, 228,
281, 336, 337, 343, 399, 449, 458, 463,
Icelandic 13, 14, 43, 55, 80, 210, 212, 242, 491, 509
245, 246, 247, 249, 250, 251, 252, 263, Iroquoian 102, 423
264, 295, 399, 472, 473, 497 Italian 11, 13, 30, 31, 39, 55, 127, 129, 140,
Illinois 131, 243, 270, 290, 309, 310, 423, 153, 198, 206, 215, 216, 224, 254, 255,
499, 501 256, 275, 277, 280, 281, 289, 296, 301,
Illyrian 47, 339, 354, 469 302, 337, 343, 346, 355, 363, 379, 399,
Indian English 327–329, 331–332, 335, 336, 404, 482
339, 501; see also South Asian English Italic 39, 40, 313, 450, 490
Indic 51, 449
Indo-Aryan 36, 50, 51, 52, 53, 54, 55, 124, Japanese 70, 95, 97, 211, 232, 247, 251, 252,
125, 228, 252, 253, 272, 305, 326, 347, 337, 338, 418, 425, 435, 505
350, 356, 357, 358, 359, 360, 361, 362, Japanese Sign Language 425
363, 401, 406, 430, 448, 449, 454, 465, Javanese 420, 421
473, 476, 481, 490, 511
Indo-European 17, 22, 25, 26, 27, 29, 32, 33, Kaliana 439
34, 35, 36, 37, 38, 40, 43, 46, 47, 48, Kalmük 218
49, 50, 51, 52, 53, 54, 55, 57, 71, 73, 79, Kannada 53, 334, 337, 347, 348, 349, 350,
87, 88, 94, 104, 105, 106, 107, 108, 109, 400, 401, 402, 420
110, 111, 115, 125, 130, 140, 141, 182, 183, Kartvelian 419, 465
205, 206, 213, 216, 217, 218, 273, 313, Kashmiri 30, 356, 400, 401
 Language index 547

Kharia 213 Malay 402–403, 420

Khasi 420 Malayalam 53, 420, 439
Khmer 53, 356, 420 Malayo-Polynesian 338, 403, 416, 420, 506;
Khoi 422 see also Austronesian
Khoisan 422 Manchu 418
Korean 11, 95, 97, 98, 99, 102, 418, 425 Manx 396
Korean Sign Language 425 Maori 420
Kurdish 52 Marathi 30, 54, 98, 347–350, 400–401, 402,
Kurux 439 430, 438, 494, 502
Kushitic 440 Mayan 66, 70, 71, 88, 90–91, 95, 102, 424,
Kutenai 439 491, 506
Kwa languages 415, 505 Medieval French 9
Kwakwala 439 Medieval Greek 224
Medieval Latin 224
Lapp 417 Melanesian Pidgin English 372, 379, 381,
Latin 1, 2, 4, 7, 8, 9, 13, 14, 15, 16, 29, 30, 31, 503; see also Neo-Melanesian and Tok
32, 33, 35, 36, 38, 39, 40, 44, 46, 47, 52, Pisin
53, 79, 80, 85, 87, 102, 105, 106, 107, Mende 103
108, 109, 110, 117, 118, 120, 132, 140, Menomini 412, 423
141, 153, 160, 167, 171, 172, 175, 176, 177, Micmac 271, 422
179, 182, 183, 210, 221, 225, 228, 232, Micronesian 204
233, 235, 246, 250, 252, 254, 255, 256, Middle Armenian 50
258, 259, 260, 262, 264, 271, 275, 276, Middle Chinese 247, 251, 420
282, 291, 294, 301, 302, 305, 306, 307, Middle English (ME) 9, 16, 55, 130, 149–150,
313, 314, 322, 326, 327, 336, 337, 338, 158, 201, 206, 211, 212, 213, 222, 224,
350, 355, 396, 398, 408, 409, 410, 413, 230, 256, 261, 292, 387
416, 433, 439, 450, 451, 456, 462, 463, Middle High German (MHG) 139, 213,
467, 468, 469, 470, 484, 489, 490, 496, 295
497 Middle Indo-Aryan 305, 326; see also
Latvian 45, 46, 398, 399, 440, 464 Prakrit
Lingala 421 Middle Welsh (MW) 336–337, 464
Lislakh 427, 442 Mitanni 54
Lithuanian 45, 46, 56, 314, 399, 451, 457, Mixe 439
460, 461, 463, 464, 467, 468, 469, 470, Mohave/Mojave 423, 439
490, 536 Mohawk 423
Lower Sorbian 44, 343 Mon 53, 356, 420
Luiseño 424 Mongol Mong. 218, 418–419, 435
Luwian 49, 88, 462, 472 Monguor 435
Lycian 49 Montenegrin 45, 55, 350; see also
Lydian 49 Bosnian-Croatian-Montenegrin-Serbian,
BCMS
Macedonian (ancient) 273–274, 469 Munda 53, 213, 271, 356, 361, 420
Macedonian (Slavic language) 45, 273–274, Mycenaean (Greek) 47–48, 90, 465, 485
350, 352, 354
Macro-Carib 439 Na-Dene 423
Macro-Tucan 439 Nahuatl 423–424
Malagasy 420–421 Naron 422
548 Language index

Navajo 422, 426 317, 318 315–318, 338, 461, 463, 464,
Neo-Melanesian 386; see also Melanesian 468, 469, 470, 500
Pidgin English and Tok Pisin Old Icelandic 43, 245, 263, 295; see also Old
Neo-Solomonic 386 Norse
New Guinean/Melanesian pidgin, see Old Irish (OIr.) 38, 40, 205, 216–217, 242,
Melanesian Pidgin English 336, 278, 337, 451, 462, 491, 509
Newfoundland dialect of English 144 Old Italian (OIt.) 224
New York English 114, 226 Old Low Frankish 43
NHG, see German Old Norse 43, 213, 242, 243, 261, 293, 409,
Niger-Congo 421 461, 463; see also Old Icelandic
Niger-Kordofanian 421 Old Persian 29, 44, 51, 73–75, 85–87, 88,
Nilo-Saharan 422 272, 398, 435
Norse 56 Old Polish 462
Northeast Caucasic 419 Old Prussian 45, 46
Northwest Caucasic 419 Old Saxon (OS) 43, 464
Norwegian 13, 43, 56, 288–289, 293–294, Old Spanish 41, 153, 224
325, 383–384, 388, 399, 472; see also Olmec 91
Bokmaal and Nynorsk Omaha 423
Nostratic 427, 435–436, 442, 507 Oneida 423
Nubian 422 Onondaga 423
Nynorsk 293–294, 296, 324–325, 388, 499; Osage 423
see also Norwegian Oscan 40, 205
Oscan-Umbrian 217
Ogham 38, 40, 82–83, 491 Ossetic 50, 52, 108, 400, 401, 402, 415,
Ojibwa 61–62, 412, 423 419
Old Church Slavic (OCS) 33, 44–45, 50, 56, Ostyak 271, 417
205, 216, 350, 408, 463 Ottawa 423
Old Egyptian 66–67, 72, 75, 87–88, 218, 421,
439 Pahlavi 403
Old English (OE) 5, 6, 7, 8, 10, 18, 19, 33, 43, Paiute 426
55, 57, 59, 94, 106, 108, 109, 110, 111, Palaic 49
112, 115, 118, 120, 121, 126, 130, 140, 141, Pali 52; see also Prakrit
142, 143, 145, 146, 147, 149, 150, 153, Pama-Nyungan 424
158, 159, 161, 162, 166, 167, 168, 172, Panjabi 54
176, 182, 183, 184, 185, 186, 204, 207, Papago 424
209, 210, 212, 213, 217, 225, 228, 233, Papiamentu 385
234, 242, 249, 256, 260, 261, 271, 275, Pashto 52, 356
285, 293, 304, 387, 403, 409, 410, 411, Penutian 439
416, 431, 434, 440, 461, 462, 463, 468, Persian 29, 36, 44, 48, 49, 51, 52, 53, 56,
482, 492 73–75, 85–88, 99, 224, 334, 253, 272,
Old Frankish 224 273, 334, 398, 400, 401, 403, 432, 435,
Old French (OFr.) 1, 8, 9, 131, 153, 201, 209, 491
213, 214, 215, 224, 232, 256, 278, 292, Phrygian 55, 469
293, 461 Pictish 38
Old Frisian 43, 464 Pidgin English 380, 386, 503
Old High German (OHG) 43, 105, 119, 139, Polish 31, 44, 45, 56, 324, 399, 462, 469
144–145, 213, 214, 228, 275, 315, 316, Polynesian 204, 338, 403, 416, 420, 506
 Language index 549

Portuguese 13, 24, 31, 39, 42, 56, 238, 289, Romance-based pidgins 374, 380, 382
293, 374, 375, 376, 377, 378, 380, 385, Romani 54, 56, 256, 280, 281, 350, 352, 473,
387, 399, 433, 503 511
Portuguese pidgins and creoles 377, 378, Romanian 13, 30, 39, 56, 350, 351, 352, 354,
380, 385 399, 488
Portuguese Proto-Pidgin 376–378; see also Romantsch 39, 289, 296, 346, 363
Proto-Pidgin Russian 41, 43, 45, 56, 82, 90, 119, 230, 231,
Prakrit 52, 274, 323, 473; see also Middle 275, 343, 354, 383, 384, 399, 417, 450,
Indo-Aryan, Pali 481, 494
pre-Indo-European 451, 452, 468 Russenorsk 383, 503
Proto-Afro-Asiatic 439 Ruthenian 45
Proto-Algonquian 412, 413, 423
Proto-Amerind 439 Saami 417, 439
Proto-Bantu 107, 122 São Tomé Creole 385
Proto-Dravidian 429 Sabir 376–377; see also Portuguese
Proto-Finno-Ugric 439 Proto-Pidgin
Proto-Germanic (PGmc.) 42, 108, 110, 112, Saka (East Iranian) 463
121, 125, 130, 154, 184, 211, 212, 401 Samoan 338, 420, 421
Proto-Indo-European (PIE) 17, 22, 25, 27, 29, Samoyed 417
32, 33, 34, 35, 46, 47, 49, 50, 54, 106, Sandawe 422
108, 110, 111, 115, 121, 125, 130, 134, 140, Sanskrit (Skt., also Sanscrit) 8, 13, 14, 22, 25,
141, 198, 206, 216, 315, 313, 314, 321, 27, 29, 30, 32, 33, 34, 35, 36, 46, 51, 52,
336, 350, 358, 401, 402, 408, 409, 411, 53, 56, 58, 93, 98, 99, 100, 104, 105, 106,
413, 414, 416, 431, 434, 435, 436, 439, 107, 109, 110, 111, 127, 130, 132, 134, 135,
447, 448, 449, 450, 451, 453, 456, 457, 140, 182, 183, 198, 205, 206, 216, 217,
458, 459, 460, 461, 462, 463, 464, 465, 218, 224, 231, 247, 251, 252, 253, 270,
466, 467, 468, 469, 470, 471, 472, 473, 271, 272, 297, 301, 305, 314, 322, 323,
481, 483, 484, 485, 489, 492, 493, 505, 326, 327, 334, 350, 356, 358, 359, 360,
509, 510 361, 396, 398, 401, 403, 408, 409, 410,
Proto-Nostratic 435 411, 413, 414, 416, 431, 432, 433, 434,
Proto-Pidgin 376–378; see also Portuguese 436, 440, 449, 450, 451, 452, 453, 455,
Proto-Pidgin 456, 457, 458, 460, 461, 462, 463, 464,
Proto-Romance (PRom.) 153, 214 467, 468, 470, 472, 473, 474, 476, 481,
Proto-Slavic 45, 231 490, 497, 498, 504; see also Vedic
Proto-Uralic 429 Saramaccan 377, 378
Proto-World 426, 428, 437, 443, 507 Scandinavian 40, 41, 43, 83, 162, 206, 208,
Puerto-Rican Spanish 123 242, 243, 246, 261, 278, 289, 293, 337,
490; see also Norse
Quechua 334, 424, 439, 467, 506 Scots English 9, 22, 26, 226, 229, 287, 288,
Quechumara 424 300
Scots Gaelic 38, 242, 281, 288, 343
Romance 13, 14, 29, 30, 33, 39, 41, 44, 56, Semitic 20, 26, 27, 47, 71, 73, 75, 76, 77,
118, 120, 132, 153, 224, 254, 255, 256, 78, 79, 80, 81, 86, 90, 99, 103, 218,
289, 305, 307, 316, 326, 327, 336, 337, 248, 395, 416, 417, 421, 427, 435,
338, 343, 350, 351, 352, 355, 374, 376, 440, 467, 468, 469, 470, 476, 491,
377, 380, 382, 385, 399, 484, 488, 490, 507
505, 512 Seneca 423
550 Language index

Serbian 45, 55, 56, 350, 352, 353, 399 Tagalog 403, 420, 421
Serbo-Croatian 45, 53, 56, 350; see also Takelma 439
Bosnian-Croatian-Montenegrin-Serbian, Tamil 11, 53, 98, 101, 198, 206, 253, 358,
BCMS 361, 400, 401, 420, 436, 439, 440, 448,
Setswana 421 463, 494
Shawnee 423 Telugu 53, 214, 420
Sinhala 54 Tfaltik 439
Sino-Tibetan 419, 420, 506 Thai 420
Sintashta 465, 466, 481, 511 Thracian 47, 55, 339, 354, 469
Siouan 423, 506 Tibetan 101, 419, 420, 506
Sioux 423 Tibeto-Burman 356, 358, 362, 419
Slave 102 Tlingit 422
Slavic 32, 33, 39, 43, 44, 45, 46, 47, 49, 50, Tocharian 29, 54, 56, 57, 313, 362, 400, 401,
52, 54, 55, 56, 81, 82, 213, 214, 231, 271, 416, 449, 450, 451, 459, 460, 463, 473,
273, 274, 313, 314, 337, 343, 350, 351, 490, 509
352, 354, 355, 399, 408, 450, 453, 455, Tok Pisin 386, 387, 388, 503; see
457, 461, 462, 463, 470, 490, 510 also Melanesian Pidgin English,
Slovak 45 Neomelanesian
Slovenian 45 Toma 103
Somali 421, 440 Tsakonian Greek 48, 299, 430
Songhai 422 Tübatulabal 424
Sorbian 44, 45, 343 Tunguz 418
Sotho 107, 125 Turkic 469
South Asian English 42; see also Indian Turkish 97, 132, 198, 230, 266, 334, 350,
English 352, 399, 400, 401, 418, 419, 435, 473,
Southern Arabic 436 494
Southern US English 114 Tuscarora 423
South Sea sailors’ jargon 379
Southwestern French 207 Ubykh 419
Spanish (Sp.) 11, 13, 22, 26, 30, 31, 39, 40, Ukrainian 45, 198, 231
41, 42, 56, 59, 64, 94, 110, 120, 121, 123, Umbrian 40
127, 153, 160, 198, 201, 214, 221, 223, Upper Sorbian 44, 343
240, 254, 255, 256, 264, 271, 272, 274, Ural-Altaic 418, 427, 442
275, 277, 278, 281, 289, 293, 296, 297, Uralic 43, 46, 125, 270, 356, 358, 398, 401,
306, 319, 322, 334, 336, 337, 339, 352, 417, 418, 419, 426, 427, 428, 429, 430,
354, 355, 368, 369, 377, 399, 436, 489, 431, 434, 436, 438, 439, 441, 442, 448,
497 482, 505, 507, 512; see also Finno-Ugric
Spanish pidgin 377 Urdu 53, 54, 253, 334, 344, 347, 348, 349,
Standard English 144, 165, 177, 179, 180, 350
221, 286, 294, 296, 305, 385, 388, 390, Ute 271, 424
489; see also English Uto-Aztecan 423, 506
Sudanic 422
Sumerian 63, 65, 69, 71, 73, 86, 202, 358, Vai 102, 103
419, 427, 437, 462, 465 Vedic 51, 99, 297, 323, 451, 452, 461, 474,
Surinam 377, 378, 439 477, 478, 479; see also Sanskrit
Swahili 97, 238, 293, 341, 421 Venetic 55
Swedish 13, 41, 43, 56, 261, 288, 289, 399 Votyak 460
 Language index 551

Walapai 439 Yiddish 11, 156, 226, 280, 281, 335, 393,
Welsh 38, 56, 216, 217, 336, 337, 343, 394, 396, 494
399, 494 Yoruba 327, 421
West Greek 48, 79, 80 Younger Avestan 463
West Slavic 45 Yuman 423
Winnebago 423 Yupik 439
Wiyot 423 Yurok 423
Wolof 283, 291, 386, 421, 499
Wyandot 423 Zulu 421
Zuni 426
Xhosa 107, 122, 124, 338
General index
[NOTE: See also the detailed Table of Contents. References to phonetic terms are limited to pages
where the terms are defined.]

abbreviation 68, 123, 261, 269; see also alveolar 24, 27

clipping ambiguity 157, 187, 188, 206, 207, 257
abjad 69, 70, 73, 75, 76, 77, 78, 96 American Indian 243, 270, 392, 393, 422,
ablaut 156; see also apophony 467; see also Indigenous American,
absolutive 347–348, 357–358 Native American
accent (phonetic) 22, 25, 27, 111–113, 128, American Indian boarding schools 393
184, 256, 269, 341; see also stress analogy 6, 7, 28, 138, 140, 141, 142, 145,
accent (in speech) 11, 138, 229, 293, 148, 152, 154, 156, 158, 162, 164, 165,
299–300, 302, 330, 332, 345, 394, 166, 167, 169, 170, 174, 181, 197, 207,
499 208, 209, 231, 233, 261, 265, 494;
accommodation (in language contact) 345, see also blending, contamination,
346, 360 four-part analogy, non-systematic
acoustically or auditorily conditioned analogy, proportion, regular analogy,
changes 122 rule-governed analogy, sporadic
acquisition of native speakers 384–386; analogy, systematic analogy
see also “catastrophic” creolization, analogical developments 138, 140, 142, 145,
creolization, depidginization 154, 156, 158, 162, 164–170, 174, 181,
acronymy 68 208, 261
acrophony 67, 68, 76, 77, 78, 270 analogical extension 169
adjacency 154, 179 Anatolia 38, 48, 49, 54, 208, 216, 448, 453,
adoption 11, 14, 39, 80, 135, 223, 225, 232, 464, 465, 466, 470, 471, 472, 480
233, 234, 244, 245, 246, 247, 248, 249, Ancient Near East 54, 60, 61, 62, 64, 66, 73,
251, 257, 259, 264, 322, 323, 497 75, 77, 90, 96
affix 128, 145, 151, 167, 168, 169, 170, 204, Andaman Islands 362
227, 235, 372, 389, 428 “Anglo-Saxon four-letter” words 204
affricates 21, 22, 26, 124, 125, 315 antonyms 153, 154
agrarian life 457, 458, 461, 471 apophony 156; see also ablaut
agriculture 454, 457, 458, 459, 460, 461, 471, arbitrary 190, 196, 198, 199, 200, 496
485, 510; see also farming arbitrary relation between sound and
AIT, see Aryan Invasion Theory meaning 190, 198–200, 496
Alexander the Great 48, 99, 273, 323, 340 archaeology 89, 447, 448, 452, 454, 461, 471
Algonquian ‘fire-water’ 412, 436 481, 510, 511
allegro speech 127, 128 archaeozoological evidence 464, 470, 474
alliteration 155 argots 203, 278, 279, 280, 281, 282, 283,
all-purpose preposition in pidgins and Trade 284, 291, 292, 303
Jargons 372, 383 Arkaim 481, 511
alphabet 8, 19, 25, 26, 38, 44, 45, 47, 50, 70, Armenian sound shift 107, 414
73, 75, 77, 78, 79, 80, 81, 82, 83, 85, 86, articulator 19, 23
88, 89, 92, 96, 97, 98, 99, 102, 103, 210, artifacts 228, 264, 436, 452, 455, 461
238, 270, 491, 492 artisans 99, 455
Altaic hypothesis 418, 427, 505 ārya 50, 272, 449, 477, 509

https://doi.org/10.1515/9783110613285–022
 General index 553

“Aryan” 50, 448–449, 474–477, 480, bilingual 11, 46, 86, 89, 326, 334, 343, 344,
481–483 345, 346, 365, 366, 376, 388, 393, 470
Aryan Invasion Theory (AIT) 448, 481–482, bilingual contact 345, 366
511 Bilingual education 393
Aryan Myth 448 bilingualism 46, 321, 322, 343, 344, 346,
Ashoka 98, 99 347, 350, 363, 374, 393, 501
aspiration 21–24 bioprogram (theory and controversy) 384,
assimilation 116, 117, 124, 134, 441 385, 444, 503
Athenian democracy 455 Biogenetic Law 444
attitude (as a factor in linguistic change) 90, birth of languages 392
137, 172, 245, 247, 250, 251, 289, 303, bitting (of horses) 464; see also horse
310, 382, 389 domestication
attrition 188, 343, 368, 369, 371, 373, 376, blending 7, 152, 153, 154, 155, 156, 157, 166,
393, 394, 395, 504 167, 169, 265, 268, 277, 475
augment 413 border area contact 347, 363
Australia 42, 164, 296, 424, 425 borrowed (material) 9, 15, 37, 41, 94, 97, 118,
auxiliary 183, 184, 185, 186, 322, 323, 329, 131, 142, 167, 210, 211, 213, 223, 224,
331, 353, 354, 355, 384, 386, 430, 495; 225, 227, 228, 229, 230, 232, 233, 235,
see also clitic auxiliaries 236, 238, 241, 242, 244, 247, 248, 253,
avoidance of excessive homonymy 196–197, 254, 255, 256, 258, 259, 262, 276, 279,
206, 207, 211 292, 305, 306, 401, 406, 408, 432, 435,
avoidance of synonymy 207 460, 461, 465, 466, 470, 471, 472, 497
Aztec empire 423 borrowing(s) 11, 14, 15, 28, 40, 41, 42, 45, 52,
87, 94, 110, 118, 153, 160, 167, 168, 208,
babbling 404 209, 211, 213, 215, 223, 224, 225, 226,
Baby Talk 369, 370, 376, 379, 380, 503; see 227, 228, 229, 231, 232, 233, 234, 235,
also Nursery Talk 237, 238, 239, 240, 241, 242, 243, 244,
back 8, 22, 23, 24, 26, 27 246, 247, 248, 249, 250, 251, 252, 253,
backflow (in language contact) 345 254, 256, 257, 258, 259, 263, 264, 266,
backformation 148, 150, 151, 160, 233, 265, 281, 282, 291, 292, 294, 303, 304, 305,
268 306, 308, 310, 317, 327, 333, 346, 347,
Balkans 11, 38, 39, 47, 54, 55, 343, 346, 359, 404, 405, 423, 427, 431, 432, 434,
350, 351, 352, 355, 365, 406, 455, 469, 435, 436, 462, 463, 469, 470, 471, 472,
502 473, 488, 497, 498
Barber, E. J. W. 454 boustrophedon 79
basic meaning 160, 194, 207, 211, 459; see bow-wow theory 443
also core meaning Brahmi 95, 98, 99, 100, 101, 102, 492
basic vocabulary 227, 228, 242, 243, 244, British “Anglicists” 448
255, 334, 405, 406, 412, 429, 472 British “Orientalists” 448
basic word order 183, 406; see also SOV, Bronze Age 456, 457
SVO, VSO, word order Buddha 454
beech-tree hypothesis 467, 468, 470, 510 build-up of interlanguages 345
biblical thinking 467 Bukele 102
Bickerton, Derek 384, 385, 444, 503
bidirectionality of interlanguage in calquing 235, 244, 247, 257, 263, 264, 372,
convergence 345, 349 373, 412, 466
bilabial 21, 81 Canaanite writing 76
554 General index

capital letters 84 412, 413, 415, 416, 419, 425, 427, 428,
“catastrophic” creolization 386 436, 437, 442, 443, 447, 475, 478, 479,
central 23, 24 485, 505; see also historical-comparative
centralized 64, 137 linguistics
centum – satem division 413 comparative method 453, 482, 483, 484,
chain shifts 123, 124, 125, 493 485, 505, 512
Champollion, François 88 comparative reconstruction 14, 33, 410, 411,
chance similarities 361, 400, 401, 402, 429, 442, 505
430, 432, 435, 437, 438, 463; see also compensatory lengthening 121, 493
similar(ity) competing forms 153, 165
change in connotations 212 complexity 17, 117, 167, 169, 311, 317, 376,
changes in culture and society 209 391, 445
chariot 212, 461, 463–466, 470–471, 473, compounds 158, 159, 160, 168, 204, 234,
481 245, 267, 497, 498
chariot burials 465 conditioning of change 111–113, 114, 116,
development of chariot 461, 470, 481 117–118, 122, 123, 136, 138, 142, 152,
invention of Indo-Europeans ? 162, 184, 190, 292, 360
465–466 connection between sound and meaning 200
Cherokee syllabary 73 connotations 2, 9, 139, 156, 158, 182, 202,
Chinese logographs 70 203, 209, 210, 211, 212, 213, 214, 215,
Chinese Sign Language 425 219, 220, 223, 233, 242, 243, 244, 247,
Chinese writing 63, 95, 96, 97, 251 257, 259, 266, 271, 274, 275, 277, 284,
Chinook Jargon 373, 383, 503 291, 311, 378, 404, 422
circumlocution 292, 371, 372 contact 10, 11, 32, 46, 53, 89, 102, 223, 231,
clan 62, 256, 453 242, 243, 244, 247, 261, 272, 290, 293,
classifiers 92, 373 308, 310, 317, 318, 319, 320, 322, 332,
clay tokens 60, 62 334, 335, 339, 340, 341, 342, 344, 345,
clitic 128, 168, 169, 170, 183, 184, 242, 347, 359, 360, 361, 362, 366, 371, 376, 378,
353, 495 379, 383, 391, 392, 404, 406, 470, 472,
clitic auxiliaries 184 473, 495, 497, 501, 502
code mixing 332, 333 contact-induced similarities 406
code switching 332–334, 335, 346 contamination 152, 153, 154, 155, 156, 157
cognates 87, 139, 146, 231, 255, 401, 402, contiguity 196
403, 411, 417, 420, 428, 430, 431, 432, controversy 13, 40, 46, 49, 53, 70, 75, 104,
434, 437, 438, 441, 442, 456, 457, 458 108, 334, 346, 355, 359, 385, 403, 416,
collocations 189, 226, 227, 258, 259, 327, 417, 418, 419, 421, 422, 423, 424, 426,
453 428, 429, 431, 440, 443, 444, 445, 447,
colloquial 3, 221, 267, 268, 298, 299, 300, 448, 458, 466, 470, 483, 484, 489, 507
363, 368, 381 convergence 46, 49, 52, 53, 339, 343, 344,
common ancestor 29, 30, 36, 398, 401, 406, 345, 346, 347, 348, 349, 350, 351, 352,
417, 442 354, 355, 356, 357, 358, 359, 360, 361,
comparative Indo-European linguistics 46, 363, 364, 365, 371, 374, 406, 413, 415,
49, 104, 398, 401, 414, 457; see also 427, 430, 431, 470, 502
Indo-European linguistics convergence area 351, 356, 363, 364; see
comparative law 456 also linguistic area, sprachbund
comparative linguistics 14, 19, 29, 30, 35, 36, convergence and dialect areas 353–364, 365
45, 47, 55, 57, 140, 205, 398, 401, 404, copula 390, 503
 General index 555

core meaning 2, 3, 191, 192, 194, 195, 208, diffusion 73, 90, 91, 99, 102, 363, 378, 436,
209; see also basic meaning 453, 454, 494, 500
correspondences 29, 30, 32, 33, 35, 105, diffusionist view of writing 63, 91, 511
106, 108, 110, 140, 205, 227, 230, 256, diglossia 297, 301, 302, 303, 304, 305, 307,
328, 329, 399, 404, 405, 406, 408, 410, 324, 326, 387, 388, 499
412, 417, 418, 419, 420, 421, 422, 423, diphthongization 123, 125, 310
424, 427, 429, 430, 431, 432, 434, 440, diphthong 94, 100, 126, 127, 137, 310, 311
467, 469, 472 disagreements in interpretation 415, 483,
cow 447, 458 484, 485
creole 366, 374, 378, 379, 384–388, 389, disambiguating 187
390, 391, 392, 443, 444, 503 dissimilation 115, 116, 129, 130, 131, 493
creole continuum 389 distant assimilation 131
creolization 378, 384, 386, 389, 391, 392, division within Proto-Indo-European
444; see also depidginization society 459
cuneiform 48, 52, 57, 72, 73, 75, 76, 85, 87, dog 10, 146, 168, 169, 190, 198, 199, 200,
88, 491 206, 208, 215, 248, 262, 268, 269, 276,
cursive writing 84, 101 370, 385, 433, 435, 436, 447, 458
cylinder seal 62, 64 domestication of animals 447; see also
Cyril 44, 81 donkey (domestication), horse
Cyrillic alphabet 44, 45, 73, 80, 81, 82 domestication
donkey (domestication) 204, 212, 462, 463,
Dante 29, 31 464, 465, 472, 510
daughter languages 321, 410, 413, 482, 483 donor language 83, 223, 224, 229, 235, 236,
decay 3, 4, 132, 141, 493 244, 245, 247, 254, 304, 472
decipherment 17, 48, 52, 57, 63, 73, 85–91, double negation 176, 177, 385, 496
95, 424, 490, 491 doublets 208, 314
decreolization 388–390 drag chain 125, 126, 127
default 179, 235, 238 Dravidian nationalists 448
deference 219, 220, 222 dual number 34, 35, 46
deformation 204, 205, 211, 215, 492; see Dumézil, Georges 455
also taboo, taboo-induced change
demotic 72 ease of pronunciation 116
dental 21, 24, 26 effect of Latin grammar 176
dependent clause 10, 184–186, 352–356, Egypt 42, 48, 50, 62, 63, 64, 73, 77, 85, 87,
369 88, 274, 302, 452, 454, 465
depidginization 378, 384, 386, 387, 389, Egyptian hieroglyphic writing 69, 72, 75, 76,
391 90, 491
derivational patterns 151 elaboration in vocabulary and depidgi-
devanagari 98, 101 nization 386
dialect 124, 284–306, 308–321; see also Elamitic writing 63
regional dialects in the British Isles Elamo-Dravidian hypothesis 435
dialect death 392 ellipsis 127, 128, 160, 161, 166, 167, 170, 219,
dialect diffusion 313–319, 363 266, 267, 268
dialect leveling 319–321; see also leveling (of empirical evidence 114, 127, 135, 136, 370,
dialect differences) 442, 476, 479
dialectology 308–321 English-based creoles 385
dialectology of convergence areas 363–364 English-based pidgins 371, 374, 377, 385
556 General index

English-Only movement 393, 504 fire-water 412, 413, 436

epenthesis 121, 122 (First) Germanic Sound Shift 105; see also
Ériu 449, 509 Grimm’s Law
establishing genetic relationship 400, 402, Five Nations 423
404, 406, 412, 416, 429; see also genetic folk etymology 15, 158, 159, 160, 209, 224,
relationship of languages, language 232, 233, 239
relationship, linguistic relationship Foreigner Talk 366, 367, 368, 369, 370, 379,
etymology 15, 139 155, 158, 159, 160, 209, 380, 381, 383, 385, 503
214, 224, 232, 233, 239, 260, 261, 262, forerunners of writing 59
263, 264, 266, 268, 270, 272, 274, 276, formal similarities between Foreigner Talk and
278, 280, 282, 438, 439, 440, 441, 442, pidgins 379
461, 463, 486, 488, 494, 498 four-part analogy 142, 144, 145, 146, 147,
euphemism 202, 215, 217, 496 148, 149, 150, 151, 152, 154, 156, 165,
Eurasian convergence area 358 166, 167, 168, 174, 233, 264, 265
Eurasian steppes 448, 464, 481, 510 free word order 188; see also word order
Eurocentric 90, 428 fricatives 21, 22, 26
European colonialist expansion 374 futhark 80
Evans, Arthur 89
Evans, J. 102 Gandhi 53
exaggeration 16, 201, 496 Gelb, I. J. 90, 91, 490, 491, 492
excessive homonymy; see avoidance of gemination 129
excessive homonymy gender 18, 34, 46, 145, 235, 236, 237, 238,
expansion 8, 11, 31, 36, 37, 40, 41, 52, 54, 65, 348, 389, 399, 452, 497; see also
101, 110, 195, 294, 319, 323, 325, 326, grammatical gender, semantic gender
336, 370, 374, 378, 379, 380, 453, 454 genetic affiliation of languages 257, 334,
expansion of vocabulary in depidgi- 422, 425
nization 384, 386–387 genetic groups 425
extension 2, 4, 9, 137, 144, 145, 146, 147, genetic relationship of languages 398, 400,
149, 151, 161, 167, 169, 179, 191, 192, 401, 402, 403, 404, 405, 406, 407, 408,
193, 196, 201, 208, 209, 210, 211, 212, 412, 416, 417, 418, 428, 429, 434, 437;
265, 267, 270, 276, 282, 290, 308, 401, see also language relationship, linguistic
414, 458 relationship
extreme inequality (and pidgins) 383 genomics 448, 464, 471, 472, 473, 474,
eye rhymes 93 479–480, 510–511
Germanic “racial” superiority 449
false friends 237, 427, 431, 434, 437, 438, gestural theory 446
441, 442 Glagolitic 81
family of languages 29, 420, 423; see also glottal (phonetic feature) 20, 21, 26
language family glottalic theory 49, 50, 108, 414, 415, 429,
farming 390, 458, 459, 460; see also 470, 483, 484, 485, 505
agriculture grain 460
fast speech 127, 128, 131, 390 grammatical gender 34, 236, 348
Father Sky 451–454 Grassmann, Hermann 130, 493
female deities 452, 509 Grassmann’s Law 130, 493
female dominance (in religion) 452 Great Vowel Shift (English) 126, 188, 206,
Fertile Crescent 471, 473, 480 302, 493
finite verb 184, 185, 187, 331, 430 Greek alphabet 38, 44, 50, 78, 79, 80
 General index 557

Greek dialects 48, 79, 80 idiolect 135

Greenberg, Joseph 427, 428, 437, 438, 439, immigrants 135, 273, 335, 393
440, 441, 442, 443, 506, 507, 508 imperfect learning hypothesis 374
Grimm, Jacob 104 impersonal commands and infinitive 382
Grimm’s and Verner’s Laws 130 Inca empire 59, 64, 424
Grimm’s Law 33, 104, 105, 106, 107, 108, Indigenous American languages 243, 429,
109, 110, 111, 112, 113, 114, 115, 116, 124, 441, 507; see also American Indian,
125, 126, 132, 133, 139, 188, 262, 315, Native American
338, 414, 492, 493, 500 Indo-European dialects; see also
Grotefend, Georg Friedrich 86 centum-satem
gutturals 20, 26 position of Anatolian 449–450
Gypsies 54, 202, 274, 350, 359, 455, 474; see position of Tocharian 449–450
also Rom, Roma, Roms Indo-European family 17, 50, 416, 459
Indo-Europeanists 49, 107, 132, 313, 413,
Haeckel, Ernst 444 415, 416, 440, 449, 461, 470, 471, 474,
Hall Jr., Robert 484 475, 477, 478, 479, 481, 484
Han’gŭl 97, 98, 99, 100, 103 Indo-European linguistics 46, 49, 104, 398,
heroic 454 401, 414, 449, 467, 477, 479, 481, 492,
hieroglyphic 72, 75, 76, 87, 88, 90, 91, 95, 512; see also comparative Indo-European
421, 491; see also Egyptian hieroglyphic linguistics
writing; Mayan hieroglyphs Indo-European myths 453
highly reduced structure and pidgins 375 Indo-European poetic heritage 453
historical-comparative linguistics 19, 447; Indo-European religion 451–454; see also
see also comparative linguistics Father Sky, female deities, female
homonymy 196, 197, 203, 204, 206, 207, dominance (in religion), male dominance
209, 210, 212, 215, 496 (in PIE religion), Mother Earth, Mother
horse 6, 21, 54, 114, 141, 198, 263, 265, 283, Goddess
447, 458, 461, 462, 464, 465, 466, 470, Indo-European women 454, 509
471, 472, 473, 481, 483, 484, 510, 511 Indo-Uralic hypothesis 431
horse burials 465 Indus script 99
horse domestication 461, 464, 465, 470, Indus Valley 63, 99
471, 481, 510; see also bitting of horses; inertia 148, 210, 212, 325
milking of mares infinitive as invariable verb form in
horse-drawn chariots 465, 466, 470, 471, pidgins 374, 375, 377, 380, 382, 503;
481; see also chariot see also impersonal commands and
host (of clitics) 128, 170, 184, 268, 419 infinitive
Hrozný, Bedřich 87 infinitive-loss (in Balkans) 339, 353, 354,
Hungarian nationalists 448, 482, 483 355, 501
Hungary 39, 301, 482 inflection 7, 34, 35, 46, 54, 145, 149, 167, 374
hunter-gatherers 460, 480, 511 inheritance 358, 401, 460, 463, 498
hyperbolic use 161, 219 innovation 47, 49, 80, 139, 168, 174, 183,
hypercorrection 164, 180, 309 218, 311, 361, 413, 415, 445, 458
inscriptions 38, 39, 40, 42, 43, 51, 52, 55, 63,
Iberian peninsula 39, 464, 510 64, 72, 80, 81, 84, 85, 86, 87, 88, 99,
ideographic 88, 90, 91 485, 490
ideology 253, 270, 448, 452, 455, 474, 475, institutionalization (of contact varieties,
478, 479, 481 interlanguage, or Foreigner Talk) 331,
558 General index

332, 334, 335, 376, 379, 380, 381–382, labiovelar 22, 23, 24, 27
383 Labov, William 136, 137, 138, 139, 157, 184,
interlanguage 327, 330, 331, 332, 335, 339, 487, 492, 493, 500, 503
340, 341, 342, 345, 346, 371, 379, 388, Labov’s theory of linguistic change 137–139,
501 157, 493
“international” words in Foreigner Talk and language attrition 373, 392, 393, 394, 395,
pidgins 368–369, 379 504
intimacy 219, 221 language contact 10, 11, 223, 243, 290, 341,
“intrusive r” 164 342, 366, 371, 379, 383, 392, 404, 473,
irony 219, 221, 273, 291 495, 497, 501; see also contact
Iroquois League 423 language death 392, 393, 395, 396, 397,
irregular 6, 111, 114, 115, 136, 140, 142, 145, 504; see also language attrition,
149, 162, 164, 183, 255, 261, 304, 314, language murder, language suicide
410; see also sporadic analogy, sporadic language development in children 443–444
effects of semantic change, sporadic language family 13, 17, 32, 36, 38, 47, 55,
sound change 105, 218, 334, 356, 400, 416, 419, 420,
irregular sound change 116, 127, 129; see 421, 422, 423, 425, 427, 441, 448; see
also “random” sound change, sporadic also family of languages
sound change language isolates 419, 425, 427, 437
“irregular” plurals 118, 144 language murder 392
irregular verb 149, 183 language preservation 397, 504
irregularity of sound change in progress 175 language relationship 12, 29; see also
isolates; see language isolates genetic relationship of languages,
linguistic relationship
jargons 203, 247, 278, 279, 281, 291, 292, language revival 395, 396, 397
303, 305, 383, 384, 503; see also North language suicide 393
Atlantic nautical jargon, slave-trader Lassen, Christian 86, 476
jargon laterals 22
Jewish identity 395, 396 Latin as a spoken language 396
Jones, William 29, 30, 32, 35, 36, 37, 51, 52, length 24
104, 105, 275, 305, 398, 467, 480, 481, lenition 119, 336; see also weakening
487, 492, 511 lento speech 128, 129
leveling (analogy) 142–146, 151, 152, 154,
Kazakhstan 464, 465, 481 162, 197, 260
King Sejong 97, 103; see also Korean writing leveling (of dialect differences) 319–321; see
Kisimi Kamala 103 also dialect leveling
koiné (formation) 293, 295, 339–341, 342, Lewis, Martin W. 448
349, 371, 374, 392 lexical (material) 1, 12, 33, 76, 89, 138, 142,
Korean writing 95, 99; see also King Sejong 145, 154, 157, 158, 161, 166, 183, 189,
Krahe, Hans 477, 489 190, 203, 207, 209, 216, 220, 221, 222,
Kupwar 346, 347, 348, 349, 351, 354, 359, 223, 225, 226, 227, 228, 232, 233, 235,
365, 406, 502; see also convergence 240, 249, 252, 253, 257, 259, 260, 261,
(area) 262, 263, 268, 270, 276, 279, 284, 308,
Kuz’mina, Elena Efimovna 481 310, 311, 312, 317, 326, 327, 333, 335,
347, 359, 369, 373, 376, 377, 378, 379,
labial 26, 33, 414 380, 383, 397, 398, 400, 407, 408, 410,
labiodental 21 412, 413, 417, 418, 427, 429, 431, 434,
 General index 559

435, 437, 471, 472, 473, 483, 498, 499, logographic writing 67, 70, 71, 73, 85, 87, 89,
507, 509 91, 92, 95, 96, 97, 102, 103, 251
lexical borrowing 223, 327 longer-distance comparison 429
lexical change 166, 222, 260–283, 284, 472, longer-distance relationships 427, 442
473, 498 long-standing bilingualism 344, 350
lexical comparison 435 loss 116, 119, 120, 121, 134, 151, 162, 179,
lexical reconstruction 412, 413 188, 212, 221, 260, 263, 306, 320, 339,
lexical replacement 204 346, 354, 355, 392, 396, 402, 440, 441,
lexical semantics 189, 190, 222; see also 483, 493, 498, 501, 504
semantics Lottner, Carl 110, 116, 492
lexicon 11, 150, 202, 204, 216, 219, 221, 222, lower-case letters 84
244, 248, 251, 257, 260, 263, 278, 300,
327, 368, 371, 372, 378, 387, 405, 411, main clauses 10, 186, 187
412, 489, 496 main verb 183, 185, 186, 331, 430
lexicon extremely limited in pidgins 371 male dominance (in PIE religion) 451, 452
liaison 157 maliq‘a ‘swallow’ 438–442, 507; see also tik
“life cycle” of pidgins 391, 392 ‘finger’, “world etymology”
limited communication (in pidgins, trade Malta 452, 509
jargons) 380, 384 manner of articulation 19
Linear A 89, 90 manumission 455
Linear B 47, 48, 88, 89, 491 Martha’s Vineyard sound shift 137, 138, 139,
lingua franca 43, 53, 322; see also link 290, 310
language Martinet, André 123
linguistic area; see convergence area “Mass Comparison” 427, 437, 438, 442, 443,
linguistic evidence 47, 376, 448, 449, 507
462 Mayan civilization 424
linguistic identity (preserving) 346 Mayan hieroglyphs 70, 88, 90–91, 95, 102,
linguistic palaeontology 447–485, 486, 509, 491
510 melioration 212, 213, 215
linguistic reconstruction 451 merger 119, 125
linguistic relationship 14, 32, 105, 140, Meso-America 62, 63, 64, 70, 88, 90, 91,
398–436, 483; see also genetic 102, 452
relationship of languages, language Mesopotamia 48, 57, 62, 63, 64, 71, 72, 73,
relationship 85, 86, 88, 95, 419, 421, 437
link language 53, 322, 324, 326, 327, 328, Metal Age 456
331, 335, 339, 340, 341, 342, 380, 384, metaphor 9, 66, 191, 161, 195, 196, 200–202,
386, 392, 396, 501; see also lingua 208, 211, 240, 262, 282, 392, 470, 478,
franca 496
lion 469, 470 metaphorical extension 9, 161, 196, 208,
liquid 22, 25 210, 267
literacy 9, 13, 48, 58, 59, 84, 86, 197, 231, metathesis 130, 131
270, 302, 323 Methodius 44, 81
literary traditions 85, 288, 289, 453 migration 31, 36, 40–41, 54, 319–321, 344,
Liverpool sound shift 108, 125, 133 359, 423, 450, 458, 469, 473, 476, 481,
loans 223, 242, 243, 244, 257; see also 483, 511
borrowing military conquest 453–454
logic (appeal to) 177 milking of mares 464
560 General index

minority languages 343, 394 Nursery Talk 129, 369; see also Baby Talk
missionaries 32, 102, 249, 387 nursery words 266, 275, 402, 404, 427, 436
mixed pastoral/agricultural society 458
mnemonic devices 59, 60, 61, 62, 64 Occam’s Razor 410, 414, 415, 470
morphological change 166–169, 181, 188, Ogham (writing) 38, 40, 82, 83, 491
472, 494, 495 Old Persian syllabary 74, 75, 86
morphology in analogical change 148, 150, one meaning – one form 143, 197, 208
151, 157, 234, 413 onomatopoeia 155–156, 199–200, 205–206,
Morse Code 82 208, 211, 368, 402, 404, 427, 436, 463
Mother Earth 452–454 onomatopoetic replacement 206
Mother Goddess 451–452 oo-shortening 138, 139
“Multilateral comparison” 427, 438 Ø-plural 146, 147, 149
mutual accommodation 360 Oppert, Jules 86
mutual convergence 350 optimal closeness for comparative linguistics
mutual intelligibility 287–290, 301, 321, 331, 412, 416, 436
339 oral language 17, 238–239, 425, 428, 437
oral tradition 47, 51–52, 57–59, 99
narrowing of meaning 8, 192, 195 origin of human language 437, 443
nasal 20, 21, 24, 25, 27 original home of Indo-European
nationalism 45, 294, 448, 481, 482, 483, 511 (general) 459, 466–472, 480–481
nation-state, monolingual 343, 363 original home (hypotheses)
Native American 102, 273, 383, 468; see also Anatolia 448, 453
American Indian, Indigenous American Balkan 469
languages Caucasus 469
naturalness in reconstruction 409, 414, 415, Central Europe 467–468
483 (Eurasian) steppes 471, 473, 511
Nazi holocaust 248, 396 India 467
Nazism 50, 248, 393, 396, 448, 449, 474, Iranian plateau 63, 467
477, 480
Neanderthal 445, 508 palaeogenomic research; see genomics
negative spread; see double negation palatal 20, 22, 25, 26
negative verb 430 palatalization 22–23, 24, 27, 118, 230, 441
neogrammarians 113, 115, 116, 123, 127, 129, paradigm 109, 142, 143, 144, 167, 176, 374,
132, 133, 135, 136, 137, 141, 142, 162 380, 388, 494
Neolithic 456–457 paradigmatic alternation 109, 118, 142–143,
neo-Nazis 448, 480 408
Niebuhr, Carsten 85, 86 paronomasia 202
Njoya 103 passive 4, 11, 34, 35, 181, 183, 188, 292, 371
nomadism 457, 471 past tense, replacement of, by perfect 363
nonstandard 7, 130, 161, 162, 165, 177, 227, pastoralism 458–461, 471, 473, 480, 510
230, 285, 299, 326, 340, 374, 390, 489 patriarchal nature of PIE society 453
non-oral means of communication 59, 445 peaceful spread of agriculture
non-systematic analogy 151, 174, 189, 215; (hypothesized) 471
see also sporadic analogy pejoration 212–215, 403
Normans 224, 243 Pereltsvaig, Asya 448
North Atlantic nautical jargon 378 peripheral meanings 2, 195
numerals 92, 153–154, 489 Persia 29, 51, 85, 86, 398
 General index 561

pharyngeal 20 productive, productivity 6, 145, 147–151, 155,

phonetic indicators (in writing 156, 168, 183, 187, 207, 234, 264, 354,
systems) 66–67, 70, 91, 95, 96 411
phoneticization of writing system 67 progress in linguistic change (theory) 4, 132
phonetic similarities 93, 400, 406, 417, 434, pronouns 34, 42, 88, 177, 178, 179, 180, 181,
436 182, 219, 220, 221, 222, 227, 228, 235,
phylum 429 236, 242, 334, 358, 373, 431, 434, 483
pictographic writing 61–64, 66–67, 71–72, proof of genetic relationship 408
75–77, 89, 90, 95 proportion 47, 147, 148, 150, 155, 165
pidgin 11, 180, 339, 366–391, 443, 444, prothesis 121
503 proto-language 198, 408, 409, 410, 412, 413,
pidgin origins 374–380; see also Imperfect 414, 415, 416, 447, 482, 489
learning hypothesis, “racial inferiority” prototypical meaning 209
hypothesis, Proto-Pidgin hypothesis, purism 158, 247, 253; see also linguistic
Foreigner Talk nationalism
pidginization 339, 392 push chain 125, 127
pidgins as administrative auxiliary
languages 386 Quakers 220
pidgins as missionary languages 386 questionable “syntactic” changes 171–172
pidgins as institutionalized forms of Foreigner quipus 59–61, 64
Talk 379–380, 383
pied-piping 177 “racial-inferiority” hypothesis
place of articulation 19, 20, 21, 23, 99 (problematic) 375–376; see also pidgin
plough 93, 458, 459, 460 origins
politeness 114, 127–128, 171, 204, 219–221, racism, racist(s) 50, 374, 375, 376, 379, 449,
367 475–479
polysemy 191–192, 197, 212, 257, 372 radical reduction of vocabulary in pidgins 371
pooh-pooh theory 443 radical simplification of structure in
popular etymology 15, 159, 160, 209; see pidgins 371, 391
also folk etymology “random” sound change 109–110, 140,
Portuguese words in pidgins and 395; see also irregular sound change,
creoles 377–379 sporadic sound change
Portuguese Proto-Pidgin hypothesis 376 Rask, Rasmus 33, 86, 104, 105, 107, 108, 113,
prehistory 46, 47, 143, 363, 423, 437, 447, 124, 492
448, 461, 475, 478, 491, 509 Rawlinson, Henry 86
prejudice and semantic change 212–214; see r-deletion 163, 185
also pejoration reanalysis 154, 156, 157, 158, 171–172, 277
preliterate 58; see also literacy rebus principle 67
prestige 3, 41, 135, 164, 165, 225, 231, 232, recomposition 159
235, 240, 241, 242, 243, 244, 246, 247, reconstruction 25, 408–416, 420, 423, 429,
248, 251, 252, 257, 259, 271, 285, 293, 435, 447, 451–453, 456, 458–463, 468,
299, 300, 301, 302, 303, 305, 306, 308, 483–485, 512
309, 310, 311, 312, 320, 322, 324, 326, recurrent similarities and comparative
327, 340, 344, 347, 388, 389, 392, 394, linguistics 4, 401, 405, 407
489, 504 reduction in structure and vocabulary
pre-writing 62, 64, 65 (pidginization) 383
primate language abilities 445 reduplication 266, 276, 404, 411, 463, 465
562 General index

regional dialects in the British Isles 390 rule-governed analogy 162–164

regional innovations in Indo-European 49, runes 15, 42, 43, 73, 80–81, 83, 210, 490,
413, 470; see also Indo-European 491; see also futhark
dialects
registers 291–292 Schlegel, Friedrich 36, 467, 475, 476, 479,
regular analogy 162–164; see also 511
rule-governed analogy Second or High German Sound Shift 105
regularity hypothesis 113, 115, 116, 127, 129, second position (syntax) 183, 184, 185, 186
139–141, 409, 492, 493 Sejong; see King Sejong
regularity of sound change 6, 107, 117, selection of a language as link language 323,
127, 138–143, 159, 162, 164, 205–207, 324, 346
260–261, 318, 406, 482 selective simplification (in language
reinterpretation 43, 70, 84, 151, 156–157, contact) 341, 349, 354, 371, 383
161, 164, 168, 172, 184–186, 208–210, semantically induced changes 208, 211
220, 308, 389, 494 semantic change 8, 146, 189–223, 279–280,
relative chronology 112, 113 283, 440, 449, 464, 496
relexification 376, 377, 378 semantic differentiation 207, 208, 211
relics 143, 150, 207, 209, 283, 390 semantic extension 193, 265, 458
religion; see Indo-European religion semantic fading 203
Renaissance 15, 31, 43, 175, 249, 253, 267, semantic fields 212, 215
277, 294 semantic fuzziness 8, 189, 190–200, 215,
Renfrew, Colin 453, 454, 471, 507, 509, 511 222, 280
replacements of homonyms 215 semantic gender 348
resurrection of a language 395 semantic indicators 66, 70, 95, 96
retroflex 24, 27 semantic overlap 191–196
retroflex/dental contrast 320, 328–333, 338, semantic range 8, 194, 195, 196, 208, 257
356, 359–362 semantic reinterpretations 210
rhyme 93, 155–156; see also eye rhyme semantics 67, 68, 142, 189, 190, 212, 222,
rhyming formation 155–156, 260, 282, 494 251, 253, 260, 261, 401, 440; see also
r-insertion 163, 164, 165, 185 lexical semantics
Rom, Roma, Roms 359, 455, 469, 474; see semantic specialization 258, 460
also Gypsies semantic transference 65
Roman alphabet 17, 38, 45, 73, 79, 80, 82, semi-speakers (in language attrition) 395
96 Semitic identity of Jewish people 395
root (morphology) 29, 32, 34, 75, 80, 87, 89, semivowels 22, 26
97, 104, 106, 110, 112, 115, 117, 128, 134, Sequoyah 102
138, 139, 143, 162, 167, 183, 218, 227, sexist attitudes 213; see also pejoration
235, 314, 374, 377, 382, 389, 398, 401, Shakespeare, William 126, 127, 151, 161, 169,
402, 403, 411, 412, 430, 438, 440, 456, 171, 175, 177, 178, 180, 213, 223, 260,
458, 457, 459, 461, 462, 463, 464, 468, 285
483, 484, 497, 498 shared idiosyncrasies and comparative
root (of family tree) 334 linguistics 407, 430–431
round 23, 27 Shaw, George Bernard 16, 18, 500
royal or editorial we 220 shift of language loyalty 350
Ruhlen, Merritt 438, 439, 507 shift of the idealized “target” (theory of sound
rule 156, 162–164, 176–180, 185, 258, 239, change) 133
297, 310, 329, 330, 395 sibilant 21, 25–26, 33
 General index 563

sign 497, 506, 508 social setting for change 9, 128, 209–214,
sign languages 17, 119, 121, 156, 238–239, 220, 228, 240, 263, 305, 309, 323, 326,
425, 428, 437, 444–446, 487–488, 492, 370–371, 374, 382–383, 388, 473, 502,
497, 506, 507 503
silent barter 368 social significance of dialect and regional
similar(ity) 29, 32, 62, 64, 100, 110, 151, 214, accents 299–300
217, 256, 261, 355, 398, 400, 401, 402, social structure of Indo-European 448, 451,
427, 430, 432, 438, 440; see also chance 453–456, 458–459, 461, 483, 509
similarities, contact-induced similarity, sociolinguistic criteria for distinguishing
phonetic similarities, recurrent language and dialect 289
similarities sociolinguistic factors 296, 310, 336, 379,
simplification; see also selective simpli- 380, 381, 382
fication in language contact, radical Socrates 454
simplification of structure in pidgins, “Somerset” voicing 305, 457
variable simplification in Trade Jargons, sound change 6, 7, 8, 12, 19, 25, 28, 50,
word-final simplification 104–140, 141, 143, 145, 148, 151, 158,
simplification in grammar or structure 12, 159, 162, 164, 165, 166, 167, 170, 175,
169, 179, 221, 369, 371–374, 375, 376, 181, 184, 190, 198, 205, 206, 207, 209,
379, 383, 503 211, 256, 260, 261, 262, 271, 272, 292,
simplification in the lexicon 12, 376, 379, 293, 303, 304, 313, 314, 318, 361, 406,
383, 389, 503 409, 410, 411, 430, 440, 449, 482, 493;
simplification as result of sound see also regularity hypothesis, regularity
change 117–118, 119, 124, 134 of sound change vs. “random” sound
simplified, Foreigner Talk variety of change, irregular sound change or
Portuguese 380 sporadic sound change
Sintashta Culture 465, 466, 481, 511 sound symbolism 215
slang 1, 8, 140, 203, 265, 267, 268, 269, 277, sound system 50, 93, 94, 411, 413, 414
278, 279, 280, 281, 282, 284, 286, 290, South Asia 11, 24, 27, 36, 52, 53, 84, 273,
291, 303, 394, 499 331, 334, 343, 344, 346, 356, 358, 359,
slavery, slaves 43–44, 213–214, 274, 283, 360, 362, 406, 419, 420, 449, 457, 464,
378, 379, 380, 383, 385, 386, 387, 388, 465, 470, 481
390, 455, 456 SOV 183, 186, 356, 358, 361, 406, 495; see
slave-trader jargon 378 also word order
social attitudes and language change 212, Sprachbund; see convergence area
244–253, 376, 388 speakers’ attitudes, importance of 3, 16,
social connotations 242, 277, 378 112, 139, 158, 159, 163, 170, 189, 198,
social considerations 137, 326 199, 200, 240, 259, 278, 284, 347, 388,
social dialects 290–293, 388–389; see also 428
standard languages speech community 138, 297, 320, 331, 392,
social distance as factor in 393, 394, 395
pidginization 380, 382, 383 speech error 116, 131, 139, 153–154, 493
social factors in politeness 221–222 speech sounds 17, 18, 19, 22, 23, 77, 119,
social motivation of change 136–139, 146, 120, 121, 129, 190, 198, 199, 205, 445
157, 175, 199, 201, 204–205, 212–215, spelling 6, 15, 16, 17, 18, 25, 28, 67, 68, 87,
278–279, 308–309, 326, 344, 361, 384, 89, 91, 93, 94, 98, 129, 158, 159, 206,
386, 393, 394, 493; see also Labov’s 230, 231, 238, 239, 261, 265, 270, 276,
theory of linguistic change 277, 282, 297, 320, 331, 400, 488, 508
564 General index

spelling pronunciation 158, 297 syllabary 68–70, 73, 74–76, 85, 86, 87, 89,
spoonerism 131, 154 102, 103, 270, 491, 492
sporadic analogy 151–162, 165, 174, 189, syllabic liquids and nasals 25, 27, 34
209, 215, 309, 494 syllable 23, 25, 27, 73, 96, 102, 111, 112, 231,
sporadic effects of semantic change 215 232, 251, 262, 269, 306, 320, 341, 404
sporadic sound change 127–131, 139, symmetry (in American Sign Language) 119
493; see also irregular sound change, synonymy 12, 193, 196–197, 199, 203, 207,
“random” sound change 237
spread of Indo-European speakers 465 syntactic change 10, 12, 170, 161, 166,
standard language 161, 175, 290, 296, 297, 168–169, 170–188, 219–222, 226,
298, 299, 300, 302, 305, 319, 387, 388, 328–333, 335, 339, 347–348,
390 352–356, 356–358, 369, 371; see also
state language 38, 39, 53, 347, 350 questionable “syntactic” changes
statistical models derived from genomic syntactic reconstruction 412–413
research 471, 472 systematic analogy 142–151
steady-state dynamic equilibrium 188 systematic effects of semantic
stimulus diffusion 73, 99, 102 change 215–222
Stone Age 456
stop consonants 20, 26 taboo 114, 127, 199, 200, 202–205, 211, 215,
stranding of prepositions 177 234, 274, 278, 282, 492, 496
stress 25, 27, 111; see also accent taboo-induced change 203, 204, 205, 211,
style 170, 171, 180, 232, 291, 300, 324, 282; see also deformation
487 Tamil script 98, 101
subgrouping 46, 50, 55, 274, 309, 350, 398, Tarzanian 12, 366, 370
399, 418, 420, 449 technical registers 405, 456; see also North
substitution 15, 123, 160, 165, 225, 229, 230, Atlantic nautical jargon, slave-trader
231, 232, 235, 329 jargon
substrate 243–244, 336, 342, 463, 468 Tidewater Area 390
substratum 335–359, 338, 339, 345, 357, tik ‘finger’ 438, 439, 507; see also maliq‘a
358, 359, 361, 365, 451, 501 ‘swallow’, “world etymology”
substratum hypothesis 358–359, 361, 364 tongue twisters 116, 131
suffix 5, 7, 76, 88, 97, 117, 118, 146, 149, 154, TOPIC (in syntax) 182–184, 186–188, 335
158, 166, 168, 169, 225, 226, 235, 237, Tower of Babel 426
253, 265, 277, 325, 347, 354, 401, 402, Trade Jargons 381, 383–384, 503
434, 440, 488 transfer (in language contact) 11, 329–331
Sumer 62, 91, 466 transhumance 457, 460–461, 471, 481, 510
super-families 422, 429 tree diagram 449–450
superstrate 243–244, 336
suppletion 407 Ukraine 43, 464
supraregional dialects 291–300; see also umlaut 117, 118, 141, 144, 145, 225, 244
standard languages understatement 201, 496
SVO 183, 186, 361, 406, 495; see also word unmarked order (syntax) 181–182; see also
order word order
Swat Valley 465 “unrelated” (i.e. without established
Sweet, Henry 18, 25 relationship) 262, 419, 425, 436–437
switch language or dialect 164, 382, 387, Ural-Altaic hypothesis 418, 427
392, 393, 396; see also code switching Urheimat; see original home of Indo-European
 General index 565

utterance-final position 119–120; see also wanderwort, wanderwörter 224, 460

word-final position warlike nature of Indo-Europeans 453, 455
uvular 20, 22, 26 Watkins, Calvert 453, 510
weakening 116, 119, 120, 121, 129, 134, 289,
Vai 102, 103 293, 336, 337; see also lenition
variability (in sound change in wheel 212, 265, 266, 411, 416, 432, 461,
progress) 137–138 463, 464, 465, 466, 471, 484, 510,
variable simplification in Trade Jargon 383, 511
390 wheel invented by the Indo-Europeans
velar 20, 21, 22, 23, 26 (problematic claim) 465
Ventris, Michael 89, 90 word-final position (in sound
vernacular 7, 31, 40, 148, 149, 165, 175, 177, change) 118–119, 120, 162–163, 389;
179, 250, 271, 284, 287, 294, 296, 297, see also utterance-final position
298, 299, 300, 301, 302, 303, 305, 306, word order 10, 14, 43, 45, 181, 184, 186, 187,
324, 327, 328, 374, 385, 387, 388, 390, 188, 356, 495
484, 489 “world etymology” 438–442; see also tik
Verner, Karl 111 ‘finger’, maliq‘a ‘swallow’
Verner’s Law 108, 111–113, 116, 119, 142, writing, development of 61, 62, 68, 70, 77,
143, 144 90, 91, 95, 202
Victorian era 203, 204 writing systems 38, 40, 44, 47, 57–103,
vocabulary expansion in depidginization 202, 210, 251, 423, 424, 490–492,
378, 384 497
vocabulary heterogeneous in Trade Jargons wrong timing and sound change 122
383 xenophobia 382, 393
voice, voiced 21, 33, 35 Xinjiang 52, 362
voiceless 21
voicing 21, 22 yo-he-ho theory 443
vowel 18, 20, 23, 24, 27 Yu. V. Knorosov 90
VSO 406; see also word order
vulgar expressions in pidgins (of European Zarathushtra, Zoroaster 51
origin) 381 Zionism and Hebrew revival 395–396