SEMANTICS: 2

BLANK PAGE
 Volume 2

JOHN LYONS
Trinity Hall, Cambridge

43 CAMBRIDGE
Sg) UNIVERSITY PRESS
Published by the Press Syndicate of the University of Cambridge
The Pitt Building, Trumpington Street, Cambridge CB2 1RP
40 West 20th Street, New York, NY 10011-4211, USA
10 Stamford Road, Oakleigh, Melbourne 3166, Australia

© Cambridge University Press 1977

First published 1977

Reprinted 1978, 1979, 1983, 1985, 1986, 1988, 1990, 1994

Library of Congress Cataloguing in Publication Data

Lyons, John.
Semantics.
Includes bibliographies and Indexes.
1. Semantics. 2. Semiotics.
3. Grammar, Comparative and general. I. Title.
P325.L96 410 76-40838
ISBN 0 521 21560 9 hardback
ISBN 0 521 29186 0 paperback

Transferred to digital printing 2001

 Contents to Volume 1

Typographical conventions x Preface xi

1 Introduction: some basic terms and concepts I
1.1, [he meaning of ‘meaning’ I
1.2. Use and mention 5
1.3. Object-language and metalanguage 10
1.4. Type and token 13
1.5. Forms, lexemes and expressions 18
1.6. Theories, models and data 25
2 Communication and information 32
2.1. What is communication? 32
2.2. A simple model of communication 36
2.3. The quantification of information 43
2.4. Descriptive, social and expressive information 50
3, Language as a semtotic system 57
3.1. Verbal and non-verbal signalling 57
3.2. Paralinguistic phenomena 63
3.3. Language and medium 67
3.4. The design features of language 74
3.5. The origin of language 85
44.1.
Semiotics 96
Signification 96
4.2. Symbols, icons, indices and symptoms 99
4.3. Nominalism, realism and conceptualism 109
4.4. Syntactics, semantics and pragmatics 114
5 Behaviourist semantics 120
5.1. General attitudes 120
5.2. More particular features of behaviourism 123
vi Contents to Volume 1
5-3. Behaviourist theories of meaning 125
5.4. Evaluation of behaviourist semantics 133
66.1.
Logical semantics 138
The formalization of semantics 138
6.2. Propositional calculus IAI
6.3. Predicate calculus 147
6.4. The logic of classes 154
6.5. Time, modality and possible worlds 161
6.6. Model-theoretic and truth-conditional semantics 167
7 Reference, sense and denotation 174
7.1. Introductory
7.2. Reference 177174
7.3. Sense 197
7.4. Denotation 206
7.5. Naming 215
7.6. Reference, sense and denotation in language- 224
acquisition
8 Structural semantics I: semantic fields 230
8.1. Structuralism 230
8,2. The Saussurean dichotomies 239
8.3. Relativism and functionalism 245
8.4. Semantic fields 250
8.5. Syntagmatic lexical relations 261
8.6. General evaluation of the theory of semantic 267
fields
g Structural semantics II: sense relations 270
9.1. Opposition and contrast 270
9.2. Directional, orthogonal and antipodal opposition 281
9.3. Non-binary contrasts 287
9.4. Hyponymy 291
9.5. Hierarchical structure in the vocabulary 295
9.6. Lexical gaps 301
9.7. Marked and unmarked terms 305
9.8. Part—whole relations 311
9.9. Componential analysis 317
Bibliography 336
Index of subjects 357
Index of personal names 369
 Contents to Volume 2

Typographical conventions x Preface xi

10 Semantics and grammar I 373
10.1. Levels of analysis 373
10.2, Grammaticality 379
10.3. Generative grammar 386
10.4. Grammatical ambiguity 396
10.5. Generative semantics 409
11 Semantics and grammar IT 423
classes 423
11.1, Parts-of-speech, form-classes and expression-

: 11.2. Subjects, predicates and predicators 430

actions : 438
11.3. The ontological basis: entities, qualities and

11.4. Determiners, quantifiers and classifiers 452

12 Semantics and grammar III 467
12.1. Kernel-sentences and sentence-nucle1 467
12.2. Predicative structures 469
12.3. Locative subjects 475
12.4. Valency 481
12.5, Causativity and transitivity 488
12.6. Participant-roles and circumstantial roles 494
12.7. Theme, rheme and focus 500
13, The
13.1. Lexicon
Lexical entries512
512
13.2. Complex lexemes 521
13.3. Compound lexemes 534
13.4. Homonymy and polysemy 550
vill Contents to Volume 2
14 Context, style and culture 5'70
14.1. The context-of-utterance 570
14.2. Communicative competence 573
position 592
14.3. Conversational implicatures and presup-

14.4. The contextual theory of meaning 607

14.5. Stylistic, dialectal and diachronic variation 613
14.6. Sentences and texts 622
15 Detxis, space and time 636
15.1. Person-deixis 636
15.2. Demonstratives and the definite article 646

discourse 657
15.3. Deixis, anaphora and the universe-of-

15.4. Tense and deictic temporal reference — 677 |

15.5. Spatial expressions 690
15.6, Aspect 703
15.7. Localism 718
16 Mood and illocutionary force 72.5
16.1. Speech-acts 725
16.2. Commands, requests and demands 7A5
16.3.
16.4. Questions
Negation 753
768
17 Modality 787
16.5. The performative analysis of sentences 778

17.1. Necessity and possibility 787

17.2. Epistemic modality and factivity 793
17.3. Tense as a modality 809
17.4. Deontic modality : 823
exemption 832
17.5. Obligation, permission, prohibition and

17.6. A tentative synthesis 841

Bibliography 850
Index of subjects 881
Index of personal names 293
 Figures

VOLUME I
1. A model of communication page 36
2. The triangle of signification 96
3. Truth-table for the one-place negation connective 143
4. Truth-tables for the two-place connectives 144
tion of classes 158
5. Venn diagrams illustrating the union and intersec-

6. Various kinds of diachronic change 256

7. A model of a hierarchically organized vocabulary 295

VOLUME 2

8. A sample phrase-marker 390

9. The so-called standard theory A412
10. The so-called generative semantics theory 413
11. Schematic representation of a lexical entry 517
 Typographical conventions

SMALL CAPITALS
For sense-components and other more abstract elements, or correlates,
of meaning (cf. 9.9). :
Italics
1. For forms (as distinct from lexemes or expressions: cf. 1.5) in their
orthographic representation.
2. For certain mathematical and logical symbols, according to standard
conventions.
Single quotation-marks
1. For lexemes and expressions (cf. 1.5).
2. For the citation of sentences (i.e. system-sentences: cf. 1.6).
3. For titles of articles.
Double quotation-marks
1. For meanings (cf. 1.5). )
2. For propositions (cf. 6.2).
3. For quotations from other authors.
Asterisk
For technical terms when first introduced and occasionally thereafter to
remind the reader of their technical sense.
Notes
1, When a term has been furnished with an asterisk, single quotation-
marks are not used.
2. Single quotation-marks are omitted when a sentence, expression or
lexeme is numbered and set on a different line; but italics and double
quotation-marks are still used in such circumstances.
3. In quotations from other authors, the original typographical conven-
tions have usually been preserved. Occasionally adjustments have
been made in order to avoid confusion or ambiguity.
 | Preface

When I began writing this book six years ago, it was my intention to
produce a fairly short one-volume introduction to semantics which
might serve the needs of students in several disciplines and might be of
interest to the general reader. The work that I have in fact produced is
far longer, though in certain respects it 1s less comprehensive, than I
originally anticipated; and for that reason it is being published in two
volumes.
Volume 1 is, for the most part, more general than Volume 2; and it is
relatively self-contained. In the first seven chapters, I have done my
best, within the limitations of the space available, to set semantics
within the more general framework of semiotics (here defined as the
investigation of both human and non-human signalling-systems); and I
have tried to extract from what ethologists, psychologists, philosophers,
anthropologists and linguists have had to say about meaning and
communication something that amounts to a consistent, if rather
eclectic, approach to semantics. One if the biggest problems that I have
had in writing this section of the book has been terminological. It is
frequently the case in the literature of semantics and semiotics that the
same terms are employed in quite different senses by different authors
or that there are several alternatives for what is essentially the same
phenomenon. All I can say is that I have been as careful as possible in
selecting between alternative terms or alternative interpretations of the
same terms and, within the limits of my own knowledge of the field, in
drawing the reader’s attention to certain terminological pitfalls. At one
time, I had hoped to be able to follow the practice of never using non-
technically any word that was also employed anywhere in the book in
some technical sense or other. I soon had to abandon this rather quixotic
ambition! Some of the most ordinary words of English (e.g. ‘case’,
‘feature’, ‘aspect’) are employed in a highly specialized sense in lin-
Xil Preface
guistics and related disciplines; and, however hard I tried, I found it
impossible to get by without them. I trust that the context (and the
device of using asterisks for introducing technical terms) will reduce, if
it does not entirely eliminate, ambiguity and the possibility of mis-
understanding.
The last two chapters of Volume 1 are devoted to structural semantics
(or, more precisely, to structural lexicology). This is a topic that I have
been concerned with, on and off, for the best part of 20 years; and,
although the so-called structuralist approach to semantics is no longer
as fashionable among linguists as it once was, I still believe that it has
much to contribute to the analysis of language.
Volume 2 may be read, independently of Volume 1, by anyone who 1s
already familiar with, or is prepared to take on trust, notions and dis-
tinctions explained in Volume 1. In Volume 2, which (apart from the
chapter on Context, Style and Culture) is concerned with semantics
from a fairly narrowly linguistic point of view, I have been tempted to
do something more than merely clarify and systematize the work of
others; and this accounts for the fact that the book, as a whole, has taken
me far longer to write than I had expected it to take. Five of the eight
chapters in Volume 2 -two of the three chapters on Semantics and
Grammar, the chapter on Deixis, Space and Time, the chapter on Mood
and Illocutionary Force, and the chapter on Modality — contain sections
in which, unless I am mistaken, there are a few ideas of my own.
Caveat lector ! | !
As I have said, the book is, in certain respects, less comprehensive
than I intended. There is nothing on etymology and historical seman-
tics, or on synonymy; and there is very little on the structure of texts
(or so-called text-linguistics), or on metaphor and style. If I had dealt
with these topics, I should have had to make my book even longer.
Sometimes one must stop even if one has not finished! :
As I write this Preface, I am all too conscious of having just moved
from Edinburgh where I have now spent twelve years, in one of the
finest Departments of Linguistics in the world. Throughout this time I
have benefited, in my writing and in my teaching, from the advice and
criticisms of my colleagues in several Departments. Many of them have
helped me, as far as the present book is concerned, by reading sections
of it for me in draft and commenting upon them or by discussing (and
in some instances originating) the ideas that have found their way into
my text: John Anderson, R. E. Asher, Martin Atkinson, Gillian Brown,
Keith Brown, John Christie, Kit Fine, Patrick Griffiths, Stephen Isard,
 Preface X11
W. E. Jones, John Laver, Christopher Longuet-Higgins, J. E. Miller,
Keith Mitchell, Barry Richards, and James Thorne. Ron Asher and
Bill Jones have been especially helpful: each of them has read the whole
typescript; and Bill Jones has undertaken to do the index for me. Apart
from these Edinburgh and ex-Edinburgh colleagues, there are many
others to whom I am indebted for their comments on drafts of parts of
the book: Harry Bracken, Simon Dik, R. M. Dixon, Francoise Dubois-
Charlier, Newton Garver, Gerald Gazdar, Arnold Glass, F. W. House-
holder, Rodney Huddleston, R. A. Hudson, Ruth Kempson, Geoffrey
Leech, Adrienne Lehrer, David Makinson, P. H. Matthews, G. A.
Miller, R. H. Robins, Geoffrey Sampson, the late Stephen Ullmann,
Anthony Warner. There are doubtless many errors and inadequacies
that remain but without the aid of so many friends, whose specialized
knowledge in many of the relevant fields is far greater than my own, I
should have gone astray more often than I have done.
Like all teachers, I have learned more from my students over the
years than they have learned from me. It has been my privilege to con-
duct several research seminars and to supervise a fair number of Ph.D.
dissertations on semantics during the period when I was writing this
book. T'wo of my students I must mention by name, since I am very
conscious of having derived directly from them some of the points that
appear in the book: Marilyn Jessen and Claudia Guimaraes de Lemos.
I have no doubt, however, that others of my students are also responsible
for much of what IJ think of as being original in the second volume.
I owe a special debt of gratitude to Rena Somerville who, as my
secretary in the last few years (the best secretary that I have ever had),
has typed so many versions of certain sections of my manuscript that she
could probably reproduce at least the gist of them from memory! Much
of this work she has done at home in the evenings and at the week-end:
I trust that her family will forgive me for the time that I have stolen
from them in this way.
Without the specialized assistance provided by the Cambridge
University Press this book would never have seen the light of day.
Jeremy Mynott read both volumes in typescript and made many valuable
editorial suggestions. Penny Carter was responsible for the sub-editing
and had to cope with far more inconsistencies and handwritten changes
in the typescript than an author should have been allowed to make. I am
grateful to both of them for their help and their forbearance.
Finally, I must record my gratitude to my wife and children for their
willingness to put up with my frequent bouts of depression, ill-temper
XIV Preface
or sheer absent-mindedness while I was writing the book and the
postponement of so many promised outings and holidays. More par-
ticularly, I wish to thank my wife for the love and support that she has
always given me, in my writing as in everything.

Falmer, Sussex J.L.

February 1977
 IO
‘Semantics and grammar I

10.1. Levels of analysis

Most linguists distinguish at least three levels* of structure in their
analysis of sentences: the phonological, the syntactic and the semantic.!
To these three they may or may not add morphology to serve as a
bridge between the syntax and the phonology in particular languages.
Looked at from the point of view of its phonological structure, every
sentence may be represented as a sequence of phonemes with a certain
prosodic contour superimposed upon it (cf. 3.1). The phonemes of a ,
language are conventionally represented by means of letters enclosed
within a pair of oblique strokes. For example, there is in English a
phoneme /b/ which occurs in the initial position of the forms bed,
bread, boil, etc., and is pronounced as a bilabial, voiced, non-nasal stop;
and this phoneme, like all the other phonemes of English, has a charac-
teristic distribution throughout the word-forms of the language. It is
part of the phonologist’s job to list, for the language that he is des-
cribing, all the phonemes that occur in that language and to specify the
principles which determine their co-occurrence, or combination, in
actual and potential word-forms. He will tell us, for example, that the
combination of /b/ with /n/ is impossible in the first two positions of
English word-forms; and he may account for this in terms of the more
general principle that stop consonants do not precede nasal consonants
in English at the beginning of a syllable. Not only is there no actual
word-form which, if it did occur, might be written bait. The existence
of such a form (in any dialect or accent of English) is prohibited by the
phonological regularities of the language.
In contrast with such phonologically impossible forms as /bnit/, there
are very many forms whose non-occurrence in English is, from the
phonologist’s point of view, inexplicable: /blit/, /prek/, /stin/, etc. They
are potential word-forms of English that have not been actualized.

1 For the use of asterisks, see the list of Typographical Conventions, p. x.

 374 Semantics and grammar I
The phonologist must not only account for the phonological accept-
ability of the totality of potential word-forms in the language that he is
describing. He must also account for such prosodic features as stress
and intonation. Every sentence of English, if it is produced as a spoken
utterance, must be uttered with one of a limited set of stress-patterns
and intonation-patterns; and these patterns (as well as a variety of other
features that we have described as paralinguistic: 3.2) play an essential
part in the interpretation of spoken utterances in all languages.
Whether stress and intonation are more appropriately handled as part
of the structure of sentences or as part of another layer of structure that
is superimposed upon sentences in the course of their utterance is a
question that we need not go into here. Like most linguists we take the
view that at least some part of what is covered by the term ‘prosodic’
should be handled in describing the structure of sentences. Since sen-
tences are cited here in their standard orthographic form, which does
not allow for the representation of stress and intonation, it must be
constantly borne in mind that every sentence is assumed to have
associated with it an appropriate representation of its prosodic structure.
For convenience, and without making any attempt to justify this
terminological decision on theoretical or methodological grounds, we
will allow for the possibility that the same sentence may have several
different prosodic patterns superimposed upon it. If the reader prefers
to think of a set of different sentences, rather than a single sentence
associated with a set of distinct prosodic patterns superimposed upon it,
he is free to do so: none of the theoretical points made in this book rests
upon our adopting one view of sentences rather than the other.
It is more difficult to say what syntax is without getting involved in
irrelevant theoretical controversies than it is to give a rough-and-ready
account of what comes within the scope of phonology. The boundary
between syntax and semantics has long been, and remains, the subject
of dispute. It is interesting, in this connexion, to note that linguists have
never experienced the same kind of problem in drawing a distinction
between phonology and syntax. They have argued, at times, about the
necessity or possibility of describing the phonological structure of
utterances without reference to their syntactic structure or their mean-
ing. But the arguments have been very largely methodological; and the
adoption of one methodological position, rather than another, does not
radically affect our view of the scope of phonology. No linguist would
seriously maintain, for example, that such strings of forms as the mouses
has came (in an appropriate phonemic representation) are phonologically
 10.1. Levels of analysis 375
unacceptable in English. Each of the word-forms is an actual form of
English (cf. mouse’s, louses, louse’s); and there is no way of ruling out
this string in terms of permissible and impermissible combinations of
phonemes. The point is that the distinction between phonology and
syntax depends upon the acknowledged properties of duality* and
arbitrariness* that are found, to a greater or less degree, in all human
languages (cf. 3.4). We could, in principle, change the phonological
structure of every word-form in a language without affecting in any way
at all the distribution of the resultant word-forms throughout the sen-
tences of the language or the meaning of the sentences; and this is done,
commonly and successfully, for the written language, by means of
_ simple codes and ciphers based on the principle of substitution. What
cannot be done, it would appear, is to change the distribution* of all the
word-forms in a language whilst holding constant the meaning of the
lexemes of which they are forms or to change the meaning of the lexemes
without affecting the distribution of the associated word-forms (cf.
Householder, 1962).2 The theoretical conclusion to be drawn from this
fact is that there is an intrinsic connexion between the meaning of words
and their distribution; and it is for this reason that it is difficult to draw
the boundary between syntax and semantics.
But we have still not said what syntax is. Let us adopt, for the
moment, the following definition: by the syntax* of a language is to be
understood a set of rules which accounts for the distribution of word-
forms throughout the sentences of the language in terms of the per-
missible combinations of classes of word-forms. This definition, it will
be observed, does not say anything about the nature of the rules or
whether they make any appeal to the meaning of lexemes. These ques-
tions will be taken up later. For the present, it 1s sufficient to note that
a syntactically acceptable sentence is a string of word-forms which
satisfies the following two conditions: (1) that each of the word-forms is
a member of some form-class* ; (ii) that the word-forms occur in posi-
tions that are defined to be acceptable for the form-classes of which they
are members. Let us assume, for example: (1) that the is a member of the
form-class Article (Art), boy is a member of the form-class Singular
Noun (NSing), runs is a member of the form-class Present-Tense,
Third-Person Singular, Intransitive Verb (VIn3SingPres) and fast is a
member of the form-class Adverb of Manner (AdvMann); and that (11)
the syntactic rules of English define the string of form-classes
2 The distribution of a unit is the set of contexts in which it occurs throughout
the well-formed sentences of the language (cf. Lyons, 1968: 7off, 143ff).
 376 Semantics and grammar I
| Art-+NSing+ ViIn3SingPres-+ AdvMann
to be syntactically well-formed. If, and only if, these two conditions are
satisfied is The boy runs fast defined to be a syntactically acceptable
sentence of English.
Form-classes should not be confused with parts of speech: nouns,
verbs, adjectives, etc.3 The parts of speech are classes of lexemes (“boy’,
‘sing’, ‘pretty’, etc.), not classes of forms (boy, boys; sing, sings, sang,
sung; pretty, prettier, prettiest; etc.). What then, it may be asked, 1s the
relationship between the two kinds of classes? There is, unfortunately,
no standard and universally accepted answer to this question. Much will
depend upon whether the linguist who is describing English, or what-
ever the language happens to be, recognizes in addition to the levels of
syntax and phonology a level of morphology that serves as a bridge
between them. It 1s arguable that languages fall into different types
(isolating*, agglutinating*, fusional*, etc.: cf. 3.4); and that for certain
languages, though not for others, it is necessary to set up a separate level
of morphological analysis. But it is always possible to draw a theoretical
distinction between morphology and syntax, on the one hand, and
between morphology and phonology, on the other; and this is what we
will do here. This will enable us to discuss the relationship between
semantics and grammar in a relatively non-technical manner and with-
out prior commitment to any of the currently available theories of
grammar.
We have said that the syntax of a language is a set of rules which
accounts for the distribution of word-forms throughout the sentences of
a language; and we have seen that this definition presupposes the assign-
_ment of every word-form to one or more form-classes. How do we know
that runs, for example, is a member of the form-class Present Tense,
Third-Person Singular, Intransitive Verb? The form runs will not
appear in any conventional dictionary of English. What we will find is
an entry for the lexeme ‘run’, listed under the conventionally accepted
citation-form* run. Now it so happens that the citation-form of most
lexemes in English can also be regarded as the stem-form, to which
3 The traditional term ‘part-of-speech’ is not as widely employed nowadays
by linguists as it used to be, but the terms ‘form-class’ and ‘word-class’,
which are used in preference to it, are hardly more precisely defined in the
literature. The distinction that is drawn here between form-classes and
parts-of-speech would seem to be both useful and workable. For a useful
discussion of the issues involved cf. Matthews (1967) and other articles in
the same volume. The term ‘word-class’ is used, and discussed in relation to
‘part-of-speech’, by Robins (1971).
 10.1. Levels of analysis 377
various inflexional suffixes may be added (-s, -ed, -zng) to construct the
other forms of the same lexeme. In so far as this holds true, English is
what is called an agglutinating language. We can treat the word-form
runs as being composed, at the morphological level of analysis, of run-
and -s. Provided that the dictionary lists ‘run’ as an intransitive verb,
we can substitute for the form-class label VIn35ingPres the morpho-
syntactic* word [‘run’: 3SingPres]. (This is no more than an ad hoc
symbolic representation of the traditional formulation “third person
singular of the present (indicative) of (the verb) ‘run’’’.) We can then
take from the dictionary the stem run and (in default of any information
to the effect that the form of the third person singular of the present
indicative is morphologically anomalous) we can apply the morpho-
logical rule, which forms the third-person singular present-indicative of
all regular verbs, whether transitive or intransitive, by adding the sufhx
-s to the stem-form.
Readers familiar with current work in linguistics will appreciate that
the above account of the relationship between the syntax, the dictionary
and the morphology begs a number of important questions. In so far as
they are relevant to a general discussion of semantics they will be taken
up later. At this point, we are concerned solely with clarifying the terms
‘syntax’ and ‘morphology’ as they are used in this book. But before we
leave the subject of morphology, it is important to introduce and
exemplify the notion of suppletion*.
The term ‘suppletion’, as it is generally employed by linguists, pre-
supposes the existence of a certain number of regular patterns of forma-
tion with reference to which a minority of the forms are said to be
anomalous and suppletive. For example, better and worse are morpho-
logically anomalous with reference to taller, nicer, longer, etc. Generally
speaking, the comparative form of English adjectives results from the
addition of the suffix -er (and the superlative form from the addition of
the suffix -est) to a stem-form which is identical with the so-called
simple, or absolute, form. The comparative form of ‘good’, however, is
better, rather than gooder: i.e. the suppletive* form bett- replaces, or
does duty, as it were, for good-. A certain amount of suppletion is found
in many languages; and it is semantically relevant in so far as the ques-
tion whether such-and-such a morphologically irregular form is a form |
of lexeme X or lexeme Y must, in many cases, be decided in terms of
meaning. It is doubtful, for example, whether better could be assigned
to ‘good’ and worse to ‘bad’ except in terms of the semantic analysis of
better as ‘‘more good” and of worse as “‘more bad”’.
 378 Semantics and grammar I
: There is no human language, we can be sure, that is totally suppletive
in its morphological structure.* But it would be easy enough to construct
one. All we would have to do is to take the morpho-syntactic words of
English and the set of actual word-forms and to map the former set onto
the latter set at random. A moment’s reflexion will show that we could
still refer to each lexeme by means of a conventional citation-form (the
singular form for nouns, the infinitive form for verbs, etc.), but there
would be no reason to call the citation-form a stem or to think of it as
being part of, or as underlying, the other forms of the same lexeme. The
fact that there is no language that is totally suppletive in this way can no
doubt be accounted for in terms of some concept of semiotic efficiency.
It is easy to see that a language-system of this kind would be difficult
for children to learn and would impose a considerable burden upon
memory.’ But this does not justify our excluding in principle the
possibility of there being such a language-system; and contemplation of
this unactualized possibility reveals a little more clearly the nature of the
relationship between morphology and syntax in the language-systems
with which we are familiar.
There are many linguists nowadays who use the term ‘grammar’ to
subsume everything in language that is amenable to systematic descrip-
tion: i.e. phonology, morphology, syntax and semantics. Throughout
this book, however, it is being used in a much narrower sense: to cover
, only morphology and syntax. This is simply a matter of terminological
decision. But it implies that we can, in principle, distinguish grammatical
acceptability from both phonological and semantic acceptability; and it
is to the discussion of this question that we now turn.
4 As the terms ‘suppletion’ and ‘suppletive’ are normally employed, the phrase
‘totally suppletive’ might be held to involve a contradiction; but the pheno-
menon to which it is intended to refer is one that we would recognize without
difficulty if we came across it in our investigation of the world’s languages.
* There is some evidence to suggest that children do nevertheless begin by
treating their native language (English, Russian, etc.) as if it were totally
suppletive. It has often been pointed out that the analogical regularization of
such suppletive word-forms as went and mice (resulting in goed and mouses)
usually comes later in the acquisition of language than does the correct use of
the irregular suppletive forms themselves. This is plausibly interpreted as
implying that at first the child treats, not only go and went or mouse and mice,
but also open and opened or boy and boys, as if they were morphologically un-
related. As soon as the child realizes that opened is derivable from open, and
boys from boy, by means of a morpho-syntactic (or morpho-semantic) rule,
he will start to apply the same rule in the formation of goed, mouses, etc.
 Semantics and grammar I 379
10.2. Grammaticality
Acceptability, to which we have appealed in the previous section, is a
pre-theoretical notion: its theoretical correlate in the linguist’s model of
the language-system is well-formedness*. In what follows, we shall be
concerned primarily with grammatical and semantic well-formedness;
and we shall take for granted the notion of phonological well-formed-
ness. That there are sequences of English word-forms which constitute
grammatically and semantically acceptable utterances (when they have
superimposed upon them an appropriate prosodic pattern) is un-
controversial: He went to town to-day, I’m sorry I butted in when I did,
etc. That there are sequences of word-forms which are both grammatic-
ally and semantically unacceptable, or deviant, is also something that we
may take to be uncontroversial: Butted to when in did sorry he town, etc.
The question that we must now discuss is whether there 1s any difference
between grammatical and semantic acceptability; and, if so, how it 1s
established.
There is an obvious pre-theoretical correlate of semantic well-
formedness: namely, the intuitive notion of making sense or being
comprehensible. This intuitive notion stands in need of theoretical
explication. But it is an everyday notion; and we can at least start by
saying that a semantically acceptable utterance is one that native
speakers can interpret or understand. The term ‘grammatical’, how-
ever, is both in origin and in everyday usage a theoretical, rather than a
pre-theoretical, term. Furthermore, when the ordinary native speaker
of English is asked whether an utterance 1s grammatically acceptable or
not, he will usually decide the question by referring to some rule or
principle that he learned at school. He may tell us, for example, that
It’s me or I ain’t seen him is ungrammatical. In so far as the native
speaker’s judgements of grammatical acceptability are based upon a
normative tradition or indeed upon any rules that he has been taught at
school or elsewhere, they are both unreliable and irrelevant in the present |
connexion. What we are after is some intuitive notion of grammatical
acceptability which native speakers have by virtue of their recognition of
principles that are immanent in their own language-behaviour; and this
is something that we cannot get at directly by asking them whether a
putative sentence is or is not grammatical. Nor does there seem to be
any other way of formulating the question that is not open to similar
objections.
The best indication of grammatical unacceptability is what is some-
times referred to as corrigibility* : an ungrammatical utterance, according
380 Semantics and grammar I
to this criterion, is one that a native speaker can not only recognize
as unacceptable, but can also correct. For example, John rang up me
(with me unstressed) is unacceptable and corrigible as fohn rang me up;
that table who(m) we bought is corrigible as that table which we bought
The boy are here is corrigible as either The boys are here or The boy ts
here; and so on. But corrigibility is no more than an indication. It may
well be that the native speaker is basing his judgement of unaccept-
ability and his corrective procedure upon some normative rule that he
has been taught at school or elsewhere. For example, he might correct
It’s me to It ts I, or [ ain't seen him to I haven't seen him. However, it is in
principle decidable whether the native informant 1s operating with a
normative rule that is itself incorrect; and he may very well concede that
some people regularly say what, in his view or in terms of the gram-
matical principles that he learned at school, they should not say (e.g.,
It’s me or I ain’t seen him). We will have nothing more to say about the
very real methodological problems involved in testing for, and dis-
counting, the influence of normative grammar.
Corrigibility is of itself neither a necessary nor a sufficient condition
of grammatical unacceptability. It will fail to define as ungrammatical a
host of such sequences of word-forms as Butted to when in did sorry he
town. Sequences like this are unacceptable, but incorrigible: there are no
acceptable utterances with reference to which a native informant can
correct them. It is worth pointing out, however, that it is only since the
advent of generative grammar (which sets itself the goal of partitioning
all sequences of forms into two complementary subsets, the one identi-
fied as the set of sentences and the other as the set of non-sentences: cf.
10.3) that sequences such as Butted to when in did sorry he town have
been cited by linguists as obvious examples of ungrammatical, rather
than nonsensical or otherwise unacceptable, sequences of word-forms.® _
Since we are adopting the viewpoint of generative grammar, we will
treat such grossly incorrigible sequences as ungrammatical, noting only
that they are of little value for the purpose of exemplifying what is
meant by the native speaker’s intuitive notion of grammatical accept-
ability. Their ungrammaticality would be determined as such by any
generative grammar of English that succeeded in accounting satisfac-
torily for the unacceptability of sequences like that table who(m) we
bought in an auction or The boy are here.’ We need not be very concerned,
6 And generative grammarians have been criticized for this: cf. Bazell (1964).
7 Arguably, however, the selection of ‘which’, rather than ‘who’, like the
selection of ‘it’, rather than ‘he’ or ‘she’, is not a matter of grammar. Any
 10.2. Grammaticality 381
therefore, about the fact that corrigibility fails as a necessary condition
of ungrammaticality.
But corrigibility also fails as a sufficient condition of grammatical
unacceptability. We can discount the fact that many unacceptable, but
corrigible, utterances are to be classified as phonologically, rather than
grammatically, unacceptable, since there are other ways of determining,
pre-theoretically, the difference between phonological and grammatical
unacceptability: we can ask our informant whether it is our pronuncia-
tion of the forms in the utterance that is at fault, whether we have put
the stress in the right place, and so on. The problem lies rather in the
fact that it is often possible for a native speaker to correct an utterance,
especially if it is produced in some normal context of use, on the basis of
the syntagmatic relations that hold between lexemes (cf. 8.5). For
example, if a foreigner were to say Milk turns rotten very quickly in this
weather, he might well be understood as having intended to express the
proposition that is more normally expressed by saying Milk turns sour
very quickly in this weather. His utterance could therefore be corrected
by the substitution of ‘sour’ for ‘rotten’. Similarly, the phrase a flock of
elephants might be corrected to a herd of elephants (on the assumption
that the mistake that has been made resides in the selection of ‘flock’,
rather than in the selection of ‘elephant’). But it would be unusual to
classify either Milk turns rotten or a flock of elephants as ungrammatical.
Although corrigibility is perhaps the criterion which comes closest to
capturing our pre-theoretical notion of what constitutes an ungram-
matical utterance, it clearly requires supplementation or refinement. The
most obvious way of supplementing it is to consider what kind of
phenomena in languages are such that, in general, they are covered by
the criterion and then to define such phenomena to be a matter of
grammar if, and only if, they can be brought within the scope of statable
rules. Since it is not generally the case that the violation of the colloca-
tional restrictions that hold between particular lexemes (on the assump-
tion that there are such restrictions) results in the production of corri-
gible utterances, we can build this into our theoretical definition of
grammaticality; and this is what is done, in effect, by those linguists who
native speaker of English who believes that tables are persons (or is composing
a story in which he deliberately personifies the table to which he is referring)
will say, correctly, such things as the table who(m) we bought in an auction.
Corrigibility in cases like this is relative to our assumptions about the
speaker’s view of the world. There are presumably no circumstances under
which a native speaker of standard English could correctly say either The boy
are here or The boys ts here.
382 Semantics and grammar I
draw the boundary between grammaticality and other kinds of accept-
ability, as we shall do, at the point at which it is drawn in conventional
grammars of English and other languages.
On the assumption that such categories as nouns, verbs, adjectives,
etc., are syntactically justifiable categories in the analysis of particular
languages, we can say that it is the function of the syntactic rules in any
model of a language-system to account for well-formed combinations of
nouns, verbs, adjectives, etc., and furthermore to specify the morpho-
syntactic properties of any lexeme which occurs in any of these well-
formed combinations. That one lexeme rather than another can be, or
must be, selected in a given position in order to produce an acceptable
sentence is something that falls outside the scope of syntax. That the
lexeme must belong to a particular part of speech, however, does fall
within the scope of syntax; and the fact that the phonological realization
of a particular morpho-syntactic word is such-and-such a form falls
within the scope of morphology.
For example, the sentence ‘Milk turns rotten’ will be treated as a
grammatical sentence in our model of the language-system, on the
following grounds. There is a well-formed combination of categories,
which may be represented here as Nominal-++Copulative Verb+-
Adjective and which underlies one kind of copulative sentence in English
(cf. 12.2). The lexeme ‘milk’ is an uncountable noun, and may therefore
occur in a nominal, or noun-phrase, without any determiner; ‘turn’ is
one of the small class of copulative verbs that can occur in this con-
struction (cf. Quirk et al., 1972: 821); and ‘rotten’ is an adjective.
Furthermore, there is a particular kind of interdependence between the
subject nominal and the verb (traditionally called concord*, or agree-
ment) such that, in the present tense, if the subject is in the third-person
singular, one form of the verb is selected, and, if the subject is in the
plural, a different form of the verb is selected. The morphological rules
of English tell us that turns is the third-person singular, present-tense
form of ‘turn’: i.e. that it realizes one of the morpho-syntactic words
that the syntactic rules allow as possible in the construction Nominal--
Copulative Verb-+ Adjective. ‘Milk turns rotten’, like ‘Milk goes sour’
or ‘Mary fell sick’, but unlike ‘Milks turns rotten’ or ‘Milk turn sour’,
conforms to all the rules that we would incorporate in the grammatical
section of our model of the language-system; and it is thereby defined
to be grammatical, regardless of its corrigibility.®
8 The situation is not always as clear as it is in this instance. On the assumption
that ‘sincerity’ is listed in the lexicon as a noun which (like ‘ milk’) can occur
 10.2. Grammaticality 383
One objection to our implied definition of grammaticality 1s that it is
crucially dependent upon the possibility of drawing a clear-cut distinc-
tion between lexemes and morpho-syntactic words; and, it is well known
that there are many languages in the analysis of which no such distinc-
tion can be drawn (e.g., Classical Chinese and Modern Vietnamese).
This objection can be met by reformulating the distinction between
grammatical and collocational acceptability, as many linguists have
done, in terms of a distinction between open and closed classes of
morphemes*. (Morphemes, in the sense in which the term 1s being used
here, are minimal forms: i.e. forms that cannot be analysed into smaller
forms.) For example, the suffixes -s, -ed and -zng and the word-forms
the, my, from and or are morphemes that belong to closed classes of
small membership. Morphemes like boy, small and stop, on the other
hand, belong to open classes of large membership. Given that this dis-
tinction between open and closed classes of morphemes can be drawn,

in a nominal without a determiner, that ‘John’ is listed, or assumed to be, a

proper name (which can also occur without a determiner) and that ‘admire’
is listed as a transitive verb, but not as one that can take a de-sentential
transform as its subject (cf. 10.3), the grammar will admit as grammatical
such strings of forms as Sincerity admires Fohn and presumably Peter’s
sincerity admires Fohn (cf. Peter’s friend admires Fohn), whilst excluding, as
ungrammatical, Peter’s being sincere admires Fohn. This might seem to be
counter-intuitive, in that Peter’s sincerity admires Fohn and Peter’s being
sincere admires John are more or less equally incorrigible; and, at first sight at
least, they are equally uninterpretable. But Peter’s sincerity cannot but admire
- and, albeit grudgingly, admit that it admires — Fohn’s deviousness is surely
more acceptable than Peter’s being sincere cannot but admire . . . John’s being
devious (and far more acceptable than That Peter is sincere cannot but admire
... that Fohn is devious); and Sincerity always admires deviousness is both
acceptable and interpretable, whereas Being sincere admires being devious is
not. The grammatical status of strings like Sincerity admires John has been in
dispute among generative grammarians ever since Chomsky (1965) proposed
a formal technique for excluding them as ungrammatical (cf. 10.3, 10.5). As
we have just seen, the unacceptability of Sincerity admires Fohn (on the
assumption that it is unacceptable) cannot be accounted for by introducing
into the grammar a rule which prevents an abstract noun (like ‘ sincerity’)
from occurring as the head of the subject of the verb ‘admire’ (for this rule
will also prevent the grammar from generating the system-sentence ‘Sin-
cerity admires deviousness’, etc.). Furthermore, there is a pre-theoretical
difference, though it is not immediately obvious, between Sincerity admires
...and Being sincere admires ... This difference is readily accounted for in a
grammar of English which draws a distinction between various kinds of de-
sentential nominal transforms (cf, 10.3). It is not so readily accounted for, it
should be noted, in terms of purely semantic principles: there are many con-
texts in which ‘sincerity’ and ‘being sincere’ are intersubstitutable without
any evident change of meaning.
384 Semantics and grammar I
in languages for which it is impracticable or unprofitable to recognize a
separate morphological level of analysis, we can define lexical morphemes
to be minimal forms that belong to open classes and we can define
grammatical morphemes to be minimal forms that belong to closed
classes. We can account for the grammaticality or ungrammaticality of
utterances in terms of the combination of subclasses of lexical mor-
phemes with or without co-occurring grammatical morphemes. For
example, the grammaticality of The dog barked fiercely at the cat would
be accounted for in these terms with reference to a formula like the+
N-+V,-ed+A-ly+at+the+N (where N stands for the class of mor-
phemes containing dog and cat, V, stands for a subclass of verbal
morphemes containing bark and A stands for the class of adjectival
morphemes containing fierce). We need not go into the details of this
approach to the explication of grammaticality. It suffices for our pur-
poses to note that, although this morpheme-based approach may look
very different from the more traditional approach that we shall be
adopting, it yields much the same kind of results as far as English is
concerned.
Other criteria for the pre-theoretical determination of grammaticality
have been proposed by linguists. So far, however, there are no univer-
sally accepted criteria which can be used to supplement corrigibility;
and it is doubtful whether any such criterion will be found that will
enable us to decide for all sequences of forms whether they are or are not
grammatically acceptable. How then, it may be asked, do linguists
actually set about the task of writing grammars for particular languages?
One might of course assume that the linguist or his informant has some
unanalysable, but reliable, intuitive appreciation of what constitutes
grammatical acceptability, as something distinct and separable from
other kinds of acceptability. But there is nothing in the collective
experience of linguists that would justify this assumption.
The only satisfactory way of determining the limits of grammaticality
would seem to be the one that we are adopting; and it is arguable that
this is the solution to the problem that is implicit, though it is never —
made explicit, in traditional grammars of English and other languages.
Having postulated rules which will account satisfactorily for those cases
of grammatical and ungrammatical utterances that are pre-theoretically
decidable as such in terms of corrigibility, we can then let the grammar .
itself decide for us whether a given utterance is or is not grammatical,
according to whether it conforms to the rules that have already been
established. In other words, we formulate the rules of the grammar in
 10.2. Grammaticality 385
such a way that they admit all the sequences of forms that are clearly
acceptable (and thus define them to be grammatically well-formed) and
prohibit all the sequences that are not merely unacceptable, but gram-
matically unacceptable in terms of corrigibility (and thereby define them
to be grammatically ill-formed). Each of the strings that is pre-
theoretically indeterminate with respect to grammatical acceptability
will be defined to be grammatically well-formed or grammatically ill-
formed according to whether it is admitted by the rules that have been
established for the pre-theoretically determinate strings.°
What the grammarian tries to do when he is describing the language-
system employed in a particular speech-community is to bring the set of
system-sentences into correspondence with that subset of actual and
potential utterances which he considers to be grammatically acceptable
on pre-theoretical grounds. If he says, as he may do, that he is trying
to describe all and only the sentences of English (Chinese, Amharic,
Quechua, etc.), it must be constantly borne in mind that, in so far as the
expression ‘all and only the sentences of English’ refers to a set of text- |
sentences* (i.e. a subset of actual or potential utterances: cf. 1.6, 14.6),
this set may be, in part at least, pre-theoretically indeterminate.
The pre-theoretical status of the notion of grammaticality is one of
the fundamental issues upon which linguists, at the present time, are
divided. But we shall be operating throughout with illustrative utter-
ances whose pre-theoretical status with respect to grammatical accept-
ability can reasonably be assumed to be determinable. 'The question of
pre-theoretical indeterminacy, important though it is in any general
discussion of the nature of language, becomes critical only when a
linguist proposes to introduce into his model of the language-system
some rule which is designed to account for empirically questionable data
and is otherwise unnecessary.
One point that is implicit in our approach to the definition of gram-
maticality must however be made explicit and given due emphasis. If
we adopt the criterion of corrigibility (supplemented in the way that has
been suggested), we can immediately draw the conclusion that there will
be many sentences defined by the model to be grammatical which
correspond to nonsensical or otherwise uninterpretable utterances.
® This is what Chomsky (1957) originally proposed. It might be argued, of
course, that the linguist’s model of the language-system should not obliterate
(by generating what are thereby defined to be all and only the sentences of the
language) the distinction between what is determinate and what is indeter-
minate, Since we are not concerned with grammatical theory as such in this
book, we will not pursue this possibility.
386 Semantics and grammar I
Among them we would presumably find, for English, such classic
examples as the following: ‘Colourless green ideas sleep furiously’,
‘Quadruplicity drinks procrastination’, ‘Sincerity admires John’. That
the set of semantically well-formed sentences in any language constitutes
a proper subset of the set of grammatical sentences in that language is
something that most linguists and philosophers have generally taken for
granted. ‘I’his does not imply, however, that much of what we describe
as the grammatical structure of languages is not determined, in part at
least, by semantic principles. It seems quite clear, in fact, that this 1s so.
The parts-of-speech, and such grammatical categories as tense*, num-
ber*, gender* or mood*, are obviously associated with particular kinds
of semantic function (cf. 11.1); and differences of grammatical structure
can serve to express differences of meaning (cf. 10.4).1°

10.3. Generative grammar

The most important development in recent linguistic theory has been
Chomsky’s (1957, 1965) formulation of the principles of generative
grammar*
In essence, a generative grammar is simply a system of rules which
operates upon a non-empty set of elements and defines a non-empty
subset of the total set of possible combinations of the elements to be
grammatically well-formed and the complement of this subset to be
grammatically ill-formed.
Before we proceed, a number of points should be made about this
definition: (1) it does not say what the elements are or where they come
from; (11) it does not say whether the set of elements is finite or infinite;
(iii) it employs the term ‘combination’, rather than ‘set’, ‘string’, or
‘permutation’; (iv) it presupposes that every combination is either
grammatical or ungrammatical and that no combination is both or
neither; (v) it does not specify that the set of ungrammatical combina-
tions should be non-empty (i.e. it allows for the possibility that all the
combinations might be defined to be grammatically well-formed); (vi) it

10 The question of grammaticality has been intensively discussed by linguists in

recent years: cf. Al (1975), Bar-Hillel (1967b), Bazell (1964), Bolinger (1968),
Fromkin (1971), Haas (1973a), Hill (1961), Householder (1973), Hymes
(1971), Katz (1964), Lakoff (1971b, 1972), McCawley (1968, 1973), Sampson
(1975), Ziff (1964).
11 Of the many introductions to Chomskyan grammar currently available
Huddleston (1976) is perhaps the best. For the more technical details: cf.
Levelt (1974), Wall (1972). Chomsky’s 1955 manuscript has now been
published: cf. Chomsky (1976a).
 10.3. Generative grammar 387
sets no upper limit upon the number of elements in the well-formed or
ill-formed combinations; (vii) it does not restrict the applicability of the
notion of generative grammar to the description of human languages;
(vill) it does not say explicitly that the well-formed combinations are
intended to represent, or stand in any kind of correspondence to, utter-
ances or any other pre-theoretically determinable units of language-
behaviour. Our definition allows for the possibility that there may be
several different kinds of generative grammars. It is Chomsky’s major
contribution to linguistic theory to have made this point clear and to
have shown that certain kinds of generative grammars are more ade-
quate than others as models of language-systems.
Chomsky begins by assuming, as most linguists do, that in all human
languages there is a subset of actual and potential utterance-signals
which are grammatically acceptable and may be identified as text-
sentences (cf. 10.2). He makes the further assumption that spoken
utterance-signals can be regarded as strings* (i.e. sequences) of discrete
forms and that each form is identifiable as a token of a certain type in
terms of its phonological shape (cf. 1.4). This second assumption is not
essential to the notion of generative grammar. But it will simplify the
presentation if we do assume that (due allowance being made for certain
instances of overlapping and discontinuity which can be readily handled
as specifiable deviations from the norm) grammatical utterances are
composed of an integral number of forms, each of which is discrete and
of constant phonological shape.
For any language-system, L, the elements will be correlates of the
minimal forms that are identifiable as tokens of the same type in
utterances of L; and the well-formed combinations that are generated*
by the rules of the grammar will be strings of such elements. For
simplicity of exposition, we will assume, initially, that the minimal
forms are words (rather than morphemes) and that they are represented
as strings of phonemes; and we will temporarily abandon our notational
distinction between sentences and utterances (cf. 1.6). By convention,
we will use an orthographic, rather than a phonemic, representation: in
doing so we draw upon the principle of medium-transferability (cf. 1.4,
3.3). Each of the well-formed strings that is generated by the grammar
will be a system-sentence* of L; and it will be correlated with a text-
sentence* of L. The correspondence-relation between system-sentences
and text-sentences is something that we will discuss later (cf. 14.6). At
this point, however, we will operate with a purely intuitive and un-
defined notion of what constitutes a system-sentence. Roughly speaking,
 388 Semantics and grammar I
we can say that a system-sentence is what any conventional grammar
would regard as a sentence.
Ideally, no grammatically unacceptable string will be defined by the
grammar to be a text-sentence of L; and no pre-theoretically acceptable
text-sentence of L will fail to have its correlate in the set of system-
sentences. Needless to say, it is very difficult, and it may even be
- impossible, to realize this ideal in the description of any natural lan-
guage. However, to the degree that any generative grammar approxi-
: mates to this ideal (the generation of all and only the sentences of L:
cf. 10.2) it may be described, using Chomsky’s earlier term, as weakly
adequate.}
Since various kinds of generative grammars may be constructed, it is
in principle possible that one kind may be appropriate for the descrip-
tion of one class of languages and another kind for a different class of
languages. At the same time, it is reasonably clear that attested lan-
guages, although they vary considerably in their grammatical structure,
do not vary to such a degree that we cannot hope to describe them all
within the same kind of generative grammar. To say this of course is to
say nothing, until we start distinguishing various kinds of grammars by
building into them rather more specific properties than we have done so
far.
One such property is recursion*. In many languages, and possibly in
all, no upper limit can be set to the length of pre-theoretically acceptable
text-sentences, where length is measured in terms of the number of
constituent minimal forms. But it is not so much the length of sentences
that is at issue; it is rather the fact that what are traditionally called
phrases and clauses may be juxtaposed or embedded* one within
another within the same sentence. For example, there is no assignable
upper limit to the number of noun-phrases that may be conjoined* (i.e.
co-ordinated) within a larger noun-phrase in English: cf. the table (and)
the chairs (and) the pictures . . . and the piano. Or to the number of ad-
verbial phrases that may be adjoined* (i.e. attached in a particular way)
to noun-phrases to create a larger noun-phrase: cf. the book on the table
in the bed-room on the second floor of the house in the park . . . on the other
12 "This distinction between weak and strong adequacy is one that Chomsky
drew in his earliest work, pointing out that two grammars might be weakly,
but not strongly, equivalent, in that each might generate exactly the same set
of strings, without however associating with each string the same structural
description. In his later work, notably in Chomsky (1965), he draws a more
controversial three-way distinction between observational, descriptive and
explanatory adequacy. _
 10.3. Generative grammar 389
side of town. Similarly, there is no assignable upper limit to the con-
joining of clauses: cf. He came in (and he) took off his coat . . . (and he)
mixed himself a very dry martini and (he) settled himself down in his
favourite armchair. Or to the embedding of one relative clause within
another: cf. the dog that worried the cat that killed the rat . . . in the house
that Fack built. "These and many other constructions must be handled |
within any finite rule-system by means of recursive* rules: i.e. rules
which apply indefinitely many times to their own output.
Furthermore, it is the case, in certain languages at least, that there are
interdependencies that hold between non-adjacent forms and that these
non-adjacent interdependent forms may themselves be separated by a
phrase or clause containing another pair of non-adjacent interdependent
forms, and so on ad infinitum. ‘This means that the rules of the grammar
must make appeal to such units as phrases or clauses. We cannot, in
principle, categorize all the minimal forms and then generate all and
only the sentences of the language as unstructured linear sequences of
classes of minimal forms.
As Bloomfield (1935: 161) put it, in his discussion of the English
text-sentence Poor fohn ran away, any native speaker of English will
recognize intuitively that this utterance is composed of two syntactic
constituents, poor John and ran away, and that each of these constituents
is itself composed of two constituents poor and Fohn, on the one hand,
and ran and away, on the other. Furthermore, poor ohn is a con-
stituent of the same category — let us call it NP (noun-phrase) — as Fohn,
my friend, the Sultan’s favourite odalisk, he, the rat that killed the cat, etc.
And ran away is a constituent of the same category — let us call it VP
(verb-phrase) — as died, went to London, potsoned his{her mother-in-law,
etc. Each of these NPs and VPs is either simple, consisting of a single
minimal form, or complex. There is an infinite number of syntactically
and morphologically well-formed NPs and VPs in English; and the
grammar must generate each of these with its correct phrase-structure*
analysis in terms of the grouping of the forms within the NP or the VP
and the categorization of the groups and subgroups. |
13 An alternative to the use of recursive rules is the use of so-called rule-
schemata — e.g., NP — NP (and NP)” — which have the formal advantage, as
far as conjoining is concerned, that they generate truly co-ordinative struc-
tures without assigning to them excessive layering or internal bracketting.
Arguably, a strongly adequate generative grammar of English will contain
both recursive rules and rule-schemata. What ts said here about recursion as
a property of language-systems necessarily involves some simplification of the
more technical details.
390 Semantics and grammar I
Various notations have been used by linguists to represent the phrase-
structure of sentences. Nowadays tree-diagrams* are commonly used
for this purpose. For example, employing 5 to stand for ‘sentence’ and
using this symbol to label the root of the tree, we can represent the
phrase-structure, or constituent-structure, of The Sultan’s favourite
odalisk potsoned her mother-in-law by means of what Chomsky calls a
phrase-marker*, as in figure 8, which assigns to the sentence a particular
labelled bracketing.

DetYN 'a eo
' a o\
'
Det N V NP
! i
i''{!'t'‘‘ttt'i‘ (t3ttN ' t ! '
|
I‘t‘t‘it$;’t
NP Poss Adj N Det N
{

the Sultan s favourite odalisk poisoned her mother-in-law

Figure 8. A sample phrase-marker

There is much in the proposed analysis of this sentence that 1s

questionable. But we are making use of it solely for the purpose of
illustration.4 Most linguists recognize the validity of the notion of
phrase-structure. One cogent reason for doing so is that many sentences
of English and other languages are interpreted differently according to
whether one phrase-structure analysis rather than another is assigned to
the string of forms: cf. He arrived late last night (“It was last night that
he arrived late” vs. ‘It was late last night that he arrived’’); Tom or
Dick and Harry will go (“Either Tom will go or Dick and Harry will
go” ws. “Either Tom or Dick will go and Harry will go”). Another
reason is that often, though not always, there are prosodic differences
144 All that needs to be said about figure 8 is that Det (for ‘determiner’) is a
category that includes, but is wider than, the class of articles (cf. 11.4). The
phrase the Sultan’s is given the same classification as her because they have
the same distribution, even though they are shown as differing in their
internal structure. It is of course arguable that her is the morphological
realization of ‘she’+Poss (where Poss stands for the possessive morpheme
and ‘she’, like ‘he’ or ‘I’, and ‘the Sultan’, are NPs). We are not concerned
with such details; nor, at this stage, with the more important fact that
Adj+N has been categorized as N.
 10.3. Generative grammar 391
which correlate with differences of phrase-structure. The utterance He
arrived late last night would not normally be ambiguous in spoken
English, since the grouping of Jate with either arrived or last night would
normally be clear from the intonation and stress. On the other hand, an
utterance like He hit the man with a stick is not necessarily disambiguated,
in normal conditions, by means of prosodic signal-information, though
it may be so disambiguated on occasion. That differences of phrase-
structure should correlate, at least partly, with differences of meaning
and with prosodic differences in the vocal-signal is something that is
perhaps only to be expected. It is important to realize, however, that the
notion of phrase-structure can be justified independently on purely
syntactic and morphological grounds.
One semantically important principle associated with the notion of
phrase-structure is endocentricity*. A phrase is said to be endocentric
if it 1s syntactically equivalent to one of its immediate constituents. (It
follows from this definition that endocentric constructions, under the
strictest interpretation of distributional equivalence, are necessarily
recursive: the combination XY will not be distributionally equivalent
to X unless Y can be combined with XY.) For example, NPs containing
an embedded relative clause are endocentric: cf. the man who came to tea,
which has the same distribution throughout the well-formed. sentences
of English as the man. Similarly, favourite odalisk is syntactically equiva-
lent to odalisk; and it is for this reason that it was classified as N in
figure 8. Both of these examples illustrate the kind of endocentricity that
is referred to as subordinative*, in which the endocentric phrase is
composed of a head* and a modifier*: the head is syntactically equiva-
lent to the whole phrase, and the modifier is the syntactically subordinate
constituent which modifies*, or qualifies*, the head.
Co-ordinative* phrases are also classified in the standard treatments
as endocentric, their characteristic property being that they have more
than one head and no modifier. But it must be recognized that what is
generally described as co-ordination covers a number of distinguishable
syntactic processes. Disjunction and conjunction operate differently in
English with respect to the determination of number-concord ( fohn and
Mary are coming vs. John or Mary is coming); a co-ordinate noun-phrase
must sometimes be interpreted collectively, sometimes distributively
and sometimes reciprocally (John and Mary are a lucky couple vs.
John and Mary are happy vs. fohn and Mary are similar); and there
are various other factors which affect the syntactic subclassification
of co-ordinative phrases and their distributional equivalence, or
392 Semantics and grammar I
non-equivalence, with one or both of their conjoined constituents. We
shall be primarily concerned with subordinative endocentric phrases
in any appeal that we make subsequently to the notion of endocentricity.
Any phrase that 1s not endocentric is by definition exocentric*.
Obvious examples in English are adverbial phrases composed of a
preposition and a noun-phrase, such as at the house, which is syntactic-
ally equivalent to locative* adverbs (adverbs of place) like here or there;
or adverbial clauses, such as when he got home, which are again syn-
tactically equivalent to adverbs or adverbial phrases like then or at that
time. Using X, Y and Z as variables which take as their values particular
syntactic categories and a dot to indicate the combination of one category
with another (regardless of the relative order or contiguity of the two
categories), we can symbolize endocentricity by means of the formula
X.Y = X, and exocentricity by means of the formula X.Y = Z (where
Z# Xand Z + Y, but the identity or non-identity of X and Y is left
undetermined). Noun-phrases in English, such as the boy or my friend,
are generally regarded as endocentric, the noun being taken as the head
and the article or determiner as the modifier. This is obviously in-
correct, as far as countable common nouns such as boy or friend are
concerned: the boy or my friend are not syntactically equivalent to boy or
friend (i.e. they are not intersubstitutable throughout the grammatically
well-formed sentences of English). Such noun-phrases as the boy or my
friend are distributionally equivalent to proper names and personal
pronouns: we will return to this point later (11.2).
Our account of endocentricity, and indeed our whole treatment of
generative grammar so far, has been rendered less precise than it might
otherwise have been by our failure to take into consideration the possi-
bility of distinguishing between forms, lexemes and expressions (cf.
1.5). Linguists in the Bloomfieldian tradition (including Chomsky) have
operated with forms and categories of forms; and they have, at most,
drawn a distinction between grammatical forms (like zf, he, the, etc.) and
lexical forms (like boy, beautiful, arrive, etc.). It is obvious, however,
that a generative grammar could be constructed in which categories like
N, V or Adj (adjective) are categories of lexemes and the so-called lexical
forms are introduced into sentences after the operation of all the syn-
tactic rules, but prior to the operation of the rules of the morphological
component. Since it is our intention, as far as possible, to make our
treatment of semantics independent of any particular system of genera-
tive grammar, we will not go into this question. We must, however, say
something about expressions. For it is to expressions, rather than to
 10.3. Generative grammar 393
either forms or lexemes, that we have assigned the semantic functions of
reference and predication; and it is arguable that, in so far as it is
semantically relevant, endocentricity also applies to expressions.
Consideration of a system-sentence like ‘Tom and Dick swim
beautifully and so does Harry’ shows that its interpretation clearly
depends upon the possibility of identifying the predicate of the first
clause with the predicate of the second clause. Assuming that the sen-
tence is uttered to make a true statement, we can infer from it three
distinct true propositions: that Tom swims beautifully, that Dick swims
beautifully, that Harry swims beautifully. One way of analysing the
sentence grammatically is to say that there is a rule of English syntax
which enables us to substitute so does Harry for Harry swims beautifully
under some condition of identity between what is realized as the form
swims beautifully in the second clause and as the form swim beautifully
in the first clause. But this is clearly not a condition of identity between
the actual forms (cf. swim vs. swims). What we want to be able to say is
that there is some syntactically identifiable unit, ‘swim beautifully’,
which is constant under its various realizations and that it is this unit —
or expression — which has predicative function in both clauses. Some of
the morpho-syntactic properties that are realized in the forms of this
syntactic unit are irrelevant to its status as an expression (notably its
being a singular or a plural form) and are determined by the rules of
concord and government. We also want to say, to take another example,
that the same expression occurs as subject, object and indirect object in
the following three Latin sentences: ‘Amicus meus mortuus est’ (“My
friend is dead’’), ‘Meum amicum interfecit’ (“He has killed my
friend’’), ‘Meo amico librum dedi’ (“I gave the book to my friend’’).
The fact that ‘meus amicus’ occurs in the nominative, accusative or
dative case and that these are all realized by different forms is from this |
point of view irrelevant.
Expressions such as ‘swim beautifully’ (and possibly ‘meus amicus’)
are endocentric expressions; and the endocentricity of the form swims
beautifully is a consequence, but not a necessary consequence, of the
endocentricity of the underlying expression.* If it happened to be a
rule of English that the third-person present-tense form of ‘swim’ was
does swim whenever it was modified by an adverb (so that ‘John swims’

18 Whether ‘swims beautifully’ is truly endocentric or not is perhaps debatable.

It is not clear, however, that there is an absolute prohibition upon the addi-
tion of a further subordinated adverbial modifier: cf. They all swim beauti-
fully, but some swim beautifully effortlessly and others do so only with difficulty.
 394 Semantics and grammar I
and ‘John does swim beautifully’ were well-formed sentences, but not
‘John does swim’ nor ‘John swims beautifully’) szzms and does swim, as
forms, would differ in their distribution throughout the sentences of
English; and yet we should still wish to say that the predicative expres-
sion ‘swim beautifully’ was endocentric and contained the lexeme
“swim ’ as its head. This means that our grammar of English must either
distinguish between a syntactic and a morphological level of representa-
tion or must recognize, within syntax, between deeper and more super-
ficial layers of structure; and it may indeed do both.
Generative grammars fall into several types: phrase-structure gram-
mars, dependency grammars, categorial grammars, etc. We shall not
discuss the differences between them or their relative advantages and
disadvantages. More important, for our purposes, is the difference
between grammars which distinguish deep structure* from surface
structure* and those which do not. We will refer to the former class of
grammars, in a somewhat loose use of the term, as'transformational* and
the latter class as non-transformational*; and we will say that it is
characteristic of a transformational grammar that it generates sentences
in two stages. First-stage rules we will call base-rules*; and second-
stage rules we will describe as transformational*. Many systems of
syntactic analysis, some of them radically different from Chomsky’s
(1965) system in other respects, can be brought within the scope of this
very broad notion of transformational grammar. What. unites them is
their acceptance of the view that superficially distinct sentences and
phrases may be derived from the same underlying structure and con-
versely that superficially identical sentences and phrases may be derived
from distinct underlying structures; and this is, up to a point, a view
that many traditional grammarians have held.
For example, corresponding active and passive sentences in English,
such as ‘The guerillas tortured the prisoner’ and ‘The prisoner was
tortured by the guerillas’, differ considerably in their surface structure.
It is when we come to account for the nature of the correspondence that
| holds between these two sentences (and indefinitely many pairs of other
sentences in English) that we invoke the notion of deep structure; and
it is here that we find the greatest difference between transformational
and non-transformational grammars, on the one hand, and also between
various kinds of transformational grammars, on the other. A trans-
formational grammar (in the deliberately general and rather loose sense
that we have given to this term here) would either derive one of these
sentences from the other (and presumably the passive from the active)
 10.3. Generative grammar 395
or derive both from wholly or largely identical underlying pre-sentential
structures of a more abstract kind. In Chomsky’s (1965) system, the deep
structure of a sentence is explicitly defined (for all sentences) as a pre-
sentential phrase-marker (i.e. a tree-structure with the root labelled as S
and the nodes labelled with such category-symbols as NP, VP, N, V,
etc.); and the surface structure of the sentence is derived from this by
the application, in sequence, of a set of transformational rules, each of
which has (in principle) some precisely defined effect upon the phrase-
marker. In Harris’s (1968) system, a distinction is drawn between
kernel* sentences, derived by base-rules, and non-kernel* sentences,
derived from kernel sentences by the operation of transformational
rules. The notion of deep structure is clearly applicable to non-kernel
sentences in Harris’s system, and to a variety of clauses and phrases that
are derived from kernel sentences by transformational rules; and we can
generalize the notion of deep structure, for our purposes, by defining the
deep structure of one of Harris’s kernel sentences to be identical with its
| surface structure. Other linguists operate with, and have made more or
less precise, a notion of deep structure (whether they use the term or
not), according to which the deep structure of a sentence is a path
through a network of syntactic choices; and the surface structure is
derived from this complex of selected syntactic features by rules that we
shall again describe, loosely, as transformational. The notion of deep
structure has played a prominent part in recent discussions of the rela-
tion between syntax and semantics.
Traditional grammar can also be seen as a system of grammatical
analysis within which we can formalize, in principle, the traditional dis-
tinction between simple sentences, on the one hand, and compound and
complex sentences, on the other. But the traditional distinction also
rests upon the prior distinction of clauses and phrases. This is not
always pre-theoretically sharp, even in English; and it is still less so in
many other languages. We will not therefore operate, except informally
in referring to illustrative sentences and parts of sentences, with this
distinction.
Let us introduce instead the theoretically neutral concept of a
desentential transform*. This is not neutral, of course, in so far as it
presupposes a commitment to some kind of transformational syntax.
But it is intended to be neutral with respect to the distinction of clauses
and phrases (in the traditional sense of this distinction or more recent
formulations of it); and also with respect to the interpretation of S as a
_ sentence or a pre-sentential structure. A desentential transform will be
396 Semantics and grammar I
any syntactically determined constituent of a sentence that is derived
from an instance of 5 in the deep structure of the sentence by means of
one or more transformational rules.
Among the transformations to which we shall be appealing in the
course of this work, and always informally, there are two that may be
mentioned here: nominalizations* and adjectivalizations*. By ‘nomina-
lization’ is to be understood the transformation of an S into an NP; this
NP may be either a clause or a phrase, and it may, in certain circum-
stances, be a single word. Examples of nominalized transforms, or
nominalizations, are such expressions as ‘the killing of Sister George’ ,
and ‘that the moon is made of blue cheese’. The term ‘adjectivalization’
is less commonly employed than ‘nominalization’ in current versions of
transformational grammar. We will use it to refer to the process whereby
attributive adjectives and adjectival phrases and clauses (including
relative clauses) are derived from a variety of predicative structures.
Examples of constructions which involve adjectivalization are ‘barking
dog’ (interpreted as “dog which barks/is barking’), ‘book on the
table’, ‘girl with green eyes’, ‘friend of my father’, ‘man who came to
tea’.
This completes our account of generative grammar. We have deliber-
ately adopted as neutral a position as possible on several controversial
issues. In the present state of grammatical theory, it would be unwise to
do otherwise; and it is our belief that much can be said about the
semantic structure of sentences and utterances without commitment to
one system of grammatical analysis rather than another. At the same
time, it must be emphasized that there are many aspects of the inter-
dependence of grammatical and semantic structure which cannot be
treated precisely except within the framework of some general theory of
the whole structure of the language-system and an explicit and detailed
account of the syntactic relations between different sentences. There are
certain important topics that have been extensively discussed recently
by linguists which we shall be obliged to deal with only cursorily and
very informally. It 1s hoped that sufficient background information has
been given in this section for the non-linguist to be able to appreciate the
significance of most of the points that will be made.

10.4. Grammatical ambiguity

It is a universally recognized and demonstrable fact that many of the
acceptable utterances of English and other languages are ambiguous*:
they can be interpreted in two or more different ways. Frequently,
 10.4. Grammatical ambiguity 397
though not always, their ambiguity passes unnoticed in everyday
language-behaviour, because the context is such that all but one of the
possible interpretations are irrelevant or relatively improbable. For
example, the following utterance-signal
(1) They passed the port at midnight
has at least two distinct interpretations, according to whether the form
port is taken to be a form of the lexeme ‘port,’ meaning “‘harbour”’ or
of the lexeme ‘port,’ which denotes a certain kind of fortified wine.
Which of these two interpretations is intended by the speaker would
generally be clear from the context in which the utterance occurs. The
utterance-signal itself, however, is inherently ambiguous; and the
linguist must describe it as such.
The reader will have noted that (1) has been referred to as an am-
biguous utterance-signal. We have been careful, at this point, not to
classify it as an ambiguous sentence. As the term ‘sentence’ is tra-
ditionally employed, two utterances will count as utterances of the same
sentence if and only if (1) they are identical at the grammatical and
phonological (or orthographic) levels of representation and (ii) the forms
of which they are composed are forms of the same lexemes. On the
assumption that ‘port,’ and ‘port,’ are distinct lexemes, the utterance
They passed the port at midnight would be associated with (at least) two
different sentences of English; and this would be so, in terms of the
traditional definition of ‘sentence’, regardless of whether the two lexic-
ally distinct, but phonologically identical, strings of forms have assigned
to them the same grammatical structure or not. It follows from this way
of looking at the relationship between utterances and sentences (which
is consistent with the view that is taken throughout this book) that much
of the ambiguity of utterances is to be accounted for by putting them
into correspondence with what the grammar and the lexicon will jointly
define to be distinct, though perhaps related, sentences.
Having made this point, we must immediately qualify it by saying
that, as the term ‘sentence’ is normally defined in generative grammar,
the system-sentence correlated with They passed the port at midnight
would normally be described as a single sentence with two (or more)
meanings: i.e. as an ambiguous sentence. This usage has certain practical
conveniences; and we will adopt it from now on in referring to various
examples of ambiguous strings. Apart from anything else, the term
‘sentence’ is less cumbersome than ‘grammatically well-formed string
of minimal forms’ or whatever alternative term we might use in order
398 Semantics and grammar I
to avoid prejudging, in every instance, the vexed question of homonymy
and polysemy (cf. 13.4).
We must begin by distinguishing linguistic from non-linguistic
ambiguity. T'wo spoken utterances are linguistically ambiguous if their
ambiguity is such that it can be explicated in terms of identity of repre-
sentation at some level of analysis in the correlated system-sentence.
Linguistic ambiguity depends solely upon the structure of the language-
system, whereas other kinds of ambiguity, actual or potential, are to be
accounted for in other ways. For example, the linguist will not be con-
cerned, in general, with the referential ambiguity of proper names,
personal and demonstrative pronouns, or definite descriptions (cf. 7.2):
e.g., with the fact that the expression ‘they’ in They passed the port at
midnight might refer to indefinitely many different groups of people.
Referential ambiguity is held to be linguistic only in so far as it depends ,
upon distinctions (e.g., reflexive vs. non-reflexive) which are gram-
maticalized in the language-system: for example, the sentence “John
Smith thinks that he has failed the examination’ might well be described
as linguistically ambiguous according to whether ‘he’ is construed as
being reflexive or not (cf. 15.4). Also classified as non-linguistic are
ambiguities that are introduced into utterance-signals by channel-noise
(cf. 2.2), by deficiencies in the language-user’s competence or per-
formance (cf. 14.2) or by the particular contexts in which the utterances
occur (cf. 14.6).
We have already seen that what we have classified as prosodic signal-
information is generally held to fall within the scope of linguistic
analysis. Forms may be systematically distinguished, one from another,
not only in terms of their phonemic composition, but also in terms of
some associated prosodic feature. There are many languages in which
different forms of the same lexeme are distinguished prosodically; and
there are languages in which forms of quite unrelated lexemes are dis-
tinguished by stress or tone. Word-forms, then, are not just strings of
phonemes: they are strings of phonemes upon which there may be
superimposed various kinds of prosodic (or suprasegmental*) features.
Not only word-forms, but also phrases, may be distinguished, as forms,
by means of prosodic features; and this is much more commonly the
case throughout the languages of the world.
16 Phonological identity, it must be emphasized, is not a pre-theoretical notion,
though it is constrained to some considerable extent by the intuitive concept
of type-token identity with which native speakers operate in deciding whether
a proposition like ‘‘John said X and so did Mary’’ is true or false (cf. 16.1).
 10.4. Grammatical ambiguity 399
Prosodic features also have another kind of function, which is usually
dealt with by linguists in terms of the stress-patterns and intonation-
patterns that are associated with the utterance as a whole, rather than
with the particular forms of which it is composed. For example, the
following utterance
(2) I’ve seen hér, not him
contains two heavily stressed forms, hér and him. But the heavily
stressed her or him are taken to be the same forms as the unstressed her
or him. Since the same sentence may have superimposed upon it several
different prosodic patterns (cf. 10.1), it follows that two or more |
prosodically distinct (and therefore non-ambiguous) utterances may be
mapped on to what we will refer to a single sentence with two or more
meanings. This may be regarded as simply a matter of terminological
convenience without any implications for the important question as to
how or whether stress-patterns and intonation-patterns are to be
accounted for in a sentence-generating grammar.
Various kinds of junctural* phenomena may occur at the boundaries
between forms. For example, the elision* of the final vowel of the forms
le and Ja in French means that an utterance like Fe l’arme beaucoup (“I
like him/her very much’’) may be analysed either as je-+-/e-+-atme+
beaucoup or je+-la+aime-+ beaucoup. Elision, of itself, does not con-
stitute a serious theoretical problem. Nor does the converse pheno-
menon, which is traditionally described as liaison* in French.
Theoretically more troublesome than elision and liaison are junctural
features which serve, optionally or obligatorily, to indicate the boun-
daries between contiguous forms in utterances. For example, such pairs
of phrases as an aim and a name, an ice-bucket and a nice bucket, the grey
tape and the great ape, and many other pairs of complex forms in
English can be distinguished, in fairly slow and careful speech, at
least, by certain transitional features, whose phonetic status is somewhat
problematical, but which are nonetheless systematic and perceptible.
The same is true of pairs of French forms like qu’al aime (que+il-+ aime,
‘““whom he loves’’) and qui l’atme (qui+le/la+-aime, “who loves
Phonological identity may be defined differently in different theories. For
simplicity, we are operating throughout this book with the assumption that
the minimal units of phonological structure are phonemes and that all forms
are to be represented in the linguist’s model of the language-system as strings
of phonemes. In doing so we are begging certain questions that are of central
importance for the phonologist. None of these, however, is crucial in the
present connexion.
 400 Semantics and grammar I
him/her’’). Junctural phenomena of this kind may serve to disambi-
guate many utterances that would be classified as ambiguous according
to current versions of generative grammar; and they correlate highly,
not just with the boundaries between simple forms, as our examples so
far might suggest, but with higher-level syntactic boundaries. In what
follows, we will discount the optional junctural features by which
utterance-tokens may, on occasion, be distinguished. We do so, how-
ever, without prejudice to the question whether they are to be mapped
on to one or several sentences in the language-system; and from now on
we will refer, on occasion, to sentences, rather than to utterances.
We are now in a position to define and to discuss grammatical
ambiguity*. A grammatically ambiguous sentence is any sentence to
which there is assigned (by a generative grammar of the language-
system) more than one structural analysis at the grammatical level of
analysis. Three points should be made immediately about our definition
of grammatical ambiguity. The first of these is that not every gram-
matically ambiguous sentence will in fact be interpretable in more than
one way. In this respect, grammatical ambiguity is like lexical ambiguity.

tence .
For example, just as the lexically ambiguous sentence ‘They drank the
port at midnight’ is presumably not interpretable as containing ‘port,’,
on any occasion of its utterance, so the grammatically ambiguous sen-

(3) He shot the man with a stick

would not normally be interpreted as meaning ‘‘He used a stick to
shoot the man’’. The second point is that the definition of grammatical
ambiguity that we have given makes it dependent upon some particular
grammatical model of the language-system: it follows that there might
be sentences which in terms of one analysis of a language-system are
grammatically ambiguous, but which would not be so described with
reference to a different analysis of the same language-system. The third
point to which the reader’s attention is drawn is that there is nothing in
our definition of grammatical ambiguity which excludes the possibility
that a sentence may be both lexically and grammatically ambiguous.
That this is more than just a possibility will be demonstrated in the
course of the discussion.
The least controversial kind of grammatical ambiguity is that which
| can be explicated in terms of phrase-structure (with or without an
. associated difference of stress, intonation or juncture). This has already
been illustrated by means of the following utterances (cf. 10.3):
 10.4. Grammatical ambiguity 401
(4) He arrived late last night
(5) Tom or Dick and Harry will go
(6) He hit the man with a stick.

To these we may add:

(7) You can’t get fresh fruit and vegetables these days
(8) They are eating apples.
That these utterances each have at least two interpretations is obvious
enough. It is also clear that their ambiguity does not depend upon
homonymy or polysemy, but is of the kind that is naturally accounted
for in terms of the notion of phrase-structure. Any reasonably compre-
hensive generative grammar of English would automatically assign to
the system-sentences in correspondence with these utterances, and to
indefinitely many sentences like them, at least two different phrase-
structure analyses. Their grammatical structure is to this extent un-
controversial.
Not all aspects of the phrase-structure of (4)-(8) are uncontroversial.
Phrase-structure, in certain systems of formalization at least, involves
both grouping and categorization (cf. 10.3). There is no problem, from
this point of view, in either (5) or (7), which demonstrate the fact that
grammatical ambiguity may, in certain cases at least, be determined
solely by differences in the way in which the forms are grouped to-
gether. But (6) is not like (5) and (7) in this respect. Under one inter-
pretation of (6) ‘with a stick’ is an adjectival phrase modifying the noun
‘man’. In the other interpretation, it is what would be traditionally
described as an adverbial complement of the verb-phrase. In both cases,
however, the phrase ‘with a stick’ is composed of a preposition and a
noun-phrase. The question is, therefore, whether the phrase should be
categorized differently in the two phrase-markers in the one case as an
adjectival phrase and in the other as an adverbial phrase. If this is done,
then (6) will be grammatically ambiguous in terms both of grouping and
of categorization. Furthermore, although we have assumed that (6) is
not lexically ambiguous, it is obviously arguable that the preposition
‘with’ differs in meaning in the two cases. We do not propose to argue
that ‘with a stick’ is or is not ambiguous in terms of its syntactic
categorization or in terms of the meaning of ‘with’. The point that is
being emphasized here is that grammatical ambiguity is at least partly
dependent upon the way in which the language-system is analysed; and
this is especially so in respect of the categorization of constituents in the
402 Semantics and grammar I

about (6). :
phrase-markers assigned by a generative grammar. Much the same
comments can be made about (4) and (8) as have just been made

This is a point that is often not mentioned at all in discussions of

certain kinds of grammatical ambiguity. One of the most famous of
Chomsky’s (1957) examples, which is often quoted in support of the
distinction between deep structure and surface structure, Is
(9) Flying planes can be dangerous.
This sentence, and others like it, 1s commonly discussed under the tacit
assumption that it has a unique surface structure and that its ambiguity
can be accounted for only in terms of a difference in the deep structure
(cf. 10.3). It is not obvious, however, that the ambiguity cannot be
handled in terms of a difference in the labelling of ‘flying planes’ in (9).
Under one interpretation, the form flying is a participle with adjectival
function; under another interpretation, it is what would be traditionally
described as a gerund. Since the distributions of participles and gerunds
throughout the sentences of English overlap, but are not identical, they
might very well be labelled differently in a generative grammar of the
language. Furthermore, the ambiguity of (9) depends crucially upon
two other factors. Modal verbs (like ‘can’, ‘may’, ‘must’, etc.) are not
subject to singular/plural concord in English. Hence the fact that under
one interpretation ‘flying planes’ is a singular NP and under the other
interpretation it is a plural NP is not reflected in the form can: this is
nonetheless a difference which is highly relevant to the distribution of
the phrase ‘flying planes’ in very many sentences (cf. “Flying planes
is/are dangerous’; “Flying planes can be dangerous, can’t they?’ vs.
‘Flying planes can be dangerous, can’t it?’). The other factor that is
responsible for the ambiguity of (9) is the possibility of using the verb
‘fly’ either transitively or intransitively. If the distinction between
transitive and intransitive verbs is taken to be criterial for their classifi-
cation as distinct lexemes (and this is perhaps arguable), then (9) would
be held to exhibit lexical as well as grammatical ambiguity: for ‘fly,’
and ‘fly,’ would then be partial homonyms. The sentence would also be
lexically ambiguous if ‘fly’ were classified in the lexicon as a single
polysemous verb with two meanings (cf. 13.1).
We will here assume that (9) is indeed grammatically ambiguous
(whether or not it is also lexically ambiguous); and that furthermore it is
transformationally ambiguous. It does not follow, however, that,
because (9) is transformationally ambiguous, it is not also grammatically
 10.4. Grammatical ambiguity 403
ambiguous in terms of surface structure (or lexically ambiguous in
terms of homonymy or polysemy). For surface-structure identity in-
volves both grouping and categorization. The surface structure of a
sentence is not something that can be determined by inspection, without
reference to the rules and grammatical categories of some particular
model of the language-system. This is even more obviously the case, of
course, in relation to the deep structure of a sentence.
Transformational ambiguity, then, neither excludes nor implies
surface-structure ambiguity. A sentence is transformationally ambiguous
(with respect to a given transformational grammar) if and only if it is
derived from two or more distinct underlying structures; and we can
reasonably assume that this will be so in the case of (9g). The syntactic
processes of adjectivalization and nominalization by means of which the
phrase ‘flying planes’ can be derived from the sentences ‘Planes fly’
and ‘Someone flies planes’, or (in Chomsky’s theory of transformational
grammar) from the structures underlying these sentences, are very
general, and possibly universal, processes in language, which can be
justified independently of semantic considerations. The transforma-
tional account of the ambiguity of a phrase like ‘flying planes’ is
semantically attractive, however, because the transformational rules
relate such sentences as (9g) to several non-ambiguous sentences, each of
which can be said to paraphrase (9) under a particular interpretation:
(10) Planes which are flying can be dangerous
(11) To fly planes can be dangerous.
There are several additional possible underlying structures, accounting
for other interpretations of (9): cf. ‘Planes for flying can be dangerous’.
But this does not affect the main point that is being made here.
It is not essential to the notion of transformational ambiguity that
there should, in all instances, be non-ambiguous transforms of the same
underlying structures. The transformational explication of ambiguity
would of course lose much of its force if it turned out to be the case that,
in a significantly large number of instances, what the grammatical rules
define to be differences in the deep structure of sentences could not be
correlated with different interpretations in terms of semantically non-
equivalent and non-ambiguous transforms of the several underlying
structures. But transformational grammar, as such, does not stand or
fall according to its capacity to handle ambiguity or the semantic struc-
ture of sentences in general.
What has just been said is in conflict with one of the fundamental
404 Semantics and grammar I
principles that has inspired much of the more recent work in generative
grammar: the principle that a sentence has exactly as many distinct
interpretations as it has deep-structure analyses (cf. 10.5). But this
principle rests upon several questionable assumptions.
One of these is that every sentence of a language does in fact have an
empirically determinable number of interpretations. None of the tests
for ambiguity that have been proposed so far by linguists lends much
support to this assumption (cf. Zwicky & Sadock, 1975). Two possible
tests for ambiguity may be ruled out immediately: translation and para-
phrase. Neither of these techniques is of itself sufficient to distinguish
ambiguity from generality of sense. The fact that ‘brother-in-law’ can
be translated into Russian by any one of four non-synonymous lexemes
or paraphrased in English as ‘wife’s brother’, “husband’s brother’,
‘sister’s husband’, etc., does not prove that ‘brother-in-law’ has several
meanings (cf. 9.2). Similarly, the fact that He went to school may be
translated into French as I est allé a P’école or Il allait a l’école and that
Il allait a l’école may be translated back into English as He used to go to
school, He was going to school and He went to school is consistent with the
view that both the English sentence ‘He went to school’ and the French
sentence ‘II allait 4 l’école’ are non-ambiguous. It is also consistent of
course with the view that they are ambiguous. The point is that transla-
tion and paraphrase are of themselves inconclusive, though they may be
indicative of what are shown to be ambiguities by other tests.
The same holds true of contextualization, It is sometimes argued that
a particular sentence is ambiguous because it might be uttered in quite
different contexts. The difficulty with this criterion is that there is
probably no way of applying it without begging the very question it is
intended to resolve. For there is no reason to rule out the possibility
that two utterances with the same meaning can occur in different con-
texts or that two utterances with a different meaning can occur in the
same context. Similarly, the argument that a sentence is ambiguous
because it might be uttered to describe several distinct states-of-affairs
(and thus would have several distinct sets of truth-conditions: cf. 6.5) is
vacuous. For almost any sentence that can be used to make a statement
will be descriptive of indefinitely many states-of-affairs.
Some of the criteria for grammatical ambiguity that have been pro-
posed rest upon theoretical assumptions about the relationship between
semantics and grammar which are themselves controversial or upon the
assumed validity of questionable syntactic processes. ‘There are others,
however, that are more widely accepted. One of these is co-ordination.
 10.4. Grammatical ambiguity 405
It is a necessary, if not sufficient, condition for co-ordination that con-
joined forms should have the same distribution in other constructions
throughout the sentences of the language. This is a principle of such
generality that it might almost be regarded as a theory-neutral test for
grammatical ambiguity; and its deliberate or unintentional violation
results in what is traditionally known as zeugma*. For example, the
following utterance contains an obviously zeugmatic instance of co-
ordination

(12) We heard your voice and him slam the door.

This much one can say without commitment as to the way in which the
forms your voice and him slam the door are generated in any particular
model of the language-system. We can also say that your votce (or the
expression of which it is a form) is a noun-phrase composed of a deter-
miner and a noun, whereas him slam the door is a desentential transform
of some kind, composed of the object-form of the personal pronoun ‘he’
and a verb-phrase in which the verb is uninflected. As far as it goes, this
is a relatively uncontroversial account of the difference, from a gram-
matical point of view, between your voice and him slam the door; and
there are indefinitely many other pairs of forms in English which differ
in the same way and whose co-ordination as objects of any transitive
verb (not only ‘hear’) would result in zeugma. There is one form, how-
ever, which can be substituted for either your or him in constructions of
the kind we are concerned with here: namely, her. And there are very
many forms which can be substituted for either voice or slam the door:
e.g., shout, cry for help, tap at the window. The substitution of her for
either your or him, or for both simultaneously in (12), nonetheless
results in zeugma. So too does the substitution of any one of the set
{shout, cry for help, tap at the window, . . .} for either voice or slam the
door, or for both simultaneously. But if her shout, her cry for help, etc.,
are substituted for either your voice or him slam the door in (12) the
resultant utterance is perfectly acceptable.
Distributional tests of this kind, when the results are as clear-cut as
they are in this instance, tell us that such utterances as
(13) We heard her cry for help,
unlike either
(14) We heard him|you cry for help
or
406 Semantics and grammar I
(15) We heard his/your cry for help,
are grammatically ambiguous. Whether it is also lexically ambiguous or
not is another matter. If verbs and nouns, whether they are formally and
semantically related or not, are held to be different lexemes, then (13) is
lexically ambiguous simply by virtue of the way the term ‘lexeme’ is
defined (cf. 11.1). More interestingly, if the verb ‘cry’ is polysemous,
‘“‘weep’”’ vs. “shout” (or alternatively, if there are two homonymous
verbs, ‘cry,’ and ‘cry,’), and if ‘cry’ in (13) can be interpreted in either
way (as it obviously could be if we dropped for help), then (13) and (14),
but not (15), are lexically ambiguous.
But are “weep”’ and “shout” distinct senses? Or is the verb ‘cry’
simply more general in sense than ‘weep’ and ‘shout’, as ‘red’ is more
general than ‘scarlet’ and ‘crimson’? To decide such questions, it has
been suggested, we can apply essentially the same co-ordination test as
we have just used to prove a case of grammatical ambiguity. Consider
the following utterances:
(16) Mary was wearing a red sweater and skirt
(17) Mary and Ruth were wearing red skirts
(18) Mary was wearing a red skirt, and so was Ruth.
Let us now suppose that Mary’s sweater was crimson and her skirt
scarlet, and that Ruth’s skirt was crimson. It is clear that (16)-(18) can be
used appropriately to describe this state-of-affairs; and the fact that this
is so can be attributed to the generality, or non-specificity, of ‘red’. In
the case of the following sentences, however,
(19) We heard Mary and Ruth crying
(20) Mary and Ruth cried
(21) Mary cried, and so did Ruth,
the verb ‘cry’ must be construed to mean either “‘weep”’ or “‘shout’’,
but not something which is neutral between the two. (20) and (21) can
be interpreted as meaning, roughly, either “Mary shouted and Ruth
shouted”’ or “Mary wept and Ruth wept’’, but not “‘Mary wept and
Ruth shouted’’. If the co-ordination test is taken to be decisive, then
(19)-(21) are ambiguous; and their ambiguity is lexical, rather than
grammatical.
The co-ordination test, as we have seen, does not distinguish between
grammatical and lexical ambiguity. The fact that
(22) She arrived in a taxi and a flaming rage
 10.4. Grammatical ambiguity 407
or
(23) I was wounded in the desert and the right shoulder
will be recognized as zeugmatic (and, according to one’s taste, as
humorous or not) does not of itself prove that ‘in a taxi’ and ‘in a
flaming rage’ or ‘in the desert’ and ‘in the right shoulder’ belong to
syntactically distinct subclasses of adverbial phrases. Nor does it prove
that ‘in’ has several meanings. These two propositions (which are not of
course mutually exclusive) must be argued for, if they are argued for at |
all, in relation to their implications for the description of the language-
system as a whole. Neither syntactic nor semantic parallelism is a
sufficient condition for non-zeugmatic co-ordination, though each of
them, we may assume, is necessary.
There are, however, two further questions which must be answered,
before we can apply the co-ordination test with complete confidence:
(1) Are syntactic and semantic parallelism jointly sufficient to guarantee
the acceptability of any particular instance of co-ordination? (ii) Is the
co-ordination test reliable, in that native speakers will always agree that
a particular utterance is or is not acceptable or that it necessarily has
either two or four interpretations? Both of these questions, it seems,
must be answered in the negative. Consider the following two utterances:
(24) Fohn likes brunettes
(25) JFohn likes marshmallows.
Each of them has a more natural and a less natural interpretation in
terms of the cultural practices and conventions operative in societies in
which English is normally used. In default of any information to the
contrary, we would normally assume that the person referred to by the
name ‘John’ in (24) and (25) is not a cannibal (or that, if he 1s, his taste
is unlikely to be determined by hair-colour) and that he is unlikely to
have developed an amorous or sentimental predilection for a particular
kind of sweets. The problem is whether (24) and (25) are ambiguous or
non-specific. But however unnatural might be the interpretation of (24)
as ‘John likes eating brunettes”’ or of (25) as “John likes to spend his
time chatting up marshmallows’’, this unnaturalness clearly has nothing
to do with the grammatical or lexical structure of English.
It is perhaps arguable, however, that the verb ‘like’ is polysemous.
The first question, then, is whether such utterances as
(26) Fohn likes brunettes and marshmallows
(27) Fohn likes brunettes more than marshmallows
 408 Semantics and grammar I
(28) John likes brunettes and Bill marshmallows

are acceptable. Most speakers of English would no doubt find them

unnatural and might even reject them. There is undoubtedly a tendency
to impose upon the verb-phrases ‘like brunettes’ and ‘like marsh-
mallows’ the same more specific interpretation when they are conjoined
in the same sentence with the same subject; and this tendency conflicts
with the tendency to give to each its more natural interpretation. But it
is easy enough to construct contexts in which (26)-(28) would be judged
acceptable; and English speakers can be brought to see that this is so,
even though their first reaction was to dismiss the utterances out of
hand. The lesson to be drawn from putting the first question with
respect to utterances like these is that all sorts of non-linguistic factors
are likely to jeopardize the reliability of the native speaker’s spontaneous
judgement of their acceptability. If, therefore, the co-ordination of
‘brunettes’ and ‘marshmallows’ is rejected in (26) for non-linguistic
reasons, might this not also be the case, in part at least, for many of the
sentences which are held to violate grammatical and semantic con-
straints?
Given that (26)-(28) are likely to be judged unnatural anyway, it may
seem almost pointless to put the second question: does the verb ‘like’
necessarily have the same more specific interpretation in each of the
underlying conjoined sentential structures? But it would seem to be the
case that, if (26) and (28) are set in some appropriate context, they are
acceptable enough with ‘like’ taking two different interpretations. But
(27) is decidedly odd if it is construed in this way. It is also possible to
envisage, and to describe by means of (26), various rather unusual
states-of-affairs in which the more natural interpretation can be com-
bined, in either way, with the less natural. In such circumstances (27) is
presumably neither more nor less peculiar than (26), or indeed than
YFohn likes brunettes more than blondes or Fohn likes marshmallows more
than macaroons; and the same goes for (28). What this suggests is that
‘like’ is not polysemous. So too does the fact that the substitution of
such more or less synonymous expressions as ‘be crazy about’, ‘have a
weakness for’, etc., for ‘like’ seems to yield the same results. But the
whole procedure is, to say the least, of doubtful validity. Once we get to
the point of convincing ourselves that, with a little imagination, we can
interpret utterances like (26)~(28), it is easy to start doing the same with
utterances like Mary and Ruth were both crying: one was weeping pro-
fusely and the other was screaming blue murder: cf. (19)-(21).
 10.5. Generative semantics 409
The view that is taken here, and throughout this book, is that the
distinction between ambiguity and generality (or non-specificity), like
the distinction between the grammatical and the ungrammatical, is pre-
theoretically clear in many cases, but not in others. So too is the
distinction between the meaningful and the nonsensical. Diagnostic
tests, such as the possibility of non-zeugmatic co-ordination, will be
applicable, and are useful, in so far as they do not involve native speakers
in making judgements about the acceptability and interpretation of
sentences they would be unlikely to utter or meet in their everyday use
of the language.”

10.5. Generative semantics

It is a widely held view nowadays that the linguist’s model of a language-
system should not only generate all and only the well-formed system-
sentences of the language, but should assign to each system-sentence
both a phonological representation (PR) and aijsemantic representation
(SR). The PR is to be thought of as a representation of the way in which
the system-sentence would be pronounced (if it were uttered as a text-
sentence and transmitted in the vocal-auditory channel) and the SR as
a representation of its meaning. Looked at from this point of view, the
model can be seen as an integrated system of grammatical, phonological
and semantic rules relating sound and meaning; and this is how genera-
tive grammars (in the broadest sense of the term ‘grammar’) are now
commonly described.
When Chomsky first put forward his theory of generative grammar
(in a version that has since been substantially modified), he had little to
say about the possibility of integrating phonology, morphology, syntax
and semantics within a unified model of a language-system (cf. 10.3).
The illustrative partial description of English that he used in his earliest
work did not contain any rules for the semantic interpretation of sen-
tences; and he took the view that the grammatical rules could be
established and formalized without making any appeal to sameness and
difference of meaning or to any other semantic notions. In this respect,
grammar was held to be autonomous and independent of semantics. It
17 'There is a particularly useful discussion of the kind of tests that linguists have
used to distinguish ambiguity from non-specificity in Zwicky & Sadock
(1975). Various types of grammatical and lexical ambiguity are exemplified
in Agricola (1968) and Kooij (1971). A classic work which, unlike most
recent treatments of ambiguity by linguists, emphasizes the positive com-
raunicative
1953). value of multiple meaning in language-utterances is Empson
410 Semantics and grammar I
was always recognized, however, that there were certain systematic
connexions between syntax and semantics and that, in so far as the
choice between two grammatical analyses was otherwise indeterminate,
semantic criteria should be used to resolve the indeterminacy. To this
extent at least, Chomskyan generative grammar has always taken
account of the systematic connexions that were held to exist between
syntax and semantics. In particular, it has always been concerned with
the fact that certain kinds of ambiguity could be regarded as gram-
matically explicable (cf. 10.4); and Harris, if not Chomsky, has from
the outset emphasized the fact that some part of the meaning of a sen-
tence remains constant under transformation. This point must be
stressed in view of the very considerable confusion that now surrounds
the thesis that grammar, and more especially syntax, is autonomous.
Chomsky, like Harris and other post-Bloomfieldian linguists (cf. 8.1),
has continually professed his methodological commitment to the
principles of autonomous syntax. But he has been paying more attention
recently, as have other generative grammarians, to the integration of
syntax and semantics.
The first explicit proposals for the integration of syntax and semantics
within a Chomskyan framework were made by Katz and Fodor (1963).
Their proposals were subsequently clarified and extended by Katz and
Postal (1964) and taken over by Chomsky (1965) in the construction of
what has now come to be called the standard version of Chomskyan
transformational-generative grammar. What Katz and Fodor did, in
effect, was to add to the grammar a dictionary, providing semantic and
syntactic information for each of the lexemes that it contained, and a set
of projection-rules*, whose function it was to associate with every
semantically well-formed sentence at least one semantic representation.
The general orientation of the Katz and Fodor approach to the integra-
tion of syntax and semantics is evident from their famous slogan:
“linguistic description minus grammar equals semantics’’ (in which
‘grammar’ is to be understood to cover, not only syntax and morphology,
but also phonology). As far as well-formedness was concerned, semantics
was residual: ‘“‘semantics takes over the explanation of the speaker’s
ability to produce and understand new sentences at the point where
grammar leaves off’’. Given that a particular string of forms was
defined by the grammar to be syntactically ill-formed, the question
whether it was semantically well-formed or ill-formed simply did not
arise. It was only with respect to grammatical sentences that the pro-
jection-rules had any role to play. This view of semantics as purely
 10.5. Generative semantics 4II
residual has had the effect that “research has been biased heavily in
favour of syntactic solutions to problems” (Jackendoff, 1972: 2).
A further point to be noted about the Katz and Fodor proposals is
that they imposed a spurious parallelism upon phonology and semantics.
The syntactic part of the integrated model of the language-system, con-
sisting of base-rules and transformational rules, was held to be central,
not only in that it came between the phonological and the semantic parts
of the model, but also in that it contained all the generative capacity of
the whole integrated model. In contrast with the rules of syntax, and
more particularly with the base-rules, both the semantic rules (i.e. the
projection-rules) and the phonological rules were held to be non-
generative and interpretive*. Their function was to take as input
syntactically structured strings of forms generated by the syntactic part
of the model and to interpret these in terms of allegedly universal
elements of meaning and of sound. What is to be noticed here (apart
from the alleged universality of the elements of sound and meaning: cf.
9.9) is the curious use of the term ‘interpret’, according to which both
the pronunciation and the meaning of a sentence constitutes an inter-
pretation of it. Apart from the terminology that is employed, the Katz
and Fodor model is strikingly similar, at this point, to the so-called
glossematic* version of structuralism developed by Hjelmslev and his
collaborators some years earlier (cf. Spang-Hanssen, 1954); and it is
open to the same objections.
There is an inherent connexion between grammar and semantics
which does not hold between grammar and phonology (cf. 10.1); and
this fact should be captured in anything that purports to be a model of a
language-system. It is, to say the least, obscured in the Katz and Fodor
model, as it is in any model that treats the phonological and the semantic
representations associated with sentences as being comparable theoreti-
cal constructs. Henceforth, we will avoid using the terms ‘semantic
representation’, on the one hand, and ‘phonological interpretation’ (or
‘phonetic interpretation’), on the other. We will talk instead of the
phonological representation of a sentence (on the assumption that it is
realized in the phonic medium: cf. 3.3) and of its semantic interpreta-
tion (or interpretations). Incorporating these terminological modifica-
tions into the Katz and Fodor model, we can formulate the relationship
between the several parts of their integrated model of the language-
system by means of the diagram in figure 9g.
It will be noted that the base* has been distinguished from the other
three sets of rules. This is intended to take account of one of the
412 Semantics and grammar I ,
principal changes that Chomsky (1965) made in his formulation of the
so-called standard version of transformational grammar: the inclusion
of the lexicon* (or dictionary: cf. 13.1) as a sub-component of the base.
We are not concerned with the reasons for this change or its implica-
tions. Another change that has been incorporated in figure 9, in order to

HES
bring it into line with the standard version of Chomskyan transforma-
tional grammar, is the introduction of the notions of deep structure and

HES
Figure 9. The so-called standard theory
surface structure. Apart from the base, there are three boxes of rules:
transformational rules (‘I’-rules), semantic rules (S-rules) and phono-
logical rules (P-rules). The output of the base is a set of deep structures
(DS), to which the S-rules (Katz & Fodor’s projection rules) apply
and yield a set of semantic interpretations (SI). The output of the T-
rules, on the other hand, is a set of surface structures, to which the
phonological rules apply and derive for each sentence its phonological
representation (PR).
The general conclusion towards which Katz and Fodor (1963), and
more especially Katz and Postal (1964), were working was the thesis
that (apart from certain rules that were responsible for what was held to
be purely stylistic variation) all of the T-rules were obligatory; and
this thesis was taken over and made part of the standard version of
transformational grammar by Chomsky (196s). It carries as an imme- _
diate corollary, by virtue of the semiotic principle that meaningfulness
implies choice (cf. 2.1), the proposition that transformations do not
affect meaning.!8 It is only in so far as this proposition is held to be true
18 The implications of this proposition and the looseness with which it was
formulated are discussed in Partee (1971).
 10.5. Generative semantics 413
that one can maintain the principle that all the information that is
relevant to the semantic interpretation of a sentence is present in its
deep structure. Acceptance of this principle is made explicit in figure 9,
it will be observed, by the absence of any path from SS to SI.
It is tempting, having accepted that deep-structure identity is a
sufficient condition of semantic identity, to take the further step of
making it a necessary condition also. In effect, this is what is done by
those calling themselves generative semanticists (cf. Lakoff, 19714), who
argue that there is no need to postulate any distinction between the deep
structure of a Sentence and its semantic interpretation. Their approach
: to the construction of an integrated model of linguistic description 1s
shown in figure 10.9

= +OLEKS Figure 10. The so-called generative semantics theory

The terms ‘interpretive semantics’ and ‘generative semantics’,

which have been widely employed to refer to the alternative conceptions
of the relationship between semantics and syntax that are diagrammed in
figure g and figure 10 respectively, are quite inappropriate for this pur-
pose. Any model of a language-system that generates a set of semantic-
ally well-formed sentences must rest upon a theory of semantics that is
properly described as generative. The difference between the alternative
conceptions of the relationship between semantics and syntax is not
therefore that one rests upon a theory of generative semantics and the
other does not. They both presuppose the existence or possibility.of a
theory of generative semantics. Indeed, in so far as these two concep-
tions of the relationship between semantics and syntax have been put
forward within the general framework of Chomsky’s theory of trans-
formational grammar, they have both taken for granted a very particular
kind of generative semantics: they have both accepted that a model of
linguistic description should not only generate the set of semantically
well-formed sentences, but should also associate with each a semantic
19 Neither figure 9 nor figure 10 is intended to capture all aspects of the two
models that they, very sketchily, represent. In particular, it should be noted
that figure 10 says nothing about the rules which generate SI: clearly such
rules are needed to perform the functions that are jointly performed, accord-
ing to the so-called standard theory, by the base-rules and the rules of seman-
tic interpretation.
414 Semantics and grammar I
interpretation in terms of a universal inventory of sense-components.
Componential analysis, as we have already seen, is theoretically suspect
on a number of counts (cf. 9.9). What we are concerned with here is
whether any kind of generative semantics 1s viable, independently of its
association with componential analysis, on the one hand, or with various
kinds of transformational grammar, on the other.
The difference between the alternatives shown in figures 9 and 10 is
that the former draws a distinction, which the latter does not, between
the deepest syntactic analysis of a sentence and its semantic interpreta-
tion. Another way of expressing this difference is to say that, whereas
the former is syntactically based*, the latter is semantically based*. A
syntactically based model operates, as we have seen, according to the
principle of the autonomy of syntax; a semantically based model does
not. Various kinds of syntactically based and semantically based models
are conceivable. But most of the discussion of the difference between
them that has taken place in recent years has been centred upon the role
that is assigned to deep structure in Chomsky’s standard theory of
transformational grammar.
According to Chomsky (1965) the deep structure of a sentence is a
phrase-marker which contains all the lexemes whose forms appear in the
surface structure of the same sentence; and it is in terms of the topology
of the deep-structure phrase-marker that the semantically relevant
notions of subject*, object* and predicate* are defined and selection-
restrictions* are accounted for. The so-called generative semanticists
take the view that lexicalization* is a particular kind of transformational
process.2° For example, the lexeme ‘kill’ might ‘be taken from the
lexicon (cf. 13.1) and substituted for an underlying structure containing
the sense-components CAUSE, BECOME, NOT and ALIVE (cf. 9.9); and the
operation whereby this substitution is carried out would be one,
among many, of the transformations involved in the generation of any
sentence containing the lexeme ‘kill’. Furthermore, lexicalizing trans-
formations of this kind do not operate in a block, it is argued, prior to
the operation of other transformations: they must be interspersed with
what would be conventionally regarded as purely syntactic transforma-
tions; and it is principally for this reason that the Chomskyan notion of
deep structure is rejected. The so-called generative semanticists also
20 [_exicalization is performed by means of a transformational rule of substitution
in Chomsky’s (1965) model too. But it is a less powerful kind of trans-
formational rule, which has no effect upon the phrase-marker to which it
applies other than that of substituting a lexical form for a dummy-symbol.
 10.5. Generative semantics A415
reject the treatment of selection-restrictions proposed by Chomsky
(1965); and, like many other linguists, they deny that what the standard

of sentences. |
theory of transformational grammar defines to be deep-structure sub-
jects, objects and predicates play any role in the semantic interpretation

In view of the confused, and at times acrimonious, discussion of the

rival merits of syntactically based and semantically based models of
linguistic description that has been taking place among transformational
grammarians, it is as well to emphasize that the point at issue is a highly
technical one that cannot even be formulated except within the frame-
work of a particular formalization of the structure of language-systems.
Unless we make the initial assumption that the semantic interpretation
of a sentence is, or may be represented as, some kind of formalizable
entity, it hardly makes sense to enquire whether the semantic inter-
pretation of a sentence is or is not identical with the deepest underlying
phrase-marker that is postulated by the transformational grammarian
in order to account for what he takes to be syntactic well-formedness.
Far from being radically different alternatives, the syntactically based
model of figure g and the semantically based model of figure 10 have so
much in common, in terms of the meta-theoretical assumptions that
support them and the formalism that they employ, that they are more
properly seen as relatively minor, and perhaps ultimately indistinguish-
able, variants of the Katz and Fodor approach to the integration of
syntax and semantics.
One of the most striking features of the presentation by Katz and
Postal (1964) of the thesis that transformations do not change meaning
was the looseness with which the term ‘meaning’ was employed. No
account was taken of the fact that the semantic relationship between a
declarative sentence and an interrogative sentence, or between a
declarative sentence and an imperative sentence, was a different kind of
semantic relationship than that which holds, or may hold, between two
declarative sentences. Furthermore, no distinction was drawn between
the meaning of a sentence and the meaning of an utterance; and what
was held to be purely stylistic variation was thereby classified as seman-
tically irrelevant.
It is now more widely recognized, both by transformationalists and
by non-transformationalists, that there are different kinds of meaning
to be accounted for in the analysis of language-systems. As far as the
research that has been carried out by transformationalists is concerned,
this may have done little so far to resolve the question whether a
416 Semantics and grammar I
semantically based model or a syntactically based model is preferable.
But it has had the effect of concentrating the attention of semanticists
upon a range of topics (negation, quantification, pronominal reference,
presupposition, etc.) whose importance transcends the theoretical and
meta-theoretical differences that divide various schools of linguists.
Chomsky (1972) has now abandoned the standard theory of trans-
formational grammar in favour of what he calls the extended standard
theory. A model of a language-system constructed in accordance with
the extended standard theory is still a syntactically based theory; but it
allows for the possibility that the semantic interpretation of a sentence
should be determined jointly by its deep structure and its surface
structure. That the propositional content of sentences is held constant
under transformation has always appeared to be a more defensible thesis
than the original Katz and Postal thesis that transformations have no
effect upon the meaning of sentences or the even stronger thesis that all
the information relevant to the semantic interpretation of a sentence is
present in deep structure.*!
The various kinds of meaning that have been mentioned in the pre-
vious paragraph are all discussed elsewhere in this book in a framework
that is intended to be as neutral as possible with respect to alternative
theories of grammatical structure. Without saying any more about the
difference between semantically based and syntactically based trans-
formational grammars, we may now turn to a consideration of whether
it is necessary or feasible for the linguist’s model of a language-system to
generate all and only the semantically well-formed sentences of the
language, regardless of whether the semantically ill-formed sentences
are excluded by the rules of the base or by projection-rules of the kind
proposed by Katz and Fodor (1963). In what follows, the terms
‘anomalous’ and ‘deviant’ will be used rather loosely, as pre-theoretical
terms, to cover both semantic unacceptability and certain other kinds of
abnormality or aberrance.
There are two classes of sentences which philosophers and linguists
have generally treated as anomalous and which are clearly of concern to
the semanticist: sentences which (when they are used to make state-
ments) express tautologies* and sentences which express contradictions*.

21 By the propositional content of a sentence is meant that part of the meaning

of a sentence to which the terms ‘true’ and ‘false’ are applicable on particular
occasions of the utterance of that sentence. For example, ‘John came in’,
‘Did John come in?’, ‘It was John that came in’, etc., differ in meaning, but
have the same propositional content (cf. 12.7, 16.3).
 10.5. Generative semantics 417
Both tautologies and contradictions are, in principle, uninformative: a
tautology tells the addressee nothing that he does not know, or could not
deduce, by virtue of his knowledge of the language; and a contradiction
fails to tell him anything that he can accommodate in his state-
description of the world (cf. 2.3). But to say that tautologies and contra-
dictions are uninformative is not to say that they are meaningless or
semantically unacceptable. If they were meaningless, they could not
have a truth-value; and their status as tautologies or contradictions rests
upon their being necessarily true or necessarily false, respectively. There
can be no question, therefore, of excluding sentences that express
tautologous or contradictory propositions from the set of semantically
well-formed sentences.
Tautologies are of not infrequent occurrence in everyday language-
behaviour. They are very commonly used, metalinguistically, in order to
explain the meaning of an unfamiliar word. For example, ‘Abiogenesis
is spontaneous generation’ can be understood as expressing, indirectly,
a proposition about ‘abiogenesis’ (rather than the proposition about
abiogenesis which it more directly expresses): in which case it gives the
addressee information about the language-system. ‘Tautologies are also
uttered, although this usage is more characteristic of formal deduction,
in order to make explicit one of the steps that would normally be left
implicit in the development of an argument. More interesting, tauto-
logies may be uttered to express what the addressee is expected to
recognize as a self-evident truth and from which he is to draw some
relevant conclusion (e.g., ‘Business is business’, ‘He is his father’s
son’), It is important to realize that, although the particular interpreta-
tion given to such utterances may vary from context to context, the
meaning of the sentence itself is constant. There is no need to invoke
any notion of metaphor or connotative meaning in order to account for
their interpretability. What the addressee does, upon hearing and under-
standing a tautologous utterance, is to say to himself, as it were: “There
must be some reason for the speaker to tell me what he knows I know to
be true. What can this reason be?”’ The addressee assumes, in default
of any evidence to the contrary, that the speaker is not indulging in
irrelevant platitudes.
The addressee makes a somewhat different assumption in the case of
what appear to be contradictions. He then says to himself: ‘The
speaker cannot be asserting what is patently a contradictory preposition.
What interpretation can I impose upon his utterance that would remove
the apparent contradiction?’’ Suppose, for example, that the speaker,
418 Semantics and grammar I
being asked whether a certain person is married, replies with the
utterance He is and he isn’t and that it is clear, in the context, that this
utterance is intended to be both informative and interpretable. There
are several ways in which the addressee might interpret the two con-
joined clauses as being logically compatible with one another. But it is
perhaps most plausibly interpretable as meaning “From one point of
view (or in certain respects) he is married and from another point of
view (or in other respects) he is not married”’. More specifically, it might
be interpreted to mean that the person being referred to is married under
the laws of one country, but not married under the laws of the country
in which he is domiciled; that he is married according to the rites of the
religion to which he or his wife subscribes, or in common law, but has
never contracted a legally valid civil marriage; that he is in fact legally
married, but does not conduct himself as a married man normally does
and should; and so on. We will not go into all the various possibilities.
The general point is that the deviance of contradictions is different from
that of tautologies. Whereas tautologies can be taken at face value, what
are at first sight contradictions are usually reinterpreted in such a way
that they are seen as merely paradoxical rather than as logically in-
consistent. In both cases, however, their interpretation, in context, 1s
subject to the application of procedures, or strategies, which derive
from the assumption that the speaker must have had some reason for
uttering a platitude or paradox.
In our discussion of the limits of grammaticality earlier in this chapter
(cf. 10.2), we said that there was no pre-theoretical notion of gram-
matical acceptability comparable with the everyday notion of semantic
acceptability or making sense. The question we are now discussing 1s
how much, if any, of what is covered by the pre-theoretical concept of
making sense is to be accounted for in the analysis of the semantic
structure of particular languages. This question is frequently decided,
in principle, in terms of a distinction between the native speaker’s
knowledge of his language and his knowledge or beliefs about the world.
For example, the sentence ‘My mother is younger than I am’ might be
held to express a proposition which describes a biologically impossible
state-of-affairs; and the speaker’s categorization of this sentence as
nonsensical (if he does so categorize it) is readily accounted for in terms
of this fact. It is unnecessary and undesirable to classify such sentences
as semantically ill-formed in terms of rules in the linguist’s model of the
language-system. Indeed, one good reason for not trying to account for
the anomaly or deviance of utterances like My mother 1s younger than I
 10.5. Generative semantics 41g
am in this way is that, although the native speaker might say that they
are absurd or nonsensical, he must first be able to interpret them in
order to classify them as nonsensical. Asked to explain why My mother
is younger than I am does not make sense, he would not say that he does
not know what it means, but rather that the proposition that it expresses
could not be true. Furthermore, although most native speakers would
probably say that My mother is younger than I am is anomalous (on the
assumption that ‘my mother’ refers to the speaker’s genetic or uterine
parent) they might be persuaded to agree, upon reflexion, that there are
imaginable, if biologically impossible, situations which could be cor-
rectly described by the proposition “ My mother is younger than I am”’.
All we have to do is to envisage the possibility of arresting or reversing
the biological process of ageing; and many works of science-fiction take
this possibility for granted. It requires a little more ingenuity to en-
visage the possibility of a child being born before its mother. But even
this is conceivable, provided that we interpret ‘mother (of X)’ as
meaning ‘“‘female genetic parent (of X)’’, rather than “ person who has
given birth to X’’; and any proposition that describes a logically possible
situation must be allowed as meaningful. It follows that, not only must
the sentence expressing this proposition be generable in the linguist’s
model of the language-system, but also that there is no reason whatsoever
for the sentence to be regarded as other than perfectly well-formed.
It must be constantly borne in mind that informants cannot be asked
to supply interpretations for the system-sentences that the linguist’s
model generates. System-sentences are theoretical constructs, which
have no existence outside the model (though they may of course be rep-
resented and distinguished one from another in terms of some appro-
priate notation). What the linguist puts to his informant (or to himself
as an informant) are actual or potential utterances; and they will always be
interpreted, if they are interpretable, in the light of the informant’s
beliefs and assumptions. If we draw a clear distinction between system-
sentences and utterances, we will the more easily avoid falling into one or
other of two common misconceptions of semantic well-formedness.
The first 1s to suppose that we can test directly for this as a property
of the language-system. We can put to an informant an actual or
potential utterance and, without telling him anything about the context
in which it might be uttered, ask him to say whether it is anomalous or
not. There is no reason to believe, however, that we are thereby tapping
: the informant’s intuitive knowledge of the language-system, as distinct
from his ability to interpret utterances in relation to the contexts in
420 Semantics and grammar I
which they occur or the plausibility of the situations that they describe.
It follows that, even if all our informants were to agree that, of two
utterances considered out of context, one is deviant and the other is
non-deviant, our model of the language-system need not reflect this fact
by generating a system-sentence isomorphic with the non-deviant
utterance and by failing to generate a system-sentence isomorphic with
the deviant utterance. When our informants tell us that a particular
utterance is deviant, anomalous, bizarre, etc., they may simply mean
that they cannot immediately imagine the circumstances under which
they would produce it. But if there are any circumstances at all under
which the utterance, if produced, would be readily interpretable by native
speakers, the utterance itself must be treated as semantically acceptable
and the corresponding system-sentence as semantically well-formed.
The second misconception consists in assuming that, because the
informant’s judgements about the semantic deviance or non-deviance of
utterances are relative to the contexts in which the utterances might be
imagined to occur and to the informant’s labile and variable beliefs
about the world, the notion of semantic well-formedness, in so far as it
applies to system-sentences, must also be made relative to the beliefs,
presuppositions and expectations of speakers. The language-system
postulated by the linguist may or may not be separable from a person’s _
other perceptual and cognitive faculties. But whether this is or is not
the case, it is obvious that we bring to bear the whole of our cognitive
ability in the interpretation of language-utterances. To attempt to build
into the linguist’s model of the language-system all the factors which
determine our capacity to interpret utterances would be to nullify the
very concept of a language-system.
Many of the sentences that linguists have chosen to regard as seman-
tically ill-formed in recent discussions of the question would seem to be
perfectly well-formed: that is to say, the corresponding utterances
would seem to be semantically acceptable. For example, ‘His type-
writer has bad intentions’ (cf. Bierwisch, 1970) correctly describes a
state-of-affairs, the existence of which is generally held to be prohibited
by the nature of typewriters: it expresses a proposition that we can
rationally debate, even though we cannot perhaps rationally hold it to
be true without surrendering other beliefs to which we are committed.
Similarly, ‘The horse miaowed’ (cf. Leech, 1974) is surely to be regar-
ded as semantically well-formed, on the grounds that it expresses a
proposition that we could, not only rationally discuss, but even verify.
We should be surprised, of course, if we actually found in the world in
 10.5. Generative semantics 421
which we live our everyday lives a horse that miaowed, rather than
neighed. But that is beside the point. We could identify a horse miaowing
if we ever came across one.
One of the reasons that is often given for including in a model of a
language-system rules that are sensitive to so-called selection-restric-
tions is that such rules make it possible to account for the native speaker’s
. ability to infer propositions that are presupposed or implied, rather than
asserted. For example, from That person is pregnant, one would nor-
mally infer that the person referred to is female; and, on the vast
majority of occasions on which this utterance might be produced, the
inference would no doubt be correct. But the proposition expressed by
the sentence ‘That person is female’ is certainly not entailed by the
proposition expressed by ‘That person is pregnant’: the inference is in
principle no more than probabilistic, since it is possible to envisage a
world in which men could be pregnant. (For the same reason, there is no
violation of the rules of the language-system involved in the sentence
“That man is pregnant’: cf. Jackendoff, 1972: 21.)
Another reason that is given for having rules which are sensitive to
the collocational restrictions holding between particular lexemes is that
they explicate the alleged fact that a phrase or sentence may be un-
ambiguous even though it contains one or more homonyms or poly-
semous lexemes in positions that the purely grammatical rules specify
as permissible for them. For example, it might be argued that, whereas
both ‘ball,’ (“spherical or ovoid object used in certain games’’) and
‘ball,’ (“‘elegant kind of party featuring social dancing’’) are permissible
in noun-phrases governed by a transitive verb, the sentence ‘The man
hit the ball’ cannot be interpreted as containing ‘ball,’, by virtue of the
requirement associated with the verb ‘hit’ that its object should refer
to a physical object (cf. Katz & Fodor, 1963). Granted that The man hit
the ball (i.e. some token of the utterance-type that is isomorphic with
the sentence ‘The man hit the ball’) will, in all probability, be inter-
preted by native speakers of English as containing ‘ball,’, rather than
‘ball,’, if the utterance is put to them out of context, it does not follow
that the sentence ‘The man hit the ball’ must be treated as non- |
ambiguous. What has to be demonstrated is that there are no circum-
stances under which The man hit the ball could be construed as con-
taining ‘ball,’. Even if this could be demonstrated, or safely assumed, it
still does not follow that rules must be formulated within the linguist’s
model of the language-system to exclude the possibility of taking ball as
a form of ‘ball,’ in this instance. Other cognitive abilities besides
422 Semantics and grammar I |
knowledge of the language-system are involved in the recognition and
interpretation of utterances.
It is difficult to avoid the conclusion that no convincing case has yet
been made for the thesis that a linguist’s model of a language-system
should be such that it generates all and only the semantically well-
formed sentences of the language (as a proper subset of the gram-
matically well-formed sentences), Nor indeed is it as obvious as it is
often assumed to be that a generative model of a language-system should
associate with each well-formed sentence one or more interpretations in
some appropriate notation. Semantic interpretations, considered as
representations of the meaning of sentences, are theoretical constructs,
which must be justified (if they can be justified) in terms of their
explanatory value; and there is nothing of what everyone would agree
has to be explained in terms of the structure of the language-system
(e.g., synonymy, antonymy, tautology, contradictoriness, entailment and
paraphrase) that cannot be explained in terms of relations defined over
sentences without the postulation of such intermediate theoretical con-
structs as semantic interpretations.
The criticisms that have been made here of the underlying assump-
tions of generative semantics should be read in conjunction with the
criticisms that were made of componential analysis in an earlier section
(9.5). Though generative semantics is logically independent of com-
ponential analysis, it undoubtedly derives much of its attraction from
the fact that it is commonly presented in association with the assump-
tion that it is possible to analyse the semantic structure of all languages
in terms of a set of universal sense-components; and this assumption is,
to say the least, questionable. Having made this point, however, it must
be emphasized that, independently of the soundness of their underlying
assumptions, both generative semantics and componential analysis have
been of immeasurable importance in recent years in that they have
obliged their practitioners to present their analyses in a precisely speci-
fied format; and this has brought them within the range of constructive
criticism and emendation which has undoubtedly increased our under-
standing of the complexity of the issues involved.??
22 For further discussion of the issues treated in this section reference may be
made to Bartsch & Vennemann (1972), Bierwisch (1970), Bolinger (196s),
Bonomi & Usberti (1971), Botha (1968), Chafe (1971), Chomsky (1972),
Dougherty (1975), Drange (1966), Dubois-Charlier & Galmiche (1972),
Fillmore (1972), Fodor (1977), Galmiche (1975), Hasegawa (1972), House-
holder (1973), Jackendoff (1972), Katz (1970, 1971), Kempson (1977),
Lakoff (1971a), Leech (1974), McCawley (1968, 1971b), Postal (1974),
Sampson (1973, 1975), Wierzbicka (1975).
 I]
Semantics and grammar II

11.1, Parts-of-speech, form-classes and expression-classes

In this and the immediately following sections, we shall be concerned
primarily with two questions: (1) Do all languages have the same
parts-of-speech (i.e. nouns, verbs, adjectives, etc.)? (11) To what degree
are semantic considerations relevant to the definition of such terms as
‘noun’, ‘verb’ or ‘adjective’? These two questions, as we shall see, are
intrinsically connected. Curiously enough, they are only rarely dis-
cussed nowadays. And yet they are crucial in any treatment of the
relation between grammar and semantics.
Although most of the published grammars and dictionaries of par-
ticular languages make use of the traditional terms ‘noun’, ‘verb’,
‘adjective’, etc., the standard definitions of such terms have long been
criticized by linguists as being unsatisfactory in several respects. It has
been argued that they are circular; that they depend upon a mixture of
morphological, syntactic and semantic criteria, which do not necessarily
coincide in particular instances; and that they are inapplicable to lan-
guages whose grammatical structure differs significantly from that of the
classical Indo-European languages. The approach which, in outline
only, we present here concedes that there is considerable force in these
criticisms, but also gives due recognition to those aspects of the tradi-
tional theory which are relevant to the central concerns of this book and,
with certain qualifications and clarifications, are of enduring validity. In
doing so, we shall attempt to hold the balance between the two extremes
of universalism and relativism (cf. 8.1).
The terms ‘noun’, ‘verb’, ‘adjective’, etc., are commonly used to
cover both lexemes and forms: for example, the lexeme ‘come’ is said
to be a verb; so too are the forms came or comes. In what follows, we will
restrict the term part-of-speech*, and also the terms ‘noun’, ‘verb’,
‘adjective’, etc., to lexemes and expressions. We will assume that every
word-lexeme is assigned, in the analysis of any language-system, to one,
and only one, such class. In making this assumption, we are adopting
424 Semantics and grammar IT
the traditional view, according to which the noun ‘love’ and the verb
‘love’, for example, are different lexemes: we shall have more to say
about this later (13.4). Nothing will be said in this section about the
assignment of compound lexemes*, and various kinds of idioms, to
particular parts of speech (cf. 13.3). It is our assumption that this
depends upon the prior classification of word-lexemes.
The term form-class* (which is often used in modern linguistics in
place of the traditional term ‘part-of-speech’) will also be restricted in
its application. Forms, whether they are simple or complex, can be
grouped together in several different ways. In view of the fact that several
conflicting definitions of ‘form-class’ are to be found in the literature,
we will arbitrarily opt for a definition of ‘form-class’ in terms of
syntactic equivalence: two forms, f, and f;, are members of the same
form-class F, if and only if they are intersubstitutable (i.e. have the
same distribution) throughout the sentences of the language (cf. 10.1).
Given that ‘come’ is a verb and that come, comes, coming and came are its
forms, we will say that they are all verb-forms; and we will use the terms
‘noun-form’, ‘adjective-form’, etc., similarly. It will be obvious that
the set of all English verb-forms (or the set of all English noun-forms)
is not a form-class in terms of our definition. Nor are sets of morpho-
syntactically equivalent forms in English, such as {wrote, came, .. .}:
wrote and came are morphosyntactically equivalent, in that they realize
morphosyntactic words that have the same morphosyntactic properties
(they are both past-tense forms). But they are not syntactically equiva-
lent (i.e. intersubstitutable); the one is a form of a transitive verb and
the other a form of an intransitive verb. (In so far as the definition of
these various sets of forms other than form-classes depends upon the
distinction between word-forms and morphosyntactic words, it will be
inapplicable, of course, with respect to languages of the so-called
isolating type.) It should be noted that the form-classes of a language
(unlike the parts of speech) will not necessarily be non-intersecting sets:
the same form might belong to two or more different form-classes. We
will return to this point in our discussion of homonymy (cf. 13.4).
\As we have defined form-classes in terms of intersubstitutability, so
we will define expression-classes*: two expressions, e,; and ¢,, are
members of the same expression-class EZ, if and only if they can be
substituted one for the other throughout the sentences of the language.
It will become clear in the course of our discussion that expression-
classes are of particular importance in any discussion of the relationship
between syntax and semantics. It will also become clear that a distinc-
 11.1. Parts-of-speech, form-classes and expression-classes 425
tion must be drawn between noun-expressions and nominal expressions
(or nominals), between verb-expressions and verbal expressions (or
verbals), and so on. The question whether all languages have nouns and
verbs is, as we shall see, distinct from, but related to, the question
whether all languages have nominals and verbals. It is somewhat tire-
some to have to make all these terminological distinctions. But nothing
but confusion will result if we do not distinguish, terminologically or
symbolically, between nouns, noun-forms and nominals; between
verbs, verb-forms and verbals; and so on. As far as the distinction
between nouns and nominals is concerned, we have already seen that,
whereas nouns have denotation, nominals (i.e. noun-phrases, as they are
usually called in current versions of generative grammar: cf. 10.3) have
(or may have) reference: the denotation of a noun like ‘man’ is quite
different from the utterance-bound reference of a nominal like ‘that
man’, ‘he’ or ‘John’ (cf. 7.2). Attention should also be drawn, in this
connexion, to the possibility of confusion that results from the use of
such inherently ambiguous terms as ‘nominalization’, ‘adjectivaliza-
tion’, etc. It has not been judged necessary, in this book, to distinguish
terminologically between the derivation of nouns (typically, in English,
by means of a nominalizing sufhx: cf. 13.2) and the formation of
nominals by means of a syntactic transformation (cf. 10.3). The con-
texts in which we have employed terms like ‘nominalization’ should, in
each instance, make it clear whether we are talking of the morphological
derivation of lexemes or of the creation of expressions by means of
syntactic transformations. These two senses are of course connected;
and more will be said about this in a later chapter (cf. 13.2).
At least three different strands must be unravelled in the rather
tangled skein which makes up the traditional theory of the parts-of-
speech: the morphological, the syntactic and the semantic. To illustrate
this point, it may be helpful if we first quote, and comment briefly upon,
a pair of representative definitions, taken from a particularly good and
authoritative dictionary of English (Urdang, 1968).
| Noun: “‘any member of a class of words distinguished chiefly by
having plural and possessive endings, by functioning as subject or
object in a construction, and by designating persons, places, things,
states, or qualities”’.
Verb: “‘any member of a class of words that function as the main
elements of predicates, typically express action or state, may be
inflected for tense, aspect, voice and mood, and show agreement with
subject or object’.
426 Semantics and grammar IT
These definitions have been framed with considerable care. But it will
be seen immediately that they are vulnerable to all the criticisms that
have been directed against the traditional definitions: circularity; the
mixing of potentially non-coincident morphological, syntactic and
semantic criteria; inapplicability to certain languages. It may also be
noted, in the light of our insistence upon the necessity of distinguishing
between nouns and nominals, between verbs and verbals, etc., that the
use of the term ‘word’ in these definitions makes it impossible to decide
what kind of linguistic unit is in fact being defined. Nor is it made clear
whether the conditions specified in the definitions are severally suffi-
cient or jointly necessary. These are points that would need to be
clarified before we could use these definitions in order to answer the
questions that were posed at the beginning of this section. At the same
time, the definitions are helpful (and they are better than the definitions
to be found in most conventional dictionaries of English) in that they
give some indication of the kind of criteria that are generally held to be
relevant.
The morphological parts of the definitions that we have quoted are
‘having plural and possessive endings’? and “may be inflected for
tense, aspect, voice and mood’’. The first of these conditions, unlike the
second, looks as if it has been formulated with English in mind. Yet
there are many lexemes in English, conventionally classified as nouns,
which have no plural or possessive form (cf. ‘significance’, etc.); and
there are others whose plural form is not made up of a stem and a
pluralizing suffix (e.g., sheep, mice, etc.), We can make the morpho-
logical condition for nouns rather more generally applicable by sub-
stituting “‘being inflected for number and case’’. But, if this is proposed
as a universally applicable condition, we must first decide whether it is
intended to be necessary, sufficient, or both; so too for the proposed
morphological condition for verbs — “being inflected for tense, aspect,
voice and mood’’. Since these conditions make an appeal to the notion
of inflexion, they are obviously inapplicable in the analysis of languages
of the so-called isolating type, in which each lexeme has but one,
morphologically unanalysable, form. Furthermore, they presuppose
some general definition, not only of inflexion, but also of notions like
case, number, tense, mood, aspect and voice. Granted that these cate-
gories can be satisfactorily defined, it does not follow that any or all of
them will be found in every inflecting language. It cannot therefore be
taken as a necessary condition for nouns, even in inflecting languages,
that they should be inflected for number and case; or for verbs, that they
 11.1. Parts-of-speech, form-classes and expression-classes 427
should be inflected for tense, mood, aspect and voice. We cannot even
take them as sufficient conditions, since there are languages in which
nouns are inflected for tense and there are languages in which verbs are
inflected for number. In saying this, of course, we are assuming that
there are other overriding criteria which determine the classification of
lexemes as nouns and verbs; and this 1s an assumption that few linguists
would challenge. Morphological criteria, then, cannot be used in any
definition of the parts-of-speech that purports to be universally applic-
able.
The conditions that we cited in the previous paragraph, “being
inflected for number and case’’ and “being inflected for tense, aspect,
voice and mood’’, are not in fact purely morphological (though they are
usually so described): they are morphosyntactic. Purely morphological
definitions would make no reference to such syntactic categories as case,
number, tense, etc.; and they would be even more obviously language-
specific. For example, a distinction might be established between two
parts-of-speech, X and Y, on the grounds that the members of X each
had only one form associated with them and the members of Y more
than one form; or more generally, on the grounds that the members of
X each had m forms and the members of Y had n forms (m + n). This
is one kind of purely morphological definition; and it is clear that, even
if it is readily applicable (as well it might be) in the analysis of particular
languages, it cannot be used to distinguish, say, nouns from verbs unless
there are supplementary non-morphological criteria for saying that the
members of .X, say, are nouns rather than verbs.
Morphosyntactic and purely morphological criteria (which figure
prominently in many published grammatical analyses of particular lan-
guages) should not be discounted, however, as irrelevant. Although
there is no reason, in principle, why the morphological structure of a
language (if it has one) should be related to its syntactic and lexical
structure, it is an empirically verifiable fact that it is; and there tends to
be a more or less high degree of correlation between the parts-of-speech,
as they are defined morphosyntactically or morphologically, and the
parts-of-speech, as they are defined with reference to other criteria. In
any general theory of the parts-of-speech, morphological and morpho-
syntactic considerations are of secondary importance. But in the analysis
of particular languages, to the degree that they support the more widely
applicable criteria that define the parts-of-speech in the general theory,
they may be not only relevant, but in some instances decisive. For
example, of the two words meaning “white” in Russian one is a verb
428 Semantics and grammar II
(‘beletj’) and the other an adjective (‘belyj’) on morphosyntactic
grounds.
Syntactic definitions of the parts-of-speech rest ultimately upon the
possibility of grouping simple and complex expressions into expression-
classes in terms of the distribution of the forms of each expression. It is
now generally agreed that the operation of substituting one form for
another in the same environment throughout a representative sample of
the sentences of a language cannot be used as a mechanical and self-
sufficient discovery procedure in syntactic analysis. But this is a separate
question: we are not concerned here with discovery-procedures (cf.
Chomsky, 1957). We are simply assuming that any generative grammar
will define (for the language it generates) a set of expression-classes
(nominals, verbals, etc.) and will make use of this classification of ex-
pressions in characterizing the well-formed sentences of the language.
We must also assume, of course, that in languages in which there is,
characteristically, a one-many relationship between expressions and
forms, the grammar will account for this by means of morphosyntactic
and morphological rules. English is one such language. The question
that concerns us here is whether any or all of the labels that are assigned
to the expression-classes in the structural analyses of the sentences of
particular languages (NP, VP, etc.) are such that their assignment can
be determined, non-arbitrarily, on purely syntactic grounds, Why do we
say, for example, that ‘John’, ‘he’, ‘my father’, ‘that old man’, etc., are
members of the category NP and that ‘be (a) dentist’, ‘come home’,
‘love one’s wife’, etc., are members of the category VP? It is obvious
that, if we were to switch the assignment of these category-labels and
make the necessary consequential changes in the grammatical rules and
the lexicon, our generative grammar of English would still generate
exactly the same set of sentences. Moreover, unless we make appeal to
some more general definition of NP and VP, we cannot say that there is
any significant difference in the structural analyses assigned to the sen-
tences of English by the two grammars.
Several proposals have been made in the literature that are relevant to
this question. The first is that there is an intrinsic connexion between
the syntactic function of being the subject* of the sentence and the
syntactic category NP; and that there is similarly an intrinsic connexion
between the function of being the predicate* of the sentence and the
category VP. Chomsky’s proposals for a language-independent defini-
tion of deep-structure subjects and deep-structure objects depend upon
the assumption that there is such a connexion (cf. Chomsky, 1965). So
 11.1 Parts-of-speech, form-classes and expression-classes 429
too does much of what has been said by logicians in the Aristotelian
tradition.’ For the moment, we will neglect the logical interpretation of
such terms as ‘subject’ and ‘predicate’. ‘The question is whether there
is any purely syntactic definition of ‘subject’ and ‘predicate’ that can be
applied across languages to determine the assignment of NP and VP.
Before embarking upon the discussion of this question, we draw the
reader’s attention to the syntactic conditions specified in the two
dictionary-definitions cited above: “functioning as subject or object in
a construction”’ and “‘that function as the main elements of predicates’’.
As we shall see it is not nouns, but nominals, that function as subjects or
objects; and it is verbals, not verbs, that function as predicates. Whether,
and in what sense, verbs are the principal constituents (‘‘the main
elements’’) of verbals is a question we shall come to after we have
investigated the syntactic basis of the distinction between subjects and
predicates, on the one hand, and between subjects and objects on the
other.
It is generally accepted by linguists that, although the traditional
theory of the parts-of-speech (noun, verb, adjective, etc.) is inapplicable,
in all its details, to languages whose grammatical structure differs
significantly from that of the classical Indo-European languages, the
distinction between nouns and verbs at least is universal. Furthermore,
it is generally accepted that this distinction is intrinsically bound up
with the difference between reference and predication. Sapir made the
point in a well-known passage, as follows: “There must be something
to talk about and something must be said about this subject of discourse
once it is selected .. . The subject of discourse is a noun. As the most
common subject of discourse is either a person or a thing, the noun
clusters about concrete concepts of that order. As the thing predicated
of a subject is generally an activity in the widest sense of the word...
the verb clusters about concepts of activity. No language wholly fails to
distinguish noun and verb though in particular cases the nature of the
distinction may be an elusive one”’ (Sapir, 1921: 117). This passage, as
will be obvious, is rather loosely written: it uses the term ‘subject’ in at
least two senses (for both the referent and the referring expression) and,
what 1s more important, it fails to draw a distinction between nouns and
nominal expressions, on the one hand, and between verbs and verbal
expressions on the other. It does not follow from the fact that all lan-
guages draw a syntactic distinction between nominal expressions and
1 For a convenient summary of more traditional views, cf. Sandmann (1954).
Strawson (1974) contains much that is relevant.
 430 Semantics and grammar II
verbal expressions (if this is a fact) that they must also draw a distinction

discusses. :
between nouns and verbs. This point, as we shall see, is particularly
relevant in connexion with one of the languages (Nootka) that Sapir

It will be our principal aim, in the following section, to examine the

implications of what Sapir and many others have said about the univer-
sality of nominals and verbals, on the one hand, and of nouns and verbs,
on the other. In so far as any general grammatical definition of nouns
can be given, potentially applicable to all languages, it is as follows: a
noun is a lexeme which may occur as the sole or the principal open-class
constituent in a nominal and is syntactically or morphosyntactically dis-
tinguishable from other lexemes that function as open-class constituents
(i.e. verbs or adjectives) in the same positions of occurrence. The term
‘principal open-class constituent’ is rather vague. It would be simpler
to appeal, as many linguists would at this point, to the notion that the
noun is the head of the nominal. Unfortunately, this criterion will not
work for what we may regard as the most typical nominals (NPs) in
English: if the head of a construction is defined on the basis of endo-
centricity, ‘boy’ is not the head in ‘the boy’ (cf. 10.3). The term ‘open-
class constituent’ is intended to exclude such lexemes as definite
articles, demonstratives and classifiers, which occur in nominals in many
languages and might well be held to function, syntactically, as the heads
of the constructions in which they occur (cf. 11.4). Verbs, as we shall see
in the next section, cannot be defined in quite the same way: the rela-
tionship between verb and verbal is different from the relationship
between noun and nominal; and this fact too has tended to introduce
confusion into the discussion of the connexion between verbs and
predicates.

11.2. Subjects, predicates and predicators

Looked at from the syntactic point of view, the distinction between
subject and predicate, as it is usually explained, rests upon the assump-
tion that the nucleus* of a simple sentence (in any language for which
the distinction holds) is composed of two immediate constituents, one
of which is a nominal (NP) and the other a verbal (VP). For example,
‘John’, ‘he’, ‘my father’, ‘that old man’, etc., belong to one expression-
class, X, and ‘be (a) dentist’, ‘come home’, ‘cross the road’, etc.,
belong to a different expression-class, Y, in English. Sentences like
‘John is a dentist’ and ‘He loves his wife’, we will assume for the
moment, would be analysed in any generative grammar of English as
 11.2. Subjects, predicates and predicators 431
containing (in addition to such non-nuclear components as tense) an
expression of category X and an expression of category Y. But how do
we know which of the two expressions is the subject and which is the
predicate? Alternatively, how do we decide, on purely syntactic grounds,
that X is the class of nominals and Y the class of verbals?
It is obvious that these two questions are unanswerable unless we
make some further assumptions about the distribution or internal
syntactic structure of nominals and verbals throughout the languages of
the world. One such assumption might be that there can be more than
one NP, but only one VP, in the nucleus of a simple sentence; and more
specifically that an NP can occur as part of a VP. For example, the
English expression-class that we have labelled arbitrarily as Y can be
subdivided in terms of the internal structure of its members into several
subclasses. Two of these are traditionally distinguished as intransitive
and transitive, respectively; and they differ in that, whereas the mem-
bers of one subclass are composed of intransitive verbs, members of the
other subclass are composed of transitive verbs combined with an object
NP. What this means in effect is that an X can not only combine with a
Y to form the nucleus of sentence but can also combine with a Z to form
a Y (when Z 1s a transitive verb); and this is the syntactic basis of the
distinction between the subject of a verb and its object (or complement).
It can be used in any language that has syntactic constructions of this
kind, in order to assign the labels NP, VP and V in the constituent-
structure of sentences.
At first sight, the diagnostic procedure that we have just outlined
might appear to be blatantly circular; and so it would be, if (i) it could
not fail, logically, to yield a decisive result or (ii) there were no indepen-
dent criteria with which to evaluate the results. But neither of these
conditions holds. It is logically possible that there should be a language
in which, let us say, there are just two kinds of simple sentences, of
structure A+B and C+D-+E (where A, B, C, D and E are non-
intersecting classes of expressions). The proposed procedure would
obviously fail to identify either A or B as an NP in a language of this
kind (though it might well be possible to do so on semantic grounds).
For any language in which we can identify NP-expressions, VP-
expressions and V-expressions by means of the suggested diagnostic
procedure, we can go on to enquire whether these classes, so defined,
satisfy other syntactic or non-syntactic criteria. The procedure is not
therefore circular; and, in so far as it yields results which correlate
positively with any semantic definitions of subject and predicate, or of
432 Semantics and grammar II
nominal, verbal and verb, it would tend to support the traditional view
that the distinction between nominals and verbals is intrinsically con-
nected with the distinction between subject and predicate.
_ However, we do not need to assume the universal validity of the :
bipartite subject-predicate analysis of simple sentences in order to
identify the nominals and the verbs in a language; and it is important to
emphasize this point, in view of the fact that many linguists have
challenged the universality of the syntactic category VP. Given that we
have, in English, intransitive structures, X-+- Y, and transitive structures,
X+Z-+X, as well as several other sentence-nucleus structures con-
taining more than one X, the fact that X is the class of nominals and
that Y and Z are different (but overlapping) classes of verbs follows
directly from the more general diagnostic principle that there may be
more than one NP, but only one verb, in the nucleus of a simple sen-
tence. Indeed, it suffices that the X-constituent that combines with a
Y-constituent should have, in general, a wider distribution throughout |
the simple sentences of a language for us to be able to identify the
X-constituent as a nominal. We do not therefore have to establish or
assume the distributional identity of Z-+-X with Y.
At this point, we may look briefly at one of the languages that is most
frequently cited as an example of a language which has no parts-of-
speech (in the traditional sense of this term). The language is Nootka,
of which it is said that “normal words do not fall into classes such as
noun, verb, adjective, preposition, but all sorts of ideas find their
expression in the same general type of word, which is predicative or non-
predicative according to its paradigmatic ending”? (Swadesh, 1939:
78). The distinction between normal words and particles that is drawn
in the article from which this quotation comes is irrelevant in the present
connexion: we are concerned solely with what Swadesh calls normal
words. Now, it is.in the discussion of languages like Nootka, that it is
particularly important to maintain the distinctions that we have estab-
lished between forms, expressions and lexemes. The procedure that we
have outlined above, it will be recalled, enables us in principle to identify
nominals and verbs, and possibly verbals, but not nouns. So far we have
not said anything about the syntactic definition of nouns. The first
point that must be made, therefore, is that there appears to be no
problem about identifying asymmetrical bi-partite X-+-Y constructions
in Nootka in terms of their component expression-classes (cf. Sapir,
1921: 134). Forms of the X-constituent take a definite or deictic suffix,
whereas forms of the Y-constituent take a modal suffix in simple ©
 11.2. Subjects, predicates and predicators 433
declarative sentences. Furthermore, it appears that there can be more
than one X-constituent, but only one Y-constituent, in a simple sen-
tence. The procedure outlined above would, therefore, seem to apply
satisfactorily to Nootka (as also to Kwakiutl and other languages in the
same family); and Swadesh’s distinction between predicative and non-
predicative function corresponds to the distinction that we have drawn
between the verbal (VP) and the nominal (NP).
It is when we go on to try to draw the further distinction between
nouns and verbs as lexemes that we see why Swadesh said of Nootka
that “normal words do not fall into classes such as noun, verb, adjective,
preposition”’ (and Boas of Kwakiutl that “all stems seem to be neutral,
neither noun nor verb”’: cf. Boas, 1911). With the exception of certain
proper names and what we may refer to here (without attempting to
justify the labels) as various pronouns and adverbs (cf. Swadesh, 1939:
78), all lexemes may occur freely in either X-constituents or Y-con-
stituents. Since we have defined the verb in terms of its occurrence in
Y-constituents, all lexemes in Nootka (apart from those mentioned as
exceptions) would be classified syntactically, in terms of our criterion,
as verbs. To say that all the lexemes are syntactically verbs is tantamount
to saying that they are neutral with respect to the distinction of noun,
verb, adjective, etc.; i.e. that they all belong to the same part of speech.
Now it may well be that a more refined syntactic analysis of Nootka,
Kwakiutl, and languages of similar structure, would bring out various
differences in the co-occurrence of subclasses of lexemes which might
lead us to reconsider this verdict (and there is the outline of what is
described as a “semantic classification’’ in Swadesh’s analysis which is at
least suggestive in this respect). Let us grant, however, that it is at least
plausible that there should be languages in which no syntactic distinc-
tion can be drawn between nouns and verbs.
The reason why we say that in English and in the vast majority of the
world’s languages there is a grammatical difference, not only between
nominals and verbals, but also between nouns and verbs, is that in such
languages there are distributional differences between the two classes of
lexemes in question with respect to their occurrence in nominals and
verbals respectively. In English, for example, we can say The woman 1s
coming and The one who 1s coming ts a woman: we cannot say (as we might
if English were like Nootka) The coming is (a) woman (or The come
womans). Nouns can occur in verbal expressions, but, when they do,
they require to be combined with what we will describe below as a
. copulative* verb and, if they are countable nouns in the singular, they
434 Semantics and grammar II
must be preceded by the indefinite article; conversely, verbs can occur
in nominal expressions, but, when they do, they must be incorporated in
a relative clause or alternatively be used, in their participial form, as the
modifier of a noun (i.e. adjectivally). In short, the distinction between
nouns and verbs in English is supported by a variety of syntactic and
inflexional differences. We have no difficulty, therefore, in saying that
rains is a verb-form in ‘It rains a lot in the Highlands’ whereas razn is a
noun-form in ‘There is a lot of rain in the Highlands’. But the distinc-
tion between the verb ‘rain’ and the noun ‘rain’ could not be drawn, it
should be noted, on purely semantic grounds. We will come back to this
point later. The two sentences that we have just given are of a kind that
is particularly interesting in connexion with the alleged universality of
the subject—predicate analysis of sentences. As we shall see, there is an
alternative analysis of the underlying syntactic structure of a sentence
like ‘It rains a lot in the Highlands’, which does not depend upon the
assumption that the nucleus of every simple sentence is necessarily
composed of at least one nominal and a verbal.
So far we have assumed that the notion of a predicate is necessarily
bound up with the bipartite analysis of sentence-nuclei in terms of their
immediate constituents, one of these constituents being the subject and
the other the predicate. That all sentences (or at least all non-elliptical
simple declarative sentences) can be divided exhaustively in this way
into a nominal subject and a verbal predicate is a view that is strongly
represented in traditional grammar. It also has its correlate in what we
have referred to loosely as Aristotelian logic. There is, however, an
alternative analysis of the structure of propositions that is formalized, as
we have already seen, in the first-order predicate calculus (cf. 6.3).
According to this view, the predicate is an operator with one or more
arguments: an intransitive verb is formalized as a one-place operator
which takes an NP as its sole argument; a transitive verb is a two-place
operator which relates one NP to another, and so on, The term ‘predi-
cate’ is sometimes used in linguistics in much the same sense. In order
to avoid unnecessary confusion, however, we will introduce the term
predicator* to bear this rather different sense of ‘predicate’. We can say
that ‘play’ in ‘Caroline plays the guitar’ is a two-place predicator
independently of whether we also say that ‘play the guitar’ is a predi-
cate. According to this conception of the syntactic structure of sentences
(which we shall look at below in connexion with the notion of valency*:
cf. 12.4) the predicator is an element which combines with a single NP
or relates an NP to something which may or may not be an NP: it is the
 11.2. Subjects, predicates and predicators 435
pivot, as it were, of the sentence-nucleus.? In certain systems of syn-
tactic analysis, the verb is taken to be the pivot upon which all other
constituents of the sentence-nucleus depend and by which they are
determined.
What we have referred to as the pivotal status of the verb correlates,
in many languages, with several more particular syntactic phenomena
that are handled traditionally in terms of concord* (or agreement) and
government*. The first of these is explicitly mentioned in the definition
that we have quoted above: verbs tend to “show agreement with sub-
ject or object’’. The second is perhaps implied when it is said that verbs
‘function as the main elements of predicates’. Neither concord nor
government, as these notions are traditionally understood, is found in all
languages: but they are both very widespread. That the verb must
agree with either the subject or the object (in number, gender, person,
etc.) and that it governs its object (in terms of case or the selection of a
particular preposition) is a statement that figures in the grammatical
descriptions of very many unrelated languages. This point may be
illustrated in relation to a set of three kinds of sentence-nuclei, all of
which are distinguishable, on purely syntactic or morphosyntactic
grounds, in many unrelated languages which otherwise differ consider-
ably in their grammatical structure.
The set of nuclear structures (to which we will add others in the next
chapter) is as follows:

(1) NP+V (intransitive)

(2) NP+V+NP (transitive)
(3) NP(+V)+N (ascriptive)
As far as English is concerned, these structures are exemplified by sen-
tences such as the following:
(1a) That boy works (hard) (nowadays) (at school)
(2a) Caroline plays the guitar (in the evening)
(3a) He is an American.
We will temporarily disregard what is enclosed in parentheses in sen-
tences (1a)-(3a): all that needs to be said about them here is that they
are assumed to be adjuncts* (i.e. syntactically optional or peripheral

2 This sense of ‘pivot’ is not intended to be directly relatable to the sense in

which the same term has been employed in Braine (1963) and other recent
work in language-acquisition. But there may be some connexion.
436 Semantics and grammar IT
expressions); and it will be observed that they do not correspond to any
symbol in the formulae for sentence-nuclei in (1)—(3)..
If we look at (1)-(3), we see that every formula has an NP as its left-
most constituent; that (1) differs from both (2) and (3) in that it is
composed of only two constituents; and that (2) differs from (3) in that
the verb-symbol is bracketted in (3), but not in (2), and the right-most
constituent of (3) is N, rather than NP. On the basis of our previous
discussion, we can identify the expression which combines with the NP
in (1) as a verbal, and consequently the lexeme ‘work’ as a verb, since
this lexeme can function, without being combined with any other
lexeme, as a verbal. It is therefore identified as a verb on exactly the
same grounds as we classified all so-called normal lexemes in Nootka as
verbs. The reason why ‘play’ in (2a) and ‘be’ in (3a) are also classified
as verbs is what we are now concerned with; and we will also explain
why (3a) 1s said to have a noun, rather than a nominal, as its right-most
constituent. We cannot of course say that ‘play’ and ‘be’ are the main
elements of the predicate, if V-+NP and (V)-+N are not recognized as
predicates; and, if they are recognized as predicates (i.e. as verbals
distributionally equivalent with ‘work’ in (1a)), we must explain what
is implied by saying that the verb is the main element in cases where the
predicate is a composite expression.
In English the second element of the structures given in (1)-(3) must
agree in terms of number and person (in certain tenses) with the left-
most NP; and, if the NP in the third position of (2) is a personal
pronoun it is governed in the object-case by the predicator in the second
position (the form zm rather than he occurs, them rather than they,
etc.). Furthermore, as in many (though not all) languages, tense* (which
is not in itself a sentence-nucleus category: cf. 15.4) and to some extent :
mood* (cf. 16.2) are realized by inflexion of the predicator. It is pheno-
mena of this kind which lead us to say that ‘play’ in (2a) and ‘be’ in (3a)
are verbs: they are pivotal with respect to concord and government and
the inflexional realization of tense, in the same way that ‘work’ in (1a)
is; and ‘work’ is by definition a verb, since its most characteristic func-
tion is that of a one-place predicator.
If the notion of government is extended to cover not only the selec-
tion of particular cases in the traditional sense of the term ‘case’
(nominative, accusative, genitive, etc.), but also the selection of par-
ticular prepositions (or postpositions) and particular kinds of sub-
ordinate clauses, it is arguable that government (though not concord) is
to be found in all languages; and what is nowadays referred to as case-
 11.2. Subjects, predicates and predicators 437
grammar* (cf. 12.4), in so far as it is intended to provide a universal
framework for syntactic analysis, depends upon this assumption. How-
ever that may be, phenomena of the kind that we referred to in the
previous paragraph will serve, for the present, to illustrate what is
implied by saying that the verb, in many languages at least, is the pivotal
element in the nucleus of simple sentences.
We will not go further into the distinction between intransitive and
transitive verbs, except to say that it is by no means as straightforward
as might appear from our somewhat superficial discussion of English.
The whole basis of the distinction, and with it the necessity of drawing
far more distinctions than are traditionally recognized in the gram-
matical descriptions of English and other languages, has been extensively
discussed in the recent literature. It suffices, for our purpose, that in
very many languages at least, a subclass of the one-place predicators and
a subclass of the two-place predicators can be identified syntactically as
verbs in terms of criteria of the kind that we have discussed above.
We will now look more closely at the structure that we have called
ascriptive (3). Two points are worth noting about this structure: (i) the
fact that the verb-symbol has been put in brackets; and (ii) the fact that
it is N rather than NP that occurs in the third position. The reason why
~ we have put the verb-symbol in brackets in (3) is that there are many
languages in which structures of this kind lack any element that would
be classified as a verb (comparable with the verb ‘be’ in English). That
there are such verbless sentences in several of the world’s languages -
invalidates the assumption that the verb 1s an indispensable element of
the sentence in all natural languages. ‘The lexeme ‘be’ is classified as a
verb in English because, with respect to concord and the realization of
tense, it is pivotal in the way that ‘work’ and ‘play’ are pivotal. Given
that such lexemes as ‘work’ and ‘play’ are verbs and that they are
inflected for such morphosyntactic categories as person, tense and
number, ‘be’ is also a verb with respect to any rules in the grammar
which account for the distribution of the inflexional forms of verbs. It is
important to realize, however, that these rules are morphological and
morphosyntactic rather than purely syntactic. It is not its copulative
function as such that makes ‘be’ a verb (cf. 12.2). If we discount the
so-called absolute existential use of ‘be’, which is more or less confined
to theological and philosophical writings (cf. the Biblical J am who am
and the Cartesian I think: therefore I am) and is parasitic upon the
structure of other languages, there is no convincing syntactic or semantic
reason for classifying ‘be’ in English as a verb.
438 Semantics and grammar II
The reason why the form an American has been treated as being the
form of a noun (N) rather than of a nominal (NP) is that it is being
assumed that there is here no underlying referring expression, ‘an
American’, as there is for example in ‘She married an American’, If it
were not for the fact that English, unlike many languages, puts an
indefinite article with countable nouns when they occur in the position
occupied by N in (3), ‘American’ would be classifiable as either a noun
or an adjective: i.e. ‘He is American’ would be syntactically ambiguous,
as ‘He is French’ and ‘He is Frenchman’ would not.
What has been said in this section should not be taken to imply that
the formulation of diagnostic procedures for the definition of nominals,
verbals and verbs in purely syntactic terms is, of itself, a worthwhile
pursuit. The theoretical interest of the endeavour is that, in so far as
such diagnostic procedures can be formulated and yield definite results,
the expression-classes and parts-of-speech which they establish can be
examined to see whether their members satisfy independently applicable
semantic criteria (based on such notions as reference and predication, or
the distinction between entities, properties, actions, relations, etc.). The
fact that there appears to be a positive correlation in all languages
between syntactically defined and semantically defined expression-
classes would tend to support the traditional view that there is a high
degree of interdependence between the syntactic structure of sentence-
nuclei and the semantic function of their constituent expressions. Des-
pite what has been said at times by certain linguists there is no reason to
doubt that the traditional view is, to this extent at least, well-founded.®
On the other hand, it must be emphasized that this correlation between
syntax and semantics is not perfect; and, as we shall see, what is from a
semantic point of view indeterminate may be determined syntactically
to be a member of one part-of-speech or expression-class rather than
another. To this extent the thesis that our ontology is determined by the
language that we speak may not be without foundation either.

11.3. The ontological basis: entities, qualities and actions

In this section, we shall be concerned mainly with the possibility of
defining nouns, verbs and adjectives in semantic terms. Something will
also be said about adverbs.
Semantic definitions of syntactic categories rest, in part, upon such
3 Reaction against the traditional view reached its peak in the period imme-
diately following the Second World War: cf. Firth (1957a), Fries (1952),
Harris (1951), Joos (1957).
 11.3. The ontological basts 439
notions as reference, predication and denotation; and we will take these
notions for granted in what follows. We will also take for granted the
semantic relevance of the syntactic relation of modification and its
connexion with predication. There are problems attaching to the notion
of modification. Since they do not affect the argument, we will adopt the
conventional view, according to which the attributive adjective is the
modifier of the noun with which it is combined, and the adverb is the
modifier of the verb or adjective with which it is combined, in endo-
centric expressions. There are many subclasses of adverbs and some
adjectives for which this statement is definitely not valid; and there are
other adverbs and adjectives for which its validity is questionable. In so
far as the generalization that has just been made does hold, however, it
explains the traditional terms ‘adjective’ and ‘adverb’: the adjective 1s
typically the modifier of a noun and the adverb is typically the modifier
of a verb or adjective.
More important in the present connexion, it is the basis for what such
scholars as Hjelmslev (1928) and Jespersen (1929) have seen as a
difference of rank* (in a particular sense of this term):* one expression,
e;, can modify another expression, e;, only if the modifying expression
is lower than, or equal to, the modified expression in terms of its rank.
Looked at from this point of view, nouns are said to be of higher rank
than verbs and adjectives, and adverbs of lower rank than verbs and
adjectives. This notion of rank will not be elaborated further. It suffices,
for our present purpose, to point out that there is a correlation between
the ranking of the parts-of-speech in terms of what they can modify and
their semantic definition in terms of what they denote. Nouns are
traditionally said to denote entities, verbs and adjectives to denote what
we may refer to here as first-order properties (in a very general sense of
the term ‘property’), and adverbs to denote second-order, or even
higher-order, properties. This distinction between first-order properties
and second-order properties 1s implicit in the formalization of predicate-
calculus and standard interpretations of it: first-order properties may be
ascribed to individuals (i.e. first-order entities), second-order properties
may be ascribed to first-order properties, and so on (cf. 6.3).5
4 ‘The term ‘rank’ is used in a different sense by Halliday (1961): cf. Huddles-
ton (1965), Lyons (1968: 206).
5 The notion of syntactic rank is made explicit in so-called categorial grammars,
which many logicians now favour for the semantic and syntactic analysis of
natural languages (cf. Lewis, 1972; Montague, 1974). For an accessible
account of the principles of categorial grammars, which goes back through
Ajdukiewicz to Legniewski, cf. Bar-Hillel (1964). What a categorial grammar
440 Semantics and grammar II
The semantic part of the traditional definitions of the parts-of-speech
presupposes the possibility of identifying entities, properties, actions,
relations, etc., independently of the way in which these are referred to or
denoted in particular languages: i.e. it presupposes the acceptance of
some neutral ontological framework. For example, unless we can identify
persons and things independently of their being denoted, in particular
languages, by lexemes that we wish to call nouns, it will not do to say in
the definition of ‘noun’ that nouns denote persons and things. Semantic
definitions of the parts-of-speech, if they are interpreted as giving the
necessary and sufficient conditions for the membership of particular
parts-of-speech, are readily shown to be either circular or inapplicable
in a vast number of instances; and this is one of the principal criticisms
that has been made of such definitions. For example, if the only reasons
that we have for calling beauty a thing are that ‘beauty’ is a noun and
that such utterances as Beauty is a wonderful thing are normal in English,
we cannot say that the reason why ‘beauty’ is a noun is that it denotes a
thing. It is no part of our purpose to defend the indefensible. The thesis
that will be maintained here is that the semantic, or ontological, part of
the traditional definitions of the parts-of-speech define for each part-of-
speech, not the whole class, but a distinguished subclass of the total
class. Each such semantically defined subclass is focal within the larger
class in much the same way that, according to the Berlin and Kay
hypothesis that we looked at earlier (8.3), a particular area within the
total area denoted by a colour term is focal.
It is because there is an intrinsic connexion between syntax and
semantics with respect to the definition of the focal subclasses, which
contain the most typical nouns, verbs, adjectives and adverbs, that we
can ask sensibly whether all languages have nouns, verbs, adjectives and
adverbs. When we say that there are adjectives, for example, in such-
and-such a language we mean that there is a grammatically definable
class of expressions whose most characteristic syntactic function is that
of being the modifier of the noun in an endocentric construction and
whose most characteristic semantic function is to ascribe properties to
entities. It does not follow from the statement that two languages both
have adjectives that every adjective of the one language is translatable by
an adjective in the other language, and conversely. Nor does it follow, as

defines to be a syntactic category (whether basic or derived) was described by

the originators of categorial grammar as semantic. The reason for this is that
categorial grammar, as it is now called, was originally developed as a tool for
logical analysis.
 11.3. The ontological basis 441
we have already seen, that all (or indeed any) of the adjectives in either
language will be lexemes: it is in principle possible that some (or indeed
all) of them should be formed by productive grammatical processes
from lexemes belonging to other parts-of-speech. If we allow that what
has been said about Nootka and such languages is correct, it is con-
ceivable that there should be nouns, verbs and adjectives in a language
without there being any parts-of-speech (as we have defined ‘part-of-
speech’) in that language. Nouns, verbs and adjectives might be defin-
able as members of syntactically and semantically distinct expression-
classes; and yet the items listed in the lexicon might be neutral with
respect to this tripartite classification. How this might be handled will
be made clearer in our discussion of the lexicon in a later chapter (13.2).
In what follows here we will not explicitly take into account the possi-
bility that there might be expression-classes, none of whose members are
lexemes.
The semantic, or ontological, parts of the dictionary-definitions for
‘noun’ or ‘verb’ quoted above are “designating persons, places, things,
states, or qualities’’ and “[which] express action or state’’, respectively
(cf. 11.1). Since there is no reason to. believe that ‘designate’ and
‘express’ are being used in such a way that the designation of a state
would be recognizably different from the expression of a state, we note
immediately that both nouns and verbs are related to the ontological
category of state by what appears to be the same semantic relation,
which we are calling denotation (cf. 7.4). As speakers of English, we can
readily imagine that what the compilers of the dictionary had in mind
was the existence in English of such lexemes as ‘peace’ and ‘know’,
each of which can be said to denote a state, though they belong to
syntactically and morphosyntactically distinct expression-classes.
More traditional definitions of the noun make no reference to either
states or qualities, but only to persons, places and things; and the more
traditional definitions of the verb make no mention of states, but only of |
actions. It is, in fact, adjectives that are traditionally said to denote
states and qualities; and their connexion with states is frequently not
mentioned or is seen as secondary. For the present, therefore, we will
neglect states: the fact that it is difficult sometimes to distinguish states
from qualities, on the one hand, and from actions and processes, on the
other, would suggest that states have a certain ontological ambivalence;
and this is reflected in the various ways in which they are given lexical
or grammatical recognition in languages. As for actions, all that needs
to be said at this stage is that the traditional term ‘action’ must be
442 Semantics and grammar IT
construed rather broadly, so that it covers, not only the acts and activi-
ties of responsible agents, but also such events and dynamic processes as
cannot be attributed to agents. We shall see later that neither the
logician’s term ‘property’ nor the grammarian’s term ‘action’ is suit-
able for the purposes to which they are commonly put (cf. 12.4).
Finally, it should also be mentioned that nothing will be said, or
implied, in this section about any connexion between nouns and places.
As we shall see later, places, like states, are ontologically ambivalent, and
in a more interesting way than states are (cf. 12.3). The ontological
categories with which we shall be initially concerned, then, are those
which comprise persons, things, actions (including events and processes),
and qualities.
We obviously cannot operate with categories of this kind without
making some minimal ontological assumptions: i.e. assumptions about
what there is in the world. The ontological assumptions that we will
make (and we will take them to be minimal and relatively uncontrover-
sial) are those of naive realism. Our first and most basic assumption is
that the external world contains a number of individual persons,
animals and other more or less discrete physical objects (cf. 6.3). That
it is difficult to draw the line precisely between what counts as a discrete
physical object and what is not is unimportant, provided that it is
possible to identify a sufficient number of what are indisputably in-
dividual physical objects: it is the lexical and grammatical structure of
particular languages that draws the line for us in the unclear instances
(e.g., with respect to the ontological status of mountains, rivers, etc.).
Physical objects are what we will call first-order entities*. Within the
class of first-order entities persons occupy a privileged position; and the
distinction between persons and non-personal entities is lexicalized or
grammaticalized, in various ways, in many, and perhaps all, languages.
It may be observed, in passing, that the distinction between persons and
non-personal entities is often represented, and arguably misrepresented,
by linguists as a distinction between human and non-human entities:
the alleged semantic deviance of sentences like ‘His typewriter has bad
intentions’ depends upon the fact that typewriters are not normally
categorized as persons (i.e. as entities to which one ascribes conscious-
ness, intention and will); and in order to impose an interpretation.on any
utterance of this sentence we have to personify, rather than humanize,
typewriters.® It should also be noted that there would seem to be
6 The parenthetical gloss in the text, ‘‘ entities to which one ascribes conscious-
ness, intention and will’’, is intended to be no more than indicative of what |
 11.3. Lhe ontological basis 443
operative in many languages, if not in all, a hierarchy within the
classificatory scheme that is employed to describe or refer to first-order
entities such that persons are more strongly individualized than animals,
and animals more strongly individualized than things (cf. 11.4). Al-
though we will work throughout this section with a notion of first-order
entities which draws no distinction between them in respect of their
ontological status, it must not be forgotten that this hierarchical dif-
ferentiation does exist, and it may be of considerable importance when
it comes to the description of the grammatical and lexical structure of
particular languages. However, it is characteristic of all first-order
entities (persons, animals and things) that, under normal conditions,
they are relatively constant as to their perceptual properties; that they
are located, at any point in time, in what is, psychologically at least, a
three-dimensional space; and that they are publicly observable (cf.
Strawson, 1959: 39ff).’ First-order entities are such that they may be
referred to, and properties may be ascribed to them, within the frame-
work of what logicians refer to as first-order languages (e.g., the lower
predicate-calculus: cf. 6.3).
The ontological status of what we will call second-order and third-
order entities 1s more controversial; and it may well depend crucially
upon the structure of the languages that we use to talk about them. We
shall have more to say about second-order and third-order entities later;
and we shall restrict ourselves here to drawing the distinction in general
terms. By second-order entities* we shall mean events, processes, states-
of-affairs, etc., which are located in time and which, in English, are said
to occur or take place, rather than to exist; and by third-order entities* -
we shall mean such abstract entities as propositions, which are outside
space and time. This distinction between three kinds of entities is such
that it corresponds only in part with the traditional distinction between
concrete and abstract entities, upon which the classification of nouns

is involved in personification. Clearly, the attribution of feelings and inten-

tions to animals or machines does not of itself constitute personification in the
full sense; and to say that ‘ This typewriter has bad intentions’ is interpretable
only under the assumption that the typewriter being referred to, on some
occasion of the utterance of this sentence, has been fully personified is per-
haps too strong. The most striking evidence of personification, as far as
English is concerned, would be in the selection of ‘who’, rather than ‘which’,
for the relative pronoun: e.g. ‘this typewriter, who has bad intentions .. .’
(cf. Quirk et al., 1972: 861).
’ "They are what Strawson refers to as basic particulars. The influence of
Strawson (1959) will be evident throughout this section.
444. Semantics and grammar IT
and nominals depends. Second-order entities, though they may be
denoted by what are traditionally called abstract nouns, are clearly not
abstract in the sense that something that has no spatiotemporal location
‘is abstract.
Second-order entities differ from first-order entities in several ways. |
In certain languages at least, they may be identified and referred to as
individuals. For example, in English we can no less readily say Fust look
at that sunset (or Just look at the sun setting) than we can say Just look at
that dog. But second-order entities are much more obviously perceptual
and conceptual constructs than first-order entities are; the criteria for
re-identification are less clear-cut, and the ability to refer to them as
individuals depends, to some considerable degree, upon the gram-
matical process of nominalization. That the criteria for re-identification :
are less clear-cut is obvious, if we contrast, from this point of view, the
following two utterances:
(1) The same person was here again to-day
(2) The same thing happened again to-day.
The reference of ‘the same person’ is constrained by the assumption of
spatiotemporal continuity and by the further assumption that the same
person cannot be in two different places at the same time. Either or both
of these assumptions may be suspended in particular contexts; but, in
general, they are simply taken for granted within the metaphysical
framework of naive realism.
This is not true, however, with respect to second-order entities. The
same event can occur or be occurring in several different places, not only
at different times, but at the same time. What this means, in effect, is
that there is no sharp distinction to be drawn between an individual
situation and a generic situation. There is no clear semantic distinction,
in other words, between ‘the same situation’ and ‘the same kind of
situation’, as there so obviously is between ‘the same person’ and ‘the
same kind of person’; and ‘the same thing’ will go like ‘the same
situation’ or ‘the same person’ in this respect, according to whether it
refers to a first-order or second-order entity. We can adopt a more or
less inclusive notion of what constitutes individual identity in the
interpretation of the referring phrase ‘the same thing’ in (2). On the
maximally inclusive interpretation, according to which every individual
situation is unique and unrepeatable, (2) expresses a logical contradic-
tion.
The distinction between second-order and third-order entities (both
 11.3. The ontological basis 445
of which would be traditionally described as abstract) is no less impor-
tant, semantically, than is the distinction between first-order and
second-order entities. Whereas second-order entities are observable and,
unless they are instantaneous events, have a temporal duration, third-
order entities are unobservable and cannot be said to occur or to be
located either in space or in time. Third-order entities are such that
‘true’, rather than ‘real’, is more naturally predicated of them; they can
be asserted or denied, remembered or forgotten; they can be reasons,
but not causes; and so on. In short, they are entities of the kind that may
function as the objects of such so-called propositional attitudes* as
belief, expectation and judgement: they are what logicians often call
intensional objects. Reference to both second-order entities and third-
order entities is made most commonly, both in English and in other
languages, by means of phrases formed by the process of nominaliza-
tion. But there is a fairly clear difference in English between the set of
nominalizations that is appropriate for the one purpose and the set of
; nominalizations that is appropriate for the other (cf. Vendler, 1968).
First-order entities we take to be more basic than either second-order |
or third-order entities in that their ontological status is relatively un-
controversial and the process of nominalization, which is used to form
nominals that refer to second-order and third-order entities, operates
characteristically upon sentence-nuclei that contain nominals whose
reference is to first-order entities. For example, ‘John’s arrival’ (which
belongs to a subclass of derived nominals that can refer to either second-
order or third-order entities) in such utterances as I witnessed Fohn’s
arrival or John’s arrival has been confirmed is transformationally relatable
to the nucleus of ‘John arrived’. To say that something is an entity is to
say no more than that it exists and can be referred to; and we will
assume that the notion of existence applies primarily to first-order
entities and that what is traditionally referred to as the hypostatization*
of higher-order entities depends crucially upon the structure of
particular languages. | ,
Looked at from a semantic point of view, nominals are referring
expressions. To be more precise, they are expressions which have a
certain potential for reference and, when they occur in utterances, are
invested with reference by the utterer (cf. 7.2). The intra-propositional
relations that nominals contract with other expressions in the nuclei of
sentences are, we assume, intrinsically connected with their referential
function in the most characteristic use of such sentences. As we can
distinguish first-order, second-order and third-order entities, so we can
446 Semantics and grammar IT
distinguish first-order, second-order and third-order nominals in terms
of their characteristic referential function. Proper names, pronouns and
descriptive noun-phrases that are used, characteristically, to refer to
first-order entities may be described as first-order nominals*. Similarly,
expressions that refer, characteristically, to second-order entities may be
called second-order nominals* ; and expressions that refer, characteris-
tically, to third-order entities may be called third-order nominals*.
Whether a language has second-order and third-order nominals and, if
so, how they differ, grammatically, from first-order nominals is clearly
a matter for empirical investigation.
Given that nouns, both as lexemes and as more complex expressions,
may be identified in terms of their occurring as open-class constituents
in particular positions of occurrence in nomuinals (and their being
syntactically or morphosyntactically distinguishable from other open-
class constituents: cf, 11.1), they can be classified as first-order, second-
order and third-order nouns according to whether they occur (in
specifiable positions of occurrence) in first-order, second-order or third-
order nominals. What are traditionally referred to as common concrete
nouns (e.g., ‘boy’, ‘cat’, ‘table’, in English) are by this criterion first-
order nouns: they are lexemes that denote classes of first-order entities;
and, as such, they are what we are taking to be the most typical nouns.
Most second-order nouns and third-order nouns in English (e.g.,
‘arrival’, ‘death’, ‘amazement’, ‘house-keeping’) are complex or com-
pound, rather than simple (cf. 13.2); and this may well be true in all
languages that have second-order and third-order nouns (cf. Kahn,
1973: 76ff). Examples of what are presumably to be regarded as simple
second-order and third-order nouns are ‘event’, ‘process’, ‘state’ and
(in the intended sense) ‘situation’, on the one hand, and ‘reason’,
‘proposition’, ‘theorem’ and (in certain uses) ‘idea’, on the other. Some
of these words obviously originated in what was once (in Latin or Greek)
a productive process of nominalization.
We will not enter here upon a full-scale classification of nouns and
nominals; and nothing more will be said at this point about either
second-order or third-order nominals, It must be emphasized, however,
that our threefold classification is not intended to be exhaustive.
Nothing has been said, for example, about the ontological status of
numbers, sets, etc., or about the expressions that refer to such entities:
logicians have been much concerned with this question, but it is of
secondary importance for the semanticist whose main interest lies in
describing the structure of natural languages. No attempt has been made
 11.3. The ontological basts 447
-to draw a distinction between various kinds of third-order entities:
between psychological and non-psychological entities; between com-
municable and non-communicable entities; and so on. Distinctions of
this kind must clearly be drawn if we are to use terms like ‘fact’ and
‘proposition’ (not to mention ‘idea’, ‘feeling’, ‘sentiment’, etc.) with
any degree of precision.
In what follows, the only nouns with which we shall be concerned are
first-order nouns. It is our assumption (although this does not follow as
a logical consequence of anything that has been said in this section) that
no language will have second-order or third-order nouns that does not
also have first-order nouns. Whether first-order nouns, in the languages
in which they exist, differ syntactically or morphosyntactically from
other subclasses of nouns is something that varies from one language to
another. In particular, languages vary as to whether, and how, they
grammaticalize the distinction between proper names and common
nouns, on the one hand, and between count nouns and mass nouns, on
the other (cf. 11.4).
It may now be objected that, not only first-order nouns, but what are
traditionally classified as qualitative adjectives (which we are taking to
be the most typical adjectives), denote classes of first-order entities. As
we have already seen, the distinction between properties and classes is
one that many logicians reject (cf. 6.4). But we will assume that, within
the framework of naive realism, it is possible to draw a distinction, at the
extremes at least, between the relatively simple perceptual properties
which are distributed among individuals and the more complex con-
junctions and disyunctions of properties in terms of which individuals
are categorized as members of particular classes (cf. Strawson, 1959:
168ff). At one extreme, we have properties like redness, roundness or
solidity; at the other extreme, we have whatever might be the conjunc-
tions or disjunctions of properties in terms of which we categorize
entities into what are traditionally called natural kinds* (cf. Putnam,
1970). But there is much that falls between these two extremes; and any
purely semantic definition of ‘noun’ and ‘adjective’ that is based on the
difference between property-denoting and class-denoting expressions is
correspondingly weakened.
It would seem, in fact, that qualitative adjectives fall, semantically, |
between the most typical nouns and the most typical verbs; and in
particular languages they may be assimilated, grammatically, to either
nouns or verbs. In Latin, for example, nouns and adjectives are much
more similar from a grammatical point of view than they are in English.
448 Semantics and grammar II
In Chinese, on the other hand, adjectives may be regarded as a subclass
of verbs (cf. Kratochvil, 1968: 113). The term ‘adjective’, as we have
already seen, implies the primacy of syntactic considerations in the
definition of the part-of-speech or expression-class that it denotes; and
it is interesting to note that, whereas the dictionary-definitions of ‘noun’
and ‘verb’ quoted above (11.1) each includes a semantic condition, the
definition of ‘adjective’ in the same dictionary is purely syntactic (‘‘any
member of a class of words functioning as modifiers of nouns, as
‘good’, ‘wise’, ‘perfect’: Urdang, 1968). Adjectives are lexemes or
other expressions whose most characteristic feature is that they can
occur more freely than any other open-class expressions as modifiers of
nouns within nominals: hence our use of the term ‘adjectivalization’ for
any transformational process that converts a predicative expression into
a noun-modifying expression within a nominal (cf. 10.3). The standard
transformationalist view, for English and for other languages, that
nominals containing attributive adjectives are derived, in general, by
means of an embedding transformation has the advantage that it enables
us to account for the semantic relationship between all kinds of attribu-
tive and predicative expressions in the same way; and we will accept this
view. But it may be assumed that the embedding of a quality-denoting
expression is more normal than the embedding of either a class-denoting
or an action-denoting expression. There is a connexion, therefore,
between the semantic and the syntactic definition of the most typical
adjectives; and we should be surprised, to say the least, if we came
across a language in which quality-denoting expressions could occur in
predicative, but not attributive, position, whereas the most typical nouns
and the most typical verbs could occur freely in both positions,
We have now discussed in sufficient detail, for the present purpose,
the traditional semantic criteria in terms of which we might distinguish
the most typical nouns, verbs and adjectives in particular languages:
concrete common nouns, action-denoting verbs and qualitative adjec-
tives. In relation to these three subclasses of nouns, verbs and adjectives
the semantic criteria traditionally invoked are applicable without evident
circularity; and it suffices that we can define semantically what we are
taking to be the most typical nouns, verbs and adjectives for us to be
able to enquire, in respect of any particular language, whether it
grammaticalizes the distinction between any two or all three of these
parts-of-speech or expression-classes. Given that a particular language
distinguishes only two rather than three of the expression-classes that
we have been concerned with, it is predictable that these will be such
 11.3. Lhe ontological basis 449
that class-denoting expressions are distinguished from action-denoting
expressions (and quality-denoting expressions grouped with either the
one or the other), rather than that class-denoting and action-denoting
expressions are grouped together and sharply distinguished from quality-
denoting expressions. This, presumably, is one part of what linguists,
like Sapir and Bloomfield, had in mind when they said, rightly or
wrongly, that all languages draw a distinction between nouns and verbs.
In so far as the semantic criteria with which we have been operating
are logically independent of the morphological, morphosyntactic and
syntactic criteria referred to in earlier sections of this chapter, it is an
empirical question whether there will be any positive correlation
between grammatically defined and semantically defined expression-
classes in particular languages. The answer to this question would seem
to be that, in all languages that have been investigated and reported
upon, there is a correlation between the grammatical and the semantic
classification of expressions. Furthermore, the fact that there is a high
degree of correlation between the grammatical and the semantic classifi-
cation of expressions obviously facilitates the child’s acquisition of his
native language (cf. 12.4). But the correlation need not be perfect; and,
more important, what is ontologically indeterminate may be determined
differently by the grammatical categories of particular languages. For
example, states may be grouped with actions, with qualities or with
entities; and this accounts for the fact that, even in languages in which
adjectives are distinguished grammatically from both nouns and verbs,
particular nouns or verbs as well as adjectives, may be stative* (cf.
‘peace’ and ‘know’). It follows that the grammatical structure of lan-
guages may be partly, though not wholly, determined by semantic
distinctions; and that semantic distinctions of the kind that are relevant
to the definition of parts-of-speech and expression-classes may be them-
selves determined by ontological distinctions that are, in part at least,
independent of the structure of particular languages. )
We will not pursue this question further. It should be added, how-
ever, that a more detailed subclassification of expression-classes in
particular languages will frequently reveal other correlations between
semantically defined and grammatically defined subclasses. Looked at
from a purely grammatical point of view, proper names may or may not
be distinguishable from common nouns: semantically, they function as
entity-referring, rather than class-denoting, expressions. Mass nouns are
closer, semantically, both to qualitative adjectives and to proper names
than countable nouns are; and, as we have already seen, stative verbs are
 450 Semantics and grammar LI
closer to qualitative adjectives than action-denoting verbs are. In a more
comprehensive account of the semantic definition of syntactic categories
a multiplicity of points like these would need to be discussed. So too
would the fact that, in English and other languages, there is a correlation
between the syntactic subclassification of adjectives in terms of their
normal order of occurrence in nominals and their semantic sub-
classification in terms of the kind of qualities they denote (e.g., age,
colour, shape, material, etc.).®
It now remains to refer very briefly to adverbs. As we have already
seen, the traditional view is that the adverb is typically or characteris-
tically the modifier of a verb or adjective, or of another adverb; that it is
a tertiary category whose function it is to modify either a secondary or
another tertiary. However, it has long been recognized by grammarians
that there are many syntactically and semantically distinguishable sub-
classes of what are conventionally classified as adverbs, and that several
of these subclasses are such that their members cannot be said to modify
an adjective, a verb or another adverb.
One class of adverbs that does satisfy the traditional syntactic cri-
terion is exemplified by such words as ‘very’, ‘quite’, ‘extremely’ in
English: adverbs of degree, as they are often called. Nothing need be
said about these other than that there is frequently, if not always, a
transformationally explicable relationship between an expression con-
taining an adjective modified by an adverb of degree, on the one hand,
and a second-order noun modified by an adjective, on the other. For
example, ‘outstandingly beautiful’ and ‘outstanding beauty’ correspond
in this way; and they exemplify the principle that, as secondaries are
made into primaries under nominalization, so their accompanying
modifiers are made into secondaries (cf. Jespersen, 1929: 171; 1937:
§39.4). This is a particular instance of a principle that is elegantly for-
malized in Shaumyan’s (1965; 1974) system of transformational gram-
mar. Its reflexion can also be seen in the kind of analysis that might be
provided for a proposition like ‘‘Alice is outstandingly beautiful’’ in
terms of the higher-order predicate calculus. Roughly speaking, this
involves the individualization of degrees (or kinds) of beauty, the
existential quantification of one of these individual degrees (or kinds) of
beauty as Alice’s, and the ascription to it of the second-order property
of being outstanding. The proposition 1s analysed as if its logical struc-
® On this question and on adjectives in general cf. Bolinger (1967b), Bowers
1975)
1968).Givon (1970), Kénig (1971), Ljung (1970), Sussex (1974), Vendler
 11.3. The ontological basis 451
ture were identical with that of “There is a (certain) beauty which Alice
has and which is outstanding”’ (cf. Reichenbach, 1947; Parsons, 1972).
So-called adverbs of manner, such as ‘beautifully’ in ‘Alice dances
beautifully’, would also seem, at first sight, to satisfy the traditional
syntactic criterion; and once again there is the same transformationally
explicable relationship between the second-order noun modified by an
adjective and the corresponding verb modified by an adverb (cf.
‘beautiful dancing’: ‘dance beautifully’). Recent work by linguists and
logicians has shown, however, that adverbs of manner cannot always be
interpreted as modifying just the verb. That some adverbs of manner, if
not all, can modify larger constituents within the sentence appears
clearly from the comparison of two such sentences as (cf. Thomason &
Stalnaker, 1973):
(1) He slowly tested all the bulbs
(2) He tested each bulb slowly.
These two sentences have different truth-conditions; and the difference
is of the kind that logicians usually formalize in terms of the scope* of
operators (cf. 6.3). It is suggested, therefore, that adverbs of manner
should also be thought of as having narrower or wider scope relative to
negation and quantification. How this might be formalized is something
that need not concern us here (cf. Richards, 1976). The point is that |
there appears to be some correlation between the position in which an
adverb of manner occurs in English sentences and what it modifies; and
the same ts true of adverbs of means and instrumental adverbs, which
are usually composed of a preposition and a nominal in English (cf. ‘by
airmail’, ‘with a knife’).
Far more striking than the possibility of using instrumental adverbs
and adverbs of manner and means with variable scope within simple
sentences is the fact that there are many adverbs in English which,
especially when they occur in initial position, can hardly be said to
modify syntactically anything at all within the sentence, unless it is the
rest of the sentence: e.g. ‘frankly’, ‘fortunately’, ‘possibly’, ‘wisely’,
as in
(3) Frankly, he doesn’t stand a chance
(4) Fortunately, no-one was hurt
(5) Possibly, it will rain
(6) Wisely, he said nothing.
Words like these are commonly described as sentence-adverbs. Looked
452 Semantics and grammar II
at from a semantic point of view, they can generally be seen as having
some kind of evaluative function. They are used by the speaker in order
to express, parenthetically, his opinion or attitude towards the proposi-
tion that the sentence expresses or the situation that the proposition
describes. Further distinctions can be drawn within this class of paren-
thetical sentence-adverbs (cf. Greenbaum, 1969; Jackendoff, 1972).
Here we may simply note that many of them express what we will later
discuss under modality* (cf. 17.1).
Enough has been said to illustrate the heterogeneity of what are
traditionally classified as various kinds of adverbs. ‘There are many kinds
of adverbs that we have not mentioned and will not mention.® In con-
clusion, however, the reader’s attention may be drawn to the fact that
there are two subclasses of adverbs which, though they may occur as
adjuncts (1.e. syntactically omissible expressions), may also occur as
complements in copulative sentences (cf. 12.2): these are locative and
temporal adverbs, such as ‘here’ or ‘outside’ and ‘then’ or ‘tomorrow’.
We shall have much to say about such expressions in later sections.

11.4. Determiners, quantifiers and classtfiers

The term determiner* is currently used by linguists to label a class of
words which includes the definite and indefinite articles, the demonstra-
tive adjectives and a variety of other words that have much the same
distribution as the definite article in sentences of English and certain
other languages.
This statement is not intended to serve as a definition, but merely as a
rough-and-ready indication of the way in which the term ‘determiner’
has come to be employed recently. It is characteristic of at least the most
typical determiners, including the definite article, that their primary
semantic function is that of determining (i.e. restricting or making more
precise) the reference of the noun-phrases in which they occur: hence
the term ‘determiner’. It has already been pointed out that, in English
and in some, but by no means all, other languages, countable nouns in
the singular cannot be used in referring expressions, unless they have
combined with them a determiner (or its syntactic equivalent).
It will be noted that, in the rough-and-ready explanation of the
meaning of the term ‘determiner’ that has just been given, determiners
were described as words, rather than as forms or lexemes. The reason is
® Apart from works referred to in the text, the following may be mentioned:
Allerton & Cruttenden (1974), Bartsch (1972), Bowers (1975), Cresswell
(1974), Dik (1975), Hartvigson (1969), Lehrer (1975), Steinitz (1969).
 11.4. Determiners, quantifiers and classifiers 453
that the distinction between forms and lexemes is rather difficult to draw
with respect to closed-class items. Furthermore, the decision might go
differently for different members of the class of determiners in the same
language or for syntactically and semantically comparable determiners
in different languages. For example, the definite article in English might
be treated as a form, but the demonstrative adjectives (and pronouns)
as lexemes, on the grounds that, whereas there is syntactically relevant
variation between thzs and these or that and those (such that the former
pair might be said to be forms of ‘this’ and the latter pair to be forms of
‘that’), the definite article is invariably the. In many other European
languages, however, such as French, German or Italian, the definite
article has several different forms associated with it; and, by certain
criteria at least, it would be classified as a lexeme. Our notational con-
ventions force us to decide the question in one way or the other when-
ever we have occasion to cite a determiner in this section (or elsewhere) ;
and, in certain instances, our decision will be somewhat arbitrary. The
point is that, as has been emphasized before, the distinction between
forms and lexemes is one that depends upon the adoption of some
particular theory of grammar; and we want our account of semantics to
be as neutral as possible with respect to alternative theories of grammar.
To refer to determiners as words creates the related but somewhat
different problem, that there are many familiar languages (e.g., Danish,
Rumanian and Bulgarian) in which the definite article is not a word, but
a suffixed element, whose form is dependent upon much the same fac-
tors (involving concord of gender and number) as is the form of the
definite article in, say, German. Problems of this kind should not be
discounted; though we shall not go into them here, we mention them in
order to emphasize that the term ‘determiner’ is one whose status in
general grammatical theory is rather uncertain. The same point might
also be made with respect to the terms ‘quantifier’ and ‘classifier’.
The main reason why such words as ‘the’, ‘this’, ‘that’, ‘some’, ‘a’,
‘each’ and ‘every’ are grouped together as determiners in descriptions
of English is a grammatical, rather than a semantic, reason; they cannot
co-occur (and they are intersubstitutable), within the same noun-phrase
(cf. Quirk et al., 1972: 137). For example, ‘a my friend’ (or ‘my a
friend’) is not a well-formed noun-phrase (in contrast with, say, “un
mio amico’ in Italian). That some of these restrictions on co-occurrence
are not governed, in any obvious way at least, by semantic factors is
evident from the fact that the same restrictions do not always hold in
other languages.
 454 Semantics and grammar II
If we are looking for a general definition of the term ‘determiner’, we
can hardly do better than start from the notion of definiteness of
reference (cf. 7.2). We can say that a determiner is any element whose
function it is to enter into the structure of referring expressions and to
determine their reference as definite rather than non-definite. Given
that in such-and-such a language there are forms or lexemes that would
be classified as determiners by virtue of this criterion, we can add the
further criterion of substitutability: anything that is substitutable for a
determiner (within a noun-phrase and without changing the syntactic
properties of the noun-phrase) is also a determiner. This definition, it
should be noted, is comparable with the kind of definitions that have
been proposed for ‘noun’, ‘verb’, ‘adjective’, etc., in previous sections
of this chapter. It allows for the possibility that in certain languages
there will be no determiners; and it makes no presuppositions about the
universality of the definite article (either in deep structure or in surface
structure). In many languages that do not have a definite article (e.g.,
Russian or Latin), the demonstrative adjectives would satisfy our cri-
terion for determiners.
As we have already seen, the term quantifier* is used by logicians to
refer to particular operators, especially the operators of existential and
universal quantification, whose function it is to bind the variables that
come within their scope (cf. 6.3). It has recently been employed by
linguists with reference to such words as ‘all’, ‘some’, ‘each’, ‘every’
and ‘any’ (as well as ‘many’, ‘few’ and ‘several’), which, in certain of
their uses at least, can be said to have much the same function as the
logician’s quantifiers; and this is the sense in which the term ‘quantifier’
is being used here. It must be emphasized, however, that several of the
items that we shall refer to as quantifiers are intersubstitutable with the
articles and demonstrative adjectives in English. They might just as well
be described, therefore, as determiners; and they are so described in
many standard works (cf. Quirk et al., 1972). Conversely, both the
definite and indefinite articles in English have certain functions (in-
cluding that of generic* reference: cf. 7.2), which are similar to, though
not identical with, those of ‘all’ and ‘some’. Since English is by no
means untypical in these respects, it is not surprising that there should
be considerable confusion attaching to the terms ‘determiner’ and
‘quantifier’ in the recent linguistic literature.
, The distinction between determiners and quantifiers may be drawn,
informally but well enough for the purpose, as follows: determiners are
modifiers which combine with nouns to produce expressions whose
 11.4. Determiners, quantifiers and classifiers 455
reference is thereby determined in terms of the identity of the referent;
quantifiers are modifiers which combine with nouns to produce expres-
sions whose reference is thereby determined in terms of the size of the
set of individuals or in terms of the amount of substance that is being
referred to. In other words, a determiner tells us which member of
which subset of a set of entities is being referred to; a quantifier tells us
how many entities or how much substance is being referred to. With
respect to this distinction, imprecise though it is, ‘this’ is clearly a
determiner in the phrase ‘this man’; and ‘many’ and ‘much’ are clearly
quantifiers in the phrases ‘many men’ and ‘much bread’.
The word ‘some’ is usually taken to be the English-language equiva-
lent of the existential quantifier. As we have already seen, however,
expressions containing ‘some’, like expressions containing the indefinite
article, may be used, in certain contexts, either with specific reference or
non-specifically (cf. 7.2). This distinction is relevant to the semantic
distinction between determiners and quantifiers, in that ‘some’ is
clearly a determiner, rather than a quantifier, when it occurs in expres-
sions with specific reference, whereas its status, from this point of view,
is far less clear when it is used non-specifically. If the phrase ‘some
students’ is in implicit or explicit contrast with ‘other students’, ‘some’
is a determiner; if ‘some students’ is in contrast, whether explicit or
implicit, with ‘all (the) students’, ‘some’ is a quantifier. This difference
between the two interpretations of ‘some’ in English comes out im-
mediately when we are faced with the task of translating a phrase like
‘some students’ into languages which lexicalize or grammaticalize the
difference between an indefinite determiner and an indefinite quantifier.
In Russian, for example, ‘nekotorye studenty’ contains the determiner
‘nekotory}’; and ‘neskoljko studentov’ contains the quantifier ‘nes-
koljko’, which can itself be paraphrased as ‘nekotoroe Cislo’ (“‘some —
specific or non-specific — number”’). The difference between the two
interpretations, or translations, of “some students’ is all the more
striking in that the determiner ‘nekotoryj’ is an adjective, whereas the
quantifier ‘neskoljko’ is a pronoun (more precisely a pro-nominal: cf.
15.3) which governs a partitive genitive. Furthermore, the fact that
statements containing ‘nekotoryj’ answer (or are presupposed by: cf.
16.3) questions containing ‘kotoryj’ (“Which . . .?’’), whereas state-
ments containing ‘neskoljko’ answer questions containing ‘skoljko’
(“How much/many .. .?’’), is evident from the morphological corres-
pondence between the indefinite adjectives and pronouns, on the one
hand, and the interrogative adjectives (and pronouns), on the other.
 456 Semantics and grammar II
There are many languages in which the same kind of morphological
correspondence supports either the specific vs. non-specific distinction
or the determiner vs. quantifier distinction.
At first sight, the situation with respect to ‘all’ is more straight-
| forward. It can be combined, in the same noun-phrase, with the non-
quantifying determiners, such as ‘the’, ‘this’ and ‘that’ (cf. ‘all those
students’); and this fact would suggest that it is more of a quantifier
than a determiner. But ‘all’ can nonetheless be used appropriately in
answering “‘ Which’’-questions; and in such contexts it is in explicit or
implicit contrast with the non-quantifying determiners. For example, a
child, to whom one puts the question Which sweets do you want?, may
reply, both truthfully and appropriately, in terms of the grammatical
and semantic structure of English, All of them. In contexts like this,
‘all’ is as much a determiner as a quantifier; and there is even less reason
to say that [ want all of them is an ambiguous utterance, having one
meaning rather than another according to whether it answers the ques-
tions How many do you want? or Which ones do you want?, than there is
to say that J want some of them is an ambiguous utterance. It follows that,
if ‘some’ is in explicit or implicit contrast with ‘all’ in a particular con-
text, it is not necessarily shown thereby to be a quantifier, rather than a
determiner, as we supposed earlier.
We will not pursue this question any further. What has been said is
sufficient to demonstrate the difficulty of drawing a sharp distinction be-
tween determiners and quantifiers on semantic grounds. In so far as there
is a distinction to be drawn, it rests, as we have seen, upon the difference
between ‘‘ Which (one/ones)?’’ and “How much/many?’’. This is the
distinction that 1s of central concern; and we will return to it presently.
Despite the enormous concentration of attention that there has been
on the so-called quantifiers by both linguists and logicians in recent
years, there is much that remains unclear (cf. Hintikka, 1974). No agree-
ment has yet been reached, for example, on the semantic interpretation
of expressions containing ‘any’. One of the most obvious facts about
‘any’ is that it tends to occur in syntactically definable contexts (roughly
speaking, in negative, interrogative and conditional clauses) in which
‘some’ tends not to occur. Conversely, ‘some’ tends to occur in syn-
tactically definable contexts in which ‘any’ does not occur. This being

English to treat }
so, it is more or less standard practice in conventional grammars of

(1) John didn’t see anyone

 11.4. Determiners, quantifiers and classtfiers 457
and

(2) Did John see anyone?

as the corresponding negative and interrogative versions, as it were, of
(3) John saw someone.
That (1) and (2) stand in the same semantic relationship to (3) as ‘John
didn’t read the book’ and ‘ Did John read the book?’ do to ‘John read
the book’ is intuitively obvious. At the same time, it is no less obvious
that both
(4) John didn’t see someone
and

(5) Did John see someone?

are to be regarded as grammatically well-formed sentences of English.
What then is the difference between (1) and (4)? And is it the same
difference as that which holds between (2) and (5)?
One relevant point is that ‘someone’, unlike ‘anyone’, may be used
with specific reference. But this fact goes only part of the way towards
accounting for the difference between (1) and (4), on the one hand, and
between (2) and (5), on the other. Specificity of reference is but one con-
dition for the utterance of (4) or (5), rather than (1) or (2). Another
relevant consideration is whether the speaker is positively, rather than
negatively or neutrally, disposed towards the propositional content of
the utterance (cf. R. Lakoff, 1968). If the speaker either expects or
hopes that John saw someone (regardless of whether he intends to refer
to some specific person who might have been seen by John), he will tend
to use ‘someone’, rather than ‘anyone’, in questions.
_ If linguists have been mainly concerned with the semantic and
syntactic relationship between ‘some’ and ‘any’, logicians and philo-
sophers seem to have been more inclined to relate ‘any’ to ‘each’,
‘every’ and ‘all’: i.e. they have thought of ‘any’ as being equivalent to
the universal, rather than the existential, quantifier. Consideration of
such pairs of sentences as
(6) Anyone can win
and

(7) Everyone can win

 458 Semantics and grammar II

and ,
or
(8) When I was in prison, I ate anything that was put in front of me

(9) When I was in prison, I ate everything that was put in front of me
shows that it is at least initially plausible to relate ‘any’ to ‘every’. But
there are problems here too.
There are many contexts in which there is a very clear semantic
difference between ‘any’ and ‘every’. For example,
(10) I don’t know anyone here
and
(11) I don’t know everyone here
differ far more obviously in their truth-conditions than (6) and (7) do.
But this can be accounted for, it is suggested, in terms of the relative
scope of the operators of negation and quantification, (10) being analysed
aS

(10a) (x)( ~(I know x)): “For all values of x, it is not the case that I
know x”’

and (11) as
(11a) ~((x) (I know x)): “It is not the case that, for all values of x,
I know x”’.

More generally, it has been proposed that ‘any’ is (or is the English-
language equivalent of) a quantifier of universal quantification which
has wider scope than any other operator of negation, quantification or
| modality. This principle accounts, not only for (10) and (11), analysed
along the lines of (10a) and (11a), but also for many other more complex
combinations of logical operators and their alleged English-language
equivalents.
However, there would seem to be certain exceptions to the principle
that ‘any’ always takes the widest possible scope in any clause in which
it occurs. For example, the following two sentences do not differ in
meaning as obviously as (10) and (11) do:
(12) Some remedies can cure anything
(13) Some remedies can cure everything.
Under one of its possible interpretations, (12) can be analysed, from a
 11.4. Determiners, quantifiers and classtfiers 459
logical point of view, as meaning roughly “There is/are some x such
that, for all values of y, x can cure y’’; and this interpretation (in which
‘some’ is to be construed as having non-specific reference) cannot be
accounted for in terms of the principle that ‘any’ always has wider scope
than ‘some’.
There are also certain rather subtler differences between ‘any’ and
‘every’ which cast doubt upon the identification of ‘any’ with the
universal quantifier. First of all, ‘any’ and ‘every’ differ in certain con-
texts with respect to existential presupposition. For example, whereas
(14) I don’t know everyone with yellow lips
would not normally be uttered (except as a context-dependent denial:
cf. 16.4) unless the speaker was prepared to commit himself (if challen-
ged, on this score) to a belief in the existence of people with yellow lips,
(15) I don’t know anyone with yellow lips
is quite neutral with respect to the existence or non-existence of people
with yellow lips.
More serious than the problem of existential presupposition, how-
ever, is the fact that the truth-conditions of sentences containing ‘any’,
unlike the truth-conditions of otherwise identical sentences containing
‘every’ or ‘some’, are rather obscure. If it is asserted that every member
of a particular set of mountaineers, {a, b, c, d}, is capable of climbing
Everest, the person making this assertion 1s committed to the truth of the
following four propositions:
(16) “‘a can climb Everest’’
(17) “db can climb Everest”’
(18) “ec can climb Everest”’
(19) “d can climb Everest”’.
Anyone asserting that some member of the team is capable of climbing
Everest (where ‘some’ is to be construed non-specifically) is committed
to the truth of the disjunction of the same four propositions. At first
sight it might appear that the truth-conditions of
(20) Any member of the team can climb Everest
are the same as the truth-conditions of
(21) Every member of the team can climb Everest.
But, arguably, this is not so. Unlike (21), (20) neither says nor implies
460 Semantics and grammar II
that there is any more than one member of the team who is capable of
climbing Everest. In this respect, it is like
(22) Some member of the team can climb Everest.
But (20) differs, nonetheless, from both (21) and (22). The proposition
expressed by (21) entails, but is not entailed by, the proposition expres-
sed by (20); and the proposition expressed by (20) entails, but is not
entailed by, the proposition expressed by (22). So much is clear enough.
The problem is to formulate the truth-conditions of (20) in such a way
that they bring out these semantic differences; and no-one appears to
have done this yet.
Vendler (1967: 70-96) has emphasized the fact that when we utter
sentences like (20) what we do in effect is to throw down a challenge:
‘‘Pick some member of the team and I assert of the member that you
pick that he is capable of climbing Everest’”’. This analysis of the
meaning of ‘any’ cannot be accommodated, in any straightforward way,
within the framework of truth-conditional semantics. But it does have
the advantage that it accounts very naturally for the fact that ‘any’ tends
to occur in a variety of what may be referred to, loosely, as modal
contexts; and the particular grammatical or quasi-grammatical relation-
ship that holds between ‘some’ and ‘any’, which was mentioned earlier,
reflects this tendency.
Limitations of space prevent us from discussing the equivalents, or
near-equivalents, of ‘some’, ‘any’, ‘every’, ‘all’, etc., in other lan-
guages. All that can be said here is that, although they may not present
exactly the same problems of analysis as the English determiners and
quantifiers, there is no reason to believe that they are any more satis-
factorily analysable in terms of the existential and universal quantifiers
than the English determiners and quantifiers are.
Semanticists have devoted far less attention to classifiers* than they
have to determiners and quantifiers. The reason, no doubt, is that,
although very many of the world’s languages make use of classifiers, the
more familiar Indo-European languages do not. What is meant by the
term ‘classifier’ (in the sense in which it is being used here) is best
explained by means of an example.
If one wanted to translate into Tzeltal (a Mayan language spoken in
Mexico) the English phrases ‘three trees’ or ‘four men’, one would
have to use, in each instance, a three-word phrase, rather than a two-
word phrase (cf. Berlin, 1968: ‘o8-tehk te?’ (‘three trees’), ‘Can-tul
winik’ (“four men’’)). The first word in these phrases (‘o8’, ‘Can’) is a
 11.4. Determiners, quantifiers and classifiers 461
numeral; the third (‘te?’, ‘winik’) is a common noun; and the second
(‘tehk’, “tul’) is a classifier. Classifiers are syntactically obligatory
elements in all such expressions; and the classifier that is employed
depends upon the nature of the entity or set of entities that is being
referred to. The classifier for plants is ‘tehk’; the one for human beings
is ‘tul’; the one for animal is ‘koht’; and so on.
Tzeltal is typical of what we may refer to, loosely but conveniently, as
classifier-languages, in that classifiers are obligatory in phrases con-
taining numerals. (Hence the alternative term, ‘numeral classifier’,
which 1s frequently used in the literature.) There are many languages,
including Mandarin Chinese and Vietnamese (cf. Chao, 1968; Emeneau,
1951), in which classifiers are also obligatory with demonstratives: i.e.
in phrases which might be translated into English as ‘that tree’, ‘this
man’, etc. In such languages, it is commonly, if not universally, the case
that there is a special pluralizing classifier, which occurs with demonstra-
tives (but not with numerals) and replaces the semantically appropriate
classifier that would be used in non-plural constructions (cf. Greenberg,
1972). For example, (i) ‘one book’, (ii) ‘three books’, (iii) ‘this book’,
and (iv) ‘these books’ are translated into Mandarin Chinese as (i) ‘i ben
shu’, (ii) ‘san ben shu’, (itt) “che ben shu’ and (iv) ‘che hsie shu’. The
word ‘ben’ is the classifier used, in general, for flat objects, whereas
“hsie’ can be used for any kind of plurality or collectivity. One way of
bringing out the difference is by translating the Chinese expressions

tivity book’. ,
into a kind of Quasi-English as follows: (i) ‘one flat-entity book’, (ii)
‘three flat-entity book’, (iii) ‘this flat-entity book’ and (iv) ‘this collec-

One further generalization may be made about classifiers. This is that

in most, if not all, classifier-languages there is, in addition to the
semantically specialized classifiers used in referring to particular kinds
of entities (e.g., human beings, animals, plants, flat objects, round
objects, etc.), a semantically neutral classifier, which may be employed
(instead of the appropriate semantically specialized classifier) with
reference to all sorts of entities. The word ‘ge’, for example, is used in
this way in Mandarin Chinese. It is as if, in Quasi-English, we had such
phrases as ‘three entity book’ (“three books’’) and ‘three entity man’
(‘three men’’) as alternatives to ‘three flat-entity book’ and ‘three
human-entity man’. In many classifier-languages the semantically
neutral classifier is restricted to non-personal, or even inanimate,
entities, so that ‘thing’, rather than ‘entity’, would be its Quasi-English
equivalent (cf. Greenberg, 1972).
 462 Semantics and grammar IT
We cannot go further into the syntax of classifier-constructions in
various languages. What has been said about Tzeltal, on the one hand,
and Mandarin Chinese, on the other, will suffice for the present pur-
pose. What we must do now is make explicit a number of semantically
relevant points, some of which are implicit in our technique of transla-
tion into Quasi-English.
Classifier-constructions are very similar, both syntactically and seman-
tically, to the construction exhibited by such phrases as ‘fifty head of
cattle’, ‘three sheets of paper’, or ‘that lump of iron’, in English. Words
like ‘head’, ‘sheet’ and ‘lump’, in constructions of this kind, serve
exactly the same function — that of individuation and enumeration — as
do the classifiers of Tzeltal, Chinese, Burmese (cf. Friedrich, 1969,
1970), etc. The difference between English and the so-called classifier-
languages is that in English, as in the Indo-European languages gener-
ally and in some other, but by no means all, language-families, there is
a grammatical distinction between countable and uncountable nouns.
Looked at from a semantic point of view, the grammaticalization of
countability rests upon the encapsulation of the component ENTITY
within the meaning of whatever lexemes are treated grammatically as a
countable noun: ‘boy’, ‘dog’, ‘tree’, ‘table’, etc. It is important to
realize that the grammatical category of countability, like the gram-
matical category of number (singular vs. plural, etc.), is but one of
several interconnected devices used in language in the construction of
referring expressions. What all these devices have in common is that
they are based upon, or presuppose, the possibility of individuation and
enumeration.
As was pointed out in an earlier chapter (cf. 7.6), most of the nouns
in classifier-languages are like the noun ‘salmon’ in English, which in
such utterances as I like salmon can be construed as referring (either
distributively or collectively) to a class of individuals (cf. I like herrings)
or to a stuff or substance (cf. J like meat). It is worth noting in this con-
nexion that there are many languages (e.g., Classical Arabic) in which
countable nouns may have a collective form, which is distinct from their
plural form. There is an obvious semantic parallel between nouns
denoting amorphous stuff or substance (e.g., ‘gold’, ‘ water’) and nouns
denoting undifferentiated collections or aggregates of individuals (e.g.,
‘cattle’). Furthermore, the plural form of countable nouns in English
is frequently used in the same constructions as uncountable nouns and
collective nouns are (cf. I like cows/cattle/beef); and for this reason it
is semantically, though not formally, unmarked (cf. 9.7).
 11.4. Determiners, quantifiers and classifiers 463
Classifier-constructions of the kind that we are concerned with here
fall into two broadly distinguishable types. These may be referred to as
sortal* and mensural*. A sortal classifier is one which individuates
whatever it refers to in terms of the kind of entity that it is. The Tzeltal
classifiers ‘tehk’, ‘tul’ and ‘koht’ are all sortal in this sense. A mensural
classifier is one which individuates in terms of quantity. The function of
mensural classifiers is comparable with that of such words as ‘pound’
or ‘pint’ in English (cf. ‘two pounds of butter’, ‘three pints of milk’).
In terms of this distinction between sortal and mensural classifiers,
we can account for the ambiguity of such phrases in English as ‘three
whiskies’, in which what is normally regarded as an uncountable noun,
‘whisky’, is treated as countable. Subject to conditions of contextual
appropriateness all such phrases are to be construed as if they contained
either a sortal classifier meaning “‘kind/sort/type’’ or a mensural classi-
fier meaning “‘quantum’’. For example, We only stock three whiskies
would presumably be construed as meaning ‘‘ We stock only three kinds
of whisky’’, whereas I only drank three whiskies in the course of the
evening would probably, though not necessarily, be understood as
meaning “I drank only three quanta of whisky . . .”’ (where what counts
as a quantum for whisky is fixed by general convention).
It is not only phrases containing what are normally thought of as
uncountable nouns that are ambiguous in this way. So too, strictly
speaking, are such phrases as ‘three tables’, which, in addition to the
more obvious meaning “three table-entities’’, can also have the mean-
ing ‘“‘three kinds of table(s)”’ (cf. We only stock three tables). Of these
two interpretations, the latter (“‘three kinds of table(s)’’) is matched by
one of the two interpretations of ‘three whiskies’; and it involves
generic reference. The other interpretation, “three table-entities”’,
might appear, at first sight, to be quite different from the second of the
two interpretations of ‘three whiskies’, “three quanta of whisky”. On
further reflexion, however, it will be seen that there is a parallelism. An
entity is a quantifiable unit of the class, or classes, to which it belongs;
an amount or quantum of some substance like water, gold or whisky
may also be regarded as an individuated, re-identifiable and enumerable
unit. Languages which grammaticalize the distinction between entity-
denoting nouns and mass-denoting nouns tend to draw a sharp syntactic
distinction between phrases like ‘three men’, on the one hand, and
‘three glasses of whisky’, on the other. Classifier-languages do not: they
treat enumerable entities and enumerable quanta in much the same way.
Indeed, the most appropriate meaning that one might assign to the
 464 Semantics and grammar II
semantically neutral sortal classifier is perhaps “‘unit”’, rather than
“entity”.
Mensural classifiers of various kinds are probably to be found in all
languages: they approximate to, and indeed in many instances merge
with, quantifiers (cf. Jackendoff, 1968). Although their analysis presents
certain semantically interesting problems, we will say no more about
them here (cf. Parsons, 1970). We will simply note that there are many
classifiers that do double duty, as it were, operating simultaneously as
both mensural and sortal classifiers. For example, Three lumps, please is
a linguistically appropriate answer to the question How much sugar do
you want in your tea? But ‘lump’ is primarily a sortal classifier in terms
of our distinction: a lump is an aggregate of a particular, though some-
what indeterminate, kind. By convention, however, sugar is produced
and sold in lumps of standard size (and shape), so that ‘lump’ may be
used to mean “‘quantum”’.
Sortal classifiers supply, or presuppose, a principle for individuating
entities and grouping them into kinds. In this respect, they are com-
parable with the most general common nouns, such as ‘person’,
‘animal’, ‘bird’, ‘fish’ or ‘tree’ in English. Such nouns are frequently
‘combined with the definite article or one of the demonstrative adjectives
to form definite descriptions; and classifiers tend to be used in the same
way in classifier-languages. Indeed, the vast majority of sortal classifiers
(in at least the most familiar classifier-languages) are nouns, albeit nouns
of a particular subtype; and it is this fact, more than any other, which
motivates :the distinction that is usually drawn between classifier-
languages and noun-class languages, such as the Bantu languages and
some Amerindian and Australian languages..A further notable charac-
teristic of sortal classifiers is that (in at least the most familiar classifier-
languages) they can be used with pronominal, or quasi-pronominal,
function in deictic and anaphoric reference (cf. 15.3). One way of ex-
pressing this point is to say: “the head noun may be deleted either when
it has been previously mentioned or can be supplied from the non-
linguistic context’’ (Greenberg, 1972). However, it 1s arguable that, in
many instances, the classifier is the head, rather than the modifier, in the
constructions in which it occurs. The fact that this is so makes sortal
classifiers rather like determiners. For determiners, despite their con-
ventional treatment as modifiers of the noun with which they occur, may
often be regarded, from a syntactic point of view, as heads rather than
modifiers (cf. 15.2).
In many other respects also, sortal classifiers are like deterrhiners; and
 11.4. Determiners, quantifiers and classifiers 465
further discussion of the way in which they are used in various lan-
guages would reinforce the conclusion that, just as there is a semantic
and syntactic connexion in English between mensural classifiers and
quantifiers, so there is, in many languages, a syntactic and semantic
connexion between sortal classifiers and determiners. It is, to say the
least, a defensible point of view that the distinction between deter-
miners, quantifiers and classifiers is one that is drawn differently, not
only in the surface structure, but also in the deep structure of different
languages.?°
It is interesting to note, however, that, in so far as the semantic
principles which determine the association of particular sortal classifiers
with particular subclasses of nouns are identifiable, these principles
appear to be much the same the world over. One major principle depends
upon the recognition of such large groupings of potential referents as
persons, animals, birds, fishes, trees, plants, etc. The entities that fall
into such classes all belong to what would be traditionally regarded as
natural kinds (cf. Putnam, 1970). Furthermore, viewed from the stand-
point of naive realism their ontological status is determinate, indepen-
dently of the language that is used to refer to them (cf. 11.3); and the
way in which they are grouped into kinds is explicable in terms of the
partly universal and partly culture-dependent basis of language (cf. 8.3).
Much of the phenomenal world, however, is not composed of entities
that belong to natural kinds. In so far as individuals are identified within
it they must be grouped into kinds in terms of their location or their
physical or functional properties. As we shall see later, location is in-
volved in the deictic identification of individuals (cf. 15.2): it also
operates in the selection of sortal classifiers in certain languages. But by
far the most common principle of sortal classification, for entities that do
not belong to natural kinds, is shape (cf. Friedrich, 1969; Greenberg,
1972; Allan, 1977). In many different classifier-languages throughout
the world one classifier is used for long thin entities, another for flat
entities and a third for round or bulky entities. Size constitutes another
common principle of sortal classification, so too does texture (i.e.
whether what is being referred to is hard or soft, solid or liquid).
Functional principles of sortal classification are less easy to identify and
to compare across languages, since they may be very largely culture-
10 That there is such a deep-structure difference between languages does not,
of course, follow from anything that has been said here. The possibility is
mentioned, however, in view of the common tendency nowadays to assume
the contrary.
466 Semantics and grammar II
dependent. But there is at least one culture-independent functional
property that serves for sortal classification in a strikingly large number
of languages: edibility.
To some extent, then, the principles of sortal classification, in so far
as they are discernible, appear to be universal, being based upon the
ontological salience of natural kinds and the perceptual or functional
salience of certain criterial attributes. How the distinctions that these
principles of classification establish are grammaticalized or lexicalized
varies considerably, however, across languages. In this section we have
dealt with them primarily in connexion with the way in which they
operate in classifier-languages: we might equally well have discussed the
gender-distinctions of noun-class languages (cf. Dixon, 1968) or the
so-called classificatory verbs of the Athapaskan languages (cf. Hoyer,
1945; Haas, 1967; Krause, 1969; Carter, 1976), where the same general
principles are operative. ‘The reason why particular attention has been
paid here to sortal, and to a lesser extent mensural, classifiers is that they
have been very largely neglected, by contrast with determiners and
quantifiers, in theoretical semantics.
 12
Semantics and grammar III

12.1. Kernel-sentences and sentence-nuclet

We are assuming throughout this work the validity of transformational
grammar and its universal applicability for the syntactic analysis of
natural languages. We are not committed, however, to any specific
formalization of transformational grammar (cf. 10.3); and in this sec-
tion as elsewhere in the book, we shall be somewhat eclectic in the use
that we make of the terminology and concepts associated with particular
systems of transformational analysis.
One notion that we will appeal to is that of the kernel-sentence*. T'wo
rather different conceptions of kernel-sentences have been formalized in
transformational grammar: one by Harris and the other by Chomsky.
For Harris, a kernel-sentence is a sentence that is not derived from any
other sentence (or pair of sentences) by means of a transformational
rule; for Chomsky, as he originally defined the notion, a kernel-sentence
is one that is generated in the grammar without the operation of any
optional, as distinct from obligatory, transformations. In later versions
of Chomsky’s system, in which the role of optional transformations has
been greatly reduced, the notion of the kernel-sentence has lost much of
its original significance; and kernel-sentences, as such, play no part in
either the generation or interpretation of sentences (cf. Chomsky, 1965:
18). Nevertheless, kernel-sentences are still definable in Chomsky’s
system (though not as simply or as elegantly as before).
The difference between these two conceptions of kernel-sentences
need not concern us. It largely depends upon the way in which trans-
formations are defined in the two systems of formalization. In Harris’s
system they operate (in general) upon sets of sentences, whereas Chom-
sky defines a transformation as an operation which converts one under-
lying abstract structure (of a particular kind) into another such structure.
We take no stand on this issue. Under either of the two definitions, the
kernel-sentences of a language are intended to constitute a subset of
what would be traditionally described as simple sentences. Granted that
 468 © Semantics and grammar III
the traditional concept of the simple sentence can be formalized in
terms of the derivation of that sentence from a single underlying sen-
tence or sentential structure, a kernel-sentence may be defined, for our
purposes, as a simple sentence which is unmarked in mood, voice and
polarity and does not contain any optional, or omissible, expressions. As
far as English is concerned, we assume that the unmarked mood is the
indicative (rather than the imperative: cf. 16.2), that the unmarked
voice is the active (rather than the passive) and that the unmarked term
in the category of polarity is the affirmative (rather than the negative).
In so far as these distinctions of mood, voice and polarity are gram-
maticalized in any language, it is perhaps reasonable to assume that the
kernel-sentences of the language will also be affirmative, active and

assumption. :
indicative. However, nothing of what is said below depends upon this

The importance of the notion of the kernel-sentence is that it holds out

to the semanticist the prospect of his being able to account for the mean-
ing of all the sentences of a language on the basis of the meaning of a
relatively small number of them. This prospect is all the more attractive
if it can be assumed, not only that the kernel sentences of a language
will be relatively short and simple in structure, but also that the vocabu-
lary of which they are composed will be restricted, for the most part, to
morphologically simple lexemes with a concrete meaning. We will make
this assumption throughout the present section, without attempting to
justify it. In a later chapter, however, we shall see that there are many
morphologically complex lexemes whose meaning and distribution
cannot be fully accounted for in terms of their transformational deriva-
tion from kernel-sentences (or the structures underlying kernel-
sentences) containing morphologically simpler lexemes (cf. 13.2).
Kernel-sentences, as we have defined them, correspond fairly closely
to what many philosophers in the empiricist tradition have thought of as
elementary propositions whose function it is to describe states of some
actual or possible world (cf. 6.5). But propositions are not sentences; and
sentences are not propositions. Propositions, under one interpretation of
this term, are abstract entities which may be asserted or denied by
making statements; and statements are made characteristically, though
not necessarily, by uttering declarative sentences in the indicative mood
} (cf. 16.1). Furthermore, in all the standard logical calculi, propositions
are taken to be timeless or tenseless, whereas declarative sentences in
many languages (including English) must include some indication of
temporal reference. We want to be able to say, for example, that the
 12.2. Predicative structures 469
sentences ‘It is raining’, ‘It was raining’ and ‘It will be raining’ all
express the same proposition (namely “It be raining’’); and that they
do so by virtue of their containing what we will refer to as the same
nucleus*. There are various reasons, both syntactic and semantic, for
treating tense (in those languages that have tense: cf. 15.4) as an extra-
nuclear category, even though tense-distinctions are most commonly
represented by variations in the form of verbs or by particles closely
associated with verb-forms in the surface structure of sentences. Aspect*
is another grammatical category that we will leave out of account,
though it is arguably less peripheral than tense or mood (cf. 15.6). We
will therefore say that, not only ‘It is raining’ and ‘It was raining’, but
also ‘It has been raining’, ‘It (never) rains’, ‘It used to rain (every
day)’, etc., contain the same sentence-nucleus.
In so far as the nuclei of kernel-sentences express propositions, we can
refer to the syntactic and semantic relations which hold between the
constituents of these sentence-nuclei as intra-propositional relations;
and it is intra-propositional relations of various kinds with which we are
concerned ih the following sections.

12.2. Predicative structures

Ideally, our discussion of the syntactic structure of the nuclei of kernel-
sentences should be conducted within the framework of some generally
accepted and universally applicable system of grammatical analysis.
Unfortunately, no such system exists. Since we cannot, in the space
available, go into the several alternatives, we shall operate with a set of
terms and: concepts which come partly from traditional grammar and
partly from more recent grammatical theory.
We will begin by listing a set of sentence-schemata* (three of which
were introduced and discussed earlier: cf. 11.2). Some or all of the
following sentence-schemata (nuclear structures in the sense of 11.2)
would appear to be identifiable, on purely grammatical grounds, in very
many unrelated languages:
(1) NP+V (intransitive)
(2) NP-+V-+NP (transitive)
(3) NP(+Cop)+NP = (equative)
(4) NP(+Cop)+N/A (ascriptive)
(5) NP(+Cop)+ Loc _ (locative)
(6) NP(+Cop)+Poss (possessive)
The symbols that occur in these sentence-schemata are to be interpreted
470 Semantics and grammar III
as follows: NP = noun-phrase (or nominal: cf. 11.3), V = verb,
Cop = copula, N = noun, A = adjective, Loc = locative (adverbial)
expression, Poss = possessive (adverbial) expression. It will be observed
that the optional element in the ascriptive structure has been classified
here as a copula, rather than (as in 11.2) a verb. We will take up this and
other points implicit in the above classification presently. For the
moment, we will simply assume that every distinct symbol stands for a
syntactically justifiable class of expressions in whatever language the
schemata are identifiable. As far as English is concerned, the schemata
are exemplified in the following sentences:
(1a) Kathleen works (hard) (nowadays) (at school)
(2a) That boy plays the piano (in the evening)
(3a) The chairman is Paul Jones
(4a) He’s a (clever) boy/He was (very) intelligent
(5a) They were in the attic (half-an-hour ago)
(6a) This bicycle is John’s.
The expressions enclosed 1n parentheses are all syntactically omissible
modifiers (i.e. adjuncts*) of various kinds; and some of the expressions
which function as constituents of particular nuclei may be derivable by
transformations from other structures. We will treat (1a)-(6a), without
the parenthesized adjuncts, as kernel-sentences of English; and we will
say nothing, in this section, of the possibility that the definite article and
demonstrative adjectives should be derived by the adjectivalization of a
locative predicate and the possessive adjectives, in some cases at least, by
the adjectivalization of a possessive adverbial expression (cf. 10.3, 15.2).
All the propositions expressed by (1a)-(6a), with the exception of
(3a) under one of its interpretations, may be described (in terms of the
logical distinction of reference and predication) as predicative. They
identify a referent and say of the referent that it does something or
other, that it has a certain property or is a member of a certain class,
that it is in a certain place, and so on. The referent is identified (in any
appropriate utterance of these sentences) by the NP-expression which
occurs as the left-most constituent of the nucleus; and this we will call
the subject*. What is said about the referent is expressed by the
predicate* (or predicative expression) that is combined with the subject
in the nucleus. All this is straightforward enough; and, apart from the
fact that we do not consider tense, mood and aspect to be components
of the predicate, it is in accord with the traditional bi-partite analysis of
both sentences and propositions. Nothing of consequence turns upon
 12.2. Predicative structures 471
our acceptance so far of the traditional view that the nuclei of transitive
sentences, like (2a), are analysable into two immediate constituents, of
which one is the subject and the other the predicate. Later in this
chapter, however, we shall consider an alternative analysis.
It will be noted that the copula-symbol in (3)-(6) has been enclosed
in parentheses. This is intended to indicate that there are certain lan-
guages in which some or all of these structures lack any element that
might be classified as a copulative verb. As we saw earlier, the reason
why the English lexeme ‘be’ is classified as a verb is simply that with
respect to concord and the realization of tense it is pivotal in the way
that transitive and intransitive verbs are pivotal (cf. 11.2). It is a
meaningless lexeme whose syntactic function it is to convert whatever it
combines with into a verbal (i.e. predicative) expression. Whether the
copula should be generated as an element in the deep structure of sen-
tences or introduced by means of transformational rules in certain
syntactically definable positions is a separate question, and one that we
need not go into here.
There are many languages in which some or all of the structures listed
as (3)-(6) have no copula; there are others in which there is an optional
or obligatory copula, but where the reasons for classifying the copula as
a-verb are not as compelling as they are in the Indo-European languages.1
It is also important to note that, although English uses the same copula
in all four structures, there are languages in which the ascriptive copula
is different from the equative and the locative copula, the locative dif-
ferent from the possessive copula, and so on. But even when the same
copula (or no copula) is employed in these four structures, there may
still be syntactic grounds for drawing the several distinctions that we
have recognized above; and, with the possible exception of the distinc- ,
tion between (5) and (6), there are extremely important semantic
distinctions which correlate with them. We will illustrate this point
briefly in relation to English.
The syntactic distinction between equative and ascriptive sentences
in English rests, principally, upon two facts: (i) the expression-class
which occurs in the third position of equative sentences is not co-
extensive with the expression-class that occurs in the third position of

* On the copula in various languages, cf. Asher (1968), Christie (1970), Ellis &
Boadie (1969), Kahn (1973), Kiefer (1968), Lehiste (1969), Li (1972), Lyons
(1967) and several articles in Verhaar (1967-73). Of particular interest is
Kahn (1973), which is much broader than the title would suggest and includes
a valuable discussion of the philosophical issues.
472 Semantics and grammar ITI
ascriptive sentences; (ii) the subject and the complement of equative,
but not ascriptive, sentences are freely permutable. The difference
between equative and ascriptive sentences is clear enough if we com-
pare the sentences ‘John is the chairman’ with ‘John is intelligent’: ‘the
chairman’ is a nominal (NP) and ‘intelligent’ an adjective; and the
nominal, but not the adjective, is permutable with the subject-NP. (The
utterance Intelligent 1s fohn is of course acceptable; but it is stylistically
restricted, and it 1s associated with a very particular intonation-pattern.
Whether one should generate a system-sentence ‘Intelligent is John’ to
account for it is, to say the least, debatable.) Adjectives cannot occur as
equative complements (in sentences that have an NP-subject). It is the |
difference between the sentences ‘John is the chairman’ and ‘John 1s a
writer’ that is both less obvious and (it must be admitted) more contro-
versial; and this difference has been obscured in many treatments of
English by classifying both kinds of complements as nominal (in what
is arguably an equivocal use of the term ‘nominal’). ‘The occurrence of
the indefinite article form a@ is a purely automatic consequence of the
fact that the subject-NP is singular and ‘writer’ is a countable noun.
The complement in the ascriptive sentence is not, therefore, the NP-
expression ‘a writer’, but the N-expression ‘writer’.
Generally speaking, an equative complement (in a sentence with an
NP-subject) can be a proper name, a pronoun or a definite noun-phrase,
but not an adjective; and an ascriptive complement can be a noun or an
adjective, but not a pronoun or a proper name. True, there is the prob-
lem that definite noun-phrases may also occur as ascriptive comple-
ments. But this problem is solved, we assume, by invoking the notion of
grammatical ambiguity (cf. 10.4). We take ‘John is the writer’, and more
obviously ‘John is the author of this book’, to be grammatically am-
biguous in terms of the distinction between equative and ascriptive
sentences. This distinction also accounts, it should be noted, for two
different kinds of apposition*: cf. “The chairman, John Smith, proposed
a wote,of thanks’ ws. ‘The chairman,,a prominent local author, proposed
a vote of thanks’.
othe semantic distinction between equative and ascriptive structures
is/that the former are used, characteristically, to identify the referent of
one expression with the referent of another and the latter to ascribe to
the referent of the subject-expression a certain property. The equative
copula is, therefore, the linguistic correlate of the identity-operator in
mathematics or logic; and, whether or not it is “‘a disgrace to the human
race’, as Russell once remarked, that the same copula is used in many
 12.2. Predicative structures 473
languages in both equative and ascriptive sentences (cf. Kahn, 1973: 4)
it is certainly important, in the semantic analysis of statements, to dis-
tinguish those which answer the question of the form What ts fohn?
from those which answer the question Who ts fohn? 'The sentence ‘John
is the chairman’ can be used to answer either of these two questions.
Similarly, Russell’s (1905) famous sentence ‘Scott is the author of
Waverley’ can be used either to answer the question Who 1s Scott? (and
this is the sense in which Russell took it: cf. 7.2) or, to make the point
very crudely, to comply with the instruction Tell me something about
Scott. The fact that an equative sentence with an NP-subject must have
an NP-complement is a natural, if not inevitable, reflex of the fact that
such sentences have as their most characteristic function that of identi-
fying an entity referred to by means of one expression with an entity
referred to by another expression. So too is the fact that the subject-NP
and the complement-NP are permutable.
Let us now turn our attention to sentence-nuclei containing a locative
complement: cf. (5) and (5a) above. Locative expressions are not
generally recognized as constituting a major class of sentence-con-
stituents on a par with nominals and verbs. In traditional grammar, they
are treated as just one of several subclasses of adverbs or adverbial
phrases. Unlike all other adverbials, however, locative expressions may
be used predicatively (as complements of the copula in English) with
first-order nominals as their subjects. That locative expressions may be
used in this way is hardly surprising. The location of the persons,
animals and things with which we interact in everyday life is no less
interesting and important to us than their actions and physical or other
properties. Where is X? 1s as natural a question as What is X doing? or
What is X like?; and the grammatical structure of English and other
languages reflects this, in that we can say of an entity where it is (or where
it has been, was or will be) without saying what it is like, what it is
doing, what is happening to it or anything else about it. Locative
adverbials may be used as readily as verbs, adjectives and nominals in
the nuclei of kernel-sentences; and they may also be used (like various
other kinds of adverbials) as extra-nuclear adjuncts.
The distinction between locative and possessive complements is, at
first sight at least, straightforward enough in English. It is worth noting,
however, that the term ‘possessive’, as it is traditionally employed by
linguists, is somewhat misleading: it suggests that the basic function of
the so-called possessive constructions that are found in many languages
is the expression of possession or ownership. Generally speaking, however,
474 Semantics and grammar III
a phrase like ‘X’s Y’ means no more than “‘the Y that is associated
with X’’; and the kind of association holding between Y and X is
frequently one of spatial proximity or attachment. It can be argued that
so-called possessive expressions are to be regarded as a subclass of
locatives (as they very obviously are, in terms of their grammatical
structure, in certain languages). We will not press this point here (cf.
Lyons, 1968: 388ff). We are more concerned to discuss the distinction
between locative adverbials and nominals and, with it, the distinction
between equative structures like (3) and predicative locative structures
like (5). In doing so, we shall see that there is a further distinction
to be drawn between entity-referring nominals and place-referring
nominals,
That there is a semantic difference between identifying the referent of
one nominal with the referent of another nominal, on the one hand, and
saying of the referent of a nominal that it is located in a certain place, on
the other, is obvious enough, provided that the referent of the nominal
is an entity, rather than a place. But nominals in English may also refer
to places (e.g., ‘London’, ‘that field over there’); and, just as it is
possible to identify two entities, so it is possible to identify two places
(cf. London 1s the capital of England). It is also possible to say of one
place that it is contained within (and is therefore part of) another place
(cf. London is in England). |
The difference between locative adverbials and place-referring nomi-
nals is not, in fact, clear-cut in all syntactic positions in English. For
example, the demonstrative adverbs ‘here’ and ‘there’ and the demon-
strative pronouns ‘this’ and ‘that’ are equally appropriate as substitutes
for ‘this place’/‘that place’ in an utterance like Thzs/that place is where we
agreed to meet. Furthermore, there are many utterances of this kind that
are ambiguous according to whether they are construed as having an
equative or a predicative nucleus. Indeed, since the verb ‘be’ serves as
both an equative and a locative copula in English, it is difficult at times
to be sure that a particular sentence expresses the proposition “‘X be
(identical with) Y”’ or “X be at/in Y”’, if both X and Y are places. For
example, London ts where I met him, unlike My friend is where I met him,
is in principle ambiguous, although it would normally be understood as
equative (cf. “London is the place where . . .’’ vs. “‘ London is in the
place where .. .’’). Of course, the difference between “‘X be (identical
with) Y”’ and “X be at/in Y”’ disappears in the limiting case in which X
and Y are spatially co-extensive; and it is because the boundaries of the
referents of such locative adverbials as ‘here’ or ‘there’ (or ‘where I met
 12.3. Locative subjects 475
him’) are indeterminate that we cannot say that a sentence like ‘London
is where I met him’ necessarily has an equative nucleus.
The difference between entities and places is something that we shall
come back to in a later chapter (cf. 15.5). It is sufficient for our present
purpose to emphasize the fact that there are many nominal expressions
in English which can be understood as referring either to entities or to
places according to the context in which they are used. For example,
‘the church’ or ‘the house’ may refer to a physical entity, which,
though it is normally located in a particular place, would still be
identifiable as the same thing if it were moved to another place. But the
same expressions may also refer to places (or spaces) within which other
entities are located: cf. fohn is in the church. The question, therefore,
arises whether all such expressions (unlike those which may be used to
refer only to entities or only to places) should be regarded as inherently
ambiguous. If they are so regarded, their ambiguity might be most
satisfactorily accounted for by recognizing a deep syntactic distinction
in English between locative and non-locative nominals. Without pur-
suing this point any further for the present (cf. 15.5), we turn now to the
more controversial topic of locative subjects.

12.3. Locative subjects

It is an important fact about locative expressions that, even when they
occur as predicative complements, they are nonetheless referring expres-
sions; in this respect they are more like equative, than ascriptive, com-
plements.? So far, the only position in the nucleus of a kernel-sentence
that we have recognized as being possible for locative expressions is that
of a predicative complement with a first-order nominal subject: cf. (5)
in the previous section. (To avoid confusion we will make the numbering
of sentence-schemata in this section compatible with that of the pre-
vious section. Both sets of sentence-schemata are to be taken as belong-
ing to the same larger set of nuclear structures.)
We must also allow for equative sentence-nuclei containing two
locative expressions, as in
2 The topic dealt with in this section is one that is rarely, if ever, discussed in
these terms by linguists. It is for this reason that it is dealt with here at some
length. The reader should be warned that the views put forward here about
locative subjects and the difference between place-referring and entity-
referring expressions are somewhat idiosyncratic. In view of the thesis put
forward in this section Kuno’s paper (1971) is of considerable interest. So
too are the works of Prague School linguists, who have frequently drawn
attention to the necessity of paying particular attention to locative phrases in
the determination of functional sentence perspective (cf. 12.7).
476 Semantics and grammar III
(7) Loc (+Cop)-+ Loc.
Such structures will underlie sentences like ‘This place is London’ and
‘Here is where I met him’. The expression ‘this place’ is a nominal,
rather than an adverbial, in English. It is, however, a locative nominal;
and it has the same meaning as the demonstrative adverb ‘here’.
More interesting than (7) are structures with a locative subject and an
ascriptive complement. Once again, in English the locative subject, in
surface structure at least, must be a nominal: cf. ‘London is cold’,
“This place is cold’. It is not possible (as it is in certain languages) for
adverbials to occur in subject-position: there are no such sentences as
‘In London is cold’ and ‘Here is cold’. There are, however, such sen-
tences as “It is cold in London’ and ‘It is cold here’, which (although
they may differ slightly in meaning from ‘London/This place is cold’)
also express a proposition in which a property is being ascribed to a
place. Such sentences have always been regarded as problematical in
terms of the traditional assumption that the subject of a sentence is
necessarily a nominal. It is sometimes suggested that the pronoun ‘it’
is to be understood as a substitute for some such deep-structure nominal
as ‘the weather’. But there is no need to postulate an underlying
nominal, if we are prepared to grant that place-referring adverbials, as
well as place-referring nominals, may occur in subject-position in the
underlying nuclei of sentences. The pronoun ‘it’ can be inserted trans-
formationally as a purely surface-structure subject. It does not have to
be interpreted as a referring expression at all. Taking this point of view,
we may add to our list of sentence-schemata the following:

cold’. , :
(8) Loc (+Cop)-+A.
If this is the structure that underlies ‘It is cold in London’, the question
now arises whether it is also the structure that underlies ‘London is

There are two different ways in which ‘London is cold’ might be

derived from structures like (8). One is by recognizing both place-
referring nominals and locative adverbials as different subclasses of the
same deep syntactic category. The other is by treating ‘London is cold’
as a transform of a structure in which there is an adverbial subject: i.e. by
deriving both ‘London is cold’ and ‘It is cold in London’ from the
same underlying nucleus. It is the second of these solutions, as we shall
see, that is generally adopted in current versions of what is commonly
called case-grammar* (cf: 12.4). But-4London is cold’ is also derivable
from the ascriptive structure ~ —
 12.3. Locative subjects 477
(4) NP (-+Cop)+A,

provided that NP is held to cover, not only first-order nominals, but

also nominals that refer to places; and (4) is presumably the structure
that underlies such sentences as “London is huge’. Places may have
physical dimensions, just as first-order entities do; and some of them
have shape (cf. ‘That field is square’), Furthermore, place-referring
expressions are intersubstitutable with entity-referring expressions (and
may be conjoined with them) in many sentences. There is nothing
peculiar, for example, about the sentence ‘I like London, but not the
people who live there’. In any language in which it is possible to conjoin
place-referring and entity-referring expressions as subjects or objects of
the same verb they must, presumably, be given the same syntactic
classification. In at least some contexts, then, place-referring expres-
sions must be classified as nominals.
Locative expressions, it would appear, have of their very nature a
certain syntactic and semantic ambivalence; and it is not surprising to
find that there are sentences containing such expressions whose deep
structure is somewhat indeterminate. It is a highly significant fact that
sentences like ‘London is cold’ or ‘This place is cold’ (unlike ‘It is cold
here’, on the one hand, and ‘ London is huge’ or ‘ London is in England’,
on the other) can be analysed in terms of two different underlying struc-
tures. As we shall see later, there is also an inherent syntactic and
semantic ambivalence in definite noun-phrases in English containing
the definite article or a demonstrative pronoun; and this is explicable on
the assumption that it is not always clear (and in some cases it makes
little difference to the meaning of an utterance) whether an entity or a
place is being referred to (cf. 15.2).
A language-system imposes its own classificatory system of categories
upon the entities, places and other phenomena that we refer to and
describe; and different language-systems may, to some degree at least,
categorize the phenomena differently in terms of the grammatical and
lexical distinctions that they impose. It is easy to see, for example, that
the ascription of a property to a place can be looked at from two dif-
ferent points of view. We can think of the property as being associated
with the place exactly as one of the relatively constant perceptible
properties of a first-order entity is associated with that entity. This way
of looking at the relationship between a place and a property is fostered,
if it is not in fact created, by the use of sentences like ‘This place 1s
cold’, in which the place is referred to as if it were an entity (by means
478 Semantics and grammar III
of a nominal expression in subject-position). Alternatively, we can think
of the relationship between a property and a place as being comparable
with that which holds between an entity and a place. Looked at from
this point of view, the state-of-affairs described by ‘This place is cold’
might be described, somewhat unnaturally in English, by means of a
sentence like ‘There is a cold(ness) here’. The fact that we can bring
out the difference between these two points of view by means of a para-
phrase depends, of course, upon the fact that English, like many other
languages, draws a fairly sharp grammatical distinction between nouns
and adjectives and has available to it, as distinguishable structures,
NP (+Cop)-+-A and NP (-+Cop)-+Loc; and these structures serve as
templates, as it were, for the categorization of states-of-affairs in which
the distinction between entities and properties is, a priori, unclear.
It is not only sentences like ‘It is cold here’ that create problems for the
traditional assumption that every sentence must have a nominal subject.
We also have to reckon with the so-called impersonal* sentences that
are found in many languages. They fall into several subclasses, only one
of which concerns us here. It may be exemplified by means of the
English sentence ‘It is raining’, which (like ‘It is cold here’) contains a
dummy pronominal subject and is therefore not so obviously impersonal
as, say, the Italian or Spanish equivalents, ‘Piove’ and ‘Llueve’. That
the pronoun ‘it’ is not a referring expression in sentences like ‘It 1s
raining’ is obvious from the fact that it would normally be considered
nonsensical to enquire what is raining; and there are no acceptable
paraphrases in which expressions like ‘the weather’ can be substituted
for the dummy pronominal subject. We can, however, add to the
minimal sentence ‘It is raining’ a locative adjunct like ‘here’ or ‘in
London’, and it is arguable that there is always implicit in any utterance
of such a sentence some place-referring expression of this kind. If any-
thing is to be identified as the underlying subject (referring to that of
which the expression ‘be raining’ is predicated) it is surely this place-
referring expression. So we can add yet another possible sentence-
schema to our list in terms of the syntactic categories with which we are
operating:
(9) Loc+V.
This is comparable with (1) NP+-V, as (8) Loc (-++-Cop)+A is com-
parable with (4) NP (+Cop)-+A. It differs from (1), however, in that
the locative adverbial in (9) is syntactically dispensible (i.e. it is an
adjunct), as the locative adverbial in (8) is also syntactically dispensible
 12.3. Locative subjects 479
and thus differs from the NP in (4). But there are many languages in
which subject nominals are syntactically dispensible in all the structures
that we have been discussing. Paradoxical though it may seem to treat
the subject of a sentence as an adjunct, rather than as an essential
constituent of the nucleus, there is obviously nothing in principle
against doing this, if the grammatical structure of the language we are
describing suggests such a treatment.
We have just seen that there are structures like (8) and (9) in which
locative expressions (whether they are nominals or adverbials) can be
plausibly interpreted as subjects. We can go further than this in our
rather untraditional extension of the traditional notion of being the
subject of the sentence. If sentences like ‘This place is London’ are
analysed in terms of the equative structure
(7) Loc (+Cop)+ Loc,
comparable with
(3) NP (+Cop)+NP,
sentences like ‘This place is a city’ and ‘London is a city’ can be
analysed in terms of
(10) Loc (+Cop)+N,
comparable with
(4) NP (+Cop)+N.
The noun (N) that occurs in (10) must of course be a place-denoting
common noun. But in this respect (10) is no more restricted than an
ascriptive structure in which an entity-referring nominal (NP) is
coupled with a predicative noun (N): the predicative noun in such
cases must be one that denotes a class of entities.
Furthermore, the very fact that it 1s possible, in certain languages at
least, to construct definite referring expressions which we recognize as
being nominals, rather than adverbials, depends upon the possibility of
there being structures like (10). It 1s because the lexemes ‘place’,
. ‘city’, etc., in English are quite clearly common countable nouns, as
lexemes like ‘person’, ‘boy’, etc., are, that we are able to construct
expressions like ‘this place’, ‘this city’, etc., and conjoin them with
what are indisputably entity-referring expressions. It is not incon-
ceivable, after all, that there should be a language in which there were
no place-denoting common nouns, but only demonstrative adverbs (like
480 Semantics and grammar III
‘here’ and ‘there’) and place-referring proper names. It 1s also con-
ceivable that a language should have a set of place-denoting lexemes
which are distinguished syntactically or morphologically from the set of
entity-denoting lexemes. In which case, we would presumably have no
reason to call the place-denoting lexemes nouns. This point should be

things. |
borne in mind in any consideration of the traditional definition of com-
mon nouns as lexemes which denote classes of persons, places and

There remains but one.further sentence-scheme containing a locative

expression that we wish to add to our list:
(11) Loc (+Cop)+NP.
It will be observed that (11) is the mirror-image of _
(5) NP (+Cop)+ Loc.
Granted that we can have sentences whose characteristic function it is
to ascribe properties to places, it is natural to speculate about the
possibility of there being sentences, in any or all languages, whose
function it is to ascribe entities to places. Is it possible, in other words,
to treat an entity’s being in a place as a property of that place, rather
than treating the location of an entity as a property of the entity in
question? Sentences which can, though they need not, be interpreted in
this light include the following:
(11a) There is a book on the table
(11b) The table has a book on it.
Both of these sentences, it will be noted, contain an indefinite noun-
phrase, ‘a book’, which is most naturally interpreted as being non-
specific in reference (on any occasion of utterance). In this respect, they
may be contrasted with
(11c) The book is on the table,
which is much more naturally interpreted as expressing a proposition
about the book that is being referred to (on some occasion of utterance).
But neither (11a) nor (11b) can be satisfactorily analysed as expressing
a proposition which describes the entity that is referred to by ‘a book’
(even if we concede that ‘a book’ is a referring expression: cf. 7.2).
Standard transformational accounts of (11a) and (11b), however, which
derive them from structures like (5), would imply that ‘a book’ is the
deep-structure subject in (11a) and (11b) as ‘the book’ is in (11c). The
 | 12.4. Valency 481
derivation of (11a) and (11b) from a structure like (11), in which there is
a locative subject, would at least have the advantage that it brings the
postulated deep-structure subject into correspondence with the actual
surface-structure subject; and it is semantically plausible.
The syntactic analysis of sentences like (11a) and (11b) within a
transformational framework is highly controversial (cf. Jenkins, 1975);
and we cannot go further into this question here. Our purpose in this
chapter is the much more general one of presenting a number of dis-
tinctions which, regardless of the way in which they are handled in the
systematic analysis of particular languages, are semantically important.
Some of these distinctions, as we have emphasized throughout, are
difficult to draw in particular instances; and some of them may be drawn
more sharply in some languages than in others. This is true, for example,
of the distinction between nouns, verbs and adjectives; of the distinc-
tion between equative and ascriptive sentences; of the distinction
between nominals and adverbials or entities and places; and of the tradi-
tional distinction between subject and predicate.
In conclusion, we should perhaps make explicit and give particular
emphasis to one of the points that arises from our discussion of sen-
tences containing adverbial subjects. Our reasons for describing such
expressions as subjects were partly syntactic: such expressions tend to
be intersubstitutable with nominals that satisfy the normal conditions.
But we also appealed to the general logical, or ontological, distinction
between a property and what the property inheres in or is ascribed to.
One might argue, however, that relatively few of the sentences that we
have discussed can be said to have as their function the ascription of
properties to either entities or places. It is straining the term ‘property’
considerably to say that the activity in which some entity happens to be
engaged or its location in a particular place at a particular time should
count as one of its properties. It would be straining normal usage even
more to describe an event that is occurring in a particular place as a
property of that place. In short, property-ascribing propositions are but
a small subclass of the propositions that we have to deal with; and it is
perhaps one of the principal deficiencies of many standard logical
treatments of predication that they do not deal satisfactorily with any-
thing other than property-ascribing propositions.

12.4. Valency
Our discussion of the structure of sentence-nuclei has been deliberately,
though tacitly, restricted so far by the assumption that it is the function
 482 Semantics and grammar III
of propositions to ascribe properties to entities. As we saw towards the
end of the last section, however, it is only by stretching the sense of the
term ‘property’ considerably beyond its normal usage that we can refer
to the location of an entity as one of its properties (i.e. as one of its
distinctive attributes). In fact, it is only for a relatively small subset of
the propositions expressed by the sentences of English (and other
natural languages) that the notion of ascribing properties to entities is at
all satisfactory. We are more often concerned to describe the events and
processes in which persons, animals and things are involved than we are
to describe either the essential or contingent qualities of persons,
animals and things. Furthermore, much of what can be described in
terms of static properties or relations ~ in terms of the physical attri-
butes of an entity or its involvement in some state-of-affairs — can also
be described in the more dynamic terms of potentiality for action or
interaction. ‘To say that something has the physical property of hard-
ness, for example, is to say that it will resist pressure; to say that
something is located in such-and-such a place is to indicate where one
must go or direct one’s gaze in order to obtain or find the thing in
question.
It has been plausibly suggested, in several recent works on the
acquisition of language by children, that the earliest and most basic
grammatical constructions that are identifiable in children’s speech can
be accounted for as the product of what Piaget has called sensorimotor
intelligence (cf. 3.5).2 According to Piaget, cognition develops on the
basis of the child’s interaction with the persons and things in his
environment: it is by virtue of his operation upon them and their opera-
tion upon him that he comes to know their properties and to categorize
them conceptually. By the time that the child reaches the final stage in
the development, or maturation, of sensorimotor intelligence (when he is
between 18 and 24 months old), he can not only draw attention to, or
comment upon, persons and things in the environment that engage his
interest, but also comment upon their absence or disappearance and ask
where they are. From this it may be inferred that he has acquired a
conception of the existence and continuous identity, through time, of
what we are calling first-order entities. Among these, there are some that
he now knows to be, like himself, self-moving and others (himself, his
parents, etc.) that he knows to be both self-moving and capable of
operating, in various ways, upon other entities; and this, it is suggested,
. 3 'The Piagetian point of view is well explained in H. Sinclair (1972, 1973).
Cf. also Bates (1976), Brown (1973), Nelson (1974).
 12.4. Valency 483
provides the basis for the child’s earliest intuitions of such semantically
and grammatically important notions as animacy and agency. An agent
is, initially at least, any entity that is capable of operating upon other
entities, effecting some change in their properties or their location; an
animate being is one that is able to move itself without the intervention
of any external agency.
The child’s conception of animacy and agency may or may not
develop in this way. An alternative view is that it is innate. But whether
it is innate or not, there is little reason to doubt that it is universal in
men; and it constitutes an important part of naive realism (cf. 11.3).
The conceptual framework within which we organize and describe our
perceptions of the physical world, whatever language we speak, is one in
which we can identify, not only states-of-affairs of shorter or longer
duration, but also events, processes and actions. |
There is, unfortunately, no satisfactory term that will cover states, on
the one hand, and events, processes and actions, on the other. We will
use the term situation* for this purpose; and we will draw a high-level
distinction between static and dynamic situations. A static situation (or
state-of-affairs, or state*) is one that is conceived of as existing, rather
than happening, and as being homogeneous, continuous and unchanging
throughout its duration. A dynamic situation, on the other hand, is
something that happens (or occurs, or takes place): it may be momen-
tary or enduring; it is not necessarily either homogeneous or continuous,
but may have any of several temporal contours; and, most important of
all, it may or may not be under the control of an agent. If a dynamic
situation is extended in time, it is a process* ; if it is momentary, it is an
event*; and, if it is under the control of an agent, it is an action*.
Finally, a process that is under the control of an agent is an activity*;
and an event that is under the control of an agent is an act*.
What precisely is involved in the notion of agency is a difficult ques-
tion; and one that we will not go into here.t We may think of the
paradigm instance as being one in which an animate entity, X, inten-
tionally and responsibly uses its own force, or energy, to bring about an
event or to initiate a process; and the paradigm instance of an event or a
_ process in which agency is most obviously involved will be one that
results in a change in the physical condition or location of X or of some
other entity, Y. Each of the features that have been singled out for
* On agency and causality, cf. Cruse (1973), Davidson (1967), Fodor (1970),
Givoén (1975), Kastovsky (1973), Kenny (1963), Kholodovié (1969), Miller &
Johnson-Laird (1976), Wierzbicka (1972).
484 Semantics and grammar ITI
mention here — animacy, intention, responsibility and the use of its own
internal energy-source — is separable, in non-paradigm instances, from
each of the others. It is a fair assumption, however, that languages are
designed, as it were, to handle the paradigm instances; and it is only to
be expected that the applicability of notions like agency should be
unclear in non-paradigm instances.
If it is the case that the child first operates, at the level of practical
intelligence, with a concept of agency as physical manipulation, he very
soon learns that he can obtain things that he wants, not only by grasping
them himself, but also by using language to get others to pass them to
him. Indeed, the child’s realization that he can influence the behaviour
of other agents by emitting appropriate vocal and non-vocal signals
indicative of his interests and desires comes long before there is any
evidence that he is beginning to master the grammar or vocabulary of
any particular language-system; and it is upon this basis that the child
develops his understanding of the instrumental function of such
utterance-acts as requests and commands.
We shall be going into this question in some detail in a later chapter
(16.2). Here we are concerned to point out that language-behaviour
itself is activity, in the course of which acts of various kinds are per-
formed, over and above the purely physical, or physiological, acts in-
volved in the production of utterance-signals. What is now commonly
referred to as the theory of speech-acts* rests upon this fact (cf. 16.1). :
So too does our decision to apply the term ‘situation’, not only (as is
customary) to the contexts, or settings, in which utterances are produced
(cf. 14.1), but also to the states, events and processes that are described
by utterances. We must be careful, of course, to make it clear what is
being referred to when we correlate a particular situation with a par-
ticular utterance. It is important to realize, however, that there is no
terminological equivocation involved. The sense in which we have
introduced the term ‘situation’ in this section is broader than, but it
subsumes, the sense in which it is used to refer to the context, or setting,
of an utterance. To make a statement is to engage, as an agent, in a
particular kind of situation, which is related, by virtue of the descriptive
function of language, to another situation; and the situation that is
described by the propositional content of a statement may itself be that
of making a statement or some other kind of utterance (cf. He said that
it was raining, He toldlasked me to open the door).
Of particular interest, in the present connexion, is the fact that
languages can be used, not only descriptively, but also instrumentally:
 12.4. Valency 485
as tools with which to operate upon situations and change them. What
this means is that certain utterances may operate as components of
situations that are causally related to other situations. For example, the
command Open the door! may operate in some antecedent situation of
which the consequent situation of the door being open is the effect.
Generally speaking, the instrumental function of language involves
indirect, rather than direct, causation: the effect that is achieved is not
brought about directly by the utterance itself (as it is in the case of
utterances that are held to have a magical or sacramental character — cf.
Ali Baba’s Open sesame!), but by a secondary agent, to whom the
utterance is addressed. The initiating agent imposes his will upon the
effective agent by uttering an appropriate request or command. There
are, of course, various ways of indirectly causing something to happen
as an initiating agent. What is being emphasized here is the fact that the
instrumental function of language is explicable within the more general
account of agency, instrumentality and causation presented in this
chapter.
The distinction between static and dynamic situations is relevant to
the analysis of the grammatical category of aspect* in many languages
(cf. 15.6). It is also lexicalized in English in the opposition between such
verbs as ‘be’ and ‘have’, on the one hand, and ‘become’ and ‘get’, on
the other. The progressive aspect in English has as one of its semantic
functions that of representing a situation, not simply as existing, but as
happening, or developing, through time; and when it has this function,
it cannot be associated with a verb denoting a static situation (e.g.
‘know’). In so far as they both have a temporal extension, however,
states and processes (including activities) are similar. The fact that it
makes sense to enquire, with respect to a state or a process, “‘ How long
did it last?”’ is reflected by the fact that verbs denoting states or pro-
cesses may be used freely with such temporal adverbials as ‘for a long
time’. In this respect verbs denoting states and processes are in con-
trast with verbs denoting events (including acts), which occur with
punctual, rather than durative, adverbials; and in those languages in
which a syntactic distinction can be established between adjectives and
verbs, the majority of verbs, if not all, denote processes and events,
whereas the majority of adjectives denote states (cf. 11.3). There are
many languages, however, in which what is traditionally described as the
perfect tense of a dynamic verb may be used to represent a state as
having resulted from an antecedent event or process (cf. 15.6).
English, as we have seen, draws a fairly sharp distinction, within
 486 Semantics and grammar III
dynamic situations, between processes or events that are under the
control of agents, and those that are not. All processes and events are
happenings, but only actions (i.e. activities or acts) are doings. To ask,
in English, What ts happening? is to presuppose merely that some
process is taking place, and this process may or may not be an activity.
To ask, on the other hand, What ts X doing? or What ts being done
(by X)? is to presuppose that the situation in which one is interested is
an activity. This distinction between non-agentive and agentive hap-
penings is based, we may assume, on a universal appreciation of what
constitutes agency in paradigm instances; and it may well have its
reflex in the grammatical and lexical structure of all languages. But
languages differ, to some degree at least, with respect to the way in
which the semantic notion of agency is grammaticalized and extended to
non-paradigm situations, This point will be explained and illustrated in
the following section. But first we must introduce some further gram-
matically relevant dimensions 1n terms of which we can classify situa-
tions.
For this purpose, we will make use of the general concept of valency*,
which derives from Tesniére (1959: cf. also Heger, 1971; Helbig, 1971)
and has now been quite extensively employed (especially in recent
Soviet work: cf. Kholodovi¢, 1969, 1974; Apresjan, 1974) in the typo-
logical comparison of different language-systems. The concept of
valency can be seen, as far as its ancestry within linguistics is concerned,
as something which takes over and extends the more traditional, but
more restricted, notions of transitivity and government. But it is also
quite clearly relatable to the predicate-calculus classification of predica-
tors in terms of the number of arguments that they take in well-formed
formulae (cf. 6.3): a one-place predicator could be described, from this
point of view, as having a valency of 1, a two-place predicator as having
a valency of 2, and so on. What is traditionally described as a transitive
verb is a verb which has a valency of 2 and governs a direct object.
But valency covers more than simply the number of expressions with
which a verb may or must be combined in a well-formed sentence-
nucleus. It is also intended to account for differences in the membership
of the sets of expressions that may be combined with different verbs. For
example, ‘give’ and ‘put’, in their most common uses, both have a
valency of 3, but they differ with respect to one of the three expressions
which (in the extended sense of ‘government’) they may be said to
‘govern: ‘give’ governs a subject, a direct object and an indirect object;
and ‘put’ governs a subject, a direct object, and a directional locative.
 12.4. Valency 487
We will therefore say that they differ in valency: they are associated
with two distinct valency-sets*.
The number of distinct valency-sets in any language-system is quite
restricted ; and there would seem to be few, if any, verbs in any language,
with a valency of greater than 3. But there are in most languages, and
probably in all, grammatically productive mechanisms for decreasing or
augmenting what might be referred to as the intrinsic valency of a verb.
For example, transitive verbs in English are intrinsically bivalent; but
when they occur in the passive they are, like intransitive verbs, mono-
valent. The adverbial phrase ‘by John’, which occurs in the passive
sentence ‘The door was opened by John’ (in contrast with the nominal
‘John’ in the corresponding active sentence ‘John opened the door’) is
an adjunct and, as such, it does not belong to the sentence-nucleus. ‘The
passive of the transitive verb ‘open’, it will be observed, has the same
valency as the active of the intransitive verb ‘open’. The difference
between ‘The door was opened’ (without the adjunct ‘by John’) and
‘The door opened’ is that the former represents the situation as an act
in which the agent is not referred to, whereas the latter represents the
situation as an event (which may or may not be an act). The passive
voice, in languages in which this category is identifiable, is generally, if
not always, associated with a decrease of valency; and there are said to
be many languages in which the passive cannot be employed if the agent
is specified. One of the principal functions of the passive, in fact, would
seem to be that it provides for the description of an act or an activity
without specification of the agent or, alternatively, for the description of
a state which 1s represented as the result of some antecedent act.5
The converse process, whereby the intrinsic valency of a verb is
augmented rather than decreased, is found most obviously in those
languages in which there is a productive causative* construction (e.g.,
Turkish, Japanese, Georgian). This has the effect of increasing the
valency of the verb by 1, so that intransitive verbs become transitive, as
it were, and transitive verbs become trivalent. What is particularly
remarkable about these constructions is that the resultant derived
valency-set is usually identical with the intrinsic valency-set of other
verbs. For example, the valency-set of the causative of an intransitive
verb will be the same as the intrinsic valency-set of transitive verbs; and
the valency-set of the causative of a transitive verb will be the same as
the intrinsic valency-set of a verb which, like ‘give’ in English, takes

5 These several points emerge clearly from Kholodovié (1974).

488 Semantics and grammar III
both a direct and an indirect object.6 We will not go further into the
syntax or semantics of causative constructions (cf. Kholodovit, 1969).
What is of importance, in the present connexion, is the fact that the
augmentation of valency is accounted for by the provision of an addi-
tional place for an expression referring to the initiating agent; and this
expression then functions as the subject of the derived causative verb
(unless the whole construction is put into the passive). Although English
has no productive morphological causative, the same general effect may
be achieved by using one of the many verbs that denote different kinds
of agentive initiation (e.g., ‘make’, ‘get’, “‘persuade’): cf. ‘John came’
vs. ‘Peter made John come’, ‘John opened the door’ ws. ‘Peter per-
suaded John to open the door’.
It is obvious that there is a considerable degree of interdependence
between the meaning of a verb and its valency; and several different
attempts have been made recently to account for the valency of verbs
within the framework of what has come to be called case-grammar (cf.
Anderson & Dubois-Charlier, 1975). The term ‘case’, in this context,
is extended beyond its traditional application in much the same way
that the term ‘government’ is extended beyond its traditional applica-
tion in connexion with the notion of valency: ‘case’ here denotes such
semantic roles as those of agent, patient, cause, effect, source and goal.”
In order to avoid the necessity of distinguishing all the time between
two senses of the term ‘case’, we will refer to these as valency-roles*.
What are traditionally called cases (e.g., the nominative, accusative,
genitive, dative, etc., in Latin, German or Russian) would correlate only
imperfectly with valency-roles, which, in certain formulations of case-
grammar at least, are held to be universal components of various kinds
of states, events and processes.

12.5. Causativity and transitivity

The syntax and semantics of causative constructions have been exten-
sively discussed recently in connexion with the hypothesis of lexical
decomposition (cf. 9.9). According to what is probably the most widely
accepted formulation of this hypothesis, both the valency and the mean-
6 For example in French ‘Je lui ai fait manger sa soupe’ (‘‘I got him to eat his
soup’’), which is derivable by means of a productive causative construction
from the structure underlying the bi-valent ‘I] mange sa soupe’ (“‘ He eats his
soup’’), is parallel as far as its valency-set is concerned, with ‘Je lui ai donné
sa soupe’ (‘‘I gave him his soup’’).
? For a discussion of some of these roles, cf. Abraham (1971), Anderson (1971,
1975), Fillmore (1968, 1970), Huddleston (1970), Nilsen. (1972, 1973).
 12.5. Causativity and transitivity 489
ing of the transitive verb ‘kill’, for example, would be accounted for in
terms of the embedding of an intransitive structure containing the verb
‘die’ (more precisely, a complex predicator meaning ‘‘ come-to-be-not-
alive’) as the object of the abstract verb causE (cf. Dowty, 1972b).
The meaning of CAUSE would also enter into, though it would pre-
sumably not exhaust, the meaning of such English verbs as ‘cause’,
‘make’, ‘get’, which denote various kinds of agentive initiation and
take a variety of complements. Its subject in the underlying semantic
representation would be a nominal referring to an agent; and its object — ,
the embedded intransitive structure (with its own subject and predi-
cate) — would refer to the situation that is brought into being as a result
of the agent’s activity. Letting pie stand for the complex predicate
(meaning “‘come-to-be-not-alive’’) which occurs in the embedded
complement of CAUSE, we can say that X CAUSE (Y DIE) is transformed
into X CAUSE-DIE Y by an operation of pre-lexical predicate raising
and that CAUSE—DIE is lexicalized as ‘kill’ (cf. 10.5).
We are not concerned here with the hypothesis of lexical decomposi-
tion as such. Granted that causE does not have the same meaning as
the English verb ‘cause’ (and may not have exactly the same meaning
as any lexeme in any language), it is by no means clear that one can
argue for or against this analysis of the meaning of ‘kill’ on empirical
grounds. It is obvious that there are semantic differences to be accoun-
ted for between ‘He killed her’ and ‘He caused her to die’, ‘He made
her die’, ‘He got her to die’, etc. But these differences could be accoun-
ted for, in principle at least, by taking CAUSE to be more general in
sense than any of the verbs that denote agentive initiation and by
drawing distinctions between the propositions expressed by these several
sentences in terms of additional notions: direct vs. indirect causation,
coercive vs. non-coercive causation, etc. We will not go further into this
question (cf. Babcock, 1972; Fodor, 1970; Kastovsky, 1973; Lakoff &
Ross, 1972; Shibatani, 1972, 1973). The point is that there are pairs of
morphologically unrelated verbs in certain languages (cf. ‘die’ and
‘kill’, or ‘see’ and ‘show’, in English) which stand in the same semantic
relation to one another as do pairs of verbs that are related by means of
a productive morphological construction in other languages. It is also
thought that the vast majority of trivalent and bivalent verbs in all
languages are most commonly used with an agentive subject and that
their meaning is generally, though not always, causative. For example,
the bivalent verb ‘kill’ is causative in relation to the monovalent verb
‘die’, the trivalent verb ‘give’ is causative in relation to the bivalent
490 Semantics and grammar II]
verb ‘have’; and so on. What we are emphasizing here is this general
connexion between causativity and augmented valency.
But what do we mean by causativity*? The abstract predicator
CAUSE, employed in the previous paragraph, takes a first-order nominal
in its subject and a second-order nominal as its object (or complement).
It most naturally reflects, therefore, the notion of causality, according
to which agents are seen as the causes of the situations which, by their
actions, they bring into existence. But one can also talk, as we did earlier,
of one situation causing another; and this involves a somewhat different
conception of causality, which 1s compatible with, but does not pre-
suppose, agency. The relationship between these two different concep-
tions has long been, and still is, philosophically controversial. We need
not go into this question. What is of importance from the linguist’s
point of view is the fact that, although causality conceived as a relation
between two situations is logically distinguishable from agency, there is
what would appear to be a natural connexion between them; and both
the grammatical and the lexical structure of English (and other lan-
guages) reflect this connexion in several ways. We can say of a given
situation that it was produced, or brought about, by an agent. But we
can also say, no less naturally, that it was produced by his action; or
indeed, that it was produced by some prior event or process in which
there was no agent involved. We can say, in English, either fohn killed
Bill or Excessive drinking killed Bill (not to mention John/Excessive
drinking caused/brought about the death of Bill, etc.); either The umpire
stopped play or Rain stopped play; and so on. There is then a natural, and
presumably universal, tendency to identify causality with agency; and
we do not have to invoke some notion of primitive animism, in order to
explain the fact that, in English, nominal expressions referring to
reified physical forces (‘the wind’, ‘fire’, ‘rain’, etc.) are intersubstitut-
able with nominals referring to agents.
Causativity involves both causality and agency (in so far as they are,
in fact, distinguishable). It also depends upon the fact that the distinc-
tion between a single temporally extended situation and two distinct,
but causally connected, situations is not something that is given in
nature, as it were. Let us suppose for example that X picks up a knife
and stabs Y and that Y immediately falls to the ground dead. It is
obvious that what is assumed to have happened can be described as a
single event, as a process that is extended (albeit minimally) in time or
as a sequence of two or more situations (events, states or processes). By
using the verb ‘kill’, we can describe what has happened as a single
 12.5. Causativity and transitivity 4gI
event, in which there is an agent (X) and a patient (Y). The agent’s
action, however, is described solely in terms of its resultant effect upon
the patient. If we were to use the verb ‘stab’, we would necessarily
incorporate the further information that X acted directly upon Y in a
particular way and used some kind of instrument. Alternatively, we
could say X killed Y with a knife. This would imply that X’s action
involved the use of a knife; but, strictly speaking, it would not neces-
sarily imply that the knife was used to stab Y. In short, within the
limits imposed by the lexical and grammatical structure of English,
there are indefinitely many ways of picking out elements of what is to be
described and presenting them as components of a single situation. ‘The
fact there are so many transitive verbs with the same valency as ‘kill’,
not only in English and the Indo-European languages, but possibly in
all languages, would suggest that, as human beings, we are particularly
interested in the results of our purposive actions and in the effects that
our actions have upon patients.
It is important to realize, in this connexion, that the situation describ-
ed by X killed Y (if it is as we have supposed it to be) can be analysed
in terms of two different valency-schemata*. Looked at from one point
of view, ‘kill’ is what we will call an operative* verb: killing is an opera-
tion that is performed upon, and affects, the patient. Looked at from
another point of view, it is what is commonly called a factitive* verb: it
denotes a process or event whereby a cause produces an effect (or
result). The two schemata, therefore, in terms of which we can analyse
the situation of X’s killing Y are:
(1) AFFECT (AGENT, PATIENT) (operative)
(2) PRODUCE (CAUSE, EFFECT) (factitive).

Furthermore, by virtue of the connexion between agency and causality,

we have a third possible schema, which combines elements of both (1)
and (2). This 1s
(3) PRODUCE (AGENT, EFFECT) (operative—factitive).
It is easy to see that in what we have taken to be paradigm instances of
agentive situations (i.e. those in which the action results in a change in
the physical condition or location of the patient) all three schemata are
relevant. It is also easy to see that the causative account of the valency
and meaning of ‘kill’ (in which CAUSE is an abstract predicator, rather
than a nominal referring to a second-order entity, as it is in (2)) 1s
closer to (3) than to either of the others. What was represented earlier,
492 Semantics and grammar IIT
rather loosely, as X CAUSE (Y, DIE) can be reformulated as PRODUCE (X,
DIE (Y)), where X is the agent and (DIE (Y)) refers to the second-order
entity (Y’s death) which is the effect, or result, of X’s action. But the
proposition expressed by ‘X killed Y’ can also be understood as saying
that X did something to Y: i.e. it can be understood as an instance of
AFFECT (AGENT, PATIENT).
Of the three basic schemata introduced in the previous paragraph it
is (1) which most directly reflects the traditional notion of transitivity,
as a property of some, but not all, bivalent verbs. It is an important fact
that many of these verbs are also causative, in that they can be inter-
preted in terms of (3). Any verb whose valency can be accounted for in
terms of two (or more) different valency-schemata we will call am-
bivalent*.
There are many transitive verbs in English that are not ambivalent
with respect to (1) and (3). For example, ‘hit’, as it is used in ‘John hit
Bill’, cannot plausibly be analysed as an operative-factitive verb. We
can, of course, say that, in so far as some change is wrought in the
condition of Bill, John’s action results in a new state. But English does
not provide us with a monovalent predicator denoting such states (as it
provides us with ‘die’ and ‘dead’, which denote, respectively, the pro-
cess and state resulting from the action denoted by ‘kill’).
There are also transitive verbs which are factitive or operative-
factitive, but not purely operative: e.g., ‘make’, ‘produce’, ‘create’,
‘cause’. Such verbs are traditionally said to take an object-of-result*.
In terms of our analysis, this may be either a first-order or a second-
order nominal: cf. ‘God created Adam’, ‘John created a disturbance’.
But the first-order entity that is referred to by the object of a factitive
verb is clearly not a patient: it does not make sense to say What God did
to Adam was to create him. Furthermore, it is always possible to treat
factitive verbs as causative, even when they occur with a first-order
nominal as their object. For example, the proposition expressed by
‘God created Adam’ is related to the propositions expressed by ‘Adam
existed’ and ‘Adam came into existence’, in the same way that the
proposition expressed by ‘John killed Bill’ is related to the propositions
expressed by ‘John was dead’ and ‘John died’. One might perhaps
argue, therefore, that ‘God created Adam’ is a non-kernel-sentence,
which results from the embedding of a desentential transform within an
operative-factitive structure. However that may be, the operative-
factitive schema is clearly relevant to the derivation and interpretation
of a very wide range of constructions in English: not only such as are
 12.5. Causativity and transitivity 493
exemplified by ‘John made Bill scream’ and ‘You make me sick’, but
also such as are exemplified by ‘They elected him president’, ‘She
rubbed him dry’, ‘He drank himself silly’, etc. (cf. Halliday, 19674).
Factitive schemata in which the cause is a second-order identity (i.e.
a situation rather than an agent) are obviously relatable to what are
traditionally referred to as complex causal sentences: e.g., ‘They
stopped playing, because it had started to rain’ (cf. ‘Rain stopped play’
and ‘They stopped playing because of the rain’). Furthermore, causal
sentences, in the traditional sense of the term, are semantically related,
on the one hand, to conditional sentences and, on the other, to temporal
sentences. The nature of this relationship is controversial. But it is
obviously no accident that, in many languages there are parallels
between causal, conditional and temporal constructions: the utterance
of a sentence like ‘Bill fell to the floor, when John stabbed him’ will
normally be taken to imply that John’s action was the cause of Bill’s
falling. Similarly, a sentence like ‘Water boils if/when you heat it to a
temperature of 100°C’ will generally be taken to imply that being heated
is the cause of the water’s boiling. Whether or not our conception of
causality is innate or is based (wholly or partly) upon inductive inference
from our experience of pairs of temporally ordered situations, the asser-
tion that two situations succeeded one another in time will frequently be
intended, and understood, to imply that they are causally connected.
We cannot, in the space available, go any further into the semantics of
causality. But one final point may be made before we move on to a
consideration of some other valency-schemata. This is that there is a
distinction to be drawn, though once again it is not always sharply
applicable in practice, between causes and reasons. When we say Rain
stopped play, for example, we are presumably implying that, in the
opinion of those responsible, the occurrence of rain was a sufficient
reason for them to suspend the activity of playing. No direct physical
link is being postulated between the event of its raining and the event of
the cessation of play; and yet the sentence ‘Rain stopped play’ is
grammatically indistinguishable from one that might be used to assert a
causal connexion between the two events. Causes, under our analysis of
causality, are second-order entities. Reasons, however, being proposi-
tional in nature,are third-order entities. There is a distinction to be
drawn, too, between real and alleged reasons. The sentence Helen was
upset because I forgot her birthday might be used in order to give what
the speaker knew or believed to be the real reason. But it might also be
uttered, in other circumstances, in order to tell us what Helen herself
494 Semantics and grammar III
had alleged to be the reason. The truth-conditions of the propositions
expressed by ‘Helen was upset because I forgot her birthday’ are quite
different under the two interpretations; and there are languages (e.g.,
Latin) in which this sentence would be translated differently (with the
subjunctive mood rather than the indicative) if the speaker were posi-
tively indicating that the reason that he is giving is not necessarily the
real reason.

12.6. Partictpant-roles and circumstantial roles

The valency-schemata, (1)—(3), listed in the previous section, will cover
many, though not all, of the dynamic bivalent verbs that are tradition-
ally described as transitive; and, as we have seen, the class of transitive
verbs may be extended, in many languages, by means of a productive
grammatical process of causativization.
Verbs-of-motion — 1.e. verbs denoting a process in the course of which
some entity changes its physical location — constitute another important
subclass of dynamic bivalent verbs. Typical members of this class are
‘come’ and ‘go’ in English, which may take as their complement an
expression referring to either the source (‘from Edinburgh’) or the goal
(‘to Edinburgh’) of the locomotion, In order to handle the valency of
verbs-of-motion, we therefore need to add to the valency-schemata
discussed in the previous section the following
(4) MOVE (ENTITY, SOURCE)
(5) MOVE (ENTITY, GOAL).

Since all locomotion necessarily involves both a source and a goal, (4)
and (5) can be combined to yield
(6) MOVE (ENTITY, SOURCE, GOAL).

Furthermore, since an entity may be moved from its source-location to

its goal-location by an agent, the situation described by either (4) or (5),
or both, may be treated as the effect in an operative-factitive schema
(3a) PRODUCE (AGENT, (MOVE (ENTITY, SOURCE)))
(3b) PRODUCE (AGENT, (MOVE (ENTITY, GOAL)))
(3c) PRODUCE (AGENT, (MOVE (ENTITY, SOURCE, GOAL))).

Most of the verbs in English whose meaning can be analysed in terms

of these three schemata are generally classified as transitive verbs (i.e.
as being bivalent, rather than trivalent): ‘remove’, ‘bring’, ‘take’, etc.
The reason why they are so classified is simply that, whereas there must
 12.6. Participant-roles and circumstantial roles 495
be an expression referring to the entity that is moved, both the source
and the goal may be left unspecified, in grammatically well-formed
sentences: expressions referring to the source or the goal, being syn-
tactically optional, are therefore regarded as adjuncts, rather than
complements. The.verb ‘put’, however, requires both a direct object and
a further complement referring to the goal. Neither ‘the book’ nor ‘on
the table’ can be deleted from ‘John put the book on the table’ (in the
way that ‘from the table’, but not ‘the book’, may be deleted from
‘John removed the book from the table’) without destroying the gram-
maticality of the sentence. Directional* schemata of the several kinds
listed here are relevant to much else in languages over and above the
analysis of verbs-of-motion (cf. 15.7).
There is little reason to doubt that each of the valency-roles dis-
cussed so far is universal; and they have figured prominently in various
versions of what is commonly called case-grammar (cf. 12.4). They may
also be ontogenetically basic, since expressions which fulfil these roles
are readily identifiable in some of the earliest utterances produced by
children after they have passed through the holophrastic stage (cf.
Brown, 1973). Indeed, if we add to schemata (1)-(6), which have been
set up here for propositions describing dynamic situations, the following
two valency-schemata to handle static situations, we can account for all
but a very small fraction of young children’s utterances:
(7) BE (ENTITY, ATTRIBUTE/CLASS)
(8) BE (ENTITY, PLACE).

The valency-role of attribute/class in (7), it should be noted, allows for

the possibility that, despite the interconvertibility of properties. (or
attributes) and classes in most systems of logic, there may be an impor-
tant difference between them in many, if not all, language-systems. As
we have seen, this difference partly depends upon and partly supports
the distinction between adjectives and nouns.
In the set of nine valency-roles with which we have now furnished
ourselves ENTITY and PLACE may be regarded as being unmarked, or
neutral, in relation to the other six more positive roles associated with
the schemata (1)-(6). Of these six, AGENT and PATIENT are roles that are
assumed by first-order entities (typically persons); CAUSE and EFFECT are
roles fulfilled by second-order entities; and SOURCE and GOAL are roles
fulfilled by places.
Most recent treatments of case-grammar tend to give the impression
that only nominals may fulfil valency-roles in the propositional nuclei
496 Semantics and grammar III
of sentences. This is not so. Locative (and directional) adverbs may also
occur as the complements of the appropriate verbs in structures that
conform to the valency-schemata (4), (5) and (8). Phrases like ‘in
London’, ‘from London’ and ‘to London’ in English contain a nominal
(‘London’), but they are themselves adverbials. Although places are not
entities, in certain constructions (and in certain languages to a greater
degree than in others) they may be treated as entities (cf. 12.3); and the
converse is also true. Entities maybe treated as places. The so-called
localist* version of case-grammar (cf. Anderson, 1971, 1975) depends
upon this fact.
At first sight, there would appear to be a sharp distinction between a
situation in which one entity affects another entity and a situation in
which an entity moves to or from a place. But such transitive verbs as
‘hit’ and ‘kill’, which we have associated with (1) and (3) respectively
are traditionally described in terms which suggest that the agent is the
source of the action and that the patient is its goal. Indeed, the very
term ‘transitive’ derives from this conception of the way the agent not
only operates upon, but directs his action at, the patient; and, as far as
verbs like ‘hit’ (or ‘grasp’) are concerned, the traditional association of
transitivity with goal-directed activity is clearly quite appropriate. The
entity that is referred to by means of the expression that functions
syntactically as the direct object is both the patient, which (as traditional
terminology puts it) suffers the effect of the action, and also the goal of
movement. Just as there are verbs that are ambivalent with respect to
(1) and (3), so there are verbs that are ambivalent with respect to (1)
and (6) or (3) and (6). For example, John’s hitting Bill can be seen in
terms of John’s movement towards Bill (or of John’s fist moving away
from him towards Bill). It can also be seen, however, in terms of John’s
action being the cause of whatever effect is produced in Bill. In so far as
verbs like ‘hit’ are typical of the class of transitive verbs taken as a
whole, there is therefore a natural connexion between agency, causation
and the source of movement, on the one hand, and between suffering
the effect of an action and being the goal of movement, on the other
hand. To say that an entity is either the source or the goal of movement
is to treat that entity as a place.
As we have seen, the syntactic distinction between nominals and
adverbials correlates, though only imperfectly, with the syntactic dis-
tinction between the subject or complements of a verb and its various
adjuncts. This latter distinction also correlates, though again imper-
fectly, with a further distinction that is commonly drawn between the
 12.6. Partictpant-roles and circumstantial roles 497
valency-roles, or participant-roles*, and the circumstantial* roles
associated with a situation (cf. Halliday, 1970b). If we are describing an
action in English, we may tell our interlocutor, not only who did what
to whom (or what), but also when, where, how or why he did it. Gener-
ally speaking, however, we are not obliged by the grammatical and
lexical structure of English to give this circumstantial information
(except in so far as we must use one tense rather than another and thus
relate the situation we are describing to the time of utterance). These
circumstances are normally referred to by means of syntactically
optional adverbs or adverbials, whereas valency-roles are associated, in
what we may take to be the kernel-sentences of English, with nominals
(and, in certain instances, place-referring adverbials) functioning as the
subjects or complements of the verb.
As far as English and many other languages are concerned, it would
seem that there is a hierarchical ordering within the valency-roles and
the circumstantial roles associated with particular kinds of situation and
this hierarchical ordering determines, in part at least, which expressions
will be included in the sentence-nucleus and whether they will function
as subjects, direct objects, indirect objects or as complements of some
other kind. For example, the instrument with which an agent performs
some action is normally referred to by means of an adjunct in English:
e.g., ‘with a knife’ in ‘John opened the letter with a knife’. It 1s possible,
however, to promote the instrumental expression to nuclear status by
employing the transitive verb ‘use’ whose valency is such that it takes
an AGENT and an INSTRUMENT as its arguments. ‘John used a knife to
open the letter’ expresses a complex proposition in which the adverbial
phrase ‘to open the letter’ is an adjunct rather than a constituent of the
nucleus. The instrumental expression can, in certain instances, also be
used as the subject of a verb which would normally take as its subject an
expression referring to the agent. But the promotion of an expression
referring to the instrument with which an action is performed (or more
generally of an expression referring to one of the circumstances of a.
situation) from adjunct status to that of subject or complement in the
. gentence-nucleus always constitutes a deviation from what 1s the most
usual and the most neutral way of describing a situation.
We have been assuming that there is a neutral or normal way of
categorizing situations from which deviations can be made for the pur-
pose of emphasis or contrast. A fuller treatment of this question would
involve us in discussing, not only the characteristic association of
particular valency-roles with particular syntactic functions in sentence-
 498 Semantics and grammar III
nuclei, but also the role of word-order, stress and intonation, and a
variety of other devices that are used in context-dependent, emphatic or
contrastive statements. ‘he point to be emphasized here is that, in many
languages if not in all, there are good reasons for saying that sentence-
schemata of the kind that we have been discussing, in so far as they are
used to describe dynamic situations, are generally filled by expressions
that provide the answers to such questions as ‘‘ What happened to X?”’,
or “What did X do to Y?” (where X and Y are first-order entities),
| rather than to such questions as “‘What was used (by X) in doing
what?”’, ‘““Where/when did what happened take place?”’, “How did
what happened come about?” and so on.
This presumably reflects our greater interest, in general, in partici-
pants rather than circumstances. After all, it is easy enough to conceive
of a language-system in which sentences necessarily included an ex-
pression referring to the time, place, manner and purpose of an activity
and only optionally included, as adjuncts, expressions referring to the
human participants. It is doubtful whether there are any such languages.
All languages, however, may well provide the means whereby what
would be an adjunct in a kernel-sentence is promoted to nuclear status
in a non-kernel-sentence: ‘The reason for my being late is that I missed
the train’, ‘The way they escaped from prison is by tunnelling under
the wall’, etc., in contrast with ‘] missed the train because I was late’,
‘They escaped from prison by tunnelling under the wall’, etc. Some-
thing more will be said about the various ways in which a situation can
be described in other than the most neutral way in the following section.
~ Here it should be noted that non-kernel-structures of the kind that have
just been illustrated make use of the process of nominalization to-create
| second-order nominals and then fit these expressions into equative or
predicative structures of the kind that contain first-order nominals in
the participant-roles in kernel sentences. In this respect, therefore, the
valency-schemata that we have been discussing may be said to reflect
the most basic and most neutral way of conceptualizing and describing
a situation. An expression referring to the agent will tend to be made the
subject of the verb, an expression referring to the patient will tend to
be made the object, an expression referring to the instrument will tend
to be excluded from the nucleus and made into an adjunct, and so on.
At the same time, there will probably be a small number of verbs (like
‘use’) whose valency is such that they permit an alternative categoriza-
tion of the situation within which a circumstantial may be treated as a
participant.
 12.6. Partictpant-roles and circumstantzal roles 499
As there is a hierarchy among the valency-roles and circumstantial
roles, such that an expression higher in the hierarchy will tend to be
selected as subject (unless there are special reasons of emphasis or
contrast for it not to be), so there are similarities between the roles, such
that we see it as natural that two distinguishable roles should be gram-
maticalized or lexicalized in the same way in particular languages. We
have already seen that there is a natural association of AGENT, SOURCE
and CAUSE; and this accounts for the fact that the same case or prepo-
sition may be used to denote these roles in particular languages. But
there are all sorts of equally natural connexions between other valency-
roles and circumstantial roles that have been identified in the several
versions of case-grammar, and it is this which casts doubt upon the view
that there is a fixed set of universal roles, which are identifiable and
distinct in all situations, but which may be grouped in various ways in
different languages. The hypothesis that this is so has led to the prob-
lem that it is very difficult to say in particular instances whether a given
expression is fulfilling one role to the exclusion of another or both
simultaneously.
It is perhaps more plausible to assume that, instead of there being a
set of universal valency-roles and circumstantial roles for all languages,
there are certain universal principles of cognition and perception (which
may or may not be innate) and that the application of these principles
to the situations that are described by language permits a considerable
range of variation in the way in which these situations can be categor-
ized. For example, if someone uses a tool to perform some action upon
something else, the tool may be regarded (like the effective agent in a
causative situation) as an intermediary (i.e. as a kind of secondary
agent); it may be seen as the path through which the action travels; it
may simply be seen as a concomitant entity or as something that the
agent is holding in his hand. The instrumental role as such need not be
distinctively represented by means of a particular case or a particular
preposition; and whether it is so represented will vary from one lan-
guage to another.
The acquisition of the grammatical and lexical structure of a language
would appear to be part of a developmental process in which successively
more abstract structures are built upon the basis of more concrete
structures. In the course of this process, syntactic patterns that are
originally used for a more restricted set of situations will serve as tem-
plates, as it were, for the description of a progressively wider set. Their
extension to this wider set, however, will not necessarily proceed on the
 500 Semantics and grammar III
basis of the same analogies in all languages. As we have just seen, there
are several ways in which situations may be categorized. It follows that,
even though there might be much in the grammatical structure of
languages that is universal, there will be much, and perhaps far more,
that is not. At the time of writing, grammatical theory is in considerable
flux; and there is far from being any kind of consensus as to the status
of the valency-roles and circumstantial roles that have been mentioned
in this section.

12.7. Theme, rheme and focus

In this section we shall be concerned with what is commonly referred to
. nowadays as the thematic* structure of utterances: the way in which an
utterance is organized, grammatically and phonologically, as a signal
encoding a particular context-dependent message (cf. Halliday, 1970b:
16off).®
We may begin by considering the several interconnected, but dis-
tinguishable, senses in which the term subject* (vs. predicate*) is used.
There is, first of all, a distinction to be drawn between expressions and
their referents (cf. 7.2). In terms of this distinction we might say, with
respect to the following utterance
(1) John ran away,
either (i) that the expression ‘John’ is the subject or (1i) that the referent
of ‘John’ (i.e. John) is the subject. Generally speaking, the term
‘subject’ is applied by linguists to expressions rather than to their
referents. In what has been said so far about subjects and predicates we
have adhered to this terminological convention (cf. 11.2); and we will
continue to do so. It follows from another terminological convention
introduced in an earlier chapter, according to which properties are
ascribed* to entities by predicating* expressions of entities, that the
predicate is not predicated of the subject, but of the referent of the sub-
8 What is referred to as thematization* in this section is often called topicaliza-
tion*, Fillmore (1968) distinguishes between primary and secondary topi-
calization, the former having to do with the processes whereby the gramma-
tical subject of the sentence is determined. His comment on the potential
implication of Gruber’s (1967) study of topicalization in child-language is
worth quoting: ‘‘It may be that when one device for topicalization becomes
habitual, it freezes into a formal requirement and the language must then
call on other processes for motivated topicalization”’ (Fillmore, 1968: 58). It
should be noted that Chomsky (1976: 149ff) and Jackendoff (1972: 29ff)
employ the term ‘thematic’, unfortunately, for what we are calling valency
relations.
 12.7. Theme, rheme and focus 501
ject (cf. 6.3). There is a good deal of variation, not to say inconsistency,
in the literature with respect to what is said to be predicated of what; and
it is not uncommon to meet statements to the effect that an utterance
like (1) is being used to say something about the subject, rather than
about the referent of the subject.
According to the earliest formulation of the distinction between sub-
ject and predicate in the Western grammatical tradition, the subject is ,
the expression that is employed by a speaker to identify what he is
talking about and the predicate is the expression that is used to say what
he wishes to say about it. This notion of subject and predicate is implicit,
it will be recalled, in the passage quoted from Sapir in the preceding
chapter (cf. 11.1). It is sometimes referred to by means of the distinction
between topic* and comment*: ‘The speaker announces a topic and
then says something about it .. . In English and the familiar languages of
Europe, topics are also subjects and comments are predicates”’ (Hockett,
1958: 201). The subject, then, is the expression which refers to and
identifies the topic and the predicate is the expression which expresses
the comment. Needless to say, the topic-comment criterion does not
apply, other than derivatively, to questions, requests or commands. But
this is not a serious problem. Given that John is the topic of (1) and that
‘John’ is the subject, we can readily identify ‘John’ as the subject of the
question
(2) Did Fohn run away?
, on the basis of its systematic grammatical relationship with (1). In so
far as it is natural to say that in making a statement the speaker is com-
menting upon some topic, it 1s also natural to say that in uttering a
question like (2) he is enquiring about some topic. It is rather less
natural to refer to John as the topic of
(3) Run away, John.
But much the same considerations of grammatical parallelism as lead us
to say that the topic-comment criterion is applicable derivatively to
questions might also lead us to say that it is applicable, again deriva-
tively, to requests and commands. We will not pursue this point. The
relationship between corresponding statements, questions, requests and
commands will occupy us, for other reasons, later (chapter 16).
It is more instructive, in the present connexion, to consider what
* Like many authors, Hockett does not distinguish expressions from their
referents. We will say that John, not ‘John’, is the topic in (1).
502 Semantics and grammar III
justification one might have for saying (as we might be inclined to say)
that in uttering (1) as a statement the speaker is making a comment
about the referent of ‘John’. Could he not be commenting upon run-
ning away? Or upon something that is not referred to or mentioned in
the utterance itself? Indeed, he could. And yet, in default of any con-
textual information to the contrary and on the assumption that (1)
carries a neutral non-emphatic intonation-contour and stress-pattern,
we would normally be happy enough to say that the speaker is com-
menting (or asserting) of John that he ran away. In doing so, we would
be appealing, whether overtly or covertly, to yet another traditional
criterion for the distinction of subject and predicate: the logical cri-
terion, that in any proposition in which a particular term is combined
with a general term, the particular term is the subject and the general
term is the predicate (cf. Strawson, 1959, 1974). This is based, ulti-
mately, upon the ontological distinction between individuals and pro-
perties (or, in Aristotelian terminology, between substances and
accidents: cf. Lyons, 1968: 337). As we have already seen, the onto-
logical distinction between individuals (first-order entities) and proper-
ties, states, processes, etc., correlates with what may well be a universal
distinction in human languages between nominals and non-nominals
(cf. 11.3). In so far as (1) the grammatical structure of the utterance is
isomorphic (in all relevant respects) with the structure of the proposition
that it expresses and (ii) the proposition that is expressed is clearly
analysable into a subject-term and a predicate-term by applying to it the
logical criterion, we are justified in distinguishing, in utterances such as
(1), what may be referred to as a logical subject and predicate.
It is customary nowadays for linguists to extend the application of the
term ‘logical subject’ beyond the point at which it can be justified in
terms of the traditional distinction of universals and particulars. As we
saw earlier, there are comparatively few propositions expressible by
means of the sentences of natural languages that are naturally thought
of as being composed of a single entity-referring expression and a single
property-denoting expression (cf. 12.4). Natural language-systems seem
to be designed, as it were, to describe dynamic, rather than static,
situations — situations in which, typically, there is an agent who is
presented as the source of the activity. In sentences that express propo-
sitions describing such situations the expression referring to the agent
is commonly called the logical subject. It is arguable that this is a very
different usage of the term ‘logical subject’ than the usage which rests
upon the traditional logical distinction of subject and predicate. How-
 12.7. Theme, rheme and focus 503
ever, since the term is currently employed, in linguistics at least, in the
extended sense, we will adopt this usage in what follows.
The logical subject and predicate may or may not be the same as the
expressions that would be identified as subject and predicate by virtue
of the topic-comment criterion, even when the topic~-comment criterion
is independently applicable, It may be assumed, however, that in general
the two criteria tend to coincide and that, when they do coincide, if
there is a thematically neutral version of an utterance (as distinct from
one or more non-neutral versions of what might otherwise be regarded
as the same utterance) this thematically neutral version will be used.
Indeed, it is by virtue of this coincidence of thematic and logical sub-
jects in particular instances that we can distinguish thematically neutral
(or unmarked*) from thematically non-neutral (or marked*) utterances.
We have been talking about (1) on the assumption that it is uttered with
normal, non-emphatic stress and intonation. But normality cannot be
defined otherwise than with reference to the very criteria that are under
discussion here.
Every statement that can be made by uttering a simple sentence
expresses a proposition, which, if it is informative (cf. 2.1), provides the
answer to either an explicit or an implicit question. If we wish to make
the question explicit in English, there is no way in which this can be
done without making certain presuppositions about the situation in
which we are interested. We must categorize it as dynamic or static; and
we must also reveal our assumptions as to whether the situation is in the
past, present or future, whether it is timeless or hypothetical, and so on.
There is no way of asking by uttering a simple interrogative sentence:
‘‘Given that some situation has been, is or will be in existence or in
progress, what kind of situation is it and what entities and circumstances
does it involve?’”’ In any question that we might put relating to the
components or circumstances of a situation, there is something that is
presupposed* and something that is in focus* (cf. Chomsky, 1969). For
example, in asking Who is X? we presuppose that X is a person and
focus our question upon his identity; in asking What happened? we
presuppose, minimally, that some event or process occurred; in asking
Why did John come home late?, we presuppose that John came home late;
and so on.
According to the explicit or implicit question that a statement is
intended to answer, so the utterance will have one rather than another
prosodic contour imposed upon it. In particular, if more than the
minimal presuppositions are made in the question that the statement
504 Semantics and grammar ITI
answers, the utterance will be pronounced with something other than
what we have been referring to as neutral, or unmarked, stress and
intonation. The phonetic details do not concern us here. The important
point is that thematic neutrality or non-neutrality is determined by the
presuppositions that the speaker makes and that one of the correlates of
thematic neutrality or non-neutrality, in English and presumably in all
languages, is stress and intonation. It is when (1) is used without the
more specific presuppositions that would be embodied in such questions
as Who ran away? or Did fohn run away? that it may be described as
thematically neutral.
So far we have distinguished two kinds of subjects: logical and
thematic. There is yet a third kind of subject to be recognized: the
so-called grammatical subject. How this is defined will vary from
language to language; and it may well be there are languages in which
there is no reason to distinguish the grammatical subject from the
thematic subject. In many languages, however, the grammatical subject
of a sentence may be identified as the nominal which determines verbal
concord; and, to a limited extent, this is so in English. Another common
indication that a nominal is the grammatical subject of the sentence in
which it occurs is its being inflected for the grammatical category of
case. Once again, this is so to a very limited extent in English: he (vs.
him) is the form of ‘he’ that occurs when ‘he’ assumes the role of
grammatical subject. A third indication that a nominal is the gram-
matical subject, as far as English is concerned, is the position of the
nominal relative to other nominals in the sentence in which it occurs.
For example, in a sentence like ‘John killed Bill’ neither verbal concord
nor the grammatical category of case, but merely its position relative to
the verb and to ‘Bill’, serve to identify ‘John’ as the grammatical
subject.
As there tends to be coincidence between the logical and the thematic
subject, so there tends to be coincidence between the logical and the
grammatical subject, on the one hand, and between the thematic and the
grammatical subject, on the other. It is well known, however, that the
logical subject (i.e. the expression referring to the agent) is distinct
from the grammatical subject in passive sentences, such as
(4) Bill was killed by an unknown assassin.
In the standard theory of transformational grammar proposed by
Chomsky (1965), the logical subject of a sentence like (4) would be its
deep-structure subject and the grammatical subject would be its surface-
 12.7. Theme, rheme and focus 505
structure subject (cf. 10.3). Many linguists would deny that the dis-
tinction of subject and predicate is relevant at the deepest level of
grammatical analysis. It would be generally agreed, however, that, in so
far as the distinction between the logical and grammatical subject is a
matter of syntax, sentences in which the logical and the grammatical
subject coincide are transformationally simpler than sentences in which
the two kinds of subject do not coincide. This thesis is compatible with,
but does not imply, the stronger thesis, maintained by Chomsky (1957,
1965), Harris (1968, 1976), and others, that the structures in which the
logical and the grammatical subjects do not coincide are transforma- ,
tionally derived from structures in which the two kinds of subjects do
coincide. If it is assumed that (4) expresses the same proposition as
(5) An unknown assassin killed Bill,
and that the subject-term of the proposition is whatever term correlates
with ‘an unknown assassin’ (rather than the term that correlates with
‘ Bill’), it may also be assumed that the syntactic structure of (5) is more
similar to the logical structure of the proposition expressed by both (4)
and (5) than the syntactic structure of (4) is.
It may be noted at this point that whereas we have been talking of
utterances in connexion with the notion of the thematic subject and
predicate, we have switched from utterances to sentences in order to
introduce the distinction between grammatical and logical subjects. The _
reason for this switch is that, in terms of the distinction that is drawn in
this work between sentences and utterances (cf. 1.6, 14.6), thematic
subjects are, first and foremost, utterance-constituents. Grammatical
subjects, on the other hand, are established by virtue of criteria that
apply primarily to system-sentences and only derivatively to utterances
that we put in correspondence with system-sentences. We can say that
‘John’ is the grammatical subject in the utterance John ran away
because it is the grammatical subject in the corresponding system-
sentence ‘John ran away’. But we cannot say that ‘John’ is the thematic
subject in the system-sentence ‘John ran away’: this system-sentence is
in correspondence with several prosodically (and paralinguistically)
distinct utterances, in some of which ‘John’ would not be the thematic
subject.
As the notion of the grammatical subject. applies primarily to sen-
tences and only derivatively to utterances, so the notion of the logical
subject applies primarily to propositions and derivatively, but inde-
pendently, both to sentences and to utterances. That an expression may
506 Semantics and grammar Il]
be identified as the logical subject of an utterance independently of its
status in the corresponding system-sentence will be evident from what
has been said already: we do not have to relate John killed Bill to ‘John
killed Bill’ in order to decide that in the former, as in the latter, ‘John’
is the logical subject. All we need to know is that ‘John’ refers to the
agent in the situation described by the proposition that is asserted.
We have seen that the logical subject and the grammatical subject will
not necessarily coincide: cf. (4). That the grammatical subject need not
coincide with the thematic subject in English utterances is apparent
from the consideration of such examples as
(6) John Smith I haven't seen for ages.
Here the grammatical subject is ‘I’, but the thematic subject is ‘John
Smith’. However, English, in contrast with many other languages
(including German and Czech), shows a very definite tendency to iden-
tify the thematic and the grammatical subject; and it has often been
pointed out that one way of doing this is by employing a passive, rather
than an active, construction. Utterances like (6) are relatively uncommon
in Modern English; and they are even more uncommon perhaps when
the grammatical subject is something other than a personal pronoun.
The triple distinction of logical, grammatical and thematic subject,
with which we have been operating, emerged in the course of the nine-
teenth century (cf. Sandmann, 1954). It is linguists of the Prague
School, however, who have so far done most to elucidate and elaborate
the difference between the thematic subject and other kinds of subjects,
in work that started in the 1920s and still continues (cf. Vachek, 1964,
1966; Firbas, 1964, 1972; DaneS, 1968).!° One of the most important
characteristics of the Prague School, which in the heyday of structural-
ism distinguished it most strikingly from other schools of structural
linguistics, was its emphasis on functionalism* (cf. 8.3); and the interest
that Prague School linguists showed in thematic structure was but one
aspect of their concern with the way language-systems are designed, as it
were, to perform their communicative functions (cf. Sgall et al., 1973).
It is to the Prague School that we are indebted for the terms theme* and
rheme*, in the sense in which they are being used here: ‘theme’ is of
course quite widely employed outside linguistics in what can be seen as
a related sense (though in ordinary usage the theme is what one is
10 For related work, cf. Bolinger (1952), Chomsky (1969), Halliday (1967b),
Halliday & Hasan (1976), Huddleston (1971: 315ff), Kirkwood (1969, 1970),
Kuno (1972b).
 12.7. Theme, rheme and focus 507
talking about, not the expression with which one identifies, or announces,
what one is talking about); the term “‘rheme’ goes back to the Greek
word ‘rhéma’ (‘‘what is said’’), which, by way of the Latin ‘verbum’,
is the source of the term ‘verb’ and its correlates in other languages —
‘rheme’ is employed by Prague School linguists to refer to the ex-
pression which contains the information which the speaker wishes to
communicate.
The theme, we have said, is the expression used by the speaker for
what he announces as the topic of his utterance: it is the thematic sub-
ject. Not surprisingly there is a very high correlation, not only in
English, but in all languages, between occupying initial position in the
utterance and being thematic, rather than rhematic. As far as English is
concerned, Halliday (1967b, 1970b) makes initial position in the clause a
necessary condition of thematic status, saying that, whereas in
(7a) Fohn saw the play yesterday
‘John’ is the theme, in
(7b) Yesterday Fohn saw the play
‘yesterday’ is the theme. In uttering (7a), the speaker gives notice, as it
were, that he is talking about John; and in uttering (7b), that he is making
yesterday’s events the topic of his utterance: the theme is “‘the peg on
which the message is hung”’ (Halliday, 1970b: 161). It will be observed
that (6) also conforms to the general principle, according to which the
theme occurs in initial position, and that one of the differences between
corresponding active and passive sentences in English is that different
noun-phrases occur in initial position. Whether the correlation between -
thematic status and initial position is ever so high, even in English, that
an expression can be said to be thematic if and only if it occurs initially
is debatable. But it is certainly true that the processes that different
languages make available for the thematization* of one expression rather
than another frequently involve putting the expression earlier rather
than later in the utterance. This, as we have said, is not surprising, if the
theme is the expression whose function it is to serve as the point of
departure in the communication process — “‘the peg on which the
message is hung’’.
What we have been referring to as the theme, or thematic subject, is
sometimes called the psychological subject. One of the questions that
nineteenth-century linguists and psychologists were much concerned
with was whether, and to what degree, the order in which expressions
508 Semantics and grammar ITI
occur in the utterance reflects the order in which their psychological
correlates pass through the mind of the speaker in the cognitive process.
The psychological subject was by definition the expression referring to
the cognitive point of departure — the entity or topic that the speaker
had in mind when he formulated the intention to produce an utterance.
To many scholars it has seemed natural that the cognitive point of
departure and the communicative point of departure should coincide:
hence the identification of the theme with the so-called psychological
subject.
Granted the plausibility of the notion of the psychological subject, it
is reasonable to suppose that what the speaker takes as the cognitive
point of departure will depend upon its psychological salience for him
at the time — upon its being uppermost in his mind, as 1t were. And one
factor which will influence the psychological salience of particular
entities or situations is whether they already exist in the universe-of-
discourse* or not (cf. 15.3). Looked at from the addressee’s point of
view, what already exists in the universe-of-discourse will serve better
as the communicative point of departure than will something that is
unknown or unfamiliar. This accounts for the fact that the theme is
commonly defined as the expression which refers to what is given* and
the rheme as that part of the utterance which contains new* informa-
tion. As Halliday (1967b, 1970) points out, however, the speaker need
not, though he usually will, choose to announce as his topic something
that is given, or known, rather than something that is new, or unknown.
Very often there is nothing that is given, or known, which can serve as
the communicative point of departure; and, even when there is, the
speaker can, if he chooses, decide to make thematic an expression
referring to something other than what is given 4
According to Halliday, a distinction is to be drawn between thematic
structure and information-structure*; and it is information-structure
that is determined by whether something 1s given or new. Information-
structure, in English at least, is primarily a matter of stress and intona-
11 Bach (1971) points out that “‘in Japanese the theme... is marked by wa,
translating in many contexts as definite in corresponding English sentences”’
(cf. also Kuno, 1972b). (He also notes that ‘‘Interrogatives cannot occur with
wa’’ and draws attention to the fact that in Japanese, as in many languages,
the interrogative and indefinite pronouns are related: cf. 16.3.) Definiteness
of reference correlates highly with existence, or being given, in the universe-
of-discourse. In many languages in which there is no definite article, e.g.,
Russian, 2 nominal occupying initial position in the utterance is usually
thematic and it is often most appropriately translated into English by means
of a nominal containing the definite article.
. 12.7. Theme, rheme and focus 509
tion. Roughly speaking, expressions that convey new information are
stressed, and expressions conveying information that the speaker presents
as given, or recoverable from context, are unstressed. ‘he terms ‘focus’
and ‘presupposition’, which we introduced earlier in this section, relate,
therefore, from this point of view, to aspects of information-structure.
We will not go into the details. It suffices for us to note that in so far
as the correlation holds between information-focus and stress, it is
predictable that it should be as it is, not only in English, but in all lan-
guages in which variation in acoustic prominence serves a communica-
tive function. As we saw earlier, signal-information is inversely correla-
ted with semantic information; and semantic information can be
quantified, up to a point at least, in terms of the notion of novelty or
surprise-value (cf. 2.3).
So far we have taken the view that theme and rheme are complemen-
tary, in the sense that whatever is not thematic is rhematic and whatever
is not rhematic is thematic. Firbas (1964, 1972) takes a different view.
He operates with a notion of communicative dynamism (CD), defined
in terms of the degree to which an expression advances, or fails to ad-
vance, the process of communication. The theme is, by definition, the
expression with the lowest degree of CD; the rheme is the expression
that carries the highest degree of CD; and there may be several transi-
tional expressions that are neither thematic nor rhematic. This distribu-
tion of CD over the expressions that occur in linear sequence in an
utterance is accounted for in terms of the Prague School notion of
functional sentence perspective (FSP).
We will not go further into these matters. Whether thematic structure
is distinguished from information-structure or not, it is clear that there
is, in practice, a high degree of interdependence. Without prejudice to
the question whether they are, or should be, separable we will use the
term ‘thematic structure’ to subsume both. For the purpose of this
chapter, which is concerned with the semantic relevance of grammatical
structure, there are just two further points that need to be made.
The first is that, in terms of the distinction that is drawn here between
sentences and utterances, much of what counts as thematic structure
would not necessarily be accounted for in the analysis of sentences. We
are committed by the definition of ‘sentence’ and ‘grammar’ with
which we have been operating to the view that any two strings of forms
that differ with respect to type-token identity must be accounted for in
terms of their correspondence with two different system-sentences. It
follows that the type-token distinction of John I know and I know ohn
 510 Semantics and grammar III
is a sufficient condition for the postulation of the two system-sentences
‘John I know’ and ‘I know John’. The difference between John I know
and It’s John (that) I know would also be accounted for in terms of a
grammatical difference between two system-sentences. But when the
difference in the thematic structure of two utterance-tokens is simply a
| matter of their prosodic superstructure, as 1t were, the situation 1s not so
clear (cf. 3.1). It is arguable that the prosodic differences are a matter of
the contextualization of the system-sentence under one set of circum-
stances rather than another. But other linguists might, quite reasonably,
take a different view. No attempt will be made here to justify the
methodological decision to draw the boundary between what is and
what is not to be ascribed to the structure of system-sentences at one
place rather than another (cf. 14.6). What must be emphasized is that
languages vary considerably with respect to whether, and how, they
grammaticalize differences of thematic structure. These differences are
well known to translators (cf. Callow, 1974; Nida & Taber, 1969). They
are sometimes such as to cast doubt upon the possibility of translating
even the propositional content of an utterance, both accurately and
naturally, from one language into another.
The second point that must be made here has to do with the frequency
with which the logical and the thematic subjects coincide, in English and
other languages, in thematically neutral. or unmarked, utterances. This
is presumably to be accounted for in terms of psychological salience. It
may be assumed, and it has often been asserted, that among the infinity
of potential referents that may engage our attention some are intrin-
sically more salient than others, just as certain potential distinctions
upon which the classification of phenomena might be based are, by
virtue of our biological endowment, intrinsically more salient than
others are (cf. 8.3). What is known is of course, almost by definition,
more salient than what is unknown; and, other things being equal, the
more recently that something has been mentioned and put into the
universe-of-discourse, or the more familiar that something is to the
participants in a conversation, the greater will be its psychological
salience.
Independently of these context-dependent considerations, however,
we may be assumed, as human beings, to be more interested in persons
than we are in animals, to be more interested in animals than we are in
inanimate entities, and so on. It follows that in any one-clause utterance
in which reference is made both to a person and to an animal or in-
animate entity, the expression referring to the person will be made
 12.7. Theme, rheme and focus 511
thematic, unless there are special reasons for doing otherwise. For
example, other things being equal, the passive
(8) A man was stung by a bee in the High Street to-day
is Aa more normal utterance than the active
(9) A bee stung a man in the High Street to-day.
Generally speaking, though not in the case of the situation described by
(8) and (9), in any process which involves a person and an animal or
inanimate entity the expression referring to the person will be the logical
subject. The reason why this is so was made clear in our earlier dis-
cussion of the ontological basis of grammatical categories and the
importance of valency-schemata in many, if not all, languages (cf. 11.3,
12.4-12.6). The vast majority of transitive verbs are such that, when
they are used in the active voice, the grammatical subject is an expres-
sion which refers to the agent in a dynamic situation. Since agents are
usually persons, the tendency to make expressions that refer to persons
thematic will generally have the effect of making the thematic subject
coincide with both the grammatical and the logical subject. Further-
more, in so far as transitivity and causativity are associated with motion
from a source to a goal, there may well be grounds for believing, as many
scholars have done, that in referring first to the agent one is adopting as
the communicative point of departure what is also the more natural
cognitive point of departure.!* Many nineteenth-century linguists took
this view (cf. Sandmann, 1954); and it would seem to have at least some
foundation in the facts.

12 'To the extent that “‘the order of elements in language parallels that in physical
experience or the order of knowledge’”’ (Greenberg, 1963: 103; cf. Friedrich,
1975) a language is iconic*, rather than arbitrary (cf. 3.4). Gruber (1967,
1975) has argued that subject—predicate constructions develop, ontogenetically
(and in some, but not all, languages), out of topic-comment constructions;
and he has linked this with the development of constative* out of prior
performative* constructions (cf. 16.1). Much the same view is taken by Bates
(1976) ; and it is relatable to earlier speculations about the origins of grammar.
It is arguable that grammar, and more especially syntax, develops by virtue
of the “freezing’’ of what was originally iconic into what is subsequently an
arbitrary “‘formal requirement”’ (see p. 500, n. 8 above) and the progressive
decontextualization of utterances (cf. 14.6).
 13
The Lexicon

13.1. Lexical entries

Conventional dictionaries are essentially lists of what might be called
lexical entries*.1 Each of these entries is introduced by a head-word* in
its standard orthographic representation; and the lexical entries are
alphabetized in terms of their head-word. Alphabetization is of course
no more than a technique for listing the entries according to a con-
veniently applicable, but theoretically irrelevant, principle. The con-
ventional dictionary can, for our purposes, be thought of as an unordered
set of lexical entries, each of which is indexed by means of its head-word.
The fact that the head-word 1s represented orthographically (and may
or may not be furnished with a phonetic or phonological transcription)
is something that will not concern us here. We should not forget, how-
ever, as linguists, that most adult native speakers of English are accus-
tomed to thinking of word-forms as relatively stable written entities
whose pronunciation may be somewhat variable. Homonymy”*, to
which we will return presently, 1s traditionally based upon orthographic
type-token identity (cf. 13.4); and this is something that the lexico-
grapher cannot but be concerned with, since the organization of the
conventional dictionary depends upon it. Faced with the fact that the
noun ‘bank’ (whose written forms are bank, banks, bank’s and banks’)
has several different meanings, he must decide how many lexical entries
(all indexed by the head-word bank) he will put into his dictionary. No
such organizational problem presents itself in respect of the verbs ‘sow’
and ‘sew’. They must necessarily be given separate lexical entries and
indexed under distinct head-words without reference to any criterion
other than that of orthographic form. That their forms are, in fact,
homophonous is of secondary importance and will not be represented
directly, as a linguistically important fact about them, in the dictionary.
1 For some discussion of the practical and theoretical problems associated with
the compilation and editing of conventional dictionaries, cf. Dubois & Dubois
(1971), Householder & Saporta (1962), Rey (1970).
 13.1. Lexical entries 513
If we were compiling a dictionary of spoken English and indexing our
lexical entries by means of phonetically transcribed head-words, we
should have the same problems with ‘sow’ and ‘sew’ as we have with
the noun ‘bank’. In what follows, we will talk about the conventional
dictionary as if the entries were indexed by means of phonological
representations of their head-words. We will, however, continue to cite
these in their conventional written form.
What we have been calling the head-word of a lexical entry in con-
ventional dictionaries of English is typically both the citation-form of the
lexeme and also the stem-form, to which various suffixes may be added
in order to produce other inflexional forms of the same lexeme.” For
example, the head-word Jove is simultaneously the conventional citation-
form and also the stem-form of the verb ‘love’. The verb ‘love’ is
morphologically regular in this respect. It is assumed that anyone using
the dictionary will either know the morphological rules of English or ,
will have access to them in some standard grammatical description of the
language. It follows that, if a lexeme is morphologically regular, there
will be no need to include any morphological information as such in the
lexical entry. But there are also many morphologically irregular lexemes
in English: i.e. lexemes, some of whose forms at least cannot be obtained
by simply adding the appropriate regular suffixes to the citation-form
treated as a stem. For example, the verb ‘ride’ has ride, rides and riding
among its forms, but rode and ridden rather than vided; and ‘go’ has go,
goes and going, but went and gone rather than goed. 'The verb ‘ride’ is not
of course morphologically unique, as ‘go’ (or ‘be’) is. It belongs to a
particular inflexional subclass of the so-called strong verbs, other mem-
bers of this subclass being ‘drive’, ‘strive’, ‘write’, etc.; and there are a
number of other inflexional subclasses among the strong verbs.
We will come back to this point in a moment. But let us first note that
a conventional dictionary of English might handle morphological
irregularity of the kind we have just exemplified by putting into the
dictionary as separate head-words all the irregular forms and associating
with each of them a lexical entry whose content is purely morphological.
The entry for rode would tell us that it is the past-tense form of ‘ride’ ;
2 'The term ‘stern’ is used here for forms (either simple, like boy, or complex,
like monstrosity) to which inflexional, rather than derivatiorial, affixes are
added (cf. 13.2). It thus differs from ‘root’, which is restricted to simple
forms (i.e. to forms which are not further analysable), and from ‘base’, which
is more general than either ‘stem’ or ‘root’ and subsumes both. The citation-
form of a lexeme is the form that is employed in order to refer, metalinguis-
tically, to the lexeme (cf. 1.2).
514 The Lexicon
and the main lexical entry indexed under vide might well include the
same information viewed, as it were, from the other end. That this
information should be given twice is obviously convenient, since the
person using the dictionary might be consulting the dictionary with
either of two rather different questions in mind: (i) “Of what lexeme is
rode a form (and what morphosyntactic word does it realize)?’’; (11)
‘“What is the past-tense form of ‘ride’ (1.e. of the lexeme that will be
indexed, by convention, under the citation-form ride)?”
Now, one way of looking at the dictionary, or lexicon*, in relation to
the grammatical description of a language is to regard it as a kind of
appendix to the grammar — an appendix in which we find, appropriately
indexed, all the information that we need to know about particular
lexemes or their associated forms and cannot derive from anything else
that the grammatical or phonological analysis of the language tells us
about them.? For the moment we are concerned solely with morpho-
logical information. Let us return, then, to the verb ‘ride’ and its irregu-
lar forms rode and ridden. As was pointed out above, ‘ride’ is a member
of a particular subclass of strong verbs and, as such, although we describ-
ed it as irregular in relation to the most general inflexional rules of
English, it manifests a more restricted kind of morphological regularity.
Let us say that it belongs to inflexional subclass X; and that the other
members of this subclass are ‘drive’, ‘strive’, ‘write’, etc. Given the
information that ‘ride’ belongs to this subclass, we can obtain any of the
forms by means of rules which make reference to the phonological
structure of the citation-form now treated as a stem.
It might have been the case, though in fact it is not, that all verbs
whose present-tense stem-form (and conventional citation-form) is of
the same phonological structure as that of ride (however this is specified)
would be found to belong to inflexional subclass X. In which case, there
would be no need to put this information in the lexical entries for ‘ride’,
‘drive’, ‘write’, etc. It would be derivable by rules operating upon a
phonological representation of the stem-form. There are, however,
verbs like ‘hide’, which belong to a different subclass of strong verbs (cf.
hid, hidden) and, more important, verbs like ‘glide’, ‘dive’ (in most
dialects of English), and ‘site’ which conform to the very general rules
of suffixation. This being so (unless we postulate a rather abstract level
of phonological representation in terms of which glide differs from ride,
dive from strive, and so on), we can do either of two things. We can
3 For this way of looking at the lexicon, cf. Gleason (1962), who refers to
Bloomfield (1935: 274).
 13.1. Lexical entries 515
introduce a rule into the grammar of the language to the effect that all
lexemes of subclass X form their past tense and past participle in such-
and-such a way. Alternatively, we can list the verbs to which these rules
apply in the grammatical rules themselves. If we adopt the second
solution, there will be no need to repeat this morphological information
in the lexicon; and it is arguable that, for English at least, this second
solution is defensible. It is, in effect, what is done in current generative
grammars of the language.4 Whether it is an appropriate way of handling
the morphology of all languages 1s a question that we need not go into.
The point we have been making so far is that our conventional dic-
tionaries are based, explicitly or implicitly, upon a prior grammatical
description of the language, but, for reasons of practical convenience and
by virtue of the principle of alphabetic listing, they may include a num-
ber of lexical entries whose head-word 1s, in some sense, morphologically
irregular. If we think of the lexicon as an appendix to the grammar and
if we assume moreover that we are able to find the main lexical entry for
each lexeme, indexed by means of its citation-form (which may or may
not be a stem from which we can generate all the other forms), there is
no need for the lexicon to contain these purely morphological lexical
entries. We will henceforth disregard them as being theoretically redun-
dant. Furthermore, we will assume that the lexicon contains, for each
lexeme, all the morphological information that is required, but no more
than is required, when the lexicon 1s used in association with a particular
generative grammar of the language. Minimally, and commonly for
English and many other languages, it will be sufficient to give a single
stem-form without any further information about the inflexional class
to which the lexeme belongs. But in the description of some languages
it may well be necessary (as an alternative to a wholesale listing of
lexemes or stem-forms in the grammar) to associate with each lexeme, in
its lexical entry, an indication of its inflexional class. This is what is done
in conventional dictionaries of French, Latin or Greek, which classify
lexemes according to their so-called declension or conjugation or,
equivalently, associate with each lexeme a subset of its forms (its so-
called principal parts), sufficient to determine, for the user of the
dictionary who knows the inflexional rules of the language, all the other
forms of the lexeme.
All that we have said so far, in our discussion of conventional dic-
tionaries, might suggest that there is some intrinsic connexion between
4 For some discussion of the role of the lexicon in Chomskyan generative
grammar, cf. Botha (1968), Hudson (1976).
516 The Lexicon
what are two logically separable functions of the head-word: (1) its use
as a metalinguistic referring expression (cf. 1.3); and (ii) its function as
the bearer of morphological information. It should be obvious, how-
ever, that we could, in principle, use any name whatsoever in order to
refer to particular lexemes provided that the name we employ is known
to be the name of the lexeme in question. Suppose, for example, that we
listed all the lexical entries in some arbitrary order and numbered each
entry according to its place on the list. We could then say, for example,
that lexeme 673 1s a morphologically regular noun, whose stem is boy
(and this might be represented either orthographically or phonologically,
according to whether we are concerned with the written or the spoken
language), and that the lexeme means such-and-such. The important
point here is that our numerical index is neutral with respect to the
morphological, syntactic and semantic information that is included in
the lexical entry. The lexeme itself is defined in terms of the information
in the entry; and our numerical index merely serves as the address of the
location in which this information is stored. The terms ‘address’,
‘location’ and ‘storage’ are borrowed from computer science; and the
reader may find it helpful to think of the lexicon in this way. The
address need not, however, play any part in relating the lexical entries to
the rules of the grammar or to any other component of the linguist’s
model of the language-system. As we said earlier, the lexicon is to be
regarded as an intrinsically unordered set of lexical entries. The
implications of this point will be made clear presently.
The information that is found in a typical lexical entry in a conven-
tional dictionary is of three kinds: morphological, syntactic and seman-
tic. Many dictionaries will also include, as an addendum to the lexical
entry proper, some more or less detailed account of the etymology of the
lexeme, in so far as this is known or can be reconstructed. Since the
etymology of a lexeme is, in principle, irrelevant to its pronunciation
(more precisely, to the pronunciation of its forms), to its distribution
throughout the sentences of the language and to its current meaning,
etymological information as such will not be included in the linguist’s
synchronic model of the language-system. But we shall certainly need to
include all the other kinds of information: morphological, syntactic and
semantic. Just how this information is encoded in the lexicon will
depend upon the formalization of phonology, grammar and semantics
that has been adopted for the description of the language-system in
question.
We will operate here with a relatively informal specification of the
 13.1. Lexical entries 517
morphological, syntactic and semantic characteristics of lexemes; and,
in doing so, we will make use of traditional terminology. Readers who
are conversant with more recent work in theoretical linguistics should
have no difficulty in converting the statements made below into what-
ever terminological or notational framework they prefer to work with
and in making the necessary adjustments. Our discussion is intended to
be neutral with respect to several possible systems of formalization. We
have started by considering conventional dictionaries because they can
be assumed to be familiar to everyone.
It will be helpful at this point to introduce a schematic representation
of a lexical entry: this is given in figure 11. The first point to be noted

Figure 11. Schematic representation of a lexical entry

in connexion with this diagram is that, although we have been talking

of three kinds of information that must be associated with a lexeme
(morphological, syntactic and semantic), we have provided four
separate boxes in our schematic lexical entry. We have allowed for the
possibility that the inflexional class of a lexeme should not be deducible,
by rule, from either its syntactic properties or the phonological structure
of its stem. This possibility is not, in fact, actualized in all languages. In
languages of the so-called isolating* type there are of course no inflex-
ional classes at all: in any lexicon associated with a generative grammar
of such languages, box (ii) would be empty. As far as non-isolating
languages are concerned, the situation is more complicated. There are
languages for which the notion of inflexional class is applicable only
vacuously; in the sense that all the nouns belong to the same inflexional
class, all the transitive verbs belong to the same inflexional class, and so
on. Once again, box (ii) would be empty, since all the morphologically
 518 The Lexicon
relevant information is derivable from box (iii). There are also languages
in which the inflexional class of a lexeme is determined jointly by its
syntactic properties and by the phonological structure of the stem; and
this is, to some considerable degree, the case in the richly inflecting
Indo-European languages. Even the two hundred or so irregular verbs of
English fall into inflexional classes that are partly; though not wholly,
determined by the phonological structure of their stem. Just how many
inflexional classes are to be distinguished in the lexicon will depend upon
the morphological and phonological rules that are established for
deriving all the so-called irregular forms in some particular grammatical
description of English. Since we are not concerned with inflexional
morphology as such, we will not go further into this question. It suffices
for our present purpose that we have given theoretical recognition to the
fact that, in certain languages at least, the inflexional class of a lexeme
may not be derivable from the phonological form of its stem, from its
syntactic properties or from its semantic specification.
The relationship between boxes (i) and (iv) has been discussed earlier.
As we have seen, it 1s generally accepted that, due allowance being made
for certain sporadic and unpredictable instances of onomatopoeia and
sound-symbolism, the relationship between the form (or forms) of a
lexeme and its meaning, in all languages, is arbitrary and conventional
(cf. 3.4). We will say no more about this.
The relationship between boxes (i) and (iii), on the other hand, is
similar to the relationship between (i) and (ii). That is to say, there are
languages in which noun-stems are characteristically distinct from verb-
stems, or verb-stems from adjective-stems, etc., in terms of their
phonological structure; and there are other languages in which no such
correlation exists. The latter class of languages can be further sub-
divided into (i) those in which many, if not all, of the morphologically
simple stems, or roots, are associated with several different sets of
syntactic properties and (ii) those in which none, or very few, of the
roots are associated with more than one set of syntactic properties.
English belongs to the former of these two subclasses. Not only is it the
case that one cannot predict from the phonological form of a morpho-
logically simple stem whether it is a noun-stem, verb-stem, adjective-
stem, etc., but it is also the case that very many of these morphologically
simple stems function as both noun-stems and verb-stems (bank, man,
jump, move, etc.), as both verb- and adjective-stems (warm, empty, open,
shut, etc.), and so on. There are many stems, too, that are the stems of
both transitive and intransitive verbs (move, open, etc.), of both countable
 13.1. Lexical entries 519
and mass nouns (fish, paper, sound, etc.), and so on. The fact that this is
so creates a problem for the lexicographer. Should he put a single lexical
entry in the dictionary for, let us say, ‘jump’ and associate with this, in
box (iii), the syntactic information that this lexeme can be used either as
a noun or a verb? Or should he treat ‘jump,’ (a noun) as being distinct
from ‘jump,’ (a verb), indicating in the lexical entry for ‘jump,’
that it can be used both transitively and intransitively? ‘There is no
generally accepted solution to this problem.
Let us now turn briefly to the relationship between boxes (iii) and
(iv). The question that confronts us here is whether there is, in any or
in all languages, a correlation between the meaning of a lexeme and its
syntactic properties. As we have seen, traditional definitions of the
parts-of-speech in terms of their denotation of persons, places or things
(nouns), of actions, processes and states (verbs), or of qualities (adjec-
tives), etc., are based on the assumption that each of the major syntactic
categories — nouns, verbs, adjectives, etc.—has a characteristically
different kind of meaning associated with it (cf. 11.3). So too are tradi-
tional definitions of such secondary syntactic categories as gender,
number and tense. For example, in all the Indo-European languages
whose nouns must be classified in terms of grammatical gender it is
generally the case that, if a noun denotes a class of male persons, it will
be masculine and that it will be feminine if it denotes a class of female
persons.5 There is therefore a correlation, as far as nouns denoting
human beings are concerned, between the grammatical gender of a
lexeme and the sex of its denotata. But there are certain nouns whose
gender is exceptional in terms of this general correlation. They fall into
two subclasses, which may be exemplified by ‘Madchen’ (“‘girl’’) and
‘Weib’ (‘‘woman’’), respectively, in German. Each of these nouns is
neuter. The difference between them is that ‘Madchen’ has as its stem
a morphologically complex form ending with the so-called diminutive
sufhx -chen, whereas the stem of ‘Weib’ is morphologically simple. Now
all nouns whose stems are formed by the sufhixation of -chen (or -lein) in
German are neuter regardless of their meaning. Provided that the
morphological composition of the stem is specified in the lexical entry
and the rules of the grammar make reference to this information, there
is clearly no need to indicate the gender of ‘Madchen’ in the lexicon.
5 On the grammatical category of gender and its semantic basis, cf. Wienold
(1967). The point that is made here with reference to gender might be made
equally well, for many languages, with respect to the selection of sortal
classifiers (cf. 11.4).
520 The Lexicon
Nor is there any need to indicate the gender of any of the masculine
nouns denoting males or feminine nouns denoting females, provided
that the grammatical rules can make reference to the semantic specifica-
tion in each lexical entry. It is only in the case of nouns like ‘Weib’
(whose gender cannot be derived by rule from the morphological or
semantic information given in the lexical entry) that it is necessary to
include an explicit indication of their gender in the lexicon.
There are considerable differences between languages in the degree
of independence or interdependence that holds between the morpho-
logical, syntactic and semantic properties of lexemes; and full recogni-
tion should be given to these differences in the description of particular
languages. In so far as there is a correlation between the meaning of a
lexeme and its syntactic properties, on the one hand, and between its
syntactic properties and the phonological form of its stem, on the other,
this correlation will of course reduce the arbitrariness that generally
holds in language between form and meaning. Consider, for example, a
hypothetical language in which all the nouns denote discrete physical
objects (including persons) and all the noun-stems (and no other stems)
are of the form CVC (consonant-vowel-—consonant: kep, tok, gup, etc.).
Given that this is so, there would be no need to include in the lexical
entries for ‘kep’, ‘tok’, “gup’, etc., the information, in box (iii), that
they are nouns. Furthermore, a rule of lexical insertion could operate
indifferently upon either the phonological form of the stem or the seman-
tic specification of the lexeme. Since there is an intrinsic connexion, in
all languages, between the meaning of a lexeme and its distribution
throughout the set of well-formed sentences, whereas there is no such
intrinsic connexion between phonological form and syntactic distribu-
tion, the linguist would no doubt integrate the grammatical rules with
the lexicon in such a way that they derived the syntactic classification of
a lexeme from its meaning rather than from the phonological form of its
stem. It would nonetheless be an important fact about our hypothetical
language that there is some correlation (apart from the usual kinds of
onomatopoeia and sound-symbolism) between form and meaning. The
only thing that is unrealistic about our hypothetical language is that we
have assumed a perfect match between phonological, syntactic and
semantic structure. There is probably no actual language in which there
is a perfect match of this kind. But there do exist languages in which
there is some degree of correspondence.
 The Lexicon 521
13.2. Complex lexemes
So far, in our discussion of the lexicon, we have restricted our attention
to what we will call simple* lexemes: 1.e. lexemes whose stems are
morphologically unanalysable. There are two other classes of lexemes
that we must now consider: complex* and compound* lexemes. The
term ‘complex lexeme’ is here introduced to cover what is commonly
referred to by linguists as derivation* : the formation of a morphologic-
ally more complex stem, Y, from a morphologically simpler stem, X, by
attaching to X a particular derivational afhx or by systematicaliy modi-
fying the form of X in some way. For example, the sufhx -ly may be
attached to certain noun-stems in English, e.g., man, friend, in order to
form the stems of the corresponding derived, or complex, adjectives:
manly, friendly. The form -ly is a derivational affix. More specifically, it
is a denominal adjectivalizing sufhx: i.e. it forms adjectives from nouns.
The prefix un- is attached to adjective-stems (e.g., sure, friendly) to form
morphologically more complex adjective-stems (unsure, unfriendly): it
is a de-adjectival adjectivalizing affix. Suffixation and prefixation are the
most common, but by no means the only, derivational processes to be
utilized throughout the languages of the world. ‘They may be sym-
bolized by means of the formulae
(1) X+a->Y
(2) 64+ XxX >Y
respectively. In these formulae a and b stand for particular afhxes, and
X and Y for classes of lexical stems.® Thus (1) says that the afhx, or
formative, a may be suffixed to a stem of class X to form a stem of
class Y; and (2), that b may be prefixed to a stem of class X to form a
stem of class Y.
The distinction between derivation and inflexion has long been
regarded as controversial. In so far as there is a clear-cut distinction it 1s
this: inflexion produces from the stem (or stems) of a given lexeme all

6 A distinction is often drawn between class-changing and class-maintaining

derivation. In terms of this distinction the sufhixation of -/y 1s class-changing
(N -++ ly — Adj), whereas the prefixation of un- is class-maintaining (un + Adj
— Adj). It is worth noting that what are normally referred to as class-main-
taining derivational processes rarely, if ever, result in true endocentricity*
(cf. 10.3). The stem of the noun ‘manhood’ is derived from the stem of the
noun ‘man’ by the suffixation of -hood (i.e. N + hood — N), but ‘man’ and
‘manhood’ belong to syntactically distinguishable subclasses of nouns.
Strictly speaking, even ‘unsure’, ‘unfriendly’, etc., are not endocentric, since
un + Adj > Adj is not recursive.
 522 The Lexicon
the word-forms of that lexeme which occur in syntactically determined
environments; derivation, on the other hand, results in the formation of
what is traditionally considered to be a different lexeme. For example,
the suffixation of -s to the form friend creates the plural or possessive
form of the lexeme ‘friend’, whereas the suffixation of -ly to the same
stem creates a new stem friendly (to which the inflexional suffixes -er and
-est may be added: friendlier, friendliest), and this is the stem of a
different lexeme. It is for this reason that derivation is traditionally
referred to as a kind of word-formation* (i.e. lexeme-formation).
One point should be emphasized in this connexion. In what follows,
we will frequently say that one lexeme is derived from another: e.g.,
that ‘friendly’ is derived from ‘friend’. It should not be forgotten,
however, that lexemes are abstract entities: what is derived by means of
prefixation, suffixation, etc., is the stem-form of a lexeme, and it is
: derived from another, morphologically simpler, stem-form. The
lexemes ‘friendly’ and ‘friend’ are formally (i.e. morphologically) rela-
ted by virtue of the derivational relationship,

3) X+h->Y,
which holds between their stems:

(4) friend + ly — friendly.

As a lexeme, ‘friendly’ is neither syntactically nor semantically more
complex than such simple (i.e. non-derived) lexemes as ‘good’ or ‘nice’.
Provided that this point is borne in mind, we can continue to talk, as
linguists commonly do, of deriving one lexeme from another. Further-
more, we can interpret X and Y in (1) and (2) above as variables which
range over classes of lexemes rather than classes of forms. For example,
the following formula

(5) Nx + ly > Ay
may be read as saying “‘lexemes of Class Ay are derived from lexemes of
class Nx by the suffixation of -/y (to the appropriate stem)”’, where Nx
and Ay are arbitrarily labelled subclasses of nouns and adjectives.
There are many morphologically simple forms in English which
function as stems for both verbs and nouns (doubt, answer, skin, knife,
etc.) or both adjectives and verbs (dirty, clean, dry, etc.). These can be
brought within the scope of the notion of derivation by recognizing
conversion*, or zero-derivation* (1.e. derivation by means of the afhixa-
 13.2. Complex lexemes 523
tion of an identity-element) as a morphological process.’ For example,
the nouns ‘release’ and ‘attempt’ might be said to be derived from the

(6) V, + 9->N, |
verbs ‘release’ and ‘attempt’ in accordance with the formula

(where @ stands for the identity-element). The reason why these nouns
are said to be derived from the corresponding verbs, and by means of
sufhixation, is that they belong to the same subclass of nouns as ‘exten-
sion’, ‘justification’, ‘arrangement’, etc., which are clearly deverbal and
derived by suffixation: deverbal nominalization is characteristically a
matter of suffixation in English. The formula given above as (6) can
therefore be seen as a particular instance of
(7) V+ VNq > Na,
where VNg is a class of deverbal nominalizing suffixes (of which the
identity-element, or zero, is one and -zon, -al, -ment, etc., are others).§
Conversion, or zero-derivation, is very productive in English; and it 1s
usually, though not always, clear which of the pair of lexemes related by
conversion is simple and which is complex in terms of the general
patterns of derivation manifest in the language.
So much by way of general background to the notion of derivation.
The formulae are purely ad hoc; but they will serve the present pur-
pose, and we do not want to go into more detail than is strictly necessary.
We are concerned with the theoretical status of complex (i.e. derived)
lexemes. Should they be listed in the lexicon and, if so, what informa-
tion should be associated with them? Two extreme views might be
maintained on this issue: (i) that no complex lexemes should be included
in the lexicon; (11) that every complex lexeme should be listed separately
in the lexicon and provided with its own lexical entry.
The arguments in favour of listing complex lexemes individually in
the lexicon, rather than deriving them by rule in the grammar, are well

* “Conversion’ is the term used by Quirk et al. (1972: 1009ff). Arguably, the
term ‘conversion’ — “‘the derivational process whereby an item is adapted to
a new word-class without the addition of an affix”’ — carries different implica-
tions from the term ‘zero-affixation’, which can be understood as implying
the addition to the stem of the identity-element functioning as an affix. For
discussion of the criteria that might be applied in deciding between these
alternatives in particular instances, cf, Haas (1957).
8 The reader is reminded of the ambiguity of the term ‘nominalization’:
(i) ‘‘creation of nouns’’; (ii) ‘“‘transformational process whereby nominals
(NPs) are constructed” (cf. 10.3, 11.1).
 524 The Lexicon
known (cf. Matthews, 1974): derivational rules are characteristically less
productive than inflexional rules; and their syntactic and semantic effect
is, in many instances, unpredictable. To take the question of restricted
productivity first: what is of crucial importance here is the fact that
certain derived lexemes, which one might expect to exist and to be in
_ current use, not only have never been attested, but are rejected by
native speakers, even though they are morphologically regular and would
satisfy the appropriate formula. For example, there is no derived noun
in English that is related, syntactically and semantically, to the verb
‘salute’ as ‘dilution’ is related to ‘dilute’, ‘pollution’ to ‘pollute’, etc.
(cf. Matthews, 1974: 50). Not only is there no noun whose stem is
salution: there is no noun at all that fills this lexical gap. Conversely,
there are many lexemes in English which are morphologically, syntac-
tically and semantically similar to various kinds of complex lexemes, but
which cannot be derived synchronically from existing lexemes. For
example, ‘doctor’ and ‘author’ are reasonably classified as agentive
nouns (like ‘actor’, ‘painter’, etc.) and their stems are such that they
might be held to contain the agentive suffix -er/-or. But there is no verb
whose stem is doct- or auth-. Examples of derivational gaps of this kind
could be multiplied almost indefinitely.®
Let us now turn to the semantic problems involved in the generation
of complex lexemes within the grammar. It has often been pointed out
that the meaning of very many complex lexemes is more specialized than
that of the lexemes from which they appear to be derived. The reason
for this would seem to be that complex lexemes are like simple lexemes,
in that, once they are created or introduced into the language-system
and pass into general currency, they may be institutionalized and, by
virtue of their use in particular contexts, develop more or less specialized
senses. For example, the noun ‘recital’ is morphologically parallel to
such other deverbal nouns of action as ‘refusal’, ‘approval’, ‘acquittal’,
etc. The form of its stem and its function as a noun of a certain kind can
be accounted for by means of the formula
(8) Vr + al > Na,
where Vr is the class of verbs whose stems may take this particular
nominalizing suffix, Ng is the class of action nouns, and (8), like (6), is
but one of a set of formulae, all of which may be subsumed under (7)
® The existence of these two derivational gaps is readily explained in terms of
the historical development of English. The lexemes ‘doctor’ and ‘author’
come, ultimately, from Latin, as do many other such nouns,
 13.2. Complex lexemes 525
above. Now there are contexts in which the syntactic and semantic
relationship between ‘recital’ and ‘recite’ is parallel to that which holds
between ‘refusal’ and ‘refuse’, ‘selection’ and ‘select’, ‘arrangement’
and ‘arrange’, etc.; and it is easy to see that, if the syntactic rules of
English were to generate a set of expressions of the form
(9) the N, of NP, (by NP,)
from structures underlying transitive sentences of the form
(10) NP, V NP,
and if the lexical entries for ‘recite’, ‘refuse’, ‘release’, ‘arrange’,
‘select’, etc., contained enough morphological information for us to
know that ‘refuse’ and ‘recite’ were members of V;r, ‘release’ was a
member of Vp, and so on, then it would be possible to generate ‘the
recital of the Lord’s Prayer (by the congregation)’, ‘the release of the
prisoners (by the terrorists)’, etc., within the grammar. So far so good.
But there are obvious problems which arise if we eliminate the noun
‘recital’ from the lexicon and generate it by means of a transformational
rule in the grammar. If Lawrence Olivier 1s billed to give a Shakespeare
recital, he will indeed recite Shakespeare, but if Yehudi Menuhin gives
a Mozart recital he will play, rather than recite, the music of Mozart.
Furthermore, it is only certain kinds of music that are played at recitals:
we would not expect to hear the Jupiter Symphony played at something
that was advertised as a Mozart recital. What constitutes and is referred
to in English as a recital is determined by accepted cultural conventions;
and one cannot be said to know the meaning of ‘recital’ unless one has
some knowledge of these conventions.
There are syntactic problems too. The expressions ‘the Shakespeare
recital’ or ‘the Olivier recital’ do not of course conform to the pattern
set forth in (9). The expression that would be accounted for by (8) and
(9) is ‘the recital of Shakespeare (by Olivier)’; and this is of doubtful
acceptability. Let us suppose then, it might be suggested, that there are
two lexemes, one of which is derived by nominalization (of restricted
productivity) and appears in such expressions as ‘the recital of the
Lord’s Prayer’ and the other of which is simple, like the noun ‘concert’.
Which of these two lexemes is found in ‘the poetry recital’? This ex-
pression is surely not ambiguous; and yet it can be related equally well
to ‘the recital of the Lord’s Prayer’ (cf. ‘the recital of poetry’), on the
one hand, and to ‘the sonata recital’, ‘the jazz concert’, etc., on the
other. It seems perverse to say that there are two distinct lexemes,
 526 The Lexicon
‘recital,’ and ‘recital,’ ; and it is not even clear that there are two sharply
distinguishable senses involved. There is the further difficulty that, in
addition to ‘recital’, we also have ‘recitation’, which satisfies (7). But
‘recital’ and ‘recitation’ are not intersubstitutable in all contexts.
‘Recitation’ is perhaps more readily generated from sentential struc-
tures containing the verb ‘recite’ than is ‘recital’; but ‘recitation’ also
has its own specialized and institutionalized senses, which must be
accounted for in the lexicon.
Enough has been said to give some indication of the difficulties which
arise for the proposal that all derived lexemes should be generated by a
combination of syntactic and morphological rules. What then of the
alternative proposal, that every complex lexeme should be listed
separately in the lexicon? First of all, it should be noted that, in one way
or another, we must relate, syntactically and semantically, those simple
and complex lexemes which do enter into paired sets of expressions and
sentences like ‘John’s refusal of the job’: “John refused the job’, ‘their
solution of the problem’:‘They solved the problem’, etc. Chomsky
(1970) has suggested one way of doing this. In effect (and we need not
go into the details of the formalism upon which it depends), it provides
for the generation, by the base rules of the grammar, of both the
nominal and the sentential structures, NP,’s X of NP, and NP X NP,
where X is realized by a noun (with its appropriate complements) in the
one structure and a verb in the other. Certain lexemes would then be
listed in the lexicon as having alternative stem-forms refusal and refuse,
solution and solve, destruction and destroy, etc., according to whether
they occur in the nominal or the sentential structure. What is being
proposed, then, is to simplify the transformational rules of the grammar
by extending the rules of the base and by handling derivation within the
lexicon. Chomsky’s proposal is made, of course, within the framework
of his own theory of generative grammar. The validity of his criticisms
of what he calls the transformationalist, as opposed to the lexicalist,
account of derived lexemes is independent, however, of this fact; and
the same points as he makes have been made by scholars of a quite
different theoretical persuasion. What is currently referred to as the
controversy between transformationalists and lexicalists is a particular
version of the more general controversy, of longer standing, between
those who wish to account for the distribution and meaning of complex
lexemes by means of productive syntactic and morphological rules and
those who favour the listing of all such lexemes in the lexicon. Let us
grant that Chomsky’s proposal or some alternative formulation of the
 13.2. Complex lexemes 527
lexicalist hypothesis is technically feasible. The question we are con-
cerned with here is whether it is desirable to include all the derived
lexemes of a language in the lexicon.
The answer to this question depends upon the answer to a prior
question: is it possible, or in practice feasible, to list all the derived
lexemes of a language? As we have seen, it is difficult, if not impossible,
to arrive at a satisfactory pre-theoretical test of grammatical accept-
ability (cf. 10.2). Our model of the language-system will inevitably
generate, as well-formed sentences, many strings of word-forms which
would be regarded as unacceptable by some native speakers: e.g.
‘Football was played by him yesterday’. So even the most obviously
productive syntactic rules of English, such as the rule (whatever its
precise formulation) which relates active and passive sentences, give
rise to problems of this kind: for the sentence ‘He played football
yesterday’ is unquestionably acceptable. The relevance of this point to
the status of complex lexemes ts that there appears to be no difference of
kind, pre-theoretically, between the productivity of what are universally
regarded as syntactic processes and the productivity of at least some
derivational processes.
The native speaker is as free to construct de-adjectival abstract nouns
with stems ending in -ness, for example, and to use them in certain
syntactically specifiable positions as he is to form passive sentences from
underlying active structures. It is not even clear that the existence of a
generally accepted alternative stem-form (whether this is also derivable
by means of a more or less productive rule or not) inhibits the operation
of the more general rule for deriving abstract nouns from adjectives.
However that may be, we cannot rely upon the existence of a particular
derived lexeme in some corpus of actually attested utterances of English
as either a necessary or a sufficient condition of the existence of that
lexeme in the language-system. Nor can we argue very convincingly
that the native-speaker’s formation of a derived lexeme, on those occa-
sions on which he does form one himself by applying the productive
derivational principles inherent in the language-system (rather than
looking it up, as 1t were, in his own internalized lexicon), results from
the exercise of some peculiarly creative ability. At least some part of
what is customarily held to fall within the scope of derivation appears to
be rule-governed in the same way that the construction of grammatically
acceptable utterances is.
We are faced, then, with a dilemma. Neither of the two extreme pro-
posals referred to earlier seems to be theoretically justifiable. This being
528 The Lexicon
so, it is reasonable to consider listing in the lexicon only those derived
lexemes that are morphologically, syntactically or semantically idio-
syncratic in some way and excluding from the lexicon any lexeme whose
stem-form and whose distribution and meaning can be accounted for by
means of productive rules. There will be problems, of course, in the
application of this criterion. One problem, as we have seen, is that it is
not always possible to assign determinate limits to the productivity of
certain derivational processes; and yet, if we do not restrict the condi-
tions under which the derivational rules apply, we will certainly generate
a host of unacceptable lexemes. This problem is inherent in the whole
process of constructing a generative model of the language-system. It
may well be that further research will lead to refinements in the specifi-
cation of the conditions under which the derivational rules operate, so
that ultimately the linguist will be able to claim, with greater justification
than he can at present, that his model generates all and only the pre-
theoretically acceptable complex lexemes of the language that he is
describing. It is only to be expected, however, that, just as there are very
many strings of word-forms of indeterminate acceptability, so there will
be a number of morphologically complex stems whose pre-theoretical
status with respect to acceptability is equally indeterminate. In such
cases, we can decide, as a matter of methodological principle, to let the
model itself resolve the question for us.
Consider, for example, all the adjectives in English whose stems end
in the suffix -able or -ible. Many of these, though by no means all, can be
accounted for in terms of a synchronically productive process of de-
verbal adjectivalization. In so far as this process is productive in present-
day English, it is restricted to transitive verbs. Let us begin, therefore,
by setting up the formula
(11) Vir + able > Az
where Vir is the class of transitive verbs and A; 1s an arbitrarily labelled
subclass of adjectives. This formula will account for the morphological
relationship between ‘read’ and ‘readable’, ‘drink’ and ‘drinkable’,
etc. Now it is well known that the unrestricted application of a formula
like (11) will yield a certain number of lexemes that would be rejected by
perhaps the majority of English speakers: ‘gettable’, ‘fetchable’, etc.
The first question that would confront us in relation to these putative
lexemes is that of deciding whether they are definitely unacceptable or
not in the dialect of English that we are describing. Let us assume, for
the sake of the argument, that some of them are definitely unacceptable
 13.2. Complex lexemes 529
and that others are pre-theoretically indeterminate with respect to
acceptability. The next question is whether there is any phonological,
morphological, syntactic or semantic property in terms of which we can
predict the applicability of (11) to particular transitive verbs in the
lexicon. A priori, any one or any combination of several factors might be
relevant: whether a particular verb-stem has a certain phonological
structure or not, whether it is itself morphologically simple or complex,
whether it is recognizably of Latin or Germanic origin, whether the
verb belongs to a syntactically or semantically restricted subclass of
transitive verbs, and so on. There is, in fact, no obvious single property
or combination of properties of this kind in terms of which we can
predict the applicability or non-applicability of the derivational formula
Vir + able -» Az.!° At the same time, it is clear that the process that the
formula is intended to account for is extremely productive; and there
are certain morphologically specifiable subclasses of transitive verbs
(e.g., those whose stems end in -zge or -if/y) to which the formula seems
to apply without restriction, in the sense that none of the resultant
lexemes is definitely unacceptable. This being so, it would be un-
reasonable to take the view that all the adjectives whose stems end in
-able should be listed in the lexicon.
The simplest solution to the problem, though not necessarily the
most satisfactory, is to let the formula Vir + able + Az apply without
restriction in our model of the language-system. Alternatively, we might
decide to mark a certain number of transitive verbs (e.g., ‘get’, ‘fetch’)
as exceptions to this derivational process, allowing it to operate in all
other instances. If the rule which derives the members of Az is left un-
restricted, it will of course cover transitive verbs whose stems end in
-1ze or ~-ify, provided that they are listed in the lexicon as transitive
verbs or that they are themselves derived by rule and their syntactic
function is assigned to them as part of the process of derivation. But,
even if we were to shrink from admitting into the grammar of English,
with or without specific exceptions, the very general rule Vir -+ able —> Az,
we could still include a rule which made reference to the morphological
composition of the verb-stems ending 1n -zze and -zfy and the syntactic
10 Hasan (1971: 152) suggests: ‘‘this suffix -able can be used with that set of
verbs which can realize the process ‘reaction’ in an active transitive clause
where two participants are required but where the role ‘affected’ can be
mapped only onto the subject ... This explains why one may say Fim is a
likeable fellow but not Jim is a puzzleable fellow’’. Some such principle may
be operative; but it is not always clear what one may or may not say, on the
one hand, and what is here covered by the term ‘affected’, on the other.
530 The Lexicon
information associated with the lexemes of which they are to be stems.
The point is that it is, in principle, feasible to handle by rule as much of
the derivational regularity in a language as is empirically justifiable.
Any derived lexeme that is syntactically or semantically irregular, in
that its distribution or meaning is unpredictable by general rule from
the lexeme whose stem is the synchronic source of its morphological
derivation, must of course be listed in the lexicon. But it will not neces-
sarily be provided with a full lexical entry. There are different kinds and
degrees of derivational irregularity; and it is frequently the case that part,
though not all, of the syntactic function and meaning of an irregular
derived lexeme is predictable by rule. To return to the adjectives whose
stems end in -able. Many of these adjectives, and perhaps all of those
that we would wish to regard as being completely regular, can be inter-
preted in terms of a modalized* passive predicative phrase, the modality
in question being that of possibility or ability (cf. 17.1). For example,
‘His anger is justifiable’ means “‘His anger can be justified’’, or more
precisely ‘‘His anger is such that it can be justified’’; “His assets are
unrealizable’ means “ His assets (are such that they) cannot be realized”’;
and so on. Let us assume, then, that all the adjectives whose stems end
in -able and whose meaning and distribution is regular in terms of a
transformational rule, which derives them from an underlying modalized
sentential structure of the appropriate form, containing the transitive
verb whose stem is the form to which -able is suffixed, are removed from
the lexicon. Granted that this 1s the norm, we can distinguish various
kinds of derivationally irregular adjectives.
One adjective that is morphologically and syntactically regular in
terms of the formula Vir + able —- Az and only partly irregular from a
semantic point of view is ‘readable’. Anything that is readable is such
that it can be read. The adjective ‘readable’, however, is commonly
used, and perhaps most commonly used, to imply rather more than
‘capable of being read’’: a readable novel, for example, is normally
understood to be a novel that one can read with pleasure or interest.
This sense of ‘readable’, we may assume, results from its institutionali-
zation and may not be derivable by rule. ‘Readable’ must therefore be
provided with its own lexical entry and this specific sense of the lexeme
accounted for in box (iv): cf. figure 11. But no special morphological or
syntactic information need be given in the lexical entry. What we want
to say about ‘readable’ is that it is derived from ‘read’ (1.e. that its stem
is derived from the stem of ‘read’) by the suffixation of able. At first
sight, the most obvious way of doing this is to put into box (i) or (11)
 13.2. Complex lexemes 531
something like read + able (where the plus-sign indicates the process of
suffixation). But there is nothing here to tell us that read is the stem of
‘read’. Suppose, therefore, we were to leave box (i) empty and to put
into box (ii) the address of ‘read’ (rather than its stem), together with an
indication of the process of suffixation: 1.e. 798 + able. Associated with
this method of representing the morphological composition of the stem
one would have, appropriately formalized, the convention that, in
default of any information to the contrary, the lexeme is morpho-
logically and syntactically regular. Looking up entry 798, we will find
the transitive verb ‘read’ with the stem read and, since there is no
specific information in box (1) of the entry for ‘readable’ indicating that
its stem is in any way phonologically irregular, we will form the stem
readable; and, in default of any further information to the contrary in
boxes (ii) and (iii), we will infer that it does indeed belong to the syn-
tactic class Az, that it forms abstract nouns with a stem ending in -zty
(cf. readability), and so on. Furthermore, the regular, though perhaps
less common, sense of ‘readable’ need not be assigned to it at all in the
lexicon, since it is derivable by the more general rule.
There are very many other adjectives with stems in -able for which a
similar treatment would seem to be required, if we are to capture, for-
mally, both their grammatical regularity and their semantic idio-
syncrasies. What has been outlined in the previous paragraph may not
be the most appropriate way of doing this. The point that we wish to
emphasize here is that by including in the entry for ‘readable’ a cross-
reference to the lexeme ‘read’, rather than simply to the form read, we
can in principle make use of this in giving the semantic information
associated with ‘readable’ in box (iv). It may well be that there are
other morphologically and syntactically regular adjectives of class Az
whose meaning is related to the verbs from which they are derived as
the meaning of ‘readable’ (in its more specialized sense) is related to the
meaning of ‘read’. If so, we can, and presumably should, account for this
in our formalization of the semanticinformation associated with ‘ readable’
in box (iv); and, even if the specialization of meaning involved in this
case is peculiar to ‘readable’, it is after all specialization. ‘Readable’ does
not mean, for example, “‘tasty”’ or “‘capable of rational demonstration ”’.
The native speaker’s knowledge of the meaning of ‘readable’ is pre-
sumably based upon his knowledge of the meaning of the verb ‘read’
and is supported, to the degree that ‘readable’ is derivationally regu-
lar, by the general grammatical and lexical structure of the language.
We have assumed that all the semantically regular adjectives with
532 The Lexicon
stems ending in -able can be paraphrased by means of a modalized
passive predicative phrase: ‘can be obtained’, ‘can be justified’, etc.
There are many such adjectives, however, which allow, or require, an
interpretation in which the modality of the associated predicative phrase
is that of necessity or obligation, rather than possibility. To say that
something is valuable, preferable, commendable, deplorable, enviable,
detestable, etc., is to say that it ought to be, rather than that it can be,
valued, preferred, commended, deplored, envied, detested, etc. All the
verbs from which these adjectives are derived are verbs of evaluation. It
is conceivable, therefore, that the derivation of the adjectives in question
might be handled within the grammar by virtue of a rule that is sensitive
to a distinction between transitive verbs of evaluation and other transi-
tive verbs. This would then be another class of semantically regular
adjectives with stems in -able for which no distinct lexical entries are
required. But there would be exceptions to the proposed subregularity.
‘Criticizable’, for example, differs in this respect from ‘deplorable’ and
‘detestable’. Let us suppose, therefore, that only one kind of complete
derivational regularity is allowed. We will, then, put ‘deplorable’,
‘enviable’, etc., into the lexicon and handle them in the way suggested
for ‘readable’. But in this case we should certainly wish to give recogni-
tion to the fact that the semantic specialization involved is found
throughout a significant number of lexemes and that it depends upon
the relationship between the modalities of necessity and possibility, or
obligation and permission, which is of importance elsewhere in the
grammatical and lexical structure of English (cf. 17.1). Just as an im-
perative sentence, like ‘Sit down!’ or ‘Come in!’, may be used, in the
appropriate circumstances, either to issue a command or to grant per-
mission, so there are certain derived adjectives with stems in -able which
are interpretable in terms of either necessity or possibility according to
the context. One example is ‘payable’. This bill 1s payable immediately
would normally be understood to imply the necessity, rather than the
possibility, of immediate payment; whereas This bill 1s payable at any
post office is paraphrasable as ‘‘ This bill.can be paid at any post office’.
If the imperative is normally associated with the modality of necessity,
but may in certain circumstances be associated with possibility or per-
mission, the converse appears to be the case as far as derived adjectives
with stems in -able are concerned. This point is mentioned here, simply
to give the reader some idea of the way in which a fuller and more
systematic account of the meaning of one class of semantically regular
derived lexemes might proceed.
 13.2. Complex Lexemes 533
All the lexemes with stems ending in -able that have been referred to
so far have been morphologically and syntactically regular in terms of
the formula Vir + able -> Az. There are others that are morphologically
irregular in that there is no corresponding verb-stem from which their
stems can be derived. They fall into several subclasses. One such sub-
class is exemplified by ‘feasible’, ‘legible’, ‘edible’, ‘intelligible’, etc.
Although there are no verbs in English whose stems are feas-, leg-, ed-,
intellig-, etc., it is arguable that the adjectives whose stems are formed
from these bound* roots do in fact satisfy the formula Vir + able > Az
(-zble being a variant of -able);4! and some of them at least are seman-
tically regular. For example, ‘edible’ is related semantically to ‘eat’ as
‘justifiable’ is to ‘justify’ or ‘obtainable’ is to ‘obtain’; and it is less
specialized in meaning than the morphologically regular “eatable’. One
way of accounting for the distribution and meaning of ‘edible’ (and
also for the fact that it is morphologically regular with respect to
nominalization: cf. ‘edibility’) might be to put ed +- zb/e in box (i) and
‘eat’ + able in box (ii) of the lexical entry for ‘edible’. Given the
appropriate conventions, everything else is derivable by rule. “Legible’
is like ‘edible’, except that its meaning is somewhat more specialized
than ‘“‘can be read’’, but specialized in a different way from that of the
morphologically regular ‘readable’; and this would need to be indicated
in box (iv). ‘Edible’, ‘legible’, and many other lexemes, then, are
morphologically and syntactically regular in terms of the formula
Vir + able — Az.
There are, however, several lexemes with stems in -able (or -zble) that
do not satisfy this formula: ‘horrible’, ‘knowledgeable’, ‘reasonable’,
etc. In each case, the lexeme is an adjective, as the form of its stem
suggests; but none of them belongs to the class of adjectives which are
nominalized by virtue of the rule Vir + able/ible +- ity — Nz, as are all
the regular adjectives with stems in -able. Furthermore, many of them
are morphologically and semantically idiosyncratic in relation to the
lexeme from which they appear to be derived. For example, the stem of
‘knowledgeable’ is patently analysable as knowledge +- able. But know-
ledge never functions elsewhere in English as a verbal stem; and, even if
we were to recognize knowledge in this instance as a bound verbal stem

11 For present purposes, a free* form may be defined as one that may function
as a word, phrase or complete utterance, and a bound* form, in contrast, as
any form that is not free. The distinction between free and bound forms is of
particular importance in work that derives, directly or indirectly, from
Bloomfield (1935).
534 The Lexicon
(meaning “‘know”’ or whatever), the modality of possibility or necessity,
which we have associated with Vir-+-able —> Az, is irrelevant to the mean-
ing of ‘knowledgeable’. Since N + able > A is certainly not a syn-
chronically productive rule of modern English, there is little point in
treating ‘knowledgeable’ as anything other than a simple lexeme in the
lexicon, despite the fact that it is obviously related semantically to
‘knowledge’. The same holds for all the other adjectives with stems in
-able which do not satisfy the formula Vir + able > Az.
We will not proceed any further with our discussion of derived, or
complex, lexemes. What we have tried to do here is to show, with
reference to just one class of derived lexemes in English, that the prob-
lem which faces the linguist is not simply that of deciding whether a
particular lexeme can or cannot be generated by rule. He must account,
as systematically as he can, for various kinds and degrees of derivational
regularity. Little progress has yet been made towards the solution of
this problem within the framework of generative grammar, or indeed
within any coherent theory of the structure of language.!? As we saw
earlier, Chomsky has argued for what he calls the lexicalist treatment of
complex lexemes. But he has also accepted the possibility, in principle,
of “‘a compromise solution that adopts the lexicalist position for certain
items and the transformationalist position for others’’ (1972: 17). Such
proposals as we have made in our discussion of the question, informal
though they have been, are consistent with this kind of compromise
solution.

13.3. Compound lexemes

We may now turn our attention to what is traditionally referred to as
compounding*. A complex (or derived) lexeme, as we have seen, is one
whose stem, Y, is formed from a simpler stem, X, by affixation or some
other kind of morphological modification (the limiting case being that of
zero-modification, or conversion). A compound lexeme, or compound”*,
on the other hand, is one whose stem is formed by combining two or

12 Householder (1959), and later Halle (1972), propose that there should be a
sub-grammar, within the grammar of English and other languages, whose
function it will be to define the set of derivationally well-formed complex
lexemes, a subset of which will be listed, in the lexicon, as having been actuali-
zed. For a range of views, cf. Brekle (1970), Chapin (1967), Chomsky (1970),
Dubois (1962), Guilbert (1975), Lees (1960, 1970), Lipka (1972), Ljung
(1970), Newmeyer (1971). An important work that antedates the rise of
generative grammar is Kurylowicz (1936). For various patterns of derivation
in English cf. Adams (1973), Marchand (1969).
 13.3. Compound lexemes 535
more stems (with or without morphological modification). In so far as
the distinction can be drawn, in particular languages, between word-
forms and combinations of word-forms, a corresponding distinction can
be established between word-compounds and phrasal compounds. ‘The
distinction is less clear in English than it is in many other Indo-European
languages; and the inconsistency with which spaces and hyphens are
employed in written English reflects this fact. The principal criterion for
drawing the distinction, in spoken English, between word-compounds
and phrasal compounds is that of stress. Generally speaking, each word-
form in English (if it is a form of a lexeme belonging to one of the major
parts-of-speech) has a single primary stress; and the position of primary
stress in word-forms of more than one syllable is determined by the
morphological composition of the stem. Given that both simple and
derived stems in English have a distinctive stress-pattern, compound
stems which have a single primary stress (e.g., screwdriver, blackbird, boy-
friend, window box) may also be classified as word-stems, regardless of
whether they are conventionally hyphenated or not in the written
language.
But we are less concerned here with the distinction between word-
compounds and phrasal compounds, than we are with the distinction
between compound lexemes*, on the one hand, and what, for want of a
better term, we will call syntactic compounds*, on the other. Syntactic
compounds are like completely regular derived lexemes in that their
meaning and distribution can be accounted for in terms of the produc-
tive rules of the language-system; and, for that reason, they need not be
listed in the lexicon. Indeed, unlike derived lexemes, they could not in
principle be listed in the lexicon, since, in certain languages at least, they
are infinite in number.
Compound lexemes frequently originate as syntactic compounds and,
having become institutionalized, acquire a more or less specialized mean-
ing. An obvious example is ‘country house’, which is regular enough in
terms of the syntax of English, but which, as a compound lexeme in
British English, denotes a much smaller class of dwellings than does the
expression ‘house in the country’. If it were not for the fact that
‘country house’ has, ‘for historical reasons, come to be associated in
Britain with what once were and in some cases still are the non-
metropolitan residences of the aristocracy, there would be no reason to
treat it as a compound lexeme. It is interesting to compare ‘country
house’ in this respect with the French ‘maison de campagne’, which one
might expect to be its translational equivalent. ‘Maison de campagne’
536 The Lexicon
is also institutionalized as a ready-made or fixed expression, in so
far as it is regularly employed by those who normally live in the town in
order to refer to the house in the country in which they might spend their
week-ends and their holidays. But its meaning is less specialized than
that of “country house’ (the closest equivalent to which is probably
‘chateau’ — conventionally, but in most cases inappropriately, translated
into English as ‘castle’); and it is, for this reason, less obviously a com-
pound lexeme than ‘country house’.
The process by which syntactic compounds are institutionalized as
lexemes has been aptly called petrification (cf. Leech, 1974). This
metaphorical.term is intended to suggest two distinguishable aspects of
the process in question: solidification and shrinkage. As soon as any
regularly constructed expression is employed on some particular occa-
sion of utterance, it is available for use again by the same person or by
others as a ready-made unit which can be incorporated in further
utterances; and the more frequently it is used, the more likely it is to
solidify as a fixed expression, which native speakers will presumably
store in memory, rather than construct afresh on each occasion. In this
respect, frequently used syntactic compounds are like frequently used
regular derived lexemes. Solidification, then, is a natural consequence
of the normal use of language; and, just as naturally, though by no
means inevitably, it leads to the other aspect of the process of petrifica-
tion, shrinkage or semantic specialization. Just as a simple lexeme may,
by virtue of its use in particular contexts, become more restricted in its
sense and denotation than it was in some earlier period, so too may
derived lexemes and syntactic compounds. This is a more or less in-
evitable consequence of the normal use of language; and it creates both
practical and theoretical problems for the lexicographer. How does he
decide whether the process of petrification has-gone far enough, in any
particular instance, to justify the inclusion of a separate lexical entry?
And what kind of information should be associated with compound
lexemes in the lexicon?
It is easy enough to formulate the general criteria for inclusion in the
lexicon: a lexical entry is required for compound lexemes (and it is this
property which makes them lexemes) if and only if they are phono-
logically, morphologically, syntactically or semantically idiosyncratic. In
practice, this criterion is difficult to apply, because it is hard to draw a
sharp distinction between regularly constructed, but institutionalized,
syntactic compounds and petrified compound lexemes. We have assumed
that ‘country house’ 1s, on semantic grounds, to be classified as a com-
 13.3. Compound lexemes 537
pound lexeme in British English. But it is indistinguishable in terms of
its phonological, morphological and syntactic characteristics from in-
numerable endocentric noun expressions in which the head-noun is
modified by another adjectivalized noun, like ‘week-end cottage’, ‘car
radio’, or “garden furniture’ (cf. 10.3); and many of these are certainly
institutionalized, in that they denote classes of things that have a more
or less distinctive role in present-day life. It is for cultural reasons that
“week-end cottage’ is of more frequent occurrence than, say, ‘week-day
cottage’. But if anyone should choose to live in town at the week-end
and in the country during the week, he would be quite free to refer to
his country residence (if it were a cottage) by means of the expression
‘week-day cottage’. This expression 1s fully acceptable and semantically
interpretable in terms of the productive rules of the language-system;
but it has not been institutionalized. It requires but little reflexion to see
that institutionalization, like petrification, is a matter of more or less,
rather than yes or no. Not only 1s ‘country house’ syntactically endo-
centric, in that its distribution is identical with that of ‘house’, but it is
semantically regular to the extent that its sense is related to that of its
component head-noun in terms of hyponymy. In this respect it differs
from ‘public house’ (in British English), and still more from ‘green-
house’ (which is identifiable, phonologically, as a word-compound). The
native speaker’s understanding of the sense and denotation of ‘country
house’ is presumably supported by, though it cannot be completely
explained by, his recognition of its internal syntactic structure and his
knowledge of the meaning of both ‘house’ and ‘country’. Furthermore,
the existence of ‘country house’ as a compound lexeme does not abso-
lutely prevent the construction by native speakers of the corresponding
syntactic compound whose distribution is identical: a sentence like ‘I
don’t like country houses’ is presumably ambiguous.
Granted that ‘country house’ is a compound lexeme, but that it is to
be related both grammatically and semantically to the simple lexemes
‘country’ and ‘house’, we come up against the theoretical problem that
there is no obvious way of doing this satisfactorily within the framework
of generative grammar. Conventional dictionaries sometimes adopt the
practice of incorporating lexical entries for compound lexemes within the
entries for one or other of the component simple lexemes, and this can
be seen as an informal way of indicating that there is some kind of
relationship between the compound and its lexical components. But the
nature of this relationship is not made explicit. Conventional dictionaries
trade very heavily, and justifiably in view of their aims, upon their users’ ~
 538 The Lexicon
intuitive knowledge, not only of the grammatical structure of the lan-
guage, but also of the kinds of things that the language is normally used
to describe or to refer to. If we are to make explicit the degree to which
the meaning and distribution of the compound lexeme ‘country house’
is determined by the meaning and distribution of ‘house’ and ‘country’,
“we must first of all have some means of representing within its lexical
entry the fact that it is composed of ‘house’ and ‘country’ and that it is,
as far as its distribution is concerned, a regular endocentric noun-
compound. We must then be able to use this information in the semantic
part of the lexical entry to indicate the relation of hyponymy that holds
between the compound as a whole and the simple lexeme ‘house’. It is
not possible to do more than make some very tentative suggestions here
as to the way in which this might be done; and the principal reason is
that the treatment of syntactic compounds within the framework of
generative grammar is, if anything, even more problematical than is the
treatment of derivational morphology.
We have been assuming that endocentric noun-expressions like
‘week-end cottage’, ‘garden furniture’, ‘car radio’, etc., are to be
generated by the productive rules of the language-system; and this
would appear to be a reasonable assumption. It has frequently been
pointed out, however, that the semantic relationship between the head-
noun and the modifying noun in such phrases is extremely diverse. If
they are to be transformationally derived from some underlying senten-
tial structure in which the head-noun is part of the subject and the
modifying noun is part of the predicate, we must allow for the trans-
formational derivation of the same denominal adjectival modifier from
many different predicative expressions. This is not in itself an objection
to the transformational derivation of syntactic compounds. Indeed, the
fact that many regular syntactic compounds are, in principle, ambiguous
is naturally accounted for by deriving them from several distinct under-
lying structures. For example, “the London train’ might refer to a train
which is going to London or coming from London (cf. Has the London
train left yet? vs. Has the London train arrived yet?). The expression
‘London taxis’, on the other hand, will normally be understood to refer
to taxis which operate in London, as will ‘London buses’. But ‘the
London bus’ can also be used to refer to a bus which is going to or
coming from London. It is inconceivable that the syntactic and semantic
subclassification of either ‘London’ or ‘train’, ‘bus’ and ‘taxi’ in the
language-system could be such that these differences could be accounted
for by rule. In any case, it is surely only our knowledge of the fact that
 13.3. Compound lexemes 539
trains normally operate between towns, taxis within towns and buses
both between and within towns that leads us to say that one interpreta-
tion is more normal than another in any particular instance. Once this
point is conceded, it is difficult to avoid the conclusion that every
syntactic compound is, in principle, highly ambiguous, although one
interpretation rather than the others, whether it is institutionalized or
not, may seem more natural or more normal in particular utterances.
In this respect, syntactic compounds are no different from very many
sentences, whose ambiguity usually goes unnoticed because they are
interpreted within a framework of shared ontological or contextual
assumptions. As we have seen, it is a debatable point whether sentences
can be said, in general, to have a determinate and integral number of
meanings (cf. 10.4). It is also matter for dispute how much of the
potential ambiguity of sentences is to be excluded by selection restric-
tions*, formalizable in our model of the language system, and how much
should be held to fall outside the scope of linguistics entirely (cf. 10.5).
Both of these questions are obviously relevant to the analysis of syntactic
compounds.
What we are concerned with here is a rather different question. On the
assumption that syntactic compounds are to be generated, within the
grammar, from underlying sentential structures, and that any given
syntactic compound may be derived from several sources, are any or all
of the ambiguities preserved in the formal representation of the syntactic
structure of the resultant compound? Is ‘London bus’, for example,
assigned the same syntactic analysis in the surface structure of sentences
regardless of whether it means “‘bus from London’’, “bus to London”’,
‘bus in London’’, etc.? As far as the interpretation of regular syntactic
compounds, like ‘London bus’, is concerned, this question might not
seem to be very important. After all, the different interpretations are
satisfactorily accounted for in the deep structure representation; the
deeper syntactic differences are not relevant to the distribution of the
noun expression ‘London bus’ throughout the sentences of the lan-
guage; and there are no correlated morphological or phonological
differences. What reason, then, is there to preserve the deeper syntactic
differences in the surface structure representation?
One reason is that it would enable us to classify compound lexemes in
the lexicon according to the subtype of syntactic compound with
which they are structurally identical. For example, on the assumption
that ‘country house’ as a compound lexeme, is structurally identical
with syntactic compounds like ‘country cottage’ under one of its
540 The Lexicon
interpretations, which are derived from a sentential structure in which
the adjectivalized noun occurs in a predicative locative expression (cf.
‘cottage which is in the country’ — ‘country cottage’), it could be
classified in the lexicon as an endocentric noun compound of the par-
ticular subtype composed of ‘house’ and ‘country’ (these two simple
lexemes being referred to by their lexical addresses). Part of the meaning
of ‘country house’, as well as its distribution, would then be predictable
by rule, in much the same way as the meaning and distribution of regu-
lar complex, or derived, lexemes is predictable by rule. If the meaning of
‘country house’ is (roughly) “‘house in the country belonging to (or
having once belonged to) an aristocratic family”’, all that would need to
be explicitly represented in the semantic part of the lexical ‘entry is
‘belonging to... an aristocratic family’’. Just how this part of the sense
of ‘country house’ might be represented,.in terms of universal or
culturally specific semantic components or otherwise, is of course
problematical (cf. 9.9). But this problem is there anyway, even for simple
lexemes. It is not produced as an artefact of the proposals that we are
making in relation to compound lexemes.
‘Country house’ will serve as an exemplar of what is a very large class
of compound lexemes. As we have seen, it is completely regular as far as
its phonological, morphological and syntactic properties are concerned;
and its status as a lexeme depends solely upon its idiosyncratic and un-
predictable semantic specialization. Its sense, we have decided for the
sake of the argument, is the product of three components “X.”’, ‘“Y”’
and ‘‘Z’’, where “‘X”’ is the sense of ‘house’, “Y”’ is the sense of the
expression “‘in the country” and “Z”’ is the idiosyncratic residue. It
would be possible, of course, to disregard the fact that ‘country house’
is a compound lexeme in the semantic part of its lexical entry. But we
are making the surely plausible assumption that the native speaker’s
knowledge of the meaning of ‘country house’ is determined, in part, by
his knowledge that it is composed of a noun-head, ‘house’, and a noun
modifier, ‘country’; and that these two simple lexemes are combined
according to the productive rules of the language to yield a particular
kind of endocentric noun compound, each lexeme having the sense, or
one of the senses, that it has elsewhere in the language. Let us assume,
therefore, that the compound lexeme ‘country house’ is represented in
the lexicon as a combination of the two lexemes 731 and 1321 (731 being
the address of ‘country’ and 1321 being the address of ‘house’) and
furthermore that it is characterized in the syntactic part of the lexical
entry as belonging to a particular subclass of endocentric noun expres-
 13.3. Compound lexemes 541
sions, which we will arbitrarily label N,. This information is sufficient
to account for the phonological, morphological and syntactic regularity
of ‘country house’ in relation to its constituent simple lexemes.
What is now required is some convention, whereby we can interpret
this information, together with what other information is given in the
semantic part of the lexical entry, in order to derive its sense from the
sense of ‘country’ and ‘house’. We know from the grammar that N, 1s
a bi-partite phrasal construction N ++ N, in which the first constituent
is an adjectivalized locative modifier and the second constituent is the
head. If there were no entry for ‘country house’ in the lexicon, it would
be interpreted semantically, as an instance of N,, as an expression mean-
ing “‘house in the country”’; and this part of its sense we want to be able
to derive by rule. Let us assume, therefore, that for every compound
lexeme composed of n simple lexemes, there are m + 1 spaces set aside in
the semantic part of the lexical entry; that one of these is reserved for
the idiosyncratic part of the sense of the component; and that each of
the other spaces is associated with one of the constituents of the com-
pound. In the present instance, there will be three spaces for semantic
information. We will refer to them as the H-space, the M-space and the
S-space (where H, M, and S stand mnemonically for ‘head’, ‘modifier’
and ‘specialization’). Since ‘country house’, under the assumptions we
are making, is regular in so far as its sense is a function of the senses of
‘country’ and ‘house’ when they are combined in an expression of the
class N,, we will leave the H-space and the M-space empty in the seman-
tic part of the lexical entry. But we will put in the S-space, in whatever
format is adopted for this purpose, the information that ‘country house’,
as a compound lexeme, includes in its sense the component “ belonging
to an aristocratic family’’.
Granted that the meaning of the compound lexeme is the product of
“X”, “Y” and “Z”’, the values of X and Y will be determined (by
virtue of the convention that the absence of any information to the con-
trary implies regularity) as the sense of lexemes 731 and 1321, respec-
tively; and these will be combined with ‘‘Z’’, which is given in the
S-space.
Several points may now be made on the basis of the technique that
has just been outlined. First of all, it is obvious that there are, in prin-
ciple, several ways in which the components of the sense of a compound
lexeme may be combined to yield the sense of the compound as a whole.
The proposals that have been made here depend very heavily upon the
assumption that there is a limited and determinate number of ways in
542 The Lexicon
which the senses of lexemes may be combined, or amalgamated, and
that each of these may be associated systematically with particular kinds
of syntactic relations. It may well be, of course, that the meaning of even
regular syntactic compounds is too loosely related to that of their con-
stituent lexemes for the meaning of the compound to be predictable by
rule. But, if this is so, it will be impossible to generate by means of a
finite set of rules operating (in association with the grammar) upon a
finite lexicon the meaning of all the sentences of English. For the set of
regularly derivable syntactic compounds in English (and presumably in
all languages) is indefinitely large. The assumption that the meaning of
an expression is a function of the meaning of its constituents is taken for
granted in much of the most recent work in theoretical semantics; and it
is difficult to see how this assumption could be abandoned without
simultaneously abandoning the attempt to formalize the semantic struc-
ture of sentences.
Our proposals also depend upon the assumption that the way in which
the idiosyncratic part of the meaning of a compound lexeme is amalgama-
ted with the meanings of its constituent simple lexemes ts precisely
specifiable. That this assumption is perhaps untenable becomes clear if
we compare ‘country house’ with ‘washing machine’. The fact that
“washing machine’ in English (like ‘machine a laver’ in French) is now
a compound lexeme, or at least is well on the way to having acquired
lexemic status in the language-system, is suggested by the fact that such
utterances as Is that a washing machine or a dish-washer?, or even Is that a
washing machine or a dish-washing machine?, appear to be completely
acceptable. Structurally, ‘washing machine’ can be related to an in-
definitely large class of expressions (many of which are institutionalized):
‘reading lamp’, ‘gardening jacket’, “swimming costume’, etc. (We will
here discount the possibility that ‘washing machine’ is structurally
ambiguous, being comparable also with ‘running water’, ‘sleeping
partner’, ‘standing committee’, etc.). The meaning of ‘washing
machine’, in so far as it is regular, is “machine (used/usable) for wash-
ing’’. If the sense of ‘clothes’ is to be amalgamated with this, however,
it is obvious that it must be combined in some way as the object of
‘washing’: cf. ‘clothes-washing machine’, ‘machine for washing
clothes’. It follows that this information must be given in the S-space.
Alternatively, if what is put in the S-space is no more than “‘(used/usable)
for clothes”’ and if the convention is held to apply, according to which
the idiosyncratic part of the sense is combined with “‘machine for
washing”? in the same way that “for washing”’’ is combined with
 13.3. Compound lexemes 543
‘‘machine”’, the resultant conjunction “‘ washing machine (used/usable
for clothes’) will be underspecified. But this is, once again, a more
general problem: ‘tennis dress’ normally means “‘dress for playing
tennis (in)”’, ‘bread-knife’ means “knife for cutting bread (with)’’; and
so on. In the case of ‘washing machine’ there is at least present in the
forms of the compound a form of the verb ‘wash’; and if “dress for
tennis” and “knife for bread’’ are considered specific enough to serve
as the meaning of ‘tennis dress’ and ‘bread knife’ (which, though they
are institutionalized as expressions, are surely not to be regarded as
lexemes), there is perhaps no reason why “machine for washing for
clothes” should not be regarded as a satisfactory analysis of the sense of
the compound lexeme ‘washing machine’. We can, however, leave open
the possibility that it may be necessary to specify in the S-space of a
compound that the idiosyncratic part of its meaning combines with that
of one, rather than another, of the constituent lexemes and that it does
so in a particular way.
Before we continue with the discussion of compound lexemes, it
should be pointed out that, if the way in which the idiosyncratic part of
the meaning of a compound lexeme is too loose and too diverse to be
brought within the scope of rules of the kind that operate in the deter-
mination of the meaning of regular syntactic compounds (as has often
been suggested), this point holds equally well for simple and complex
lexemes. They are no less likely to acquire, by institutionalization and
subsequent petrification, idiosyncratic restrictions than are compound
lexemes. What is at issue, in fact, is the whole question of lexical de-
composition: is it possible to represent the sense of lexemes, without
residue, as a compositional function of sense-components and, if so,
what are the combinatorial principles? ‘This is one of the most funda-
mental, and controversial, questions of theoretical semantics at the
present time. If it cannot be answered in the affirmative for compound
lexemes, it seems clear that lexical decomposition, as such, must be
rejected completely; and we have already seen that lexical decomposition
is suspect on other grounds (cf. 9.9).
A further point that must be made explicit before we proceed has to
do with the relevance, in relation to compound lexemes, of the distinc-
tion between homonymy* and polysemy* (cf. 13.4). A sentence like ‘I
hate the country’ is ambiguous in English: so too is ‘I hate the town’.
Under one interpretation, ‘the country’ and ‘the town’ are singular
definite referring expressions, and the lexemes ‘country’ and ‘town’
occurring in them denote such classes of entities (or places) as England,
 544 The Lexicon
France, Germany, etc., and Manchester, Birmingham, Stratford-upon-
Avon, etc. Under the other interpretation, ‘the country’ and ‘the town’
_ do not refer to entities (or places) of this kind at all; their mode of
reference is similar to, if not identical with, that of singular mass nouns.
Now, it is clear that the compound lexeme ‘country house’ (unlike
‘house in the country’) is not ambiguous. If the ambiguity of ‘I hate the
country’ is accounted for by recognizing two distinct lexemes ‘country,’
and ‘country,’ (the first of which is not pluralizable and never occurs
except with the definite article), there is no problem, in so far as our
proposed technique for handling the relationship between ‘country
house’ and ‘country’ is concerned. ‘Country,’ and ‘country,’ will have
two different lexical addresses; and it will be the address of ‘country,’
that is given in the M-space. But if ‘country’ is treated as one lexeme
with two meanings, “‘country,’’ and “country,’’, we shall have to be
able to identify in the lexical entry for ‘country house’ which of the two
senses is involved. The distinction between homonymy and polysemy,
as we shall see later, is very difficult to establish on general grounds, and
may indeed rest upon ultimately untenable assumptions about the
discreteness of the senses of lexemes. However that may be, the point
that we have raised in this paragraph is an important one.
Having discussed, at some length, the nature of such compound
lexemes as ‘country house’, we can now move on to deal more briefly
with other kinds of compound lexemes, whose meaning cannot be
accounted for, even in principle, as the product of the meaning of a
regular syntactic compound, on the one hand, and of a more specialized,
idiosyncratic component, on the other. First of all, there are several
classes of compound lexemes which are phonologically and grammatic-
ally regular, but semantically irregular with respect to one or more of
the simple lexemes of which they are composed. For example, ‘public
school’ in British English is semantically regular with respect to its head:
‘public school’ is a hyponym of ‘school’, and this can be indicated in
the lexicon in the way that the hyponymy of ‘country house’ and .
‘house’ is indicated. But the sense of the adjective ‘public’ does not
enter into the meaning of the compound lexeme ‘public school’ as the
sense of ‘country’ is included in the sense of ‘country house’. ‘Public
school’ denotes a subclass of the class of institutions in Great Britain
denoted by ‘private schools’ (which, like ‘State school’, but unlike
‘grammar school’, ‘prep school’, etc., is a regularly derivable expres-
sion).
It might be argued, of course, that ‘public’, in this case, has a meaning
. 13.3. Compound lexemes 545
which it never has in any other context. Suppose we put ‘public,’ into
the lexicon as a homonym of ‘public,’ (the more common and more
freely combinable adjective) and associate with ‘public,’ whatever it is
in the sense of ‘ public school’ which distinguishes it from ‘State school’,
‘grammar school’, ‘prep school’, etc. ‘Public school’ would then be
regularly derivable; and it would not need its own lexical entry. But we
should have to indicate in the lexical entry for ‘public,’ the fact that it
can occur only as a pre-noun modifier and, moreover, that the only noun
it can modify is ‘school’. (Adopting this treatment would be like putting
a verb ‘auth’ into the lexicon with the information that its stem auth
must necessarily be combined with the derivational agentive sufhx to
form the stem of the noun ‘author’: cf. 13.2.) Another alternative is to
opt for polysemy rather than homonymy, saying that ‘public’ has (at
least) two distinct senses “‘public,’’ and “‘public,’”’. Once again, we
should have to indicate in the lexical entry for ‘public’ that, when it
means ‘‘public,’’, it must be combined with ‘school’. The theoretical
implications of adopting one, rather than the other, of these solutions
should not be overlooked. In the first case, by taking ‘public,’ as a
distinct lexeme and restricting its syntactic distribution in the way that
is required, we should be setting up, in effect, a one-member subclass of
adjectives in English. In the second case, we should be making the syn-
tactic distribution of ‘public’ a function of its meaning. There would
seem to be little point in thus complicating the syntactic description of
the language in order to handle what can be handled equally well by
putting into the lexicon an entry for ‘public school’, which relates it
semantically to ‘school’, but not to ‘public’. How might this be done?
Let us recall that the convention suggested for ‘country house’ was
that it should be regarded as being composed of two nouns, identified by
their lexical addresses, and that the semantic part of the entry for
‘country house’ should have three spaces: an H-space for the head, an
M-space for the modifier, and an S-space for the idiosyncratic part of
its meaning. Adopting the same format for ‘public school’ (and assum-
ing that it is identified as an endocentric compound), we would leave the
H-space empty (thus accounting for the hyponymy of ‘public school’
and ‘school’) and we would indicate by a special symbol, the identity-
symbol or zero (%), in the M-space that the modifier is semantically
vacuous. ‘The third space would then carry the idiosyncratic part of the
meaning of ‘public school’. The meaning of ‘public school’ would be
thus determined as “school which is and which Z’’. We have here used
the identity-symbol, or zero, to represent the semantically vacuous
546 The Lexicon
component: i.e. the part contributed by the adjective ‘public’. Several
questions arise in relation to this treatment.
First, it might be asked why, if ‘public’ makes no contribution to the
meaning of ‘public school’, it is nonetheless recognized as a constituent
of the compound lexeme. The answer is that we need to get the form
public, just as we need to get the form school, from the lexicon; and we
must either put the form public in the entry for ‘public school’ or obtain
it from the lexical entry for ‘public’. If we identify ‘public school’ as
public + 523 (i.e. as a combination of a form and a simple lexeme — 523
being the address of ‘school’), we are saying that there is no relationship
of any kind, other than accidental coincidence of form, between the
simple lexeme ‘public’ and the modifier of ‘school’ in the compound
lexeme ‘public school’. The alternative, which we have adopted, is
based on the assumption that the existence in the vocabulary of English
of a simple lexeme ‘public’ and the fact that ‘public school’ is of the
same syntactic type, and has the same stress-pattern, as regularly
derivable phrases composed of an adjective and a noun, is sufficient to
justify this treatment. It must be admitted, however, that in default of

somewhat arbitrary.
any positive morphological reasons in favour of this solution, it is

A second question has to do with the status and function of the

identity-symbol. It might be objected, with some justification, that what
we have in fact done by means of a purely technical device is to transfer
the meaning of the modifier ‘public’ (which is in paradigmatic contrast
with ‘grammar’, ‘State’, etc., when they are combined with ‘school’)
to the otherwise unnecessary third part of the semantic entry. Is it not
preferable to associate the part of the sense of ‘public school’ that we
have put in the S-space — let us again call it “‘Z’’ — with the adjective
‘public’? This is easily done in either of two ways: (i) by putting “Z”’
directly in the M-space; or (11) by leaving “Z’’ in the S-space and em-
ploying a distinctive symbol, say S, to indicate that the modifier has the
specialized meaning given in the S-space. There is perhaps little
difference between (i) and (ii), though (ii) might be thought of as the
diachronic precursor of (1). The solution which we have adopted, how-
ever, is based on the view that there are no positive reasons to associate
the distinction between ‘public school’ and ‘State school’ with the
adjective ‘public’ rather than with the compound as a whole.
The lexeme ‘ public house’ (in British English) exemplifies yet another
subclass of phonologically and grammatically regular, but semantically
irregular, compound lexemes. It does not stand in a relationship of
 13.3. Compound lexemes 547
hyponymy with ‘house’; ‘public’, however, does bear the sense here
that it has elsewhere. There is of course an obvious historical connexion
between the adjective ‘public’ and ‘public house’. But one can no more
deduce the specialized meaning of ‘public house’ from its components
than one can deduce the meaning of the Russian ‘ publiényj dom’ (which
is used to refer to brothels) from the meaning of the adjective ‘ publi¢nyj’
(“public”) and the noun ‘dom’ (“house’’). In the case of ‘public
house’, there is even less reason to associate its specialized meaning
with one of the two simple lexemes of which it is composed than there
is in the case of ‘public school’. What we can do therefore is to put the
identity-symbol, or zero, in both the H-space and the M-space. The
whole of the sense of ‘ public house’ would then be given in the S-space.
Yet another subclass of endocentric compound lexemes is exemplified
by ‘motor car’ which, in so far as it is still current in the language-
system, is synonymous with ‘car’ (in one of its senses). What has
happened in the diachronic development of these two lexemes is some-
thing that happens very frequently as a consequence of cultural changes.
The adjectivalized noun ‘motor’ was once significant in the compound
‘motor car’ (as it still is in “motor boat’). The sense of ‘car’ was then
specialized to the point that the modifier became redundant. On the
assumption that the lexicon contains both ‘motor car’ and ‘car’, and
that the latter is now petrified in the sense we are concerned with here,
it would seem to be natural to represent the meaning of ‘motor car’ by
putting zero in the M-space and leaving the H-space empty. In this
case, and in many others that are similar, the proposed notational
technique neatly reflects the diachronic process whereby one of the
constituents of a compound lexeme transfers its meaning to the other
and, in doing so, itself becomes redundant.
All the compound lexemes that we have looked at so far have origina-
ted, we assume, as syntactic compounds that are derivable by the
productive rules of the language-system. There are at least two other
classes of compound lexemes which are not of this kind; and they fre-
quently manifest various kinds of irregularity. The first class is made up
mainly of word-compounds, which are fossilized (rather than being
merely petrified) in that the rule by which they are derived from the
simple lexemes of which they are composed is no longer productive in
the present state of the language-system. ‘Two such examples are ‘ pick-
pocket’ and ‘turn-coat’ which are synchronically irregular (in contrast
with, say, ‘tooth-pick’, on the one hand, or ‘safe-breaker’, on the other)
in that it is not possible for a native speaker of English to construct
 548 The Lexicon
compounds composed of a verb-stem and a noun-stem (in that order)
from an underlying sentential structure in which the noun (or an
expression containing it) is the object of the verb. (The contrast with
modern French in this respect is striking: cf. ‘ouvre-boite’, “can-
opener”’, ‘allume-gaz’, “‘gas-lighter’’, etc.) Since fossilized compounds,
by definition, do not belong to synchronically productive classes, we
will assume that they are listed in the lexicon as if they were simple
lexemes.
The second major class of compound lexemes which do not conform
to the productive rules of the language-system is theoretically more
interesting. What is involved, in this case, is the application of deriva-
tional principles which do not so much violate the syntactic rules of the
language-system as creatively extend or transcend them. Metaphor*
and metonymy* are the traditional terms under which most of the com-
pounds we are concerned with here can be accounted for.!5 An example
of a compound that is obviously based on metaphor is ‘wet blanket’ in
the sense in which it denotes someone who inhibits others in their
enjoyment or enthusiasm; another is ‘live wire’. Both of these are
phonologically and, up to a point, grammatically regular: He zs a very
wet blanket and He 1s a very live wire are perfectly acceptable utterances.
But as compound lexemes they are not fully endocentric: ‘wet blanket’
and ‘live wire’, unlike ‘blanket’ and ‘wire’, belong to the subclass of
animate nouns. An example of a compound lexeme based on metonymy
- is ‘red cap’, which in American English denotes a porter and in British
English a military policeman (in each case for obvious reasons). Once
again, this is not fully endocentric, since its syntactic distribution is
different from that of ‘cap’ (cf. The red cap who was on duty last night
got drunk).\4 In what follows we will use the term ‘metaphor’ to include
metonymy.
Metaphor is not of course restricted to the formation of compound
lexemes. Many simple lexemes can be used metaphorically and have, for
that reason, acquired more or less institutionalized senses which need to
be put into the lexicon. What is theoretically interesting about metaphor
is that, although it cannot be brought within the scope of a deterministic
3 For a rich collection of literary metaphors, with discussion, cf. Brooke-Rose
(1958). For some recent discussion of the implications of metaphor and
figurative usage with reference to its implications for linguistic theory, cf.
Cohen & Margalit (1970), Loewenberg (1975), McIntosh (1961), Weinreich
1966).
14 tneaving this, I am assuming that the selection of who(m), rather than which,
is a matter of grammar. Arguably it is not (cf. 10.2).
 13.3. Compound lexemes 549
system of generative rules and is normally discussed under the rubric of
stylistics*, rather than semantics (cf. 14.5), it is by no means restricted
to what is often thought of as the more poetic use of language. If a dis-
tinction is drawn between productivity (a design-feature of the language-
system: cf. 3.3) and creativity (the language-user’s ability to extend the
system by means of motivated, but unpredictable, principles of abstrac-
tion and comparison), we can draw a corresponding distinction, with
respect to both the production and the interpretation of language-
utterances, between rules and strategies.
The term ‘lexicalization’, as it is currently used in linguistics, is
ambiguous or equivocal from this point of view. It may refer to what is
formalized in generative grammar as lexical insertion: the selection from
the lexicon of pre-existing lexemes according to their syntactic and
semantic specification. This process is naturally accounted for in the
linguist’s model of the language-system in terms of rules. But ‘lexicali-
zation’ may also refer to the creation of new lexemes. This is the sense
in which we have employed the term in previous chapters; and it is the
sense that concerns us here.
We have assumed that lexicalization, in the sense of the creation of
, lexemes, cannot be accounted for in terms of generative rules. This does
not mean that it is not subject to the constraints imposed by particular
language-systems and perhaps also to more general constraints which
govern all language-systems. But if we are correct in assuming that the
lexicalization of compounds by means of metaphorical extension is a
normal process in the everyday use of language and that it can only be
accounted for in terms of strategies, rather than rules, this casts doubt
upon the validity of yet another assumption with which we have so far
operated: the assumption that all the lexemes of a language can, in
principle, be listed in the lexicon. Conventional dictionaries do no more ,
than list a subset of the compound lexemes that native speakers are
likely to produce; and it is difficult to see how any lexicon could do more
than this.
The question of metaphorical extension has been raised here in con-
nexion with compounds because in this case it has obvious implications
for the actual listing of lexemes. Exactly the same problems arise, of
course, when it comes to the metaphorical extension of the sense of a
1 Bazell (1964) draws a distinction between constraints imposed by the
language-system and restraints to which users of the language-systems will
normally conform; and Haas (1973a) draws a similar distinction between rules
and tendencies.
 550 The Lexicon
simple lexeme, which is no more predictable by rule than is the creation
of a new compound lexeme by means of the same strategies. It is not
generally thought of as extending the vocabulary of the language. Cur-
rent attempts to formalize the semantic structure of language-systems
and to generate all and only the possible interpretations of sentences are
based upon the assumption that, not only the number of lexemes in any
language, but also the number of senses associated with each lexeme, is
finite and enumerable. Metaphor constitutes a very serious theoretical
problem for any theory of semantics that is based on such assump-
tions.1¢

13.4. Homonymy and polysemy

As we saw in an earlier section, there are two kinds of lexical ambiguity,
one of which depends on homonymy* and the other on polysemy*. The
difference between homonymy and polysemy is easier to explain in
general terms than it is to define in terms of objective and operationally
satisfactory criteria!’ Let us begin by asking what are the criteria that
linguists and lexicographers actually apply in coming to the decision
that ‘port,’ (“harbour’’) and ‘port,’ (‘‘kind of fortified wine”’), for
example, are distinct, but homonymous, lexemes, but that ‘mouth’ is a
simple polysemous lexeme — i.e. one lexeme with several different senses
(“organ of body’’, “‘entrance of cave’’, etc.).
One criterion, which is made explicit in the etymological information
that is appended to many dictionary entries, is the lexicographer’s
knowledge of the historical derivation of words. It is generally taken to
be a sufficient, though not a necessary, condition of homonymy that the
lexemes in question should be known to have developed from what were
formally distinct lexemes in some earlier stage of the language. For
example, ‘ear,’ (“the organ of hearing’’) and ‘ear,’ (“part of such
cereal plants as wheat and barley’’) are treated as homonymous lexemes
by virtue of the etymological criterion (because the Old English words
from which they derive were formally distinct and the forms of these
two lexemes merged in Middle English). In practice, the etymological
criterion is not always decisive. First of all, there are many words even
in English (which has written records going back hundreds of years)
16 A further problem, not dealt with in this book, is the existence in all language-
systems of various kinds of idioms: cf. Chafe (1968), Fraser (1970), Makkai
(1972). Idioms frequently originate, of course, in metaphor.
1” ‘Tests of the kind that have been carried out by Lehrer (1974) suggest that
native speakers are in agreement over a fair range of examples of homonymy
and polysemy, but that there is a considerable residue of borderline cases.
 13.4. Homonymy and polysemy 551
about whose historical derivation we are uncertain. Secondly, it 1s not
always clear what is meant by etymological relationship in this context.
The lexeme ‘port,’ (meaning “harbour’’) derives from the Latin ‘ portus’
(which, if we go back far enough in Indo-European reconstruction, 1s
itself related to what in modern English is ‘ford’ and the German verb
‘fahren’). ‘Port,’, on the other hand, came into English fairly recently
and derives from the name of the city in Portugal from which the
particular kind of wine it denotes was exported. But the name of this
city ‘Oporto’ derives in Portuguese from an expression (‘o porto’)
which originally meant, simply, “the harbour’’; and the Portuguese
‘porto’ comes from the same Latin lexeme from which the English
‘porti’ derives. Whether we say that ‘port,’ and ‘port,’ are etymologi-
cally related, therefore, depends upon how far we are prepared to go,
when we have the evidence, in tracing the history of words. The
criterion of etymological relationship is not therefore as straightforward
as it might appear at first sight.
But this is not the main criticism that can be directed against the
etymological criterion. Useful though it may be to have readily accessible
in our standard dictionaries whatever information is available about the
origins and history of particular words, this information is, or should be,
irrelevant in the synchronic* analysis of languages (cf. 8.2). For the
native speaker is generally unaware of the etymology of the words that
he uses and his interpretation of them is unaffected (except when he is
being pedantic or exploiting certain aspects of their etymology for
stylistic purposes) by whatever knowledge of their historical derivation
he may happen to possess. In so far as the etymological meaning of a
lexeme differs from its usual synchronic meaning and is stylistically
relevant, this can be taken into account in the analysis of particular texts
(cf. 14.5). But it should play no part in the definition of homonymy.
The second major criterion that is traditionally invoked by linguists
and lexicographers in drawing the distinction between homonymy and
polysemy is unrelatedness vs. relatedness of meaning (cf. 1.5); and it is
clear that this is a relevant and important consideration. Indeed, it is
arguable that it is the only synchronically relevant consideration. In so
far as the distinction between homonymy and polysemy is pre-
theoretically determinable, it would seem to correlate with the native
speaker’s feeling that certain meanings are connected and that others are
not. For example, all speakers of English would probably agree that the
noun ‘mouth’ is a single lexeme with several related senses (i.e. that it is
polysemous). ‘They would not of course use such theoretical terms as
552 The Lexicon
‘lexeme’ and ‘polysemous’. But they might tell us that in such expres-
sions as ‘the mouth of the river’ and ‘the mouth of the bottle’ it is the
same word, ‘mouth’, that is being used as in an utterance like Don’t
speak with your mouth full; and they might well account for their feeling
that it is the same word by saying that the basic, or literal, meaning of
‘mouth’ is something like ‘‘aperture in the face (through which men
and animals take food, breathe, emit vocal signals, etc.)’’ and that this
meaning has given rise, by some discernible process of metaphorical or
figurative extension, to the use of the same word in referring to other
kinds of openings or apertures. If it is the case that most native speakers
do see a metaphorical connexion between the different senses of what
they take to be the same word, we are perhaps justified in saying that the
word in question is a single polysemous lexeme by virtue of the pre-
theoretical criterion of relatedness of meaning.
There are several problems, however, which arise when the pre-
theoretical criterion of relatedness of meaning is made the basis for the
distinction between polysemy and homonymy. The first of these is that
relatedness of meaning appears to be a matter of degree; and it has yet
to be demonstrated, and may not in fact be demonstrable, that the
intuitions of native speakers coincide sufficiently for it to be worthwhile
looking for some universally applicable and clear-cut distinction between
polysemy and homonymy in the language-system. It has often been
pointed out that some native speakers will claim to see a connexion
between an ear of corn and the part of the body that is denoted by the
noun ‘ear’, whereas other native speakers will deny that any such
connexion exists. Faced with what appears to be an instance of pre-
theoretical indeterminacy like this, we find ourselves in some difficulty.
If it could be shown that the two groups of speakers differ systematically
in their use and interpretation of ‘ear’ or ‘ear,’/‘ear,’, we would be
justified in saying that they speak slightly different dialects of the same
language. But this has not been shown to be so. The question that arises,
therefore, for the descriptive semanticist is whether he can or should
take account of the native speaker’s intuitions of relatedness of meaning
in deciding between polysemy and homonymy. Until it has been
demonstrated that intuitions of this kind correlate with empirically
decidable differences in the use of words, the linguist might well decide
that it 1s preferable to leave the theoretical status of the distinction
between homonymy and polysemy unresolved; and this is what we
propose to do in what follows.
Attempts have also been made to explicate the notion of relatedness
 13.4. Homonymy and polysemy 553
of meaning in terms of a componential analysis of the senses of lexemes
(cf. 9.9). But it is fair to say, without prejudice to the possibility that
this approach to the question will ultimately prove viable, that all such
attempts have so far failed. It is easy enough of course to select particular
lexemes, like ‘bachelor’ for example, which standard dictionaries treat
as being polysemous and to interrelate their senses in terms of a set of
semantic components. The problem lies in justifying the componential
analysis of sense for the vocabulary as a whole and showing how it can
be used, in other than a few relatively clear-cut examples, to separate
homonyms from single polysemous lexemes. ‘The componential approach
to semantics, as we have seen, is one that commends itself to many
linguists on general grounds; but it is fraught with serious theoretical
and methodological difficulties. As far as the present issue is concerned,
the possibility or impossibility of decomposing the senses of lexemes
into a (structured or unstructured) set of semantic components is
irrelevant, unless we can specify just how many components, or alter-
natively what kind of components, two senses must share in order for
them to meet the criterion of relatedness of meaning. Should we say, for
example, that the two senses must have z components in common? And,
if so, do we assign some fixed numerical value to 2 or do we make the
value of ” proportional to the total number of components recognized
for both senses? Or should we weigh the totality of semantic components
throughout the vocabulary as a whole, discounting for the purpose of
establishing the requisite degree of relatedness of meaning such very
general components as ANIMATE or PHYSICAL OBJECT and giving greater
weight to such components aS ADULT or MARRIED? Until such questions
have been answered and whatever measure of semantic relatedness is
selected has been shown to produce results that are consistent with
native speakers’ judgements, in cases where there is a consensus of
opinion among them, we must treat with caution any suggestion that the
technique of lexical decomposition, upon which the componential
analysis of the meaning of lexemes is based, provides us, even in
principle, with a decision procedure for distinguishing between poly-
semy and homonymy.
There are two possible ways of circumventing, rather than solving, the
problem of drawing a sharp distinction between polysemy and homo-
nymy in the analysis of particular language-systems: one is to maximize
homonymy by associating a separate lexeme with every distinct meaning
(‘mouth,’, ‘mouth,’, ‘mouth,’, etc.: ‘ear,’, ‘ear,’, ‘ear,’, etc.); the
other is to define the lexeme solely in terms of its associated forms and
554 The Lexicon
their syntactic function. If we adopt the first approach, which has been
proposed by certain linguists, we will end up with many more lexical
entries than are recognized in the standard dictionaries of the language
we are describing. This is not in itself a very damaging criticism. But
many of these entries will duplicate the phonological and grammatical
information that is contained in other entries. For example, on the
perhaps questionable assumption that there are four senses associated
with the lexeme(s) of which bachelor and bachelors are forms, we will list
four distinct lexemes (‘bachelor,’, ‘bachelor,’, ‘bachelor,’, ‘bachelor,’)
in the lexicon (cf. Weinreich, 1966; McCawley, 1968); and we will
encode the information, in whatever formalism is being used for the
analysis, that each of the lexemes is a countable noun whose stem is
bachelor and that it is morphologically regular (taking the suffix -s in the
plural). ‘The methodological maximization of homonymy will therefore
lead to considerable redundancy in the dictionary (cf. Hudson, 1976).
More serious, however, is the fact that distinctions of sense can be
multiplied indefinitely. Does ‘mouth’ have the same meaning in ‘the
mouth of a river’, for example, as it has in ‘the mouth of the tunnel’ or
‘the mouth of the jar’? Does the verb ‘play’ have the same sense in such
utterances as the following?
(1) She plays chess better than she plays the flute
(2) He’s never played Hamlet
(3) L’m playing scrum-half next Saturday
(4) Can I go out to play now, Mummy?
How indeed can we decide such questions? As we saw in an earlier
section, ambiguity tests based on co-ordination are of limited applic-
ability (cf. 10.4). Could we delete the second occurrence of the form
plays in (1)? And what about (5)?
(5) He played scrum-half in the afternoon and Hamlet in the evening.
It may well be that the whole notion of discrete lexical senses is ill-
founded; and, if it is, there is no hope of defining lexemes on this basis.
The equally radical alternative is to maximize polysemy. This will
have the effect of producing a lexicon with far fewer entries than are to
be found in our standard dictionaries. But there is little doubt that, on
methodological grounds, if for no other reason, it is preferable. Same-
ness and difference of form (in either the phonic or the graphic medium:
cf. 3.3) is, in general, something that is readily decided (by virtue of the
design-features of duality, discreteness and arbitrariness: cf. 3.4); and
 13.4. Homonymy and polysemy 555
formal identity is involved anyway, whether we set out to maximize
either homonymy or polysemy. Let us therefore attempt to make more
precise than we have done so far the formal and syntactic criteria for
identifying word-lexemes. Other kinds of lexemes will be left out of
account for the present.
We will assume, for simplicity of exposition, that the language-system
is one for which the distinction between morphology and syntax is
justifiable and, with it, the distinction between word-forms and the
morphosyntactic words that they realize (cf. 10.1). Nothing of conse-
quence, however, depends upon this assumption. We will also assume ,
that the lexemes of the language are to be assigned to parts-of-speech
(nouns, verbs, adjectives, etc.) and to various subclasses of the parts-of-
speech in terms of such distinctions as proper vs. common and countable
vs. mass (for nouns), transitive vs. intransitive (for verbs), and so on (cf.
11.1). Given such a classification for all the lexemes in the language-
system, we will say of any two lexemes, Li and L;, that they are syn-
tactically equivalent (Li = L;) if and only if they belong to exactly the
same subclasses. This definition, it should be noted, does not rest upon
the presupposition that there are in fact any syntactically equivalent, but
distinct, lexemes in any particular language-system. Every lexeme is
syntactically equivalent with itself: lexemic identity (Li = L;) implies
syntactic equivalence (Li = Lj). This proposition we will take to be
axiomatic under any formalization of the relationship between the
lexical and the grammatical structure of languages. The converse pro-
position, that Li = L; implies L; = Lj, is one that we will discuss
presently in connexion with the notion of homonymy.
The notion of syntactic equivalence, as we have just defined it, is a
particularly strong notion (n.b. “exactly the same subclasses”’). It can
be weakened, however, by relativizing it to any subset of the syntactic-
ally relevant distinctions in terms of which lexemes are subclassified. For
it is an important fact about the syntactic subclassification of lexemes
that the syntactic properties customarily recognized by linguists are to a
considerable degree independent of one another. Each lexeme must be
cross-classified in terms of its membership of several intersecting sub-
classes. It is in principle possible, therefore, that, in terms of their syn-
tactic subclassification in some particular model of the language-system,
no two distinct lexemes (their distinctness being established on other
grounds) will satisfy our definition of syntactic equivalence. We might
still wish to say, however, that they are syntactically equivalent with
respect to such-and-such a subset of syntactic properties.
556 The Lexicon
To say that two lexemes are syntactically equivalent implies that they
are intersubstitutable throughout the grammatically well-formed sen-
tences of the language: i.e. that they have the same distribution*. Since
lexemes as such do not occur in sentences, the substitution of one lexeme
for another must be understood in the sense of substituting a form of one
lexeme for a form of another lexeme: e.g., went (a form of ‘go’) for came
(a form of ‘come’), or conversely, in such sentences as ‘He came/went
home last night’. The distribution of word-forms throughout the gram-
matically well-formed sentences of a language (and hence the inter-
substitutability of any pair of word-forms in any subset of these sen-
tences) is only partly determined by the syntactic classification of the
lexemes of which they are forms. The distribution of a word-form is also
determined, in general, by the inflexional properties of the morpho-
syntactic word (or words) that it realizes. For example, came is the past-
tense form of the intransitive verb ‘come’, as went is the past-tense form
of the intransitive verb ‘go’. These two forms are intersubstitutable in
any position that the syntactic rules define to be a possible environment
for intransitive verbs in the past tense. Now, intransitivity (in this sense
of the term) is syntactically, but not morphologically, relevant in
English: it is a property of lexemes (with respect to which ‘came’ and
‘go’ are syntactically equivalent), Past tense, on the other hand, is both
morphologically and syntactically relevant: the past-tense form of
‘come’ is distinct from the two present-tense forms come and comes; and
the three forms differ in their distribution throughout the well-formed
sentences of English. Past tense and present tense are inflexional, rather
than lexical, properties.
When we say that came is the past-tense form of the verb ‘come’,
what we mean is that it is associated by the morphological rules (operat-
ing upon whatever information is given in the lexicon: cf. 13.1) with a
morphosyntactic word which has the inflexional property of past tense.
Every morphosyntactic word is made up of two parts: a lexical com-
ponent (which, in many languages, is realized by a root* or stem*) and a
set of inflexional properties. Given that this is so, we can define two
kinds of partial identity between morphosyntactic words: lexical and
inflexional. ‘Two distinct morphosyntactic words, W; and Wj, are
lexically identical if they share the same lexical component; and they are
inflexionally identical if they have exactly the same inflexional proper-
ties. (If they are both lexically and inflexionally identical, they are not
distinct morphosyntactic words: i.e. Wi = W;.) These partial identities,
lexical and inflexional, are relations which hold, it should be empha-
 13.4. Homonymy and polysemy 557
sized, between morphosyntactic words, and not between forms. But
they are defined in terms of properties which are postulated in order to
account, ultimately, for the distribution of word-forms.
For the definition of the traditional notion of homonymy, we shall
need to invoke a particular kind of grammatical equivalence, which is
based partly upon the relation of syntactic equivalence holding between
lexemes and partly upon inflexional identity. Let us say that two
morphosyntactic words, W; and Wj, are grammatically equivalent (i.e.
Wi = W)j) if and only if:
(i) the lexemes with which they are associated, Li and Lj, are syn-
tactically equivalent (Li = Lj); and
(11) the morphosyntactic words in question, W; and W,, are inflexion-
ally identical.
The first of these conditions may be referred to as a condition of lexical
- equivalence (of which lexical identity is a special case). Grammatical
equivalence defined in this way as a relation between morphosyntactic
words is a sufficient, but not necessary, condition for the syntactic
equivalence of the word-forms that realize the morphosyntactic words in
question. Suppose, for the sake of the argument, that there was no verbal
or adjectival concord in English, but that there was nonetheless a dis-
tinction of singular and plural manifest in the forms of countable nouns,
such that sentences like ‘This boy is our friend’ (“This boy is our
friend’’) and ‘'This boys is our friends’ (‘‘’T’hese boys are our friends’’)
were well-formed. The distribution of the singular form boy would now
be identical with the distribution of the plural form boys. But the
morphosyntactic words realized by boy and boys would still be distinct
in terms of the inflexional property of number. Generally speaking,
inflexionally distinct morphosyntactic words will differ in distribution.
But we must allow for the possibility that they will not do so in every
particular case.
There is one further criterion that is involved in the definition of
homonymy: that of formal identity. Two word-tokens are formally
identical in the phonic medium if they have the same phonological
representation. They are formally identical in the graphic medium if
they have the same orthographic representation. In languages that are
conventionally written with an alphabetic or syllabic orthography, both
kinds of formal identity generally coincide. But they are in principle
completely independent of one another (cf. 3.3).
We can now extend this notion of formal identity to lexemes and to
 558 The Lexicon
morphosyntactic words. Every lexeme is associated with a set of forms.
Let us say therefore that two lexemes, Li and Lj, are formally identical
if and only if each 1s associated by the rules of the language-system with
the same set of forms: Li* = L;* (where Li* 1s the set of forms associa-
ted with Li, L;* the set of forms associated with Lj, and so on). The set
of forms associated with a lexeme may be a one-member set. But
typically in languages for which the distinction between morphology and
syntax is justifiable, it will have more than one member; and it is for this
reason, as we shall see, that the formal identity and the syntactic equiva-
lence of lexemes are necessary, but not sufficient, conditions for
homonymy.
We will define the formal identity of morphosyntactic words some-
what differently. Given that W = {W,, W., ..., Wm} is the set of
morphosyntactic words and F = {F,, F,, ..., Fn} the set of morpho-
logically simple or complex forms, we will say that W; is formally
identical with Wj, if and only if they are realized by the same form (in
either medium: cf. 3.3). The rules of the language-system will map the
set W into the set F and, in doing so, will establish a relation of formal
, identity between the members of particular subsets of W. In English (if
we discount the relatively few cases in which there are alternative
realizations, in the same dialect, of the same morphosyntactic word: cf.
dreamt|/dreamed, etc.), the relationship between the members of W and
the members of F is a many—one correspondence; and this is commonly
the situation in languages. For example, the past-tense form of the
modal verb ‘will’ is formally identical, in the phonic medium, with the
singular form of the noun ‘wood’, the past-tense form of the verb ‘read’
is formally identical, in the graphic medium (but not in the phonic
medium), with the present-tense form of the same verb; the past-tense
form and the past participle are formally identical in both the phonic and
the graphic medium, for all the so-called regular (or weak) verbs in
English.
The formal identity of lexemes and of morphosyntactic words (like
the type-token identity of forms, upon which it rests) is a medium-
dependent notion; and it is for this reason that a distinction 1s tra-
ditionally drawn between homophones* (like the verbs ‘sow’ and ‘sew’
in English) and homonyms. The usage of the term ‘homophony’ 1s, if
anything, less consistent in the literature than 1s that of “homonymy’;
and the parallel term ‘homography’ is only rarely employed. If we wish
to be precise in our definition of homophony* and homography* (within
the more or less traditional framework with which we are operating),
 13.4. Homonymy and polysemy 559
there are several ways of restricting their application. We could define
them as medium-dependent relations of identity holding between forms
(type-token identity): in which case, we would say that the past-tense
form of the verb ‘read’ is homographic (but not homophonous) with one
of the present-tense forms of the same verb (read = read, but /red/ +
/ri:d/); that the simple form of the adjective ‘red’ is homophonous (but
not homographic) with the past-tense form of the verb ‘read’ (/red/ =
/red/, but red 4 read); and that the present-tense form of the modal
verb ‘can,’ (“‘be able’’) is both homographic and homophonous with
one of the present-tense forms of the transitive verb ‘can,’ (“put into a
can’’), Alternatively, we could define homophony and homography as
medium-dependent relations holding between morphosyntactic words,
rather than forms. In this case, the definitions might be as follows: -
Given that W is the set of morphosyntactic words, that F is the set of
phonologically represented forms and G the set of graphically represen-
ted forms, W; is homophonous with W, if and only if Wi and Wj are
realized by the same form Fx and W; is homographic with W, if and
only if W; and Wj are realized by the same form Gi. It will be obvious
that there is little practical difference between defining homophony and
homography as relations which hold between forms or relations which
hold between morphosyntactic words.
There is, however, another way of defining homophony and homo-
graphy (which ties these notions to the traditional notion of homonymy);
and this to define them as relations between lexemes. When the term
‘homophony’ is used in traditional grammar with respect to what are
taken to be distinct lexemes, e.g. ‘sew’ and ‘sow’, it stands implicitly, if
not explicitly, in contrast with ‘homonymy’. The forms of ‘sew’ and
‘sow’ are spelled differently, but have the same pronunciation; and this
is considered to be something exceptional, which is worthy of termino-
logical recognition. If the forms of ‘sew’ and ‘sow’ had both the same
spelling and the same pronunciation, but were on other grounds taken
to be different lexemes, they would normally be described as homonyms,
rather than homophones. Clearly, it is simply a matter of terminological
convenience whether we decide to restrict the application of the term
‘homophony’ in this way or not; and the same goes for ‘homography’.
What we propose to do, in fact, is to begin by defining a notion of
absolute, or complete, homonymy, as a relation between lexemes, and
then to distinguish homophony and homography as two kinds of partial
homonymy. This is in accord with traditional practice, though it is in-
consistent with other definitions that will be found in the literature.
560 The Lexicon
At least the following three conditions are necessary for the absolute
homonymy of two lexemes, Li and L;;:
(i) Li ~ Ly; (lexemic distinctness),
(ii) Li = Lj (syntactic equivalence),
(iii) Ly* = Lj* (formal identity).
Each of these requires some comment; and we shall see presently that
the conditions of formal identity and syntactic equivalence are not
strong enough for what is required of them. Most, if not all, of the kinds
of partial homonymy that are actually found in languages can be defined,
however, in terms of some qualification of conditions (ii) and (iii), each of
which is independent of the other. It is convenient, therefore, to start with
a set of conditions that are severally necessary, if not jointly sufficient.
Little need be said about lexemic distinctness. We could obviously
drop this condition, if we wished to say that every lexeme is an absolute
homonym of itself. But this seems rather pointless; and it would be
perverse to allow for the possibility that a lexeme may or may not be
absolutely homonymous with itself. At the same time, it must be
emphasized that all the other conditions for absolute homonymy that we
shall be considering are also conditions for lexemic identity. It is an open
question, therefore, whether L; 4 Lj ever holds when all the other
conditions are satisfied. The maximization of polysemy, at the expense of
homonymy, which we referred to earlier as a defensible methodological
principle, would lead us to say, as we shall see, that there are no absolute
homonyms in language, but only various kinds of partial homonymy.
The condition of syntactic equivalence has already been discussed; and
it has been pointed out that this is relative to a particular analysis of the
language-system and a matter of degree. At the grossest level of classifi-
cation, we assume, lexemes will be distinguished according to the part-
of-speech they belong to. Independently of any other consideration,
therefore, no two lexemes can be absolutely homonymous if they are
members of different parts-of-speech; and this is widely, if not univer-
sally, taken to be the case in traditional treatments of homonymy.
Problems begin to arise, however, when we start taking into account
finer syntactic distinctions. The nouns ‘port,’ (“‘harbour’’) and ‘port,’
(‘kind of fortified wine’’), for example, differ in that the former is a
countable, and the latter a mass, noun. (The fact that ‘port,’ can be
used, in expressions like ‘two ports and a madeira’, as a countable noun,
is covered by a very general principle which, we assume, need not be
accounted for in the lexicon: cf, 11.4.) These two lexemes, therefore, are
 13.4. Homonymy and polysemy 561
only partially homonymous; and the same is true of such pairs of transi-
tive and intransitive verbs as ‘run,’ and ‘run,’, ‘move,’ and ‘move,’,
etc. At this point, we are already classifying as distinct lexemes many
nouns and verbs which would be identified as the same lexeme in con-
ventional dictionaries. In doing so, we bring them within the scope of
the condition of lexemic distinctness (and thereby rule out the applic-
ability of ‘polysemy’); simultaneously, however, we classify them in
such a way that they fail to satisfy one of the conditions for absolute
homonymy. It is easy to see that a finer and finer subclassification of
lexemes will lead to the recognition of more and more distinct lexemes.
For example, ‘realize,’ (in ‘He realized his assets’) might be distin-
guished in the lexicon from ‘realize,’ (in ‘He realized that he was mis-
taken’). Which of these two lexemically distinct verbs is it then which
occurs in a sentence like ‘He realized his mistake’? It is obviously
related more closely to ‘realize,’ than it is to ‘realize,’ in terms of its
meaning; and it may be identified with ‘realize,’ under a particular
transformational analysis of English. The point that is being made here
is simply that absolute homonymy, as we have defined it, is relative to a
certain syntactic analysis of the language-system, because the notion of
syntactic equivalence is itself relative, explicitly or implicitly, to a
particular set of rules. It might well turn out to be the case that, under a
very comprehensive sub-classification of lexemes, no two distinct
lexemes are syntactically equivalent (cf. Gross, 1975). However that may
be, much of what is traditionally considered to be homonymy is only
partial homonymy with respect to the condition of syntactic equivalence.
Let us now look briefly at the condition of formal identity. Consider
the following sets of word-forms in English: X = {can, could}, Y =
{can, cans}, Z = {can, cans, canning, canned}. X will be associated, we
assume, with the modal verb ‘can,’ (“be able’’), Y with the noun ‘can,’
(‘‘a kind of receptacle’’) and Z with the transitive verb ‘can,’ (“put in
a tin/can’’). The three lexemes in question would of course be dis-
tinguished in terms of their syntactic non-equivalence; and it is generally
the case in English, as it is in most, if not all, languages with inflected
word-forms, that the syntactic non-equivalence of lexemes implies their
formal non-identity. But this is clearly a matter of empirical fact. Sup-
pose, for example, there were a language in which the noun-stems and the
verb-stems took exactly the same set of inflexional affixes. Two syn-
tactically non-equivalent lexemes might well be formally identical; and
the fact that this was so would not necessarily produce any instances of
grammatical ambiguity. Partial formal identity of lexemes, as exemplified
562 The Lexicon
by ‘can,’ vs. ‘can,’ vs. ‘can,’ above, is by no means uncommon
in languages; and it gives rise to different kinds of partial homonymy, as
we shall see presently, according to whether the lexically distinct
morphosyntactic words are grammatically equivalent or not. But first we
will give an example from English of two lexemes, which might reason-
ably be held to be syntactically equivalent, but which (in some dialects
at least) are formally non-identical; these are ‘hang,’ (‘‘suspend”’) and
‘hang,’ (“‘execute by suspending by the neck’’). These two lexemes
differ formally (in certain dialects) in that ‘hang,’ is associated with the
set {hang, hangs, hanging, hung} and ‘hang,’ with the set {hang, hangs,
hanging, hanged}. The forms hung and hanged, it will be noted, each
realize two distinct morphosyntactic words (on the assumption that
‘hang,’ is syntactically equivalent to ‘hang,’); and we have two in-
stances of grammatical equivalence.
Enough has been said to demonstrate the theoretical independence of
syntactic equivalence and formal identity and their relevance in the
definition of homonymy as a relation between lexemes. But the reader
will have realized that we have deliberately not taken account, in the
immediately preceding discussion, of the fact that formal identity is a
medium-dependent notion. This deficiency is easily remedied. If we are
concerned solely with either the written language or the spoken language,
we will interpret the condition of formal identity accordingly. But, if we
wish to make our description of the language neutral, as far as possible,
with respect to medium-dependent distinctions and identities, we will
take condition (ili) to imply formal identity in both mediums. Lexemes
will be absolutely homonymous if they satisfy all the other conditions
and are formally identical in both the phonic and the graphic medium.
We can then restrict the application of the terms ‘homophony’ and
‘homography’ (much as the term ‘homophony’, but not ‘homography’,
is commonly restricted in practice) to cases where there is a lack of 1so-
morphism, in this respect, between the written and the spoken language.
This is simply a matter of terminological convenience. We could just as
well look upon homophony and homography as independent, but com-
patible, notions.
It is obvious that, just as homonymy can be absolute or partial, so too
can homophony and homography. An example of absolute homophony
has already been given: ‘sew’ and ‘sow’ (on the assumption that they
are syntactically equivalent) are absolute homophones that are not even
partial homographs. Partial homophony (of various kinds) is found in
many languages. Examples from English are ‘read’ vs. ‘red’, ‘great’ vs.
 13.4. Homonymy and polysemy 563
‘grate,’ and ‘grate,’ (which are themselves partial homonyms), etc. In
fact, most of the lexemes which are traditionally described as homo-
phones in English are only partial homophones. Partial homography is
also quite common in English. For example, the verbs ‘put’ and ‘putt’
(“hit a golf-ball in a certain way’’) are syntactically equivalent in so far
as they are transitive and both take locative complements (the one
obligatorily and the other optionally): cf. the ambiguous utterance He 1s
putting the ball on the green. But they differ syntactically in other ways.
And their formal identity in the graphic medium is only partial: cf.
{put, puts, putting} vs. {putt, putts, putting, putteds. The utterance He is
putting the ball on the green is therefore lexically (and perhaps gram-
matically) ambiguous in the written language (cf. 10.4). But in the spoken
language (in those accents of English in which the vowels of butcher and
butter are phonologically distinct), the written utterance-type He 1s
putting the ball on the green would be in correspondence with two
phonologically distinct utterance-types; and the past-tense He put the
ball on the green differs from He putted the ball on the green in both the
graphic and the phonic medium. Other examples have been given earlier
in this section. Absolute homography may not exist at all in languages
that are conventionally represented, in the written medium, by means of
an alphabetic or syllabic system of writing.
It now remains to show that syntactic equivalence and formal identity
are not strong enough to serve jointly, with lexemic distinctness, as
sufficient conditions for absolute homonymy (or indeed for absolute
homography and homophony). We will demonstrate that this is so by
means of a simple hypothetical example. Let us take one of the most
commonly cited instances of what is generally regarded as homonymy in
English: ‘bank,’ (‘‘side of a river”) and ‘bank,’ (‘‘financial institu-
tion’). We will assume that these are syntactically equivalent (though
this assumption might be challenged). That they are formally identical
in both the phonic and the graphic medium is indisputable. Whether
they are absolute homonyms or not depends, therefore, upon the condi-
tion of lexemic distinctness. Let us leave this question (homonymy vs.
polysemy) unresolved. Instead we will assume, for the sake of the
argument, that, in a language otherwise identical with English, whereas
the singular and plural forms of ‘bank,’ are bank and banks (as they are
in English), the singular and plural forms of ‘bank,’ are banks and bank,
respectively. The two lexemes ‘bank,’ and ‘bank,’ are, under the
assumptions we have made, both syntactically equivalent and formally
identical. But we should certainly not wish to call them absolute
564 The Lexicon
homonyms; and the reason is obvious. Expressions like ‘this bank’ and
‘these banks’ (not to mention ‘this banks’ and ‘these bank’) would not
be ambiguous; nor would sentences like ‘The bank is/are invisible from
here’. Furthermore, if we were confronted with an actual case of the
kind we have envisaged, we would definitely regard the two lexemes as
distinct. The question of homonymy vs. polysemy would simply not
arise, any more than it arises with respect to syntactically non-equivalent
or formally non-identical lexemes in any of the standard treatments of
this topic.
What is at issue is the grammatical equivalence of morphosyntactic
words as we defined this earlier. If we combine this condition (which
subsumes the syntactic equivalence of lexemes) with formal identity, we
will have a sufficiently strong definition of absolute homonymy; and this
definition will exclude cases like our hypothetical ‘bank,’ and ‘bank,’
(if there are any such cases in any natural language). Let us therefore
substitute, for the conditions given above, the following:
(i) Li 4 L; (lexemic distinctness)
(ii) Li* = Lj* (formal identity)
(111) (x, y) (we Li® & ye Li* &w = y & R(x, Wi) & Ry, Wj)) >
(W; = Wj) (grammatical equivalence)
Condition (ii), in which R symbolizes the relationship of realization
that holds between a form and a morphosyntactic word, as will be clear
from the preceding discussion, still requires some amplification and
emendation since it presupposes a one-to-one correspondence between
forms and the morphosyntactic words they realize. Moreover, it does not
restrict the values of Wi and Wj; to morphosyntactic words associated,
respectively, with Li and Lj. Provided that we make these necessary
adjustments, it will serve our purpose; and it is perhaps easier to see its
import, if we leave it as it stands. All the requisite notions have been
introduced informally above and they have been satisfactorily formalized
elsewhere (cf. Matthews, 1967).
Conditions (i)—(iii), interpreted as we have suggested, would seem to
be both necessary and sufficient, taken jointly, to define the relation of
absolute homonymy; and it 1s arguable that, in formulating them more
or less precisely, we have done no more than make explicit what is
implicit in conventional lexicographical practice, on the one hand, and
traditional discussions of homonymy, on the other. What we have not
done, however, is to specify what degree of syntactic equivalence
between lexemes (and consequently what degree of grammatical equiva-
 13.4. Homonymy and polysemy 565
lence between morphosyntactic words) is necessary or sufficient for
absolute homonymy. If the condition Li = Ly, is taken in its strongest
possible sense, relative to a generative grammar which draws a large
number of syntactically relevant distinctions, there will be fewer in-
stances both of absolute homonymy and polysemy than are recognized
in the standard dictionaries of English and other languages. The reason
that this is so is that, as was pointed out earlier, the syntactic equivalence
of Li and L; is normally taken to be a condition, not only of homonymy,
but also of lexemic identity. If Li ~ Lj, the question of polysemy does
not arise; and if L; 4 Lj, Li and Lj cannot be absolute homonyms in
terms of our definition.
In traditional lexicographical practice, it is a comparatively weak
notion of syntactic identity that is invoked in deciding whether L; and
L; are identical or not. Usually no more is taken into account than the
classification of lexemes in terms of the part-of-speech to which they
belong. If Li and LL; are formally identical and if both are nouns, verbs,
adjectives, etc., they will normally be handled within the same lexical
entry, provided that they are not distinguished on etymological or
semantic grounds as homonyms. It is frequently the case, however, that
distinctions of sense correlate, on the one hand, with syntactically rele-
vant differences and, on the other, with such non-inflexional morpho-
logical differences as are customarily described in terms of word-
formation* (or derivation*) (cf. 13.2).
For example, what would normally be classified as a single verb,
‘act’, with transitive and intransitive uses is morphologically related to
the noun ‘actor’. But the sense of ‘actor’ is more restricted than that of
the verb ‘act’ considered as a single lexeme. ‘Actor’ means, roughly,
‘““someone who plays a role’’; and ‘‘play a role”’ is one of the senses of
the verb ‘act’. Furthermore it is only when it is used in this sense that
‘act’ may be used transitively: cf. “He acted the part superbly’. If two _
different lexical entries are put into the dictionary, one for ‘act,’ (‘“‘play
a role’) and for ‘act,’ (“behave”’), we can relate the noun ‘actor’
systematically to ‘act,’, instead of cross-referencing it, as is usually done
in conventional dictionaries, to one of the senses of the lexically un-
differentiated verb ‘act’; and we can restrict the syntactic property of
transitivity to ‘act,’. (It is in fact a particular kind of transitivity. The
verb ‘act,’ takes what is traditionally described as a cognate object, as do
such verbs as ‘sing’, ‘play’, etc.: cf. ‘She acted the part of Ophelia’,
‘She sang an aria’.) The ambiguity of ‘He acts well’, in contrast with
the non-ambiguous ‘He is a good actor’, will then be accounted for in
566 The Lexicon
terms of partial homonymy, rather than polysemy. Considerations of
this kind would probably lead us to recognize other verbs which are
partial homonyms of ‘act,’ (‘‘play a role’’) and ‘act,’ (“‘behave”’). There
are syntactic differences to support further subclassification; and there
are the nouns ‘act’ and ‘action’ to be accounted for (not to mention the
derivationally irregular ‘activity’). But we will not pursue the example
in detail. The important point is that it is by no means untypical.
Earlier in this section, it was proposed that the maximization of
polysemy might be accepted as a methodological principle. It will now
be clear that, in so far as polysemy differs from absolute homonymy
solely in respect of condition (i), what is meant by maximizing polysemy
(relative to a particular grammatical analysis) is making it a matter of
methodological decision that conditions (ii) and (iii) jointly imply the
negation of condition (i). And this would eliminate in principle all cases
of absolute homonymy.
Many linguists would reject the proposal that polysemy should be
maximized. They would say that, however difficult it might be to for-
malize the notion of relatedness of meaning, upon which the distinction
of homonymy and polysemy depends, the criterion of relatedness of
meaning is one that native speakers draw upon in their intuitive judge-
ments of what constitutes lexical identity; and this is undoubtedly true.
However, the obliteration, by methodological fiat, of the whole basis for
the distinction between polysemy and homonymy has the advantage that
it is more readily applicable (by virtue of the design-features of discrete-
ness and duality) than is the alternative principle of admitting absolute
homonymy defined in terms of some as yet unexplicated global concept
of unrelatedness of sense.
It might, however, be argued in a more positive vein that, since the
ability to extend the sense and denotation of lexemes by a process of
metaphorical transfer is an integral part of every speaker’s linguistic
competence and is demonstrably involved in the child’s acquisition of
his language, our methodological principle does no more than give
recognition to something that is of central importance in language-
behaviour. It has already been mentioned that speakers of a language will
often look for, and discover, a metaphorical connexion between what
are, by the synchronically irrelevant etymological criterion, quite clearly
homonyms: cf. ‘ear,’ and ‘ear,’. That they should do this is 1n part,
presumably, a consequence of their intuitive appreciation of the syn-
chronic importance of metaphorical extension; and that they should
frequently disagree among themselves about the nature of the putative
 13.4. Homonymy and polysemy 567
metaphorical connexion is only to be expected. For they will also dis-
agree about the precise interpretation of many so-called figurative
expressions. The methodological elimination of absolute homonymy
could perhaps be regarded, therefore, in so far as it has any psychological
implications at all, as an idealization of metaphorical creativity. However
that may be, it is important to realize that, even if we grant that there is a
pre-theoretically valid distinction to be drawn between polysemy and
absolute homonymy, clear instances of the former are far more numerous
in English, and no doubt in all languages, than are clear instances of the
latter, no matter how stringent we are in the application of condition
(iii). Furthermore, polysemy — the product of metaphorical creativity —
is essential to the functioning of languages as flexible and efficient
semiotic systems. Homonymy, whether complete or partial, is not.
The criteria that we have outlined and exemplified for identifying L;
and L; as the same lexeme are essentially distributional criteria. It must
be clearly understood that they are not intended to provide the linguist
with a procedure for grouping word-forms into sets, X = {x,,..., Xn},
Y = {1 Voy - ++ Yn}, etc., such that, independently of other considera-
tions, the members of X can be said to be forms of Lj and the members
of Y to be forms of L;. The question we have been concerned with is
whether Lj is identical, or not, with Lj. We are assuming that the assign-
ment of forms to L; and L; has already been made, together with the
syntactic subclassification of Li and L;, in some generative grammar of |
the language-system.
It should also be emphasized that the proposed criteria do not pre-
suppose that any or all of the forms of a lexeme (if it has more than one
form) should be morphologically related. We must obviously allow for
suppletion*. But it may be assumed that suppletion, in so far as it exists
in languages, 1s something exceptional. It has already been suggested
that it might be impossible to group better and best with good and worse
and worst with bad, without taking into account the meaning of ‘good’
and ‘bad’ (cf. 13.1). Although these particular instances of suppletion
are not especially troublesome in so far as the assignment of good, better
and best to one lexeme and bad, worse and worst to another lexeme is
concerned, it is obvious that suppletion as such constitutes a problem;
and there would be little point in trying to define the lexeme generally
in terms of the distribution of its forms throughout the sentences of the
language, if it were not possible to do this in most instances without
appealing to the criterion of relatedness of meaning.
Our discussion throughout this section has been in terms of a fairly
568 The Lexicon
traditional conception of the grammatical structure of languages for
which the distinction of morphology and syntax is justifiable. The
definition of lexical identity in terms of conditions (ii) and (iii) is none-
theless applicable to languages (like Vietnamese or Classical Chinese) of
the so-called isolating* type. In such cases, of course, condition (ii)
would hold over one-member sets, and the syntactic equivalence of
lexemes would be sufficiently strong as condition (iii). For, in isolating
languages (which to the degree that they are of this type have no in-
flexional variation) there is no point in drawing a distinction between the
morphosyntactic word and the lexeme with which it is associated. The
distinction between forms and lexemes, however, is no less important in
isolating languages than it is in so-called inflecting and agglutinating
languages (like Latin and ‘Turkish, respectively).
Nor should it be thought that our treatment of homonymy and lexem ic
identity stands or falls according to whether the morphosyntactic word
is recognized as a unit at some particular level of analysis in a generative
grammar of English and other languages. The fact that current versions
of generative grammar operate directly with forms and do not postulate
the occurrence of morphosyntactic words in the surface structure of
sentences is irrelevant to the validity of the various distinctions that we
have been drawing. The problems of lexemic identification are essen-
tially the same, whatever theoretical and terminological framework we
adopt. If morphology is not recognized as a distinct level of analysis,
what we have discussed in terms of morphological differences between
lexemes must be accounted for in either phonology or syntax. As we
have already seen in our discussion of lexical entries, for certain lan-
guages at least there is good reason to distinguish systematically between
the phonological, morphological and syntactic information that must be
included in the lexical entry (cf. 13.1).
Nothing has been said about the possible identity of the roots or
stems of word-forms associated with different lexemes; and little, in fact,
need be said, except that it is only indirectly relevant to the notion of
homonymy. It so happens that in some languages, but not others, the
stem-form of the lexeme may be itself a word-form. This is so for all
nouns and verbs in English. For example, the form girl is both the stem-
form, to which the pluralizing suffix -s may be added to yield girls, and
it is also the singular form of ‘girl’; the form come is the stem-form to
which -s and -zng are added to yield the third-person singular present-
tense form comes and the present participle coming, and it is itself a
present-tense form; and so on. But this does not mean that, even in
 13.4. Homonymy and polysemy 569
English, the lexeme can be identified with its stem-form: itis the formal -
identity or non-identity of lexemes that is important. ‘I'wo lexemes.
whether they are syntactically equivalent or not, may have the same
stem-form without having a single word-form in common: they may
belong to what are traditionally described as different conjugations and
declensions. Conversely, it is in principle possible for one lexeme to be
formally identical with another (and even for each of the morpho-
syntactic words associated with one lexeme to be inflexionally identical
with one, and only one, of the morphosyntactic words associated with
the other lexeme) without the two lexemes having the same stem. Con-
sider the following hypothetical example: Li* = {szmulat, simulatin}
and L,* = {stmulat, stmulatin}. L, we will assume is a transitive verb,
whose stem 1s mulat; and L; an intransitive verb, whose stem 1s silat.
Our reasons for analysing the forms in this way might be that transitive
verbs, in general, form their present-tense form by prefixing s7- and
intransitive verbs by infixing -mu-, and that both classes of verbs add the
further suffix -zm to form their past-tense forms. Although complete
formal identity between lexemes that do not have the same stem-form
may be rare, partial formal identity is not; and it falls within the scope
of our definition of partial homonymy.
In approaching the question of homonymy in the way that we have __.
done in this section, we have not been indulging in a pedantic taxonomic
exercise. Our purpose has been to elucidate, in relation to a theoretically
ideal notion of absolute homonymy, different kinds of partial homonymy.
The manner in which these are classified, and the terminology that is
used, 1s of secondary importance; and we have not in fact put forward a
set of terms to label the kinds of partial homonymy that we have
exemplified. There is no reason why the terms ‘homonymous’ and
‘“homographic’ should not be employed with respect to forms, morpho-
syntactic words or lexemes when it is convenient to do so, provided that
it is made clear what is being referred to. The importance of recognizing
different kinds of partial homonymy lies in the fact that it tends to pro-
' duce ambiguity (if it produces ambiguity at all) only in certain contexts;
and these contexts can be specified in terms of the structural analysis
that is assigned to sentences by the grammar that generates them. In so
far as partial homonymy creates ambiguity in sentences, it creates
ambiguity that is both lexical and grammatical (cf. 10.4); polysemy, on
the other hand, like absolute homonymy (if we admit its existence),
produces purely lexical ambiguities.
 14
Context, style and culture

14.1. The context-of-utterance

Any utterance-token that is produced on some particular occasion is an
actual utterance (cf. 1.6). In certain situations, the utterance that is
produced (as a token of a particular type) is very highly determined by
factors which we may describe, loosely for the moment, as contextual.
For example, the utterance of Hello when answering the telephone or of
Good morning upon entering a shop at a certain time of day is highly
determined by the social role that the utterer is playing and his recogni-
tion of what utterance-types are appropriate to this role and by a.
variety of more particular contextual features. Generally speaking, how-
ever, we can say that actual utterances are in contrast with indefinitely
many potential utterances which might have been actualized on the
occasion in question, but were not. |
Every actual utterance is spatiotemporally unique, being spoken or
written at a particular place and at a particular time; and, provided that
there is some standard system for identifying points in space and time,
we can, in principle, specify the actual spatiotemporal situation of any
utterance-act (which has as its product an actual utterance-signal: cf.
1.6) by giving its spatiotemporal co-ordinates within the framework of
the standard system. We can say, for example, that a particular utterance-
token was produced by X at 12 noon on 6 January 1971, in Edinburgh;
and we can be more or less precise than this in our specification of the
spatiotemporal co-ordinates of the utterance-act.
That languages provide the means, when this is necessary, of making
explicit reference to the time and place of utterance, as they also provide
the means of referring to events that are removed in space and time from
the actual situation of utterance, is an important fact; and we will come
to it presently. The spatiotemporal co-ordinates are, however, only one
part of the actual situation of utterance. Other components can also be
described in purely external observational terms (cf. 1.6): the appear-
ance, bearing and attitude of the various participants* in the language-
 14.1. The context-of-utterance 571
event (or language-process) of which the utterance in question is a
constitutive part; preceding, concomitant and subsequent activity; other
events taking place in the vicinity; and so on. Not all the observable, or
observationally salient, components of the actual situation of utterances
are linguistically relevant, and in some cases very few of them are. More-
over, the linguistic relevance of much of what is observable is apparent
only to those who are familiar with a given language-system and culture:
it becomes observationally salient, and is then describable in some
neutral metalanguage, by virtue of its linguistic and cultural rele-
vance.
This is an important point. It is not being denied that some correla-
tions between certain features of utterances and components of actual
situations are discoverable by external observers; nor indeed that some
utterance-tokens can be grouped, at least tentatively, into utterance-
types, and some actual situations into situation-types. Linguists and
anthropologists in the field may start by doing this. Subsequently, how-
ever, they work from within the culture, and more or less successfully in
proportion to their success in identifying the culturally and linguistically
relevant distinctions. Children acquiring their native language may also
begin by matching observationally identifiable components of utterances
and situations: this much of the behaviourist theory of semantics we have
granted as plausible, provided that it is combined with the postulation of
a richer set of innate propensities for cognitive development than the
radical behaviourist at least would normally postulate (cf. 5.4). But
no-one has yet justified a more extensive appeal than this to the matching
of utterances and situations. It is pointless, as far as descriptive semantics
is concerned, to argue whether it is in principle possible to discover all
the relevant correlations between bits of utterances and features of
situations by observation alone. What Chomsky (1957: 51) called dis-
covery procedures are no more attainable in semantics than they are in
grammar. It suffices that the descriptive semanticist can, if necessary,
identify and describe the correlations in terms of language-specific and
culture-specific distinctions.
It is intuitively obvious (and nothing but an empiricist bias would
prompt us to deny this) that there is considerable variation in the degree

1 Language-events (like other events, processes and states) are second-order

entities (cf. 11.3). What is here referred to as an event (or process) might also
be described, in so far as it is agent-controlled, as an act (or activity). The
participants in a language-event are performing both deictic roles and valency
roles (cf. 14.2, 15.1).
572 Context, style and culture
of interdependence holding between actual situations and actual utter-
ances. Most utterance-tokens, if not all, can be identified as tokens of a
given type independently of the actual situations in which they occur,
their identification being made on structural, rather than functional,
grounds (cf. 1.6). Tokens of the same utterance-type can occur in actual
situations which will be described as instances of quite distinct situation-
types. For example, It ts raining may be uttered in innumerable situa-
tions which have little in common. Conversely, and much more ob-
viously, many actual situations can be grouped into types independently
of the utterances which occur in them; and the utterances which occur
in the tokens of a particular situation-type may instantiate quite dif-
ferent utterance-types. To assume the contrary, as we saw in our dis-
cussion of behaviourist semantics, is both unnecessary and stultifying
(cf. 5.3).
We will make no further use of the pre-theoretical notion of the actual
situation of utterance. Nor will we go into the question of grouping
actual situations into situation-types; it may well be that this can be done
only in a relatively small number of cases. Of greater importance for the
semanticist is the theoretical notion of the context-of-utterance*.
Context, it must be emphasized, is a theoretical construct, in the
postulation of which the linguist abstracts from the actual situation and
establishes as contextual* all the factors which, by virtue of their in-
fluence upon the participants in the language-event, systematically
determine the form, the appropriateness or the meaning of utterances.
It is important to stress the qualifying term ‘systematically’. All random
variation is to be discounted in terms of the distinction of competence*
and performance* (cf. 1.6). The theoretical notion of the context-of-
utterance is based of course upon a pre-theoretical notion of context
(which is intuitive rather than observational: cf. 1.6) — a pre-theoretical
notion to which we constantly appeal in the everyday use of language.
Asked by a child or a foreigner what a particular word means, we are
frequently unable to answer his question without first getting him to
supply some information about the context in which he has encountered
the word in question. We will also say, pre-theoretically, that a certain
lexeme, expression or utterance is appropriate or inappropriate, or that
it is more or less effective than another, in a certain context. The prob-
lem is to explicate this pre-theoretical, intuitive, notion of context in a
theoretically satisfying way.
Many philosophers have said that context is a matter of pragmatics
rather than semantics. This, as we have seen, was the view that Carnap
 14.2. Communicative competence 573
took in his earlier work; but he always maintained that pragmatic con-
siderations were essential for the analysis of language (cf. 4.4). Among
linguists, two fairly extreme positions have been defended on this ques-
tion. At one extreme, Katz and Fodor (1963), though they did not deny
that contextual factors were relevant to the interpretation of actual
utterances, argued that descriptive semantics should be concerned with
the meaning of sentences considered independently of their utterance in
actual situations. At the other extreme, we find scholars like J. R. Firth,
who built up his whole theory of semantics upon the notion of context,
describing what he referred to as his “technique” for the analysis of
meaning in language as “‘a serial contextualization of our facts, context
within context, each one being a function, an organ of the bigger context
and all contexts finding a place in what might be called the context of
culture”’ (1935: 33). The views of Firth, and of others who have insisted
upon the necessity of incorporating the notion of context within seman-
tics, will be discussed in a separate section (14.4).

14.2. Communicative competence

One way of embarking upon the analysis of context is to ask what kinds
of knowledge a fluent speaker of a language must possess in order to
produce and understand contextually appropriate and comprehensible
utterances in that language. Hymes (1971), in an important and in-
fluential discussion of this subject, has introduced the term communica-
tive competence* to cover a person’s knowledge and ability to use all the
semiotic systems available to him as a member of a given socio-cultural
community. Linguistic competence, or knowledge of the language- __
system, is therefore but one part of communicative competence. Fur-
thermore, as we have already seen, much that is involved in language-
behaviour is excluded by methodological decision from the linguist’s
model of the language-system and is thereby defined as non-linguistic
(cf. 1.6, 3.1). What might be referred to as language-competence is
therefore broader than and includes linguistic competence.
Hymes (1971) raises four questions, which, he suggests, are relevant
for language and for other forms of communication: “‘1. Whether (and
to what degree) something is formally possible; 2. Whether (and to what
degree) something is feasible in virtue of the means of implementation
available; 3. Whether (and to what degree) something is appropriate
(adequate, happy, successful) in relation to a context in which it 1s used
and evaluated; 4. Whether (and to what degree) something is in fact
done, actually performed, and what its doing entails’’. It is the third of
 574 . Context, style and culture
these questions that concerns us here. If we think of the linguist’s model
of the language-system as a set of rules which generates all the well-
formed system-sentences of a language, we can conceive of this as being
incorporated within a more comprehensive model of language-compe-
tence, which contextualizes* these system-sentences according to certain
conditions of appropriateness. No one person of course has a perfect
mastery of any language; there are degrees of fluency, and there are
variations of different kinds in any language-community. Our model of
language-competence, however, will be based upon the knowledge
possessed by what might be described as an ideal omnicompetent speaker
of a language, where ‘omnicompetence’ implies, not only perfect mastery
of the rules which determine the well-formedness of sentences, but also
the ability to contextualize them appropriately in terms of the relevant
variables,
Some of these contextual variables may be identified, in a preliminary
way at least, by asking what kinds of knowledge the participants in a
language-event must possess, over and above their knowledge of the
phonological and grammatical rules of the language-system and the
sense and denotation of lexemes, in order to produce and understand
contextually appropriate utterances.” Much of this additional knowledge,
we may assume, is of a very general nature, which is not restricted to the
use of language, but is relevant to all kinds of semiotic behaviour. Under
this head we can include an understanding of certain universal logical
principles and of the general conditions of appropriateness that Grice
(1975) has called conversational implicatures* (cf. 14.3). We are not
concerned with these at this point. What we have in mind is knowledge
of the kind that determines particular phonological, grammatical and
lexical options within the language-system in particular contexts of
language-use. Let us list some of these.

(i) Each of the participants must know his role* and status*. Lin-
guistically relevant roles are of two kinds: deictic and social. Deictic
roles derive from the fact that in normal language-behaviour the speaker
addresses his utterance to another person (or other persons) who are
2 According to Goffman (1964): “‘It hardly seems possible to name a social
variable that doesn’t show up and_have its little systematic affect upon speech
behaviour: age, sex, class, caste, country of origin, generation, region,
schooling; cultural cognitive assumptions; bilingualism, and so forth’’. For
exemplification and discussion, cf. Bauman & Sherzer (1974), Bright (1966),
Fishman (1965, 1968, 1971, 1972a, b), Giglioli (1972), Gumperz & Hymes
(1971), Hymes (1964, 1974), Pride (1970), Pride & Holmes (1972).
 14.2. Communicative competence 575
present in the situation and may refer to himself, to the addressee(s) or
to other persons and objects (whether they are in the situation or not),
not by means of a name or description, but by means of a personal or
demonstrative pronoun, whose reference is determined by the participa-
tion of the referent in the language-event, at the time of the utterance.
Deictic roles are grammaticalized in many, though not all, languages in
what is traditionally called the category of person*. We will discuss this
in more detail in the chapter dealing with deixis* (15.1). Here it is
sufficient to say that in English the use of ‘I’ (and ‘we’) is determined,
in normal language-behaviour, by the speaker’s assumption of the role
of speaker in relation to the addressee(s) and by his referring to himself
as the person fulfilling this deictic role. The addressee must be able to
identify the referent of ‘I’ and also the referent of ‘you’; and this
implies that he knows that he is being addressed. Many of the non-vocal
paralinguistic phenomena which accompany and are integrated with
spoken utterances have this vocative* function of inviting a particular
person to assume the role of addressee; and names, titles or special
terms of address based on social status may be used, and in some situa-
tions are obligatory, in order to identify the addressee (cf. 7.5).
Social roles are culture-specific functions, institutionalized in a society
and recognized by its members: for example, the function of being a
doctor, a parent, a teacher, a customer, a priest. These roles are typically
reciprocal: doctor-to-patient and patient-to-doctor, parent-to-child and _.
child-to-parent, and so on. The most obvious effect of social role, as a
contextual variable, lies in its determination of terms of address: as when
‘Sir’, ‘Doctor’ or ‘My lord’ (in the courtroom) are used with vocative
function in English. The speaker in using such expressions accepts, and
shows that he accepts, his role vis-a-vis the addressee. In many lan-
guages there is a richly differentiated set of terms of address which the
speaker must control if he is to produce appropriate utterances in
various situations. Social role may also determine the selection of per-
sonal pronouns and associated components of the grammatical structure
of utterances. A clear instance of this is the use of the so-called royal
first-person plural pronoun by a monarch, the Pope or a bishop in a
number of European languages (“‘ We have taken unto ourself [sic] . . .’’),
and, in Japanese, the use of a special first-person pronoun by the Em-
peror. Generally speaking, however, it would seem to be status, rather
than role, which is the determining factor in the selection of pronouns.
For example, the fact that in the Russian army, before the Revolution,
an officer would address a private soldier as ‘ty’ (roughly comparable
576 Context, style and culture
with the French ‘tu’ and the German ‘du’), but be addressed by the
soldier as ‘vy’ (cf. French ‘vous’, German ‘Sie’), is explicable in terms
of more general principles based on status (cf. Friedrich, 1966). Role
normally implies status. There are, however, many aspects of language-
behaviour that are systematically determined by social role: the use of
various characteristic expressions by a judge addressing the jury or a
preacher addressing the congregation, by lovers in situations of inti-
macy, by a person saying his prayers, and so on. Role may also be the
primary determining factor in the switch from one dialect to another, or
even from one language to another, in situations of diglossia* —a
phenomenon that will be referred to later in this section.
By social status is meant the relative social standing of the partici-
pants. Each participant in the language-event must know, or make
assumptions about, his status in relation to the other; and in many
situations status will also be an important factor in the determination of
who should initiate the conversation. The participants may not agree
about their relative status; each speaking to the other as superior-to-
inferior, or more commonly perhaps (and in a way that is often con-
ventionalized in language by means of an accepted code of politeness) as
inferior-to-superior; or one treating the other as an equal, while he is
himself addressed as a superior or inferior. Societies vary considerably,
of course, in the degree to which status is explicitly recognized as such
and institutionalized in dress, titles and so on; and the degree to which
language-behaviour is determined by status also varies from one
language to another. But there is probably no language for which it is —
totally irrelevant.
Once again, the most obvious correlate of social status in language-
behaviour, as far as the utilization of the language-system is concerned,
is in the use of particular terms of address and personal pronouns. It is
supported and confirmed by such paralinguistic phenomena as eye-
movements, gestures, posture and physical contact or proximity (cf. 3.2).
The importance of status in the selection of certain terms of address in
American English has been demonstrated in a now classic paper by
Brown and Ford (1961); and their work has been carried further by
Ervin-Tripp (1969). In many European languages, though not in
Modern English, the selection of a second-person singular pronoun is
determined, partly at least, by relative social status: the particular deter-
mining factors vary, however, from one language to another, and indeed
from one social group to another within the various language-communi-
ties. In Japanese and Korean social status and deictic role jointly
 14.2. Communicative competence 577
determine the selection of all the personal pronouns; and status (to-
gether with other factors) governs the selection of particular forms of
certain verbs (cf. Martin, 1964; Harada, 1975). But status, like role, also
determines, and probably in all languages, the selection of a wide range
of stylistic factors in phonology, grammar and vocabulary; and the socio-
linguistic literature contains many illustrations of this from all over the
world.
Sex and age are so often determinants of, or interact with, social
status that they may be conveniently mentioned here. The terms of
address employed by a person of one sex speaking to a person of another
sex, or by a younger person speaking to an older person, may differ from
those which would be employed in otherwise similar situations by people
of the same sex or of the same age. This phenomenon is so pervasive and
SO apparent even to the casual observer of language-behaviour that
exemplification is unnecessary. The Women’s Liberation movement has
recently drawn attention to some of the linguistic difficulties which stand
in the way of their achieving social equality with men: notably, to the
fact that few of the major languages of the world provide a general
term of address for a woman which is not determined by her marital
status. The sex of the participants is grammatically relevant in many
languages. In Thai men employ one first-person pronoun and women
another, and there are other systematic differences of grammatical
structure; and in a number of other languages in various parts of the
world there are more extensive grammatical differences, as well as
differences in phonology and vocabulary, between the language of men
and women (cf. Haas, 1944; Grootaers, 1952). In the Romance and
Slavonic languages the sex of the participants determines the form of
certain adjectives and certain verb forms according to the category of
gender; and this, it should be noted, unlike the gender agreement which
holds between third-person pronouns or noun phrases and verbs or
adjectives, is wholly a matter of contextual appropriateness. For example,
Je suis heureux and Fe suis heureuse in French (“‘I am happy’”’) are both
grammatically well-formed; the first utterance, however, would normally
be produced by a man or boy, the second by a woman or girl. The
qualification implied by the use of the word ‘normally’ is, as always,
necessary. What counts is not, in principle, the actual sex of the partici-
pants, but the sex that is ascribed to them or they ascribe to themselves
in the situation. A man might be playing a woman’s part in a play, for
example; and there are other obvious situations in which a man might
appropriately say Fe suzs heureuse.
578 Context, style and culture
(ii) The participants must know where they are in space and time. At
first sight, this might appear to be an unnecessary condition to impose
upon the appropriateness of utterances. Consider, however, an utterance-
token like We are having a fine summer here in Queensland this year pro-
duced by someone in Edinburgh in December. It is grammatically and
semantically well-formed, but situationally inappropriate; and it is for
this reason that it is uninterpretable (except, again, under rather special
circumstances). One cannot be having a fine summer during winter in a
place where one is not. The situational inappropriateness of the utter-
ance derives from the fact that ‘here’ is a deictic adverb which refers to
the place where the speaker is (or believes himself to be) at the time of
utterance, and the tense of the verb, as realized in the form are having,
refers to a period of time which contains the point of time at which the
utterance is made. ‘The speaker of a language must control and be able to
correlate at least two different systems of spatiotemporal reference: one
is the deictic system whose co-ordinates are created by the act of utter-
ance itself (cf. 15.1); the other is a culture-specific system for referring
to time and place that is lexicalized in the language he is speaking.
The appropriate use of greetings such as Good afternoon! or Happy
Christmas! is similarly dependent upon the speaker’s knowledge of the
time at which he is producing them. In order to be able to employ them
correctly the speaker must know (in addition to certain other facts) what
counts as afternoon or Christmastide and whether it is indeed the after-
noon or Christmastide at the time of utterance. He can of course
deliberately violate the normal conditions governing the use of such
greetings. For example, he might say Good afternoon! in the middle of
the morning to a colleague arriving late to work; and his utterance will
be understood as situationally appropriate, but ironical. Irony depends
upon and presupposes the participants’ knowledge of the normal con-
ditions of situational appropriateness.
The speaker and addressee are normally in the same spatiotemporal
location; and it is probably true to say that all languages are designed, as__
it were, to operate in such circumstances. Problems of spatiotemporal
reference arise when the participants are separated in space and time.
We have only to think of the difficulties we encounter in this respect
when we make a long-distance telephone-call (e.g., from Great Britain
to the United States). The speaker can either adopt the spatiotemporal
co-ordinates of his own location (greeting the addressee, let us say, with
Good afternoon!) or he can project himself into the spatiotemporal loca-
tion of the addressee (saying Good morning!). But the speaker is not
 14.2. Communicative competence 579
completely free with respect to the possibility of projection into his
addressee’s spatiotemporal location: there are restrictions. For example,
if we are in London, speaking (in English) to someone in New York, we
can say, appropriately, either We are going to New York next week or
We are coming to New York next week. We can also say We are going
there next week, and even We are coming there next week (where the deictic
adverb ‘there’ refers to New York). What we cannot say without |
violating the rules which govern the use of ‘here’ is We are coming here
next week (with ‘here’ referring to New York). We are coming here next
week is a perfectly grammatical utterance (more clearly so perhaps than
We are coming there next week, which some speakers of English find un-
acceptable). But it is situationally inappropriate. The use of ‘come’,
unlike the use of ‘here’, allows the speaker to project himself into a
deictic context centred on the addressee.
The conditions under which deictic projection* is permitted (if I may
introduce a term for the phenomenon just illustrated) would seem to
vary, to some degree at least, from one language to another. For example,
the French ‘venir’ and the Italian ‘venire’ (“to come’’) cannot be used
in deictic projection as freely as the English ‘come’ can. Similarly, in
Classical Latin it was possible, when writing letters, to use the so-called
epistolary past tense in referring to events taking place at the time of
writing; and this practice is explained in terms of the writer’s projection
of himself into the situation the receiver would be in when the letter
arrived. The past tense cannot be used in this way in English.
The non-deictic system of spatiotemporal reference was described
above as culture-specific. It is important to realize that there may be
alternative, and even conflicting, systems used by different groups
within a language-community. The Jewish New Year and the Christian
New Year do not coincide; Christmas is celebrated somewhat later by
the members of some of the orthodox churches than it is by the members
of other Christian sects; and so on. The interpretation of phrases like
‘over the New Year’ and ‘at Christmas’, in terms of some external and
neutral system of temporal reference, may vary accordingly. Even more
striking are the discrepancies which arise in different parts of the
English-speaking world in the correlation of local seasonal reference
(e.g., ‘this summer’) with standard calendar reference (e.g., ‘in July’ or
‘in December’). The situational inappropriateness of We are having a
fine summer here this year, said in Edinburgh in December, depends in
part on its violation of the system of local seasonal reference. Our ideal
omnicompetent speaker of English must be able to control and inter-
 580 Context, style and culture
relate, appropriately, the deictic system, and a whole set of secular and
religious holidays or feasts. Whether knowledge of the kind should be
included within linguistic competence is a moot point (cf. Leech, 1969:
118). But it certainly belongs to language-competence as this is manifest
in the appropriate or inappropriate use of English.

, (iii) The participants must be able to categorize the situation in terms of

its degree of formality*. Joos (1962) has postulated five degrees of for-
mality in English, each of which is said to correlate with systematic
phonological, grammatical and lexical differences: his terms for the five
kinds of situation and the styles* of English appropriate to them are
‘frozen’, ‘formal’, ‘consultative’, ‘casual’, ‘intimate’. Whether the
scale of formality in English can be categorized as neatly as Joos suggests
in terms of five distinct styles is open to doubt (cf. Crystal & Davy, 1969:
74). But it is intuitively clear that there is a scale of formality, not only
in English, but probably in all languages. We all recognize that certain
utterances would be phonologically, grammatically and lexically stilted if
used in certain informal or intimate situations; and, conversely, that
there are utterances that are appropriate in informal situations, but
would be judged by most speakers to be too colloquial for formal
occasions.
In many language-communities two or more distinct dialects of the
same language are regularly employed by educated speakers, the use of
the one or the other depending upon the formality of the situation (and
also upon other factors). Ferguson (1959), in his classic paper on
diglossia*, illustrates this phenomenon with reference, primarily, to
Arabic, Swiss German, Haitian Creole and Modern Greek, but he also
refers to Tamil, Medieval Latin and Chinese in the same connexion. It
has since been extensively discussed and illustrated for many language-
communities throughout the world (cf. Hymes, 1964; Fishman, 1968;
Gumperz & Hymes, 1971; Pride & Holmes, 1972). The ability of mem-
bers of such language-communities to pass from one dialect or variety
of the language to another according to the situation-of-utterance may
be referred to as code-switching*.
Code-switching is by no means restricted to language-communities in
which two or more recognizably distinct dialects (or languages) are
regularly employed. As recent research has shown, there is no qualita-
tive or functional difference between diglossia (the situationally deter-
mined employment of different dialects or languages within the same
language-community) and the language-behaviour of so-called mono-
 14.2. Communicative competence 581
linguals, who switch from one style to another under similar determining
conditions (cf. Sankoff, 1972). It is in any case very often difficult to
draw a sharp distinction between styles, dialects and languages (cf. 14.5).
Code-switching is particularly striking, and obvious to the most casual
observer, when a Puerto-Rican executive and his secretary shift from
English to Spanish and then back again to English in the course of a
single conversation, Spanish being used for the casual and friendly dis-
cussion of a topic that has arisen in connexion with a letter that is being
dictated and English, not only for the letter itself, but for all the more
formal sections of the conversation (cf. Fishman, 1969). But there would
probably be discernible differences of style in a similar conversation
between a monolingual English-speaking businessman and his secretary.
As Hymes says (1967): “‘Cases of bilingualism par excellence . . . are
salient, special cases of the general phenomena of variety in code reper-
toire and switching among codes. No normal person, and no normal
community, is limited in repertoire to a single variety of code’.

(iv) The participants must know what medium* is appropriate to the

situation. As we have already seen, this is not simply a matter of being
able to control the peripheral transmitting and receiving mechanisms
involved in speech and writing: the medium ts to be distinguished from
the channel* (cf. 3.3). No more need be said about this question at this
point, except to emphasize that there are medium-dependent differences
of grammar and vocabulary that have a bearing upon the situational
appropriateness of particular utterances. By virtue of the frequent and
longstanding association of the graphic medium with more formal, and
the phonic medium with less formal, situations in many cultures,
medium-dependent variations of grammatical and lexical structure
correlate highly with variations based on formality. For example, a judge
addressing the jury or pronouncing sentence in English will use the
graphic medium, as far as grammar and vocabulary are concerned, even
though his utterance is transmitted along the vocal-auditory channel
(and conforms to the phonological structure of the phonic medium). His
utterance will be in a formal style, and it may contain elements peculiar
to his particular role and status.

(v) The participants must know how to make their utterances appro-
priate to the subject-matter; and the importance of subject-matter as a
determinant in the selection of one dialect or one language rather than
another in bilingual or multilingual communities has been stressed by
582 Context, style and culture
such writers as Haugen (1953), Weinreich (1953) and Fishman (1965).
More recently, however, Fishman (1972c) has pointed out that the
greater appropriateness to subject-matter of one language rather than
another in multilingual settings ‘may reflect or be brought about by
several different but mutually reinforcing factors”; and he has suggested
that the selection of one language rather than another may be simply a
consequence of the fact that “certain socio-culturally recognized spheres
of activity are, at least temporarily, under the sway of one language or
variety’.
Crystal and Davy (1969) introduce the term province* for ‘‘the
features of language which identify an utterance with those variables in
an extralinguistic context which are defined with reference to the kind of
occupational or professional activity being engaged in’’; and they make
the point that “‘subject matter, in so far as this is a question of the use of
distinctive vocabulary, is but one factor among many which contributes
to a province’s definition, and in any case has predictive power only
in a minority of extremely specialist situations”. This is undoubtedly
correct.
It does not follow, however, that the semanticist should not be con-
cerned with subject-matter as a contextual variable. Its importance is
revealed as soon as we consider the practical problems of disambiguating
utterances which contain lexemes with more than one sense: e.g., That
plant is an eyesore. If the conversation in which this utterance occurs is
concerned with the layout or appearance of a garden it will presumably
be taken to have a different meaning from the meaning that the same
utterance (i.e. as a token of the same type) would have in a conversation
devoted to the architectural merits of a group of factory buildings. Ad-
mittedly, other situational variables might suffice, in particular instances,
to disambiguate such utterances. But, in principle, our omnicompetent
speaker can talk about anything, whatever occupational or professional
activity he happens to be engaged in at the time and whatever social role
he happens to be performing. The fact that his choice of vocabulary will
be very largely determined by subject-matter may well imply that the
selection of one word rather than another falls outside the scope of
stylistics (“the description of the linguistic characteristics of all situa-
tionally-restricted uses of language’’: Crystal & Davy, 1969: go). But
we cannot, as semanticists, neglect the fact that the speaker can assume,
and normally does so unconsciously, that particular lexemes will be
interpreted by the addressee in one sense rather than another by virtue
of the subject-matter of the utterance in question and previous utter- _
 14.2. Communicative competence 583
ances in the conversation. So far, however, little progress has been made
in giving a theoretically satisfying account of this phenomenon.
When research in machine-translation was being actively pursued in a
number of centres in different countries throughout the world some
years ago, it was suggested by certain scholars that homonymous or
polysemous lexemes could be disambiguated by means of a computer-
program which would scan a text and determine its subject-matter in
terms of the occurrence in the text of a preponderance of lexemes from a
certain area of the vocabulary; and this technique is now regularly
employed, with a fair measure of success, in automatic indexing and
information retrieval. In its most sophisticated and_ linguistically
most interesting form, the proposal to disambiguate homonymous or
polysemous lexemes in this way presupposes an analysis of the lexi-
cal structure of the language-system on thesaurus*, or field-theory*
principles (cf. 8.2). It might be assumed, for example, that the noun
‘plant’ would be shown in the thesaurus as belonging to at least two
fields, the one field containing such lexemes (in one of their senses) as
‘vegetable’, ‘bush’, ‘flower’, ‘lawn’, ‘garden’, ‘grow’, ‘prune’,
‘weed’, and the other field containing such lexemes as ‘factory’,
‘machine’, ‘manufacture’, ‘equipment’, ‘building’. The idea under-
lying this approach to the contextual resolution of lexical ambiguities is
intuitively attractive. It is doubtful, however, whether any purely
mechanical, or algorithmic, procedure for disambiguation can be devised
along these lines, even presupposing the existence of an ideal thesaurus
(cf. Bar-Hillel, 1964: 178). Nonetheless, it seems to be an inescapable
fact that the participants’ awareness of the subject-matter is a potential
and frequently relevant disambiguating factor in everyday language-
behaviour, whether this can be accounted for in terms of the co-
occurrence in a text of a relatively large number of lexemes from the
same semantic field or not.
There is another aspect of subject-matter, which relates to the ex-
pressive* function of language (cf. 2.4). This is the selection by the
speaker of elements which make the utterance appropriate to his attitude
towards, or his emotional involvement in, what he is talking about. He
may be ironical, enthusiastic, sceptical, reserved, scornful, sentimental ;
and so on. Although the speaker’s attitude towards the subject-matter
may be influenced by such other situational factors as degree of formality
and the interpersonal relations subsisting between him and the addressee,
it is, in principle, distinguishable from these other factors. For example,
some speakers might avoid using what are generally regarded as obscene
584 Context, style and culture
words in more formal situations, and in informal situations when
addressing a member of the opposite sex, but might use them quite
freely, in relation to the same subject-matter, when talking informally to
some of their own sex; and their employment of such words might be
indicative of their attitude towards the subject-matter, as well as having
the particular social function of promoting solidarity.

(vi) The participants must know how to make their utterance appro-
priate to the province* or domain* to which the situation belongs. ‘The
term ‘province’ has already been introduced, under (v), with its defini-
tion by Crystal & Davy (1969). The term ‘domain’ is taken from Fish-
man (1965), who defines a domain as a “cluster of social situations
typically constrained by a common set of behavioral rules”’ and relates
it to “those ‘generally termed’ spheres of activity which have more
recently been independently advanced by others interested in the study
of acculturation, intergroup relations and bilingualism’’. Yet a third
term that is quite widely used in the recent literature of linguistics and
stylistics is register*, which has been defined in terms of systematic
variation “‘by use in relation to social context’? (Leech, 1966: 68; cf.
Halliday, McIntosh & Strevens, 1964: 77; Strang, 1968: 21). ‘Register’,
however, is commonly held to subsume, not only the phenomena
covered by ‘province’ and ‘domain’, but also subject-matter.
Scholars who have been concerned with systematic variation of the
kind that we are referring to here would be among the first to admit that,
whatever technical terms they may employ, their theoretical discussions
and classification of the phenomena are tentative and provisional. Fish-
man (1965) relates the concept of the domain of language-behaviour, on
the one hand, to subject-matter, and on the other, to locale* and role-
relations. He points out that “most major social institutions are associa-
ted with a few primary locales’’. For example, the domain of the family
is primarily associated with the home; the domain of religion is primarily
associated with the church; the domain of employment is primarily
associated with the office or factory; and so on. Within each domain a
variety of characteristic reciprocal role-relations (and their converses)
can be identified: mother-to-father, wife-to-husband, parent-to-child;
priest-to-parishioner; secretary-to-boss; etc. The locale of the utterance
and the role-relations of the participants tend to be mutually reinforcing
and congruent; and they also tend to be congruent with the subject-
matter. But they can be incongruent; and, in such cases, one can
investigate which of the components, if any, is dominant in the determina-
 14.2. Communicative competence 585
tion of the structure of the utterance. “If one meets one’s clergyman at
the race track the impact of the locale on the topics and role-relation-
ships that normally obtain is likely to be quite noticeable’? (Fishman,
1972c: 22). Fishman is mainly concerned with the establishment and
validation of a theoretical framework within which one can describe, and
perhaps explain, systematic variation in language-behaviour (code-
switching*) in diglossic or multilingual communities. The contextual
variables that he and other sociolinguists have discussed in connexion
with the notion of domain are equally important, however, in the analysis
of the situational appropriateness of utterances in what are normally
regarded as monolingual communities.
‘Province’ (as used by Crystal & Davy, 1969: 71ff) is narrower in
scope than ‘domain’; and it fits into a somewhat different analysis of the
major situational variables. Province features are defined “‘ with reference
to the kind of occupational or professional activity being engaged in’’,
and they are said to “provide no information about the people involved
in any situation — about this social status or relationship to each other,
for example’’. Conversation is regarded as a province, but the point is
made that “conversation is different from all other provinces in that it
is the only case where conventional occupational boundaries are irrele-
vant’’. Other provinces in English include the language of public
worship, advertising, newspaper reporting, science and law; and samples
of texts in some of these provinces are discussed in detail by the authors.

Six different kinds of knowledge or competence which have a bearing

on the situational appropriateness of utterances have been listed and
briefly discussed in this section; and there are others that might have
been mentioned. Many of the phenomena that have been referred to
would generally be held to fall within the scope of sociolinguistics* or
stylistics*, rather than within linguistics proper. The question that now
confronts us is whether a distinction can usefully be drawn between
sociolinguistics or stylistics and what, for convenience, we will call
microlinguistics* ; i.e. the branch of the study of language that is con-
cerned with the analysis of the phonological, grammatical and semantic
structure of system-sentences. :
In a much-quoted passage, Chomsky has said (1965: 3): “‘ Linguistic
theory is concerned primarily with an ideal speaker-listener, in a com-
pletely homogeneous speech community, who knows its language [1.e.
the language of the community] perfectly and is unaffected by such
grammatically irrelevant conditions as memory limitations, distractions,
586 Context, style and culture
shifts of attention and interest, and errors (random or characteristic) in
applying his knowledge of the language [i.e. the language-system] in
actual performance”. Chomsky’s use of the term ‘performance’, to
cover everything that does not fall within the scope of a deliberately
idealized and theoretically restricted concept of linguistic competence,
was perhaps unfortunate (cf. Hymes, 1971). But this does not alter the
fact that we can draw a distinction, in principle at least, between the
language-system as a set of sentences (which no individual speaker in any
language-community ever controls) and the appropriate (or inappro-
priate) use of these sentences in situations-of-utterance by the ideal
omnicompetent user. Idealization is inevitable; and it is as much in-
volved in the sociolinguistic or stylistic analysis of the situational
variables which are postulated as the determinants of appropriate use as
it is in the microlinguistic description of the structure of the language-
system as such.
There are of course very considerable methodological problems in-
volved in the idealization of language-behaviour and the postulation of
an underlying system to account for it. We can in fact distinguish three
rather different kinds of idealization, each of which has its own charac-
teristic practical difficulties. The first kind of idealization may be called
regularization*. Under this head, we can discount all slips of the tongue,
mispronunciations, hesitation pauses, stammering, stuttering, etc.: in
short, everything that Chomsky, in the quotation given above, attributes
to the influence of such microlinguistically irrelevant factors as memory
limitations, distractions, shifts of attention and interest, and the mal-
functioning of the physiological and neurological mechanisms involved
in language-behaviour. The distinction of (underlying) competence and
(actual) performance finds its most obvious, and least controversial,
application with respect to this kind of idealization: the regularization
of actual utterances by means of the elimination of what may be called
performance-phenomena.
Such performance-phenomena are far more frequent in everyday con-
versation than is generally appreciated. The participants may not even
notice them during the conversation itself, since there is usually suffi- |
cient redundancy* to compensate for the noise* that errors and other
performance-phenomena introduce into the signal (cf. 2.3). But in many
instances at least, speakers will accept, or readily volunteer the informa-
tion, that they have produced an incorrect utterance, if they are con-
fronted with a transcript or recording of it afterwards. In many cultures,
including our own, there are certain grammatical and phonological
 14.2. Communicative competence 587
norms with which many speakers have become acquainted during their
formal schooling; and the knowledge of these norms tends to influence
their judgements of correctness. This 1s a serious problem, which
linguists cannot afford to dismiss. But the principle of regularization
is not in doubt.
The investigation of performance-phenomena is by no means devoid
of theoretical interest. The incidence and nature of errors provides
important evidence for the study of the mechanisms involved in speech-
production (cf. Laver, 1970); and social psychologists attach great im-
portance to performance-phenomena as being symptomatic of the
emotional state, or more enduring personality traits and attitudes, of a
speaker (cf. Argyle, 1972). Looked at from this latter point of view, even
errors may be regarded as meaningful: they convey indexical* informa-
tion (cf. 4.2). Performance-phenomena, however, are excluded from the
linguist’s model of the language-system; as are various other components
of utterances which the linguist defines, on other grounds, to be non-
linguistic (cf. 1.6).
The second kind of idealization may be called standardization* ; and
it is more relevant to our present concerns. When we say that two people
speak the same language (e.g., English), we are, whether we are aware of
it or not, abstracting from all sorts of systematic differences in the
language-systems which underlie their language-behaviour. Some of
these differences are covered by the terms dialect* and accent* (cf.
14.5). Others are attributable to such factors as sex, age, social status,
social role, professional occupation, many of which have been described
in this section as contextual variables. There is a sense in which it is true
to say that everyone we normally describe as a native speaker of English
speaks a different English: he has his own language-system, distinct to
some degree in vocabulary, grammar and phonology. Indeed, every
native speaker of English speaks many varieties of English and uses them
in different situations.
It would be absurd to hope to describe, or even to determine, all these
differences within what we call, pre-theoretically, English. What the
linguist does, in practice, is to discount all but the major systematic
variations in the language-behaviour of the community whose language
he is describing; and this is what is meant by standardization. For
example, he would usually exclude from his model of the language-
system any feature of phonology, grammar or vocabulary that was
peculiar to a single individual; and he would probably exclude also any
feature characteristic of the language-behaviour of a small subset of the
588 Context, style and culture
members of the language-community, if this subset did not constitute a
recognizable geographically or socio-culturally determined group within
the community. Of course, there may be particular reasons for a linguist
to be concerned with some restricted variety of English; and he will
normally restrict his description to some pre-theoretically distinct dia-
lect of the language. But there will always be some degree of standardiza-
tion. Empirical questions arise in connexion with the delimitation of
language-communities and the degree of standardization practised by
the descriptive linguist in his postulation of an underlying language-
system. It is pointless to argue, however, that there is no such thing as a
homogeneous language-system underlying the language-behaviour of
the whole language-community. This is true, but irrelevant. The ques-
tion is whether it is useful to assume, as most linguists have assumed in
the past, that there is some kind of overall system underlying those
utterances which most members of the language-community would
accept as being relatively neutral with respect to minor differences of
dialect, situation, medium and chronological period. The empirical
validity of some such concept of an overall language-system, however
vaguely determined might be the notions of minor differences and rela-
tive neutrality inherent in it, is proved by the practical usefulness of the
grammars, phonological descriptions and dictionaries that are produced
by descriptive linguists.
The third kind of idealization that is involved in the postulation of an
underlying language-system may be referred to as decontextualization* ;
and, like standardization (which might be regarded as a distinguishable
part of the general process of decontextualization), it is highly relevant
to the central theme of this chapter. We have said that the linguist’s
model of the language-system can be conceived as a set of rules which
generates all (and only) the system-sentences of a language; and that the
ideal omnicompetent user of a language will not only know all the rules
which determine the well-formedness of system-sentences, but also
possess the ability to contextualize* them appropriately in terms of the
relevant variables. We are now concerned with what might be regarded
as the inverse of this process of contextualization; and we can restrict the
scope of the term ‘language-system’, in the light of our discussion of
standardization, to that of ‘overall language system’. System-sentences
are idealized utterances in the particular sense of the term ‘idealization’

utterances. |
that is implied by ‘decontextualization’: they are derived from utter-
ances by the elimination of all the context-dependent features of
 14.2. Communicative competence 589
Spoken utterances of everyday conversation tend to be heavily con-
text-dependent, as well as being characterized by errors and other
performance-phenomena, which, we are assuming, are eliminable by
regularization. One aspect of context-dependence is manifest in what is
traditionally called ellipsis. A conversation consisting entirely of gram-
matically complete text-sentences would generally be unacceptable as a
text; and it is part of the language-competence of a speaker of the
language (if not of his linguistic competence in the narrower sense) that
he should be able to produce grammatically incomplete, but con-
textually appropriate and interpretable, sentence-fragments*. For
example, the utterance As soon as I can (produced with the appropriate
stress pattern and intonation) might occur in a text in reply to an
utterance (intended and taken as a question) such as When are you
leaving? The grammatical structure of the context-dependent sentence-
fragment As soon as I can, and at least part of its meaning, can be
accounted for by describing it as an elliptical, appropriately con-
textualized, version of the utterance I’m leaving as soon as I can. Ellipsis,
then, is one of the most important and one of the most obvious effects of
contextualization; and decontextualization, in the case of sentence-
fragments such as the one just illustrated, consists in supplying some
element or elements from the preceding co-text.®
Ellipsis is not the only phenomenon to be taken into account in the
decontextualization of text-sentences or sentence-fragments. There is a
whole range of other phenomena, including the use of pronouns, the
definite article, word-order, sentence connectives and such prosodic
features as stress and intonation. Any of these features may suffice to
make a text-sentence or sentence-fragment context-dependent. For
example, the text-sentence I haven’t seen him before cannot be inter-
preted unless the referent of the pronoun ‘he’ can be correctly identified
by the hearer; and the referent will normally have been mentioned in
the preceding co-text. The different, but related, text-sentence I haven't
seen him before (where the pronoun ‘he’, in its form him, bears heavy
stress) is also context-dependent; but the referent of ‘he’ need not have
been mentioned in the co-text. The referent might be some person in
the situational context, who is identified paralinguistically by the speaker
as he makes the utterance (e.g., with a gesture of the hand or a movement
of the head). There is some disagreement among linguists as to how
3 It is not being suggested that everything that would be described traditionally
as a sentence-fragment is to be treated as the product of ellipsis (cf. Allerton,
1975; Shopen, 1973).
590 Context, style and culture
many of these phenomena should be accounted for as part of the struc-
ture of system-sentences. Here we are concerned to emphasize that the
grammatical and semantic coherence of text-sentences and text-frag-
ments within a text is but one aspect of the global problem of contextual
appropriateness; and that it cannot be handled without taking into
account situational factors and the non-linguistic features of utterances
and their co-text.
If linguistic semantics is taken to be that branch of semiotics which
deals with the way in which meaning (of all kinds) is conveyed by lan-
guage, it must be accepted that a comprehensive theory of linguistic
semantics will need to be based upon, or include, a theory of contextual
appropriateness, It is arguable, however, that, at the present time at
least, the construction of such a comprehensive theory of linguistic
semantics is too ambitious a task. There are various ways in which we
can set about constructing a partial theory of linguistic semantics, or a
set of partial theories, each of which will abstract from, or take for
granted, facts which other theories try to systematize and explain. One
such partial theory, which might be appropriately described as a theory
of microlinguistic semantics, would be restricted to the analysis of the
meaning of maximally decontextualized system-sentences. It would be
concerned with the sense and reference of linguistic expressions, with
the range of semiotic functions (or speech-acts: cf. 16.1) that can be
performed by the utterance of particular sets of sentences, with the
implications and presuppositions which hold between the propositions
expressed by sentences (assuming that the sentences are uttered under
certain standard conditions) and with the validation of these proposi-
tions in terms of truth-conditions holding in some actual or possible
world. It would not be concerned, except incidentally and minimally,
with socio-culturally determined variation, with textual coherence or
with the other aspects of contextualization mentioned in this section.
Much of the recent work in the formal analysis of meaning in language
falls within the scope of microlinguistic semantics as we have just
defined this field. Provided that it is appreciated that the distinction of
microlinguistics from sociolinguistics or stylistics is a purely methodo-
logical distinction, based upon the linguist’s regularization, standardiza-
tion and decontextualization of utterances, there 1s much advantage to
be gained from the deliberate neglect in microlinguistic semantics of
contextual appropriateness. Within the restricted framework of micro-
linguistic semantics, we can give a satisfactory account of the sense of
most lexemes in the vocabularies of languages and, no less important,
 14.2. Communicative competence 5g!
we can investigate the way in which the grammatical structure of
system-sentences determines their meaning and their characteristic
semiotic function in utterances; and this is clearly one of the central
tasks of linguistic semantics.
Microlinguistic semantics, as it has been developed so far, deals
primarily with descriptive meaning. Language, however, is not merely
an instrument for conveying factual information; it also serves a variety
of social and expressive functions. Indeed, as we have already seen, it is
difficult, in the last resort, to draw a sharp distinction between the
descriptive and the interpersonal functions of language (cf. 2.4). No
satisfactory and comprehensive theory of semantics can afford to neglect
social and expressive meaning in language; and in doing so it must draw
fully upon the notion of contextual appropriateness. If this is held to
fall within the scope of sociolinguistics or stylistics, then at least this
part of sociolinguistics and stylistics is to be included in linguistic
semantics; and it should always be borne in mind that methodological
distinctions within linguistics do not necessarily reflect any inherent
differences in the internalized system of rules which underlie language-
behaviour.
In a previous chapter, it was pointed out that the distinction fre-
quently drawn by semioticians between semantics and pragmatics* was
of uncertain applicability in the analysis of meaning in natural languages
(cf. 4.4). One way of drawing it by definition (with respect to micro-
linguistics) is to say that microlinguistic semantics deals with the
meaning of maximally decontextualized system-sentences and that micro-
linguistic pragmatics studies the meaning that these sentences have
when they are uttered (as text-sentences) in particular classes of con- }
texts. One can perhaps study the meaning of propositions, and their
truth-conditions in relation to possible worlds, without invoking the
notion of the context-of-utterance. But one cannot get from sentences
to the propositions expressed by them (even supposing that we would
wish to do so) without taking account of certain contextual features (cf.
Stalnaker, 1972: 383). It is for this reason that we have said that
microlinguistic semantics deals with the meaning of maximally, rather
than fully, decontextualized system-sentences. The context-dependence
of many system-sentences (and hence the necessity of invoking prag-
matic concepts in the analysis of their meaning) is especially clear in the
case of sentences containing deictic elements (cf. 15.1).
 592 Context, style and culture
14.3. Conversational implicatures and presupposition
The term implicature* was introduced into the philosophy of language
by Grice in his William James lectures in 1967/8 (cf. Grice, 1975). It is
now quite widely employed, not only by philosophers, but also by
linguists.
As we have already seen, the term ‘implication’ is normally used in
philosophical semantics to refer to the truth-functional relation of
material implication; and this in turn is distinguished from strict im-
plication, or entailment (cf. 6.2). In everyday usage, the words ‘implica-
tion’ and ‘imply’ are used in what appears to be a quite different sense
from that which is associated by definition with the operation of material
implication under the standard interpretation of the propositional cal-
culus (hence the so-called paradoxes of material implication) and in
: what is certainly a much broader sense than that borne by the terms
‘entailment’ and ‘entail’ in philosophical semantics. Grice’s notion of
implicature 1s intended to cover at least some of the difference between
the broader, everyday, notion of implication and the narrower, philoso-
phical, notion of entailment. He is also concerned to show how implica-
tures co-operate with, and supplement, material implication in the
everyday use of language: we will not go into this part of his programme.
The notion of implicature rests upon a distinction between what is
actually said and what is implied (but not entailed) in saying what is
said. As we shall see later, there are various senses in which the verb
‘say’ can be interpreted. At least two of these are relevant in the present
connexion: “‘say,’’ and “‘say,”’ (cf. 16.1). For example, if someone says,
(i.e. utters a token of the utterance-type that is conventionally repre-
sented as) It 1s cold in here he would normally be saying, (i.e. asserting
the proposition) that it is cold where he is. Considerable attention has
been devoted by Grice and other philosophers of the so-called ordinary-
language school (cf. 6.1) to the analysis of the conditions under which
in saying X one can be held to have said,, and to have meant, that p
(where X is an utterance-signal and.p is a proposition). They have also
been much concerned with explicating the notion of meaning in terms
of which it is reasonable to assert that one can say (i.e. “‘say,’’) that p
without meaning that p. It turns out that in one sense of ‘mean’ we
cannot say that p without meaning that p; and the reasons why this is so
will occupy us later (cf. 16.1). There is a different, but undoubtedly
related, sense of ‘mean’, however, in which it is possible, in saying that
p, to mean that q (p % q), instead of, or in addition to, p. For example,
in saying, that it is cold where one is (by saying, Jt 7s cold in here) one
 14.3. Conversational implicatures and presupposition 593
might mean, or be implying, that the heating should be turned up, that
one’s host pays more heed to his fuel bills than to the comfort of his
guests, and so on. Given that certain conditions are satisfied, we shall be
entitled to say (in terms of Grice’s notion of implicature) that these
various additional propositions are implicated*, though not asserted:
they are implicata* of the utterance It 1s cold in here (under certain con-
textual conditions).
Grice distinguishes two kinds of implicature: conventional and con-
versational. The difference between them is not always clear-cut in
particular cases. In principle, however, the difference seems to be that,
whereas a conventional implicature depends upon something additional
to what is truth-conditional in the normal (i.e. conventional) meaning of
words, a conversational implicature derives from a set of more general
conditions which determine the proper conduct of conversation. It is the
so-called conversational implicatures with which we are concerned here;
and henceforth the terms ‘implicature’ and ‘implicate’ will be used
without qualification in this narrower sense.
The conditions from which implicatures derive are formulated by
Grice as maxims, grouped under the four headings of quantity, quality,
relation, and manner.
Quantity. (i) Make your contribution as informative as is required (for
the current purposes of the exchange); (ii) do not make your contribu-
tion more informative than is required.
Quality. Try to make your contribution one that is true: (1) Do not
say what you believe to be false; (i1) Do not say that for which you lack
adequate evidence.
Relation. Be relevant.
Manner. Be perspicuous: (1) Avoid obscurity of expression; (11) Avoid
ambiguity; (iii) Be brief (avoid unnecessary prolixity); (1v) Be orderly.
All of these maxims relate in a fairly obvious way to the more general
purpose of promoting the efficient communication of propositional
information. They are inherently restricted, therefore, to what we have
identified as the descriptive function of language. But much, if not most,
of the semantic information contained in everyday language-utterances
is social and expressive, rather than descriptive (cf. 2.4). In so far as
Grice’s maxims are inapplicable in the analysis of utterances whose
function is something other than that of augmenting the addressee’s
store of propositional knowledge, they need to be supplemented and
qualified in various ways. It has been pointed out, for example, that
politeness and consideration for the feelings of one’s addressee may
594 Context, style and culture
impose requirements that are in conflict with any or all of Grice’s
maxims (cf. R. Lakoff, 1973).
The usefulness of Grice’s maxims is further reduced by the generality,
not to say vagueness, with which they are formulated. We have already
looked at some of the problems that arise in connexion with the quanti-
fication of semantic information (cf. 2.3). As for relevance and perspi-
cuity, it is, if anything, far more difficult to evaluate utterances in terms
of these two properties than it is to quantify the amount of semantic
information in an utterance. The fact that Grice’s maxims have not
been, and perhaps cannot be, fully formalized makes his notion of
implicature rather less precise than a logician would like it to be.* It is
undeniable, however, that, whether they are fully formalizable or not,
the pre-theoretical notions that Grice has dealt with in his formulation
of the maxims of quantity, quality, relevance and manner have an im-
portant explanatory role to play in the semantic analysis of texts.
By appealing to the maxim of quantity, for example, we can account
for the fact that, if X says to Y .
(1) Have you finished your homework and put your books away?
and Y replies
(2) I have finished my homework,
X can reasonably infer that Y has not put his books away. Presented
with the conjunction of p and q, Y has deliberately chosen to assign a
truth-value to just one of the conjuncts, p, when he might have assigned a
truth-value to the whole conjunction, p & g (by saying yes), if not only p,
but also g, were true. Given that X has no reason to believe that Y is
violating the maxim of quantity (or any of the other maxims), X is
entitled to assume that g is false. At the same time, it is obvious that p
does not entail ~g. Nor can Y be held to have asserted ~g (or, alter-
natively, to have denied q: cf. 16.4). He has merely implicated ~q; and
he has done so by his failure to assert g (in a context in which he could
be expected to assert q).
Taken together, the maxims of quantity and quality can be invoked,
as we shall see later, to. account for the fact that, if someone says J think
it’s raining or It may be raining, he can be held to have implied that he
does not know for certain that it is raining (cf. Caton, 1966; Ducrot,
1972). According to the maxim of quantity, we should be as informative

4 They are partially formalized, and fully discussed in relation to the notions of
presupposition and implicature, in Gazdar (1976).
 14.3. Conversational implicatures and presupposition 595
as we need to be. The proposition “It may be raining”’ is less informa-
tive than the proposition “It is raining’’, since it is compatible with both
“It is raining”’ and “It is not raining’’. The speaker would presumably
have said It 1s raining, without qualifying in any way his own commit-
ment to the truth of the proposition “It is raining’’, if he had known for
certain that it was raining. For knowledge that p is true constitutes
adequate evidence for asserting p. It follows that, by saying either J think
it’s raining or It may be raining, the speaker can normally be held to
imply (i.e. to implicate) that he does not have the evidence that would
enable him to make the more informative assertion It 1s raining. On the
other hand, if the speaker, having said It is raining, is asked Why do you
think it’s raining?, he can quite reasonably, though at first sight illogic-
ally, reply, I don’t think it’s raining: I know tt is, It is interesting to note,
in this connexion, that, in the everyday use of language, not only Jt may
be raining and I think it’s raining, but also It must be raining and I know
it’s raining, involve a weakening of the speaker’s commitment to the
truth of the proposition “It is raining’? (cf. 17.2). This too can be
explained in terms of the Gricean maxims: if the speaker’s evidence is
unimpeachable or his commitment to the truth of p so firm that there
is no doubt at all in his mind that p is true, he will not feel obliged to
make explicit the fact that this is so. By being more informative, in this
respect, than he need be, he draws the addressee’s attention to the
possibility that the evidence for p is not as strong as it ‘might be.
It is characteristic of implicatures that derive from the maxim of
quantity, though perhaps not of others, that they can be explicitly can-
celled or qualified by the speaker ; and in this respect they differ sharply
from entailments. For example, if X said to Y
(3) I tried to telephone Fohn yesterday,
it would normally be reasonable for Y to infer that X had failed to con-
tact John. But this implicature can be cancelled without contradiction.
If Y asks X

(4) Did you try to telephone Fohn yesterday?

X can reply
(5) Yes, and I got through straightaway
or

(6) f not only tried, but I succeeded.

596 Context, style and culture
Again, a statement like the following,
(7) Most languages have at least one sibilant,
would normally be held to implicate the proposition “Some languages
do not have any sibilants’’. But there is no contradiction involved in the
utterance of
(8) Most, 1f not all, languages have at least one sibilant,
where the implicature is explicitly qualified by the speaker. This kind of
qualification is very common.
If the distinction between implicatures and entailments is, for the
most part, fairly sharp, this is far from being the case as far as the dis-
tinction between implicatures and presuppositions* is concerned. What,
then, are presuppositions? How do they differ, on the one hand, from
entailments and, on the other, from implicatures?
There are, in fact, several senses in which the term ‘presupposition’
has been used, more or less technically, by philosophers and linguists
(cf. Cooper, 1974; Garner, 1971; Kempson, 1975; Wilson, 1975). First
of all, there is a sense in which Strawson (1950) employed the term in his
criticism of Russell’s (1905) analysis of statements like
(9) The King of France ts bald.
According to Russell, the proposition expressed by (9) is (roughly) of the
form
(10) “There is one, and only one, King of France and he is bald’’.
Given that there was no King of France in 1905, as there is not to-day,
anyone saying The King of France is bald in 1905, or to-day, would be
asserting something that was false; namely the existential proposition,
(11) “There is a King of France”’.
As we have seen, Strawson objected to Russell’s analysis on the grounds
that it fails to draw a distinction between assertion and presupposition
(cf. 7.2). What we should say, according to Strawson (who, in this
respect, agrees with Frege), is that (9) is neither true nor false, because
one of its presuppositions, (11), is not satisfied.
Strawson’s view of the proposition asserted in (9g) 1s that it has no
truth-value. An alternative view, which some proponents of so-called
presuppositional logic have adopted, is that it does have a truth-value:
the somewhat peculiar truth-value of neither-true-nor-false, distinct
 14.3. Conversational implicatures and presupposition 597
from the two values, true and false, of the standard propositional calculus
(cf. 6.2). Yet another view, based on much the same notion of pre-
supposition, might be that (g) cannot be said to assert any proposition at
all unless its existential presupposition is satisfied. What these various
theories of presupposition have in common, despite these differences, is
their acceptance of the principle that if p is a necessary condition of the
truth of both g and ~g, then p is a presupposition of g. They are all
truth-conditional* theories of presupposition (cf. 6.6). It will be noted
that (11) is a necessary condition, not only of the truth of the proposi-
tion expressed by (9), but also of its falsity: i.e. it is a necessary condition
of the proposition expressed by
(12) The King of France 1s not bald.
This, at least, is the view taken by Strawson and his followers; and, for
the moment, we may leave it unchallenged and without qualification.
(In a later chapter we will draw a distinction between the assertion of
~p and the denial of p: cf. 16.4.) The point to be stressed here is that,
if p is a necessary condition of the truth of both g and ~g, then p cannot
be simply an entailment of g. For g => p (“gq entails p’’) is consistent
with, but does not imply, ~qg. This kind of presupposition, unlike entail-
ment, remains constant, then, under negation.
It also remains constant under the conversion of a simple statement
into the corresponding question. For example,
(13) Ls the King of France bald?
carries with it the same existential presupposition as (9) or (12) does.
Somewhat different from the existential presuppositions associated
with the use of definite referring expressions are the presuppositions of
what we will later refer to as x-questions (cf. 16.3). As we shall see, a
question like
(14) What did Fohn do?
carries the presupposition that John did something; and anyone
answering the question must accept this presupposition.® But this kind
of presupposition does not remain constant under negation. For
(15) What didn’t Fohn do?
makes the presupposition, not that John did something, but rather that
* I am assuming that Nothing is not an answer, though it is an appropriate
response (as I don’t know is), to (14). This is perhaps debatable.
598 Context, style and culture
there is something that John did not do. In so far as there is any par-
ticular statement to which (14) corresponds in the way that (13) corres-
ponds to (9), it is
(16) John didn’t do something;
and (16) expresses (or contains) the very proposition that (15) pre-
supposes. It follows that the kind of truth-conditional account of
presupposition that has been proposed for utterances containing definite
referring expressions will not do for the explication of the presupposi-
tions of x-questions.®
The notion of presupposition that applies to x-questions is similar to,
if not identical with, the notion of presupposition that we invoked,
without actually discussing it, in the section dealing with theme, rheme
and focus (cf. 12.7). As we saw, every statement can be seen as providing
an answer to either an explicit or implicit question. A thematically
marked (i.e. non-neutral) statement is one that provides an answer to an
explicit or implicit x-question that carries with it (or would carry with it
if it were made explicit) certain determinable presuppositions. For
example, the statement
(17) fohn is working in the study
(with heavy stress on the form study) answers the question
(18) Where 1s John working?
And (18) presupposes that John is working somewhere (cf. Chomsky,
1969). Very similar to (17), as far as the distinction between focus and
presupposition is concerned, are
(19) It’s in the study that John is working
(20) The study ts where Fohn 1s working
and various alternative phonologically and grammatically marked
utterances that are revealed as such by the devices (stress, intonation,
word-order, the so-called cleft-sentence construction, etc.) that English
makes available for indicating thematic structure. Thematically marked
statements have the same presuppositions, then, as do the explicit or
implicit questions to which they provide an answer; and these remain
constant under negation and interrogation (cf. John isn’t working in his
stidy, Is fohn working in his study?).

® For a different view cf. Hull (1975), Keenan & Hull (1973).
 14.3. Conversational implicatures and presupposition 599
A fourth important class of utterances in connexion with which the
notion of presupposition has been invoked recently by linguists is that
of utterances containing so-called factive* verbs (cf. 17.2). Anyone who
says
(21) fohn realizes that tt is raining
(in order to make a statement) is committed by his use of the verb
‘realize’ to the truth of the proposition expressed by the complement-
clause: he presupposes that it is raining. There are very many English
verbs, notably ‘know’, that are factive in this sense; and most of them
(but not ‘know’) can be grouped together in terms of several syntactic
characteristics that they share (cf. Kiparsky & Kiparsky, 1970). This
kind of presupposition, it will be observed, also holds constant under
negation and interrogation (except under certain conditions of context-
dependency that we will temporarily disregard). Hence the peculiarity
of an utterance like
(22) I don’t know that it 1s raining
(with don’t unstressed) — provided that it is construed as meaning
(23) “That it is raining is a fact of which I am unaware’’,
rather than
(24) “‘I am inclined to doubt that it is raining’’.’
If (22) is construed as meaning (24), it is, of course, perfectly acceptable.
The four kinds of presupposition that have been mentioned so far
differ from one another in various ways. But each of them can be seen as
involving a fairly natural sense of the pre-theoretical term ‘ presupposi-
tion’. In each case it is reasonable to say that the speaker, in making an
assertion or asking a question, assumes or presupposes that something is
so. For example, if X says to Y
(25) Why does God tolerate man’s wickedness?
’ There are circumstances in which (22) is interpretable as meaning (23): cf.
Let us suppose (for the sake of argument) that I don’t know that it’s raining, in
the utterance of which the speaker may be well aware of the fact that it is
raining and yet invite his addressee to operate, hypothetically, with the
assumption that he, the speaker, is not aware of this fact. In such circum-
stances (22) is not peculiar under this interpretation; otherwise it is. As for
(24): it is perhaps only in certain dialects of non-literary English that (22)
will support this interpretation ; and when it does, it may have a characteristic
rhythm and intonation-contour.
 600 Context, style and culture
Y can quite reasonably say, in ordinary non-technical English,
(26) I do not accept the presuppositions that you are making
or
(27) I do not accept the presuppositions of your question.
In saying either (26) or (27), Y may be challenging any or all of the
following propositions (and perhaps others): ‘“‘God exists”’, ‘Mankind
is wicked’’, ‘‘God tolerates man’s wickedness’’. The difference between
(26) and (27), it will be noted, is that, whereas (26) treats presupposition
as a relation between persons and propositions (i.e. what they hold to
be true and would be prepared to assert), (27) treats presupposition as a
relation between utterances and propositions. Given that the verb
‘presuppose’, in its pre-theoretical sense, is more or less synonymous
with ‘assume’, we can perhaps legitimately infer that, in its pre-
theoretical sense at least, ‘presuppose’, like ‘assume’, denotes primarily
a relation between person and utterances (i.e. utterance-signals as tokens
of certain types: cf. 1.6). In other words, ‘presuppose’, in its pre-
theoretical sense, would seem to be primarily a verb of propositional
attitude: it seems to be more like ‘assume’ (or “believe’) than ‘entail’.
It is hardly surprising that the truth-conditional definition of pre-
supposition, in which presupposition is taken to be a relation that is
logically comparable with entailment, should fail to apply to everything
that falls within the scope of the pre-theoretical notion of presupposition.
Faced with the fact that the truth-conditional definition of presupposi-
tion, which, for the moment at least, we are assuming to be applicable to
statements like The King of France 1s bald, cannot be applied to the
other kinds of presupposition, some scholars have opted for a distinction
between so-called semantic presupposition, defined in terms of truth-
conditions, and so-called pragmatic presupposition (cf. Keenan, 1971).®
However, in attempting to formalize the so-called semantic notion of
presupposition in such a way that it 1s both coherent and distinct from —
entailment, they have been forced to extend the two-valued propositional
calculus. One extension consists in admitting a third truth-value into
the system. Another, which comes closer perhaps to formalizing the
notion of presupposition that Strawson invoked against Russell, consists
: in allowing formally for the possibility of truth-value gaps*: i.e. the
8 The term ‘so-called’ is used here simply to draw attention to the fact that in
the present work ‘semantic’ is not restricted, in opposition with ‘pragmatic’,
to what can be handled by means of a truth-conditional theory (cf. 4.4, 6.6,
14.2).
 14.3. Conversational implicatures and presupposition 601
possibility that certain statements should lack a truth-value. We will not
go into either of these alternatives. It has been cogently argued, recently,
that the so-called semantic approach to the definition of presupposition
fails to handle certain crucial cases without otherwise unmotivated
adjustments and qualifications (cf. Kempson, 1975; Wilson, 1975). It
has also been argued that much of what proponents of truth-conditional
definitions of presupposition have taken to be presupposition is in fact
entailment.
So far we have assumed that the truth-conditional definition of pre-
supposition is indeed applicable to such statements as The King of
France is bald. But Russell, of course, took the view that the proposition
it expresses is false; and there are many that would agree with him.
What is not emphasized as strongly as it ought to be in linguistic treat-
ments of presupposition is that it is rather pointless arguing whether a
statement has a determinable truth-value or not, unless we know what
statement we are in fact discussing and what its thematic structure 1s.
Not only is this rarely made clear, but the notion of presupposition has
all too often been discussed, by linguists and by philosophers, in terms of
sentences, rather than utterances. As we saw earlier, several thematically
distinct statements (as well as several thematically distinct utterances of
other kinds) may be put into correspondence with the same system-
sentence (cf. 12.7). What is being said about what — what comment is
being made about what topic — depends upon the thematic structure of
the utterance. In so far as there is any pre-theoretical dispute as to
whether the proposition expressed by The King of France 1s bald is false,
on the one hand, or neither true nor false, on the other, this can be
explained, at least partly, by the possibility of taking ‘the King of
France’, in different tokens of this utterance-type, to be thematic or not.
If it had been previously asserted by X that no currently reigning
European monarch happened to be bald and then Y said, in all serious-
ness, The King of France is bald, X could quite reasonably retort That’s
not true — there is no King of France. Even if it turned out that Y was
referring to Giscard d’Estaing, it would be reasonable, in this context, to
say that what he said was false, by virtue of the failure of its existential
presupposition. The reason is that, in the context that we have just
constructed, ‘the King of France’ is not the theme. It has been asserted
by X that the class of reigning European monarchs contains no bald-
headed member. Y’s counter-assertion that this class contains the King
of France is reasonably described as false, independently of whether the
person that Y is referring to happens to be bald or not. For the point at
602 Context, style and culture
issue, in this context, is not whether a particular person is bald or not,
but whether there are any bald-headed, currently reigning European
monarchs (cf. Cooper, 1974: 36ff). It is only when a referring expression
is thematic that failure of the existential presupposition results in what
Strawson and those who take the same view as he does would call a
truth-value gap; and Strawson himself appears to have come, more
recently, to the same conclusion (cf. Strawson, 1964b).
One further point is worth making about The King of France is bald.
This is that, if ‘the King of France’ is construed, not as a definite
description, but as a title whose relationship to its bearer is like that of a
name to its bearer (cf. 7.5), the existential presupposition is one that
cannot be captured by means of the proposition “‘There is a king of
France’’.® The fact that this is so further limits the usefulness of truth-
conditional definitions of presupposition, even in respect of the existen-
tial presupposition of referring expressions, for which, at first sight, they
seem to be especially appropriate. Names are not true of their bearers in
the way that expressions that denote the defining property of a class are
true of the members of that class. And yet, when names (and titles that
are arbitrarily associated with their bearers) are used as referring expres-
sions, they do not differ from definite descriptions as far as their existen-
tial presuppositions are concerned. What counts is whether there is a
referent that is appropriately referred to by means of the expression in
question. This is a more general condition than is the satisfaction of a
particular set of truth-conditions.
We will say no more about such classic examples as The King of
France is bald. 'The existential presuppositions of referring expressions
were fully discussed in a previous chapter (7.2): it suffices here to re-
emphasize the importance of drawing a distinction between correct and
successful reference and to insist upon the fact that reference is always,
in principle, context-dependent.
So too, it would seem, is any linguistically useful notion of presupposi-
tion. Given that the truth-conditional definition of presupposition is, to
say the least, of very restricted coverage and cannot be applied to actual
or potential utterances unless certain assumptions are made about the
thematic structure of the utterances and about the contexts in which

® Suppose, for example, that, although France was still a republic, X had
conferred upon him (not necessarily by any official institution in France) the
courtesy-title ‘the King of France’. X could then be correctly referred to by
means of this expression and the proposition “‘X is the King of France”
would be true, but not “X is (a) king of France’”’, and still less ‘‘ X is (a) king’’.
 14.3. Conversational implicatures and presupposition 603
they occur, there would seem to be little point in drawing a theoretical
distinction between two kinds of presupposition in terms of the dis- :
tinction between semantics and pragmatics — a distinction which is, in
general, of doubtful value as far as the analysis of the structure of natural
languages is concerned (cf. 4.4).
In addition to the four kinds of presupposition that have been men-
tioned so far, there are others that have been discussed recently by
linguists, which further extend the notion. McCawley (1968) has said
that the adjective ‘buxom’ carries with it the presupposition that who-
ever it is applied to is female, so that
(28) My neighbour is buxom
will be understood as implying that the referent of ‘my neighbour’ is |
female. Similarly, it has been suggested, the meaning of ‘bachelor’ can
be split into two parts: what is presupposed, that the entity to which
‘bachelor’ ts applied is male, adult and human, and what is asserted,
that the entity in question 1s not married (cf. Fillmore, 1971a). G. Lakoff
(19714) has said of
(29) John told Mary that she was ugly and then shé insulted him
(where she and him bear heavy stress) that it carries the presupposition
that to tell someone that she is ugly is to insult her. Keenan (1971) says
of the French utterance
(30) Tu es dégotitant
(“You are disgusting’) that it (pragmatically) presupposes that ‘‘the
addressee is an animal, child, socially inferior to the speaker, or per-
sonally intimate with the speaker’’. Fillmore (1971b) has said of |
(31) John accused Harry of writing the editorial
that it presupposes that John regarded the writing of the editorial as
something reprehensible; and of
(32) Please open the door
that it presupposes that, at the time of the utterance, the door is shut and
the addressee is in a position to comply with the request that is addressed
| to him. |
This is a somewhat heterogeneous set of examples. In each case it is
reasonable to say that the term ‘ presupposition’ is being used in a way
that is consistent with its everyday pre-theoretical sense. But any
 604 Context, style and culture
theoretical concept of presupposition which covers all these cases of
what may be pre-theoretically classified as presupposition is likely to be
too broad to be of any real value. It might be suggested, for example, that
the presuppositions of an utterance are the conditions that it must
satisfy, if it is to be interpretable and appropriate, in the context in
which it occurs. This kind of definition would certainly cover everything
that has been classified in the recent literature as presupposition. But it
would cover much else besides, including everything in the context that
determines the form or interpretation of an utterance.
One way of narrowing the definition is by talking of propositions to
whose truth the speaker is committed, rather than of conditions that the
utterance must satisfy. We might say, for example, that an utterance
presupposes a proposition p if and only if the speaker assumes that p is
true and assumes that the addressee also assumes that p is true (cf.
Karttunen, 1973). The problem then arises as to what is meant by
assuming that a proposition is true. For example, one can presumably
assume that one’s addressee 1s one’s social inferior and demonstrate by
one’s behaviour that one has made this assumption without having
entertained at any time the specific proposition “The addressee is my
social inferior’’. The point is that there is a distinction to be drawn, in
principle, between the belief that something is so and the belief that a
certain proposition is true. We will not go into the problems of making
this distinction precise. It is worth pointing out, however, that, in saying
(21) fohn realizes that it 1s raining, the speaker 1s committed to the truth
of the specific proposition ‘It is raining’’, which 1s part of the propo-
sitional content of the utterance. But, even if we feel entitled to say that
by uttering (29) John told Mary that she was ugly and then shé insulted
him the speaker commits himself to the truth of a proposition, we cannot
be sure what proposition this is. Is it “’T’o tell someone that he is ugly
constitutes an insult” or “To tell a girl that she is ugly constitutes an
insult’’? Or is it some other proposition, more general or more specific?
There is no way of telling from the utterance itself and the meaning of
_ its verbal and non-verbal component.
_ Most of the definitions of presupposition to be found in the recent
literature take the presuppositions of an utterance to be a set of propo-
sitions. An alternative (though not necessarily incompatible) view is that
they are the conditions that must be satisfied before the utterance can be
used felicitously to perform its function as a statement, a question, a
promise, a request, etc. (cf. Fillmore, 1971b). This notion of the
felicity-conditions* of an utterance is something that we shall come back
 14.3. Conversational implicatures and presupposition - 605
to (cf. 16.1). Here it is sufficient to make two points. The first is that the
felicity-conditions of an utterance need not be described as propositions
_ to whose truth the speaker subscribes, though they can be, and fre-
quently are, so described. The second, and more important, point is
that, in saying that the presuppositions of an utterance are necessary
conditions for its felicitous use, we are still operating with a very broad
notion of presupposition, unless we distinguish between various kinds
of felicity-conditions. It has been argued by Cooper (1974) that the
conditions that count as presuppositions are all ontological, in that they
have to do, not necessarily with existence, but with whatever kind of
ontological satisfaction is appropriate to the entity, state-of-affairs, event,
process, etc., in question. ‘I'he existential presuppositions of expressions
that refer to individuals (first-order entities: cf. 11.3) would thus be no
different in kind from the presuppositions of utterances containing fac-
tive verbs. Referring expressions presuppose that certain entities exist;
and existence is the ontological condition for first-order entities: factive
utterances presuppose that certain states-of-affairs obtain, that certain
events occur, etc.; and obtaining, occurrence, etc., is the ontological
condition for states-of-affairs, events, etc. (which, in so far as they may
be referred to as entities in particular languages, are second-order
entities: cf. 11.3).
This view of presupposition has the advantage that it provides a.
unified and theoretically motivated account of most of what has been
considered, pre-theoretically, to be a case of presupposition. It does so
by emphasizing reality rather than truth. For example, instead of saying
that (29) presupposes the truth of some specific proposition, it says that
(29) presupposes that a certain event (viz. John’s insulting Mary) took
place. Similarly, instead of saying that (g) presupposes the truth of
‘There is-a King of France’’, it says that (9) presupposes the existence
of some entity that is identifiable (in context) by means of the expres-
sion ‘the King of France’; and, as we have seen, this is a more defensible
point of view.
What has just been said of presuppositions can also be said of im-
plicatures. Earlier in this section we described the implicata of an
utterance as propositions. But it will now be obvious that it is often more
plausible to say that a speaker implicates that something exists, is so or
has occurred than that he implicates some determinable set of specific
propositions. What then is the difference between implicature and
presupposition?
Pre-theoretically, the difference would seem to be that, whereas what
 606 Context, style and culture
is presupposed is what the speaker takes for granted and assumes that
the addressee will take for granted as part of the contextual background,
what is implicated 1s what the addressee can reasonably infer, but is not
necessarily intended to infer, in the context in which the utterance
occurs, from what 1s said or 1s not said. There is nothing in this pre-
theoretical account of the difference between them, it will be observed,
to prohibit the possibility of one and the same fact being both pre-
supposed and implicated. Hence the various attempts that have been
made recently to subsume presuppositions under the notion of implica-
ture and to account for their presence in terms of Grice’s maxims of
quantity, quality, relation and manner. So far, however, there is far from
being general agreement as to the feasibility of accounting for pre-
supposition in this way.
It is generally agreed that implicatures can be cancelled or qualified
in particular contexts. If it is conceded that presuppositions cannot (and
our pre-theoretical characterization of presupposition would suggest
that they cannot), there would be at least this difference between
implicatures and presuppositions. This difference has been challenged
by several scholars (cf. Wilson, 1975). It has been challenged, however,
on the basis of a truth-conditional theory of semantics, a controversial
view of negation and entailment, and the failure, or refusal, to draw a
distinction between sentences and utterances, on the one hand, and
between propositions and facts, on the other. Needless to say, most of
the argumentation in this area (including such argumentation as there
has been in this section) is very heavily theory-dependent. That being so,
it is almost impossible to compare one view of presupposition with
another, within a common terminological and conceptual framework,
without thereby prejudicing the decision one way or the other. What has
been said about implicature and presupposition in this section is no
more than a very general, non-technical and, for the most part, pre-
theoretical introduction to the two notions. There is by now a quite

these notions.1° ,
considerable technical literature devoted to the problems of formalizing

10 Entry to this literature may be made by means of such works as Cooper

(1974), Ducrot (1972), Franck & Petéfi (1973), Garner (1971), Gazdar (1976),.
Karttunen (1973, 1974), Katz (1973), Kempson (1975), Wilson (1975), Zuber
(1972). Of particular importance in the more technical treatments of pre-
supposition has been the so-called projection-problem (cf. Langendoen,
1971): 1.e. the problem of determining the presuppositions of complex pro-
positions in terms of the presuppositions of their component simple propo-
sitions.
 Context, style and culture 607
14.4. The contextual theory of meaning
There are several senses in which theories of meaning might be classified
as contextual. The term ‘the contextual theory of meaning’ is being used
here, as it is often used by linguists in Great Britain, with particular
reference to the so-called Firthian theory of meaning: 1.e. to the theory
developed by J. R. Firth, initially in association with the famous
anthropologist Malinowski (1930, 1935 — cf. Firth, 1957b), and further
elaborated by his followers.4
There are those who would deny that Firth ever developed anything
systematic enough to be described as a theory; and Firth himself, who
was suspicious of what purported to be systematic and well-articulated
theories of the structure of language, might have been only too happy to
agree, without however taking this to be a criticism of his approach to
the study of language. We will not stop to debate the terminological
issue of what constitutes a theory. The Firthian view of meaning has
been influential; and it has something of value to contribute to what
might ultimately count as a comprehensive and materially, as well as
formally, adequate theory of semantics. Since no satisfactory formal
theory of meaning has yet been proposed by anyone, the semanticist
cannot afford to discount the insights and suggestions of someone like
Firth, who was sceptical of the value of formalization.
The first point that must be made is that in discussing the Firthian
view of meaning one is concerned with an all-embracing functionalist
view of language, and not merely with semantics as the term ‘semantics’
is customarily interpreted. According to Firth, the most important thing
about language is its social function: “normal linguistic behaviour as
a whole is meaningful effort, directed towards the maintenance of
appropriate patterns of life’ (Firth, 1957a: 225). Every utterance occurs
in a culturally determined context-of-situation* ; and the meaning of the
utterance is the totality of its contribution to the maintenance of what
Firth here refers to as the patterns of life in the society in which the
speaker lives and to the affirmation of the speaker’s role and personality
within the society. In so far as any feature of an utterance-signal can be
said to contribute an identifiable part of the total meaning of the
utterance, it can be said to be meaningful. It follows that, not only words
and phrases, but also speech-sounds and the paralinguistic and prosodic
features of utterances, are meaningful (cf. 3.1). These meaningful
11 Cf, Ellis (1966), Halliday (1966), McIntosh (1961), Mitchell (1975), Sinclair
(1966). Much of the present section is based upon Lyons (1966). For a more
comprehensive account of Firth’s views, cf. Robins (1971).
 608 Context, style and culture
components of utterances are abstracted from the data by a careful study
of the contrasts that hold between utterances in the contexts-of-situation
in which they occur. And the meaning of each component — paralinguis-
tic, phonological, grammatical, lexical, etc. — is described in terms of its
function as an element in the structure of units of the level above. The
structures of the higher-level units are the contexts in which the lower-
level units function and have meaning. Semantics, in the Firthian use
of the term, relates utterances to their context-of-situation; but all
branches of linguistics necessarily deal with meaning. There is nothing
tautological, therefore, about the Firthian phrase ‘semantic meaning’,
and there is nothing contradictory, or otherwise anomalous, about such
phrases as ‘phonetic meaning’ or ‘grammatical meaning’.
Rather more puzzling, at first sight, are statements to the effect that
‘voice quality is part of the mode of meaning of an English boy, a
Frenchman, or a lady from New York” and that “‘it is part of the
meaning of an American to sound like one”’ (Firth, 1957: 191-2, 225-6).
Statements like this might seem to depend upon a perverse and wilful
extension of the term ‘meaning’; and there is little doubt that Firth
delighted in the shock-effect of such formulations of what he meant by
‘meaning’. But they are consistent with his general view that being
meaningful, or having meaning, is a matter of functioning appropriately
(i.e. significantly) in context. To speak with an American accent is to
indicate that one 1s an American; and, in so far as speaking with an
American accent 1s the result of one’s socialization as an American and
part of one’s present state of being an American, it makes sense to say
that in speaking with an American accent one is simultaneously being an
American and meaning that one is an American. Looked at from a social
| and behavioural point of view, one’s modes of being are one’s modes of
meaning; and one means what one is (or, alternatively and equivalently,
one is what one means) by behaving in such-and-such a way in one’s
context.
There may well be some equivocation here with the term ‘meaning’
(cf. 1.1). It should not be forgotten, however, that a number of philoso-
phers, notably Grice (1968), have taken the view that there is an
intrinsic connexion between what a person means by his utterance and
what his utterance means, the latter being explicable ultimately in
terms of the former. Firth would have taken the same view. But, unlike
Grice and most philosophers, Firth was more interested in the social and
expressive (or indexical) functions of language than he was in its
descriptive and conative (or directive) functions (cf. 2.4, 4.2). Like
 14.4. The contextual theory of meaning 609
Malinowski, he tended to treat the descriptive, and to a lesser extent,
perhaps, the conative, function of language as something that was
subsidiary to, and part of, the more basic and more general function of
maintaining the appropriate patterns of life. It is at least arguable that
Firth’s view of meaning is no more distorted, if distorted it 1s, than the
more common dualistic view that the meaning of a word or an utterance
is what the word or utterance signifies (cf. 4.1). However that may be,
Firth’s use of the term ‘meaning’, idiosyncratic though it undoubtedly
is at times, 1s not as perverse and unmotivated as it appears at first sight.
‘‘Meaning ... is to be regarded as a complex of contextual relations, and
phonetics, grammar, lexicology, and semantics each handles its own
components of the complex in its appropriate context”’ (Firth, 1957: 19).
The analysis of the meaning of an utterance consists in abstracting it
from its actual context-of-utterance (cf. 14.1) and splitting up its mean-
ing, or function, into a series of component functions. This process of
analysis 1s, on occasion, explained by way of analogy: “‘the suggested
procedure for dealing with meaning is its dispersion into modes, rather
like the dispersion of light of mixed wave-lengths into a spectrum”
(Firth, 1957: 192). The analogy, in itself, is not very helpful. But it does
serve to bring out the fact that Firth thinks of the meaning of an utter-
ance as something within which the components are blended in such a
way that they are not recognizable as distinct until they have been dis-
_ persed into modes by linguistic analysis.
The key term in the Firthian theory of meaning is, of course, ‘con-
text’. The analysis of the meaning of an utterance will consist in ‘“‘a
serial contextualization of our facts, context within context, each one
being a function, an organ of the bigger context and all contexts finding
a place in what might be called the context of culture”’ (Firth, 19574:
32). The context-of-culture, which Firth appeals to here, is postulated
as the matrix within which distinguishable and socially significant
situations occur. By invoking the concept of the context-of-culture
(which, like that of the context-of-situation, derives from his collabora-
tion with Malinowski), Firth commits himself, as many linguists of his
generation did, to the view that there is an intimate connexion between
language and culture. But he never committed himself to anything like
the Whorfian hypothesis (cf. 8.3). Neither he nor his followers have been
much concerned with epistemological and ontological questions. Their
main purpose has been to emphasize that language-utterances, like other
bits of socially significant behaviour, could not be interpreted otherwise
than by contextualizing them in relation to a particular culture. It is
610 Context, style and culture
because Firth cannot envisage such sentences as Jespersen’s ‘A dancing
woman charms’ or Sapir’s “The farmer kills the duckling’ ever being
employed in some actual context of use that he describes them as non-
sense (Firth, 1957a: 24). They cannot be contextualized: they cannot
“be referred to typical participants in some generalized context of
situation”’ (Firth, 1957a: 226). They may be grammatically meaningful ;
and yet, if they do not have what Firth refers to as the implication of
utterance in some culturally acceptable and interpretable situation, they
will not be meaningful at the semantic level of analysis.
It is no part of our purpose to defend the Firthian theory of meaning
in all its details. Indeed, it would be difficult to do so in view of the
obscurity of certain key passages in his works. For example, it is not
clear how non-deictic reference, on the one hand, and denotation, on the
other, would be handled by means of the Firthian notion of function in
context. Deictic reference is more or less plausibly accounted for in
terms of the establishment of correlations between linguistic expressions
and entities in the context-of-situation (cf. 15.2). But it is difficult to see
how this kind of account can be extended to cover a potentially infinite
set of non-deictic referring expressions without re-introducing, though
possibly in reduced measure, something of the dualism that Firth
objected to in traditional theories of meaning (Firth, 1957a: 217, 227).
Having said that, however, one must also admit that Firth’s so-called
monistic theory of meaning constitutes a healthy reaction against the
excessive and essentially empty conceptualism of traditional approaches
to semantics (cf. 4.3).
Contextualization can be looked at from two points of view. We can
think of it as the process whereby the native speaker of a language pro-
duces contextually appropriate and internally coherent utterances — a
process which, as we have seen, involves a lot more than knowledge of
the language-system (cf. 14.2). We can also think of it as a process which
the linguist carries out in his description of particular languages. In so
far as the semantic analysis of a particular language is descriptively
adequate, in Chomsky’s (1965: 27) sense, there must be some corres-
pondence between these two kinds of contextualization: the factors
identified by the linguist as contextual must be the factors that deter-
mine the native speaker’s production and interpretation of utterances in
actual situations of use. The term ‘contextualization’ is used by Firth
with respect to what the linguist does in describing a language; and, like
most linguists of his generation, he was not concerned with what would
now be called descriptive adequacy. We shall continue to use the term
 14.4. The contextual theory of meaning 611
‘contextualization’ both of what the native speaker does in the use of
language and of what the linguist does in describing the underlying
system of elements, rules and principles by virtue of which the native
speaker is able to create (and interpret) what Halliday (1970b) and others
refer to as text* (cf. 14.6).
One way of approaching the analysis of context, as we have seen, is
by asking what kind of knowledge the native speaker of a language must
possess in order for him to be able to create and understand texts in that
language (cf. 14.2). It is important to realize, however, that this know-
ledge need not be propositional. This point should be borne in mind, in
view of the tendency to define context, in work that falls within the area
of what is now commonly referred to as pragmatics, as a set of proposi-
tions (cf. Bar-Hillel, 1971). To say, as someone who is committed to a
contextual theory of meaning might say, that to know the meaning of an
utterance, a word, an intonation-pattern, etc., is to know the contexts in
which it can occur is not necessarily to impute to the person of whom it is
said that he knows the meaning of an utterance, a word, an intonation-
pattern, etc., the knowledge of a set of propositions.
The same point might also be made, incidentally, with respect to the
truth-conditional theory of meaning (cf. 6.5). Granted that it is plausible
to say that to know the meaning of a statement is to know what the world
must be like for that statement to be true, it does not follow that this
knowledge is, in all cases, propositional. There is a certain vagueness, not
to say equivocation, that frequently creeps into expositions of the truth-
conditional theory of semantics in this respect. It would seem that one
can, in principle, know that it is raining (or what the world was like
when it was last raining and what it will be like when it is next raining)
without being able to interpret, still less give one’s assent to, the propo-
sition ‘‘It is raining’’. Presumably, many species of animals demonstrate
this non-propositional knowledge of what we might be quite willing to
describe as facts. There is, therefore, a sense in which they know the
meaning of such propositions as “It is raining’’: they can distinguish
the possible worlds in which they are true from the possible worlds in
which they are false. We shall, in fact, make use of the notion of the
. propositional content of utterances in later chapters: we shall make no
assumptions, however, about the epistemological or psychological status
of propositions.
There is no conflict, in principle, between the contextual theory of
meaning and the truth-conditional theory of meaning; and it is arguable
that what is required is a more comprehensive theory which subsumes
 612 Context, style and culture
both. For it is as reasonable to say that someone does not know the
meaning of a word or an expression on the ground that he cannot con-
textualize it as it is to say that he does not know the meaning of a word
or expression on the ground that he does not know its truth-conditions.
If we accept that this is so we are unlikely to press either the contextual
theory of meaning or the truth-conditional theory of meaning too hard.
As we have seen, in a previous section, a case can be made for the seman-
tic analysis of maximally decontextualized system-sentences in terms of
their truth-conditions (cf. 14.2). This does not mean, however, that
context is of secondary importance — something to be appealed to only
when a truth-conditional account of the meaning of a sentence fails or is
inapplicable. :
In his later work, Firth introduces the notion of collocation* as part of
his overall theory of meaning (cf. Firth, 1957a: 197). It is at the so-called
collocational level of analysis, intermediate between the situational and
the grammatical, that he proposes to deal, in whole or in part, with
lexical meaning: 1.e. with that part of the meaning of lexemes which
depends, not upon their function in particular contexts-of-situation, but
upon their tendency to co-occur in texts. He tells us, for example, that
‘fone of the meanings of ‘night’ 1s its collocability with ‘dark’ and of
‘dark’, of course, collocation with ‘night’”’ (Firth, 1957a: 197). He also
talks, in the same passage, of “the association of synonyms, contraries
and complementary couples in one collocation”’; and elsewhere, of such
‘‘ordered series of words’? and “paradigms, formal scatter, so called
synonymous and antonymous, lexical groups by association, words
grouped by common application in certain recurrent contexts of situa-
tion” (Firth, 1957a: 228). Exactly what Firth meant by collocability is
never made clear. It may nonetheless be helpful to refer in this connexion
to the so-called distributional theory of meaning.
According to at least one version of the distributional theory of mean-
ing (to which Firth may or may not have subscribed) two lexemes will
have the same meaning if and only if they have the same distribution
throughout a representative sample of texts (cf. Harris, 1951). All that
needs to be said about this thesis is that no convincing reason has ever
been given for believing that sameness of lexical meaning defined in this
way will be in correspondence with what is pre-theoretically taken to be
sameness of meaning. For example, from a pre-theoretical point of view
The milk has turned and The milk has gone sour would seem to be very
similar, if not identical, in meaning. The distribution of the intransitive
verb ‘turn’ is very different, however, from the distribution of ‘go sour’.
 14.5. Stylistic, dialectal and diachronic variation 613
It might be argued, of course, that ‘turn’ has several meanings and that
it is only when it has a particular one of these meanings that it has the
same distribution and the same meaning as ‘go sour’. But this cannot
be shown to be true or false unless there is some other way of deter-
mining sameness and difference of meaning. That there should be a
fairly high correlation between sameness of meaning and sameness of
distribution is only to be expected; and the ordinary view of the matter
would be that distributional similarity is the result, rather than the
cause, of similarity of meaning. What is theoretically interesting is the
fact that the distribution of lexical items is not always fully determined
by their sense and denotation; and this fact runs counter to the dis-
tributional theory of meaning.
The distributional theory of meaning as such may be rejected. At the
same time, it must be admitted that there is frequently so high a degree
of interdependence between lexemes which tend to occur in texts in
collocation with one another that their potentiality for collocation is
reasonably described as being part of their meaning. For example, the
collocation of “bandy’ with ‘leg’ (usually in the plural) could hardly be
accounted for in terms of some specification of the meaning of ‘bandy’
which did not incorporate a mention of its collocability with ‘leg’; and
there are many such examples in all languages. At the very least the
notion of collocability is an important corrective to an excessive reliance
upon the dualistic notion of signification.
No more will be said here about the Firthian notion of collocability or
about the contextual theory of meaning in general. The importance of
giving full weight to syntagmatic lexical relations in the language-
system has been emphasized in a previous chapter (cf. 8.5); and various
aspects of contextualization are dealt with in this and the following
chapter. However, there is much pertaining to the analysis of context
that is not discussed at all in this book, which is limited, for the most
part, to microlinguistic semantics (cf. 14.2).

14.5. Stylistic, dialectal and diachronic variation

In this section we shall be concerned, though only cursorily, with the
_ semantically relevant aspects of stylistic, dialectal and diachronic
variation. We may start with the uncontroversial, but unhelpful, state-
ment that, as semantics is the study of meaning, so stylistics* is the
study of style.
The term ‘style’ is used, non-technically, in a variety of senses. It may
be used to refer to the kind of systematic variation in texts that 1s
 614 Context, style and culture
covered by such terms as ‘formal’, ‘colloquial’, ‘pedantic’, etc.; and
this sense of ‘style’ gives rise to one very broad definition of stylistics,
“the description of the linguistic characteristics of all situationally-
restricted uses of language’’ (Crystal & Davy, 1969: go). Stylistics, under
this interpretation of the term, will merge with what others may wish to
call sociolinguistics or pragmatics; but it will be subsumed under
semantics, according to the Firthian definition of meaning (cf. 14.4).
The term ‘style’ is also used to refer to those features of a text, and
more especially of a literary text, which identify it as being the product
of a particular author. We talk, for example, of the style of Jane Austen
as being characteristically different from that of Charlotte Bronté; of _
the odes of Propertius as being recognizably different in style from those
of Horace or Tibullus; and so on. The term ‘stylistics’ is frequently
restricted to the analysis of literary texts from this point of view (cf.
Chatman, 1971). Since the identification of a literary text as the work of
a particular author is not generally regarded as an end in itself, but is
usually coupled with, or made subsidiary to, the determination of those
features of the text which produce a particular effect upon the reader,
literary stylistics, under this interpretation of the term ‘stylistics’,
merges with what was traditionally called rhetoric.
It is easy to see that there are other uses of the word ‘style’ in every-
day, non-technical, English which bridge the two senses of ‘style’
distinguished above; and ‘stylistics’, at its broadest, can be held to cover
all of these. No attempt will be made to give even a summary account of
stylistic variation in language. Much of what would be counted as
stylistic variation, under a broad definition of ‘style’, has been dealt
with in terms of situationally and socially determined variation, and of
thematic structure (cf. 14.2, 12.7), What we are concerned with here is
whether a sharp distinction can be drawn between semantic and
stylistic variation, on the one hand, and between stylistic and non-
stylistic variation, on the other.
One way of drawing the distinction between semantics and stylistics
is in terms of the kind of meaning that is involved (cf. 2.4). For example,
the statement that “‘stylistics 1s concerned with the expressive and
evocative values of language’’ (Ullmann, 1962: 9) rests upon the widely
held view that the so-called cognitive and non-cognitive aspects of
meaning are analytically separable. The same view is reflected, though
it is rarely made explicit, in the distinction that Chomsky and his fol-
lowers draw between stylistic and non-stylistic variation. The thesis that
transformational rules do not change meaning, put forward by Katz and
 14.5. Stylistic, dialectal and diachronic variation 615
Postal (1964) and adopted by Chomsky (1965) in the so-called standard
theory of transformational grammar, was interpreted in terms of an
intuitive distinction between meaning and style (cf. 10.5): transforma-
tions that were postulated in order to account for what was classified as
purely stylistic variation (cf. ‘I gave John the book’ ws. ‘I gave the book
to John’, ‘John came’ ws. ‘It was John that came’, etc.) were made
optional; and they were not counted as meaning-changing (cf. Partee,
1971). Enough has been said elsewhere in this book about the distinction
between what in our terminology is called descriptive and non-descrip-
tive meaning and about the prejudice that is involved in assuming that
non-descriptive meaning is less basic than descriptive meaning. Here it
may be simply noted that, in the works of those whose primary concern
is with descriptive meaning, ‘stylistic’ is often no more than a residual
catch-all term for every kind of synchronic variation within what is
assumed to be a single dialect.
It is of considerable importance, however, to draw a distinction, at
least in principle, between the kind of stylistic variation that is deter-
mined by the communicative intentions of the speaker or writer, by his
social role and status, and by other factors in the context-of-situation and
the kind of stylistic variation that is undetermined by such factors.
Independently of jother considerations, an individual is restricted in
expressing his individuality, as distinct from indicating his membership
of one social group rather than another and his more immediate com-
municative intentions, to the choices that he makes within the range of
situationally and socially undetermined variation. As we saw earlier, it
is not only voice-quality and paralinguistic features that serve as
individual-identifying indices, but also such characteristics as the em-
ployment of a particular form or lexeme or the use of a particular
grammatical construction; and such characteristics may be either purely
indexical* (in the sense in which the term is being used in this book) or
both indexical and expressive* (cf. 4.2). If it is part of the meaning of an
American to sound like one, it is also part of his meaning to sound like
himself (cf. 14.4); and the language-system will provide the means for
him to indicate* who he is, whether he indicates this intentionally or not,
by allowing for variation that is undetermined by other functional or
social factors.
The very concept of a language-system, in the sense in which lin-
guists claim to be studying the system that underlies the language-
behaviour of members of particular communities, rests upon the kind of
idealization that we have called standardization* (cf. 14.2). This involves
616 Context, style and culture
the more or less deliberate discounting of dialectal* and diachronic*
variation within the community whose language is being described.
There is no need to enter here into a full discussion of the nature and
significance of dialectal and diachronic variation. We will simply note a
few points that are relevant to semantics.
In everyday usage the term ‘dialect’ is usually associated with regional
variation. In most countries, however, there is socially determined, as
well as regionally determined, variation, and the term ‘dialect’ is com-
monly extended by linguists to cover both. It is all the more convenient
to have a single term for both in that there is often a considerable degree
of interdependence between them. For example, what is popularly
known as Cockney is not the dialect of a particular part of London as |
such; it is the dialect of a particular social class living in a particular part
of London. It has often been pointed out that in England, more so than
in many other countries, there is very little dialectal variation in the
speech of typical members of what may be identified, in socio-economic
terms, as the upper classes. At most, there are minor differences of
accent* within what, from the point of view of its lexical and gram-
matical structure, is the same dialect: Standard English. Very often,
when the term ‘dialect’ is employed nowadays with reference to the
English of England, it is little more than accent or pronunciation that is
at issue. Gross dialectal differences of the kind that distinguished the
language-systems of different regions in the past have disappeared in all
but the remotest rural communities. Such minor differences of dialect
as remain, though they may be discounted by the linguist, in his con-
struction of a model of the overall language-system, may serve the same
function as the more striking dialect differences did in the past in
England and still do in many countries.
Differences of dialect and accent have an important indexical func-
tion; and this is the most obvious reason why they are of interest to the
semanticist. hey indicate the speaker’s membership of a particular
social or regional community — his solidarity with his fellow-members
and his difference from members of other groups within the same
language-community. The indexical information carried by dialect and
accent is usually group-identifying. But, if it so happens, as it frequently
does, that an individual comes to live or work in a community whose
speech is noticeably different from that of the community in which he
was brought up, what would normally count as group-identifying
indexical features may operate instead as individual-identifying. In-
dexical information carried by dialect and accent, like that carried by
 14.5. Stylistic, dialectal and diachronic variation 617
voice-quality, is not normally communicated in the strict sense of the
term: i.e. it is not usually transmitted intentionally (cf. 2.1). But it may
serve as the basis for true communication in certain circumstances.
Actors will commonly simulate the more salient features of some dialect
other than the one that they would otherwise speak, in order to indicate
to the audience that the characters they are playing are supposed to come
from a certain region or class. In such contexts conventionalized stereo-
types of American English, Irish English, Scots English, Welsh English,
Oxford English, or whatever it might be, serve the communicative pur-
pose to which they are put even better than would the genuine accents
or dialects of which they are stereotypes.
It might seem that acting a part professionally on the stage is purely,
derivative. But'we'all act character-parts, non-professionally, at times in
our everyday use of English: we tell jokes about the Irishman, the
Scotsman and the Welshman; we repeat more or less verbatim some-
thing that colleagues or neighbours have told us; and so on. In doing so,
we may mimic certain characteristic features of accent or dialect; and,
in so far as the accent or dialect in question is associated, in the language-
community as a whole, with certain traits of character or patterns of
behaviour, what is said in the propositional content of the utterance may
be reinforced or highlighted by these features of accent or dialect.
Although this kind of language-behaviour is hardly to be classed as
representative of language-behaviour in general, it is of some interest to
the semanticist, in that it shows that there are times when the informa-
tion that is encoded in the indexical features of utterances can be a part
of what is communicated.
More important, however, from the point of view of semantics, 1s the
phenomenon of diglossia and code-switching. It has already been pointed
out that different dialects of the same language, as well as different
languages, may be associated with characteristically distinct contexts-of-
situation (cf. 14.2), All that needs to be said here, in this connexion, is
that, when it comes to questions of diglossia, it is impossible, in the last
resort, to draw a sharp distinction between different languages, different
dialects of the same language and different styles. Standard languages, as
we know them in the modern world, are, in origin at least, no more than
dialects which, for political and cultural reasons, have acquired a certain
ascendancy and prestige. Their adoption as regional or national stan-
dards, however, means that they will tend to be used, over a wider area
and in a wider range of situations, than any of the non-standard dialects,
whose employment may come to be restricted to domestic or informal
618 Context, style and culture
situations. If what were originally different regional or social dialects of
the same language come to be restricted, throughout the language-
community as a whole, to particular kinds of situations, they are no less
appropriately described as styles than as dialects.
Nothing that has been said so far would suggest that dialectal varia-
tion is relevant as far as the descriptive function of language is con-
cerned. But there are certain respects in which it is, or may be; and this
fact is only rarely taken into account, in general treatments of semantics.
For example, the words ‘lake’ and ‘loch’ are, from a certain point of
view, denotationally equivalent, the one being part of the vocabulary of
Standard English and the other being part of the vocabulary of various
dialects of Scots English. Many speakers of Standard English, however,
will use the word ‘loch’, instead of ‘lake’, when they are in Scotland
and referring to a Scottish lake, whereas they would not use the word
‘loch’, as a speaker of Scots English would, with reference to a lake in
some country other than Scotland. There are two ways of looking at facts
of this kind. One 1s to say that the speaker of Standard English borrows
what he knows to be, and what remains for him, a word from another
dialect; the other 1s to say that the vocabulary of Standard English
contains two different lexemes, ‘lake’ and ‘loch’, one of which is more
restricted in its application than 1s the other. If we adopt the first point
of view, we can say that what we have is a rather minimal instance of
diglossia. If we adopt the second point of view, we are faced with the
question whether the two Standard English words have the same de-
scriptive meaning or not; and this question is not answerable, without
qualification, in one way rather than the other. The relationship between
‘loch’ and ‘lake’ is not such that the former can be said to be a hyponym
of the latter (cf. 9.4); and “in Scotland” is not encapsulated in the sense
of ‘loch’, as “of hunger’’, let us say, is encapsulated* in the sense of
‘starve’ (cf. 8.5). There is nothing tautological about the collocation
‘Scottish loch’. And yet the reference of a nominal like ‘that loch’ is
restricted by conditions of applicability attaching to ‘loch’ which have
the same effect as would something like the encapsulation of ‘‘in
Scotland”’ in its sense. It is at least a presupposition of the use of ‘loch’
in Standard English (under the assumptions that we are making) that, if
something is referred to as a loch, it should be in Scotland, as it is at
least a presupposition of the use of ‘bachelor’ in referring expressions
that the referent should be a male adult (cf. 14.3).
This is just one example of what is a quite pervasive phenomenon. It
is but a short step from cases like ‘loch’: ‘lake’ (or ‘burn’: ‘stream’/
 14.5. Stylistic, dialectal and diachronic variation 619
‘brook’) to cases like ‘kilt’: ‘skirt’ where, arguably, we have a clear
difference of descriptive meaning; and it is but another short step to
cases like ‘skirt’:‘shirt’, which are definitely different in sense and
denotation and would probably not be recognized as having originated
as dialectal variants by anyone but a specialist in the history of English
and the Germanic languages. Everyone knows that forms of lexemes may
be borrowed from another language or dialect and used, more or less
deliberately, to evoke certain aspects of the situation that is being
described. If we hear an English speaker of Standard English say We
went swimming in the loch or There’s an old kirk at the bottom of the
garden or We spent the nght in a ruined bothy, not only do we know
immediately that what is being referred to is in Scotland, but we are
likely to conjure up a quite different picture of the situations being de-
scribed from that which we would if the speaker had used instead the
stylistically neutral words ‘lake’, ‘church’ and ‘cottage’. It is perhaps a
little less obvious, until one’s attention is drawn to the fact, that there is
no clear line of demarcation between the purely stylistic use of dialectal
variants, whether it is deliberate or not, and the use of a lexeme from
another dialect because it is felt to have a more specific descriptive
meaning. Furthermore, in so far as the borrowed lexeme, in the borrow-
ing dialect, though not necessarily in the dialect from which it is
borrowed, is felt to have a more specific meaning than the stylistically
neutral variant, this derives as much from the contexts in which it 1s
used as it does from the salient perceptual or functional differences in
the class of things that the borrowed lexeme denotes.
Mention of a pair of lexemes like ‘skirt’ and ‘shirt’, which from a
historical point of view may be thought of as dialectal variants, brings up
the connexion between diachronic and dialectal variation. As we saw
earlier, the necessity of drawing a distinction between the synchronic
and the diachronic investigation of languages is something that was
emphasized by Saussure and is now taken for granted by linguists.
, Diachronic, or historical, semantics has so far been concerned almost
exclusively with changes in the meaning of words, except in so far as
changes in the meaning of grammatical constructions have been noted,
as causes or consequences of changes in the grammatical structure of
languages, by scholars working in the field of historical syntax. Much of
the earliest work in diachronic semantics, following the publication of
Bréal (1897), was strongly influenced by the principles of traditional
etymology, which itself drew heavily upon classical logic and rhetoric.
The attempt was made to formulate laws of semantic change which
620 Context, style and culture
would account for developments that had taken place in the meaning of
particular lexemes in much the same way that the so-called sound-laws
(Lautgesetze) of the Neogrammarians (Junggrammatiker) were intended
to account for developments that had taken place in what we would now
refer to as the phonological structure of word-forms. The laws of seman-
tic change that were proposed by Bréal and his followers, however, were
quite different in character from the Neogrammarian sound-laws, in
that they were held to be operative at all times and with respect to all
languages. Changes of lexical meaning were classified in terms of such
notions as broadening, narrowing and metaphorical transfer, on the one
hand, and, under the influence of the fast-developing science of
psychology, of the acquisition of pejorative or ameliorative associations,
on the other; and the laws of semantic change that were proposed did
little more than reflect the prior classification of the data. Most conven-
tional dictionaries, it may be observed, whether they are said to be
constructed on historical principles or not, still operate with such notions
as broadening, narrowing and metaphorical transfer in their classifica-
tion of what are held to be different, but related, meanings of polysemous
lexemes (cf. 13.4).
The three most important developments in diachronic semantics that
have taken place in the last fifty years are: (i) the application of the
principles of structuralism in tracing the history of particular semantic
fields (cf. 8.4); (ii) the implementation of the principle that the history
of the vocabulary of a language cannot be studied independently of the
social, economic and cultural history of the people speaking that
language; (iii) the realization that diachronic and dialectal variation are
ultimately inseparable. It is the third of these developments that we are
concerned with here. All that needs to be said about the first is that it
constitutes a modification of the Saussurean view, according to which
synchronic linguistics alone deals with systems (cf. Ullmann, 1962); and
about the second, that it has more than justified itself in the numerous
monographs that have been written in the spirit of the so-called words-
and-things (Wo6rter-und-Sachen) movement — “Ohne Sachforschung
keine Wortforschung mehr” (“‘ No more study of words [i.e. etymology]
without a study of things”: cf. Ullmann, 1957: 211).
The distinction between synchronic and diachronic linguistics, as we
have seen, must not be pressed too hard: regional and social dialects of
the same language spoken at the same time may be more different from
one another than diachronically distinct states of what we would con-
sider to be the same language (cf. 8.2), What is more important, how-
 14.5. Stylistic, dialectal and diachronic variation 621
ever, in the present connexion is the fact that synchronic dialectal
variation is itself the source of a considerable amount of what is sub-
sequently describable as change in the overall language-system. Speakers
of dialect A will tend to imitate speakers of dialect B if they wish to be
integrated, wholly or partially, into the community in which dialect B
is spoken. It is generally, but not always, the case that in situations of
this kind dialect B will be one that enjoys prestige throughout the
language-community as a whole (cf. Labov, 1972). Whether this is so or
not, it will often be on account of their semantic function as indices of
membership of a particular social class or social group that forms,
lexemes, collocations and grammatical constructions, as well as features
of pronunciation, are taken over and adopted, in some or all situations,
by speakers of another dialect.
There is another sense, too, in which the distinction between the
synchronic and the diachronic must not be pressed too hard. The nor-
mal Ilanguage-community will contain, at any one time, children who are
completing the process of learning their native language (in so far as
this process is ever complete) as well as old men and women who have
been speaking the language for upwards of seventy years. In so far as the
speech of one generation, including its more or less ephemeral slang and
jargon, is noticeably different from the speech of other generations
within the same language-community, there will be a kind of diachrony-
in-synchrony, of which members of that language-community may
themselves be aware. ‘The stylistic importance of this phenomenon of
diachrony-in-synchrony is something that has been strongly emphasized
by linguists of the Prague School (cf. Vachek, 1964). They have pointed
out that, at any one time, certain forms, lexemes or expressions will
strike the average member of the language-community as old-fashioned
and that other forms, lexemes or expressions may strike him as new and
not fully established. To this extent, therefore, the average member of
the language-community may be conscious of the directionality of
change in the overall system; and, if he 1s familiar with the written or
oral literature that has been composed in the past and transmitted from
one generation to the next within the language-community, his sense of
the directionality of change, of diachrony-in-synchrony, will be con-
siderably enhanced. As one can deliberately use a form, lexeme or
expression from another regional or social dialect either to evoke aspects
of the place or society that one is describing or to associate oneself with
that place or society, so one can deliberately employ forms, lexemes or
expressions from the speech of an older or younger generation for much
 622 Context, style and culture
the same purposes. Diachrony-in-synchrony, like dialectal variation, has
its indexical function.
Throughout this section we have been concerned to emphasize the
impossibility of drawing a sharp distinction (other than methodo-
logically by a process of deliberate idealization: cf. 14.2) between
stylistic, dialectal and diachronic variation. We have also emphasized
the fact that variation within the language-community serves, at least
potentially, an important semantic function, in that it provides indices
for membership of one social group rather than another. Most of what
has been said is irrelevant as far as descriptive meaning is concerned. It
should not be forgotten, however, that a particular lexeme may have a
quite different meaning in dialect A from the meaning that it has in
dialect B. If both meanings are associated with the lexeme in question,
in a model of the overall language-system that discounts the differences
between dialect A and dialect B, the lexeme will be represented in the
lexicon as polysemous; and some or all system-sentences containing it
will be counted, by the model, as ambiguous. It may well be, however,
that other purely indexical features in the utterances in which the
lexeme occurs, in particular contexts-of-situation, would forestall mis-
understanding. For example if £ never wear suspenders 1s spoken (by a
man) with an American, rather than a British, accent, it will tend to be
understood, other things being equal, as having the same meaning as in
I never wear braces in British English. To this extent at least, the indexi-
cal meaning of an utterance can contribute to the determination of its
descriptive meaning; and we have seen that as far as certain stylistically
marked dialectal variants are concerned it is hard to draw a sharp dis-
tinction between descriptive and non-descriptive meaning.

14.6. Sentences and texts

Our main concern in this section is the relationship between system-
sentences* and text-sentences*. The distinction that we have drawn
between these two kinds of entities is one that is not usually drawn by
linguists; and it requires some justification. System-sentences, it will be
recalled, are abstract theoretical constructs, correlates of which are
generated by the linguist’s model of the language-system in order to
explicate that part of the acceptability of utterance-signals that is
covered by the notion of grammaticality; text-sentences, on the other
hand, are context-dependent utterance-signals (or parts of utterance-
signals), tokens of which may occur in particular texts* (cf. 1.6, 10.3,
14.2).
 14.6. Sentences and texts 623
The term ‘text-sentence’ has been chosen to emphasize the fact that
there is a legitimate sense of the word ‘sentence’ in which some texts at
least can be said to consist of an integral number of sequentially ordered
sentences. It is in this sense of ‘sentence’, for example, that we can say
that the first sentence of the previous paragraph begins with the word-
form our, capitalized, by convention, to indicate that it is the first
word-form in a text-sentence. A recent study of the structure of texts in
English starts as follows: “‘If a speaker of English hears or reads a
passage of the language which is more than one sentence in length, he
can normally decide without difficulty whether it forms a unified whole
or is just a collection of unrelated sentences ... The word ‘text’ is used
in linguistics to refer to any passage, spoken or written, of whatever
length, that does form a unified whole. We know, as a general rule,
whether any specimen of our own language constitutes a text or not”’
(Halliday & Hasan, 1976). The authors of the work from which this
quotation is taken do not draw a terminological distinction between
system-sentences and text-sentences. But it is clear that, in so far as they
are dealing with sentences as parts of texts, they are dealing with what
we would identify as text-sentences.
Not only is it true that native speakers of a language can normally tell
the difference between texts and non-texts; it is also the case that they
can normally segment any text that consists of more than one text-
sentence into its component text-sentences. As far as formal written
prose is concerned, deciding whether a certain portion of text consti-
tutes a single complete text-sentence or not is a fairly straightforward
matter. The beginning of a text-sentence is indicated by capitalization of
the first letter of the first word-form and the end of a text-sentence is
indicated by the presence of a full-stop. Taking capitalization and the
presence of a full-stop to be necessary, if not sufficient, conditions for the
classification of a portion of text as a text-sentence, we know that the
first paragraph of this section consists of not more than three text-
sentences. But, as literate native speakers of English, conversant with the
conventions of punctuation, we also know that each of the two semi-
colons occurring in the paragraph in question might have been replaced
with a full-stop. The paragraph would still be an acceptable text, if it
consisted of five, rather than three, orthographic text-sentences; and,
arguably, it does not make any difference to what is being said whether
it is punctuated in the one way or the other. Two conclusions follow:
first, that, as far as the written language is concerned, an author is able,
within certain limits, to insert his own sentence-boundaries; second, that
 624 Context, style and culture
intersubjective agreement among literate native speakers as to what these
limits are, under what circumstances and with what effect (if any) a
full-stop may be substituted for some other punctuation-mark, shows
that it is far from being a matter of arbitrary decision how a written text
is segmented into text-sentences.
There is nothing in spoken language that corresponds directly to
capitalization or the occurrence of a full-stop in written texts. This does
not mean, however, that spoken texts cannot be segmented into text-
sentences. What it means is that the identification of spoken text-
‘sentences is rather more complex; and, because the production of
utterances in the phonic medium is not, in general, subject to constraints
of the kind that have been conventionalized for the graphic medium by
printers and editors, the identification of text-sentences in spoken texts
cannot always be carried out consistently to the point that every spoken
text is analysable, without residue, into an integral number of text-
sentences. Spoken utterances are punctuated*, as we have seen, by
prosodic and paralinguistic features ~ by stress, intonation and rhythm
(cf. 3.1). But there is no single prosodic feature that serves as a sentence-
boundary marker in the phonic medium in quite the same way that a
full-stop, a question-mark or an exclamation-mark serves to mark the
end of a text-sentence in the graphic medium.
The identification of text-sentences in the spoken language usually
involves considering, not only their prosodic and paralinguistic punctua-
tion, but also their grammatical structure. But, up to a point, it can be
done non-arbitrarily by native speakers; and the conventionalization of
the use of capital letters and punctuation-marks in written texts rests,
ultimately, upon the correspondence that holds (within the limits of
medium-transferability: cf. 3.3) between written and spoken text-
sentences. Furthermore, in so far as the spoken language is basic and the
written language derived from it, the notion of the spoken text-sentence
is logically prior to the notion of the written text-sentence.
More basic than either kind of text-sentence, however, is the system-
sentence, which is, by definition, the maximal unit of grammatical
description. In so far as a text 1s segmentable into portions each of which
is identifiable as a text-sentence, this is because, typically, there is some
kind of correspondence between particular system-sentences (correlates
of which are generated by the linguist’s model of the language-system)
and particular text-sentences. So far we have been operating with “‘the
simplifying assumption that system-sentences are sequences of words in
a one-to-one order-preserving correspondence with what would be
 14.6. Sentences and texts 625
judged, intuitively by native speakers, to be grammatically complete
text-sentences”’ (cf. 1.6). But most of the actually attested utterances of
English and other languages cannot be put into correspondence with
system-sentences in terms of a word-for-word, one-to-one, order-
preserving relation of this kind. As we have seen earlier in this chapter,
system-sentences can be regarded as maximally decontextualized ,
utterance-signals (cf. 14.2). The vast majority of the utterances that
actually occur in the everyday use of language are very heavily context-
dependent; and their context-dependence may be made manifest in the
utterance signals themselves in various ways. An utterance may be
elliptical and therefore classifiable as a sentence-fragment* (e.g., As soon
as I can: cf. 14.2); it may contain a connective, like so, but, and or
however, which relates the content of the utterance in question to what
has already been said (e.g., So we arrived late); it may contain anaphoric*
elements of one kind or another (cf. 15.3); it may have a thematically
marked word-order or prosodic structure (cf. 12.7). ‘These are but some
of the ways in which the context-dependence of an utterance may be
manifest in the utterance itself.
It is in principle possible that there should be languages in which the
contextualization of a system-sentence always involves making manifest,
in one way or another, the relationship between the utterance itself and
the context in which it occurs. It might then be the case that there was
no text-sentence that could be put into word-for-word, one-to-one,
order-preserving correspondence with any system-sentence. The rela-
tionship between text-sentences and system-sentences might be in all
instances similar to that which holds between the text-sentence So we
arrived late and the system-sentence ‘We arrived late’; and there are
other more or less plausible possibilities. The point to be emphasized
here is that the notion of system-sentence, which we take to be more
basic than that of text-sentence, does not depend for its validity upon
the occurrence, as acceptable utterances of the language in question, of
strings of words in one-to-one, order-preserving correspondence with
the strings of words that are held to be well-formed system-sentences. A
system-sentence may first have to be contextualized in one way or
another before the resultant utterance can be judged in terms of
acceptability. In what follows, however, we will assume that, as appears
to be the case for English and many, if not all, languages, there are some
text-sentences that meet the condition of word-for-word, one-to-one,
order-preserving correspondence with system-sentences. Most of the
utterances cited in this book are text-sentences that do, in fact, meet this
626 Context, style and culture
condition; and their status as text-sentences is thereby validated. Their
grammatical acceptability is accounted for by the well-formedness of the
corresponding system-sentences.
But why, it may now be asked, do we say of at least some utterances
which do not meet the condition of word-for-word, one-to-one, order-
preserving correspondence that they are, nonetheless, text-sentences?
Why do we say, for example, that So we arrived late is a text-sentence?
There are two reasons. The first is that, although it is context-depen-
dent, its context-dependence is not manifest in its grammatical structure:
it is not a sentence-fragment, rather than being a grammatically com-
plete text-sentence; and it does not contain any form or construction
that is restricted to occurrence in subordinate or dependent clauses. The
second reason is that it has the same unitary intonation contour as We
arrived late, which, by virtue of its correspondence with ‘We arrived
late’, we have classified as a text-sentence. It was pointed out earlier that
nowadays most linguists take the view that at least some part of the
prosodic structure of utterances, including their intonation-contour,
should be accounted for in the analysis of system-sentences (cf. 10.1).
The same grammatically structured string of forms may have several
different intonation-contours superimposed upon it; and some of these
may be identified, in the linguist’s model of the language-system, as
being characteristic of different kinds of system-sentences (e.g., declara-
tive, interrogative, exclamative). A particular utterance-token may
therefore be clearly identifiable as a text-sentence by virtue of its having
superimposed upon the string of forms of which it is, in part, composed
a particular kind of sentential intonation-contour.
What are traditionally described as compound* sentences (i.e. sen-
tences composed of two or more conjoined co-ordinate clauses) are not
identifiable as single sentences in terms of their grammatical structure
alone. It is only because, in English and many languages, the conjoined
co-ordinate clauses can be brought within the domain of a unitary
intonation-contour that there is any reason to recognize compound
sentences as a subclass of system-sentences. As far as their grammatical
structure is concerned there is nothing to distinguish
(1) John got up late and he missed the train
from
(2) John got up late. And he missed the train.
The fact that (1) is punctuated in the written language as a single text-
 14.6. Sentences and texts 627
sentence, whereas (2) 1s punctuated as a sequence of two text-sentences
depends, ultimately, upon the fact that, on the assumption that (1) and
(2) are being uttered under normal conditions and with a primarily
descriptive function, (1) will have a single sentential intonation-contour
superimposed upon it, whereas (2) will have two.
Similarly, in the somewhat more marginal case of clausal parataxis*
(i.e. the juxtaposition, as distinct from the conjoining, of co-ordinate
clauses) it is intonation which, if anything, brings it within the scope of
the linguist’s model of the language-system. For example,
(3) fohn missed the train: he got up late,
which is here represented, by the orthographic conventions of the
written language, as a single text-sentence, can have a single sentential
intonation-contour superimposed upon it in the spoken language. In
which case, the two clauses may be construed as being combined para-
tactically as co-ordinate constituents of a single spoken text-sentence;
and this will be so regardless of whether the linguist’s model of the
language-system generates the sentence ‘John missed the train: he got
up late’ (with a characteristic prosodic contour) or not. The difference
between (3) and the following sequence of two text-sentences
(4) John missed the train. He got up late
is, ONce again, made explicit and fully determinate, by the conventions
of punctuation in the written language. If an author writes (3), rather
than (4), he thereby makes explicit the fact that he is treating John’s
missing the train and John’s getting up late as two connected events.
That the events described in (2) and (3) are held to be causally con-
nected, rather than being connected in some other way, is no more,
perhaps, than a matter of conversational implicature (cf. 14.3). For
(5) John got up late: he missed the train
will normally be construed as implicating the same causal connexion
between the two events. This shows that the colon in the written text-
sentences and the absence of a sentence-final intonation-contour on the
first of the juxtaposed clauses in the spoken text-sentences cannot be
satisfactorily accounted for by postulating the deletion of some specific
causal conjunction. At the same time, it is difficult to envisage circum-
stances under which the implicature that the events are causally con-
nected in one direction or the other could be cancelled or qualified. And
this it is which distinguishes (3), semantically, from (4), even when it is
628 Context, style and culture
otherwise clear that the sequence of two text-sentences in (5) constitutes
the whole or a part of a single text.
As far as spoken texts are concerned, the difference between a single
text-sentence consisting of two paratactically juxtaposed clauses and a
sequence of two text-sentences connected solely in terms of their content
is not always clear-cut. Nor indeed is the difference between a compound
text-sentence like
(6) fohn got up late and he missed the train
and a sequence of two text-sentences (the second of which begins with a
connective that is identical in form with an intrasentential co-ordinating
conjunction) like
(7) Fohn got up late. And he missed the train.
Unlike phonologically distinct features of the verbal component of
utterances, prosodic and paralinguistic features (intonation, rhythm,
etc.), which serve to distinguish clear instances of the phonic equivalents
of (6) and (3) from clear instances of the phonic equivalents of (7) and ,
(4), do not have the design-property of discreteness (cf. 3.4). The con-
ventions of punctuation force a writer to make a decision, in this respect,
that a speaker is not obliged to make in producing an utterance and the
addressee is not obliged to make in interpreting it.
It follows from what has just been said that whether a certain portion
of a spoken text is a single text-sentence or not is not necessarily decid-
able. Granted that this is so, it might still be the case that every text is
segmentable (after regularization: cf. 14.2) into an integral number of
text-sentences. ‘This is an assumption that is commonly made; and, once
again, the conventions of punctuation in the graphic medium are such
that, as far as even quite informal written texts (such as chatty, personal
letters) are concerned, it holds true. Whether it holds with respect to
spoken texts, however, is largely a matter of how we choose to define the
text-sentence. It is up to us, for example, whether we count such
utterances as A friend of mine —I can’t remember his name — used to go
there every year as single text-sentences or not. It is also up to us whether
we classify sentence-fragments (traditionally described as incomplete
sentences) and such utterance-signals as Yes, No, For heaven’s sake,
with which everyday conversation abounds, as text-sentences. If we do,
on the grounds that some of them at least are functionally equivalent to
what are unquestionably text-sentences and may have a sentential
intonation-contour superimposed upon them, it will be true, as a conse-
 14.6. Sentences and texts 629
quence of this decision, that most, if not all, spoken texts (after they have
been duly regularized) are analysable into an integral number of
sequentially ordered text-sentences. But the question is very largely
definitional.
The distinction that has been drawn in this book between system-
sentences and text-sentences enables us to avoid, in principle, if not
_always in fact, much of the confusion that attaches to the term ‘sentence’
in linguistics. Generations of grammarians have tried to define the sen-
tence without being clear about what they were trying to define.
Linguists tend to spend far less time these days discussing the nature of
sentences. But this is not because there is now some generally accepted
criterion, or set of criteria, in terms of which it can be decided what is
and what is not a sentence. The reason is simply that linguists have been
less concerned recently with questions of definition. Chomsky and his
followers, in particular, have been content to operate with the assump-
tion that native speakers have an intuitive appreciation of the fact that
certain strings of forms are sentences and others are not. But they have
failed to give any account, even in principle, of the way the sentence as a
theoretical construct within the linguist’s model of the language-system
is related to the sentence as a contextualized product of language-
behaviour. This being so, there has been, and there still remains, con-
siderable uncertainty as to what is meant by saying that an observa-
tionally adequate grammar will generate, in the ideal, all and only the
sentences of the language: if there is one thing that is certain about the
system-sentences of a language, it is that they are not given as part of the
observable data. And yet it is system-sentences, rather than text-
sentences, that a Chomskyan sentence-generating grammar aims to
generate. By drawing a terminological distinction between system-
sentences and text-sentences we can at least explain what is meant by
saying that a grammar generates all and only the sentences of a language
(cf. 10.3).
We can also avoid the problems that arise in trying to work with a
definition of the sentence like Bloomfield’s (1955: 170), according to
which a sentence is an independent linguistic form not included by
virtue of any grammatical construction in some larger linguistic form.
12 Some of the two hundred or so different definitions of the sentence that have
been identified are listed and discussed, from a Bloomfieldian point of view,
by Fries (1952). Bloomfield’s own review of Ries (1931) is worth consulting
in this connexion (cf. Bloomfield, 1931). For two quite different recent
approaches to the problem of defining the sentence, cf. Allerton (1969),
Kasher (1972).
630 Context, style and culture
As a definition of the system-sentence, this is satisfactory enough: indeed
it would be hard to improve upon it. But, as it was applied by Bloom- ,
field and his followers, the definition was really intended to cover what
we have called text-sentences; and this is where it breaks down. As we
have seen, English texts cannot in general be segmented without residue
into successive and non-overlapping portions each of which is a sentence
in terms of Bloomfield’s criterion of grammatical independence; and yet
many of these grammatically non-independent segments of texts are
reasonably regarded as text-sentences by virtue of their prosodic contour
and their functional equivalence with actual or potential text-sentences
that would meet the criterion of grammatical independence. There are
many languages in which the difference between grammatically depen-
dent and grammatically independent text-sentences is more striking than
it is in English (cf. Waterhouse, 1963). The grammatically dependent
text-sentences may satisfy one part of Bloomfield’s definition: non-
inclusion by virtue of any grammatical construction in some other

tion. |
linguistic form. But they obviously do not satisfy the rest of the defini-

Throughout this section, and elsewhere in the book, we have adopted

the traditional view, that sentences are the maximal units of gram-
matical description. Not all linguists nowadays would accept this point
of view. There has recently been a considerable upsurge of interest in
what is sometimes referred to as text-linguistics, or even text-grammar
(cf. Dressler, 1972; Van Dik, 1972); and some, though not all, of the
scholars who share this interest in the linguistic analysis of texts have
argued that the relation between a sentence and the text of which it is a
component part is, in all relevant respects, comparable with the relation
that holds between a word, or phrase, and the sentence of which it is a
grammatically dependent constituent. It remains to be seen what will
come of the attempt to construct a generative theory of well-formed
texts. But it is obvious even now that, once we draw a distinction
between system-sentences and text-sentences, the case for recognizing
grammatical units larger than the sentence loses much of its force. The
notion of grammatical well-formedness applies primarily to system- .
sentences and only secondarily to text-sentences. It is text-sentences,
however, of which it makes sense to say that, by virtue of the contex-
tualization of the system-sentences from which they are derivable, they |
enter into inter-sentential and supra-sentential relations; and it has yet
to be shown that text-sentences function within texts, or within dis-
tinguishable parts of texts (e.g., units comparable with the conventional-
 14.6. Sentences and texts 631
ized paragraphs of written prose), in the way that words, phrases and
clauses function grammatically within system-sentences.
Texts, as the term ‘text’ is normally understood, are the product of
more or less conscious and controlled literary composition. They have a
determinate beginning and end; and some kind of internal coherence or
unity. Most of our everyday conversation, however, is not made up of
texts in this sense. It is arguable, therefore, that the notion of the
coherent, or well-formed, text, useful though it may be in literary

language-behaviour. |
stylistics, is not generalizable to the most typical and most basic kind of

The term ‘text’ may also be used, and commonly is so used by

linguists who do not necessarily subscribe to the notion of the well-
formed text that has been developed in text-linguistics, for the phono-
logically transcribable product of everyday language-behaviour. In this
sense of the term, the relevant question is not “Is this a text?”’, which
carries with it presuppositions of internal organic unity and determinate
external boundaries, but “‘Does this constitute text (rather than non-
text)?”’. The difference between these two questions is of considerable
theoretical and practical importance. The second (in which ‘text’ is _
used as an uncountable noun) gives due recognition to the fact that
successive text-sentences, in either a dialogue or a monologue, tend to be
connected in various ways; but it neither presupposes nor implies that
what is correctly describable as text is, or forms part of, some determ-
inate unified whole. Earlier in this section we quoted, without comment-
ing upon the point at the time, a statement from a recent important work
on cohesion* in English, to the effect that, as native speakers of any
language, ““we know, as a general rule, whether any specimen of our
language, constitutes a text or not”’ (cf. Halliday & Hasan, 1976: 1).
From the point of view adopted here, it would be preferable to say (as
the authors do in fact say on the same page) that native speakers “‘are
sensitive to the distinction between what is text and what is not’’. The
first of these statements may very well imply the second; but the second
certainly does not imply the first.
There is more to the description of a language-system than construct-
ing a set of rules which will generate correlates of all and only what are
taken to be the system-sentences of the language in question. The native
speaker’s ability to contextualize system-sentences (i.e. to produce text-
sentences) depends, in part, upon the existence, in every language-
system, of certain text-forming resources, whose function it is to relate
utterances to the context in which they are produced (cf. Halliday,
632 Context, style and culture
1970b; Halliday & Hasan, 1976: 27). These text-forming, or contex-
tualizing, resources may be lexical, grammatical or phonological. Some,
though not all, may have an exclusively text-forming function. For
example, such word-forms as however or moreover never occur in what
we take to be the system-sentences of English: their function is ex-
clusively that of relating the text-sentences in which they occur to the
preceding co-text. Such word-forms as but and and, on the other hand,
have both a contextualizing and a non-contextualizing function; and it
would be difficult, and somewhat artificial, to draw a sharp distinction,
in terms of meaning, between these two functions.
Indeed, it will be obvious that there is a certain degree of arbitrariness,
not to say artificiality, involved in the process of decontextualization
itself (cf. 14.2). There is no reason to suppose that system-sentences, as
such, play any role in the production and interpretation of utterances —
the more so, as there are no accepted criteria for deciding, with respect
to certain phenomena, whether they are to be accounted for as part of
the structure of system-sentences or not. For example, among the
resources that the English language-system makes available for the —
contextualization of utterances in terms of their thematic structure are
the possibility of using (though to a much more limited degree than
many other languages) variations of word-order; and the possibility of
superimposing one prosodic contour, rather than another, upon the
verbal component of the utterance (cf. 12.7). As far as the prosodic
contour of utterances is concerned, we have recorded it as the view of
most linguists that at least some part of this should be handled within
the sentence-generating grammar, but we have, in principle, left this
question open (cf. 3.1, 10.1). We have taken it for granted, however, that
in so far as the thematic structure of a text-sentence is made manifest by
differences of word-order and other non-phonological devices, this will
be accounted for by postulating a transformational relationship between
system-sentences. It is easy to see that there 1s an element of arbitrari-
ness in the distinction that is drawn here between phonological and
non-phonological text-forming devices.
To say that there is some degree of artificiality in the process of de-
contextualization whereby we arrive at a representative subset of the
system-sentences of a language is not to say that the notion of the
system-sentence is completely spurious. It is a theoretical construct
whose principal function in the linguist’s model of the language-system
is to define the concept of grammaticality; and, as we have seen, there
are certain pre-theoretical constraints which at least partly determine
 14.6. Sentences and texts 633
the scope of any theoretical concept of grammaticality that we might
wish to define (cf. 10.2).
Although the system-sentence is a unit of the language-system which
serves, first and foremost, as the domain of grammatical processes
(concord, government, etc.), it was argued in the first section of this
chapter that what might be appropriately described as a theory of
microlinguistic semantics would be concerned with the meaning of
maximally, though not fully, decontextualized system-sentences (cf.
14.1). We can now develop this point and, in doing so, conclude our
treatment of the relation between system-sentences and text-sentences.
It will not have escaped the reader’s notice that the terms ‘utterance’
and ‘utterance-signal’ have been used in this section, and elsewhere 1n
this book, without any attempt yet having been made to relate them at
all precisely to either “system-sentence’ or ‘text-sentence’. Indeed,
there has been until now a certain equivocation in our use of the term
‘utterance’. In the very first chapter we adopted Harris’s (1951: 14)
characterization of the utterance as a pre-theoretically identifiable unit,
as “any stretch of talk by one person, before and after which there is
silence on the part of that person”’; and we pointed out that utterances,
in this sense, might consist of several text-sentences. The vast majority
of utterance-signals cited in this book, however, have been single text-
sentences. We have tacitly assumed that within the set of what are pre-
theoretically identifiable as utterances, in terms of external observa-
tional criteria, there is a subset of particular interest — which we will
now call utterance-units* — to which such terms as ‘statement’, ‘ques-
tion’ and ‘command’ are applicable (cf. 1.6). Looked at from a logical
point of view, statements, questions, commands and exclamations, as
well as utterance-units of other kinds, may be classified as simple or
complex, according to whether they contain a simple or a complex
proposition (cf. 6.2). Let us concentrate first upon simple utterance-
units.
There is an obvious, and far from fortuitous, connexion between
simple utterance-units and what are traditionally classified as simple,
rather than compound or complex, sentences. As a simple utterance-
unit is one that contains one and only one simple proposition (whatever
else it may contain over and above its propositional content), so a simple
sentence is one that expresses one and only one simple proposition
(whatever else it may express). Simple utterance-units, 1n this sense, are
the basic units of language-behaviour. They may be heavily context-
dependent, such that it is impossible to determine which of indefinitely
634 Context, style and culture
many propositions they contain without drawing upon the information
that is given in the co-text or context-of-situation (cf. John did). They
may be relatively independent of the context in which they occur with
respect to the determination of the proposition that they contain (cf.
Fohn came). Regardless of whether they are heavily context-dependent
in this respect or not, simple utterance-units will be functionally
equivalent if, and only if: (1) they all have the same illocutionary force*
(i.e. if all of them are statements, or all of them are questions, etc.: cf.
16,1); and (ii) they all contain the same proposition. In many, if not all,
languages, simple utterance-units with the same illocutionary force tend
to have the same prosodic contour (provided that they do not differ
significantly in thematic structure). The two criteria of functional
equivalence and identity of prosodic contour are mutually reinforcing in
the delimitation of simple text-sentences; and it was for this reason that
they were invoked earlier in our discussion of the relationship between
system-sentences and text-sentences.
Not only do functionally equivalent simple utterance-units tend to
have the same prosodic contour, but, to the extent that they are gram-
matically complete (i.e. non-elliptical: cf. 14.2), they tend to be parallel
in terms of their grammatical structure. In English, for example, ques-
tions are asked, characteristically, though not necessarily, by means of
utterance-units whose grammatical structure is accounted for by
deriving them from interrogative system-sentences (cf. 16.1). There is a
tendency, therefore, for grammatical structure, functional equivalence
and prosodic structure to coincide as far as the determination of the
basic units of language-behaviour is concerned; and this coincidence of
grammatical, semantic and phonological criteria is what enables us to
identify, as readily as we do, simple text-sentences and to group them
into equivalence-classes in terms of their postulated derivation from
maximally context-independent system-sentences.
The relationship that holds between complex utterance-units and
non-simple (i.e. compound and complex) system-sentences is no differ-
ent, in principle, with respect to the coincidence of grammatical,
semantic and phonological criteria from the relationship which holds
between simple utterance-units and simple system-sentences. Whether
we say that someone has made two statements, each containing a simple
proposition, will largely depend upon whether the utterance that he has
produced is classified as two consecutive text-sentences or as a single
text-sentence. It might even be argued that the very notion of a complex
proposition is parasitic upon the existence, in certain languages, of
 14.6. Sentences and texts 635
grammatical and phonological resources for constructing non-simple
system-sentences. But we will not pursue this question.
As far as everyday conversation is concerned, a considerable number
of what are pre-theoretically identifiable as utterances (stretches of
speech by a single person) are either simple or complex utterance-units.
They are the products of what may be taken to be single speech-acts*
(cf. 16.1); and their unitary intonation-contour reflects this. Others are
composed of sequences of text-sentences, each of which constitutes a
single (simple or complex) utterance-unit; and, once again, the intona-
tion-contour of each text-sentence may be seen as a reflexion of its
status as an utterance-unit. This correspondence between utterance-
units and text-sentences, each containing a single complex proposition,
is the norm, from which there are certain deviations. There are single
text-sentences that contain more than one utterance-unit (cf. Did Fohn,
who was here yesterday, say anything about 1t?, in which a statement is
parenthetically included within a question); and there are other devia-
tions that we need not mention here. However, it is only because
utterance-units and text-sentences are normally in correspondence that
deviations from the norm are recognizable as such.
Sentences are frequently defined, in traditional discussions of this
question, in terms of the completeness of the meaning or thought that
they express. It has often been pointed out, however, that the criterion
of completeness of meaning is difficult to apply without begging the very
question that it is intended to answer. If we assume that what we are
calling utterance-units, and more particularly simple utterance-units,
are the basic units of language-behaviour, and that, in general, utterance-
units are in correspondence with text-sentences, we can give a non-
circular account of completeness of meaning for a subset of text-
sentences, in terms of their capability of being used, without any
supporting co-text, as utterance-units. How this subset of text-sentences
is related, on the one hand, to the totality of text-sentences and, on the
other, to system-sentences is something that has been dealt with at some
length in this section. In what follows we shall be concerned, for the
most part, with the meaning either of utterance-units or of system-
sentences; and we shall generally operate with utterance-units that are
in one-to-one order-preserving correspondence with the system-
sentences from which they are assumed to be derived. In doing so, we
shall be deliberately confining ourselves within the limits of micro-
linguistic semantics.
 15
Deixis, Space and Time

15.1. Person-deixis
The term ‘ deixis’ (which comes from a Greek word meaning “pointing”
or “‘indicating’’) is now used in linguistics to refer to the function of
personal and demonstrative pronouns, of tense and of a variety of other
grammatical and lexical features which relate utterances to the spatio-
temporal co-ordinates of the act of utterance.1 As employed by the
Greek grammarians, the adjective deictic* (‘deiktikos’) had the sense of
‘‘demonstrative’’, the Latin ‘demonstrativus’ being the term chosen by
the Roman grammarians to translate ‘deiktikos’ in the works of the
Stoics, of Dionysius Thrax and of Apollonius Dyscolus, which laid the
foundations of traditional grammar in the Western world. It is worth
noting that what we now call demonstrative pronouns were referred to
as deictic articles in the earlier Greek tradition and that the Greek word
‘arthron’, from whose Latin translation, ‘articulus’, the technical term
, article* derives, was no more than the ordinary word for a link or joint.
It was only in the later tradition that the Greek equivalent of ‘pronoun’
was used; and this fact is of some significance. The point is that in early
Greek, no sharp distinction can be drawn, in terms of their forms or
syntactic and semantic function, between demonstrative pronouns, the
definite article and the relative pronoun: the term ‘article’ was at first
applied to them all, and it was chosen, presumably, because they were
regarded as connectives of various kinds.
The term ‘pronoun’ carries quite different implications from ‘article’.
It suggests that the characteristic function of pronouns is to operate as
substitutes for nouns. But to say that pronouns deputize syntactically
and semantically for nouns and that this is their primary, or basic,
1 On deixis in general, cf. Antinucci (1974), Benveniste (1946, 1956, 1958a),
Buhler (1934), Collinson (1937), Fillmore (1966, 1970), Frei (1944), Hjelmslev
(1937), Jakobson (1957), Kurylowicz (1972). The account of deixis given here
draws, eclectically, upon a variety of additional sources, not all of which have
been listed in the Bibliography.
 15.1. Person-deixis 637
function is seriously misleading in two respects. First of all, it fails to
draw the distinction between nouns and nominals (cf. 11.3): pronouns
are referring expressions, and they are syntactically equivalent to
nominals, not nouns. Secondly, to say that pronouns are primarily sub-
stitutes, whether for nouns or nominals, is to imply that their anaphoric*
function is more basic than their deictic function. The difference
between deixis and anaphora, and the connexion between them, will be
discussed in the present chapter (15.3); and we shall see that it is deixis
that is the more basic of these two kinds of pronominal reference. The
term ‘pronoun’ is now so well entrenched in the technical vocabulary of
linguistics that it would be futile to attempt to dispense with it. We must
be wary, however, of its traditional implication of substitutability for
nouns (or nominals).
The fact that the Latin-based term ‘demonstrative’ has been special-
ized in linguistic terminology in the sense that the Greek grammarians
gave to ‘deiktikos’, enables us to employ the terms ‘deictic’ and
‘deixis’ in a wider sense; and this is now common practice in linguistics.
As we shall see, deixis covers not only the characteristic function of the
demonstrative pronouns, but also tense and person, and a number of
other syntactically relevant features of the context-of-utterance. Deixis
is also involved in the philosophical notion of ostension*, or ostensive |
definition* (cf. 7.6); and it is worth noting that ‘ostensive’, ‘deictic’
and ‘demonstrative’ are all based upon the idea of identification, or
drawing attention to, by pointing. So too is Peirce’s term ‘indexical’,
which has been employed in the recent philosophical literature in roughly
the sense that we are assigning to ‘deictic’ (cf. 4.2).
By deixis* is meant the location and identification of persons, objects,
events, processes and activities being talked about, or referred to, in
relation to the spatiotemporal context created and sustained by the act of
utterance and the participation in it, typically, of a single speaker and at
least one addressee.
The grammaticalization and lexicalization of deixis is best understood
in relation to what might be called the canonical situation of utterance:
this involves one-one, or one—many, signalling in the phonic medium
along the vocal-auditory channel, with all the participants present in the
same actual situation able to see one another and to perceive the associ-
ated non-vocal paralinguistic features of their utterances, and each
assuming the role of sender and receiver in turn (cf. 2.2, 3.1, 3.2). There
is much in the structure of languages that can only be explained on the
assumption that they have developed for communication in face-to-face
638 Deixis, space and time
interaction. This is clearly so as far as deixis is concerned. Many utter-
ances which would be readily interpretable in a canonical situation-of-
utterance are subject to various kinds of ambiguity or indeterminacy if
they are produced in a non-canonical situation: if they are written rather
than spoken and dissociated from the prosodic and paralinguistic fea-
tures which would punctuate and modulate them (there are limitations,
as we have seen, upon the principle of medium-transferability: cf. 3.3);
if the participants in the language-event, or the moment of transmission
and the moment of reception, are widely separated in space and time; if
the participants cannot see one another, or cannot each see what the
other can see; and so on. Some of the complications which arise in
language-behaviour by virtue of the spatiotemporal separation of the
participants were mentioned in the previous chapter (14.2).
The canonical situation-of-utterance is egocentric* in the sense that
the speaker, by virtue of being the speaker, casts himself in the role of
ego and relates everything to his viewpoint. He is at the zero-point of the
spatiotemporal co-ordinates of what we will refer to as the deictic con-
text (cf. 14.1). Egocentricity is temporal as well as spatial, since the role
of speaker is being transferred from one participant to the other as the
conversation proceeds, and the participants may move around as they
are conversing: the spatiotemporal zero-point (the here-and-now) 1s
determined by the place of the speaker at the moment of utterance; and
it is this, as we shall see, which controls tense* (cf. 15.4).
The grammatical category of person* depends upon the notion of
participant-roles and upon their grammaticalization in particular lan-
guages. The origin of the traditional terms ‘first person’, ‘second
person’ and ‘third person’ is illuminating in this connexion. The Latin
word ‘persona’ (meaning “‘mask’’) was used to translate the Greek
word for ‘‘dramatic character”’ or “‘role’’, and the use of this term by
grammarians derives from their metaphorical conception of a language-
event as a drama in which the principal role is played by the first person,
the role subsidiary to his by the second person, and all other roles by the
third person. It is important to note, however, that only the speaker and
addressee are actually participating in the drama. The term ‘third
person’ is negatively defined with respect to ‘first person’ and ‘second
person’: it does not correlate with any positive participant role. The
so-called third-person pronouns are quite different in this respect from
the first-person and second-person pronouns.
That there is a fundamental, and ineradicable, difference between
first-person and second-person pronouns, on the one hand, and third-
 15.1. Person-detxts 639
person pronouns, on the other, is a point that cannot be emphasized too
strongly. One of the questions that we raised, but did not answer, in our
discussion of reference in a previous chapter was whether personal pro-
nouns are, in principle, dispensable (cf. 7.2). As we shall see presently,
third-person personal pronouns are obviously dispensable in favour of
demonstrative pronouns; and there are many languages that do not have
third-person personal pronouns comparable with the English ‘he’,
‘she’, ‘it’ and ‘they’. There is perhaps no language, however, in which
there are no first-person and second-person pronouns. But is it possible,
or feasible, for a language without first-person and second-person pro-
nouns to operate as a natural semiotic system under essentially the same
conditions as do the actual languages that we are familiar with (cf.
4.4)!
It is clear that first-person and second-person pronouns, as such, are
not essential. Many languages grammaticalize the category of person by
inflecting the main verb. Latin will serve as a familiar example. The
sentence ‘Odi profanum vulgus’ (“I hate the common herd’’) has no
first-person pronoun in it: it is the form odi which indicates (though not
by means of any isolable segment or morpheme) that the speaker would
normally be referring to himself if he were to utter this sentence. Latin
grammaticalizes the category of person by means of morphological
variation in the verb-form only in so far as the subject of the verb is
concerned. There are other languages, however, in which the verb is
inflected for the category of person with respect to both the subject and
the object in the case of transitive verbs and with respect to the subject,
the direct object and the indirect object in the case of verbs with a
higher valency* (cf. 12.4). All these languages, it would appear, also
have first-person and second-person pronouns, which are used in cer-
tain constructions. However, let us admit, for the sake of the argument,
that personal pronouns as such are completely dispensable, provided
that the category of person is grammaticalized morphologically in the
verb-form. Let us also discount, in the present connexion, certain well-
known differences between languages as to the way in which they
grammaticalize the category of person: whether they have a distinction
between an inclusive (‘‘I and you”’) and an exclusive (“I and he/they’’)
first-person plural; whether they have different kinds of second-person
or third-person pronouns; and so on. The question that we are con-
cerned with transcends these differences between languages. What we
are asking is whether it is possible, or feasible, for a language to dispense
completely with the grammatical category of person. For simplicity of
640 Deixis, space and time
exposition, however, we will talk throughout in terms of personal
pronouns. The point to be borne in mind is that the category of person
depends crucially upon the grammaticalization of the participant-roles,
and more especially upon the grammaticalization of the speaker’s
reference to himself as the speaker.
As we have seen, there are in English three grammatically distinct
kinds of singular definite referring expressions: proper names, definite
noun-phrases and pronouns (7.2). Now it is probably true that all lan-
guages have a class of expressions (or provide the means for constructing
and using such a class of expressions) which on semantic grounds can
be described as proper names, though in many languages they cannot be
distinguished, in terms of their internal grammatical structure, from
-noun-phrases constructed according to the productive grammatical
rules of the language. Furthermore, it is intuitively clear that a language
with proper names could dispense with personal pronouns. To see that
this is so, all we have to do is to make minor adjustments to the grammar
of English, so that, in what we will call Quasi-English, someone whose
name is ‘John Smith’ will not say J am hungry, but John Smith be
hungry (it being understood that speakers normally refer to themselves
and to the addressee by name) and the addressee will respond, not with
Are you?, but with Be John Smith? It will be noted that we have put be
rather than are and ts in these Quasi-English utterances in order to
eliminate variation with respect to the category of person from the forms
of the verb. So far, so good. The obvious practical difficulty, of course, -
is that the addressee might not know the name of the speaker. But this
is soluble, in principle, in various ways. If the speaker had reason to
believe that the addressee might not know his name, he could point to
himself (or identify himself paralinguistically in some other way) whilst
making the utterance. Alternatively, he could reply to the addressee’s
enquiry Who be John Smith? by saying Fohn Smith be the person speaking,
provided that it is understood, by an existing convention, that it is in
this way that speakers identify themselves in such circumstances. It is
also intuitively clear that personal pronouns could be dispensed with in
favour of definite descriptions. Provided that the conventions exist and
are understood, John Smith might say The person speaking be hungry.
As we shall see presently, there are certain logical problems attaching
to the analysis of such utterances. But it seems clear that, given the
existence of the appropriate conventions, the expression ‘the person
speaking’ or ‘the speaker’ (or even ‘the person here’) could serve their
purpose of referring to the speaker in terms of his participant-role; and
 15.1. Person-detxis 641
‘the listener’, or ‘the addressee’, could equally well replace the second-
person pronoun ‘you’. .
Although philosophers and logicians have generally discussed the
status of personal pronouns in relation to proper names and definite
descriptions, it is more interesting for the linguist to consider this
question from a somewhat different point of view. As we have seen,
there are many languages in which participant-roles are grammaticalized
or lexicalized, at least partly, in terms of social status or social roles (cf.
10.1). Let us, therefore, construct a rather different version of Quasi-
English, in which there are neither proper names nor personal pronouns,
but a special subset of definite descriptions (included in the full set of
definite descriptions existing at present in English) whose application in
referential and vocative function is determined by social status. Since
the principle is unaffected by the number of degrees and dimensions of
status that are lexicalized in a language-system, we will, for simplicity,
admit just one dimension and two degrees: superior and inferior,
lexicalized in the opposition ‘master’: ‘servant’. T'wo points should be
emphasized at the outset: first, that none of the assumptions that we
shall make about the conventions which determine the applicability of
‘master’ and ‘servant’ in Quasi-English is at all unreasonable in the
light of what we know of the operation of actual language-systems in
particular societies; and second, that ‘master’ and ‘servant’ are ordinary
countable nouns, which (like ‘man’, ‘tree’, ‘book’, etc.) may be used
with a determiner in singular definite noun-phrases and without a
determiner as vocative expressions. This version of Quasi-English 1s
identical with ordinary English except that it lacks the grammatical
category of person.
Let us now establish the conventions for the use of ‘master’ and
‘servant’ in vocative and referring expressions. First, it may be assumed
that in most cases of social interaction it will be clear to any arbitrary
pair of participants whether they are of equal social status or not and, if
they are of unequal status, which of them is the superior and which the
inferior. Social superiority may depend upon social role (parents being
superior to their children, teachers to pupils, and so on), sex (women
being superior to men), age (an older person being superior to a younger
person), and various other factors. What the socio-cultural correlates of
status are is of no consequence, provided that they are identifiable; that
this is a plausible assumption is clear from the fact that there are many
languages (e.g., Japanese or Korean) in which status is grammaticalized
in this way. When there is conflict between any two correlates (e.g., when
642 Deixis, space and time
an older man is talking to a much younger woman), and whenever the
participants are in any doubt as to their relative status, this conflict or
doubt will be resolved by their operating with an assumption of social
equality. And the convention which determines the use of vocative and
referring expressions in cases of social equality (a convention which
operates in many languages) is that each will refer to himself as an
inferior, and will address and refer to the other as a superior.
Given these conventions: if John Smith is of superior status, he will
say The master be hungry, and his addressee will respond with Be the
master? (in place of Are you?); if John Smith is of inferior status, he will
say The servant be hungry and his addressee will say Be the servant? ; and
if they are of actual or assumed equal status John Smith will say The
servant be hungry and his addressee Be the master? So too for vocative
expressions: the English utterances It’s raining, Sir[Fohn|my friend will
be translated into Quasi-English as It be raining, master when said by an
inferior or an equal, and as It be raining, servant when said by a superior
to an inferior.
We have now constructed a sociolinguistically plausible language-
system based on English, but lacking personal pronouns. It might be
objected that the noun-phrases ‘the master’ and ‘the servant’ are in-
directly related to participant-roles; and this is true. But it does not
follow from this fact that they are personal pronouns, or even that they
grammaticalize the category of person. Under the assumptions that have
been made, the conditions that determine the reference of ‘the master’
and ‘the servant’ when they refer to the speaker or hearer are no dif-
ferent in kind from the conditions which determine their reference in
context-independent utterances. Nor can we say that ‘master’ and
‘servant’ differ in sense or denotation, according to whether the sen-
tences ‘The master be hungry’ and ‘The servant be hungry’ are
uttered in order to make an assertion about oneself (or one’s addressee)
or about some other person. These sentences, considered as system-
sentences of Quasi-English, are no more ambiguous or indeterminate in
meaning than is the English sentence ‘The master is very kind’, which
a generation or so ago, if not to-day, might have been uttered equally
well by a servant addressing the master of the house or by some other
person with reference to the master of the house.
That Quasi-English is a possible natural language would seem to be
proved by the fact that the correlates of status mentioned above interact
with person in determining the situational appropriacy of personal pro-
nouns and honorific expressions of address and reference in many
 15.1. Person-deixis 643
languages; and in certain situations, even in English, honorific expres-
sions can substitute for personal pronouns. All that we have done in
constructing this version of Quasi-English is to generalize this possi-
bility and to simplify the conventions for deciding status.
If the arguments put forward above are valid, they show that it is
possible, in principle, for a natural language to use definite descriptions
instead of personal pronouns; and furthermore, to use definite descrip-
tions which, unlike ‘the speaker’ and ‘the hearer’, do not directly
identify their referents in terms of their participant-roles. This does not
mean that the notion of participant-roles is irrelevant to the interpreta-
tion of utterances in a language of the kind envisaged: clearly they are.
But they are not grammaticalized or lexicalized in the structure of
sentences. Throughout this work, we are concerned to maintain the
distinction between sentence-meaning and utterance-meaning. One
reason for establishing this distinction in the first place derives from the
fact that the same sentence may be uttered to perform various speech-
acts (cf. 16.1). Another reason is the related fact that the utterance or the
context-of-utterance may contain non-linguistic information which
contradicts the information that 1s linguistically encoded in the utter-
ance-signal (cf. 3.1). For example, the meaning of a sentence like ‘John
is a brave man’ is not affected by its being uttered ironically (the irony
being indicated paralinguistically). The same principle applies in the
analysis of the sentences of the version of Quasi-English that we have
just envisaged. Given the conventions which determine the interpreta-
tion of utterances in context, particular text-sentences are translatable
from Quasi-English into English, and conversely, in much the same way
as particular text-sentences are translatable from any one actual language
into another. Translation between any two languages always operates, in
principle, with respect to contextualized utterances; and the fact that the
Quasi-English The master be hungry is translatable into English some-
times as J am hungry, sometimes as You are hungry and sometimes as The
master is hungry gives us no grounds whatsoever for saying that the
Quasi-English system-sentence ‘The master be hungry’ is ambiguous.
Comparison of a language like Quasi-English with such actual lan-
guages as English or French (or indeed any of the actual languages that
have been studied and described by linguists) brings out clearly the
distinctive character of person-deixis. It is tempting for logicians, and
linguists making use of formal logic in the analysis of natural languages,
to begin by attempting to eliminate from their representation of the
meaning of the sentences of particular languages all the deictic features
644 Deixis, space and time
which make the truth-value of the propositions expressed by those sen-
tences dependent upon the context-of-utterance. We have already noted
that this involves the elimination of tense from the formal representa-
tion of the structure of propositions (cf. 14.1); and we will return to this
question later. Let us concentrate here upon the elimination of person-
deixis.
Suppose John Smith says J am hungry and that he does so, in what
may be described loosely as normal conditions, in order to make a de-
scriptive statement about himself. Has he expressed the same proposition
as some other person who says, at more or less the same time and again
in normal conditions, fohn Smith is hungry? The answer to this question
turns, in part, upon the way in which we choose to define the term
‘proposition’. It is easy to envisage circumstances under which it is
reasonable to reformulate the propositional content of an utterance like
I am hungry in terms of the propositional content of an utterance like
John Smith 1s hungry: we frequently do this when we report what others
have told us. Having heard John Smith say I am hungry, we might very
well say to someone else fohn Smith is hungry and, if asked to justify this
assertion, we might say He told me so or He said that he was. But this
process of reformulation depends upon our ability to interpret the
original utterance in the light of our knowledge of the identity of the
speaker; and we cannot in general eliminate the deictic features of an
utterance-token without adding or removing information in the process
of conversion. This will become clearer in our discussion of illocu-
tionary force* and subjective modality*, to which the notion of speaker-
involvement is central (cf. 16.1, 17.2).
It may be noted at this point, however, that, although there are cogent
reasons for saying that, if John Smith says (or believes) that he is hungry
and someone else says (or believes) that John Smith is hungry, both
John Smith and the other person have said (or believe) the same thing,
there are equally cogent reasons for denying that this is so. It is arguable
that the beliefs that we have about ourselves and the propositions that
we express about ourselves are necessarily different from the beliefs that
others have about us or the propositions that they express about us. The
philosophical problems attaching to the notion of self-knowledge need
not concern us. We shall see later, however, that, as far as their semantic
interpretation is concerned, there is much in common between first-
person pronouns and reflexive pronouns. For example,

(1) John Smith intended to kill himself

 15.1. Person-deixis 645
differs in meaning from
(2) John Smith intended to kill John Smith.
Apart from anything else, their truth-conditions are different, in that
(2) might be true even if John Smith did not know that he was John
Smith; and (1) could only be true in conditions in which John Smith
might say, truly, I intend to kill myself (regardless of whether he would
also have spoken truly in saying I intend to kill John Smith).”
Connected with this fact is the further fact that the conditions deter-
mining successful reference are different for proper names and definite
descriptions, on the one hand, and for first-person pronouns, on the
other. Indeed, the distinction that we drew in an earlier chapter between
correct and successful reference cannot seriously be drawn in relation to
first-person pronouns (cf. 7.2). The speaker will correctly and success-
fully refer to himself by means of the pronoun ‘I’ in English under
normal conditions (i.e. in situations other than those in which he acts as
an interpreter or spokesman for somebody else) only if he is performing
a particular deictic role. It is his performance of this role, and not the
truth of any presupposed identifying proposition which determines the
correct reference of ‘I’.
The point that has just been made is of the utmost importance. As we
have seen, it is possible, in principle, to eliminate the first-person pro-
noun and the second-person pronoun from English by substituting for
them various definite descriptions; and in particular, by substituting
expressions like ‘the speaker’ and ‘the hearer’. It must not be thought,
however, that the meaning of ‘I’ and ‘you’ is accounted for by saying
that ‘I’ means ‘‘the one who is (now) speaking”’ and that ‘you’ means
‘‘the one who is being addressed”’. In so far as ‘the speaker’ and ‘the
hearer’ are substitutable for ‘I’ and ‘you’ in ordinary English, they are
conventionalized pseudo-descriptions which (like ‘the author’ and
‘your lordship’) depend for their interpretation upon our intuitive
understanding of how person-deixis operates. Furthermore, the pro-
posed analysis of ‘I’ in terms of some underlying definite description
meaning ‘‘the one who is (now) speaking’’, if it is pressed to the point
at which it will do the job that it is intended to do, must be relativized to
the very utterance that contains the first-person pronoun whose meaning
it, allegedly, explicates. In other words, if ‘the speaker’ is to serve as the
equivalent of ‘I’ in The speaker is hungry, the proposition that is
2 On the philosophical aspects of this question cf. Castafieda (1968), Linsky
(1971).
 646 Deixis, space and time
expressed must be understood to be, not just “The person who is speak-
ing is hungry’’, but “'The person who is uttering this very utterance is
hungry’’; and the logical status of propositions like this, which neces-
sarily involve token-reflexivity (cf. 1.3) is, if anything, even more
obscure than is the analysis of propositions containing terms that refer
to the self.
It is difficult to escape the conclusion that person-deixis in any lan-
guage that manifests it (and, as far as we know, all natural languages do)
is something that cannot be analysed away in terms of anything else.
Deixis, in general, sets limits upon the possibility of decontextualization;
and person-deixis, like certain kinds of modality, introduces an in-
eradicable subjectivity into the semantic structure of natural languages
(cf. Benveniste, 1958a).

15.2. Demonstratives and the definite article

Demonstrative pronouns and demonstrative adjectives, like the English
‘this’ and ‘that’, as well as demonstrative adverbs, such as ‘here’ and
‘there’, are primarily deictic; and, when they have this function, they
are to be interpreted with respect to the location of the participants in
the deictic context. Roughly speaking, the distinction between ‘this’ and
‘that’, and between ‘here’ and ‘there’, depends upon proximity to the
zero-point of the deictic context: ‘this book’ means “the book (which is)
here”’ or “the book (which is) near to the speaker”’; ‘that book’ means
“the book (which is) there’ or “the book (which is) not near the
speaker” or, in explicit contrast with ‘this book’, ‘the book (which is)
farther from the speaker (than the book which is nearer the speaker)”’.
This statement of the difference between the demonstrative pronouns,
adjectives and adverbs in English is very imprecise. But it will be suffi-
cient to show the connexion between the demonstratives and the
participant-role of speaker. None of the qualifications, refinements and
extensions that would be required in a fuller account of the demonstra-
tives in English, or demonstratives in other languages, would seem to
invalidate the general point that is being made.
In the Indo-European languages what are now distinguished ter-
minologically as the definite article, the demonstrative pronouns and the
third-person pronouns are all diachronically related; and we saw above
that they were classified as articles (as also was the relative pronoun) by
the earlier Greek grammarians. Without going into the details, we may
simply note that the demonstrative pronoun is the source of both the
definite article and the third-person pronouns in the Germanic and
 15.2. Demonstratives and the definite article 647
Romance languages.® In view of the historical relationship between the
forms of what we now tend to describe as three separate classes of
lexemes, it is natural to look for some general semantic and syntactic
connexions between them; and such connexions are readily found.
First of all, it should be noted that there is a component of definiteness
in the meaning of all three classes of lexemes: ‘this’ means, roughly,
‘‘the one here’’; ‘that’ means ‘‘the one there’’; ‘he’ means ‘“‘the male
one’’; and so on. As we shall see, definiteness 1s combined with the
distinction of proximity vs. non-proximity in the case of the demonstra-
tives; and with distinctions of gender, or sex, in the case of the third-
person pronouns. The second point to note is that, generally speaking,
in English ‘this’ is marked* and ‘that’ is unmarked* (cf. 9.7): there are
many syntactic positions in which ‘that’ occurs in English and is neutral
with respect to proximity or any other distinctions based on deixis. The
third relevant point is that the distribution of ‘he’, ‘she’ and ‘it’, on the
one hand, and of ‘the’, on the other, is defective by comparison with the
distribution of ‘this’ and ‘that’. ‘This’ and ‘that’ may be used either
pronominally or adjectivally; ‘he’, ‘she’ and ‘it’ cannot be employed
adjectivally; and ‘the’ cannot be used as a pronoun.
Putting together the various facts mentioned in the previous para-
graph, we can see how English might have developed (as it did in fact
develop) from a system in which there were no third-person personal
pronouns, as such, and no definite article, but a set of two demonstra-
tives, each of which had three genders and each of which could be used
either pronominally or adjectivally. Looked at from a diachronic point
of view, then, the definite article in English is a demonstrative adjective
uninflected for gender and number, and the third-person personal pro-
nouns are demonstrative pronouns, distinguished with respect to gender
and number, but, like the definite article, unmarked for proximity.
Distinctions of proximity are lexicalized or grammaticalized in the
pronominal systems of many languages; so too are distinctions of gender,
number and, as we have seen, status. Other languages lexicalize or
grammiaticalize distinctions of gender that are based, not on sex, but on
size, shape, function, texture, etc. (cf. 11.4); or spatial distinctions that
are based upon visibility, the speaker’s normal habitat, the points of the
3 Cf, Christophersen (1939), Heinrichs (1954) and, for a wider sample of lan-
guages with definite articles, Kramsky (1972).
4 Postal (1967) also sought to relate these three classes of lexemes (or forms),
within a synchronic transformationalist framework. But he did so by taking
the articles, rather than the demonstrative pronouns, to be functionally more
basic (cf. Sommerstein, 1972).
 648 Deixis, space and time
compass, some salient landmark, etc. (cf. Bloomfield, 1935: 259). The
function of the demonstrative pronoun is to draw the attention of the
addressee to a referent which satisfies the description implied by the use
of the pronoun in terms of gender, number, status, etc.
Broadly speaking, there are two ways 1n which we can identify an
object by means of a referring expression: first, by informing the
addressee where it is (i.e. by locating it for him); second, by telling him
what it is like, what properties it has or what class of objects it belongs
to (i.e. by describing it for him). Either or both kinds of information may
be encoded in the demonstrative and personal pronouns of particular
language-systems. For example, the English demonstrative pronoun
‘this’, when it is used as a referring expression, locates the referent in
relation to the speaker; the pronoun “he’, on the other hand, gives the
addressee some qualitative information about the referent, but says
nothing about its location. The meaning of demonstrative and third-
person pronouns is comparable, in this respect, with the meaning of
definite noun phrases in English: ‘this’ is roughly equivalent to ‘the one
near me’, and ‘he’ to ‘the male one’. Clearly, the more information,
whether locative or qualitative, that 1s encoded in a deictic expression
the easier it is for the addressee to identify its referent.
In order to focus more clearly upon the nature of demonstrative pro-
nouns, we will envisage a rudimentary language-system in which there
is but a single deictic element, neutral with respect to distinctions of
gender, proximity, etc.; and we will consider how this language-system,
another version of Quasi-English, might be learnt by a child and subse-
: quently extended into something that approximates to ordinary English.°
The function of the single deictic element, we will assume, is at first
quasi-referential, rather than truly referential (cf. 7.5). We can think of
this deictic as meaning something like “‘Look!”’ or ‘“‘There!’’ Such
forms as Latin ecce, French voici/voila, etc., are worth noting in this
connexion: their function is quasi-referential, rather than purely
referential; and it is not always clear whether they are being used to
draw attention to an entity or to a place. It has often been suggested that
children do in fact pass through a stage, fairly early in the acquisition of
language, at which their utterances contain a deictic element of this
kind; and that it is up to the hearer to guess what feature of the environ-
5 What follows is developed in greater detail in Lyons (1975). A similar
approach is taken by Ferch (1975). For relevant psychological work, both
theoretical and empirical, cf. Bates (1976), Bruner (1974/5), E. V. Clark
(1977), H. H. Clark (1973), Miller & Johnson-Laird (1976: 394ff).
 15.2. Demonstratives and the definite article 649
ment is engaging the child’s interest. Other words may also be used,
holophrastically, in the same way ; and witha variety of semiotic functions
(cf. 3.5).
At a later stage, when the child is producing two-word and three-
word sequences, the deictic (D) may be combined with another word in
two-word utterances which we can represent as D-+-X or X-+-D, where
D is the deictic and_X is a variable ranging over a small number of other
words. Some or all of these other words may at first have been interpreted _
by the child as names (cf. 7.5). We will assume, however, that by this
stage in the acquisition of language, the distinction between names,
common nouns and verbs is emerging. Utterances such as Book D or |
D book might be interpretable as “I want that/this book’’, “‘Give me
the book’’, ‘Look! A book’’, ‘‘That’s a book’’, and so on. It will not
always be clear, in particular instances, whether the utterance is to be
regarded as one text-sentence, two text-sentences or a sentence-
fragment, or what its semiotic function is. We are not concerned here
with the way in which the child’s developing control of a language
enables him to differentiate and make explicit various semiotic functions.
We will concentrate instead upon the development of a distinction
between the referential and the predicative function of the deictic ele-
ment in simple utterances, both of these functions arising from what we
are assuming to be its prior quasi-referential function.
The terms ‘subject’ and ‘predicate’ have been defined in various
ways in linguistics; and distinctions have been drawn between the
grammatical subject and predicate of a sentence and its logical and
psychological subject and predicate (cf. 12.7). It is perhaps reasonable
to assume, however, that in the earlier stages of language-acquisition no
such distinctions can be drawn.* The grammatical subject will be the
expression which refers to what is being talked about (which may or may
not be contextually given) and the grammatical predicate will be the
expression which says something about the referent of the subject-
expression. It is perhaps also reasonable to assume that the subject will
normally precede the predicate in utterances which can be interpreted
as statements: initial position in the utterance correlates quite highly in
many languages with the function of being the thematic subject (1.e. of
being the expression which identifies what is being talked about, cf. 12.7).
If Book D occurs and is interpretable as a single text-sentence with a
6 This section was written before the appearance of Strawson (1974); it is
gratifying to note that what is here suggested as plausible fits in well with
Strawson’s account.
 650 Deixts, space and time
subject and a predicate, it will mean ‘“‘The book is there’’. D book, on
the other hand, will mean “That is a book”’ or “That place has a book
in it”’. The deictic, it will be observed, may refer to either an entity or a
place; and this ambivalence is the source of a subsequent syntactic dis-
tinction between its use as a pronoun and its use as an adverb. As a
predicative expression it always has an adverbial function.
There is perhaps no fully developed language with a single deictic
element that operates syntactically in this way. But structures of the
kind outlined in the previous paragraph are found in many languages;
and, provided that allowance is made for the differentiation of deictics,
variously in different languages, in terms of their adverbial and pro-
nominal function, on the one hand, and of their encoding of distinctions
of descriptive or locative information, on the other, such structures
involving deictics can perhaps be regarded as universal in the onto-
genesis of languages. English, as we have seen, distinguishes two
adverbial deictics in terms of proximity (‘there’:‘here’) and two adjec-
tival deictics in terms of proximity and number (‘this’: ‘that’, with the
forms this, these: that, those): i.e. ‘this’ and ‘here’ are proximal* and
‘that’ and ‘there’ are non-proximal*. The situation with respect to
pronominal deictics is more complex: the so-called third-person singular
| pronouns (‘he’, ‘she’, ‘it’) are distinguished for gender, but not for
proximity, whereas the demonstrative pronouns are distinguished for
proximity, and number, but not for gender, and their forms are identical
with the forms of the demonstrative adjectives. The definite article
behaves syntactically like the demonstrative adjectives, but is neutral
with respect to proximity, gender and number: and it derives, his-
torically, from the non-proximal demonstrative adjective ‘that’.
Whether there is any fixed sequence in the acquisition by children of
these semantic, syntactic and morphological distinctions in English is,
in the present state of research, uncertain. It is quite conceivable, of
course, that they will be acquired at different stages by different chil-
dren. For simplicity of exposition, however, we will here assume the
following stages in the development of the English system: (i) the
distinction of formally different pronominal and adverbial deictics, D,
and D,; (ii) the distinction of proximity in both D, and D,; (iii) the
distinction of gender in D, (but not D,); (iv) the adjectivalization of D,;
(v) the development of the definite article. We will take no account of
the distinction of singular and plural, or of any of the other grammatical
categories in English.
We have made the assumption that in the earliest stage of language-
 15.2. Demonstratives and the definite article 651
acquisition there will be a single deictic (of perhaps indeterminate form)
whose function, like that of an ostensive gesture, is to draw the addres-
see’s attention either to a particular entity or to a particular region in the
environment. One reason for making this assumption is that in some
languages deictics can be employed in this way. When the deictic is
being used to refer to a person or object in the situation, it is, in tradi-
tional terminology, a demonstrative pronoun; when it is being used to
refer to a place, it is a locative adverb. Introducing this distinction into
the version of Quasi-English that we are building up, we can say that
the utterance D, nice means “‘ He/she/it/this/that is nice’’, D, nice means
“It is nice here/there”’ (or ‘“This/that place is nice’’); D, book means
‘““This/that is a book’’, D, book means “’There’s a book here/there’’;
and Book D, means “The book is here/there’’.
The next stage, we are assuming, results in the differentiation of D,
into ‘this’ and ‘that’ and of D, into ‘here’ and ‘there’. This deictic
opposition is characterized, in English, by the property of semantic
marking, ‘that’ and ‘there’ being the unmarked members of the
opposition proximal: non-proximal (cf. 9.7). The utterance This nice in
Quasi-English will therefore mean “The entity near me is nice’’, but
That nice will mean either “The entity not near me is nice”’ or ‘The
entity (whose location is unspecified) is nice’’. When ‘that’ is employed
in its deictically neutral sense as a referring expression it gives the
addressee no information about the referent other than the fact that it 1s
an entity rather than a place; and it would be natural to suppose that the
form that would be unstressed in these circumstances and its utterance
would not be accompanied by a paralinguistic gesture pointing to the
entity in question. There mice and Book there will also be interpretable
somewhat differently according to whether ‘there’ is being used in
(explicit or implicit) contrast with ‘here’ or neutrally with respect to
proximity. We will come back presently to the neutral sense of ‘there’.
We can now extend the system by introducing a distinction of gender
based on sex. In doing so, we will assume that ‘this’ and ‘that’ can no ,
longer be used pronominally to refer to persons (except in those con-
structions in which they can be so used in ordinary English) and that
‘he’ or ‘she’ is employed instead; and we will also assume that by now
‘it’ has replaced ‘that’ in its deictically neutral sense. In terms of this
analysis of the meaning and syntactic function of the third-person
pronouns, ‘he’, ‘she’ and ‘it’ are all variants, as it were, of the deictic-
ally neutral pronominal ‘that’; they differ in that ‘he’ encapsulates the
meaning “‘male”’, ‘she’ encapsulates the meaning “female’’, and ‘it’ is
 652 Deixts, space and time
a syntactically determined variant of ‘that’ which encapsulates neither
“male” nor “female’’ but something like “‘non-personal’’.”? Encapsula-
tion can, of course, be formalized, and plausibly enough perhaps in this
case, in terms of universal sense-components (cf. 9.9). The system now
forces the speaker, when it comes to referring by means of a pronoun to
an entity that is present in the environment, to decide whether that
entity is a person or not, and if it is a person to select ‘he’ or ‘she’
according to the sex of the referent and, in using ‘he’ or ‘she’, to give
the addressee no information (except paralinguistically) about the
deictic proximity or remoteness of the referent. If the referent is not a
person, the speaker will normally identify the referent for the addressee
in terms of the deictic opposition of ‘this’ and ‘that’. Under certain
conditions, however, he can refer to a non-personal entity by means of
‘it’, giving the addressee no information about its location. Since we are
not concerned to account for all the environments in which the demon-
stratives and personal pronouns occur in English, or for all the factors
which determine the selection of a particular demonstrative or pronoun,
' we will not go further into this question.
Let us now turn to the demonstrative adjectives, ‘this’ and ‘that’, as
they are employed in definite noun-phrases (such as ‘this boy’ and ‘that
boy’). The first point to be noted is that singular definite noun-phrases
are syntactically equivalent, not to common countable nouns, but to
proper names or pronouns: ‘John’, ‘he’ and ‘this/that boy’ (as well as
‘the boy’) can be substituted for one another in English sentences, and
each of them, unlike ‘boy’, can be used as a singular definite referring
expression (cf. 7.2, 11.2). In this respect the demonstrative adjectives
differ from qualitative adjectives (such as ‘good’, ‘nice’, etc.): ‘good
boy’ is syntactically equivalent to ‘boy’. On the other hand, it is arguable
that the relationship between “‘A/The boy is good’”’ and “‘the good boy”’
is the same as the relationship between ‘‘A/The boy is here’”’ and “‘this
boy” or “‘A/The boy is there”? and ‘“‘that boy’’. These facts would
suggest that the demonstrative adjectives have a certain grammatical
ambivalence. Although the demonstratives ‘this’ and ‘that’ are tradi-
tionally regarded as adjectival modifiers of a head noun in such phrases
as ‘this boy’ and ‘that boy’, one might equally well think of them, from
a semantic point of view, as pronouns combined with an appositional
noun or nominal. In fact, there are two rather different ways in which
” Needless to say, this is not intended to be a complete account of the meaning
of ‘he’, ‘she’ and ‘it’ in English. But it captures what I take to be the basic
semantic difference between them.
 15.2. Demonstratives and the definite article 653
we can interpret ‘this boy’: (i) as meaning “this (entity) - a boy” or
(ii) as meaning “‘the boy (who is) here”. The second interpretation
can be accounted for in a transformational grammar by a very general
rule which adjectivalizes a predicative expression (in this case a deictic
adverbial) and embeds it within a noun-phrase as a modifier of the noun
(cf. 10.3). The former interpretation can be explained on the basis of a
somewhat different, though still quite general, rule which takes a
predicative noun and brings it into an appositional relationship with a
deictic pronoun.® These two processes may be summarized as follows in
relation to Quasi-English: (i) NX & ND, = (DN)X (e.g., ‘Animal big’
& ‘Animal here’ = ‘This animal big’); (11) D.X & D,N = (DN)X (e.g.,
‘This big’ & ‘This animal’ => ‘This animal big’).
The effect of these proposed derivations, it will be observed, is to
make ‘this’ an adjective under the interpretation accounted for by (i).
But, as we have already noted, ‘this boy’ is never syntactically equivalent
to ‘boy’, as ‘good boy’, which is also derivable by (1), is syntactically
equivalent to ‘boy’. There are various ways in which we can remedy this
deficiency; and to discuss the question in detail would take us too far
from our present concerns (cf. Lyons, 1975). For present purposes, we
may simply opt for one possibility and assume that it is at least plausible:
namely, that noun-phrases like ‘this animal’ are derivable in a trans-
formational grammar by applying rule (1) to the output of rule (11). The
syntactic ambivalence of ‘this’ and ‘that’ is now accounted for by
treating ‘this animal’ as being syntactically equivalent to ‘this big
animal’ (i.e. as being derived, as it were, from ‘this here animal’, where
‘here’ is an adjectival modifier of ‘animal’ and ‘here animal’ is in
apposition with the pronoun ‘this’). This proposed derivation of definite
noun-phrases with demonstratives implies that the so-called demonstra-
tive adjectives include a pronominal deictic component and also an
adjectivalized predicative deictic.
The final stage in the development of this part of the grammar of
English, as we are presenting it here, involves the introduction of the
definite article into the system as a replacement for ‘that’ in those
positions in which it derives from the non-proximal demonstrative
interpretable in its neutral sense. As we shall see in the next section,

8 The status of apposition in current versions of transformational grammar is

rather uncertain (cf. Burton-Roberts, 1975). It is my assumption that,
whether apposition is to be accounted for by relative-clause reduction or
otherwise, it can be based upon either a predicative or an. equative structure
(cf, 12.2). These alternatives are allowed for in Lyons (1975).
 654 Deixis, space and time
there is some correlation between phonological stress and the deictic use
of the demonstrative and personal pronouns in English, and there is
historical support for the view that the definite article results from the
phonological reduction of the unstressed forms of what is diachronically
identifiable with ‘that’ in certain positions. We are not concerned, in
principle, with the historical origins of the English pronouns and
definite article. But their historical development is relevant in so far as
it shows that there is some plausibility in our analysis. Our intention has
been to construct part of a language-system which is similar in many
respects to English, though simpler in its grammatical structure, and to
show the semantic and syntactic relationship which holds between
demonstratives, third-person pronouns and the definite article with
respect to deixis.
According to this analysis of the function and meaning of the definite
article, it is neither a pronoun nor an adjective, but a form which
- amalgamates both a pronominal component and an adjectivalized predi-
cative component; and each of these is to be understood as being
unmarked for the deictic distinction of proximity and remoteness. But
what is the point of this analysis?
When we identify an object by pointing to it (and this notion, as we
have seen, underlies the term ‘deixis’ and Peirce’s term ‘index’: cf.
15.1), we do so by drawing the attention of the addressee to some spatio-
temporal region in which the object is located. But the addressee must
know that his attention is being drawn to some object rather than to the
spatiotemporal region. This is accounted for by the differentiation of D,
and D, in the system that we have constructed. Now it is not generally
possible in English to use a referring expression (other than a proper
name) which does not simultaneously inform the addressee that some-
thing is being referred to and give him some further information about
the location of the referent and/or about one or more of its properties.
Let us, however, envisage a system in which D, is used to point to an
object, as it were, without locating it anywhere in the deictic space.
English would be such a system, if it were possible to say not only
He] She|It is good and This/That ts good, but also The 1s good, it being
understood that ‘the’ could refer to any entity regardless of its location
or properties. The deictically neutral pronominal component of the
English definite article can be thought of as having just this function: it
informs the addressee that some specific entity is being referred to without
however giving him any locative (or qualitative) information about it.
Philosophers, as we saw in an earlier chapter, have devoted a lot of
 15.2. Demonstratives and the definite article 655
attention to the question of uniquely referring expressions and they have
emphasized the similarity, from this point of view, between proper
names and noun-phrases introduced by the definite article (cf. 7.2).
Many of them have claimed that the use of the definite article in a
singular definite referring expression implies or presupposes that there
is one and only one entity that satisfies the descriptive information con-
tained in the noun-phrase. But there is no reason to associate any
implication or presupposition of uniqueness with the definite article as
such. When the speaker refers to a specific individual, by whatever
means, he tacitly accepts the convention that he will provide any in-
formation (not given in the context) that is necessary for the addressee
to identify the individual in question. Uniqueness of reference, under-
stood in this sense, is always context-dependent; and it applies just as
much to the use of the personal pronouns and the demonstratives (and
indeed to the use of proper names), as it does to the use of the definite
article. The pronominal component in the definite article has exactly the
same function as has the same component in the meaning of the demon-
strative and personal pronouns: that of informing the addressee that a
specific individual (or group of individuals) is being referred to. When
the definite article is used, such information as is necessary for the
addressee to identify the referent is encoded in other parts of the noun-
phrase. If the participants believe that there is one and only one in-
dividual of which it is true to say that it is a unicorn, it will of course be
sufficient, in any context, to refer to it by means of ‘the unicorn’
without giving any further locative or qualitative information about it.
It does not follow from this fact, however, that the phrase ‘the unicorn’
of itself carries the presupposition or implication that one and no more
than one unicorn exists. Even such phrases as ‘the King of France’ are
interpretable, in principle, as implying or presupposing no stronger sense
of uniqueness than the context-dependent uniqueness that a phrase like
‘the cat’ implies or presupposes (cf. 14.3).
The second component in the definite article is the adjectivalized
deictic adverbial ‘there’, interpreted in its neutral sense (cf. Thorne,
1972). As we have seen, the English demonstratives ‘this’ and ‘that’,
used as deictics, can be understood as instructing, or inviting, the
addressee to direct his attention to a particular region of the environ-
ment in order to find the individual (or group of individuals) that 1s
being referred to. The definite article, when it is used deictically (with
or without any accompanying paralinguistic modulation of the expres-
sion of which it forms a part), 1s to be understood as instructing, or
656 Deixts, space and time
inviting, the addressee to find the referent in the environment, without
however directing his attention to any particular region of it. In so far
as the very fact of pointing to something commits the person who is
pointing to a belief in the existence of what he is pointing at, the use of a
deictic pronoun carries with it the implication or presupposition of
existence. The act of reference does this anyway: but there is perhaps
some reason to believe that there 1s a deeper connexion between detxis
and the presupposition of existence. When expressions containing the
definite article are used non-deictically, the adjectivalized adverbial
component of the definite article will inform the addressee that he will
find a referent satisfying the description somewhere; and the presump-
tion is that the addressee has all the information he needs in order to
find it. Just as the neutral sense of the demonstrative pronoun ‘that’ is
derived by abstraction from the gesture of pointing, so the neutral sense
of the adverbial ‘there’ is derived by abstraction from the notion of
location in the context-of-utterance. More will be said about this after
we have dealt with the relationship between deixis and anaphora (cf.
15.3).
At first sight, it might appear that the derivation of the definite
article proposed here would have the effect of making ‘The cat is here’
contradictory and ‘The cat 1s there’ tautologous. But this is not so. As
we saw in our discussion of semantic marking, a sentence like ‘That dog
is a bitch’ is not a contradiction and ‘That dog is a dog’ is not a tauto-
logy, provided that ‘dog’ is taken in the neutral sense in the subject —
noun-phrase. Nor is it the case that ‘That dog is here’ 1s necessarily
contradictory. As we shall see, the basically deictic distinction of
‘this’: ‘that’ and ‘here’: ‘there’ is extended to a variety of non-deictic
dimensions; there is no conflict therefore between the proximal and
the non-proximal expressions in the underlying structure.
The deictic function of demonstratives is far more complex than our
somewhat schematic account here might suggest. But what has been said
should be sufficient to establish at least the initial plausibility of the
hypothesis that demonstratives are more basic than either third-person
personal pronouns or the definite article in that they can all be derived
from a deictic element which might be first used and understood, in the
acquisition of language, as having quasi-referential function. It is not
being suggested that the five stages of development proposed here
correspond with five chronologically distinct periods in the child’s
acquisition of English. What 1s essential to the hypothesis is merely the
assumption that the function of demonstrative pronouns in languages is
 15.3. Deixis, anaphora and the universe-of-discourse 657
first learned in actual situations-of-utterance with reference to entities
present in the situational context. Taking this to be their basic and onto-
genetically prior function, we can see how they might later come to be
used with reference to entities removed in space and time from the
situation-of-utterance. It seems clear that the design-features of
reflectiveness* and displacement* both support and depend upon this
development in the use of the demonstratives (cf. 3.4). It also seems
clear that the hypostatization* of second-order and third-order entities,
which language makes possible (and some languages, apparently, more
readily than others), represents yet a further stage in the process of dis-
placement; and it trades, once again, upon the existence of a gram-
matical framework for referring to entities by means of definite noun-
phrases (cf. 11.3).
As we saw in a previous chapter (7.2), it 1s easier to conceive of a
language without proper names than it is to envisage a language operat-
ing successfully without the means of constructing an unlimited number
of definite descriptions. But definite referring noun-phrases, as they
have been analysed in this section, always contain a deictic element. It
follows that reference by means of definite descriptions depends ulti- |
mately upon deixis, just as much as does reference by means of demon-
stratives and (as we saw in the previous section) personal pronouns.
However that may be, it is clear that a language which does not have
demonstrative pronouns (if there is any such language) is radically
different from one that does.
The thesis that the referential function of definite descriptions and
personal pronouns cannot be accounted for except in terms of detxis
might seem to be refuted immediately by the fact that both definite
descriptions and personal pronouns have anaphoric, as well as deictic,
uses. In the next section, however, it will be argued that anaphora also
depends ultimately upon deixis.

15.3. Detxis, anaphora and the universe-of-discourse

As we have seen, the term ‘pronoun’ owes its origin to the view that
there are certain forms or expressions whose function it is to operate as
substitutes for nouns (15.1). Since the distinction between nouns
(expressions, including lexemes, of the class N) and nominals (expres-
sions, including lexemes, of the class NP) was not clearly drawn in
traditional grammar and is still not drawn in much of the more recent
work in the field of theoretical and descriptive syntax, the term ‘pro-
noun’ is used by most linguists to cover both noun-substitutes and
658 Deixis, space and time
nominal-substitutes. It is also used, as ‘noun’, ‘verb’ and ‘adjective’
are, to cover both forms and expressions (including lexemes): i.e. to
refer to he, him, they, them, etc., and also to ‘he’, ‘they’, etc. (cf. 1.5).
That this distinction between forms and expressions can be drawn, as
far as certain pronouns are concerned, is obvious enough. It is obvious,
too, that at least some pronouns (such as ‘he’, ‘she’, ‘they’ in English)
must be listed in the lexicon as lexemes. (Strictly speaking, this is
obvious, as we shall see in this section, only if it is accepted that these
pronouns are such that their occurrence in system-sentences cannot be
fully accounted for by means of a grammatical process of substitution.)
What is not clear, however, is whether everything that is a pronoun-
form is the form of some pronoun-lexeme. All the pronouns that have
been mentioned so far in this chapter are members of the expression-
class of nominals; and we can reasonably assume that they would be
listed in the lexicon. But there are forms that are traditionally described
as pronouns of which it is not at all clear that they are forms of lexemes.
For example, the form one which occurs in the sentence
(1) I want the red scarf, not the blue one
might well be introduced by means of the transformational substitution
of one for either the form scarf or the lexeme ‘scarf’; and, if the form one
actually occurs in the transformational rule itself, it does not need to be
derived by means of morphosyntactic or morphological rules operating
upon the output of a lexical insertion-rule which introduces the lexeme
‘one’. The distinction between forms and lexemes may be drawn
differently in different grammatical descriptions of the same language;
and in certain grammatical descriptions it may not be drawn at all. In
what follows, we shall frequently be forced by our notational conven-
tions to commit ourselves to one view rather than another of the status
of the individual linguistic entities that we have occasion to refer to.
Unless their status as forms or expressions, on the one hand, or as forms
or lexemes, on the other, is relevant to the point at issue, we shall not
attempt to justify one classification, rather than another, of particular
linguistic entities.
We shall be concerned almost exclusively with demonstrative and
third-person personal pronouns, which, unlike the form one in (1) above,
are nominal-substitutes, rather than noun-substitutes. In so far as it is
necessary to distinguish these two subclasses of pronouns, by means of
an appropriate terminological convention, we can do so by calling
nominal-substitutes pro-nominals* and noun-substitutes pro-nouns*,
 15.3. Detxis, anaphora and the universe-of-discourse 659
This convention has the advantage that it leaves the traditional term
‘pronoun’ and the corresponding adjective ‘pronominal’ free for more
informal reference to the forms and expressions that we shall be dis-
cussing. It has the further advantage that the technique of creating
hyphenated terms for various kinds of substitutes can be freely ex-
tended: a pro-verb will be a substitute for a verb; a pro-verbal will be a
substitute for a verbal; a pro-locative will be a substitute for a locative;
and so on. Many such terms have been created, and are now more or less
widely used, by linguists working in what may be referred to, rather
loosely, as the Bloomfieldian tradition. For it was Bloomfield (1935) and
his followers who extended and generalized the notion of substitution as
a grammatical process or relationship (cf. Crymes, 1968); and this,
as we shall see, is the historical source of what has been until very
recently the standard treatment of pronouns in Chomskyan generative
grammar.
It has already been mentioned that pronouns are traditionally con-
ceived as having two distinct, though related, functions: deixis and
anaphora*. Their anaphoric function may be illustrated by means of
utterances like
(2) fohn got home late and he was very tired
in which ‘he’ may be said to refer to its antecedent*, the expression
‘John’. The antecedent of an anaphoric pronoun is an expression which,
as the term ‘antecedent’ implies, normally precedes the correlated
anaphoric pronoun in the text or co-text (cf. 14.1). In certain languages,
and under certain conditions, the antecedent may follow the correlated
anaphoric pronoun. Some linguists, following Bihler (1934: 121), dis-
tinguish between anaphora and cataphora*, according to whether the
pronoun follows or precedes the expression with which it is correlated.
We will adhere to the more traditional usage, according to which
‘anaphora’ covers both normal backward-looking anaphoric reference
and the less normal forward-looking, or anticipatory*, anaphoric
reference. Relative pronouns, unlike demonstratives, are restricted to
anaphoric function; and the term ‘relative’, in this sense, derives in fact
from the Latin translation of the Greek ‘anaphorikos’. We shall not be
concerned explicitly with relative pronouns, though much of what is
said about the anaphoric reference of demonstratives applies also to
relatives.
Underlying the notion of anaphoric reference is the principle of
substitution, in the sense in which Bloomfield and his followers use the
 660 Deixis, space and time
term ‘substitution’. But there are, in fact, two different ways of defining
the notion of anaphoric reference. We can say, as we have done in the
previous paragraph, that the pronoun refers to its antecedent; and this 1s
perhaps the more traditional formulation of the relation between a
- pronoun and its antecedent. Here the term ‘refer’ can be traced back to
the Latin ‘referre’, which was used to translate the Greek ‘anapherein’
and, in this context, meant something like “bring back’’, ‘“‘recall”’ or
‘repeat’. Alternatively, we can say that an anaphoric pronoun refers to
what its antecedent refers to. This alternative formulation, which is
based on a quite different sense of the term ‘refer’, has the advantage of
bringing anaphoric reference within the scope of the current philo-
sophical concept of reference (cf. 7.2) and, more important, of making
it possible, as we shall see, to relate anaphora and deixis in terms of a
single notion of pronominal reference. Furthermore, by adopting this
alternative, less traditional usage of the term ‘refer’, we can avoid the
confusion that often arises in modern treatments of anaphora. Hence-
forth, then, we will not say that a pronoun refers to its antecedent but
rather that it refers to the referent of the antecedent expression with
which it is correlated.
We can illustrate the point that has just been made by considering
briefly a few English sentences. The first is
(3) My friend looked up when he came in
(where the form he bears normal, non-emphatic and non-contrastive,
stress). What does the expression ‘he’ (of which he is a form) refer to?
If he is unstressed (i.e. bears normal stress), the expression of which it
is a form will probably be anaphoric, rather than deictic: it will be co-
referential* with (i.e. have the same referent as) some antecedent
referring expression. The antecedent will be either ‘my friend’, since
this expression satisfies the conditions which determine the reference of
‘he’ (roughly, the possibility of its being used to refer to a male person
or animal), or some other expression in the preceding co-text. Within
the limits of microlinguistic semantics there is no way of deciding be-
tween these alternatives. Microlinguistic semantics is concerned with
reference only to the extent that it specifies the conditions which deter-
mine the potential reference of expressions in terms of the sense and
denotation of the expressions and the relevant grammatical and phono-
logical rules in particular languags-systems.
Let us now consider the sentences
(4) When he came in, my friend looked up
 15.3. Detxis, anaphora and the universe-of-discourse 661
and
(5) He came in and my friend looked up
(the form he being unstressed in both sentences). According to the
grammatical conditions which govern the potential reference of pro-
nouns in English ‘he’ may be co-referential with ‘my friend’ in any
utterance of (4), but cannot be co-referential with ‘my friend’ in any
utterance of (5). It is noteworthy that anticipatory anaphora does not
hold between co-ordinate clauses in compound sentences; and this
restriction seems to apply in a number of languages.® It is perhaps
reasonable to hypothesize that this is because a complex sentence is a
grammatically more cohesive unit than a compound sentence. However
that may be, anticipatory anaphora is far from being as free as the more
normal, backward-looking, anaphora.
Rather different from (4) and (5) is the sentence
(6) John looked up when hé came in
(with he bearing heavy stress). On the assumption that the assignment of
the prosodic feature of heavy stress in English is to be accounted for by
the rules which generate system-sentences (and this is the view taken by
most linguists), what can we say about the potential reference of ‘he’
in this sentence? Since he bears heavy stress, the expression ‘he’, of
which it is a form, may be either deictic or anaphoric in particular
utterance-tokens. If ‘he’ is deictic, there will usually be some con-
comitant paralinguistic feature (a nod of the head, a gesture with the
hand, etc.) which draws the attention of the addressee to the referent in
the situation-of-utterance. If ‘he’ is anaphoric, it will refer either to
John or to the referent of some other antecedent in the preceding co-text
under the normal conditions which determine the reference of anaphoric
pronouns; and if it refers to John, there must be some kind of emphasis
or contrast involved. The prosodic feature of stress is relevant to the
reference of ‘he’ only in so far as it increases the probability of a deictic
interpretation. But whether the pronoun is interpreted as having
anaphoric or deictic reference (or both) would seem to depend primarily
upon the context-of-utterance and cannot be decided within a micro-
linguistic analysis of the structure and meaning of the sentence. It is
worth noting, too, that the emphatic or contrastive function of heavy
® To say that anticipatory anaphora can never hold between co-ordinate
clauses, in English and other languages, may be to make too strong a state-
ment. Anticipatory anaphora under these conditions is certainly less normal
than it is in cases like (4).
662 Deixis, space and time
stress is independent of its deictic function; and that the one does not
exclude the other. If the pronoun is deictic and non-anaphoric, it may
or may not be also contrastive or emphatic; so too if it is both deictic and
anaphoric; it is only when the pronoun is non-deictic that stress must be
interpreted as contrastive or emphatic. Stress, then, is not a sufficient
condition of deixis; and it requires paralinguistic support. Finally, it
should be noted that intonation is also relevant to the determination of
the reference of pronouns in particular utterance-tokens: if (6) is uttered
with the highest point of the intonation-contour on he, ‘he’ will prob-
ably be taken to be deictic; but if he is pronounced on a lower pitch
than John or up, ‘he’ will probably be taken to be anaphoric (and con-
trastive). It is important to realize, however, that the sense of the pro-
noun ‘he’ is constant over all the interpretations that we have considered;
and its reference in particular utterance-tokens is determined partly by
its sense (i.e. its contrast in terms of gender or number with ‘she’, ‘it’,
‘they’ and with a variety of other expressions) and partly by the general
conditions which govern anaphora and deixis in English.
Generative grammarians have often been inclined to underestimate
the role played by deixis in the interpretation of utterances. They have
tended, until recently, to handle anaphora in terms of the pronominaliza-
tion* of an antecedent expression under a condition of lexical or
referential identity.!° It has generally been assumed, for example, that
the sentence ‘John looked up, when he came in’ is to be derived from a
deep structure or semantic representation in which the subject of the
second clause is ‘John’. In the earliest formulation of the rules of
pronominalization (in work based on Chomsky, 1957), no account was
taken of referential identity. The substitution of ‘he’ for ‘John’ in the
second clause was optional and was made dependent solely upon the
identity of the two subject-expressions in the underlying structure (cf.
Lees & Klima, 1963).
Subsequently, with the incorporation of a semantic component in
what is currently described as the standard version of transformational
grammar and with the development by Chomsky (1965) of a more
explicit notion of deep structure (cf. 10.3, 10.5), pronominalization was
made conditional upon both lexical and referential identity. The referen-
10 'The term ‘pronominalization’ is being used here in the sense in which it was
originally employed: i.e. for the conversion into a pronoun (by means of a
substitution transformation) of something (in the terminology of this book,
an expression) which, at some deeper level of grammatical representation, is
not a pronoun. The term ‘pronominalization’ is employed differently in
Jackendoff (1972).
 15.3. Detxis, anaphora and the universe-of-discourse 663
tial identity or non-identity of nominal expressions is shown by means
of referential indices* assigned to them in the deep structure of sen-
tences and informally represented as numerical subscripts. For example,
‘Johnj+looked-++up-++ when-+ Johni-++came-++in”’ is an informal repre-
sentation of the deep structure of a sentence in which ‘John’ in the first
clause is co-referential with ‘John’ in the underlying structure of the
second clause; and “‘ Johnj-+looked-+-up+ when+Johnj-+-came-+in”’ is
an informal representation of a sentence in which the two occurrences
of ‘John’ in the deep structure are not co-referential.' It must be
appreciated, in this connexion, that under Chomsky’s (1965) formula-
tion of the conditions of referential identity, the absolute numerical
value of the subscripts is irrelevant: the indices do not identify the
referents of expressions, but merely show whether the expressions are
co-referential or not. Chomsky’s proposal, therefore, does not bring
reference as such within the scope of microlinguistic syntax and seman-
tics. The so-called generative semanticists, in contrast with Chomsky,
make pronominalization conditional solely upon referential identity; and
furthermore they assign what might be called absolute values to the
referential indices (cf. McCawley, 1969). That 1s to say, they interpret
the numerical subscript attached to an expression in the semantic
representation of a sentence as designating a particular individual in the
universe-of-discourse.
We will not go into the details of these different conceptions of the ,
role of referential identity in generative grammar. Anyone who wishes
to maintain the distinction between sentence-meaning and utterance-
meaning will reject, without more ado, the proposals made by the
so-called generative semanticists. But Chomsky’s treatment of pro-
nominalization is hardly more attractive. In fact, it is extremely doubtful
whether anything more than a very restricted concept of pronominaliza-
tion is required within a theory of grammar which is restricted to the
generation of system-sentences.!?
1! The subscripts 7 and 7 are to be interpreted as variables ranging over the set
of positive integers, under the tacit further condition that 1 # 7. Technically,
within Chomsky’s formalization of the standard theory of transformational
grammar, these numerical indices are features, or properties, comparable
with the syntactic features of concreteness or countability that are assigned to
nouns by the rules of the base-component of the grammar. This, in itself,
constitutes something of a problem, since it is nominals (NPs), not nouns,
that serve as referring expressions.
12 In his most recent publications, Chomsky has been putting forward a rather
different theory of pronominalization, and more particularly of pronoun-
deletion.
664 Deixis, space and time
There are innumerable utterances of English and other languages in
which the pronouns which occur in them have (in particular utterance-
tokens) a purely deictic function: e.g., What’s that in your hand?; For
heaven's sake, he’s grown a beard!3 Pronominalization, as a grammatical
process, is obviously irrelevant to the generation of the corresponding
system-sentences, except on the unverifiable, and unnecessary, assump-
tion that individuals, even when we first encounter them, are invariably
categorized in terms of a proper name or descriptive expression (‘that
man’, ‘that thing’, etc.). There are also innumerable utterances in
which the pronouns (in particular utterance-tokens) are co-referential
with antecedent expressions in a preceding text-sentence; and pro-
nominalization is, once again, irrelevant to the generation of the
corresponding system-sentences. For example, ‘John looked up when
he came in’ must be generated from a deep structure in which ‘he’
occurs as the subject of the second clause in order to account for the
interpretation of ‘he’ (in particular utterances of the sentence) as
referring to someone other than John. Given that this is so, there is no
convincing reason why the occurrence of ‘he’ in the second clause of
‘John looked up when he came in’ should ever be accounted for in
terms of the pronominalization of ‘John’.
The conditions which determine the reference of third-person
personal pronouns in English utterances, as we have seen, are roughly
as follows: (i) a pronoun can refer deictically to any entity (or set of
entities) in the situational context that satisfies the descriptive content of
the pronoun (provided that the pronoun is shown to be deictic by some
appropriate paralinguistic modulation of the utterance and, optionally
in certain instances, but perhaps obligatorily in others, by stress and
intonation); (ii) whether the pronoun 1s deictic or not, it can refer
anaphorically to the referent of a correlated antecedent expression which
does not conflict with the descriptive content of the pronoun and which
either precedes the pronoun in the same text or, under grammatically
restricted conditions, follows it in the main clause of a complex text-
sentence. These conditions will cover anaphoric pronouns with ante-
cedents in the same text-sentence, as well as anaphoric pronouns whose
antecedents occur in a preceding text-sentence.
It might be objected at this point that the process of pronominaliza-
13 Tt is worth noting that, in cases such as this, the pronoun-form need not be
heavily stressed. It is by no means true that pronouns may occur without
antecedents only when they are the focus of contrastive stress (cf. Chafe,
1970: 260).
 15.3. Detxis, anaphora and the universe-of-discourse 665
tion is required in order to handle the distribution and interpretation of
reflexive pronouns. But this is not so. First of all, it should be noted that
it is by no means as clear as most generative grammarians appear to have
assumed that a sentence like ‘The old man killed herself’ is ungram-
matical, though the conditions under which it might reasonably be
uttered are undoubtedly unusual. For example, anyone who believed
that after a certain age all men became female, without however ceasing
to be men, might well take The old man killed herself to be an acceptable
utterance. Let us grant, however, for the sake of the argument that ‘The
old man killed herself’ is ungrammatical ; and that, under the application
of the pre-theoretical principle of corrigibility (cf. 10.2), it should be
eliminated in favour of ‘The old man killed himself’. There are at least
two ways in which this can be done within a generative grammar. One
way is to generate it by a rule which reflexivizes the pronoun ‘he’ in the
underlying structure of “The old man killed him’, optionally, but on
condition that the sense of the object pronoun is compatible with the
sense of the subject expression. The other is to generate an underlying
structure with a reflexive element undifferentiated for gender and
number (let us label it SELF) and to make this compatible with the subject
expression in terms of sense.‘ By either of these techniques the grammar
would generate ‘The old man killed himself’, ‘They killed themselves’,
“My cousin killed himself’, ‘My cousin killed herself’, etc., and would
fail to generate “’The old man killed herself’, ‘They killed itself’, etc. In
14 T have used capitals to refer to the element SELF with some hesitation. I am
reluctant to postulate a lexeme, ‘self’, of which himself, myself, etc., would
be forms and tend to think of the reflexive element, here represented as
SELF, as something more abstract than a lexeme (comparable with CAUSE, |
etc.: cf. 12.5) ~ indeed, as something that may be thought of as underlying
not only the traditionally recognized reflexives, in English and other lan-
guages, but also the first-person pronoun, ‘I’, such that the proposition
expressed by ‘I am hungry’ would be ‘sELF be hungry’. But this kind of
analysis (though, to my mind, much more appealing semantically than, say,
‘‘’The speaker be hungry’’, or even ‘SPEAKER be hungry”: cf. 15.1) is, to say
the least, unorthodox. And it has its own problems: e.g., that of accounting
for the difference, at some deeper level, between ‘John hates me’ and ‘John
hates himself’, or between ‘John said that he had been there’ (under the
interpretation according to which the proposition expressed by John was ‘‘I
have been there’’) and ‘John said that I had been there’, if first-person
pronouns and reflexive pronouns (and certain third-person non-reflexives in
English) both derive from the same underlying element, SELF. It is interesting
to note that the occurrence of the reflexive pronoun zibun in subordinate
clauses in Japanese is governed by factors which relate it closely to what are
characteristically indirect discourse constructions with verbs of propositional
attitude (cf. Kuno, 1972; Inoue, 1976; N. McCawley, 1976).
 666 Deixis, space and time
neither case, however, is any rule of pronominalization required operat-
ing upon a non-pronominal antecedent.
The use of either of these techniques in the generation of well-formed
sentences does not exclude the use of the other; and there may well be
grounds for allowing both of them to operate, as alternatives, in the
generation of what would normally be thought of as the same sentence.
For example,
(7) John nominated himself
can be uttered to assert that John performed the action of self-nomina-
tion. But it may also be uttered to assert that John nominated someone
who, as it happens, was himself (though he may not have known or
intended this). At first sight, one might be inclined to say that there is no
ambiguity involved here and that the truth-conditions of the two
propositions are identical. However, we have only to consider the truth-
conditions of such sentences as
(8) John nominated himself and so did Harry
or
(9) Only John nominated himself,
which differs very strikingly in meaning from
(10) Only John nominated John,
to see that a case might be made for deriving (7) from both “‘John+
nominate+SELF”’ and “‘John,;+nominate+he,”’ (with reflexivization of
| ‘he’ under a condition of co-referentiality) and saying that it expresses
two different propositions. This would then account for the much
more obvious ambiguity of (8), and also the difference between (9) and
(10). However that may be, such writers as Geach (1962: 132ff) have
convincingly demonstrated the necessity of allowing for the formation of
reflexive predicates in the underlying structure of sentences; and this
implies that reflexivization is at most only one of the ways in which
reflexive pronouns (or the reflexive forms of verbs in certain languages)
can be derived in a grammar. Partee (1970) has discussed some of the
syntactic and semantic problems that arise in connexion with dis-

15 T am assuming that (8) is ambiguous: i.e. that the second clause means either
‘‘Harry (also) nominated John”’ or ‘‘ Harry (also) nominated himself’’. There
may be some disagreement as to whether it can sustain the former inter-
pretation.
 15.3. Detxts, anaphora and the universe-of-discourse 667
tinguishing between such sentences as (9) and (10); and Castafieda
(1968) has drawn attention to the fact that there is a close connexion
between the third-person reflexive pronoun, as it 1s employed in propo-
sitions that attribute self-knowledge to others, and the first-person

role of speaker. ,
pronoun which the speaker uses to refer to himself in his participant-

No more need be said here about reflexive pronouns. It is obvious that

the notion of co-reference, to the limited extent that is required in
generative grammar, does not necessarily presuppose a process of pro-
nominalization. Co-reference is relevant in sentence-grammar in so far ,
as it can be invoked in order to account for the optional reflexivization of
pronouns in certain syntactic positions and the fact that, under gram-
matically definable. conditions, pronouns (which may be non-reflexive
in form) and other anaphoric expressions must be, may be or cannot be
co-referential with particular expressions in the same sentence. For
example, it falls within the scope of grammar to say that ‘he’ in ‘The
boss wants him to go’ cannot be co-referential with ‘the boss’; that the
expression underlying the reflexive pronoun in ‘John killed himself’
must be co-referential with ‘John’; and that ‘he’ in ‘John thinks that
he is amusing’ may or may not be co-referential with ‘John’. Consider-
able attention has been devoted to the specification of the grammatical
conditions for co-referentiality in recent years. But we need not go
further with this question here.1®
It has been suggested, though not so far demonstrated, that deixis is
more basic than anaphora. The link between the deictic and the ana-
phoric function of pronouns is seen in what may be called textual
deixis*. Demonstrative pronouns and other deictic expressions may be
used to refer to linguistic entities of various kinds (forms, parts of forms,
lexemes, expressions, text-sentences, and so on) in the co-text of the utter-
ance; they may even be used, in a manner which can give rise to cer-
tain well-known logical paradoxes (e.g., \ This sentence, which I am now
uttering, is false“ : cf. 1.4), to refer to the whole utterance in which they
occur.!” Consider the following text: (X says) That’s a rhinoceros (and Y
responds) A what? Spell it for me. Here the referent of ‘it’ is clearly the
form rhinoceros. The function of ‘it’ is not anaphoric, although at first

16 Cf. Bach (1970), Dik (1973), Dougherty (1969), Fauconnier (1974), Jacken-
doff (1972), Kuno (1972a), Langacker (1969), Lees & Klima (1963), McCawley
(1969), Partee (1970, 1975a), Postal (1971), Ross (1969b).
17 The raised arrows are token-quotes indicating token-reflexivity in Reichen-
bach’s (1947) sense (cf. 1.4).
668 Deixis, space and time
sight it might appear to be. It is not co-referential with any antecedent
expression; it refers to, but is not co-referential with, a preceding lin-
guistic form. Textual deixis is frequently confused with anaphora, by
virtue of the traditional formulation of the notion of pronominal
reference (according to which, as we have seen, a pronoun is said to
refer to its antecedent) and the common failure to distinguish clearly
between linguistic and non-linguistic entities. There is no need to give
further examples of this kind of textual deixis: the text of the present
book is full of them. It should be noted however that the sense in which
forms and text-sentence occur in the co-text is different from the sense
in which lexemes or expressions occur in the co-text.
At one remove from what might be called pure textual deixis, though
not as clearly distinct from it as anaphora, is the relationship which
holds between a referring expression and a variety of third-order entities,
such as facts, propositions and utterance-acts (in the more abstract sense
of ‘utterance-act’ noted in 1.6). This may be exemplified by means of
the following text: (X says) I’ve never even seen him (and Y responds)
That’s a lie. It is clear that ‘that’ does not refer either to the text-sen-
tence uttered by X or to the referent of any expression in it. Some
philosophers might say that it refers to the proposition expressed by the
sentence uttered by X; others, that it refers to the utterance-act, or
speech-act (cf. 16.1), performed by X. However, under either of these
analyses of the reference of ‘that’, its function seems to fall somewhere
between anaphora and deixis and to partake of the characteristics of
both. Let us say that its function is that of impure textual deixis. It is not
always easy to draw the distinction between pure and impure textual
deixis in particular instances.
‘This’ and ‘that’, in English, may be used deictically to refer not only
to objects and persons in the situation and to linguistic entities of
various kinds in the text or co-text, but also to refer to events that have
already taken place, are taking place or are going to take place in the
future. The conditions which govern the selection of ‘this’ and ‘that’
with reference to events immediately preceding and immediately fol-
lowing the utterance, or the part of the utterance in which ‘this’ and
‘that’ occur, are quite complex. They include a number of subjective
factors (such as the speaker’s dissociation of himself from the event he is
referring to), which are intuitively relatable to the deictic notion of
proximity/non-proximity, but are difficult to specify precisely. What
does seem clear, however, is that the use of the demonstratives in both
temporal and textual deixis, and also in anaphora, is connected with
 15.3. Detxts, anaphora and the universe-of-discourse 669
their use in spatial deixis. ‘This is more obviously so in many languages
other than English. For example, in Latin the distal demonstrative ‘ille’
(“that’’) is used anaphorically to refer to the referent of the more
remote of two possible antecedents and the proximal demonstrative ‘hic’ |
(“‘this’’) to refer to the referent of the nearer of two possible antecedents;
and they can frequently be translated (into somewhat stilted English) as
‘the former’ and ‘the latter’, respectively. The same is true of the
German ‘jener’: ‘dieser’, the Spanish ‘ése’ (‘aquel’):‘éste’, the French
‘celui-la’:‘celui-ci’, the Turkish ‘o’:‘bu’, and so on. It is the notion of
relative proximity in the co-text to the moment of utterance that con-
nects anaphora and textual deixis with temporal reference; and it is the
more general principle of localization* (cf. 15.7) that relates temporal
reference, in many languages at least, to the more basic notion of spatial
deixis.
As we saw in the previous section distinctions of proximity are
lexicalized or grammaticalized in the pronoun-systems of many lan-
guages; and they are commonly combined with other distinctions, based
on status, sex, size, shape, etc. In so far as they are used deictically, it is
the function of pronouns to draw the attention of the addressee to
referents in the situation, identifying these referents for the addressee in
terms of their position relative to the zero-point of the deictic space, on
the one hand, and of their status, sex, size, shape, etc., on the other.
What now concerns us 1s the way in which the basically deictic distinc-
tion of proximity operates in anaphora. A simple example will serve.
In English, as we have seen, the third-person pronouns are neutral
with respect to proximity, but distinguished in terms of gender. Turkish,
by contrast, has three demonstratives distinguished in terms of deictic
proximity, but neutral with respect to gender. Latin also has three
demonstratives, distinguished in terms both of proximity and of gender
(and number). Both Turkish and Latin, as was pointed out above, make
use of their proximal and distal demonstratives for anaphoric reference.
The effect of these differences in the descriptive content of pronouns is
readily seen if we consider the way in which we might translate the
pronouns in a short text in English into Latin and Turkish respectively:
John and Mary came into the room: hé (i.e. ‘the male-one’’) was laughing,
but shé (“the female one’’) was crying. The Latin version might translate
‘he’ with ‘ille’ (‘‘that-male-one’’) and ‘she’ with ‘haec’ (“‘this-female-
one”’); in Turkish ‘he’ might be translated with ‘o’ (“‘that-one’’) and
‘she’ with ‘bu’ (‘‘this-one’’), Suppose now that we reverse the order of
the conjoined nominals ‘John’ and ‘Mary’ in the English text. This has
670 Detxts, space and time
no effect upon the choice of pronouns in English: Mary and Fohn came
into the room: shé was crying, but hé was laughing. In both Latin and
Turkish, however, if we reverse the order of the antecedents, by virtue
of the lexicalization of deictic proximity in the anaphoric demonstra-
tives, we will now translate ‘he’ as ‘hic’ (“‘this-male-one’’) and ‘bu’
(‘‘this-one’’), and ‘she’ as ‘illa’ (“that-female-one’’) and ‘o’ (“‘that-
one’’).
The example that has just been given is very simple. But it does
illustrate clearly the way in which deictic distinctions can be used to
identify the antecedents of anaphoric expressions. Anaphora involves
the transference of what are basically spatial notions to the temporal
dimension of the context-of-utterance and the reinterpretation of deictic
location in terms of what may be called location in the universe-of-
discourse*. ‘The notion of previous mention, which is commonly invoked
in discussions of anaphora, depends upon the temporal relation which
holds (in a spoken text) between the anaphoric expression and its ante-
cedent. The basically deictic component in an anaphoric expression
directs the attention of the addressee to a certain part of the text or
co-text and tells him, as it were, that he will find the referent there. It is
not of course the referent itself that is in the text or co-text. The
referent is in the universe-of-discourse, which is created by the text and
has a temporal structure imposed upon it by the text; and this temporal
structure is subject to continuous modification. To say that the referent
has a textual location implies, then, that it will be found in a certain part
of the universe-of-discourse as this is structured, temporally, by the
text; and subsequent reference to this referent by means of an anaphoric
expression will identify the referent in terms of the textual location of
the antecedent. Let us suppose, for the purpose of illustration, that the
English demonstratives ‘this’ and ‘that’, in anaphoric expressions, do
no more than simply encode the distinction of temporal proximity in
relation to the moment of utterance. ‘This animal’, used as an anaphoric
expression, will direct the attention of the addressee to the most proxi-
mate referent in the universe-of-discourse satisfying the sense of
‘animal’; ‘that animal’ will refer to a textually more remote referent;
and ‘the animal’ will refer to some animal which has a textual location,
in the sense explained, but will give the addressee no information about
its textual location. No such information will be required of course, if
it is the only animal that has been previously mentioned; and no in-
formation will be required, if there is a generally accepted convention
that, in default of any specific information as to the textual location of
 15.3. Detxis, anaphora and the universe-of-discourse 671
the referent, it is taken to be the most recently mentioned entity that is
being referred to.
Things are not quite as simple as this illustration might suggest. The
anaphoric use of ‘this’ and ‘that’ in English involves other considera-
tions besides the relative proximity of the antecedent to the moment of
utterance; and ‘this’ ws. ‘that’ cannot be used, as the Latin ‘hic’ vs.
‘ille’ and the Turkish ‘bu’ ws. ‘o’ can, to mean ‘‘the latter”’ vs. ‘“‘the
former’’. One cannot say John and Mary came into the room: that person
was laughing, but this person was crying, to mean “John and Mary came
into the room: John was laughing, but Mary was crying’’. It is not
being maintained, however, that the anaphoric use of demonstratives, in
English or in any other language, is totally predictable from their deictic
use. The important point is that, independently of whether particular
languages make anaphoric use of demonstratives or not, what logicians
commonly refer to as the universe-of-discourse (or point-of-reference* :
cf. 6.6) is not simply an unstructured set of potential referents, each of
which is equally accessible throughout a text or conversation. Some of
the potential referents are more salient than others; and saliency is in
part determined by recency of mention. In so far as recency of mention
is itself a deictically based notion and is encoded, in one way or another,
in the anaphoric pronouns used 1n particular languages, anaphora rests
ultimately upon deixis.
However, it requires but little reflexion to see that the potential
referents in the universe-of-discourse cannot be indexed solely, or even
primarily, in terms of recency and relative order of previous mention.
The limitations of human memory are such that, without having imme-
diate access to a transcript of all that has been said previously (or,
alternatively, to some continually updated computer-file), we could not
operate with a system of anaphoric reference which employed expres-
sions meaning, for example, “the twelfth most recently mentioned
entity” or ‘‘the twelfth entity mentioned in the present text’’. The tem-
poral structure imposed upon the universe-of-discourse by the succession
of referring expressions in texts 1s, therefore, of very limited duration;
and the anaphoric use of the basically deictic distinction of proximity
to the zero-point of the context-of-utterance is determined by this fact.
Furthermore, salience in the universe-of-discourse is not simply a
matter of recency of previous mention. Indeed, there need not have been
any previous mention. As Isard (1975) points out, if a child reaches
towards the lion’s cage in order to pat what he takes to be a friendly big
cat, the zoo-keeper can say
 672 Deixis, space and time
(11) Be careful, he might bite you,
without there having been any previous reference to the lion. In this
case, the lion is present in the context-of-utterance; and, although the
form he would probably be unstressed, the reference of ‘he’ might well
be described as deictic by virtue of the almost inevitable paralinguistic
accompaniment of eye-gaze and gesture. Other examples can be pro-
duced, however, which show that a potential referent is salient in the
universe-of-discourse, even though it is not present in the situation-of-
utterance and has not been mentioned previously by either the speaker
or the addressee. For example, I might offer my condolences to a friend,
whose wife has just been killed in a car-crash, by saying
(12) I was terribly upset to hear the news: I only saw her last week.
Naturally enough in these circumstances, there is no need for me to
specify what news I am referring to or who the referent of ‘she’ is.
Examples like (12) show us very clearly that entities need not have been
mentioned previously in order for them to be salient in the universe-of-
discourse. If the notion of anaphora is so defined that it presupposes the
occurrence of a correlated antecedent expression in the text or co-text,
then ‘she’ is obviously not anaphoric in (12). And yet its function in
(12) appears to be no different from the function it has in
(13) I know Mrs Smith very well: I only saw her last week.
In both (12) and (13) ‘she’ refers to the currently most salient person;
and, since we know or believe that the person we are referring to is a
woman, we are obliged by the grammatical and lexical structure of
English to use ‘she’, rather than ‘he’ or ‘it’.
Many scholars, including Buhler (1934), would say that the reference
| of ‘she’ in (12) is deictic, rather than anaphoric, on the grounds that it
involves pointing to something in the intersubjective experience or
common memory of speaker and addressee, rather than to something in
the external situational context (cf. Crymes, 1968: 62-3). It is obvious,
however, that the notion of intersubjective experience, or common
memory, is the more general notion, without which anaphoric reference,
as it is traditionally conceived, cannot be explained. Such writers as
Kristeva (1969) and Barthes (1970) have insisted that what is commonly
referred to as intersubjectivity should be more properly described as
intertextuality*, in that the shared knowledge that is applied to the
interpretation of text is itself the product of other texts (cf. Ducrot &
 15.3. Deixis, anaphora and the universe-of-discourse 673
Todorov, 1972: 446; Culler, 1975: 139). Up to a point this is true; and
especially in so far as literary texts are concerned. But not all of the
intersubjective knowledge that is exploited in the interpretation of texts
derives from what has been previously mentioned; and, in the last
resort, there would seem to be no reason to deny that the reference of
‘she’ in (12) is anaphoric.
Both deixis and anaphora are far more complex than the somewhat
schematic account of them given here might suggest. What has been said
will be sufficient, it is hoped, to justify the assertion that deixis is more
basic than anaphora. Anaphora presupposes that the referent should
already have its place in the universe-of-discourse. Deixis does not;
indeed deixis is one of the principal means open to us of putting entities
into the universe-of-discourse so that we can refer to them subsequently
(cf. Isard, 1975).?8 ,
There is much in the more recent work on pronouns that we have
deliberately left on one side in our treatment of deixis and anaphora and
will do no more than mention here. The standard approach to the
analysis of pronominal reference by logicians is to treat pronouns as the
natural-language correlates of the variables that might be used instead
of constants in the well-formed formulae of the predicate calculus or of
some other logical calculus (cf. 6.3). It has been pointed out by Partee
(1975a) that although the pronouns-as-variables* analysis works well
with certain sentences, there are others for which it is far from appro-
priate. For example,
(14) No-one drives when he is drunk
expresses a proposition which (if we neglect the difference between
‘no-one’ and ‘nothing’, on the one hand, and between ‘he’/‘she’ and
‘it’, on the other) may be represented, loosely, as :
(14’) “No x drives when x is drunk”’.
This illustrates the pronouns-as-variables analysis; and in a more formal
representation of the structure of (14’) the variable x would be shown as
being bound* by the universal quantifier in both positions of occurrence
(cf. 6.3).

example: ,
To be contrasted with (14) are such sentences as Karttunen’s (1969)

18 Strictly speaking, it is not the entities themselves that are put into the
universe-of-discourse, but their intensional correlates (cf. Lyons, 1977).
674 _ Detxis, space and time
(15) The man who gave his paycheck to his wife was wiser than the
man who gave it to his mistress.
Under the interpretation of (15) that concerns us here, ‘it’ is not co-
referential with ‘his paycheck’; and there will have been no previous
mention of the entity to which ‘it’ refers. For the analysis of (15),
Partee (1975a) proposes a treatment which, following Geach (1962), she
calls a pronouns-of-laziness* treatment. It is characteristic of pronouns-
of-laziness that they can be substituted for expressions that are identical,
but not necessarily co-referential, with antecedent expressions. In this
respect, therefore, the pronouns-of-laziness analysis is the one that is
formalized by the earliest version of pronominalization in Chomskyan
transformational grammar: ‘it’ would be substituted for ‘his paycheck’,
by means of an optional transformational rule, at some stage in the
syntactic derivation of (15); and the question of co-reference would
simply not arise. Failure to apply this rule of pronominalization would
result in the derivation of

(16) The man who gave his paycheck to his wife was wiser than the
man who gave his paycheck to his mistress,
which, under the interpretation that concerns us here, is equivalent to
(15). The use of a pronoun-of-laziness may be seen as a purely stylistic
or rhetorical device, which enables the speaker or writer to avoid repeti-
tion of the antecedent.
We will not go further into the pronouns-as-variables analysis or the
pronouns-of-laziness analysis. All that needs to be said here is that
neither of them appears to be capable of handling everything that the
other can handle, although there are many sentences whose meaning can
be accounted for equally well by either. This point has been well argued
by Partee (1970, 1975a). But it must also be emphasized that neither the
pronouns-as-variables analysis nor the pronouns-of-laziness analysis is
particularly successful in handling either the deictic reference of pro-
nouns or their anaphoric reference to entities that have not been pre-
viously mentioned; and many of the sentences whose meaning can be
accounted for by the pronouns-as-variables analysis or the pronouns-of-
laziness analysis can also be handled in terms of an analysis which takes
deixis to be basic. Granted that no uniform treatment of the relation
between pronouns and their antecedents is possible, it is arguable that
the use of pronouns as deictics is more basic than their use as variables
or their use as pronouns-of-laziness.
 15.3. Deixis, anaphora and the universe-of-discourse 675
It may be added that the pronouns-as-deictics analysis can be
naturally extended to handle such sentences as
(17) John wants to catch a fish and to eat it for supper,
which have also been much discussed recently. The problem here is that
‘it’ is used in the second clause as an expression with singular definite
reference, whereas (under the relevant interpretation) ‘a fish’ in the
first clause does not refer to any specific individual and carries no
presupposition of existence. And yet the notion of co-referentiality
appears to be relevant to the interpretation of (17) in a way in which it
was not relevant to the interpretation of either (14) or (15). What seems
to be involved here is the treatment of some hypothetical entity as if it
were an actual entity. As we have already seen, there can be anaphoric
reference to entities that have not been previously mentioned, provided
that they are in the universe-of-discourse; and hypothetical entities are
treated exactly like actual entities in this respect. Even though no
previous reference has been made to some particular fish, the hypotheti-
cal entity whose existence and whose being caught by John are pre-
conditions of the actualization of the possible world-state described by
the proposition ‘‘John wants to eat the fish that he catches’’ enters into
the universe-of-discourse as a potential referent no less readily than does
some actual entity that has been introduced by means of detxis or the
use of a non-deictic referring expression.
Throughout this section we have concentrated upon pro-nominals. It
is arguable, however, that anaphora involving other kinds of forms and
expressions (with the exception of pro-locatives, to which we will come
presently) cannot be accounted for in the way that we have accounted
for the anaphoric reference of pro-nominals. For there are many other
sets of forms and expressions (not all of them classifiable as pronouns)
whose function is best accounted for in terms of the notion of gram-
matical substitution: e.g., the pro-noun one in such sentences as ‘I want
the red scarf not the blue one’; the pro-verbal ‘do’ in such sentences as
‘I will go to the party if you do’; etc.
Mention should also be made at this point of the so-called classifiers*,
whose pronominal function was referred to in an earlier chapter (cf.
11.4). Many theoretical discussions of the nature and use of pronouns
fail to draw attention to the fact that in many languages classifiers are
used anaphorically as pro-nouns in much the same way as the form one
is employed in English.
That the use of pro-nouns, pro-verbals, etc., cannot be explained in
 676 Deixis, space and time
terms of co-reference, or in terms of their reference to entities in the
universe-of-discourse, is evident from the fact that they are not referring
expressions at all. Their function is closer to, if not identical with, that
of the so-called pronouns-of-laziness mentioned above. There is, how-
ever, one other set of expressions whose function is very similar to what
we have described as the basic function of pro-nominals: such adverbs
of place and time as ‘here’ vs. ‘there’ and ‘now’ vs. ‘then’. Because of
their anaphoric function, they have frequently been classified as pro-
nominal adverbs: they are more suitably described as pro-locatives* and
pro-temporals*.
As we saw in the previous section, there is a very close connexion
between the deictic function of the demonstratives ‘this’ and ‘that’,
and the locative adverbs ‘here’ and ‘there’. The same connexion is to
be noted as far as the pro-locative, anaphoric function of ‘here’ and
‘there’ is concerned: cf.
(18) I was born in London and I have lived here/there all my life
and
(19) I was born in London and this/that is where I have lived all my
life.

In both (18) and (19) the use of the pro-locative ‘here’ vs. ‘there’ and
the pro-nominal ‘this’ vs. ‘that’ is simultaneously deictic and anaphoric,
since the selection of one expression rather than the other is determined,
under normal conditions of utterance, by whether the speaker is in
London or not at the time. Similarly, in both
(20) You mustn’t come at six: that’s when John is coming
and

(21) You mustn’t come at six: John is coming then,

the use of the pro-nominal ‘that’ and the pro-temporal ‘then’ is simul-
taneously deictic and anaphoric. For obvious reasons, however, ‘now’
is only rarely employed with anaphoric reference.
In view of the emphasis that was placed upon the syntactic ambi-
valence of place-referring expressions in an earlier chapter, it should be
noted, not only that (18) and (19) are more or less equivalent, as are (20)
and (21), but also that ‘here’, ‘there’, ‘now’ and ‘then’ can also func-
tion syntactically as nominals in English: cf. ‘in here’ and ‘in this
place’, ‘by now’ and ‘by this time’, etc. There are many such indica-
 15.4. Tense and detctic temporal reference 677
tions of the peculiarly ambivalent status of the pro-locatives and pro-
temporals.
In our discussion of the relation between deixis and anaphora in this
section we have restricted our attention to what may be thought of as
the normal interpretation of the deictic distinction of proximity and
remoteness. We have seen that, in certain languages at least, this basic-
ally spatial distinction may be transferred to the temporal dimension and
used, in anaphora, to identify referents in terms of their place in the
universe-of-discourse. In conclusion, we would draw attention to what
we will call empathetic* deixis and its role in anaphoric reference. It
frequently happens that ‘this’ is selected rather than ‘that’, ‘here’
rather than ‘there’, and ‘now’ rather than ‘then’, when the speaker is
personally involved with the entity, situation or place to which he 1s
referring or is identifying himself with the attitude or viewpoint of the
addressee. The conditions which determine this empathetic use of the
marked member of these deictically opposed demonstratives and adverbs
are difficult to specify with any degree of precision. But there is no
doubt that the speaker’s subjective involvement and his appeal to shared
experience are relevant factors in the selection of those demonstratives
and adverbs which, in their normal deictic use, indicate proximity. At
this point deixis merges with modality (cf. 17.2).

15.4. Tense and detictic temporal reference

Traditional discussions of the grammatical category of tense* do not
give sufficient emphasis to the fact that it 1s a deictic category; and they
tend to be misleading in other respects also. The semantic analysis of
tense is something that we will come back to in a later chapter (cf. 17.3).
In this section we will concentrate upon its connexion with deixis.
It is often implied, if not actually asserted, that the distinction of past,
present and future is essential to the notion of tense and that the future
is like the past, except that it follows, rather than precedes, the present
in the infinitely extensible unidimensional continuum of time. But the
future is not like the past from the point of view of our experience and
conceptualization of time. Futurity is never a purely temporal concept;
it necessarily includes an element of prediction or some related modal*
notion (cf. 17.2). This does not mean of course that languages could not,
in principle, treat predictions as being grammatically parallel with state-
ments about the past or present. But in general they do not; and the
so-called future tense of the Indo-European languages (which 1s of
comparatively recent development in many of them) and the so-called
678 Deixis, space and time
future tense of the relatively small number of other languages through-
out the world that have anything that might reasonably be called a future
tense is partly temporal and partly modal. Nor is it the case that tense
must be based upon a distinction of past and present; it could be based
instead upon a distinction of present and non-present, or upon various
degrees of proximity to the time of utterance. What is commonly
referred to as the present tense, in English and many other languages, is
in fact more satisfactorily described as the non-past tense. Normally, the
use of the past tense in simple sentences does indeed locate the situation
about which a statement is being made in the past with respect to the
time of utterance (e.g., He worked hard); but the use of the so-called
present tense does not generally imply contemporaneity with the act of
utterance (cf. He works hard). It is only in contexts of immediate report
or commentary that the English simple non-past tense, without an
accompanying adverb of time, is used to locate a situation in the present.
The most basic distinction in the English tense-system, as it is in the
vast majority of the tense-systems of other languages, is the distinction
between past and non-past.
Traditional doctrine is also misleading in that it tends to promote the
view that tense is necessarily an inflexional category of the verb. It is an
empirical fact (which may be accounted for in terms of the centrality, or
pivotal status, of the verb: cf. 12.4) that tense, like person, is commonly,
though not universally, realized in the morphological variations of the
verb in languages. Semantically, however, tense is a category of the
sentence (and of such clauses within a sentence as may be regarded as
desentential in the full sense: cf. 10.3).
It has already been pointed out (14.1) that the participants in a
language-event must be able to control and interrelate at least two
different frames of temporal reference: the deictic and the non-deictic.
Tense, in those languages which have tense, is part of the deictic frame
of temporal reference: it grammaticalizes the relationship which holds
between the time of the situation that is being described and the tem-
poral zero-point of the deictic context.
But tense, as distinct from deictic temporal reference, is not a univer-
sal feature of language. In so far as there is no sharp pre-theoretical
distinction to be drawn between grammaticalization and lexicalization,
whether a language has tense or not is a question that can be decided
only on the basis of a grammatical analysis of particular languages. When
it is said that certain languages (e.g., Chinese or Malay) do not have
tense, what is usually meant is that these languages do not obligatorily
 15.4. Tense and detctic temporal reference 679
relate the time of the situation being described to the time of utterance
by any systematic variation in the structure of the sentence. Particular
utterances of Chinese or Malay might be translated into English, pro-
vided that the necessary information is given in the context, as It is
raining or It was raining.
Though not all languages have tense, it is probably true to say that all
languages have various deictic adverbs or particles of time, comparable
with the English words ‘now’, ‘then’, ‘recently’, ‘soon’, ‘to-day’,
‘yesterday’, etc., which provide the means, when it is necessary or
desirable, for drawing deictic temporal distinctions of the kind that are
obligatory, and grammaticalized in the fullest sense as tense-distinctions,
in such languages as English. Whether all languages have words which
enable them to draw non-deictic distinctions of time is less certain. It is
noteworthy, however, that children learning English normally come to
acquire control of the non-deictic system of temporal reference (in
terms of calendar-time and clock-time) after they have mastered the use
of tense and the more common deictic adverbs. This would seem to
indicate that the deictic frame of temporal reference is basic and essen-
tial to language in a way that the non-deictic frame of temporal reference
is not.
It is important to distinguish between the tenselessness of sentences
and the tenselessness of propositions; the more so, since the terms
‘tensed’ and ‘tenseless’ (not to mention ‘sentence’ and ‘ proposition’)
are sometimes employed in a way that tends to confuse the distinction
between deictic and non-deictic temporal reference. A tenseless sen-
tence is, quite simply, a sentence without tense. As we have seen, whether
a language has tense or not is determined, in part, by the point at which
we draw the line, for that language, between grammaticalization and
lexicalization. If a language is without tense, then all its sentences willbe __
tenseless* (though some of them would contain deictic adverbs or
particles with temporal reference). If a language has tense, then in
principle some of its sentences may be tenseless and others will be
tensed*. In many languages, including English, there are no tenseless
declarative or interrogative system-sentences; and the tense of declara-
tive and interrogative sentence-fragments is recoverable from the context
(cf. 14.1). Whether imperative sentences, in English and other languages,
are tensed or tenseless is a question that we need not discuss here (cf.
16.2).
Tenselessness is sometimes confused with timelessness, especially in
philosophical discussions of the temporal or non-temporal status of
 680 Deixis, space and time
propositions. Furthermore, timelessness is not always distinguished
from omnitemporality (cf. Strawson, 1952: 151). A timeless* proposi-
tion is one for which the question of time-reference (whether deictic or
non-deictic) simply does not arise: the situation, or state-of-affairs, that
_ it describes is outside time altogether. Obvious examples of timeless
propositions are the so-called eternal truths of mathematics and theology.
An omnitemporal* proposition, on the other hand, is one that says that
something has been, is and always will be so: it is a proposition whose
truth-value is constant for all values of t;, in a finite or infinite set of
time-points or time-intervals, {t,, f, fs, ..., tn}. Obviously, there are
philosophical problems attaching to this distinction between timeless
and omnitemporal propositions; but we need not be concerned with
them. It suffices for our purposes that the distinction is seriously
defended by many philosophers who have been concerned with the
question of time; and it is a distinction to which we will appeal presently.
For the moment, it may simply be noted that the everyday use of the
English adverb ‘always’ is not a sure guide to the distinction between
timelessness and omnitemporality.
Any proposition that is not timeless will be called time-bound*.
Omnitemporal propositions are one subclass of time-bound proposi-
tions: they are time-bound, but temporally-unrestricted*. In English
and many other languages, both timeless and omnitemporal proposi-
tions are expressed characteristically by sentences in the so-called
present tense: cf. ‘God is just’, “The sun rises every day’. It is import-
ant to realize, however, that there is no intrinsic connexion between the
grammatical category of tense, and still less between any particular
tense, and the expression of either timeless or omnitemporal proposi-
tions. There are many languages in which tenses other than the so-called
present tense are used for this purpose. |
Indeed, there is considerable variation among languages with respect
to the way in which they grammaticalize the various temporal distinc-
tions with which we are concerned here. It is probably the case that the
vast majority of languages do not grammaticalize the distinction between
timelessness and omnitemporality. But there are certain languages,
apparently, that do; and we will come to this point later (cf. 15.6). It is
of considerable importance for the analysis of so-called generic* propo-
sitions, such as ‘‘Cows are herbivorous’’. Generic propositions, it
might be argued, are not merely omnitemporal, but timeless.
As the distinction between timelessness and omnitemporality is
difficult to draw in particular instances, so too 1s the distinction between
 15.4. Tense and deictic temporal reference 681
omnitemporality and a variety of time-bound notions that approximate
to, or resemble, omnitemporality. Linguists frequently employ the term
gnomic* to refer to such so-called general truths as “‘It never rains but
it pours’’, “Corruption starts at the top’’. Many of these truths (if they
are truths) are expressed in the proverbs and aphorisms that are passed
on, in all cultures, from generation to generation. The temporal status
of the propositions embodied in gnomic utterances is extremely diverse:
some are timeless and others are omnitemporal; but many of them could
hardly be said to describe anything more than tendencies, generalities
and assumed regularities. Some languages are said to have special
gnomic tenses (in a rather broad sense of the term ‘tense’). More
commonly, however, gnomic utterances will employ, as they usually do
in modern English, a tense, mood or aspect that is employed, charac-
teristically, with a rather different function.
It is in the nature of things that the term ‘gnomic’ cannot be given a
very precise definition. But it is a useful term, the more so as it is often
much easier to decide that an utterance is gnomic than it is to decide
whether it expresses a timeless or omnitemporal proposition. We will
therefore use the term ‘gnomic’ to describe both utterances of a certain
kind and the propositions that are expressed in the production of such
utterances. There is some overlap, it will be noted, between ‘generic’
and ‘gnomic’; but the two terms are by no means co-extensive.
Gnomic propositions can be looked at from several points of view;
and this accounts for the fairly wide variation that is found among lan-
guages with respect to the way in which they handle gnomic propositions.
English uses the present, or non-past, tense. This is explicable in terms
of the fact that the non-past tense is semantically unmarked in English
(as it is in many languages): but we should not be misled by the implica-
tions of the traditional term ‘present tense’ into thinking that the
so-called general truths embodied in gnomic utterances have anything
to do with present time. Their temporal status, if they are time-bound,
is non-deictic. Moreover, we can base our assertion of a general truth
upon the evidence of our past experience: hence the use of the past,
rather than the non-past, tense for the expression of gnomic proposi-
tions in certain languages. Alternatively, our belief in the validity of
some general truth may be based upon our knowledge of what is usually
the case; this makes the habitual or iterative aspect appropriate in
languages that grammaticalize aspectual distinctions of this kind.
Finally, gnomic utterances are such that they tend to be concerned with
matters of opinion, rather than fact; and this brings them within the
682 Deixis, space and time
scope of epistemic modality*, which, in some languages, is gram-
maticalized as mood (cf. 17.2).
A fairly clear distinction can be drawn in the metalanguage of general
linguistic theory between the terms ‘tense’, ‘mood’ and ‘aspect’. Not
only do particular languages differ, however, with respect to the seman-
tic distinctions that are grammaticalized in terms of the categories of
tense, mood and aspect. What is classified as a tense, a mood or an
aspect in any particular language may have a range of functions, some
or all of which may fall outside the scope of the general definition of the
grammatical category in question. For example, the past tense in
English has certain functions which are not covered by any definition
that one might give to the term ‘past tense’ in general linguistic theory;
and English is not untypical in this respect. Indeed, it is no exaggeration
to say that there is probably no tense, mood or aspect in any language
whose sole semantic function is the one that is implied by the name that
is conventionally given to it in grammars of the language. Furthermore,
it is undoubtedly the case that the terms conventionally used to describe
the functions of the tenses, moods and aspects in certain languages are
very misleading. This point must be borne constantly in mind.
As ‘tenseless’ must not be confused with ‘timeless’, so ‘tensed’ must
not be confused with either ‘time-bound’ or ‘temporally restricted’.
The crucial fact about tense, whether we are talking about sentences or
propositions, is that it is a deictic category. A tensed proposition, there-
fore, will be, not merely time-bound, or even temporally restricted: it
will contain a reference to some point or period of time which cannot be
identified except in terms of the zero-point of utterance.
It is a commonplace of the philosophy of time that there are two quite
different ways of conceiving and talking about time. As Gale puts it
(1968a: 7) there is “the dynamic or tensed way’’, according to which
‘‘events are represented as being past, present and future, and as con-
tinually changing in respect to these tensed determinations”’; and there
is ‘“‘the static or tenseless way’’, according to which “the very same
events which are continually changing in respect to their pastness,
presentness or futurity are laid out in a permanent order”. The conflict
between these two conceptions of time accounts for some apparent
inconsistencies in the use of temporal expressions; we will not go into
these (cf. Traugott, 1975). Here we are concerned to emphasize that
the dynamic conception of time is deictic and the static conception is
not.
But the distinction of past, present and future, as we pointed out
 15.4. Tense and deictic temporal reference 683
above, is not essential to the definition of tense. Provided that we can
identify the temporal zero-point of the canonical situation of utterance,
we can define a variety of potential tense-distinctions in terms of
simultaneity vs. non-simultaneity, proximity vs. non-proximity, earlier
than vs. later than, etc. Given that to is the zero-point (referred to by
the adverb ‘now’ in English): then, (i) if t; = ¢,, ¢, will refer to the
same time as fo and will therefore define the notion of present tense;
(i1) if t; 34 to, te will refer to some point or period of time that is not
simultaneous with fo, and it will define the non-present (which may be
referred to by the adverb ‘then’ in English); (iii) if t¢ < to (“ts is earlier
than ¢,’’), t; will refer to some point or period in the past; (iv) if t > to
(‘‘t, 1s later than t)’’), ¢; will refer to some point or period in the future.
By drawing a distinction between points of time and periods of time
and allowing that a point may be included in a period (e.g., ti > to), we
can establish such further potential tense distinctions as the distinction
between the punctual present and the extended present (cf. Bull, 1963).
In all cases, however, it is the deictic zero-point, to, that makes the dis-
tinction that we establish a distinction of tense. We shall make consider-
able use of such temporal indices as fy and the variables % and t; in a
later chapter (cf. 17.3).
It will now be clear that not all temporally restricted propositions are
to be regarded as tensed, even though the most natural way of expressing
them in certain languages may be by uttering a tensed sentence. For
example, the tenseless proposition ‘John Smith be ill on 13 April
1971’ would normally be expressed in English by ‘John Smith was ill
on 13 April 1971’, if the person wishing to assert this proposition
believed that ‘13 April 1971’ referred to some day in the past. Proposi-
tions which include an expression referring to some determinate point
or period of time are readily formalizable in the higher predicate calculus
in various ways; and their meaning is context-independent.
As we saw in an earlier section, there are logical problems involved in
the conversion of a context-dependent proposition containing a first-
person pronoun into a context-independent proposition (cf. 15.1). At
first sight, no such problems might seem to arise in the conversion of a
tensed proposition or a proposition containing deictic adverbs of time
into a context-independent proposition which explicitly mentions the
time of the situation that the proposition describes: e.g., in the conver-
sion of “It 1s raining’ into “It be raining (in Edinburgh) at noon on
6 January 1971’’; and there are good reasons, at times, why we should
wish to carry out this kind of conversion (cf. 14.1). It is obvious of
 684 Detxis, space and time
course that, before we can carry out the conversion we must know when
and where the tensed proposition “It is raining’’ was asserted.
What is less obvious perhaps is that, although they are generally
regarded as non-deictic (in contrast with such adverbs as ‘now’ or
‘yesterday’), even expressions like ‘at noon on 6 January 1971’, as they
are used in everyday discourse, are not entirely free of deictic implica-
tions. We cannot tell what point of time the speaker is referring to when
he says It was raining (in Edinburgh) at noon on 6 fanuary 1971 unless
we know which temporal frame-of-reference he wants us to adopt.
‘6 January 1971’ might refer to a period of time, and ‘noon’ to a point
within that period, whose limits are determined by the location of
Edinburgh. But it may also refer to a period of time whose limits are
determined by the speaker’s location in one time-zone, rather than
another, at the time of the utterance. It may even refer to a period of
time determined by where the speaker happened to be at what was
(where he happened to be at the time) noon on 6 January 1971. It is
only if we can standardize all temporal references in relation to some
arbitrary, but decidable, public frame-of-reference (e.g., Greenwich
Mean Time) that we can completely eliminate temporal deixis from the
propositions expressed by sentences containing expressions that refer to
some definite point or period of time. It is all too often forgotten that
public calendar-time and clock-time, in the use that we make of it in
everyday life, is not a chronologically stable frame-of-reference.
There is one important difference between temporal and spatial
deixis; and it is crucial to the conversion of context-dependent proposi-
tions into context-independent propositions, in so far as this process of
conversion can be carried out at all satisfactorily. This has to do with
the separability of a transient, but intersubjectively common, temporal
zero-point from the here-and-now of utterance. Let us suppose, for
example, that X and Y are having a long-distance telephone conversa-
tion, X being in London and Y in Los Angeles. Now, if X says
(1) It 1s raining here now,
‘here’ will be understood as referring to the place and ‘now’ to the
time of utterance. To this extent ‘here’ and ‘now’ are similar. The
difference is that, whereas X. and Y, knowing that they are in different
places, will adjust their use of ‘here’ and ‘there’ accordingly, they will
assume that ‘now’ refers to the same point of time in London as it does
in Los Angeles. It is for this reason that it makes sense for X to say to Y,
What time is 1t now in Los Angeles?
 15.4. Lense and deictic temporal reference 685
One of the assumptions with which we operate in our everyday lives
is that, whereas the same first-order entity cannot be in two different
places at the same time, the same first-order entity can be in the same
place at different times and in different places at different times; and
furthermore that, whereas two first-order entities cannot be in precisely
the same place at precisely the same time, they can be in different places
at precisely the same time. These assumptions are supported, if not
created, by what we will refer to as the principle of deictic simul-
taneity*: the principle that in the canonical situation-of-utterance the
temporal zero-point, fo, is identical for both speaker and addressee.
Furthermore, notwithstanding our acceptance of the Einsteinian theory
of relativity, we normally operate with the more general principle of
absolute simultaneity, according to which, regardless of where they are,
X and Y can refer to the same point of time by means of such deictic
temporal expressions as ‘now’ or ‘three seconds ago’. For example, we
will say that X and Y have asserted the same proposition, if each of them
produces, independently but at what we can identify as the same
absolute point in time, a token of the utterance-type
(2) It ts raining now in Edinburgh,
regardless of whether X and Y are in Edinburgh and regardless of
whether they happen to be in the same place or not at the time that each
produces his utterance-token.
It is worth noting, however, that utterances do not need to contain a
deictic temporal expression in order for them to be understood, in
context, as asserting (or presupposing) a temporally restricted proposi-
tion. This should be borne in mind in view of the centrality in truth-
conditional semantics of Tarski’s formalization of the notion of truth.
The sentence
(3) ‘Snow is white’ is true if and only if snow is white
is tacitly construed as if it expresses a tenseless, if not timeless, proposi-
tion and is being used metalinguistically to say something about a
tenseless sentence of the object-language (cf. 6.7). |
It is obvious enough that the Tarskian criterion for truth cannot be
applied without modification to such sentences as
(4) ‘It is raining’ is true if and only if it is raining;
and model-theoretic extensions of Tarski’s notion of truth (which :
relativize it to a so-called point-of-reference) introduce the necessary
modifications. Here we are concerned to emphasize the general point
686 Deixis, space and time
that, unless there are indications in the context which would lead us to
interpret a particular utterance-token It 1s raining as having reference to
somewhere other than where the speaker is at the time of the utterance
and to some time other than the time of the utterance, It 1s raining
would normally be taken to mean “It is raining here now”’.
It must also be emphasized that this implicit reference to the here-
and-now of the deictic context holds independently of whether the
language in question has tense or not. If we were to construct a tenseless
version of Quasi-English and to use it as other tenseless natural lan-
guages are used, we would presumably find that such sentences as
(5) It be raining,
which carry no explicit indication of their reference to past, present or
future time (or to any other deictically determined point or period of
time), would nonetheless be interpreted, on particular occasions of their
utterance, as having reference, preferentially, to the here-and-now: i.e.
as meaning “‘It is raining here now’’.
Whether (5) expresses a tensed proposition (or rather, indefinitely
many distinct tensed propositions) is a philosophically controversial
question that we need not attempt to resolve. So too 1s the more general
question whether all, or any, tensed propositions can be converted into
tenseless propositions without doing violence to their meaning. There
is no doubt that tense is something that the linguist must take into
account in his analysis of at least some languages, regardless of whether
tense is in principle eliminable from logical representations of the struc-
ture of propositions; and he must be careful, as must the logician and
the philosopher, not to confuse the tenselessness of sentences with the
tenselessness of propositions. As we have seen, tenseless sentences can
be used to express tensed propositions; and tensed sentences may, and
in certain languages must, be used to express tenseless, or even timeless,
propositions. It is only very recently that attempts have been made to
formalize some of the tense-distinctions that we have invoked in this
section by logicians who are convinced of the necessity of incorporating
the logic of tense within the framework of modal logic (cf. 17.2).
One final point should be made about tenselessness and timelessness.
This has to do with the potential ambiguity of such expressions as
‘timeless truth’ and ‘timeless proposition’, It 1s a defensible, if not
unchallengeable, view that, if a proposition is true, it is eternally (i.e.
timelessly) true and that this holds of all propositions. Furthermore, this
would seem to be part and parcel of our everyday assumptions about
 15.4. Tense and deictic temporal reference 687
facts and true propositions. And yet we have distinguished between
tenseless and time-bound propositions, and within time-bound proposi-
tions between omnitemporal and temporally restricted propositions.
This apparent contradiction is resolved by drawing a distinction between
the world (or state-description*: cf. 6.7) in which a proposition is true
and the world of which a proposition is true. A timeless proposition, if
true, is timelessly true of some timeless state-of-affairs; a time-bound
proposition, if true, is timelessly true of a time-bound situation; and so
on. In other words, the intensional world 1n which a proposition is true
is timeless; but the extensional world, actual or potential, of which a
proposition 1s true may be timeless or time-bound. It follows that, for
anyone who takes the view that truth is eternal, “It is the case that p”’
will always be timeless, regardless of whether p stands for a timeless or
a time-bound proposition. This point is of considerable importance; and
we will return to it later.
All that remains now is to say something about the difference between
the rather narrow view of tense that has been taken in this section and
the much broader view of tense that is taken in traditional grammar. As
the term ‘tense’ is traditionally employed, it covers, not only what is
here classified as tense, but a range of other time-related distinctions
which are nowadays subsumed, by linguists at least, under the term
‘aspect’. Aspect will be dealt with in a later section (cf. 15.6). But there
are three points relating to the connexion between tense and aspect that
may be conveniently made here.
The first point is that aspect, which differs from tense, as a gram-
matical category, in that it is non-deictic (having to do with such
distinctions as extension in time vs. instantaneity, completion vs. non-
completion and iteration vs. non-iteration), is probably far more wide-
spread throughout the languages of the world than tense is. Many
languages that do not have tense (e.g., Chinese, Malay, Classical
Hebrew) grammaticalize one or other of these aspectual distinctions. It
is not uncommon, however, for languages to have both tense and aspect.
In English, for example, the so-called progressive aspect, which we may
think of as expressing duration (i.e. extension in time), may be combined
with either the past or the non-past tense, such that its presence or
absence yields what would be described in traditional grammar as four
distinct tenses of the verb: the simple present (cf. ‘John works’), the
present progressive (cf. ‘John is working’), the simple past (‘John
worked’), the past progressive (‘John was working’). It is obvious that
the semantic distinction between ‘John sings’ and ‘John is singing’,
 688 Deixis, space and time
unlike the distinction between ‘John sings’ and ‘John sang’, is non-
deictic.
The second point to be made is that, although both tense and aspect
may be found in the same language, it is not uncommon for there to be
gaps and asymmetries. For example, in Russian there is a distinction
between the perfective and imperfective aspect, which operates dif-
ferently in the past tense from the way in which it operates in the non-
past tense; in Classical Greek the aorist aspect is restricted (in the
indicative mood) to the past tense; in Turkish there is a set of simple
(i.e. aspectually unmarked) past-tense forms, but there are no corres-
ponding simple non-past forms (in the indicative mood); in literary
French there is an opposition in the past tense (cf. ‘Jean travailla’ vs.
‘Jean travaillait’), which is not matched in the present tense (cf. ‘Jean
_ travaille’). Examples like this could be multiplied almost indefinitely.
Furthermore, there are languages, like Latin, which, at first sight,
appear to exhibit no such gaps or asymmetries, but which put one set of
forms to double use, and thus draw a wider range of semantic distinc-
tions in the past tense than they do in the non-past tense.
The fact that there should be such gaps and asymmetries in particular
languages is by no means inexplicable. It derives, in part, from the two
different conceptions of time — the static and the dynamic conception —
referred to earlier and, in part also, from two rather different kinds of
description: historical* and experiential*. The importance of this dis-
tinction between two kinds, or modes*, of description has often been
recognized (cf. Bull, 1963; Weinrich, 1964; Benveniste, 1966: 239;
Ducrot & Todorov, 1972: 398ff).!® But there is no established ter-
minology for it. The term used here, ‘historical’, is intended to suggest
the narration of events, ordered in terms of successivity and presented
dispassionately with the minimum of subjective involvement; and this
mode of description clearly relates to the static, non-deictic, objective
conception of time. The term ‘experiential’, on the other hand, is
suggestive of the kind of description that might be given by someone
who is personally involved in what he is describing; and this mode is no
less clearly related to the dynamic, deictic, subjective conception of
time.
It is possible, though unusual, to adopt the historical mode of descrip-
tion for the narration of successive current events, presenting each as if
19 "The term ‘mode’ has occasionally been used by linguists in a sense in which
it approximates to ‘mood’. It has other technical and semi-technical uses too
(cf. chapter 16, n. 18).
 15.4. Lense and deictic temporal reference 689
it were instantaneous: more commonly, we adopt the experiential mode
for the description of contemporary situations. It is for this reason that
aspectually unmarked simple non-past sentences in English (cf. ‘John
sings’) are only rarely construed as referring to events. As far as the
description of past situations is concerned, we can much more freely
adopt the one mode, rather than the other: we can present a situation
(e.g., John’s singing) as an event (cf. ‘John sang’) or as a process (cf.
‘John was singing’); and we can switch between the one mode of
description and the other for stylistic or rhetorical purposes. But the
historical mode is the norm from which the experiential mode constitutes
a deviation; and this is why the past progressive in English is less
frequently used than the present progressive. It also explains why the
distinction between perfective and imperfective aspect operates dif- ,
ferently in the past and non-past in Russian; why there is an aspectually
unmarked past tense in Turkish, but no aspectually unmarked present
tense; and so on. Furthermore, it accounts for the fact that there are
many languages that have what is commonly described as a special
narrative or consecutive tense: this is not a tense, in the narrower sense
of the term ‘tense’ adopted here. Its function is non-deictic; and it 1s
used in the historical mode of description to chronicle, or narrate, the
occurrence of serially ordered events, without regard to their pastness,
presentness or futurity or to any other deictic notion.
This brings us to the final point. Throughout this section we have
been concerned with what is sometimes referred to as primary tense.
Many linguists recognize, in addition to the primary tense-distinctions
based on deixis, a range of secondary tense-distinctions derived from
them. For example, what is traditionally described as the pluperfect in
English (cf. ‘John had sung’ and ‘John had been singing’) might be
classified as a past-in-the-past tense, whose function it is to express the
anteriority of one situation in the past relative to another situation in the
past. Anteriority is not of course a deictic notion: so that ‘earlier-in-the-
past’ might be more appropriate than ‘past-in-the-past’. Furthermore,
it is often very difficult to draw a distinction between secondary tense
and aspect. Anteriority is not always distinguishable from completion or
_ termination; and it is for this reason that linguists are still undecided as
to whether the so-called perfect and pluperfect in English (cf. ‘John has
sung’, ‘John had sung’, etc.) are to be distinguished from the corres-
ponding non-perfect forms in terms of tense or in terms of aspect (cf.
15.6). All that needs to be said here is that there is an obvious connexion,
not only between anteriority and completion, but also between so-called
690 Deixis, space and time
secondary tense and what was referred to earlier as deictic projection and
empathetic deixis (cf. 15.1, 15.3). The speaker projects himself back-
wards or forwards in time, as it were, into some other world, from which
events appear to him as being in the past or in the future. It is from this
point of view that we shall look at the question of so-called secondary
tense in a later chapter (cf. 17.3). But it must be recognized that at this
point there is not, and cannot be, in universal grammar any sharp dis-
tinction between tense and aspect, on the one hand, or between tense
and modality, on the other.

15.5. Spatial expressions

How do we explain to someone where an object is? And how do we
describe the spatial characteristics of particular objects — their extension
in space and their shape? These are the questions with which we shall
be concerned in this section; and, as we shall see, they are inter-
connected.”°
Looked at from one point of view, man is merely a middle-sized
physical object. But in man’s world —the world as man sees it and
describes it in everyday language — he is, in the most literal sense, the
measure of all things. Anthropocentrism and anthropomorphism are
woven into the very fabric of his language: it reflects his biological
make-up, his natural terrestrial habitat, his mode of locomotion, and
even the shape and properties of his body (cf. H. Clark, 1973; Miller &
Johnson-Laird, 1976: 375ff).
We live and move, normally, on the surface of the earth (rather than
in water or in the air); and we do so, again normally, in an upright
position. This gives us the means of identifying one of the dimensions
in a three-dimensional space; it also gives us a fixed zero-point at
ground-level. Furthermore, directionality in the vertical dimension —
i.e. the difference between upwards and downwards — is established by
our experience of the effects of the force of gravity, by the fact that,
normally, the sky is above us and the ground beneath us and by the
asymmetry of the human body in the vertical dimension. For these, and
other, reasons, verticality is physically and psychologically the most
salient of the spatial dimensions: linguistically, as we shall see, it is the
primary dimension (cf. Bierwisch, 1967).
There are two horizontal dimensions, neither of which is fixed, in the
way that verticality is, by the force of gravity or anything comparable.
20 1 have been much influenced in forming the views that are expressed in this
section by Jessen (1974).
 15.5. Spatial expressions 691
Moreover, man is mobile; and he can pivot freely in a horizontal plane.
But he is asymmetrical in one of the two horizontal dimensions, and
symmetrical in the other: i.e. he has a front and a back, and two sym-
metrical sides. He has his principal organs of perception directed
towards the region in front of him; he normally moves in the direction
in which he is facing; and, when he interacts with his fellows, he does so,
in what has been felicitously described as the canonical encounter* (cf.
H. Clark, 1973), by confronting* them. The asymmetrical front—back
dimension is less salient than the vertical dimension, but more salient
than the symmetrical right—left dimension. Linguistically, the front—
back dimension, then, is the secondary dimension.
Man is symmetrical about the front—back plane; and there is no
obvious directionality in what we may now identify as the third dimen-
sion. Recognition of the difference between right and left is dependent
upon the prior establishment of directionality in the front—-back dimen-
sion. Right could, of course, be distinguished from left on the basis of
the predominance of right-handedness in any human population; and it
is noticeable that it is the phenomenon of dexterity that has provided the
word for the right-hand side in many languages. The right hand is the
one that is used characteristically to do things that require some kind of
dexterity: it 1s the hand with which one eats, writes, grooms oneself, and
so on. The right-left dimension is nonetheless dependent upon the
front—back dimension: we classify people as left-handed, rather than as
back-to-front, if they do not happen to conform to the statistical norm
for dexterity.
In the up—down and front-back dimensions there is not only direc-
tionality, but polarity: what is above the ground and in front of us is,
characteristically, visible to us and available for interaction; what is
beneath the ground or behind us is not. Upwards and frontwards are
positive, whereas downwards and backwards are negative, in an ego-
centric perceptual and interactional space based on the notions of
visibility and confrontation. There are no such reasons, however, for
recognizing a positive and negative polarity in the right—left dimension:
dexterity provides, at best, a rather weak criterion for classifying the
right-hand side as positive and the left-hand side as negative. It has been
plausibly argued that polarity and markedness in pairs of directional
opposites derive, not only in the vocabulary of location and locomotion,
but more generally, from the natural properties of the ego-centric per-
ceptual space and the spatial orientation and physical asymmetries of the
human body (cf. 9.2, 9.7). We will restrict our attention to questions
692 Deixts, space and time
having to do with the location and spatial extension of objects; and we
shall see that all the points that have been made so far are relevant.
If we want to tell someone where something is, we need some means
of identifying direction from any one point in space to any other.
Furthermore, it is convenient, if not essential, to have units of measure-
ment (and numerals). We will not go into the question of measurement,
except to emphasize two points that are of particular importance in the
present connexion. ‘The first point is that there are two ways in which we
can conveniently measure the distance (without the use of special
instruments) between X and Y: (1) in terms of the time taken for some-
one or something to move from X to Y at a known and roughly constant
speed; and (ii) in terms of something that can be used as a ruler (of
unit-length) such that the distance can be said to be a fraction or multiple
of the length of the ruler. Using method (i), we can say that X is an
hour’s journey from Y (it being understood from the context that the
journey is to be made on foot, on horseback, by car, etc.) or, at a more
sophisticated level and for the measurement of stellar or galactic dis-
tances, that X is so many light-years from Y. Using method (ii), we can
say that X is so many inches, feet, yards or miles from Y. What is to be
emphasized with respect to method (11) is that the everyday units of
measurement with which we operate were all derived, originally, from
measurements on the rulers that we carry around with us everywhere:
our own bodies. Given that human beings are of roughly equal size and
proportions (or, at least, that there is an acceptable consensus about the
average size of hands, feet, stride, etc.) this is a perfectly satisfactory
system of measurement for most purposes. It may not meet the exacting
standards of modern science and technology; but it has other advan-
tages.
The second point to be made in connexion with the measurement of
distance is that the distance from X to Y is not necessarily the same as
the distance from Y to X. This 1s obviously so if the distance is measured
by method (1). But it may also be the case when distance is measured by
method (ii). ‘There will always be some assumption made with respect to
the route taken; and, for all sorts of reasons, the normal route from X to
Y may be shorter or longer than the normal route from Y to X. This is
an important point. For there is much to suggest that in the measure-
ment of distance an asymmetrical, and covertly dynamic, construction
like ‘‘How far is it from X to Y?”’ is more basic than the more sym-
metrical and static construction “What is the distance between X and
Y?”
 15.5. Spatial expressions 693
The physical world contains a certain number of discrete, or relatively
discrete, three-dimensional first-order entities (cf. 11.3). Of these, some
(notably human beings and animals) are self-moving and others, though
not self-moving, are moveable. ‘There are some first-order entities that
are either permanently or normally static, rather than self-moving or
moveable: but they will not count as first-order entities unless the
language so classifies them and they stand out from their environment
with respect to their colour, shape or texture. Such aggregates, collec-
tions or conglomerations of matter as cliffs, mountains, clouds, lakes,
and so on, may or may not be perceived and conceptualized as first-order
entities: their status is ontologically indeterminate; and they may be
treated differently by different languages. We shall be concerned solely
with the location and spatial characteristics of what are incontrovertibly
first-order entities according to the assumptions of naive realism.
As we saw in a previous chapter, places are not entities, though they
may be hypostatized* and treated as entities in particular languages (cf.
11.3). The distinction between entities and places will be of considerable
importance in this section. As places are not entities, so entities are not
places; but, in so far as they occupy space, entities may serve to identify
the spaces that they occupy. For example, in
(1) I'll meet you at the car
‘the car’ is used indirectly to identify a place: i.e. the space that is
occupied by the car. We will assume, therefore, that what (1) means is
‘T will meet you at the place where the car is’’; and we will also assume — |
that what
(2) John is with Peter
means is ‘‘John is where Peter is’’. We will make no attempt to justify
these analyses, noting only that they would seem to be semantically
appropriate, in that (1) and (2) may be used in reply to locative questions
like Where will you meet me? and Where is fohn? The principles which
determine the selection of ‘with’, rather than ‘at’, in (2) need not detain
us. Nor shall we go into the difference between (2) and
(3) Fohn is by Peter.
What must be emphasized is that in all such instances we are relating an
entity to a place. But we refer to the place indirectly in terms of the
entity that it contains; and this is tantamount to treating the entity as a
property of a place (cf. 12.3).
694 Deixis, space and time
In relating an entity (X) to a place (Y), in locative propositions like
(4) “X be located at Y”’
we must decide, as far as English and many other languages are con-
cerned, whether Y is to be represented as having dimensionality or not.
The simplest kind of spatial relation is the one for which, characteristic-
ally, the preposition ‘at’ is employed in English. We may symbolize this
therefore as
(5) aT (X, Y).
This relation is grammaticalized in the locative* case in many languages;
and it may be thought of as being encapsulated in the deictic adverbs
‘here’ and ‘there’ in English (so that “X 1s here/there’ means, as it
were, ‘“X is at here/there’’). Indeed, (4) and (5) may be treated as
equivalent, provided that Y is represented as a place whose size and
dimensionality is irrelevant or negligible. If Y is represented as having
dimensionality, the choice lies between
(6) on (X, Y)
and
(7) IN (X, Y).
Of these, (6) is appropriate when Y is represented as a line or a surface,
and (7) when Y is represented as an enclosed area or volume (cf. Leech,
1969: 161ff; Bennett, 1975: 65ff).
Associated with each of these positional*, or static, relations there are
two directional*, or dynamic, relations. One of them treats Y as the
goal* and the other as the source* of locomotion:
(5a) TO (X, Y)
(6a) ONTO (X, Y)
(7a) INTO (X, Y)
and
(5b) FROM (X, Y)
(6b) oFF (X, Y)
(7b) ouT oF (X, Y)
The goal-relation is grammaticalized in many languages in the allative*
(or dative*) case, and the source-relation in the ablative* case. Both of
 15.5. Spatial expressions 695
these relations have been discussed earlier in connexion with the notion
of valency and action-schemata (cf. 12.4).
As far as polarity is concerned, the goal-relation is positive and the
source-relation is negative. Distance, height and depth are measured
from the zero-point to some point along the dimension in question. The
following implicational relations justify the classification of the goal-
relation as positive and the source-relation as negative: “‘X have moved
to Y’’ + “X have moved from not-Y”’ and ‘‘X have moved to Y’”’ —>
‘“X be at Y”’ (cf. 9.2). These implicational relations also constitute
evidence for the view that the goal-relation is the dynamic correlate of
the static locative relation (cf. Anderson, 1971: 119ff). So too does the
fact that in many languages the same preposition or the same case is used
for both the locative-relation and the goal-relation (cf. French ‘a’ for AT
and To; Latin ‘in’ for IN and INTO, and for ON and ONTO; German ‘auf’
for ON and ONTO; etc.). This correlates with the fact that the source-
relation and the locative relation are very rarely conflated: ‘Where has
he gone?’ in English cannot be interpreted as meaning anything other
than ‘Where has he gone to?”’ (i.e. ““where”’ and “‘whither’”’ are con-
flated, but not ‘where’ and ‘‘whence”’). Finally, it may be observed
that there is, in English, the possibility of using source-expressions in
what appear to be, at first sight, positional, rather than directional,
constructions: cf.

(8) He is out of the house.

But (8) is by no means equivalent to

(9) He is not in the house,

though the proposition expressed by (8) entails the proposition ex-
pressed by (g). The meaning of (8) is close to, though not identical with,
that of

(10) He is no longer in the house.

In other words, (8) implies a change of state, whereas (g) does not.
The use of such terms as ‘north’, ‘south’, ‘east’ and ‘west’ exempli-
fies one method for specifying the location of objects within a stan-
dardized and, in its modern development, precise and sophisticated
frame-of-reference. It is also possible to locate objects in relation to the
rising or setting sun, to a relatively fixed star or constellation, to the
prevailing wind at such-and-such a time of year, to some prominent
 696 Deixts, space and time
landmark in the environment (the sea, a river or range of mountains,
etc.), and so on.”? Many languages lexicalize, or even grammaticalize,
orientational distinctions of this kind. But here we are concerned with
the alternative, and, in everyday life, more common way of describing the
location of objects or the direction in which they are moving, by means
of the oppositions “‘up”’ ws. ““down”’, “front”’ vs. “back” and “‘right”’
vs. “left”’. These distinctions, combined with the deictic distinction of
‘“here”’ us. ‘‘there’’, would seem to be more basic, as far as English and
many other languages are concerned, than distinctions which involve
reference to some fixed landmark or natural phenomenon.
For example, it is impossible to specify the meaning of such preposi-
tions in English as ‘above’, ‘below’, ‘behind’ and ‘beside’, or of such
so-called complex prepositions as ‘in front of’ and ‘at the side of’,
without drawing upon the oppositions “up” ws. “down” and “‘front”’
vs. “‘back”’ (cf. Bennett, 1975). ‘X 1s above Y’ means roughly, ““X is in
the space that is adjacent to the top of Y’’; ‘X is in front of Y’ means
‘““X is in the space that is adjacent to the front of Y’’; and so on. It is
interesting to note, in this connexion, that, although there is a preposi-
tion ‘beside’ (and in certain uses ‘by’) which is neutral with respect to
“right” vs. “left”, there is no preposition that 1s neutral with respect to
the directionality and polarity of the vertical dimension or the major
horizontal dimension. This is hardly surprising in view of the greater
salience of the up-down and front—back dimensions and the anthropo-
morphic asymmetries associated with them.
The principles which determine the use of the words ‘top’, ‘bottom’,
‘front’, ‘back’ and ‘side’ in English are fairly complex, but they are
essentially systematic. First of all, it should be observed that, like many
English lexemes, when expressions containing them are applied to
three-dimensional objects, such expressions may refer either to three-
dimensional parts or two-dimensional surfaces of three-dimensional
objects: cf. ‘the top of the cupboard’ in ‘The top of the cupboard is
detachable, but it’s very heavy’ and ‘It’s on the top of the cupboard’,
respectively (cf. Leech, 1969: 173ff; Teller, 1969). Furthermore, the
English word ‘part’ can be used in expressions which refer either to
entities (i.e. to physically or conceptually detachable portions of a larger
entity) or to places (i.e. to spaces enclosed by the extremities, or boun-
2l It is interesting to note that in Old Irish the words for ‘‘front”’, ‘‘back’’,
“right” and ‘“‘left”’ are used to translate ‘east’, ‘west’, ‘south’ and ‘north’,
respectively (cf. Thurneysen, 1946: 305). I am indebted to Anders Ahlqvist
for this information and the reference.
 15.5. Spatial expressions 697
daries, of an entity or of its parts). The same is true of lexemes like
‘top’, ‘bottom’, ‘front’, ‘back’, etc. Hence the ambiguity of
(11) There are some nails in the top of the cupboard.
All such nominals as ‘the top of the cupboard’ have three possible inter-
pretations, though it will generally be clear from the preposition that is
used with them (e.g., ‘on’ ws. ‘in’) or from other information in the
context which of the interpretations is relevant.
Part—whole relations of the kind that we are concerned with here are
transitive (cf. 9.8); and they preserve dimensionality, in the sense that a
three-dimensional entity or space has as its parts other three-dimensional
entities or spaces; a two-dimensional space has as its parts two-dimen-
sional spaces; and a (one-dimensional) line has as its parts (one-
dimensional) lines. The relationship between an entity (or a space) and
its extremities involves a reduction of dimensionality: a three-dimen-
sional object has as its extremities one or more two-dimensional surfaces;
a two-dimensional surface or area has as its extremities (or boundaries)
one or more (one-dimensional) lines; a line need not have extremities,
but, if it does, they will be (non-dimensional) points. These geometrical
considerations determine the interpretation, not only of ‘top’, ‘bottom’,
‘front’, ‘back’, etc., but also, as we shall see presently, of such lexemes
as ‘long’, ‘wide’ and ‘high’ or ‘deep’.
It is not only the dimensionality* of entities and spaces that is relevant
to the classification of their parts and extremities. Shape is also relevant.
Apart from considerations of symmetry and asymmetry, which have
already been mentioned, there is the distinction of angularity vs. round-
ness to be taken into account. The recognition of edges and corners
depends upon the perception of angularity; and the number of surfaces
(or boundaries) assigned to an entity (or space) is a function of the
number of edges and corners it has. Edges and corners may be regarded
as either parts or extremities: cf. ‘on the edge’ vs. ‘at the edge’; ‘at the
corner’ vs. ‘in the corner’ vs. ‘on the corner’. Other aspects of shape
may be disregarded for the moment.
Most important of all, as far as the semantic analysis of locative and
directional expressions is concerned, is the orientation* of entities (and
spaces) with respect to the up—down, front—back and right-—left dimen-
sions; and here we have to take into account, in particular instances, the
difference between the inherent, the canonical and the actual orientation
of entities. Immobile terrestrial objects (such as mountains, buildings or
trees) are inherently oriented in the vertical dimension, but in no other
698 Deixis, space and time
dimension: they may, however, be canonically* oriented in the front—
back dimension. As we have seen, human beings and animals, and in
general self-moving (rather than merely moveable) entities, have an
inherent front—back dimension, and they are canonically oriented in the
vertical dimension. Moveable entities are not inherently oriented in any
dimension (unless they are unstable in certain positions). They may,
however, have a canonical orientation in either the up-down or the
front—back dimensions.
Such expressions as ‘upside down’ and ‘back-to-front’ are instructive
with respect to the distinction between the canonical and the actual
orientation of an entity. We cannot sensibly use ‘upside down’, for
example, except of an entity that has a canonical top and bottom (i.e. a
canonical upper and lower part); and we cannot use ‘upside down’
correctly (other than by chance) unless we can distinguish the canonical
from the actual orientation on particular occasions.
The most important factor in the assignment of canonical orientation
in the horizontal plane is what was referred to earlier as confrontation*.
When two people are involved in conversation or some other kind of
interaction (in a canonical encounter) they normally confront* one
another: i.e. each turns his front to the other. It follows that anything
that is between X and Y, when they confront one another, will be in
front of both of them; but whatever is to the right of X will be to the left
of Y; and whatever is behind X will be in front of Y. The assignment of
‘right’ and ‘left’ will differ, therefore, according to whether the speaker
uses the orientation of the confronting entity (X) or the confronted
entity (Y) in order to compute the assignment. Problems also arise, as
we shall see, with respect to the interpretation of ‘behind’.
It is by virtue of the characteristic orientation of the body in the
canonical encounter that a canonical front—back dimension, and secon-
darily a right-left dimension, is assigned to a variety of immobile and
moveable entities. The front of a house is the part or extremity from
which it is normally confronted; so too is the front of a piano, a desk or
a cupboard. Given that such entities as buildings and pieces of furniture
can be said to have a canonical front and back, the ambiguity of sen-
tences like
(12) The church is behind the town-hall
is readily explained (cf. Leech, 1969: 168). Under one interpretation the
speaker is understood to be locating the church in the space that 1s
adjacent to the town-hall’s canonical rear. Under the other he is under-
 15.5. Spatial expressions 699
stood to be disregarding the canonical orientation and treating the
town-hall as he might treat a tree or some other such entity that stands
between him (or someone else whose point of view he is adopting) and
the church. These two interpretations will, of course, coincide if the
actual orientation of the town-hall with respect to the speaker (or the
person whose point of view he 1s adopting) coincides with its canonical
orientation.
In the case of most self-moving entities, such as trains, cars and ships,
it seems to be the criterion of the direction of locomotion, rather than
the notion of confrontation, which serves to identify the canonical front.
The ambiguity of ,
(13) She is standing in front of the car
is to be accounted for in essentially the same way as the ambiguity of
(12).
There is much else that would need to be discussed in a fuller treat-
ment of the dimensionality and orientation of entities and spaces. We
will make just two further points. The first is that there is an intuitive
connexion between the deictic distinction of “here”’ vs. ‘“not-here”’ and
the non-deictic distinction of “‘inside”’ vs. “‘outside’’: i.e. ‘X is here’
can be interpreted as “X is within the space which contains sELF’’. The
notion of containment, or interiority, is obviously a very basic notion;
and there may be grounds for introducing it into the analysis of the
meaning of such prepositions as ‘above’, ‘below’, etc. For example,
*X is above Y’ might be understood to mean “X is outside the space
containing Y and upwards of it’’. This analysis, it will be noted, natur-
ally accounts for the deictic interpretation of the locative adverbs

SELF’’). }
‘above’, ‘below’, etc. “X is above’, when used deictically, would be
understood to mean ‘‘X is outside and upwards of here”’ (or, equiva-
lently, as “X is outside and upwards of the space containing

The second point to be stressed is that many, if not all, of the locative
expressions involving dimensionality and orientation can be seen as
being implicitly, or covertly, directional: i.e. as being dynamic, rather
than static. It has been argued that, in order to account for the meaning
of such overtly directional sentences as
(14) Gwyneth walked through the kitchen,
we have to recognize expressions which are used to refer to the path*
(or route) that is taken on the journey from a source to a goal (cf.
700 Detxis, space and time
Bennett, 1975: 18ff). Once the notion of a path is recognized in such
obviously directional expressions as ‘through the kitchen’ in (14), it
can also be identified in the analysis of what appears to be locative
expression:
(15) Gwyneth is through the kitchen.
The place referred to is here identified by means of the route that must
be taken in order to get there. It is in this sense that a locative expression
may be implicitly, or covertly, dynamic.
This point is more significant than it might seem to be at first sight.
It has already been noted that the asymmetrical and dynamic construc-
tion ‘How far is it from X to Y?” can be seen as more basic than the
more symmetrical and static construction ‘‘What is the distance
between X and Y?” Distance is always measured from a point-of-
reference* (to use this term in a sense that is related to, though some-
what different from, the sense in which it is employed in model-
theoretic semantics: cf. 6.6). To say that X is three miles away from Y
implies that one must travel three miles from the point-of-reference, Y,
in order to arrive at X. If there is no other point-of-reference, explicit or
implicit, in the context of an utterance like
(16) The church ts three miles away,
it is generally the location of the speaker that is taken to be the starting-
point of the actual or imaginary journey. Locative expressions involving
orientation with respect to “‘up”’ vs. ““down”’ and “front” vs. “‘back”’
are similar to expressions like ‘three miles away’ in that there is always
a point-of-reference explicitly or implicitly referred to in the utterances
containing them. It is always possible to give a dynamic interpretation
in terms of a journey from the point-of-reference.
Furthermore, visual perception can be seen in terms of the metaphor
of travel: one looks into the distance at, or towards, an object; one’s
gaze travels to and reaches, or grasps, the object. The ambiguity of (12),
“The church is behind the town-hall’, which we accounted for in terms
of confrontation, has been explained in terms of the difference between
(i) ‘The church is on the other side of the town-hall from where I am
looking”’ and (ii) “The church is on the other side of the town-hall from
the road”’, with (i) involving a deictic point-of-observation (“seen from
here’’) and (ii) a non-deictic point-of-orientation (cf. Leech, 1969: 167,
182). If perception is interpreted as making a kind of visual journey from
a point-of-reference, as source, to the object of perception, as goal, the
 15.5. Spatial expressions 701
fact that a point-of-observation and a point-of-orientation, if explicitly
mentioned, are referred to by means of a source-expression (‘from
here’, ‘from the road’, etc.) is automatically explained. So too is the
fact that many locative expressions can be seen as covertly, or implicitly,
dynamic. What we may now refer to as the action-schema of the journey
involving a source, a goal and, optionally, a path is of very general
applicability in the semantic and grammatical analysis of languages (cf.
12.6, 15.7).
The shape, dimensionality and orientation of entities (and spaces) is
crucial in the analysis of the meaning of such positional and qualitative
adjectives in English as ‘long’:‘short’, ‘far’:‘near’, ‘high’:‘low’,
‘deep’: ‘shallow’, ‘wide’:‘narrow’, and ‘thick’:‘thin’. Positional
adjectives can always be regarded, from a semantic point of view at least,
as being transformationally derivable from predicative locative struc-
tures: for example, a distant building is one that is relatively far from
some point-of-reference. A long building, on the other hand, is one that
is extended in a particular spatial dimension; and its length can be
regarded as one of its physical properties independent of its location in
space. Since ‘high’ and ‘low’, unlike most of the antonymous adjectives
with which we are concerned, can be used in either kind of construction,
phrases such as ‘a high window’ are, in certain contexts, ambiguous:
(i) “a window that is relatively far up (from a point-of-reference)”’ vs.
(ii) ‘a window that is significantly extended in the vertical dimension”’.
The distinction between the positional and the qualitative sense of an
adjective like ‘high’ rests upon the distinction between distance and
extension. But these notions, as we have seen, are interrelatable by
virtue of their correlation in terms of measurement: if a street is a
hundred yards long, measured from one extremity to the other, the two
extremities are a hundred yards apart. There is also an obvious percep-
tual correlation between extension and distance; and in so far as per-
ception can be analysed in terms of the notion of a visual journey, the :
static qualitative notion of extension can be based upon the more dyna-
mic notion of distance.
In the discussion of the factors which control the selection of the
correct qualitative adjectives of extension, it is convenient to neglect,
initially, the question of orientation (inherent, canonical and actual).
This immediately eliminates ‘high’ (and ‘tall’) vs. ‘low’; but not, as we
shall see, ‘deep’ vs. ‘shallow’.
The first question to be asked in relation to three-dimensional
physical objects is whether they have a maximal dimension (1.e. whether
 702 Deixis, space and time
their extension is significantly greater in any one dimension than it 1s in
the other two). If an object has no maximal extension, it cannot be said
to have length. We do not say of a tennis ball that it is longer than a golf
ball, and we would not normally say that it is wider, broader, thicker or
deeper. Instead, we use the most general qualitative adjectives of spatial
extension, which denote overall size regardless of shape and dimen-
sionality: ‘big’ vs. ‘little’, ‘large’ vs. ‘small’.
If an object has a maximal dimension, this is identified as its length;
and the question of overall shape becomes criterial for the labelling of
the non-maximal dimensions. If the extension of the object in the other
two dimensions is negligible in relation to its length, we then collapse
these two dimensions, as it were, in the single dimension of thickness:
we talk, for example, of a long thick pole. If the object is significantly
extended in one of the other dimensions, then the opposition of ‘wide’
vs. ‘narrow’ comes into play, but only if one of the two non-maximal
extensions is significantly greater than the other; and it is this dimension
to which ‘wide’ vs. ‘narrow’ applies. The third dimension, if identified
and described, is subject to a further, and in part independent, criterion:
whether the object is regarded as hollow or not (i.e. whether it has an
interior space). If it is hollow, it may be described as being deep or
shallow; if solid, as being thick or thin. In summary, then, we can say
that any unoriented three-dimensional object whose dimensions are
identifiable and significantly different in extension will have length as its
first dimension, width as its second dimension and either thickness or
depth as its third dimension.
For two-dimensional spaces (or figures) the same general considera-
tions of shape and maximality apply to determine the first dimension as
being that of length. In labelling the second non-maximal dimension,
| however, we choose between width and thickness, according to whether
the non-maximal dimension is treated as being significant or not in its
extension. We can describe a line as long and thin (as we can so describe
a pole or stick), but not as long and narrow. To say that something is a
line is to say that it is essentially unidimensional, approximating to the
ideal geometrical line. In contrast, we would describe a street as long and
narrow, rather than as long and thin: the non-maximal dimension of the
street is, of necessity, significant in its extension.
So far we have been dealing with unoriented entities and spaces. As
soon as we take orientation into account, we must concede primacy to
the vertical dimension. The general principle is that verticality (whether
inherent, canonical or actual) is dominant over maximality. Height (and
 15.6, Aspect 703
tallness) are measured upwards from the point-of-reference, which is
commonly, though not necessarily, ground-level. Depth 1s measured
downwards from the point of reference: but it applies only to spaces or
to hollow entities; and this correlates with the use of ‘deep’ vs. ‘shallow’
for the third dimension of unoriented hollow entities.
The vertical dimension is not necessarily the maximal dimension.
Whether it is or is not, orientation is again dominant in the assignment
of length, width and depth or thickness to the non-vertical dimensions.
If the entity has an inherent or canonical front, it has width (from side
to side) and depth or thickness (from front to back); if it has no inherent
or canonical front, it has length (from end to end) and width (from side
to side). The same rectangular building may be described in terms of
width and depth, according to whether it is treated as having a canonical
front or not; and a dual-purpose piece of furniture may be described as
long and wide, when it is being used or considered as a table, but as wide
and deep when it is thought of as a desk with a canonical orientation in
the horizontal plane. The important point to note is that it is contradic-
tory to say that the width of an object exceeds its length: we simply do
not talk about its length if, by virtue of its canonical orientation, we have
assigned width to its greater horizontal extension.
This very brief account of the meaning of qualitative and positional
spatial adjectives in English is incomplete; and it may be inaccurate in
certain details. But it will serve to show how intimate is the connexion
between the lexicalization of distinctions of shape and the lexicalization
of dimensionality and orientation in English.
There is no reason to believe that English is in any way untypical in
this respect. German (cf. Bierwisch, 1967) and French (cf. Greimas,
1966: 35) are very similar, as are the other Indo-European languages. As
for shape, and more especially dimensionality, it should also be men-
tioned that there are many languages in which this appears to be one of
the principal factors determining the selection of classifiers (cf. 11.4). It
is frequently the case that one classifier is used for saliently one-
dimensional objects, another for saliently two-dimensional objects and
a third for saliently three-dimensional objects; and there are other
places in the structure of language in which these distinctions are
grammaticalized or lexicalized (cf. Friedrich, 1970).

15.6. Aspect
That the notion of aspect*, if not the term ‘aspect’ itself, is less familiar
to non-linguists than are the notions of tense and mood is largely a
704 Deixts, space and time
matter of historical accident. The Stoics had realized that, in addition
to precedence and successivity, there was another factor involved in the
determination of what Aristotle, and subsequently the Alexandrians,
referred to as tense (i.e. “time’’); and they identified this as being what
we would now describe as an aspectual distinction of completeness vs.
incompleteness (cf. Robins, 1967: 29). Unfortunately, in the later
development of the Greco-Roman grammatical tradition, which has in-
fluenced, and in many respects distorted, the grammatical analysis of the
majority of the world’s better-known languages and the way they are
taught in our schools and universities, the terms ‘perfect’ and ‘im-
perfect’ (which derive from the Latin translations of ‘complete’ and
‘incomplete’) came to be used in collocation with ‘tense’. Furthermore,
the definitions of the so-called present, perfect, imperfect and plu-
perfect tenses, not only in Greek and Latin, but also in other languages,
tend to obscure the difference between past vs. present vs. future, on the
one hand, and perfect vs. imperfect, on the other.
What is traditionally referred to as the present tense of Latin or
Greek is more appropriately described as being the present imperfect:
it is present (or, better still, both non-past and non-future) in tense and
imperfect in aspect. The so-called perfect and imperfect tenses contrast
with the so-called present tense in being present perfect and past im-
perfect, respectively; and the misleadingly named pluperfect (‘further
[in the past] than the perfect’’) 1s the past perfect. The morphological
structure of both Greek and Latin supports this two-dimensional
classification. So too, as far as Greek is concerned, does syntactic and
semantic analysis. In Latin, however, the present perfect forms were
also used in contexts in which, in the historical, as distinct from the
experiential, mode of description, Greek would use the so-called aorist
(egrapse), literary French would use the so-called past definite (2/ écrivit),
Russian (under certain rather more specific conditions) would use the
so-called past perfective (on naptsal), and so on. The fact that the
so-called perfect tense of Latin (scripsit) has this double function — in
which respect it is quite different from the so-called perfect tense of
English (he has written), literary French (il a écrit) or Standard German
(er hat geschrieben), but similar to the similarly named forms of collo-
quial French and some southern dialects of German — is responsible, no
doubt, for much of the confusion surrounding the terms ‘perfect’ and
‘perfective’ in linguistics (cf. Comrie, 1976: 16ff).
There can be no question of entering here upon a full-scale treatment
of aspect. Our purpose is merely to introduce the reader to some of the
 15.6. Aspect 705
aspectual distinctions that are grammaticalized in languages and to
emphasize the importance of these distinctions in the construction of a
general theory of the structure of language. It is, as we have said, largely
a matter of historical accident that the notion of aspect does not figure
as prominently in traditional grammar as does the notion of tense. Aspect
is, in fact, far more commonly to be found throughout the languages of
the world than tense is: there are many languages that do not have tense,
but very few, if any, that do not have aspect. Furthermore, it has been
argued recently that aspect is ontogenetically more basic than tense, in
that children whose native language has both, come to master the former
more quickly than they do the latter (cf. Ferreiro, 1971). We will not go
into this issue. But we would re-emphasize the fact that aspect has
frequently been confused with tense in the standard treatments of
particular languages; and that this has had its effect upon the more
theoretical discussions of temporal relations in language by linguists,
philosophers and psychologists. The main difference between tense and
aspect, as we have already seen, is that, whereas tense is a deictic cate-
gory, which involves an explicit or implicit reference to the time of
utterance, aspect is non-deictic (cf. 15.4). It has also been pointed out,
however, that the distinction between tense and aspect is hard to draw
with respect to what is sometimes described as relative, or secondary,
tense. We will say no more about this.
The term ‘aspect’ is currently used by linguists as the rather un-
satisfactory, but conventionally accepted, translational equivalent of the
| term that is employed in Russian (‘vid’) to refer to the opposition of
perfective and imperfective in the Slavonic languages. Usually, though
not invariably, it is extended to cover a variety of other oppositions, in
so far as they are grammaticalized in the structure of particular lan-
guages — oppositions based upon the notions of duration, instantaneity,
frequency, initiation, completion, etc. We have tacitly adopted this
relatively broad conception of aspect, according to which the opposition
| between the progressive and the non-progressive forms in English (cf.
he 1s writing vs. he writes), the opposition between the simple past and
the imperfect in literary French (il écrivit vs. il écrivatt), the opposition
between the progressive and the aorist forms in Turkish (okuyor, ‘“‘he
is writing’’, vs. okur, ““he writes regularly/habitually”’) and comparable
grammaticalized oppositions in other languages, are all legitimately
classified as aspectual.
A distinction is sometimes drawn between aspect and Aktionsart*.
The specialized employment of the German term ‘Aktionsart’ (which,
706 Deixis, space and time
in origin, meant nothing more than “kind of action’’) rests upon one or
other of two more general distinctions: (i) the distinction between
grammaticalization and lexicalization; and (11) the distinction, within
morphology, between inflexion and derivation. The fact that neither of |
these two distinctions is itself clearcut, coupled with the further fact
that, in so far as they are partially, but not wholly, coincident, some
scholars operate with the one and some scholars with the other, has been
responsible for a good deal of confusion in the use of the term “‘Aktion-
sart’ (cf. Comrie, 1976: 6). Partly for this reason and partly because
‘Aktionsart’ is in itself a very unsatisfactory term, in that (a) it is more
naturally applied to the denotata of verbs, rather than to some semantic
property of the verbs themselves, and (b) the term ‘action’ (traditional
though it is in this sense) is too narrow, we will make no further use of
the term ‘Aktionsart’. We will introduce instead the term ‘aspectual
character’. The aspectual character of a verb, or more simply its
character*, will be that part of its meaning whereby it (normally) denotes
one kind of situation rather than another. For example, ‘know’ differs
from ‘recognize’ in English, as ‘kennen’ differs from ‘erkennen’ in
German or ‘znatj’ from ‘uznatj’ in Russian, by virtue of its aspectual
character. ‘Know’ (like ‘kennen’ and ‘znatj’) normally denotes a state,
whereas ‘recognize’ (like ‘erkennen’ and ‘uznatj’) normally denotes an
event. It is generally accepted nowadays that any discussion of aspect
from a semantic point of view must also take account of what we are
referring to as the character of particular verbs. It is well known, for
example, that certain subclasses of verbs in English do not normally
occur in the progressive aspect (cf. Leech, 1971: 14ff; Palmer, 1974:
7off). Aspect and character are interdependent in this way because they
both rest ultimately upon the same ontological distinctions.
As we saw earlier, most verbs, in all languages, are inherently dynamic,
in that they normally denote either events (including acts) or processes
(including activities), rather than states (cf. 11.3). Indeed, this distinc-
tion between dynamic and static situations is an important part of what
we have taken to be the ontological basis for the syntactic and morpho-
logical distinction that some, but not all, languages draw between verbs
_and adjectives. As we also saw earlier, however, even in those languages
in which there is a sharp grammatical distinction between verbs and
adjectives, a minority of verbs may be stative, rather than dynamic, in
character: cf. ‘know’, ‘have’, ‘belong’, ‘live’, ‘contain’, etc., in English.
Stative verbs constitute the most important subclass of verbs that do not
normally occur in the progressive aspect in English. Stativity*, then, is
 15.6. Aspect 707
lexicalized, rather than grammaticalized, in English: it is part of the
aspectual character of particular verbs. It is grammatically relevant,
however, in that it is incompatible with progressivity*, which is gram-
maticalized in English, as it is in Spanish, Italian, Icelandic, Irish, etc.,
but not in French, German, Latin, Greek, Russian, etc. (cf. Comrie,
1976: 32ff). Whether a language grammaticalizes either stativity or pro-
gressivity (or both, or neither) is something that cannot be predicted in
advance of an empirical investigation of the grammatical and semantic
structure of the language. The incompatibility of stativity and pro-
gressivity is explicable, however, in terms of the language-independent
ontological distinction of static and dynamic situations. Such contrasting
pairs as the stative She has a headache and the non-stative She is having a
bath (comparable, semantically, with She 1s washing her hair) are es-
pecially instructive in this connexion. Even more instructive perhaps
is She 1s having a headache (or She is having one of her headaches), which
can be given several different interpretations according to the context in
which it occurs, but which must necessarily be construed as describing a
dynamic, rather than a static, situation.
Stativity and progressivity are but two of the semantic notions to
which reference is commonly made in general treatments of aspect (cf.
Comrie, 1976; Friedrich, 1974). Others are duration, completion,
habituality, iteration, momentariness, inception and termination. All of
them, it will be noted, are non-deictic temporal notions. It is our main
purpose in this section to show how they relate to, and depend upon, the
subclassification of situations that we first invoked in connexion with
the concept of valency (cf. 12.4). In all that follows we will take for
granted the validity of the distinction between events, states and pro-
cesses, on the one hand, and of the distinction between acts and activities
(as distinguishable kinds of actions), on the other. Events, it will be
recalled, are non-extended dynamic situations that occur, momentarily,
in time; processes are extended dynamic situations that last, or endure,
through time; states are like processes in that they too last, or endure,
through time, but they differ from processes in that they are homo-
geneous throughout the period of their existence; acts and activities are
agent-controlled events and processes, respectively (cf. 12.4). :
Certain consequences derive immediately from this classification. One
of them has been mentioned already: the incompatibility of stativity
with progressivity.22 Another is the possibility of grouping states and
22 This is a different explanation from the one given in Lyons (1968: 315f),
which is rightly criticized by Bennett (1975: 111).
708 Deixis, space and time
processes together, in contrast with events, in terms of the notion of
duration. In so far as aspect is concerned, either one of these two binary
distinctions might be taken to be more basic than the other and gram-
maticalized in particular languages. If we take the notion of markedness
into account, however, we see that there are, in principle, six, rather than
two, possibilities:
(i) stative vs. non-stative
(ii) dynamic vs. non-dynamic
(iii) stative vs. dynamic
(iv) durative vs. non-durative
(v) punctual vs. non-punctual
(vi) durative vs. punctual
We also see that the progressive vs. non-progressive distinction that is
grammaticalized in English cannot be identified with any of these six:
progressivity involves both dynamicity, unlike (iv), and durativity,
unlike (ii), Whereas progressivity is a natural enough concept, so that
states and events may be grouped together negatively in contrast with
processes, in a two-term system of aspects which grammaticalizes the
distinction between a marked progressive and an unmarked non-
progressive (as in English), it 1s difficult to see any positive reason for
bringing both events and states together in contrast with processes. To
the six possibilities just listed we may therefore add a seventh,
(vii) progressive vs. non-progressive
rather than a further three. All but (111) and (vi), it will be noted, are
privative oppositions of the kind that give rise to the structurally import-
ant phenomenon of markedness (cf. 9.7).
In the present state of our knowledge of the grammaticalization of
aspectual distinctions throughout the languages of the world, it is
impossible to say with any degree of confidence just how many of the
seven potential two-term systems actually exist. But (iv), (v) and (vi1)
would seem to be quite common. English, as we have seen, 1s but one of
several European languages that exemplifies (vii); French, and many
other languages, exemplify (iv) in the past tense at least; and, according
to what is probably the standard analysis these days, Russian exemplifies
(v), in that the so-called perfective positively represents a situation as an
event, whereas the corresponding imperfective, being the unmarked
term, simply fails to represent it as an event and therefore only nega-
tively, as it were, has anything to do with durativity. The difference
 15.6. Aspect 709
between (iv) and (v) is complicated, however, by the fact that what is the
marked member of an aspectual opposition in the experiential mode of
description may be the unmarked member in the historical mode, and
conversely (cf. 15.4). The perfective is generally taken to be the seman-
tically marked member of the perfective-vs.-imperfective opposition in
Russian and the other Slavonic languages (cf. Comrie, 1976: 112). But
its relatively high frequency of occurrence in written texts (in which
respect it is comparable with the Greek aorist and the French simple
past) would tend to suggest that, as far as the historical mode of descrip-
tion in the past tense is concerned, the perfective is unmarked (cf.
Friedrich, 1974: 30).
However that may be, the English progressive, the French imperfect,
the Russian imperfective, and comparable aspects in other languages —
comparable in that they can be accounted for in terms of (iv), (v) or (vii)
— may all be used, in the historical mode of description, to represent one
situation as a state or process within which some other situation,
represented as an event, is temporally located. Such sentences as the
following exemplify this possibility (cf. Comrie, 1976: 3):
(1) John was reading when Mary came in
(2) Jean lisait quand Marie entra (French)
(3) Ivan éital kogda Maria voSla (Russian).
Mary’s entry is represented as an event occurring at some point within
the period during which John’s reading was going on.
It is important to realize that, in languages in which the distinction
between events and processes is grammaticalized in the aspectual sys-
tem, whether a situation is represented as the one or the other does not
depend upon some absolute measure of duration. What is, both objec-
tively and as perceived by the speaker, the same situation may be repre-
sented as either a process or an event according to whether the speaker is
concerned with its internal temporal structure or not. It is for this
reason that either
(4) Ilrégna pendant trente ans
or
(5) Il régnait pendant trente ans, |
both translatable into English as
(6) He reigned for thirty years,
710 Deixts, space and time
may be used in French to describe what is objectively the same situation
(cf. Comrie, 1976: 17). Looked at from one point of view, a thirty-year
reign is just as much an event as is a sudden explosion or a flash of
lightning. It all depends upon whether the person who refers to the
situation in question 1s concerned to treat 1t in one way rather than the
other. As we saw earlier, in the historical mode of description dynamic
situations are normally represented as events and in the experiential
mode as processes; and this accounts for some of the gaps and asym- .
metries to be found in particular languages (cf. 15.4). It also accounts for
certain stylistic nuances associated with the use of one aspect rather than
another in particular languages.
The fact that there is an element of subjectivity involved in sub-
classification of situations as events, states and processes does not
invalidate the temporal distinctions upon which this subclassification is
based. If a situation is represented in one way rather than another, then
it becomes subject to the logic of temporal relations which determines
the acceptability of certain combinations of aspectual notions and the
unacceptability of others (cf. Miller & Johnson-Laird, 1976: 442ff).
Among the more important of the principles which govern the logic of
aspectually relevant temporal relations the following may be taken as
axiomatic: (i) given the undimensional directionality of time and our
punctual conceptualization of events (i.e. as second-order entities with
position, but no magnitude, in the continuum of time), two or more
events may be ordered in terms of precedence and successivity, but one
event cannot be included, wholly or partly, within another; (ii) by virtue
of our everyday assumptions about time (notwithstanding our commit-
ment to the theory of relativity), two or more events can be represented
as absolutely simultaneous; (111) since states and processes are extended
in time, but events are not, an event may be included, as a point, within
the temporal extension of a state or process; (iv) two (or more) states or
processes may be ordered, not only in terms of precedence and success-
ivity, but also in terms of co-extension or (total or partial) inclusion.
These principles, it will be noted, lend themselves very naturally to a
localistic formulation (cf. 15.7); and they go a long way towards account-
ing for the use of the major aspectual distinctions that are gram-
maticalized in the more familiar European languages.
But they do not go all the way. We must now draw upon the additional
notion of phase*, not in the rather specialized sense in which the term
‘phase’ has been employed (in contrast with ‘aspect’) by such scholars
as Joos (1964) and Palmer (1974), but in a sense which is closer to that
 15.6. Aspect 711
which it bears in everyday usage (cf. Comrie, 1976: 48). Durative situa-
tions (i.e. states and processes), unlike events, may have — and, unless
they are either omnitemporal or eternal (cf. 15.4), necessarily will have —
both a beginning and an end (at different points in time). Furthermore,
if they are temporally bounded, in that they have a beginning and an
end, they will have, between their beginning and their end, indefinitely
many temporal phases. States differ from processes (including activi-
ties), as we have seen, in that the former are homogeneous and un-
changing throughout their successive phases, whilst the latter are not.
There is a sense, however, in which processes, no less than states, can be
said to consist of successive homogeneous phases. Just as ‘‘John has
loved Mary from t; to ¢;”” entails “John loved Mary at ¢;’”’, so “John has
been running from % to t;’’ entails “John was running at t,”’, where ty
refers to any of the infinity of points or periods of time between #; and t;
(i.e. ti < te < tj). This similarity between states and processes, which
was noted by Aristotle, has been emphasized recently by a number of
scholars, including Ryle (1949), Kenny (1963) and Vendler (1967).
They have pointed out further that, within the class of processes,
there is a subclass, which Vendler calls accomplishments*, to which the
notion of completion is applicable. It makes sense to enquire with respect
to various kinds of processes “Has it stopped?’’. But the question ‘‘ Has
it finished?’ carries with it the more specific presupposition that the
process to which reference is being made 1s one that proceeds towards a
climax, or natural terminal point. For example, the process of deciding
is an accomplishment (in this rather specialized sense of the term
‘accomplishment’) which has as its terminal point the event (in this
case an act) of reaching a decision. Since deciding, unlike running or
singing (but like running a race or singing a song), is an accomplishment,
it makes sense to ask such questions as “‘How long did John take to
decide?”’. Accomplishments take time and are completed in time,
rather than merely going on and coming to an end in time. Furthermore,
whereas “‘ John is deciding”’ implies “John has not (yet) decided”, “‘John
is running” does not imply “John has not (yet) run’’, but rather “John
has run”’ (or “‘ John has been running’”’); and “ John has decided”’ implies
‘John is no longer deciding’, whereas “John has run’”’ does not imply
‘John is no longer running”’ (unless it is known that John had a certain
amount, or stint, of running to do, such that it makes sense to say “John
has finished running” or even “John has done his running”’). These
logical differences make clear the importance of taking into account
the distinction between accomplishments and non-accomplishments
712 Deixis, space and time
in theoretical discussions of the nature of aspect (cf. Dowty, 1972A).
The terms ‘perfective’ and ‘perfect’ owe their origin to the tradi-
tional view that the basic function of the aspects to which these labels
are attached in particular languages is to express completion. As we
have just seen, however, the notion of completion is inapplicable to
non-accomplishments. To this extent, therefore, the traditional terms
‘perfective’ and ‘perfect’ are potentially misleading.
Accomplishments are processes which have as their end-point an
event. Vendler (1967) also recognizes, and distinguishes from accom-
plishments, a subclass of situations which he describes as achievements*.
For example ‘arrive’ is an achievement-denoting verb. So too are
‘remember’, ‘forget’, ‘die’ and such phrases as ‘win the race’, ‘eat up
his dinner’, etc. Achievements are events, rather than processes. It
follows that an achievement-denoting verb cannot normally be used
with a period-of-time adverb; and it cannot normally be used in a dura-
tive aspect (e.g., in the English progressive). Nor does it make sense to
ask with respect to what is presented as an achievement and therefore
instantaneous, ““How long did it take?’”’. There are many apparent
exceptions to these generalizations. Most of them are to be accounted for
by the fact that particular kinds of achievements are frequently associ-
ated with particular kinds of activity whose successful performance
results in the achievement; and what appears to be an achievement-
denoting verb is to be taken, in the context, as an activity-denoting verb.
For example, John is winning might be interpreted as meaning “‘John is
performing in such a way that he is likely to win’’. What it cannot mean
is “‘John is in the process of winning’’. For winning is not a process.
Somewhat different are The train 1s now arriving at platform six and They
are now reaching the top, in which the durative verb-expressions, ‘be
arriving’ and ‘be reaching’, describe the terminal phase of processes
which will naturally issue in the achievements of arriving at platform six
and reaching the top respectively. Different again are such apparent
counter-examples as He is (always) forgetting something or other, in
which (under the most likely interpretation of this utterance) the pro-
gressive aspect has an iterative function: a series of events (in this case
achievements) is represented as if it were a process in progress.
It has now become clear that such notions as duration, completion,
momentariness, inception and termination (which were listed earlier as
being among the more common notions to which reference is made in
general discussions of aspect) are not all applicable to every kind of
situation. If each of them was grammaticalized in some distinctive way
 15.6. Aspect 713
in the aspectual system of a particular language, there would be severe
restrictions upon the combination of certain aspects with verbs having a
certain aspectual character. Some languages do have a rich set of distinct
aspects. It is not uncommon, however, for there to be no more than two or
three formally distinct aspects, the distribution of which is rather wider
than the terms that are employed to label them would tend to suggest.
It may then happen, and frequently does, that one and the same aspect
will be interpreted differently according to the character of the verb.
For example, the most typical use of the Ancient Greek aorist, in
which respect it is similar to the Russian perfective or the literary French
simple past, is to represent a dynamic situation as an event. Since static
situations cannot, in the nature of things, be represented as events, one
might suppose that stative verbs could not be used in Ancient Greek in
the aorist aspect. But some of them at least could be; and it is interesting
to note that in many such instances the aorist 1s most naturally inter-
preted as ingressive (or inchoative) — i.e. as indicating entry into the state
that the verb normally denotes. Even in English, the simple past of the
stative verb ‘know’ can be interpreted as ingressive in an appropriate
context: cf. J knew immediately what he had in mind. As static situations
cannot be represented as events, so events cannot be represented as
durative processes; and, once again, it might be expected that there
would be a prohibition upon the use of event-denoting verbs in a dura-
tive aspect. It has already been pointed out, however, that achievement-
denoting verbs may be used in the progressive aspect in English to
describe the terminal phase of a process which naturally issues in the
achievement that the verb denotes: cf. He is dying.
Many other examples could be given, from different languages, of the
occurrence of verbs in aspects that are semantically incompatible, at
first sight at least, with the aspectual character of the verbs in question.
Such examples are commonly regarded as exceptions to any general
statements that can be made about the meaning of particular aspects, or
alternatively, in atomistic rather than structural accounts, as evidence
that a particular aspect has several distinct functions. It would be
foolish to suggest that there are no arbitrary exceptional and unpredict-
able phenomena, here as elsewhere, in the use that is made of the
grammatical resources of particular languages, or, alternatively, that a
particular aspect cannot have more than one meaning. To a very con-
siderable extent, however, the more specific aspectual meaning that a
verb-form has can be seen as the product of the central, or basic, func-
tion of its aspect and its character.
714 Deixis, space and time
It may be added, in this connexion, that, although we have talked
throughout of the aspectual character of verbs, it is not only verbs, but
also nouns and adjectives, that have, or may have, a particular aspectual
character. That this is so will be obvious from our earlier discussion of
the ontological basis for the grammatical distinction between various
parts-of-speech and their syntactically relevant subclasses (cf. 11.3). The
second-order noun ‘explosion’, for example, denotes an event, whereas
‘peace’ denotes a state: hence the difference in acceptability between
The explosion occurred at three o'clock and Peace occurred at three o'clock.
Rather less obvious perhaps is the fact that, especially when we are
talking about the aspectual character of verbs, we must be careful to
make it clear that the term ‘verb’ in this context denotes expression-
classes and not merely lexeme-classes (cf. 11.1). The lexeme ‘read’, for
example, normally denotes an activity (i.e. an agent-controlled process),
whereas ‘read the book’ (containing a singular definite referring expres-
sion) will normally, though not necessarily, be construed as denoting an
accomplishment (terminating in completion): hence the difference in
acceptability between I shall have read the book by three o'clock and I
shall have read by three o'clock, in contrast with the absence of any such
difference in acceptability between I shall have been reading the book by
three o'clock and I shall have been reading by three o'clock. A more
detailed treatment of aspect and aspectual character would need to go
into the question of the interdependence that holds between aspect, on
the one hand, and number, countability and specificity of reference, on
the other (cf. Quirk et al., 1972: 148). As far as certain languages are
concerned it would also have to take into account negation (cf. Russian
Kto-to pozvonil, ‘“Someone telephoned”’ vs, Nikto ne zvonil, ‘‘ No-one
telephoned’”’: perfective vs. imperfective) and the grammatical category
of case (cf. Pettersson, 1972). Few parts of a language-system illustrate
better than its aspect-system does the validity of the structuralist slogan:
Tout se tient (“Everything hangs together’’: cf. 8.1).
But our treatment of aspect must, of necessity, leave much on one
side. We will bring it to an end (if not a conclusion!) by mentioning
briefly a number of specific points of particular importance in semantics.
The first has to do with the nature of what is traditionally called the
perfect (as distinct from the perfective) in grammatical descriptions of
particular languages (cf. Comrie, 1976: 52ff; Friedrich, 1974: 16ff).
This is commonly, and perhaps always in origin, a stative aspect, with
the more specific feature that it is used to represent the state that it
denotes as being consequent upon the completion of the process which
 15.6. Aspect 715
the verb (in other aspects than the perfect) denotes. Generally speaking,
this is true of the English perfect, as far as process-denoting verbs are
concerned: cf. [ have learnt my irregular verbs now. What is normally a
stative verb in one language is often translatable by means of the perfect
aspect of a process-denoting verb in some other language (cf. Comrie,
1976: 57). In languages that have both stative verbs and a stative
(resultative) perfect there is still a semantic difference, in principle, not
only between the perfect and the non-perfect of a particular stative verb
(cf. I know my irregular verbs now vs. I have known my irregular verbs
since I was in the fourth form or I have known my irregular verbs in the
past, but I don’t know them any more), but also between the perfect and
the non-perfect of what may be thought of, from this point of view, as
corresponding, stative and non-stative verbs. ““X knows Y”’ does not
entail ““X has learned Y’’; and, strictly speaking, ‘“‘X is dead’’ does not
entail “X has died’’, though “X has died”’ entails “X is dead’’ (pro-
vided that we exclude the possibility of resurrection or rebirth: Christ has
died, but he now lives; Fohn Smith has died several times on the operating
table, but each time he has been brought back to life and now he 1s living a
full and useful life in Surbiton). In practice, however, we tend to operate
with a less strict notion of implication, than entailment (cf. 14.3). And
one of the consequences is that, just as the perfect of ‘get’ has now come
to be used in English as if it were semantically equivalent to the non-
perfect of the stative verb ‘have’ (cf. ‘He has/has got a cold’), so what is,
from a diachronic point of view, a stative perfect may develop into a
stative non-perfect. What is lexicalized in English in the contrasts
between ‘have’ and ‘get’ and between ‘be’ and ‘have’ (if this is indeed
lexicalization: 13.1) is grammaticalized in the aspect-systems of many
languages.
It has often been pointed out that the English perfect has certain uses
which make it more like a tense than an aspect; and there is something
to be said for this view. There is a very general tendency for aspects like
the perfect, which are retrospective in the sense that they carry an
implicit reference to a point or period of time preceding that of their
primary temporal reference, to develop into simple past tenses or, more
specifically, tenses referring to the recent past (cf. Anderson, 1973b: 37).
This has happened, for example, in modern spoken French. As far as
English is concerned, however, the basically aspectual nature of the
perfect is manifest in the well-known restrictions that govern the com-
bination of the perfect with point-of-time adverbs (having definite,
rather than indefinite, reference). It is also manifest in the interesting
716 Detixis, space and time
property of the English perfect, usually interpreted in terms of the
notion of current relevance, which makes Queen Victoria has visited
Brighton, under conditions of non-emphatic stress and for anyone who
presupposes that Queen Victoria is dead, less acceptable, or less im-
mediately interpretable, than Brighton has been visited by Queen Victoria
(cf. Palmer, 1974: 53; Chomsky, 1969; McCawley, 1973).
The next point has to do with the related notions of habituality and
interaction (or frequency). The term ‘habitual’ is hallowed by usage;
but it is something of a misnomer in that much of what linguists bring
within its scope would not generally be thought of as being a matter of
habit. For example, it would be absurd to say that the apple-trees at the
bottom of the garden are in the habit of shedding their fruit in October,
as one might say of John Smith that he habitually goes to church, or
changes his shirt, on Sundays. The term ‘habitual’, then, is conven-
tionally applied by linguists to situations (and, derivatively, to the
aspects that describe such situations) to which a much broader, but
intuitively related, set of terms is applicable, including ‘customary’,
‘frequent’, ‘regular’, ‘usual’ and even ‘normal’; and ‘iterative’ or
‘frequentative’ are commonly employed in the same sense. It is often
explained: first, that the regular iteration of an event creates a series
which may be represented as a unitary durative situation with many of
the properties of a state, and that this accounts for certain uses of the
aspects in particular languages; and, second, that, in so far as “some-
times’’ shades into “‘often’’ and “often’’ may approximate asymptotic-
ally to “‘always’’, a habitual aspect is appropriately used for the expres-
sion of so-called timeless truths like “Cows eat grass’’. We have already
had occasion. to comment upon the different senses that ‘timeless’ can
bear and, in particular, upon the distinction between “‘omnitemporal”’
and ‘“‘eternal’’ and the relevance that this has to gnomic and generic
statements (cf. 15.4). It so happens that in many, and perhaps most,
languages the same aspect is, or may be, used both for habitual situa-
tions (in the broadest technical sense of ‘habitual’) and in generic state-
ments: the imperfective in Russian, the simple non-progressive in
English, the so-called aorist in Turkish, and so on. In Swahili, however,
a generic and a habitual sense of “‘Cows eat grass’’ can apparently be
distinguished in terms of a grammaticalized aspectual distinction (cf.
Ashton, 1947: 38; Closs, 1967). It is also worth noting that there
are languages in which habituality and progressivity go together (e.g.,
in the past-tense imperfect in French and Latin) and languages in
which they are grammaticalized differently (e.g., English and 'Turkish).
 15.6. Aspect 717
Mention of generic states in the previous paragraph brings us to the
final point. This has to do with what is variously referred to as a dis-
tinction between permanent and transitory, or between essential (or
absolute) and contingent, states (cf. Comrie, 1976: 103ff). Some such
distinction is grammaticalized in several languages, whether by means
of a distinction in the case of predicate nouns and adjectives (as in
Finnish, Czech and, to a relatively minor extent nowadays, Russian) or
by means of a difference in the ascriptive or locative copula that is used
in certain constructions (as in Scottish Gaelic, Irish, Spanish and
Portuguese). Brazilian Portuguese is particularly interesting in that it
has extended the ‘ser’ vs. ‘estar’ distinction, which Spanish draws with
respect to predicative adjectives (cf. Es guapa: Estd guapa, ‘‘She is
pretty”’ vs. “She looks pretty (today)’’), to a wide range of predicative
constructions including locatives (cf. Thomas, 1969: 226). Further-
more, it affords proof that, as appears to be so also in Swahili, a systema-
tic distinction must be drawn between what is contingently, though
perhaps always, the case and what is essentially, generically or (in some
non-statistical sense) normally the case.?3 If something belongs in a
certain place, even though it is not, never has been and, as far as one
knows, never will be in that place, it is possible to assert, without odd-
ness or contradiction, the following proposition: “‘X is (é) at/in Y, but
X is (estd) never at/in Y (and never has been, and never will be at/in
Y)’’. The theoretical importance of examples like this, from particular
languages, is that they demonstrate the inadequacy, not to say irrele-
vance, of universal quantification as a means of accounting for the
meaning of statements having to do with the essential properties or
normal location of things.
The treatment that has been given of aspect and of the aspectual
_ character of verbs in this section has been selective, rather than com- |
prehensive.*4 It has been biased towards the discussion of a relatively
small number of semantically relevant general points, each of which
could do with considerably more exemplification and, at times, qualifica-
tion than it has been possible, in the space available, to provide.*> We _
23 Many convincing examples to support this point are given by De Lemos
(1975), to whom I am indebted for what I have learned of the operation of the
‘ser’ vs. ‘estar’ distinction in Portuguese.
*4 Friedrich (1974) suggests that the three most basic oppositions in terms of
which a variety of aspectual systems can be analysed are durative vs. non-
durative, completive vs. non-completive and stative vs. non-stative
25 For further discussion and exemplification, in addition to the works cited in
the text, cf. Bull (1963), Heger (1963), Hirtle (1975), Johanson (1971), Klein
(1974), Schopf (1974), Verkuyl (1972), Wunderlich (1970).
 718 Deixts, space and time
turn now to the thesis of localism, which, though it can be, and has been,
defended on a much wider front, has also been put forward with
particular reference to aspect (cf. Anderson, 1973b).

15.7. Localism
The term localism* is being used here to refer to the hypothesis that
spatial expressions are more basic, grammatically and semantically, than
various kinds of non-spatial expressions (cf. Anderson, 1971, 19734).
Spatial expressions are linguistically more basic, according to the
localists, in that they serve as structural templates, as it were, for other
expressions; and the reason why this should be so, it is plausibly sug-
gested by psychologists, is that spatial organization is of central import-
ance in human cognition (cf. Miller & Johnson-Laird, 1976: 375ff).
Stronger and weaker versions of localism can be distinguished
according to the range of grammatical categories and constructions that
are brought within its scope. At its weakest, the localist hypothesis 1s
restricted to the incontrovertible fact that temporal expressions, in many
unrelated languages, are patently derived from locative expressions. For
example, ‘‘nearly every preposition or particle that is locative in English
is also temporal”’; the prepositions for, since and till, which are temporal
rather than spatial in Modern English, “derive historically from loca-
tives’’; and ‘‘those prepositions which have both spatial and temporal
use developed the temporal meaning later in all instances’”’ (Traugott,
1975). What is true of prepositions and particles is true also of very many
verbs, adverbs, adjectives and conjunctions, not only in English, but in
several languages.
The spatialization of time is so obvious and so pervasive a phenom-
enon in the grammatical and lexical structure of so many of the world’s
languages that it has been frequently noted, even by scholars who would
not think of themselves as subscribing to the hypothesis of localism. It is
more characteristic of what is clearly identifiable as localism that it
should treat the grammatical categories of tense and aspect from this
point of view. Tense, as we have seen, is a deictic category; and there is
an obvious parallel between spatial and temporal deixis. As ‘here’ and
‘there’ can be analysed as meaning “‘at this place’’ and “‘at that place”’,
respectively, so ‘now’ and ‘then’ can be analysed as meaning “‘at this
time’”’ and “‘at that time’’. Moreover, by virtue of the interdependence
of time and distance (in that what is further away takes longer to reach),
there is a direct correlation between temporal and spatial remoteness
from the deictic zero-point of the here-and-now (cf. 15.1). It is not sur-
 15.7. Localism 719
prising, therefore, that the tense-systems of various languages should
make use of what are locative, and more specifically deictic, forms or
expressions in order to draw such distinctions as past vs. non-past; and
we shall see later that what is commonly regarded as past tense (and has
been so treated in this chapter) is perhaps better analysed, in certain
languages at least, in terms of the more general notion of modal remote-
ness (cf. 17.3).
Aspectual distinctions are even more obviously spatial, or spatializ-
able, than tense-distinctions. Situations are second-order entities whose
relation to time is comparable to the relation that first-order entities
have to space. Events occur at particular points in time, whereas states
and processes endure throughout a certain period of time (cf. 15.6). Asa
point is contained within a line, so an event can be located within the
duration of a state or process: cf. Our first child was born (at a time) when
we were very hard up; He was run over (at a time) when he was crossing the
road. It is no less natural, it would appear, to say of a first-order entity
that it is in some state or process at a particular time and, in doing so, to
make use of the spatial analogy, or metaphor, of location or containment:
cf. ohn is in a state of blissful ignorance; Fohn ts in the process of cleaning
_ his teeth. There are many languages in which the aspectual notions of
progressivity or stativity (and, more especially, contingent stativity) are
expressed by means of constructions that are patently locative in origin
(cf. Anderson, 1973b; Comrie, 1976: 98ff). Indeed, the English pro-
gressive derives, diachronically, from what was once an overtly locative
construction: cf, I’ve been a-courting Mary Fane, where the dialectal, and
now rather archaic, form a-courting can be accounted for in terms of its
derivation from at (on, or in) courting.
Localists commonly treat temporal location as being less concrete
than spatial location, but more concrete than various kinds of so-called
_ abstract location (cf. Anderson, 1971: 100ff; Pottier, 1974: 56). For
example, looked at from this point of view, ‘in despair’ can be said to be
less concrete than ‘in London’. Several kinds of abstract location have
already been referred to in the course of this work. Some of them may be
mentioned again in connexion with the hypothesis of localism.
It was pointed out in chapter g that an understanding, not only of such
directional opposites as ‘up’:‘down’ and ‘front’: ‘back’, but of oppo-
sition in general, is based upon the analogical extension of distinctions
‘‘which we first learn to apply with respect to our own orientation and
the location or locomotion. of other objects in the external world”’ (cf.
9.2). The use of particular prepositions, verbs or adverbs in English |
720 Deixis, space and time
provides plenty of evidence for this view: prices are said to go up or
come down, according to whether they increase or decrease; if X is
judged to be better than Y, X is said to be above or in front of (or to
precede) Y; and so on. Much of what is commonly thought of as being
metaphorical in the use of language can be brought within the scope of
the thesis of localism.
It is less obvious perhaps that much of what is not usually thought of
as being metaphorical can also be brought within the scope of localism.
As we have seen, acquiring a certain property or entering a certain state
stands in the same semantic relation to having that property or being in
that state as arriving at (or coming into) a place does to being in that
place (cf. 9.2). To say that X has become Y or has acquired Z is tanta-
mount to saying, on the localistic interpretation, that X has passed from
the state of not being Y or not having Z to the state of being Y or having
Z. Similarly, to say that X has ceased to be Y or has lost Z is tanta-
mount to saying that X has passed from being Y or having Z to not
being Y or not having Z. The process whereby someone or something
passes from one state to another may be accounted for in terms of the
localistic notion of a journey* and, as far as the grammatical structure of
languages is concerned, in terms of the valency-schema MOVE (ENTITY,
SOURCE, GOAL), introduced in a previous chapter (cf. 12.6). Analysed in
this way, all verbs denoting a change of state may be regarded as verbs-
of-motion (cf. Miller & Johnson-Laird, 1976: 526ff).
If an entity travels from a source to a goal, it usually does so by
taking a particular route, or path*, which may or may not be referred to
(cf. such expressions as ‘through the wood’, ‘along the Embankment’,
‘by way of Beachy Head’) in descriptions of the journey (cf. Bennett,
1975: 18). Journeys are initiated by the entity’s departure from the
source and terminated by the entity’s arrival at the goal. Both departure
and arrival are achievements (in Vendler’s sense: cf. 15.6). If the source
and the goal are conceived as areas, rather than points, departure and
arrival will be achievements that have the more particular logical pro-
perty of being describable as border-crossings* (cf. Jessen, 1974). By
generalizing these localistic notions from the paradigm case of so-called
concrete locomotion (in which a first-order entity moves from one
physical location to another in some measurable interval of time) to
various kinds of abstract locomotion, the implicational relations that
hold between such pairs of prepositions as ‘‘X has learned Y’’ and
‘*X (now) knows Y”’’ or between “X has forgotten Y”’ and “X no longer
knows Y’’, on the one hand, and between “‘X has arrived in Y’’ and
 15.7. Localism 721
‘“X is (now) in Y”’ or between “X has departed from Y”’ and “X is no
longer in Y’’, on the other, can be brought together within a common
framework. By means of the achievement of learning Y (which may or
may not be the termination of some process), one crosses the border
between the ignorance and the knowledge of Y; by forgetting Y, one
crosses the border in the reverse direction (i.e. going out of, rather than
coming into, the state of knowledge). Among the many sets of expres-
sions in English whose aspectual character and implicational relations
can be accounted for within this localistic framework are ‘go to sleep’,
‘(be) asleep’, ‘wake up’, ‘(be) awake’; ‘get’, ‘have’, ‘lose’; ‘get
married’, ‘(be) married’, ‘get divorced’; ‘be born’, ‘live’, ‘die’, ‘(be)
dead’. In fact, all that was said in the previous section about aspect and
aspectual character could be reformulated in terms of the localistic
notion of travelling from one place or state to another place or state.
It will also be clear from our earlier discussion of causativity and
transitivity how these two notions relate to the hypothesis of localism
(cf. 12.5). The very term ‘transitive’, as we have seen, derives from the
paradigm instance of an agentive situation in which both the operative
schema AFFECT (AGENT, PATIENT) and the factitive schema PRODUCE
(CAUSE, EFFECT) are applicable. The hypothesis of localism has always
laid great stress upon the natural association of the valency-roles
SOURCE, CAUSE and AGENT, on the one hand, and of GOAL, EFFECT and
PATIENT, on the other — an association that is manifest in, and according
to the localists accounts for, certain apparent coincidences in the use of
such grammatical cases as the ablative* or dative*, or prepositions (or
postpositions) meaning ‘“‘from”’ or “to”’, in many unrelated languages
(cf. Anderson, 1971; Anderson & Dubois-Charlier, 1975).
Causes, according to the view of causality that we have adopted here,
are second-order entities; as such, they may be conceived, localistically,
as the sources of their effects. Similarly, at a higher level of abstraction,
the antecedent proposition, p, may be thought of as the source from
which the consequent proposition, g, proceeds in a complex conditional
proposition: i.e. “if p, then q’’ can be interpreted, localistically, as ‘gq
comes from p’’. This would suggest that the semantic and gram-
matical parallelism that holds between causal, conditional and temporal
clauses can be accounted for, not only in terms of the successivity in
time of causally related situations, but, more particularly, in terms of the
localistic notion of a journey (cf. 12.5).
| Even instrumental adverbials and adverbials of manner, which, like
locative, temporal and causal adverbials are characteristically adjuncts,
 722 Detixis, space and time
rather than nuclear constituents, in simple sentences, may be brought
together, from a localistic point of view, and analysed in terms of the
notion of a path (cf. Anderson, 1971: 171). It is noticeable, in this con-
nexion, that the form how in English (in both its relative and its interro-
gative function: cf. ‘This is how he did it’ and ‘How did he do it?’)
subsumes both instrumental adverbials and adverbs of manner (cf.
‘with a knife’ and ‘carelessly’); and in each instance it can be para-
phrased with a phrase containing the word ‘way’ (cf. ‘This is the way in
which he did it’ and ‘In what way did he do it?’). It is presumably for
this reason that Pottier (1974: 197) classifies the French adverb ‘ainsi’
(“‘thus”’’) as the notional (i.e. abstract or third-order) correlate of the
spatial deictic ‘ici’ (“here’’) and the temporal deictic ‘maintenant’
(““now’’). As we saw earlier, with respect to what is commonly, but
perhaps wrongly, thought of as the universal circumstantial role of
instrument, there may be several alternative templates for the gram-
matical and lexical categorization of what is objectively the same situa-
tion (cf. 12.6). At least one of these templates, however, would seem to
be spatial in origin.
Two kinds of states that have been discussed more frequently, per-
haps, than others in terms of the notion of abstract location are those of
possession and existence. This is because there are very many unrelated
languages throughout the world in which overtly locative constructions
are used in sentences that would be translated into English as ‘John has
a book’ (or ‘The book is John’s’) and ‘There are unicorns’, or ‘ Uni-
corns exist’ (cf. Allen, 1964; Asher, 1968; Boadi, 1971; Christie, 1970; __
Li, 1972; Lyons, 1967, 1975). In everyday usage the term ‘possession’
is more or less equivalent to ‘ownership’ (though jurists may draw a
sharp distinction between the two terms): whatever X is said to possess
may be described as his property. In traditional grammatical usage
‘ possession’ and ‘possessive’ are construed much more broadly. Indeed,
it can be argued that they are highly misleading: it is only a minority of
what are traditionally called possessive constructions that have anything
. to do with property or possession; and there is no reason to believe that
this minority constitutes a particularly important, or basic, subclass of
the total class. However that may be, there can be no doubt that the
localist interpretation of so-called possessive constructions is eminently
plausible on both semantic and syntactic grounds. The question Where
1s the book? can be answered, equally well, with either J?’s on the table or
John has it; and there is no reason to treat the verb ‘have’, here and
_. elsewhere, as anything other than a transformationally inserted variant
 15.7. Localism 723
of the locative copula. We saw earlier that there is nothing paradoxical,
untraditional though it may be, about the postulation of underlying
locative subjects (cf. 12.3). We can say that ‘John’ (or ‘at John’) is the
underlying locative subject in ‘John has a book’, as ‘on the table’ is the
underlying locative subject in ‘There is a book on the table’; in ‘John
has the book’, on the other hand, the underlying personal locative ‘at
John’ is perhaps best regarded as being predicative (cf. ‘The book is on
the table’).
Existential constructions are, if anything, even more obviously of
locative origin than so-called possessive constructions are. It can be
argued, in fact, that existence is but the limiting case of location in an
abstract, deictically neutral, space (cf. Lyons, 1975); and that it is the
deictically neutral, so-called existential, sense of the adverb ‘there’ that
is adjectivalized (and amalgamated with a pronominal element) in the
definite article (cf. 15.2). But whether we take this view of the matter or
not, the locative basis of existential constructions, in many, if not all,
languages, is hardly open to doubt and has been quite widely recognized.
Once the locative source of existential constructions is postulated, the
way lies open for a localistic interpretation, not only of such existential-
causatives* (traditionally said to manifest an object-of-result) as ‘God
created Adam’, ‘John painted a picture’ (cf. 12.5), but also of the
linguistic representation of such central logical notions as quantification,
negation, knowledge and truth. Since we are not concerned, in this work,
to push the hypothesis of localism to the limits of its coverage or to
develop it in any detail, we will leave quantification and negation on one
side (cf. Anderson, 1975) and confine ourselves to a brief, and some-
what speculative, localistic account of knowledge and truth.
The concepts of knowledge, truth and existence are interconnected in
that the proposition “‘X knows p”’ cannot be true unless the proposition
p itself is true (cf. 17.2); and p cannot be true, of some world (or world-
state), unless the situation that it describes actually exists in the world
(or world-state) in question (cf. 6.5). Now propositions, which we have
classified as third-order entities, may be treated either as purely abstract,
non-psychological, entities or, alternatively, as objects of knowledge and
belief. Truth is the third-order correlate of what for first-order entities
is existence in space; and a statement like That ts so (where ‘that’ refers
to a proposition) is structurally comparable with a statement like
X exists (where X refers to a first-order entity). This, in a nutshell, 1s
one version of a localistic theory of truth; and (although we shall not
develop this point in the present work) it will be seen later that this con-
 724. Deixis, space and time
ception of truth fits in with the logical analysis of utterances with which
we shall be operating, according to which every statement contains an
‘“it-1s-so’’ component (cf. 16.5).
If propositions are treated as psychological entities, rather than as
purely abstract third-order entities, then it is natural to treat as their
location the persons (or the minds, or brains, of the persons) who have
what philosophers might describe as a propositional attitude (know-
ledge, belief, etc.) with respect to them. It is obvious that the process of
communicating propositional information is readily describable, as is
the process of transferring possession (cf. Miller & Johnson-Laird,
1976: 558ff), in terms of the localistic notion of a journey: if X com-
municates p to Y, this implies that p travels, in some sense, from X to Y.
It does not follow, of course, that Y will believe p; and, even if Y does
believe p, he cannot be said to know p, unless p is true. At most, he can
be said to have been acquainted with p. It may be suggested, therefore,
that ‘“‘p is at X”’ (where X is a person) is the underlying locative struc-
ture that is common to “X knows p’’, “X believes p’’, ‘“X has p in
mind’’, etc. There is much in the structure of particular languages,
however, to suggest that ‘““X knows p”’ is comparable with “X has Y”’
and should therefore be regarded as the most typical member of the
class of propositions subsumable under “‘p is at X”’. We will not develop
this admittedly rather speculative point in the present work. It may be
mentioned, however, that, as the localistic notion of truth fits in with the
tripartite analysis of utterances that is to be adopted later, so the localistic
analysis of propositional knowledge that has been adumbrated here can
be extended to cover many of the constructions in which modality (both
epistemic and deontic: cf. 17.2, 17.5) is objectified, or propositionalized,
in natural languages.
At the beginning of this section the point was made that the hypothesis
of localism can be maintained in either a stronger or a weaker form. It is
only in a relatively strong version of localism that the linguistic expres-
sion of truth and modality, not to mention negation and quantification,
would be brought within its scope; and nothing that is said about these
notions elsewhere in this book depends upon a localistic interpretation of
them. There is much else that has been referred to in this section —
notably aspect, the grammatical category of case, and existential and }
possessive constructions — for which a localistic analysis would be far
more widely accepted.
 16
Mood and illocutionary force

16.1. Speech-acts
Throughout much of this book so far we have been mainly concerned
with the descriptive function of language: i.e. with the way language is
used to make statements*.! But language also serves as an instrument for
the transmission of other kinds of information. Not all the utterances we
produce are statements; and statements, as well as questions, commands,
requests, exclamations, etc., will contain a certain amount of non-
descriptive information, which may be characterized, broadly, as
expressive* (or indexical*) and social* (cf. 2.4). Furthermore, the trans-
mission of descriptive information is not usually an end in itself. When
we communicate some proposition to another person, we do so, nor-
mally, because we wish to influence in some way his beliefs, his attitudes
or his behaviour.
To produce an utterance is to engage in a certain kind of social inter-
action. This is a fact that, until recently, logicians and philosophers of
language have tended to overlook, though it has often been stressed by
linguists, psychologists, sociologists and anthropologists. One of the
most attractive features of the theory of speech-acts, which was intro-
duced into the philosophy of language by J. L. Austin, is that it gives
explicit recognition to the social or interpersonal dimension of language-
behaviour and provides a general framework, as we shall see, for the
discussion of the syntactic and semantic distinctions that linguists have
traditionally described in terms of mood* and modality*.
Austin’s theory of speech-acts* was developed over a number of
years; and in its final version (in so far as Austin himself succeeded in
producing a final, or definitive, version before his death) it is deliberately

1'The term ‘statement’ is commonly used by logicians in a rather different

sense (cf. Lemmon, 1966). My usage is intended to be closer to what I take
to be its everyday sense. In particular, it should be noted that statements are
a subclass of utterances and that they may be regarded either as acts or signals
(cf. 1.6).
726 Mood and tllocutionary force
modified and extended in the course of its presentation (Austin, 1962).
The term ‘speech-act’ is in fact rarely used by Austin; and, when he
does use it, it is not entirely clear how much of what is done, or per-
formed, in the production of the utterance he intends it to cover. We
will not go into this question.
Since the term ‘speech-act’ is now widely employed in work which
derives from Austin, and notably in the title of an influential book by
Searle (1969), we will use it in the present discussion. It should be
pointed out, however, that it is an unfortunate and potentially mis-
leading term. First of all, it does not refer to the act of speaking as such
(i.e. to the production of an actual spoken utterance), but, as we shall
see, to something more abstract.” Secondly, ‘speech-act’, in what we
may call its Austinian (or post-Austinian) sense, 1s not restricted to
communication by means of spoken language. Indeed, it is arguable that
there are certain non-linguistic communicative acts that would satisfy
Austin’s definition of speech-acts. For example, if X summons Y with
a manual gesture he may be said to have performed a particular speech-
act in the Austinian sense. Austin, it is true, developed his theory of
speech-acts with particular reference to language; and he would certainly
have accepted the principle of the priority of the phonic medium (cf.
3.3). Neither he nor his followers, however, would seem to be committed
to the view that gestures and other kinds of signals can be described
within the framework of the theory of speech-acts only in so far as they
are equivalent to, or parasitic upon, language-utterances. But this too is
a question that will not be discussed further. We will henceforth confine
our attention to language-utterances.
Austin started by drawing a distinction between constative* and
performative* utterances. Constative utterances are statements: their
function is to describe some event, process or state-of-affairs, and they
(or the propositions expressed) have the property of being either true or
false. Performative utterances, by contrast, have no truth-value: they
are used to do something, rather than to say that something is or is not
the case. For example, the sentences ‘I name this ship ‘Liberté’’ or ‘I
2 The term ‘speech-act’ (translating the German ‘Sprechakt’ of Buhler, 1934)
has often been used by linguists, and is occasionally still used, in the more
natural sense of ‘‘act of speech’’,
3 ‘The view taken here, and throughout, is that, in technical usage, the terms
‘true’ and ‘false’ apply primarily to propositions and only secondarily to the
statements expressing, or containing, such propositions. Pre-theoretically, it
is not clear what the basic senses of ‘true’ and ‘false’ are (cf. p. 734, n. 5
below).
 16.1. Speech-acts 727
advise you to stop smoking’ would be uttered, characteristically, to per-
form particular kinds of acts which, as Austin pointed out, could hardly
be performed in any other way. Roughly speaking, we can say that the
distinction between constative and performative utterances, as it was
originally drawn, rested upon the distinction between saying something
and doing something by means of language (where the expression
‘saying something’ means “asserting that something is or 1s not so’’).
It was an important part of Austin’s purpose to emphasize (1) that state-
ments, or constative utterances, constitute only one class of meaningful
utterances and (11) that performative utterances should also be brought
within the scope of logical and philosophical investigation.
Austin was in this respect challenging the restrictive view of meaning
held by the logical positivists (cf. 6.1), according to whom the only fully
meaningful utterances were empirically verifiable statements, all other
utterances being classified as emotive*. This catch-all sense of ‘emotive’
was commonly used in the hey-day of logical positivism to criticize as
meaningless what purported to be descriptive statements in such fields
of discourse as metaphysics (cf. Ayer, 1936), and it became the founda-
tion-stone of the so-called emotive theory of ethics (cf. Stevenson, 1944).
It was imported into literary criticism and stylistics by such influential
writers as I. A. Richards (1925).
Wittgenstein, who had himself been closely associated with the
founders of logical positivism, later came to renounce the simplistic
distinction of the descriptive and the emotive functions of language,
emphasizing instead the functional diversity of language-utterances.
Using language, he said, is like playing games. whose rules are learned
and made manifest by actually playing the game. One acquires one’s
command of a language, not by first learning a single set of prescriptive
rules which govern its use on all occasions, but by engaging in a variety
of different language-games, each of which is restricted to a specific kind
of social context and is determined by particular social conventions.
Describing how the world is (or might be) is but one of indefinitely
many language-games that we play as members of a particular society;
and it should not be accorded preferential status by philosophers and
logicians. Every language-game has its own internal logic (or grammar,
as Wittgenstein would have said in a somewhat extended sense of
‘grammar’) and deserves equal consideration. It is within this general
framework that Wittgenstein enunciated his famous, and controversial,
principle that the meaning of a word is revealed in its use. Without going
into the details of the relationship between Wittgenstein’s (1953) doc-
728 Mood and illocutionary force
trine of language-games and Austin’s theory of speech-acts, it will suffice
here to point out that they are similar in that they both emphasize the
importance of relating the functions of language to the social contexts in
which languages operate and insist that, not only descriptive, but also
non-descriptive utterances should be of concern to the philosopher.
Austin emphasized the fact that many declarative sentences (e.g., ‘I
name this ship ‘ Liberté’’) are employed, in certain standard contexts,
not to describe a state-of-affairs which obtains independently of the
utterance, but as a constitutive part of some action that 1s being per-
formed by the speaker. Logical positivists had wished to classify as
emotive and unverifiable pseudo-statements such utterances as It is
wrong to kill or God is good. Whether they were right or not in denying
to such utterances the status of descriptive statements, they had failed to
recognize that there is a whole range of declarative sentences which,
though they might not be satisfactorily described as emotive, were even
more obviously not being used to make statements.
Austin drew a further distinction within performative utterances
between what he called primary performatives and explicit performa-
tives. For example, we can perform the act of promising in English in
either of two ways: by saying (cf. Austin, 1962: 69)
(1) Ill be there at two o'clock
(2) I promise to be there at two o'clock.
The first of these utterances, (1), is a primary performative; the second,
(2), which contains a form of the performative verb ‘promise’, is an
explicit performative. Two points should be emphasized in connexion
with this distinction of primary and explicit performatives.
The first point is that the fact that a primary and an explicit performa-
tive may be used to perform the same speech-act does not imply that the
sentences in question have the same meaning. An explicit performative
is typically more specific in meaning than a primary performative. If
someone says, in the appropriate circumstances, I promise to be there at
two o'clock, he can hardly deny subsequently that he has made a promise.
But if he says J’/l be there at two o'clock, unless the context is such as to
exclude the possibility of any other interpretation, he might reasonably
claim that he was merely predicting, rather than promising, that he
would be there at two o’clock; and the fulfilment of his prediction might
have been conditional upon factors over which he had no control.
Secondly, it is a characteristic feature of explicit performatives in
English that they have the form of declarative sentences with a first-
 16.1. Speech-acts 729
person subject and that the performative verb is in the simple present
tense. But this is neither a necessary nor a sufficient condition of their
being explicit performatives. On the one hand, we will find explicit
performatives, such as Passengers are requested to cross the ratlway line by
the footbridge, with the performative verb in the passive. This is com-
monly the case with requests or commands that are issued by some
impersonal or corporate authority. On the other hand, we will find
performative verbs, like ‘promise’, being used in the simple present
tense with a first person subject in constative utterances. In certain
circumstances I promise to be there is interpretable as a statement. As
Austin points out, we can usually settle the question, in particular
instances, by asking ourselves whether it would be possible to insert the
word ‘hereby’. I hereby promise to be there is indubitably an explicitly
performative utterance. Generally speaking, however, in default of
‘hereby’ or something equivalent to it in the utterance or in the context
in which the utterance is produced, explicit performatives do not carry
any definitive indication, in their verbal component at least, of their
status. As far as their grammatical structure is concerned, they have the
form of declarative sentences; and this gives them, as Austin says, “a
thoroughly constative look”’.
So far we have discussed the theory of speech-acts on the basis of the
distinction between saying something and doing something with lan-
guage. But Austin soon came to realize that this is an untenable distinc-
tion. Saying (or asserting) that something is so is itself a kind of doing.
Constative utterances, or statements, are therefore just one kind of
performatives; and they too may be primary or explicit. T’o the primary
statement
(3) The cat 1s on the mat
there corresponds the explicitly performative statement
(4) I tell you that the cat is on the mat
which contains the performative verb ‘tell’. Similarly, to the primary
question
(5) Are all the guests French?
there corresponds the explicitly performative question
(6) I ask you whether all the guests are French;
and to the primary command (if, on some particular occasion of its
utterance, it 1s in fact a command)
730 Mood and illocutionary force
(7) Close the window!
there corresponds the explicitly performative command
(8) I order you to close the window.
In all these cases, it should be noted, the explicit performative has the
grammatical form of a declarative sentence; and it is more specific in
meaning than the corresponding primary performative.
In his further development of the theory of speech-acts, Austin drew
a threefold distinction between locutionary*, illocutionary* and per-
locutionary* acts, as follows.
(i) A locutionary act is an act of saying: the production of a meaningful
utterance (“the utterance of certain noises, the utterance of certain
words in a certain construction, and the utterance of them with a certain
‘“‘meaning”’ in the favourite philosophical sense of that word, 1.e. with a
certain sense and a certain reference’’. Austin, 1962: 94).
(ii) An illocutionary act is an act performed in saying something:
making a statement or promise, issuing a command or request, asking a
question, christening a ship, etc.
(iii) A perlocutionary act is an act performed by means of saying
something: getting someone to believe that something 1s so, persuading
someone to do something, moving someone to anger, consoling someone
in his distress, etc.
It would seem to follow from Austin’s definition of the locutionary act
that formally identical tokens of the same utterance-type (cf. 1.4) whose
constituent expressions differ in either sense or reference are by virtue
of this fact products of a different locutionary act; and, if this is so, the
whole basis of the distinction between locutionary and illocutionary acts
appears to collapse (cf. Hare, 1971: 100-14). This distinction, as Austin
drew it, has been the subject of considerable philosophical controversy,
which we need not go into here. We will operate instead with the dis-
tinction drawn in chapter 1 between utterance-signals (which may be
grouped as tokens of the same type on the basis of their phonological,
grammatical and lexical structure, independently of the sense and
reference of their constituents) and utterance-acts (to which the notion
of type-token identity does not apply).
Austin’s distinction between illocutionary acts and perlocutionary acts
is crucial; and it is one that has frequently been missed or blurred in
theoretical semantics. In our deliberately restricted treatment of the
notion of communication in chapter 2 and our subsequent discussion of
logical semantics in chapter 6, we took the view that the transmission of
 16.1. Speech-acts 731
propositional information from a sender X to a receiver Y had as its
purpose X’s making Y aware of some fact of which he was not previously
aware: i.e. of putting some proposition into Y’s store of knowledge. As
far as it goes, this analysis 1s satisfactory enough as an account of the
descriptive function of language. But, apart from its failure to cover
more than a small part of what we mean by the communication of in-
formation, it fails to bring out the fact that, when we make a statement,
we may do so for a variety of reasons and not simply, or even necessarily,
to augment or alter the addressee’s beliefs. We have already seen that
tokens of the same utterance-type may be used to perform a variety of
illocutionary acts: to make statements, utter threats, issue commands,
etc. What we have not so far introduced into our discussion of the
meaning of utterances (and it should be stressed that we are here talking
of utterances, nct system-sentences: cf. 1.6, 14.6) is the distinction
between their illocutionary force* and their (actual or intended) per-
locutionary effect*; and, as Austin recognized, these are independent
components of the complex act of utterance, although they are no doubt
connected in certain standard situations. By the illocutionary force of an
utterance is to be understood its status as a promise, a threat, a request, a
statement, an exhortation, etc. By its perlocutionary effect is meant its
effect upon the beliefs, attitudes or behaviour of the addressee and, in
certain cases, its consequential effect upon some state-of-affairs within
the control of the addressee. For example, if X says to Y Open the door!
investing his utterance-signal with the illocutionary force of a request or
command (and associating with it the appropriate prosodic and para-
linguistic features: cf. 3.2), he may succeed in getting Y to open the
door. Our use of the word ‘succeed’ presupposes of course that it is X’s
intention to bring about this particular effect. We must be careful
therefore to distinguish between the intended and the actual per-
locutionary effect of an utterance. It is the intended perlocutionary effect
that has generally been confused with illocutionary force.
It is especially important to distinguish between the intended or
actual perlocutionary effect of an utterance and what Austin called
illocutionary uptake*: the addressee’s recognition that a particular
illocutionary act has been performed. [llocutionary uptake is necessary,
though not a sufficient, condition of the receiver’s successful perform-
ance of the cognitive act we call understanding an utterance. It is not a
sufficient condition, because the receiver’s knowledge of the phono-
logical, grammatical and lexical structure of the language is also in-
volved. There is a sense in which understanding an utterance can be
732 Mood and illocutionary force
described as a cognitive response on the part of the receiver. It is,
however, a response which is distinct from the actual or intended per-
locutionary effect; and to call it a response would tend perhaps to blur
this distinction (cf. Searle, 1969: 42ff). If X tells Y that something is so,
he may do so because he wants Y to believe that it is so; but Y’s under-
standing of the utterance is independent of his recognition of this
intended perlocutionary effect. Y can quite legitimately say, afterwards,
that X had made a statement and that he does not know whether X in-
tended him to believe it or what effect X intended to achieve. In other
words, Y can know what X. meant without knowing, or needing to know,
why X said what he said.4
One of the questions that has been hotly debated by philosophers in
connexion with the theory of speech-acts is whether, as Austin appeared
to hold, convention is necessarily involved in the determination of the
illocutionary force of an utterance. Strawson (1964a), following Grice
(1957), has argued that such basic illocutionary acts as making state-
ments, asking questions and issuing commands are essentially non-
conventional, in the sense that they can be explicated solely in terms of
so-called natural responses involving beliefs and the recognition of
communicative intention. According to Searle, “‘some acts at least, e.g.
statements and promises . . . can only be performed within systems of
‘‘constitutive’’ rules and the particular linguistic conventions we have
in particular natural languages are simply conventional realizations of
these underlying constitutive rules’’ (1971: 9). He concedes, however,
that this is “‘one of the most important unresolved controversies in
contemporary philosophy of language’’. This being so, we will make no
attempt to pre-judge it one way or the other here. Both parties to the
controversy subscribe to some form of Grice’s (1957) analysis of mean-
ing as being crucially dependent upon the sender’s intention that the
addressee shall recognize his intention to perform a particular illo-
cutionary act.
What we may call the sender’s communicative intention turns out to

4 We are talking here of arbitrarily selected particular occasions. It is arguable

that there is nonetheless an essential connexion between knowing what X’s
utterance means and knowing what someone would normally mean by pro-
ducing a token of such-and-such an utterance-type under standard con-
ditions (cf. Grice, 195'7, 1968). In much the same way, it can be argued that
although deceit and prevarication are possible (and indeed quite common) in
everyday language-behaviour, communication depends logically upon there
being established in the community a convention of truthfulness (cf. Lewis,
1969).
 16.1. Speech-acts 733
be, on further analysis, rather more complex than it appears at first
sight; and we need not go into the details. The important point is that
meaning and understanding are correlative, and both involve inten-
tionality: the meaning of an utterance necessarily involves the sender’s
communicative intention and understanding an utterance necessarily
involves the receiver’s recognition of the sender’s communicative in-
tention. We can abstract from communicative intention, or illocutionary
force, in our discussion of the meaning of a sentence, or of the expres-
sions that occur in a sentence. At the same time, we must recognize that
in all languages sentences are systematically associated, in terms of their
phonological, grammatical and lexical structure, with the illocutionary
acts that may be performed in uttering them. There is no one-to-one
correspondence between grammatical structure, in particular, and
illocutionary force; but we cannot employ just any kind of sentence in
order to perform any kind of illocutionary act. Furthermore, we learn
the sense and denotation of lexemes and the meaning of grammatical
categories and constructions in actual utterances; and it is this fact that
relates several of the distinguishable senses of ‘meaning’ mentioned in
the first chapter (1.1).
Austin pointed out in his discussion of speech-acts that there are
various felicity conditions* which an illocutionary act must fulfil if it is
to be successful and non-defective. The felicity conditions will be
different for different kinds of illocutionary act, but they can be grouped
under three main heads, which, following Searle (1969: 57-61), we
may refer to as preparatory (or prerequisite) conditions, sincerity con-
ditions and essential conditions, respectively. Violation of each of these
sets of conditions makes the utterance infelicitous in a particular
way.
(i) Preparatory conditions. The person performing the act must have
the right or authority to do so; and, in certain cases, the occasion of his
utterance must be appropriate to the illocutionary act in question. For
example, one cannot christen a ship simply by uttering the sentence ‘I
name this ship ‘Liberté’’ regardless of the situation of utterance. The
person who performs the act of christening must be authorized to do so
and, presumably, he must produce the utterance in the course of a more
or less well established ceremony. If these preparatory conditions are
not fulfilled the act will be null and void: as Austin puts it, the act will
misfire. It is important to realize that it is not only ritualistic and cere-
monial utterances that are governed by preparatory conditions. Accord-
ing to Austin, we cannot make a valid statement unless we have evidence
734 Mood and tllocutionary force
for our assertion and have reason to believe that the addressee is unaware
of what we assert to be the case.
(ii) Sincerity conditions. If the person performing the act does so
insincerely (i.e. without the appropriate beliefs or feelings) his illo-
cutionary act will not be nullified, but he will be guilty of what Austin
calls an abuse. For example, if X makes a statement which he knows or
believes to be untrue, he thereby perpetrates the abuse that we refer to
as lying or prevarication; and, if he does so on oath in a court of law, he
commits perjury. Similarly, if X thanks Y for some gift or service, he
must, if he is sincere, feel gratitude or appreciation towards Y. There
are of course occasions when sincerity is overridden by politeness; and
these occasions are presumably determined by social convention, even
if the more basic sincerity conditions are not. We are not always expected
to tell the truth or give expression to our true feelings.®
(iii) Essential conditions. The person performing the act is com-
mitted by the illocutionary force of his utterance to certain beliefs or
intentions; and, if he thereafter produces an utterance which is in-
consistent with these beliefs or conducts himself in a way that is
incompatible with the intentions to which he is committed, he may be
judged guilty of a breach of commitment. For example, in making a
statement we commit ourselves to the truth of the proposition expressed
by the sentence uttered in making the statement. Commitment*, in this
sense, does not mean that we must believe that what we say is true; still
less does it mean that the asserted proposition is in fact true. Commit-
ment is independent of sincerity and truth; it is a matter of appropriate
behaviour. The nature of our commitment is revealed in the generally
accepted illogicality of asserting simultaneously two contradictory pro-
positions: e.g., ‘All of John’s children are bald”’ and ‘‘Some of John’s
children are not bald’’. According to Austin, violation of the law of the
excluded middle in an argument is a breach of commitment of essen-
tially the same kind as breaking a promise.
5 It is worth noting: (i) how readily the word ‘true’ is used in collocation with
such words as ‘feeling’, ‘attitude’ or ‘sentiment’; and (ii) that, in everyday
usage, the expression ‘tell the truth’ carries very strong implications of
sincerity. Telling the truth is not simply a matter of saying what is true — i.e.
of uttering a proposition which, regardless of one’s own beliefs, happens to be
true. One cannot tell the truth by insincerely and accidentally saying what is
true, but one can insincerely or accidentally say what is true without telling
the truth. Arguably, the sense of ‘true’ in which to speak truly is to give
expression to one’s true feelings is as basic a sense of ‘true’, pre-theoretically,
as is the sense in which to speak truly is to utter a proposition which happens
to correspond to some state-of-affairs.
 16.1. Speech-acts 735
We need not go further into the question of felicity conditions. The
important point to notice (and it is philosophically controversial) is that,
under this analysis, such basic illocutionary acts as making statements,
asking questions and issuing commands are made subject to the same
kinds of conditions as are the more obviously performative utterances
which Austin originally contrasted with constatives. What Austin offers
then is, in principle, a unified theory of the meaning of utterances within
the framework of a general theory of social activity. His theory of mean-
ing, like the later Wittgenstein’s, can be described as a contextual theory
of meaning, in the sense in which the theories of Firth and Malinowski
are contextual theories (cf. 14.4); and it has the advantage that it throws
a bridge over the chasm that has long existed between philosophical and
sociological or anthropological approaches to semantics. It is perhaps
fair to say also that Austin’s theory of speech-acts preserves all that 1s
valid and useful in behaviourist semantics (cf. 5.4). It is not of course a
behaviouristic theory in the strict sense, but it is not incompatible with
an extended version of behaviourism; and Austin’s distinction of illo-
cutionary force and perlocutionary effect, on the one hand, and his
analysis of the different sets of felicity conditions, on the other, points
the way to the kind of extension that is required in order to remedy the
more obvious inadequacies of behaviourist theories of semantics of the
kind we looked at earlier.®
Over and above the three sets of felicity conditions listed and exem-
plified above, illocutionary acts are governed and determined by what
we may call a general condition of meaningfulness; and it is here that
Grice’s analysis of meaning in terms of intention (which again is not
incompatible with an extended version of behaviourism) comes into
play. As Searle puts it: “The speaker intends to produce a certain
illocutionary effect by means of getting the hearer to recognize his
intention to produce that effect, and he also intends the recognition to
be achieved in virtue of the fact that the meaning of the item he utters
conventionally associates it with producing that effect’’ (1969: 60-1).
In making a promise, for example, the speaker assumes that “the seman-
tic rules (which determine the meaning of the expressions uttered) are
such that the utterance counts as the undertaking of an obligation’’.
Two further points should be mentioned before we move on from this
general discussion of speech-acts to consider how the notion of illocu-
tionary force relates to mood and modality. The first is that what we ,
6 Bennett’s (1976) broadly behaviouristic account of communication by means
of language is of considerable interest in this connexion.
736 Mood and illocutionary force
have referred to as a single illocutionary act, such as making a statement
or a promise, may involve, and typically will involve, several component
speech-acts. Suppose we make a statement in order to ascribe to a
particular entity a certain property which is denoted by a predicative
expression. Our reference to the entity in question by means of a par-
ticular referring expression is itself a particular kind of act (in Austin’s
sense of the term ‘act’). So too is predication, or the ascription to the
entity of a certain property (cf. 6.3). We may think of the propositional
content of a sentence (i.e. the proposition expressed by a sentence when
it is uttered to make a statement) as being an abstraction from a particu-
lar propositional act, and the propositional act as being composed of the
two component acts of reference and predication (cf. Searle, 1969: 22-6).
But the illocutionary force of a statement is not exhausted by its pro-
positional content: it must be associated with the illocutionary act of
assertion. And the same propositional content may be associated, as we
shall see in the next section, with a variety of different illocutionary acts
to yield such distinct speech-acts as questions, commands, requests,
etc.
The second point is that Austin’s theory of speech-acts necessarily
raises the question whether there is any upper or lower limit to the
number of illocutionary acts that need to be recognized in the semantic
analysis of natural languages. There are some hundreds of performative
verbs in English; and it is clearly unsatisfactory to have a theory which
leaves all the acts denoted by these verbs distinct and unrelated. Can
they be grouped into a relatively small number of basic classes? And, if
so, how? There are at least three ways of doing this.
(i) By studying the relationship between primary and explicit per-
formatives, on the assumption that different kinds of primary performa-
tives (whose meaning, as we have seen, is typically more general)
distinguish certain basic categories of illocutionary force. The meaning
of such performative verbs as ‘promise’, ‘predict’, ‘swear’ and
‘threaten’ might, on this assumption, derive from the encapsulated
syntagmatic modification of a more basic underlying performative verb
which particular languages may or may not lexicalize (cf. 8.5).
(ii) By studying the terms used to report instances of particular kinds
of utterances. For example, the fact that X’s utterance I'll be there at
two o'clock might be reported in English as X promised to be there at two
o’clock is an indication that sentences such as ‘I’ll be there at two
o'clock’ may be used to make promises. It is worth noting, in this
connexion, that, although, in English, the verbs that are used to describe
 16.1. Speech-acts 737
particular illocutionary acts are for the most part the same as the verbs
that are used to perform these same illocutionary acts, this is obviously a
contingent fact about the lexical structure of particular languages. It 1s
in principle possible (though it would be uneconomical) for a language
to have two distinct sets of verbs, one set for performing and the other
for describing illocutionary acts. Indeed, it would be possible for a
language to make no use of performative verbs at all, but to use instead a
set of performative particles or prosodic and paralinguistic features in
order to distinguish particular kinds of explicitly performative utter-
ances.’
(iii) By studying the felicity conditions associated with particular
kinds of speech-acts and constructing a typology of speech-acts in terms
of shared subsets of preparatory, sincerity and essential conditions. It has
been argued that such intuitively apparent relationships as hold between
promising and threatening, and between advising and warning, can be
explicated in this way (cf. Searle, 1969). But the analysis of felicity
conditions for a wide and representative sample of speech-acts is a task
that so far has barely been started and, until it is accomplished, it is hard
to say what the result will be.®
Nothing has been said so far, it should be noted, about the univer-
sality of particular kinds of speech-acts. It is perhaps reasonable to
assume that what Strawson and others have called basic illocutionary
acts — notably making statements, asking questions and issuing com-
mands or requests — are universal, in the sense that they are acts that are
performed in all human societies; and this might be so regardless of
whether they are necessarily grounded in convention or not. But there
are certain speech-acts that would seem to be dependent upon the legal
or moral concepts institutionalized in particular societies. Austin’s
example of naming a ship is presumably one such act. Others are the
acts of swearing on oath in a court of law, baptizing a child into the
Christian faith or conferring a university degree. Such acts are obviously
both conventional and specific to particular cultures. In what follows,
we shall be concerned solely with the more basic speech-acts which may
be assumed to be universal. It should not be forgotten, however, that, in
any particular society, these more basic speech-acts are integrated with,

7 For example in Ancient Greek the form émén (‘‘truly’’, ‘‘verily’’) was regu-
larly used, with or without the accompanying performative verb-form
hémnumi (‘‘I swear’’), in oaths.
§ Austin began this task himself and introduced a number of general classes of
speech-acts (cf. Fraser, 1974).
738 Mood and tllocutionary force
and governed by, felicity conditions of the same kind as those which
govern other forms of behaviour and social interaction in that society.
At this point, it is convenient to introduce the notion of what have
been called parenthetical verbs*. Verbs such as ‘suppose’, ‘believe’ and
‘think’ may be used parenthetically in the first person of the simple
present tense “to modify or weaken the claim to truth that would be
implied by a simple assertion’? (Urmson, 1952). Their function, as
described by Urmson, is illustrated by sentences like
(9) She’s in the dining-room, I think;
and it is comparable, if not identical, with what was referred to in
chapter 3 as the prosodic and paralinguistic modulation* of utterances
(cf. 3.1).
The similarity between performative and parenthetical verbs will be
obvious. In fact, it would seem to be desirable to widen the definition of
parenthetical verbs offered by Urmson so that it also includes performa-
tive verbs used parenthetically. Sentences such as the following
(10) I'll be there at two o’clock, I promise you
illustrate the parenthetical use of performative verbs. In uttering a
sentence like (10), the speaker adds to the first clause, with which he
performs the illocutionary act of promising, a second clause which
makes explicit the nature of his speech-act; and the parenthetical ‘I
promise you’ confirms, rather than establishes, the speaker’s commit-
ment (cf. [’ll be there at two o'clock — that’s a promise). Now, as (10) is
related, both semantically and grammatically, to
(11) I promise (that) I'll be there at two o’clock,
so (9) is related to
(12) I think (that) she’s in the dining-room;
and it is arguable that
(13) I promise to be there at two o’clock
is semantically, if not grammatically, equivalent to (11). Just how these
sentences are related, grammatically and semantically, is a controversial
question. It has been argued by J. R. Ross (1970) and others that all
sentences contain an underlying performative verb of saying: we will
come back to this question later (16.5).
 16.1. Speech-acts 739
The parallelism between parenthetical and performative verbs was
noted by Benveniste (1958a), independently of both Austin and Urm-
son; and Benveniste emphasized their non-descriptive role as markers of
subjectivity (“indicateurs de subjectivité’’) — i.e. as devices whereby the
speaker, in making an utterance, simultaneously comments upon that
utterance and expresses his attitude to what he is saying. This notion of
subjectivity* is of the greatest importance, as we shall see, for the under-
standing of both epistemic* and deontic* modality (cf. 17.2, 17.4).
In a related article, Benveniste (1958b) also draws attention to what
he calls delocutive* verbs. These may be defined as follows: a verb ‘x’
is delocutive if it is morphologically derived from a form «x and if it
means “‘to perform the (illocutionary) act that is characteristically per-
formed by uttering x (or something containing x)’. This definition is
hardly precise enough, as it stands (cf. Ducrot, 1972: 73ff): but it will
serve for our present purpose. The important point to note is that x is a
form that is uttered in the performance of the act that is denoted by the
lexeme ‘x’ and that there is a morphological relationship between x and
the forms of ‘x’. For example, the Latin ‘salutare’ (“to greet’’) is
morphologically related to the stem-form of ‘salus’ and thus to Salus!
(“‘Greetings!’’); the French ‘remercier’ (“to thank’’) is morphologic-
ally related to merci (cf. Merci!, “Thank you!’’); the English ‘to wel-
come’ is morphologically related to the form welcome (cf. Welcome!).
Of course, ‘salutare’ does not mean “‘to say Salus!’’, any more than
‘remercier’ means ‘“‘to say Merci!’ or ‘to welcome’ means “‘to say
Welcome!’’. But one way of greeting a person in Latin was to say Salus/,
as one way of thanking someone in French is to say Merci! and one way
of welcoming someone in English is (or was) to say Welcome! Moreover,
in each case the utterance that serves as the basis for the morphological
derivation of the verb denoting the more general act of greeting, thank-
ing or welcoming is one that is (or was) characteristically used for this
purpose. The conventionalization of the utterance of x is prior to the
creation of the lexeme ‘x’ or to the association with the pre-existing
lexeme ‘x’ of the sense ‘‘to perform the act that is characteristically
performed by uttering x’’. Up to a point, therefore, we are justified in
saying that the more general sense “‘to greet”’ developed out of the more
specific sense ‘‘to say Salus/”; and so for the more general sense of
‘remercier’, ‘to welcome’, etc.
What must be emphasized, however, is that the verb ‘say’, which we
have used in referring to the more specific sense of ‘salutare’ does not,
and cannot, simply mean “‘utter’’; and the reason why this is so is
740 Mood and illocutionary force
crucial for a proper understanding of both delocutive and performative
verbs, and of the connexion between them.
It is a commonplace of the philosophical discussion of language that
the verb ‘say’ (and more or less comparable verbs in other languages)
has several distinguishable senses. Austin himself (1962: g2ff) analyses
the act of saying (“in the full sense of ‘say’ ’’) into three component acts:
(i) the act of “uttering certain noises’’; (11) the act of “‘uttering certain
vocables or words, i.e. noises of certain types belonging to and as
belonging to a certain vocabulary, in a certain construction”’; (iii) the
act of using the product of (11) “with a certain more or less definite sense
and a more or less definite reference (which together are equivalent to
meaning)’’. It is easy to see that Austin’s analysis is, from the linguistic
point of view, either incomplete or imprecisely formulated (e.g., it is not
made clear, under (i), how much of the vocal signal is covered by the
non-technical term ‘noise’, and no attempt is made, under (ii), to dis-
tinguish between forms, lexemes and expressions); and the technical
terms that Austin does introduce at this point (notably ‘phatic’ and
‘rheme’) tend to be used quite differently by linguists. But his general
intention is clear enough; and, as far as it goes, his analysis would seem
to be on the right lines. At least these three kinds of acts are involved in
the complex act of saying.
Vendler (1972: 6ff) draws a broad distinction, as others have done,
between saying something in the full sense of the word (let us call this
“‘say,’’) and saying something in the weak sense which is “roughly
equivalent to uttering, mouthing or pronouncing”’ (let us call this
‘‘say,’’); and he points out that “‘no illocutionary act will be performed
if, for one thing, the speaker does not understand what he is saying or,
for another, he does not intend to perform such an act, that is, does not
intend his audience to take him to be performing one”’ (p. 26). It is in-
herent in the notion of performing an illocutionary act (i.e. of saying in
the sense ‘“‘say,”’) that the speaker shouJd both understand and mean
what he says (in the sense “‘say,’’). It might be argued that we are
frequently held to be responsible, in a court of law for example, for the
unintended consequences of our actions. This is true, but irrelevant. If
we are judged guilty of breach of promise by virtue of the utterance of
something that we did not intend to be taken as a promise, our guilt is
established, at law, in terms of the eminently practical principle that we
must be deemed to have made a promise if we have ostensibly performed
an act which is conventionally interpreted as making a promise and if
there is no clear indication at the time that we are not to be taken
 16.1. Speech-acts 741
seriously. But it is one thing to be deemed to have made a promise; it is
another to have actually made a promise. If we indicate clearly (i.e. in a
way that any reasonable person could be expected to interpret correctly)
that our utterance is not to be taken seriously as a promise, we shall not
only not have made a promise, but we shall not even be deemed to have
made a promise. What was said earlier about promising, and more
generally about the performance of any illocutionary act, is to be con-
strued in terms of this proviso. One cannot unwittingly or unintention-
ally say something 1n the sense “say,” merely by saying something in
the sense “say,’’. Furthermore, as Vendler points out, whereas ‘say’ in
the sense “say,” 1s an activity-verb, in the sense “‘say,’’ it is an
accomplishment-verb: it follows that the truth of the proposition “X
says,...” at a particular point in time carries no implications whatsoever
with respect to the truth of the proposition ‘‘X says, ...’’ at the same,
or any subsequent, point in time (cf. 15.6).
This distinction between “say,” and “‘say,”’ is by no means sufficient
to support all the weight that it is sometimes expected to bear: in
particular, it will not of itself suffice for drawing the distinction between
direct and indirect discourse from a semantic point of view.® Apart from
the various problems that philosophers have discussed in their attempts
to make precise all that is involved in “‘say,’’, there are quite serious
problems attaching to the interpretation of “‘say,’’; and these have not
been so extensively discussed. It is clearly of some importance, for
example, to distinguish between the type-token identity that is relevant
to the notion of repetition and the type-token identity that is relevant to
mimicry. Repetition and mimicry are two quite different kinds of
replication* (cf. 1.4). When we assert truly that X has correctly repeated
Y’s utterance, we abstract from all sorts of phonetically describable
differences in the utterance-signals. Voice-quality is certainly not rele-
vant to the specification of the truth-conditions of the sentence “Mary
repeated what John had said’. For Mary to make an attempt to replicate
John’s characteristic voice-quality in response to his request that she
should repeat what he had said would be, to say the least, supereroga-
tory. So too would be her attempt to replicate the paralinguistic, and
even some of the prosodic, features in his utterance. The truth-condi-
tions of ‘Mary repeated what John had said’ (unlike those of ‘Mary
imitated what John had said’) are presumably identical with the truth-
® For some interesting comments on the relationship between corresponding
direct and indirect discourse constructions cf. Banfield (1973), Partee (1972),
Zwicky (1971).
 742 Mood and illocutionary force
conditions of ‘John said X and so did Mary’ (where X is a form or an
utterance-signal and ‘and’ is construed to mean ‘‘and subsequently’’).
But it is remarkably difficult to establish, other than by methodological
fiat, what these truth-conditions are.!° It is presumably a necessary
condition of the relevant kind of type-token identity that the two tokens
of X should contain the same forms in the same order. But this is rarely,
if ever, a sufficient condition; and it is not clear that there is any deter-
minate set of additional conditions that would be appropriate to decide,
for all occasions of its utterance, whether ‘John said X and so did Mary’
is being used to assert a true proposition or not. In short, “say,” is far
from being as straightforward as one might think. So too is the distinc-
tion that philosophers frequently invoke between what is said (in the
sense ‘‘say,’’) and the manner of saying it.
Crude though it is, the distinction between “say,’’ and ‘“‘say,”’ may
be used to throw light on the nature of performative and delocutive
verbs, and on the nature of the relationship between them. As we have
seen, the utterance by X of J promise can never of itself be a condition of
the truth of the proposition “X promises’’. However, in so far as I
promise serves as a performative formula whose utterance (in the
appropriate circumstances) 1s associated by convention with the act of
promising (i.e. of committing oneself, under pain of dishonour, social
disapproval or some other such sanction, to some future act or course of
action), as the utterance of Hello! is associated with the act of greeting
and the utterance of Welcome! with the act of welcoming, ‘“‘X said,
I promise’ will generally be held to imply “X promised’’, just as ‘X
said Hello!/Welcome! (to Y)”’ will generally be held to imply ‘“‘X greeted/
welcomed Y’’. It is arguable, therefore, that the performative use of
I promise is logically, if not historically, prior to the descriptive use of the
verb ‘to promise’ and that the token-reflexivity of particular utterances
of I (hereby) promise . . . is a secondary consequence of this fact (cf.
Ducrot, 1972: 73ff). However that may be, the semantic connexion
between the Latin delocutive verb ‘salutare’ and the performative
formula Salus! is obviously no different, as far as the distinction between
“say,” and “‘say,’’ is concerned, from the semantic connexion that
10 By the term ‘methodological fiat’ I am referring to the more or less deliberate
process of standardization* that is an inevitable part of linguistic analysis and
description (cf. 14.2). There are of course constraints upon the linguist’s
fiat: up to a point native speakers will agree that two utterances are tokens of
the same type, the one being a repetition of the other. But dialectal and
stylistic variation are such, in most language-communities, that the question
is not always pre-theoretically decidable.
 16.1. Speech-acts 743
holds between the descriptive sense of the verb ‘to promise’ and the
performative formula J promise. The fact that the performative formula
I promise, unlike Salus! (or Hello! or Thanks!), contains the first-person
singular form of the corresponding descriptive verb, and may thus be
construed as token-reflexive (cf. 1.4), is, from this point of view, irrele-
vant.
As we have seen, in Austin’s later doctrine all utterances, including
Statements, are taken to be performative utterances. Much of the
original motivation for introducing the term ‘performative’, therefore,
disappears in the subsequent development of the theory of speech-acts.
But the distinction between explicit and primary performatives remains;
so too does the distinction between the performative and the purely
descriptive use of such verbs as ‘say’ and ‘promise’. Each of these two
points requires a final brief comment.
It is not absolutely clear on what grounds Austin draws his distinction
between explicit and primary performatives: in particular, it is not clear
whether an explicit performative must necessarily contain a performa-
tive verb. (The reader should note at this point that, whenever the term
‘performative’ is employed as a noun in this book, it is to be construed
as an abbreviation for ‘performative utterance’. In this respect, we base
our usage upon Austin’s (1962: 6). Other writers treat the noun ‘per-
formative’ as an abbreviation for ‘performative verb’; and this can
occasionally lead to confusion.)"' If we take seriously the criterion of
“making explicit (which 1s not the same as stating or describing) what
precise action is being performed”’ (cf. Austin, 1962: 61), it is obvious
that, in principle, the element that makes explicit the illocutionary force
of an utterance need not be a verb. For there is no reason to suppose that
only verbs have the function of making things explicit. Indeed, it need
not be a word, or even a particle: it could be some prosodic or para-
linguistic feature. But Austin certainly argues throughout as if the only
way in which the illocutionary force of the utterance can be made
explicit is by means of a performative verb (in the first-person singular) ;
and his examples all suggest that this is so. It very much looks, 1n fact, as
if Austin is covertly and perhaps illegitimately restricting the interpreta-
tion of “‘making explicit’’. Exegesis is rendered the more difficult in that
Austin, like most philosophers and many linguists, does not explain how

11 Our distinction between ‘sentence’ and ‘utterance’ is different from Austin’s,

for whom sentences were a subclass of utterances. We do not therefore
operate, as others (including Austin) have, with the notion of performative
sentences.
 744 Mood and illocutionary force
much of the signal-information in an utterance is to count for type-
token identity: 1.e. he does not tell us how to interpret what we have
called ‘say’ in the sense “say,”’. Presumably, it is a precondition of
something being an explicitly performative element for Austin that it
should be part of what we say (in the sense “‘say,”’), rather than part of
our manner of saying it; and he does operate with this distinction
between what is said and the manner of its being said. But the distinc-
tion itself is never made precise.
As for the distinction between the performative and the descriptive
use of verbs like ‘say’ and ‘promise’, it is frequently argued (and more
especially by those who wish to account for the meaning of all sentences
in terms of their truth-conditions: cf. 6.6) that Austin was wrong when
he said that performative utterances (in the original sense of ‘performa-
tive’) were neither true nor false. All that needs to be said on this issue
is that it is by no means as clear-cut as either Austin or his opponents
have implied. A case can'be made for assigning a truth-value to the
propositions that are expressed by sentences used to make non-constative
utterances (cf, Stampe, 1975). But it certainly should not be asserted as a
matter of commonsense (cf. Lewis, 1972: 210) that anyone saying I
declare that the earth is flat (under the appropriate conditions) has spoken
truly. The commonsense view would surely be that anyone saying this
would be asserting, somewhat emphatically, the proposition that the
earth is flat, rather than the proposition that he declares that the earth is
flat. Nor is it the case that anyone saying J am speaking would normally
expect to be taken as asserting that in the course of saying [ am speaking,
rather than before or after his utterance of I am speaking, he was
speaking: it is difficult, though not impossible, to imagine a situation in
which I am speaking could be token-reflexive. There may well be
theoretical advantages in extending the notions of description and truth
in such a way tnat, in our metalinguistic statements about performative
utterances, we can say that the speaker, in producing an explicit per-
formative, simuitaneously describes his performance (by means of a
parenthetically used performative verb) and, provided that the felicity
conditions are fulfilled, that he does so truly. But we cannot reasonably
say that this is in accordance with any everyday or commonsense use of
the terms ‘describe’ or ‘true’. Furthermore, the whole question whether
sentences used to make performative utterances do or do not have
truth-conditions is of secondary importance. As we have seen, there is a
systematic relationship between the truth-conditions of “X promised”’
and the felicity-conditions of I promise said by X. If, for theoretical
 16.2. Commands, requests and demands 745
reasons, we say that X, in saying [ promise, asserts the proposition “X
promises”’ and that the proposition 1s true provided that the felicity-
conditions are fulfilled, we are in effect labelling the felicity-conditions
as truth-conditions; and that does not absolve us from taking account of
the differences, to which Austin drew attention, between constative and
non-constative utterances.

16.2. Commands, requests and demands

In this and the following section we shall be concerned with what are
traditionally regarded as the three main classes of sentences. Most
grammars, however, do not distinguish systematically between sen-
tences and utterances. Throughout our discussion we shall maintain the
terminological distinctions that have already been introduced, using
‘statement’, ‘question’ and ‘command’ for utterances with a particular
illocutionary force and ‘declarative’, ‘interrogative’ and ‘imperative’
for sentences with a particular grammatical structure.
As far as statements are concerned, we will restrict our attention in
this section to modally unqualified, or categorical*, assertions: 1.e. to
statements that are unqualified in terms of possibility and necessity. It is
the propositions expressed by the sentences uttered in making such
statements that have been formalized in the standard two-valued pro-
positional calculus (cf. 6.2). We shall need to distinguish later between
the assertion of a negative proposition and the denial of a positive
proposition, but we can proceed, for the present, without drawing this
distinction. It will be sufficient, at this point, to remind the reader that
the typical statement will have the form of a simple declarative sentence;
and that assertion is an illocutionary act, which, when combined with a
propositional act, makes the utterance into a statement.
As used in traditional grammar, the term ‘command’ is generally
taken to cover requests and entreaties, as well as commands in the
narrower sense. In order to avoid confusing the more general and the
more specific senses of ‘command’, we will henceforth employ Skinner’s
term mand* as a general term to refer to commands, demands, requests,
entreaties, etc. Our use of the term ‘mand’ does not of course commit
us to a behaviouristic analysis of meaning (cf. 5.3). Mands, as we shall
see, are a subclass of what might be called directives* (cf. Ross, 1968);
12 For further discussion of the notion of speech-acts, from a philosophical and
linguistic point of view, cf. Cole & Morgan (1975), Ducrot (1972), Fann
(1969), Habermas (1972), Wunderlich (1972). For the integration of speech-
oe974).
meory with sociolinguistics and stylistics: cf. Giglioli (1972), Hymes
 746 Mood and tllocutionary force
that is to say, utterances which impose, or propose, some course of
action or pattern of behaviour and indicate that it should be carried out.
Mands differ from other subclasses of directives, such as warnings,
recommendations and exhortations in that they are governed by the
particular speaker-based felicity-condition that the person issuing the
mand must want the proposed course of action to be carried out: if the
speaker does not really want his mand to be obeyed or complied with, he
is guilty of what Austin would call an abuse (cf. 16.1). Not only mands,
but all personal directives, including warnings, recommendations and
exhortations, are governed also by the more general addressee-based
condition that the speaker must believe that the addressee is able to
comply with the directive. One cannot appropriately command, request,
entreat, advise, or exhort someone to perform an action, or demand that
| he perform an action, which one knows or believes he is incapable of
performing.
In many languages the difference between mands and statements 1s
grammaticalized in the form of the main verb of the sentences that are
characteristically used to perform such acts, These differences in the
inflexional forms of the verb are traditionally described in terms of the
grammatical category of mood*. For example, the second-person
singular imperative form of the Latin verb ‘dicere’ (“‘to say’’) is dic and
the second-person singular of the present indicative is dicis: cf. ‘Dic
mihi quid fecerit’ (“Tell me what he did’’) ws. ‘ Dicis mihi quid fecerit’
(“You are telling me what he did”). Latin is typical of most of the
Indo-European languages (and many other languages outside the Indo-
European family are like Latin in this respect) in that the second-person.
singular imperative carries no overt indication of person or tense (as the
vocative singular of nouns in the Indo-European languages carries no
overt indication of case). It has often been suggested that the reason for
this is that the imperative, as the principal mood of will and desire, is
ontogenetically more basic than the indicative, the mood of statement.
Whether or not this is a correct explanation of the fact that, in certain
languages, the imperative forms of the verb carry no overt indication of
tense and person, it is important to realize that commands and requests,
of their very nature, are necessarily restricted with respect to the seman-
tic distinctions that are grammaticalized, in many languages, in the
categories of tense and person. We cannot rationally command or request
someone to carry out some course of action in the past: the only tense
distinctions that we might expect to find grammaticalized in the im-
perative, therefore, are distinctions of more immediate and more remote
 16.2. Commands, requests and demands 747
futurity. For similar reasons, the imperative is intimately connected
with the second person (or vocative). It is implicit in the very notion of
commanding and requesting that the command or request is addressed
to the person who is expected to carry it out. In so far as the imperative
is the mood whose function is that of being regularly and characteris-
tically used in mands, the subject of an imperative sentence will neces-
sarily refer to the addressee. This does not mean of course that the
subject of a command or request must be a second-person pronoun. We
can transmit a command or request indirectly through an intermediary
(e.g., Let him come and see me tomorrow). More important, we can, in
certain styles, refer to an addressee in the third person (cf. 15.1); it is in
principle possible, therefore, for a language to have a true third-person
imperative.
What are traditionally described as first-person and third-person
imperatives, however, in the Indo-European languages at least, are not
true imperatives, in the sense in which the term is being used here. The
subject of these so-called imperatives does not refer to the addressee.
The fact that the subject of an imperative sentence is normally gram-
maticalized in the second person (in those languages which do in fact
grammaticalize the deictic category of person) derives from the fact that
the communication of a command or request, like the communication of
a proposition, requires both a sender and an addressee; and commands
and requests are necessarily, not just contingently, addressed to those
who are to carry them out.
At this point, the reader’s attention is drawn to an important dif-
ference between the terms ‘imperative’ and ‘interrogative’, as they are
traditionally employed by grammarians, The former, like ‘indicative’
and ‘subjunctive’, is used to refer to the mood of the verb, and only
secondarily to particular kinds of sentences: an imperative sentence,
therefore, is a sentence whose main verb is in the imperative mood, as
an indicative sentence is one whose main verb is in the indicative and a
subjunctive sentence is one whose main verb is in the subjunctive.® The
term ‘interrogative’, on the other hand, is never used in traditional
13 The term ‘mood’ is used throughout this work in its traditional, rather
restricted, sense. In view of what is said in this and the following paragraph I
now believe that it was misleading (although it is by no means uncommon) to
suggest that the difference between declarative and interrogative, like the
difference between indicative and imperative, is a matter of mood (cf. Lyons,
1968: 307). Many linguists nowadays employ the term in a much broader
sense (cf. Halliday, 1970a; Householder, 1971). So do certain logicians (cf.
Kasher, 1972; Stenius, 1967).
748 Mood and illocutionary force :
grammar to refer to one of the moods of the verb; and the reason is that
in none of the languages with which traditional grammar has been con-
cerned, and possibly in no attested language, is there a distinct mood
that stands in the same relation to questions as the imperative does to
mands.!4 The term ‘declarative’ is like ‘interrogative’ in this respect.
We have been operating with two tripartite distinctions: between
statements, questions and mands, on the one hand, and between declara-
tive, interrogative and imperative sentences, on the other. It will now be
clear, however, that this is a somewhat misleading classification, in that
‘imperative’ goes with ‘indicative’ (and ‘subjunctive’) rather than with
‘declarative’ and ‘interrogative’. As a sentence may be both interroga-
tive and indicative (but not both interrogative and declarative), so, in
principle, it might be both interrogative and imperative (but not both
indicative and imperative). What is required, then, is a term that does
stand in the same relation to mands as ‘interrogative’ does to questions
and ‘declarative’ to statements. The term that we will use for this pur-
pose is one that has occasionally been employed in something like this
sense by grammarians: jussive*. A jussive sentence, then, will be one of
a grammatically defined class of sentences that are characteristically used
to issue mands. Generally speaking, imperative sentences (in languages
that have a distinct imperative mood) will be a proper subset of jussive
sentences. In Spanish, for example, the class of jussive sentences in-
cludes both imperative and subjunctive sentences, as it does in many
other languages (though the conditions for the use of one kind of sen-
tence, rather than the other, may vary considerably across languages).
Needless to say, the term ‘imperative sentence’ is frequently employed
by other writers in the broader sense that we have here given to ‘jussive
sentence’; and this can lead to confusion.
So far we have made no attempt to distinguish between commands
and requests. It has been suggested that this difference (like the differ-
ence between offers and promises) is one of politeness or deference (cf.
Gordon & Lakoff, 1971; Heringer, 1972). But this suggestion is un-
convincing. It may well be that the notion of politeness is inapplicable to
14 ‘There may well be languages, however, with a mood whose basic function is
that of expressing doubt or qualifying the speaker’s commitment to truth;
and, as we shall see later, there are parallels between questions and dubitative,
or epistemically qualified, utterances such that it would not be unreasonable
to expect that what is basically a dubitative mood might be regularly used
both for posing questions and expressing doubt or uncertainty.
15 Even greater confusion is caused by the fact that the term ‘imperative
sentence’ is often used in place of ‘command’, ‘request’, etc.
 16.2. Commands, requests and demands 749
commands. But one can be either polite or impolite in the way in which
one makes a request; and an impolite request is not a command.
The crucial difference between a command and a request seems to be
rather that a request leaves to the addressee the option of refusal to
comply with the mand, whereas a command does not. One way in which
this option of refusal may be encoded in the verbal component of English
utterances is by adding the form please. A sentence like
(1) Open the door, please
will therefore be normally used to make a request. But, as always,
information that is encoded in the verbal component of an utterance may
be contradicted or cancelled by information that is encoded prosodically
or paralinguistically (cf. 3.1); and the difference between commands and
requests is in fact mainly conveyed, as one might expect, in the non-
verbal component of utterances. Another way of encoding verbally the
option of refusal in English is by adding a parenthetical interrogative tag
(e.g., ‘will your’, ‘won’t you?’) to an imperative clause, as in (2) and (3):
(2) Open the door, will you?
(3) Open the door, won’t you?
The tag* that is added to an imperative clause clearly indicates that the
speaker is conceding to the addressee the option of refusal. But, once
again, this concession may be contradicted or cancelled by the prosodic
or paralinguistic component of the utterance.
In his analysis of the meaning of declarative, jussive and interrogative
sentences, Hare (1970) draws a valuable terminological distinction
between what he calls the phrastic, the tropic and the neustic.’® By the
phrastic*® he means that part of sentences which is common to corres-
ponding declarative, jussive and interrogative sentences: its propositional
content. The tropic* is that part of the sentence which correlates with
the kind of speech-act that the sentence is characteristically used to
perform: it is what Hare calls “‘a sign of mood’’; and in many languages
it will in fact be grammaticalized in the category of mood. The difference
between the imperative and the indicative mood in Latin, for example,
grammaticalizes the difference in the tropics of corresponding jussive and
declarative sentences: e.g., ‘Dic mihi quid fecerit’ and ‘Dicis mihi quid
fecerit’ (to repeat the example given earlier). ‘The neustic* is what Hare
calls a ‘‘sign of subscription”’ to the speech-act that is being performed:
16 This tripartite distinction constitutes a refinement of the earlier, and perhaps
better-known, bipartite distinction of Hare (1952).
750 Mood and illocutionary force
it is that part of the sentence which expresses the speaker’s commitment
to the factuality, desirability, etc., of the propositional content conveyed
by the phrastic. Like many authors, Hare frequently used the term
‘sentence’ where it would seem to be more appropriate to use the term
‘utterance’; nor does he distinguish clearly between ‘statement’,
‘declarative’ and ‘indicative’, between ‘command’, ‘“jussive’ and
‘imperative’, and so on. We will treat the neustic, the tropic and the
phrastic as being components of the logical structure of utterances.
Hare’s distinction of the neustic from the tropic separates two of the
functions that Russell & Whitehead (1910: 9), following Frege (cf.
Dummett, 1973: 308ff), ascribed to the assertion-sign (F), which they
prefixed to a propositional variable, in order to show that the proposition
was being asserted as true, rather than merely being entertained or put
forward for consideration. As far as straightforward statements of fact,
or categorical assertions, are concerned, the tropic can be said to have
the meaning “it is so” and the neustic “I say so’’. Both of these mean-
ings are normally taken to be included in “‘it is the case that”? when we
interpret the formulae of the propositional calculus as having this phrase
prefixed to them (cf. 6.2). But they can be dissociated. When a simple
proposition (e.g., p) is embedded in a complex proposition (e.g.,
p — q); the I-say-so part of the assertion-sign (“‘it is the case that”’) is
not applicable to the component simple proposition, but only to the
complex proposition taken as a whole. The component simple proposi-
tion, however, still has associated with it what Hare calls a sign of mood
(‘it is so”). When we make a hypothetical, rather than a categorical,
assertion (e.g., If Fohn 1s working, . . .), we do not subscribe to the
factuality of the proposition expressed by the embedded declarative
sentence (‘“‘John is working’’); we nonetheless put this proposition for-
ward for consideration as a fact, and thereby associate with it the it-1s-so
component of the act of assertion. Similarly, when we embed a declara-
tive sentence as the object of a verb of saying in indirect discourse, we
associate the it-is-so component, but not the I-say-so component, with
the proposition that is expressed by the embedded sentence (cf. the
statement He says that Fohn 1s working).
The illocutionary force of a statement may be regarded as the product
of its tropic and its neustic. As we shall see later, it is in principle possible
to draw a distinction between the unqualified assertion of the possibility
of a proposition and the qualified assertion of its factuality (17.6); and
this distinction can be handled in terms of the difference between
qualifying the tropic and qualifying the neustic. English, however, does
 16.2. Commands, requests and demands 751
not systematically distinguish between these two kinds of modality; and
perhaps no language does in primary performatives.
Mands differ from statements in that their tropic is to be interpreted
as “‘so be it’’, rather than ‘“‘it is so’’. Whereas a statement tells the
addressee that something 1s so, a mand tells the addressee that something
is to be made so. Corresponding statements and mands can be said to
have the same propositional content, but to differ in their tropic. Both
categorical assertions and commands, however, contain the same un-
qualified I-say-so component, indicating that the speaker commits
himself fully to the factuality (it-is-so) or desirability (so-be-it) of what
is described by the phrastic. The difference of illocutionary force
between categorical assertions and commands is, therefore, a function of
the difference between “‘it is so”’ and “‘so be it”’.
The only kinds of mands that we have considered so far are commands
and requests. There is however a third major type of mand: demands*
(cf. Boyd & Thorne, 1969). Demands are like commands and requests in
that they are inherently restricted with respect to tense. Just as we
cannot rationally command someone to do something in the past, so we
cannot rationally demand that it be so in the past. But demands differ
from commands and requests in that they are not necessarily addressed
to those upon whom the obligation of fulfilment is imposed. In English,
primary performatives with the illocutionary force of demands will
typically contain what is traditionally described as a third-person im-
perative (e.g., Let there be light) or one of the modal verbs ‘shall’
(pronounced with heavy stress) or ‘must’ (e.g., He must be here at six,
He shall be here at six). Explicitly performative demands are typically
introduced by verbs such as ‘demand’ and ‘insist’ (e.g., J demand that
he be here at six, I insist that he come). It is worth noting that in many
dialects of English the subordinate clause in such explicitly performative
demands is grammatically distinct from the subordinate clause in
explicitly performative statements. The verb ‘insist’ can be used to
make explicitly performative statements or demands: cf. I insist that he
1s there, I insist that he be there. (Insistence, of course, is not an illo-
cutionary act: it is an emphatic qualification of the I-say-so component
that is common to both statements and demands.)
Corresponding primary performatives with the illocutionary force of
statements and demands differ characteristically in Latin (and many
other languages) in much the same way that the subordinate clause of
I insist that he ts there differs from I insist that he be there in English. To
the Latin subjunctive sentence ‘Fiat lux’ (‘‘Let there be light”’), which
752 Mood and illocutionary force
may be uttered to make a demand (whoever or whatever, in this case,
might be the addressee), there corresponds the indicative sentence
‘Fit lux’ (“Light is coming into being’’), which may be uttered to make
a statement. Although fiat is traditionally described as a present-tense -
subjunctive form, it is (by virtue of its so-be-it component) as much
future as present. The semantic opposition of past, present, and future
does not apply to demands, as it does not apply to commands and
requests. The correspondence between indicative and subjunctive sen-
tences (like the correspondence between declarative and jussive sen-
tences in general) is therefore a many-to-one, and not a one-to-one
correspondence. But this does not affect the general point being made
here that corresponding indicative and subjunctive sentences express the
same proposition. It is, in any case, possible (as we shall see later) to
treat the tense of an indicative sentence, like its mood, as something
which is analytically separable from the proposition which it expresses
(cf. 17.2).
If demands are said to be like commands in that they have the same
phrastic and the same neustic as categorical assertions do, but to differ
in their tropic, how do we account for the difference in the illocutionary
force of commands and demands? The answer that is tentatively offered
here depends upon the assumption (which might however be challenged)
that there is no difference, as far as primary performatives are con-
cerned, between imposing a command and imposing a demand upon the
addressee. The distinction can be drawn, it is true, by means of two
different explicitly performative utterances, such as
(4) I order you to free the prisoner immediately
and

(5) I demand that you free the prisoner immediately.

But the felicity-conditions attaching to the appropriate utterance of (4)
and (5) are very similar, if not identical. At most, the difference would
seem to reside in the fact that giving commands 1s something that we
associate with institutionalized authority, but issuing demands is not;
and this is not a difference which makes the speaker’s assumption of
authority when he utters a command something different from his
assumption of authority when he issues a demand. A primary performa-
tive, using the imperative, like
(6) Free the prisoner immediately
 16.3. Questions 753
could be used in circumstances in which either (4) or (5) would be
appropriate. Further support for this analysis comes from the fact that
commands and demands can both be reported by means of the same ,
kind of statement, for example,
(7) I told him to free the prisoner immediately ;
also, from the fact that constructions which are characteristically used
to make demands (the subjunctive in Latin, ‘must’ in English, ‘sollen’
in German, etc.) can, under conditions which vary from one language to
another, be used interchangeably with the imperative to issue com-
mands, provided that it is clear in context that the obligation to carry
out the mand is being imposed on the addressee. It may well be therefore
that the difference between commands and demands is not one of
illocutionary force, but something that derives solely from the nature of
social interaction and communication. It is nevertheless convenient to
have distinct terms, ‘command’ and ‘request’, for mands that are issued
to the addressee, since these are, again by virtue of the nature of social
interaction and communication (in most situations at least), the most
frequently used kinds of mands; and many languages, as we have seen,
have special forms of the verb, imperatives, whose characteristic func-
tion is that of being employed in commands and requests.

16.3. Questions
It has been argued that questions can be analysed satisfactorily as sub-
types of mands (cf. Hare, 1949; Lewis, 1969: 186). According to this
proposal Who 1s at the door? might be analysed as an instruction to the
addressee to name (or otherwise identify) the person at the door and
Is he married? as an instruction to assert one of the component simple
propositions of the disjunction “He is married or he is not married’”’.
Essentially the same proposal has been made more recently, within the
framework of generative grammar, by several linguists; and it has been
quite widely accepted. The advantage of this analysis of questions is
that it would enable us to handle the illocutionary force of the three main
classes of utterances in terms of the two primitive notions of asserting
and issuing mands. There are, however, a number of objections to the
proposal that questions should be analysed as instructions to make a
statement. None of these objections is perhaps conclusive. Taken
together, however, they point the way to an alternative, and more
general, analysis of the meaning of questions.”
17 For other approaches to the analysis of questions from a logical and lin-
guistic point of view: cf. Aqvist (1965), Bach (1971), Baker (1970), Hamblin
 754 Mood and tllocutionary force
‘The first point to note is that the grammatical structure of what we
will call yes-no* questions (1.e. questions to which we can respond
appropriately with the words ‘yes’ or ‘no’ in English, and their equiva-
lents in other languages) is, in many languages, similar to that of
declarative sentences. In fact, the difference between questions and
statements is commonly drawn solely in the non-verbal component of |
utterances; and it is one that can be associated with an intonation pattern
or paralinguistic modulation of the utterance which expresses the
speaker’s doubt. This fact would suggest that the difference between
declarative sentences and interrogative sentences (in those languages in
which such a distinction is drawn in the verbal component of sentences)
results from the grammaticalization of the feature of doubt. It would be
generally agreed that one of the felicity-conditions attaching to the
appropriate utterance of questions (other than so-called rhetorical
questions) is that the speaker should not know the answer to his ques-
tion. It is for this reason that certain authors prefer to analyse questions
as meaning, not “Assert that such-and-such is so”’, but “‘ Bring it about
that I know that such-and-such is so”’ (cf. Aqvist, 1965; Householder,
1971: 85; Hintikka, 1974b); and it is worth noting that ‘Tell me that
such-and-such is so” can be interpreted in either of these two ways.
What is at issue is whether, in uttering a question, the speaker necessarily
assumes that his addressee knows the answer. If he does not make this
assumption he can hardly impose upon the addressee the obligation to
supply the answer.
The second point to be made is that, if yes-no questions were a sub-
class of mands, one might expect that the response No would indicate the
addressee’s refusal to comply with the mand (1.e. his refusal to state
whether something is or is not so). But this is not the case. If the
addressee says No in response to a question of the form Is the door open?,
he is answering the question. But if he says No in response to what is
clearly a mand, such as Open the door, he is refusing to do what he is
being commanded or requested to do.
Finally (and this is the most important point), it does not seem to be
essential to the nature of questions that they should always require or
expect an answer from the addressee. It is true that, in normal everyday
conversation, we generally expect the questions that we utter to be
answered by our addressee. But this is readily explained in terms of the

(1973), Hudson (1975), Hull (1975), Keenan & Hull (1973), Prior & Prior
(1955), Rohrer (1971).
 16.3. Questions 755
general conventions and assumptions which govern conversation. If I
say [ wonder whether the door 1s open or I don’t know whether the door is
open, which (like the question Is the door open?) express my doubt as to
the state-of-affairs which obtains, the addressee can appropriately re-
spond to my utterance, if he is in a position to do so, by resolving my
doubt. Given that this is so, all we need to assume in order to account
for the fact that questions normally expect and obtain an answer is a
conventional association between the utterance of a question and the
expectation of an answer from the addressee. In principle, however, this
association is independent of the illocutionary force of questions.
What seems to be required, in fact, is a distinction between asking a
question of someone and simply posing* the question (without neces-
sarily addressing it to anyone). When we pose a question, we merely give
expression to, or externalize, our doubt; and we can pose questions
which we do not merely expect to remain unanswered, but which we
know, or believe, to be unanswerable. To ask a question of someone is
both to pose the question and, in doing so, to give some indication to
one’s addressee that he is expected to respond by answering the question
that is posed. But the indication that the addressee is expected to give an
answer is not part of the question itself.
The advantage of this analysis of questions is that it 1s more general
than their analysis as mands. It covers, not only information-seeking
questions, but various kinds of rhetorical and didactic questions without
obliging us to treat these as being in any respect abnormal or parasitic
upon information-seeking questions (cf. Bellert, 1972: 59-63). Ithasthe
further advantage that it puts factual questions into more direct corres-
pondence with statements and what are traditionally described as
deliberate questions (e.g., Should I wash my hair to-night?, What am I
to do?) into more direct correspondence with mands and other kinds of
directives. Corresponding statements and factual questions, on the one
hand, and corresponding mands and deliberative questions, on the
other, can be said to have the same phrastic and tropic, but to differ in
their neustic. This is not simply a difference between the presence and
the absence of an element meaning “I-say-so’’; it is the difference
between the presence of an I-say-so element and the presence of an
I-don’t-know element.
One of the inadequacies of the analysis of questions as mands which
has not so far been mentioned is its failure to account satisfactorily for
the difference between wondering whether something 1s so and asking
oneself whether something is so. According to Hare (1971: 85):
 756 Mood and illocutionary force
(1) Lf wonder whether that ts a good movie?
which he classifies as an indirect interrogative, is “very similar in
meaning”’ to
(2) I ask myself “Is that a good movie?”’
Similar in meaning they may be, but there is an important difference
between them; and one should not be misled by the fact that in certain
languages the verb used to refer to, or give expression to, an act of
wondering ts a reflexive form of the verb meaning “‘ask”’. The equivalent
of (1) in French, for example, is
(3) fe me demande si c’est un bon film.
But the expression ‘se demander’ does not normally mean “‘to ask
oneself’’; the most common French expression used to refer to, or to
perform, acts of asking oneself whether something is so is ‘se poser la
question’ (“to put the question to oneself’’). The difference between
wondering and asking oneself is the difference between simply posing a
question and putting a question to oneself as the addressee with the
intention of answering it. For one can ask questions of oneself, in
soliloquy and discursive reasoning, just as one can make statements or
issue mands to oneself; and to ask a question of oneself is to perform a
mental or illocutionary act which is governed by the same felicity-
conditions as those which govern information-seeking questions
addressed to others. If Sherlock Holmes asks himself whether his visitor
is married or single, he does so with the expectation and intention, after
considering the evidence, of answering the question which might be
formulated, in an utterance, as ,
(4) Is he married? |
If Sherlock Holmes merely wonders whether his visitor is married he
poses the same question, but he does not necessarily expect to be able to
answer it.
Wondering, like entertaining a proposition, is first and foremost a
mental act: indeed, it is one way of entertaining a proposition. In order
for wondering to be converted into an illocutionary act by means of
utterance, it must be the speaker’s intention to tell the addressee that he
has a particular proposition in mind and that he is entertaining it in
what we may refer to as the dubitative mode.'® Otherwise the utterance
18 'The term ‘mode’ is quite commonly used in this sense by philosophers. It is
related to, though distinguishable from, the sense in which it was employed
 16.3. Questions 757
is at most informative, rather than communicative (cf. 2.1); and illo-
cutionary acts, as we have seen, are necessarily communicative (cf. 16.1).
I wonder whether that 1s a good movie may or may not be used (like Do you
know whether that is a good film?, I don’t know whether that 1s a good film,
Can you tell me whether that is a good film?, etc.) to ask, indirectly, a
question of one’s addressee.
So far all the questions that we have actually discussed have been of
the yes—no type. But there is another class of question which, following
Jespersen (1933: 305), we will call x-questions*. (Jespersen’s term for
yes—no questions is ‘nexus-question’: cf. also Katz, 1972: 207.) As
Jespersen points out, in x-questions “we have an unknown quantity x,
exactly as in an algebraic equation”’ and “‘the linguistic expression for
this x is an interrogative pronoun or pronominal adverb’’. Since the
interrogative pronouns and adverbs in English are words, which, in their
written form, typically begin with wh- (who, what, when, where, etc.),
x-questions are commonly referred to in the literature as wh-questions;
and wh- is sometimes treated, by linguists, as the orthographic form of
an interrogative morpheme which, when it is combined with indefinite
pronouns or adverbs, has the effect of converting them into interrogative
elements whose characteristic function it is to be used in x-questions
(cf. Katz & Postal, 1964; Katz, 1972: 204ff).
Not only x-questions, but also yes—no questions, can be treated as
functions which contain a variable (or “unknown quantity’’, to use
Jespersen’s phrase). When we ask a question of our addressee, what we
are doing, in effect, is inviting him to supply a value for this variable. A
yes—no question, like Js the door open?, contains a two-valued variable. It
is equivalent to the bipartite disjunctive question Js the door open or not? ;
and it can be appropriately answered with either Yes (which implies the
proposition expressed in the statement The door is open) or No (which
implies the proposition expressed by The door is not open). A factual
yes—no question presupposes* (in one of the senses of this term: cf.
14.3) the disjunction of two propositions, each of which has associated
with it an it-is-so tropic. Similarly, a deliberative yes—no question (e.g.,
Shall I get up?) presupposes the disjunction of a corresponding positive
or negative proposition associated with a so-be-it tropic.
An x-question is a many-valued function, which presupposes the
earlier for the two ways of describing situations (cf. 15.4). The term ‘mode’,
like ‘mood’, derives from the Latin ‘modus’, which being a word of very
general meaning (‘‘manner’’, ‘‘way’’, etc.) acquired several distinct technical
uses.
758 Mood and tllocutionary force
disjunction of a set of propositions (positive or negative according to the
form of the question), each member of the set differing from the others
in that it supplies a different value for the variable. For example,
(5) Who left the door open?
presupposes the disjunction of the set of propositions expressed by the
statements that could be made by uttering
(6a) John left the door open
(6b) That little boy left the door open
(6c) Uncle Harry left the door open
etc.
More particularly, (5) presupposes the proposition expressed by
(7) Someone left the door open;
and the indefinite pronoun ‘someone’ (in its non-specific interpretation:
cf. 7.2) can also be thought of as a variable whose range of possible
values depends upon the universe-of-discourse. If the addressee re-
sponds to (5), uttered as a question (whether it is asked of him or merely
posed), by making the statement
(8) No-one left the door open,
he is denying (7) and thereby refusing to accept one of the presupposi-
tions of (5); he is not answering the question, but rejecting it. If, on the
other hand, he replies to (5) by uttering (7) — that is to say, by asserting
what (5) presupposes — he is evading, rather than answering, the ques-
tion (cf. Katz, 1972: 213).
We will not discuss the grammatical structure of interrogative sen-
tences in detail. One point should be made, however, in connexion with
the grammatical relationship between (5) and (7). In most of the Indo-
European languages, the forms of the interrogative pronouns and
adverbs are related etymologically to indefinite pronouns and adverbs:
cf. English who, whom, what, etc.; French quz, que, quand, etc.; Russian
kto, cto, etc.; Greek tis/tis, pote/pote, etc.; Latin quis, quando, etc. In
many languages, including English, a set of indefinite pronouns and
adverbs has been created by affixing an adjectival modifier meaning
‘“‘some’’ to the original common pronominal or adverbial element, or to
some replacement of it: cf. English someone, something; French
quelqu’un, quelque chose (where the numeral meaning “‘one”’ and a noun
meaning ‘“‘thing’’ replace the original pronominal element); Russian
 16.3. Questions 759
kto-to “‘someone (specific)”’, kto-nibudj “someone (non-specific)”’ ;
Latin, guidam “‘someone (specific)”’, aliquis ‘someone (non-specific) ”’;
English somewhere, French quelque part (where the noun ‘part’ replaces
the original adverbial element). In Classical Greek the same forms are
used as both interrogative and indefinite pronouns (tus, tz, tina, etc.); and
the difference between them is one of accentuation (the indefinite pro-
noun normally being unaccented) and their position of occurrence in the
sentence.
These various morphological relationships clearly depend upon the
grammaticalization of some semantic property which is common to what
may be regarded as corresponding statements and questions. We have
seen that the question (5) presupposes the proposition expressed by the
statement (7), with ‘someone’ taken in its non-specific interpretation.
But consider now the effect of questioning the proposition expressed by
(7), not by means of (9)
(9) Did anyone leave the door open?
(where ‘anyone’ may be regarded as a grammatically determined variant
of non-specific ‘someone’), but by means of the prosodic (and para-
linguistic) modulation of (7). This prosodic (and paralinguistic) modula-
tion we will symbolize with a question-mark:
(10) Someone left the door open?
This utterance (which might be naturally used in English to express
doubt, surprise, etc.), if it is taken as a question, might be appropriately
answered with Yes or No. But it is easy to see that it might also be con-
strued as an x-question, presupposing the proposition expressed in (7)
and expecting the addressee to respond by supplying a value for
‘someone’.
Given that this is so, it is also easy to see that there are various ways
in which languages might systematically distinguish between yes—no
questions containing an indefinite pronoun (or adverb), x-questions and
indefinite statements, like (7), without necessarily grammaticalizing the
difference between all three classes of utterances, or any two of them, in
the verbal component. Suppose, for example, that we were to associate a
falling-intonation with statements and a rising-intonation with ques-
tions, and that we were to assign heavy stress to the indefinite pronoun
(or adverb) in x-questions. This of itself would be sufficient to maintain
the distinction between the three classes of utterances. Needless to say,
the relationship between statements and the two kinds of questions 1s
 760 Mood and illocutionary force
rarely, if ever, systematically made solely in the non-verbal component
of languages. But it is important to realize that it need not be gram-
maticalized in terms of a structural difference between declarative and
interrogative sentences; and furthermore that, if the distinction between
declarative and interrogative sentences is grammaticalized, particular
languages might well employ prosodically (and paralinguistically) modu-
lated declarative sentences, containing indefinite pronouns or adverbs,
for x-questions, reserving interrogative sentences for yes—no questions;
or alternatively that they might employ non-verbally modulated declara-
tive sentences for yes—no questions and use interrogative sentences
containing special pronominal or adverbial forms for x-questions. In
other words, the distinction between yes—no questions and x-questions
is a logical, or semantic, distinction that is universal, in the sense that it
can be drawn independently of the grammatical and lexical structure of
particular languages; but the difference between two kinds of interroga-
tive sentences, and even the difference between interrogative and
declarative sentences, is not.
The morphological relationship between interrogative and relative
pronouns (and adverbs) that holds in most Indo-European languages
including English is also worth commenting upon briefly in connexion
with x-questions. The forms who, when and which (cf. also German
‘welcher’, Latin ‘qui’, etc.), which are diachronically related to, if not
identical with, the interrogative/indefinite pronouns of the earlier Indo-
European languages, are found in both restrictive* and non-restrictive*
relative clauses in most dialects of modern English: cf.
(11) That man, who broke the bank at Monte Carlo, is a mathemati-
cian
(12) The man who/that broke the bank at Monte Carlo 1s a mathema-
tician.
Non-restrictive relative clauses, like (11), which are set off by commas
in written English and are at least potentially distinguishable by rhythm
and intonation in the spoken language, do not concern us here.!® Nor do

19 Non-restrictive relative clauses may have a different illocutionary force

associated with them from that which is associated with the rest of the text-
sentence within which they occur. In this respect they are like parenthetically
inserted independent clauses (cf. 14.6). For example, (11) can have the same
range of interpretations as That man — he broke the bank at Monte Carlo — is a
mathematician; and just as we can have Is that man — he broke the bank at
Monte Carlo — a mathematician? as an acceptable text-sentence, so we can have
Is that man, who broke the bank at Monte Carlo, a mathematician?
 16.3. Questions 761
relative clauses introduced by that, which is of a quite different origin
(cf. 15.2).
Restrictive relative clauses, like (12), are used, characteristically, to
provide descriptive information which is intended to enable the
addressee to identify the referent of the expression within which they
are embedded (cf. 10.3). For example, “the man who/that broke the
bank at Monte Carlo’ tells the addressee of which person it is being
asserted that he is a mathematician. In order to bring out the semantic
relationship between restrictive relative clauses, used in this way, and
x-questions containing an indefinite/interrogative pronoun, we will con-
struct a form of Quasi-English in which ‘someone’ and ‘something’ are
employed indifferently (like the Classical Greek ‘tis’) in questions and
indefinite statements: cf. (7) and (10). A sentence like

(13) Someone broke the bank at Monte Carlo

could be used, therefore, in this kind of Quasi-English either to ask an
x-question or to make a statement. Now, just as an attributive adjective,
like ‘tall’, denotes a property which supplies a value for x in referring
expressions like ‘the x man’, so too do restrictive relative clauses (and
they are traditionally classified as adjectival clauses). Let us therefore
simply embed (13) in place of this adjectival variable, to yield, for
example, the Quasi-English sentence

(14) The someone broke the bank at Monte Carlo man is a mathema-
tician,

which is equivalent, in meaning, to (12). If this sentence were used to

make a statement, it would, like (12) in the same circumstances, pre-
suppose that someone broke the bank at Monte Carlo, the proposition
expressed by (13); that someone is a man, expressed by
(15) Someone is a man;
and furthermore that it is the same specific (rather than non-specific)
someone that is being referred to in both cases and that the addressee
should be able to identify him in terms of the properties denoted by
‘man’ and ‘having broken the bank at Monte Carlo’. What has been
outlined here is one way in which relative clauses might be formed by
grammaticalizing these presuppositional relations and associating them
with an adjectivalized interrogative or indefinite declarative sentence
construed as a predicate denoting a property; and this would seem to be
 762 Mood and tllocutionary force
the source of the ‘who’/‘which’ relative clauses in English, except that
they are more closely related, diachronically, to interrogative than to
indefinite declarative sentences.
We have seen that there is a particular kind of semantic correspon-
dence which holds between an x-question like Who broke the bank at
Monte Carlo? and an indefinite statement like Someone broke the bank at
Monte Carlo: the former presupposes* the truth of the proposition
expressed by the latter. We have also seen that a similar semantic
relationship holds between an open yes—no question and the disjunctive
proposition formed by combining the proposition (p) expressed by the
corresponding statement with the negation of that proposition ( ~):
e.g., [s John married? presupposes the truth of the disjunction of ‘‘ John
is married” (p) and “‘John is not married”’ ( ~f).
Mention should also be made, in this connexion, of disjunctive ques-
tions like Are you British or American?, which can be construed either as
restricted x-questions or as open yes—no questions, according to the
context and the nature of the propositions that are put forward in the
disjunction. By a restricted x-question* is to be understood an x-question
in which the set of possible values for x is restricted to those that the
_ speaker actually supplies in his question. If
(16) Are you British or American?
is taken as a restricted x-question, it presupposes the truth of one, and
only one, of the propositions that are put forward by the speaker: i.e.
‘“You are British”’ and “You are American’’. If it is construed as a
yes—no question, on the other hand, it presupposes the disjunction of the
two contradictory disjunctive propositions “‘ You are British or Ameri-
can’ and “‘You are not British or American’’. Generally speaking, the
two kinds of disjunctive questions are kept apart in English by intona-
tion (cf. Quirk et al., 1972: 399).
To investigate and formalize the presuppositions of different kinds of
questions is one of the central concerns of erotetic* logic (cf. Prior &
Prior, 1955; Aqvist, 1965). Another is to decide what constitutes a valid
answer to a question. That these two parts of the logic of questions are
interconnected will be clear from the fact, mentioned above, that either
to assert or to deny the presuppositions of a question is to fail to answer
it. But there are other ways in which one can respond* to a question
without answering it (cf. Hull, 1975). Responses may be appropriate or
inappropriate; and answers, complete or partial, constitute but one of
the subclasses of appropriate responses. We shall not pursue this topic
 16.3. Questions 763
any further, except to point out that J don’t know is generally an appro-
priate response to a question, though it is not of course an answer. Any
utterance which has the effect of qualifying the speaker’s categorical
assertion that something is or is not the case is also an appropriate
response; and the reason why this is so will be made clear in our treat-
ment of epistemic* modality (cf. 17.2).
There is one final point about x-questions that may be mentioned
before we move on. There is no interrogative pronoun meaning “what
entity?” in English (or in any of the more familiar languages). What we
find instead is a distinction drawn between ‘who?’ and ‘what?’ (“‘ what
person?”’ vs. “‘what thing?’’); and it is interesting to note that the same
distinction is drawn in such languages as French, German and Russian,
where it cuts across the distinctions of gender operating elsewhere in the
pronominal system, and also in a language like Turkish (“kim’ vs. ‘ne’),
which has no gender. The distinction between ‘who’ and ‘what’, used
as interrogative pronouns, matches the distinction between ‘someone’
and ‘something’ and leaves the same gap with respect to anything that
is neither a person nor a thing (cf. 11.4). This is obvious enough. Rather
less obvious is the fact that there are other presuppositional differences
between ‘what’-questions and ‘who’-questions, which are independent
of the difference between persons and things. Whereas What did you see?
makes no presuppositions with respect to specificity or definiteness of
reference and may be answered with either John’s new car or A car (and
this may be construed as having either specific or non-specific reference),
Who(m) did you see? cannot be answered appropriately otherwise than by
means of an expression with definite reference. To respond to Who(m)
did you see? with A man, even if this is construed as having specific
reference, is to evade, rather than to answer, the question. It is not clear
how general this difference is, across languages, in the presuppositions
of “‘what person?” vs. ‘‘what thing?’”’. In Turkish, where specificity of
reference is indicated by a special (so-called definite) suffix, there is a
clear distinction between the two versions of ‘“‘What did you see?”’
(‘Ne gérdiin?’ vs, ‘Neyi gérdiin?’), but there is only one version of
‘“Who(m) did you see?’”’ (‘Kimi goérdiin?’). Apparently, Persian 1s like
Turkish in this respect, whereas in Macedonian there are two versions
of ““Who(m) did you see?’’, but only one version of “‘What did you
see?’ (cf. Browne, 1970). It is nonetheless arguable that in all three
languages, as in English, a question introduced with an interrogative
pronoun meaning “what person?”’ is more specific in its presuppositions
than one introduced by an interrogative pronoun meaning ‘what
 764 Mood and illocutionary force
thing?’’; and this is hardly surprising in view of the greater salience, for
human beings, of other individual human beings.
It is interesting to note that the interrogative tags*, which may be
attached to declarative sentences in English, can be accounted for in
terms of a natural extension of the analysis of open questions that we
have outlined above. There are in fact two kinds of interrogative tags
that may be attached to declarative sentences, and their function is
rather different. ‘Tags of the first kind, which may be referred to as copy
tags* (cf. Sinclair, 1972: 75), have the same value with respect to the
distinction of positive and negative as the declarative sentence to which
they are attached: cf.
(17) The door is open, is it?
(18) The door isn’t open, isn’t it??°
Their function is to express the speaker’s attitude (surprise, scepticism,
irony, scorn, etc.) towards the state-of-affairs described by the proposi-
tion expressed by the declarative sentence to which they are attached.
Sentences like (17) and (18) may be used to pose or ask questions, but
they do so indirectly. They are like exclamatory sentences in that they
do not have any characteristically distinct illocutionary force associated
with them.
It is the second type of tags, checking tags*, with which we are pri-
marily concerned here; and these have a more definite and describable
effect upon the illocutionary force of utterances containing them.
Checking tags may be regarded as elliptical interrogative sentences
which, when they are attached to declarative sentences with the same
tropic and expressing the same proposition, produce single sentences
. whose characteristic illocutionary force is that of asking (and not simply
posing) a question. Negative tags are attached to positive sentences, and
positive tags to negative sentences: for example,
(19) The door is open, isn’t it?
(20) The door isn’t open, is it?
What we now have to explain is how (19) and (20) differ in meaning from
(21) Is the door open?
and
(22) Isn’t the door open?
20 Some native speakers of English may be doubtful about (18). I am assuming
that it is grammatical, but there may well be genuine differences of dialect or
idiolect here.
 16.3. Questions 765
Questions like (21) are open* questions in the following two senses:
(i) they are neutral with respect to any indication of the speaker’s beliefs
as to the truth-value of p; and (ii) when they are asked of an addressee,
unless they are given a particular prosodic or paralinguistic modulation,
they convey no information to the addressee that the speaker expects
him to accept or reject p. Their presupposition of the disjunction of p
and ~p is unweighted, as it were, in these two respects. Simple negative
questions, like (22), are also unweighted with respect to the speaker’s
expectation of acceptance or rejection of the proposition, ~p, that is
being questioned; and like open questions they may be merely posed
(cf. ‘‘Isn’t the door open?”’, he wondered; He wondered whether the
door wasn’t open). The speaker utters (22) rather than (21) because
there is some conflict between his prior belief that p is true and present
evidence which would tend to suggest that ~p is true. He questions
~p because it is the negative proposition that occasions his doubt or
surprise.
Utterances formed with checking tags cannot be used merely to pose
questions. Hence the normality of “‘ The door isn’t open, is 1t?”’, he asked
himself, and the abnormality of ‘‘ The door isn’t open, 1s 1t?”’, he wondered.
The reason why this is so is that the function of the checking tag 1s
expressly to solicit the addressee’s acceptance or rejection of the propo-
sition that is presented to him. A sentence like (19), when it is uttered
with its characteristic illocutionary force, puts to the addressee the
positive proposition p (which the speaker is inclined to believe is true
and assumes the addressee will accept), but at the same time explicitly
admits in the tag the possibility of its reyection. Sentence (22), on the
other hand, offers the addressee the negative proposition ~p, which he
is expected to accept as true, but may reject. There is a difference,
therefore, between the two utterances Isn’t the door open? and The door
is open, isn’t it?, although they may both be said to indicate the speaker’s
belief that p is true (cf. Hudson, 1975: 27).
Whether the checking tag is negative or positive, it may have various
intonation-patterns superimposed upon it; and so may the declarative
clause to which it is attached. For our purpose, it is sufficient to dis-
tinguish two patterns: falling (including rising-falling as a subtype) and
rising (including falling-rising). Of these: the falling intonation-pattern
may be regarded (as it generally is in English) as being neutral with
respect to indexical information (cf. 3.1, 4.2). The most neutral realiza-
tion of (19) and (20) has a falling intonation on both the declarative part
and the tag; and any variation of this intonational pattern is indicative of
 766 Mood and tllocutionary force
the speaker’s doubt, surprise, etc. We need not go further into the
details, which are complex and to some extent controversial.
The tag that is added to an imperative sentence, as we have seen, can
be interpreted as an element which explicitly concedes to the addressee
the option of refusal to comply with the mand; and it does so formally by
parenthetically questioning his willingness or ability. With negative
imperatives, whose characteristic function is that of uttering prohibitions
(i.e. commands or requests not to do something), the addressee’s ability
is not at issue, but only his intentions. For that reason, neither Don’t open
the door, can you? nor Don’t open the door, can’t you? is an acceptable
utterance. Furthermore, there is no point in telling or asking someone
to refrain from carrying out some course of action, unless we have some
prior expectation that he will or may do what we want him not to do. It
is therefore only the positive tag, will you?, that may be attached to
negative imperatives: Don’t open the door, will you?, but not Don’t open
the door, won’t you?, is an acceptable utterance, With positive impera-
tives the situation is different, and all four tags may occur: will you?,
won't you?, can you?, can’t you? (cf. Bolinger, 1967a). Of these wll you?
is the most neutral and, unless it is given a particular kind of prosodic or
paralinguistic modulation, it reveals nothing of the speaker’s beliefs
about the addressee’s willingness to comply: its function is like that of an
open question. The negative tags, won’t you? and can’t you?, however,
are like Isn’t the door open? They are used when the speaker is confronted
with some evidence (e.g., the addressee’s initial failure to respond)
which suggests that the addressee is unwilling or unable to comply with
the mand. It is for this reason that they are commonly, though not
necessarily, associated, in utterance, with a prosodic or paralinguistic
modulation indicative of impatience or annoyance.
There is a relationship between questions and one major subclass of
mands, namely requests, that we have not so far mentioned. That
requests are related to questions in the way that commands are related
to categorical assertions is suggested by the fact that in English ques-
tions and requests are reported as acts of asking, but commands and
categorical assertions as acts of telling: cf.
(23) He asked me whether the door was open
(24) He asked me to open the door
(25) He told me to open the door
(26) He told me that the door was open.
Similarly in French:
 16.3. Questions 767
(27) Il m’a demande si la porte était fermée
(28) Il m’a demandeé de fermer la porte
(29) Il m’a dit de fermer la porte
(30) Il m’a dit que la porte était fermée.

In most of the Indo-European languages different verbs are used for

reporting questions and requests, on the one hand, and commands and
statements, on the other. ‘This may or may not be taken as evidence that
the distinction between asking and telling is language-specific. We will
pursue the hypothesis that it is a universally applicable distinction: that
asking and telling are two distinguishable subtypes of saying.
We have already suggested that categorical assertions and commands
(and demands) contain the same unqualified I-say-so component, but
that they differ in their tropic (“‘it 1s so”’ vs. “so be it’’); and that posing
a yes—no question has the effect of qualifying the I-say-so component by
expressing the speaker’s inability to assign a truth-value to the propo-
sition expressed by the sentence used to pose the question. It may now
be suggested that requests contain the same neustic as questions (and
the same tropic as commands). In order to make a case for this analysis,
we must clearly look for some more general interpretation of their
alleged common neustic component.
As we have seen, a yes—no question presupposes the disjunction of a
proposition and its negation: the speaker admits the possibility that
either p or ~ is true. If he asks the question of an addressee, he does
so, normally, with the expectation that his addressee will assign a
truth-value to p by accepting or rejecting it. When the speaker issues a
request, he explicitly admits the possibility that the addressee may or
may not comply with the mand and thus, by his response, make p true
or refuse to make p true. Unlike questions and requests, categorical
assertions and commands do not explicitly leave to the addressee the
option of acceptance or rejection, though the addressee may in fact deny
the speaker’s assertion, just as he may refuse to comply with a command.
It is the option of acceptance or rejection that we propose to identify
with the act of asking.
We cannot, however, say that a request is related to a command in the
same way that an open yes—no question is related to a categorical asser-
tion. It is clearly unsatisfactory to analyse requests as meaning nothing
more than “Is it to be so that p?’’. In fact, this is the kind of analysis
that we require for deliberative questions: when the speaker poses or
asks a deliberative question (What am I to do?, Shall I open the door?,
768 Mood and illocutionary force
etc.) he expressly withholds his commitment to the desirability or
necessity of the course of action which would make p true. When he
issues a request, however, he commits himself, by virtue of the speaker-
based sincerity condition which governs all mands, to the desirability
of the proposed course of action. Requests are related to commands as
non-open* yes—no questions are related to categorical assertions. To
make the point rather crudely: The door ts open, isn’t 1t? means something
like “I think that “The door is open”’ is true: but I concede your right
to say that it is not true’’; and Open the door please means “‘I want you
to make “The door is open”’ true: but I concede your right not to make
it true’’.24 Another way of making the point is to say that in non-open
questions and requests the speaker indicates his own commitment to the
it-is-so or so-be-it component of the utterance and invites the addressee
to do the same. The addressee’s commitment to the so-be-it component
of a request may be expressed in either of two ways: (i) he may carry
out the course of action that is proposed and thereby make p true at
some later time; (11) he may say yes (or something equivalent to this),
and his utterance will count as a promise, rather than a prediction.
The analysis of questions and requests that has been put forward in
this section, it must be emphasized, is by no means a standard or
generally accepted analysis. But it does have the advantage that it seems
to give a more satisfactory account of the difference between acts of
asking and acts of telling than the alternative analysis, mentioned above,
according to which questions and requests constitute two unrelated
subclasses of mands. As we shall see in the following chapter, there is
also a relationship between factual questions and epistemically modalized
utterances that must be accounted for; and a further, and quite different,
relationship between mands and deontic modality.

16.4. Negation ,
So far we have treated the assertion of a negative proposition (‘‘it is the
case that not—p’’) as equivalent to the denial of the corresponding
positive proposition (‘it 1s not the case that p’’). Both of these are
symbolized in the propositional calculus as ~p (cf. 6.2). As soon as we
start considering propositions containing a modal operator of possibility,
however, it becomes clear that a distinction needs to be drawn between

21 Strictly speaking, one should draw a distinction between wanting the door to
be open and wanting ‘‘The door is open”’ to be true (and also, though less
obviously, between causing the door to be open and causing ‘‘The door is
open” to be true). But the distinction is not relevant to the argument.
 16.4. Negation 769
the negation of the modal operator and the negation of a simple propo-
sition within the scope of the modal operator (cf. 6.3): e.g., ~nec p (‘it
is not necessary that p’’) vs. nec ~p (“it is necessary that not—p’’).
There is also a clear difference of meaning in utterances which result
from the negation of a performative verb and the negation of its comple-
ment:
(1) Lf don’t promise to assassinate the Prime Minister
(2) I promise not to assassinate the Prime Minister.
It is only (2) that can be said in the performance of the illocutionary act
of promising; and in this case it would be a promise to refrain from
doing something. Utterance (1), on the other hand, might be a statement
with which the speaker explicitly refuses to make, or denies that he is
making, a promise. Assertion, like promising, is an illocutionary act; and,
as (1) differs from (2), so the statement
(3) The door zs not open
differs from the statement
(4) I do not say that the door is open.
There is no way of representing this difference in the propositional
calculus, which does not allow for the negation of the assertion-sign.
In modal logic, the difference between the negation of the modal
operator and the negation of the proposition within the scope of the
modal operator is commonly referred to in terms of a difference between
external and internal negation; and the difference between (1) and (2),
or between (3) and (4), has been described in the same terms (cf. Hare,
1971: 82). It is possible, of course, to negate both the performative verb
and the complement, as in
(5) I do not promise not to assassinate the Prime Minister
(6) I do not say that the door 1s not open.
And, just as ~nec ~? is not equivalent to p (but to poss p) so (5) and
(6) are not equivalent to I promise to assassinate the Prime Minister and
The door is open. In other words, if one negative is external and the
other internal, two negatives do not make a positive. The relationship
_ between I do not say that the door 1s not open and The door is open is
therefore different from the relationship that holds in the propositional
calculus between ~ ~p and p.
This much about negation is relatively uncontroversial. But it is
770 Mood and illocutionary force
arguable that there are at least two kinds of external negation; and that
the difference between them can be accounted for by assigning one to
the neustic and the other to the tropic. It is negation of the neustic (the
I-say-so component) that is exemplified by (1) and (4). We will refer to
this kind of negation as performative negation*. By negating the neustic
we express our refusal or inability to perform the illocutionary act of
assertion, promising, or whatever it might be. But to do this is in itself
to perform an illocutionary act: an act of non-commitment. Acts of
non-commitment are to be distinguished, on the one hand, from saying
nothing and, on the other, from making descriptive statements. Con- ,
sider, for example, the circumstances under which it might be appro-
priate to utter a declarative sentence like
(7) I don’t say that John is a fool.
If someone has asserted or implied that X goes around saying that John
is a fool, X can deny that this is so by uttering (7) as a descriptive
statement; and in this case don’t will bear heavy stress, as it normally
does (as we shall see presently) in denials*. X’s utterance-act might then
be reported by means of another descriptive statement such as
(8) X said that he didn’t/doesn’t say that Fohn is a fool.
But X might also utter (7), without stressing don’t, not to deny that he
goes around making a particular assertion, but to express, or indicate,
his refusal, on this very occasion, to put his signature, as it were, to
“IT say so’’; and this is a positive act, which might be reported by
(9) X wouldn’t/couldn’t say that/whether Fohn was a fool,
rather than by (8) or
(10) X didn’t say that Fohn was a fool.
The theory of speech-acts, as it has been developed so far, does not seem
to allow for acts of non-commitment. They are nonetheless of frequent
occurrence in the everyday use of language; and their perlocutionary
effect is characteristically different from that of statements. If we express
our refusal to assert that p is so, by means of an act of non-commitment,
we will often create in the mind of the addressee the belief, which he did
not previously hold, that p may in fact be true;?* especially if the situa-
tion is such that we might be expected to assert that not—p 1s so.
22 We might even be held to invite the inference that p is true: cf. Zwicky &
Geis (1971) for the cases of what they call invited inference.
 16.4. Negation 771
When we negate the tropic (the it-is-so component) of a statement,
we are still making a statement; but it is a different kind of statement
from that which is made by negating the phrastic. We are denying* that
something is so: we are rejecting a proposition (either positive or nega-
tive) that we might be expected to accept as true. One reason for
believing that we are expected to accept that p holds might be that our
interlocutor has just explicitly presented it to us in a statement or a non-
open question: but this is not the only reason. The situation may be
such, or the proposition itself of such a nature (a generally accepted
truism), that there need be no previous assertion to which our denial is
linked. Nevertheless, the proposition that is accepted or rejected will
always be, in some sense, in the context. It seems reasonable therefore
to draw a distinction between context-bound* and context-free* state-
ments, and to account for the difference between the denial of p and the
simple assertion of not—p in terms of this distinction. We will say that
to reject p by means of a context-bound statement is to deny* that p
holds and to accept p by means of a context-bound utterance is to con-
firm* that p holds. In either case p may be positive or negative. Denials*
and confirmations* are thus the two major subclasses of context-bound
statements.
Just as a negative sentence may be uttered to deny a positive propo-
sition, SO a positive sentence may be uttered to deny a negative proposi-
tion. But this does not invalidate the distinction between the negation of
the tropic and the negation of the phrastic that we have just drawn. If
(3) is uttered as a denial, it will bear heavy stress on the negative particle
(and, if not is contracted with the verb, upon isn’t). If
(11) The door 1s open
is uttered as a denial of the proposition expressed by the context-free
assertion of (3) it will have a heavy stress on the form zs. So too if it is
uttered to confirm, rather than simply to assert, that the door is open.
The distinction between context-free and context-bound statements is
systematically maintained, in English, by stress; it is not generally
represented in the verbal component of utterances.
The distinction between negation of the phrastic, which we will
henceforth refer to as propositional negation*, and negation of the tropic,
which we will call modal negation*, is something that must be taken
account of in the treatment of presupposition. According to what is
probably the most generally accepted criterion for at least one class of
presuppositions (cf. 14.3), one proposition, p, is said to presuppose
172 Mood and tllocutionary force
another proposition q, if q is entailed, or strictly implied, by both p and
its negation: 1.e. if both p entails g and ~~? entails g, then p presupposes
q (cf. Keenan, 1971). Granted that the context-free assertion of a propo-
sition, p, or its propositional negation, ~p, commits the speaker to a
belief in the truth of any proposition, qg, that is presupposed by , this
does not hold for the denial of p. If someone were to assert that the
present King of France is bald, we could quite reasonably deny this by
saying:
(12) The present King of France is not bald: there is no King of France.
Similarly, though it would be irrational for someone to utter
(13) L don’t know that the earth ts round
as a context-free assertion (‘know’ being a factive* verb which normally
presupposes the truth of the proposition expressed by its complement:
cf. 17.2), there is nothing wrong with the utterance of (13) as a denial.*8
It might further be argued that there are two kinds of propositional
negation: one of which converts the proposition into its contradictory*
and the other into its contrary*. The logical distinction between contra-
dictories and contraries has already been referred to, in a previous
chapter, in connexion with the difference between gradable and un-
gradable opposites (9.1). An utterance like
(14) I don’t like modern music
is more likely to be interpreted as expressing a proposition that is the
contrary, rather than the contradictory, of
(15) “I like modern music’’.
But both interpretations are possible. Since the conversion of a propo-
sition into its contrary seems to depend upon the negative being more
closely associated with the predicate, than with the subject—predicate
23 As was noted earlier (14.5), tokens of (13) may also be interpreted as meaning
‘‘I am not sure that the earth is round”’ or even “‘I am inclined to believe
that the earth is not round”’. In such cases don’t will not normally bear heavy
stress (though ts may if the utterance is context-bound). It is suggested later
in this section, in the discussion of (29), that examples like (13), as well as
examples containing such verbs of propositional attitude as ‘think’ and
‘believe’ (which are normally handled in terms of the notion of transferred
negation), are perhaps best analysed as containing performative negation.
There is a close parallel between the negation of a performative verb as in (1)
or (4) and the negation of a parenthetical* verb (cf. 16.1); and the same general
notion of positive non-commitment seems to be relevant to both.
 16.4. Negation 773
link, or nexus* (to use Jespersen’s term), within the proposition, we
might appropriately distinguish the two kinds of propositional negation
as predicative negation* and nexus negation*. Both types of negation are
found in
(16) I don’t not like modern music,
which is equivalent to
(17) I don’t dislike modern music.
It seems to be the case that the application of propositional negation to a
gradable expression (e.g., ‘like’) will always tend to produce a contrary,
rather than a contradictory, whether the language-system lexicalizes the
contrary (e.g., ‘dislike’) or not.
If we recognize two kinds of propositional negation, as suggested in
the previous paragraph, we are going against the traditional logical
maxim that negation of the predicate is equivalent to negation of the
proposition. Furthermore, it might be maintained that utterances like
(16) and (17) would rarely, if ever, occur as context-free assertions; and,
if this is so, there is no need to distinguish nexus negation from modal
negation. Four kinds of negation may well be too many, but it does seem
clear that at least three kinds are required in order to handle utterances
with an it-is-so tropic (1.e. statements and factual questions).
One may now ask whether the same three kinds of negation are to be
found in utterances with a so-be-it tropic. There is of course a clear
difference between explicitly refusing to tell someone to do something
and telling him not to do something. We therefore need to allow for at
least two kinds of negation: performative negation and either modal or
propositional negation. One of the points that we shall have to deal with
in our discussion of deontic modality* is how acts of non-commitment
with a so-be-it tropic are related to permissions* and exemptions* (cf.
17.4). For the present, however, the following will serve as an analysis
of the meaning of utterances in which the speaker explicitly refuses to
commit himself to the imposition of an obligation: “‘I do not say so — so
be it — (that) p”’.
The main problem that concerns us here, however, is to decide
whether prohibitions, like
(18) Don’t open the door,
contain a negated tropic or a negated phrastic. And this is a tricky
problem. Does (18), construed as a prohibition, mean “I (hereby) im-
pose upon you the obligation not to make it so that p holds”’ or “I
774 Mood and tllocutionary force
(hereby) impose upon you the obligation to make it so that not—p
holds’’? As we have already seen, prohibitions are not normally intended
or taken as instructions to carry out, but to refrain from carrying out,
some course of action. The speaker does not want the addressee to bring
about a state-of-affairs of which not—p is true; for this state-of-affairs
already exists. The reason why he issues his prohibition is that he thinks
that, in default of the prohibition, the addressee will, or may, bring
about a state-of-affairs of which p (the contradictory of not—p) would be
true. It seems preferable, therefore, to treat prohibitions as having a
negated tropic: i.e. as resulting from modal negation. This analysis, it
should be noted, groups prohibitions with denials, rather than with
context-free negative assertions. It also suggests that corresponding
positive and negative jussive sentences (e.g., ‘Open the door’ and
‘Don’t open the door’), in their most characteristic use, will be con-
strued as contradictories, rather than contraries; and, in general, this
seems to be correct.
At the same time, it must be recognized that negative jussive sen-
tences, used to issue prohibitions, are sometimes intended and taken as
directives to bring it about that not—p holds. This is especially clear
when not—p? is the contrary, rather than the contradictory, of p. For
example,
(19) Don’t trust him
can be understood as being the contrary of
(20) Trust him,
equivalent to
(21) Distrust him.
If the difference between contradictory and contrary statements 1s
associated with the distinction between modal and propositional nega-
tion, we should perhaps allow for propositional negation in directives as
well as statements. Generally speaking, however, prohibitions will
involve modal negation; and they are to be analysed as ‘I say so — let it
not be so — (that) p”’, rather than as “‘I say so — so be it — (that) not—p”’.
Two further points may now be mentioned, each of which has been
the subject of considerable discussion recently. The first is the inter-
action between negation and information-focus* (cf. 12.7). When con-
trastive stress occurs elsewhere than at the end of the clause in English,
this is an indication that the clause has marked, rather than neutral,
 16.4. Negation 775
information-focus; and, if this happens in a negative sentence, the
sentence will tend to be interpreted (on particular occasions of its
utterance) as carrying rather specific presuppositions (cf. Quirk e¢ al.,
1972: 382; Jackendoff, 1972: 252). For example, the utterance Mary
didn’t come (with contrastive stress on Mary) is comparable with Jt was
not Mary (but someone else) that came: but the two utterances are not
equivalent, since it is also possible to say J don’t know whether anyone
came (but) Mary didn’t come (i.e. it is possible to cancel what is normally
taken to be a presupposition of Mary didn’t come, the proposition
‘Someone came”’: cf. 14.3). It is obvious that stress, and also intona-
tion, must be taken into account in any comprehensive discussion of
negation; and, when they are taken into account, further complexities
become apparent. The question of information-focus is clearly of some
relevance to the notion of modal negation; but the connexion between
them is, in certain respects, obscure.
The second point has to do with what has been called transferred
negation* (cf. Quirk et al., 1972: 789). The term suggests (as do
‘negative transportation’ or ‘negative raising’ — which are the terms
most commonly used in Chomskyan transformational grammar) that the
phenomenon to which it applies involves some kind of displacement of
the negative operator. For example,
(22) L didn’t think he would do it
is rightly said to have two interpretations: roughly, (i) “It is not the
case that I thought he would do it”’ and (ii) ‘‘I thought that he would
not do it”. The former of these is unusual, however, in colloquial
English and almost inevitably requires heavy stress on didn’t: and this
fact suggests that it is a modal negation that is involved. Under the
second interpretation, which with normal stress and intonation is by far
the more common, (22) is generally said to be equivalent to
(23) I thought (that) he would not do it
and to be derivable, in a transformational grammar, from the same
underlying structure, the negative element having been transferred from
the subordinate to the main clause.
Whether (22) and (23) are equivalent is a moot point. It is important
to notice, in this connexion, that the construction in question (which is
to be found in many languages) is especially common with verbs
denoting belief and assumption; and furthermore that it is far more
common with first-person subjects than it is with third-person and
and |
776 , Mood and illocutionary force
second-person subjects. What is interesting about (22) is that, although
it has a first-person subject, it is in the past tense. In this respect it
stands mid-way between
(24) L don’t think he'll do it

(25) She didn’t think he would do 1t.

Now (25) is much closer, semantically, to
(26) She thought he wouldn't do it
than (24) is to
(27) I think (that) he won't do tt.
And the reason is that, under the most normal interpretation of (24) and
when don’t is unstressed, the verb ‘think’ is not being used descriptively,
in a constative utterance (cf. 16.1). As in
(28) I think he'll do tt,
it is being used to qualify the speaker’s commitment to the truth of “He
will do it”. Even more striking than (24) is the colloquial, but by no
means unusual, use of
(29) I don’t know that he'll do it
to mean, roughly, “I am inclined to believe that he won’t do it’’. (29)
cannot be accounted for in terms of transferred negation.
What all this suggests is that (24) exemplifies something that is closer
to, if not identical with, performative negation. Looked at from a
purely semantic point of view, it has an obvious connexion with subjec-
tive epistemic modality (cf. 17.2). However that may be, the notion of
transferred negation is by no means as straightforward as it is frequently
supposed to be.
Indeed, it might be suggested, in conclusion, that the notion of nega-
tion itself is far from being as straightforward as it might appear to be
at first sight. Much of the research that has been devoted recently to the
study of negation in natural language has, not surprisingly, taken as its
starting-point that kind of negation (propositional negation, in our
terms) which is formalized in the propositional calculus and defined in
the associated truth-tables as the difference between the values of p and
~p (cf. 6.2). Our own treatment of negation in this section might be
 16.4. Negation , a7
justifiably criticized, in this vein, for using a single negation operator, in
different positions or with variation of scope, to account for the several
kinds of negation that we have discussed. It is worth noting, therefore,
that in many languages there are several different kinds of negative
sentences, often with different negative particles, and that, if it were not
for a prior commitment to the belief that propositional negation is basic, ,
we might not be inclined to treat these several kinds of negation under
the same rubric. What is or 1s not basic is, as always, a thorny question.
But if we interpret ‘basic’ to mean ‘acquired earlier by children and
serving as the basis for further development”’, it seems fairly clear that
propositional negation is not basic. |
The following four kinds of negation have been identified by scholars
working in the field of language-acquisition (cf. Brown, 1973: 17): (1)
non-existence; (11) rejection; (111) refusal to comply; (iv) denial. What 1s
called non-existence is perhaps best described as absence or disappear-
ance: this fits the data, and it is less suggestive of propositional negation.
As for the other three kinds of negation, they can be much more satis-
factorily accounted for in terms of the more general notion of rejection
than they can be in terms of the logician’s notion of negation, definable
with reference to truth and falsity. As one rejects some physical entity
that is offered (pushing it away so that it disappears or goes away: cf.
(i)), So one may reject a proposition or a proposal. Looked at from this
point of view, modal negation would seem to be more basic than propo-
sitional negation; and assent and dissent, rather than truth and falsity,
would seem to be the notions with which we should operate in any
account that we give of the difference between the assertion and the
denial of p. If this point of view is accepted, propositional negation may
be seen as developing out of modal negation, in much the same way
as objective epistemic modality develops out of subjective epistemic
modality (cf. 17.2). In this connexion, it should not be forgotten that, as
speakers of English, we are tempted to interpret Yes and No as meaning
“That is so”? and “That is not so’’. Not all languages, however, have
forms of assent and dissent that can be interpreted in this way. For
example, Da and Njet in Russian are much more satisfactorily interpre-
ted, not as meaning “No” and “Yes”’, but “I accept (what you assert
or imply)” and “I reject (what you assert or imply)’’. Hence, the
possibility of saying what is often mistranslated into English as Yes, it’s
not raining.
 778 Mood and illocutionary force
16.5. The performative analysis of sentences
What has been presented in the previous-sections of this chapter, it
should be emphasized, is an analysis of the structure and meaning of
utterances, not sentences. However, there is commonly some corres-
pondence between the grammatical and lexical structure of sentences
and their characteristic illocutionary force. For example, explicitly per-
formative utterances in English will normally have, as the complement
of a verb of telling or asking, an embedded finite clause (e.g., that the
door is open, whether the door is open) when, under our analysis, they
contain an it-is-so tropic and an embedded infinitive or subjunctive
clause (e.g., fo open the door, that the door be open) when they contain a
so-be-it tropic. So too will statements which describe or report illocu-
tionary acts. And in both cases whether it is a verb of telling or a verb of
asking that is used will be determined by the occurrence of an I-say-so
or an I-can’t-say-so component in the neustic of the illocutionary act
that is being performed or reported.
In view of the structural correspondence that exists, in many lan-
guages at least, between the sentences used to make explicitly performa-
tive utterances (e.g., ‘I tell you that the door is open’, “I tell you to open
the door’) and the sentences used to describe or report illocutionary acts
(e.g., ‘I told you that the door is open’, ‘I told you to open the door’),
regardless of whether these illocutionary acts have been performed by
means of primary performatives (e.g., T’e door 1s open, Open the door) or
explicitly performative utterances (e.g., I tell you that the door is open, I
| tell you to open the door), it is natural to consider the possibility of deriv-
ing all sentences from underlying structures with an optionally deletable
main clause containing a first-person subject, a performative verb of
saying and optionally an indirect-object expression referring to the
addressee. That the grammatical and semantic structure of all sentences
should be accounted for in terms of the embedding of a subordinate
clause within an outer, or higher, performative main clause has been
proposed, independently, by Boyd and Thorne (1969), Householder
(1971), Lakoff (1969), Ross (1970), Sadock (1974), and others; and it is
this proposal that we are referring to as the performative analysis of
sentences. A more restricted version of the performative analysis
according to which only non-declarative sentences are to be accounted
for in terms of the embedding of a subordinate clause as the complement
of a deletable performative verb of saying has been put forward by such
scholars as Lewis (1972).
 16.5. The performative analysis of sentences 779)
The performative analysis of sentences, it should be obvious, would
tend to invalidate the distinction between system-sentences and utter-
ances, which we have tried to maintain, as consistently as possible,
throughout this work. It would also destroy, or at least lessen the
importance of, Austin’s distinction between primary performatives and
explicitly performative utterances (cf. 16.1). These would be merely
alternative surface-structure realizations of the same, underlying deep
structures, or semantic representations (cf. 10.5); and every sentence
that can be used to perform a variety of distinct illocutionary acts would
be provided by the rules of the grammar with a set of non-equivalent
deep-structure analyses. For example ‘I'll be there at two o'clock’
would be shown, presumably, as grammatically ambiguous according to
whether it is held to be equivalent in meaning to ‘I (hereby) promise to
be there at two o’clock’ or ‘I (hereby) state/predict that I’ll be there at
two o'clock’. And ‘I promise to be there’ will be shown as gram-
matically ambiguous according to whether its utterance constitutes an
act of promising (allowing for the insertion of ‘hereby’) or a statement
descriptive of some habitual action that the speaker performs. The dis-
tinction of system-sentences and utterances is of course a methodological
distinction: it is not one that has to be maintained at all costs. The ques-
tion is whether we can more satisfactorily describe the structure and use
of language by drawing it than by not drawing it; and this is a question
that is difficult, if not impossible, to answer until the performative
analysis of sentences has been developed in greater detail and applied to
a wider range of data.
There are certain considerations, however, which would suggest that
the performative analysis of sentences, as it has been formulated within
the framework of transformational grammar, is ultimately indefensible.
First of all, it should be noted that the conditions which determine the
selection of a subject expression and an indirect-object expression in the
text-sentence (S) containing the performative verb of saying in an ex-
plicitly performative utterance are quite complex; and they will vary
according to the nature and occasion of the illocutionary act that is being
performed. The speaker can use any expression that 1s appropriate to
refer to himself as the performer of the illocutionary act in question; and
this need not be, and for certain socio-culturally and ritualized acts
cannot be, the first-person pronoun. Similarly, he can use any expression
that is appropriate to refer to the addressee; it need not be a second-
person pronoun. It will not therefore be sufficient to formulate the rules
in such a way that the subject of the performative verb in English is
780 Mood and tllocutionary force
necessarily ‘1’ (or ‘we’) and its (optional) indirect object ‘you’. To say
that ‘the chair’ is a surface-structure replacement for a deep-structure
first-person pronoun in ‘The chair (hereby) rules that the last speaker
was out of order’ is surely misguided. There are circumstances when it
is appropriate for the speaker to refer to himself, in English, as ‘I’ and
there are circumstances in which it is appropriate or mandatory for him
to use another expression. In this connexion, it is worth noting that in
reporting or describing a mental or illocutionary act involving reference
to himself as the addressee, the speaker will normally use a first-person
reflexive pronoun (e.g., [ told myself not to be a fool), but in performing
the act in question he will address himself normally in the second person
(e.g., Don’t be a fool). Furthermore, whereas he will address himself in
the second person in a factual question that he puts to himself, he will
use the first-person in a corresponding deliberative question (cf. What
are you going to do? vs. What am I to do?). It is hard to see how these
differences can be accounted for within the framework of the performa-
tive analysis of sentences, which presumably treats the relationship
between a performative verb and its indirect object as being the same as
the relationship between a descriptive verb and its indirect object.
The second point, and it is connected with the first, is that the per-
formative analysis of sentences explicates our understanding of the
meaning of the underlying performative verb in terms of our prior
understanding of the meaning of verbs of saying as they are used to
describe or report illocutionary acts. But the whole purpose of Austin’s
original distinction of performative and constative utterances was to
establish the difference between engaging in an act and describing it;
and, as we have seen, he eventually came to the view that saying is a
kind of doing (16.1). It is arguable, though philosophically controversial,
that the performance of certain basic communicative acts is logically
independent, not only of the existence of language, but also of any con-
ventionalized system of communication (cf. 16.1). However that may be,
it seems perverse to assimilate the performative function of verbs of
saying to their descriptive function, rather than to assume that we come
to know the sense and denotation of verbs of saying by virtue of our
prior understanding of what is involved in the performance of illocu-
tionary acts; and that we acquire this understanding, just as we acquire
our understanding of how deixis operates, by engaging in communica-
tive acts and learning how particular language-systems conventionalize
the means for. referring to ourselves, our addressees and other com-
ponents of the situation-of-utterance.
 16.5. The performative analysis of sentences 781
None of the syntactic arguments so far produced in favour of the
performative analysis of sentences is very convincing; and much of the
data that has been cited in arguing for it is questionable (cf. Sadock,
1974). What the syntactic arguments show is that we can insert in pri-
mary performatives and explicitly performative utterances expressions
which refer to, or qualify in some way, various components of the illocu-
tionary act that we are performing. But this fact in itself does not force
us to recognize a superordinate performative clause in all sentences. In
fact, the standard performative analysis is clearly unsatisfactory in its
representation of the relationship between the performative clause con-
taining the verb of saying and the complement clause. The relationship
between the performance of an illocutionary act and the sentence that is
uttered is an instrumental relationship of some kind, as the possibility of
inserting ‘hereby’ would indicate. It is by means of the very utterance-
signal that we are producing that we perform the act of assertion, com-
manding, promising, etc. But, as we saw earlier, there are two different
senses of ‘say’ involved here: ‘‘say,’’ and “‘say,”’ (cf. 16.1). Given that S
is the explicitly performative statement I (hereby) tell you that the door is
open and that S’ is the embedded sentential complement of the per-
formative verb, the meaning of S would seem to be something like
‘(By saying, S) I say, “S’” (to you)”, where “S’”’ (i.e. the meaning of
S’) is “that (it is the case that) p’’. The optionally insertable ‘hereby’
(which has been glossed as “‘by saying, S’’) is naturally construed
as a deictic element referring either to S or to what Austin would dis-
tinguish as the locutionary act of saying S (cf. 16.1). The performative
verb, on the other hand, refers to the illocutionary act that is performed
in saying S. It does not seem right therefore to treat simple sentences
like ‘The door is open’ as transforms of the subordinate clause of
a deleted performative verb. They differ from explicitly performative
utterances in that they lack the property of token-reflexivity (cf. 1.4);
and they differ from the complements of descriptive verbs of saying in
that they do not mean “‘(that it is the case) that p”’.
The fact that explicitly performative utterances have the property of
token-reflexivity makes them rather difficult to handle from a logical
point of view. They are like such notorious utterances as \ What I am
now saying is false* or \ This sentence contains five words”. Utterances
of this kind have been analysed in various ways by philosophers; and
there is still no generally accepted solution to the logical paradoxes, or
antinomies, to which they give rise in the formalization of semantics.
Towards the end of the first section of this chapter (16.1), we intro-
782 Mood and tllocutionary force
duced the notion of parenthetical modulation. It was suggested that the
function of the parenthetical tag-clause J promise you in the utterance
Pll be there at two o'clock, I promise you was to confirm, and make
explicit, the speaker’s commitment to the illocutionary force of I'll be
there at two o'clock. The parenthetical that’s an order in Be here at two
o'clock: that’s an order has the same function. The grammatical structure
of I'll be there at two o'clock, I promise you and Be here at two o'clock:
that’s an order is, superficially at least, quite different from the gram-
matical structure of I promise you (that) I'll be there at two o’clock and
I order you to be here at two o'clock. The tag-clauses do not have the
property of token-reflexivity, and it seems gratuitous to derive them from
deep-structure main clauses. Now, it is undeniable that to be here at two
o'clock is part of the grammatical complement of the verb ‘order’ in the
explicitly performative I (hereby) order you to be here at two o'clock as it
is in the statement I ordered him to be here at two o'clock; and it is
possible that both of them should be derived by embedding an impera-
tive sentence (.S’) within a deep-structure main clause (S) containing a
verb of saying of a particular subclass (‘order’, ‘tell’, ‘command’,
‘request’, ‘ask’, etc.).
It is also possible, however, that the surface-structure status of a
performative main verb should be accounted for by a grammatical rule
which operates upon two juxtaposed, or paratactically* associated,
clauses neither of which is subordinate to the other in deep structure.
One of the semantic effects of this rule would be to produce the peculiar
property of token-reflexivity. But it would not destroy the parenthetical
character of the performative clause; and this is an important point that
must be accounted for. The illocutionary force of I promise you that I'll
be there and I'll be there, I promise you seems to be the same; and the
relationship between I promise you and the act of promising is in both
cases the same. Current versions of transformational grammar operate
with processes of subordination and co-ordination, but they have no way
of formalizing what is traditionally known as parataxis*: i.e. a looser
syntactic association of the constituents of a sentence than co-ordination.
There are, however, good semantic reasons for distinguishing parataxis
from subordination and co-ordination. |
One of the syntactic arguments that has been adduced in favour of the
performative analysis of sentences, and possibly the strongest, is that it
enables us to account for the function of at least a subclass of so-called
sentence-adverbs, like ‘frankly’ and ‘honestly’, which can be said to
modify the optionally deletable performative verb in statements. Granted
 16.5. The performative analysis of sentences 783
that the following two statements have the same meaning (and this is in
itself debatable)
(1) I tell you frankly (that) he’s a fool
and

(2) Frankly, he’s a fool,

what, in fact, is explained by the performative analysis of the gram-
matical structure of (2)? Notice, first of all, that if (2) is denied by
uttering
(3) That's not true,
the person uttering this denial is not challenging the frankness or honesty
of his interlocutor, but the assertion that the referent of ‘he’ in (2) is a
fool. The situation is less clear cut if (3) is uttered in response to (1).
What is clear, however, is that (3) cannot be used to deny the fact the
utterer of either (1) or (2) is performing an act of telling. If (3) is uttered
in response to

(4) Anne told Mary frankly that he was a fool

it may be construed as a denial either that an act of telling was per-

formed or that the person performing this act was frank in doing so; and,
under either of these interpretations, (3) may be followed in the same
utterance by She didn’t, It may also be construed as denying the fact,
which is not asserted by (4) (and to whose truth the speaker of (4)
makes no commitment), that the referent of ‘he’ is, or was, a fool: in
which case it may be followed by He isn’t (or He wasn’t). Only He tsn’t
(and not You don’t) can be added to (3) as a response to either (1) or (2).
The second point to be made in connexion with the difference
between (1) and (2), on the one hand, and (4) on the other, is that (4) is
not a natural description of an illocutionary act of the kind that would be
performed by uttering either (1) or (2). The person saying (4) is asserting
that Anne was frank in the manner in which she performed the act of
telling (whatever sentence, or sentences, may have been used in the per-
formance of this act). If the person uttering (1) or (2) is making any
claim to frankness at all, it is best construed as a parenthetical comment
on his willingness to make a blunt, unqualified statement or as an appeal
to the addressee to accept his opinion. If we substitute ‘honestly’ or
‘actually’ for ‘frankly’ in (1), (2) and (4), these differences are even
 784 Mood and illocutionary force
clearer. Furthermore, there are adverbs and adverbial phrases (e.g., to
tell you the truth, as a matter of fact) which can be used in place of
‘frankly’ in (2) and which appear to have a similar function, but which
cannot be used to modify either a performative or a descriptive verb of
saying. The performative analysis of sentences contributes nothing to
our understanding of the function of these.
What these various arguments show is that there are serious semantic
objections to the postulation of a deleted main clause in the underlying
structure of primary performatives. They also show that, from a seman-
tic point of view, the function of the main clause in an explicitly per-
formative utterance is, in some sense, parenthetical; and that the
relationship between a performative verb and its modifiers appears to be
different from the relationship between a descriptive verb of saying and
its modifiers.
: One of two conclusions can be drawn. The first is that, if the per-
formative analysis of sentences is to be made semantically revealing, it
must somehow capture the notions of parenthetical modulation and
token-reflexivity: this is not done in current versions of the performative
analysis.
The alternative conclusion, which is more in accord with traditional
conceptions of the nature and limits of grammatical analysis, is that there
is a distinction to be drawn between the grammatical structure of a
sentence and its meaning, on the one hand, and a different distinction to
be drawn between the meaning of sentences and the meaning of utter-
ances, on the other. Nothing that has been said in this section forces us
to draw one conclusion rather than the other. But the performative
analysis of sentences, as currently formulated, is clearly unacceptable;
and it has not yet been demonstrated that an amended version would in
any way simplify the task of linguistic analysis or solve any problems
that are insoluble within a more traditional framework. We will continue
to draw a distinction, therefore, between the grammatical structure of
sentences and the logical, or semantic, structure of the speech-acts that
these sentences may be used to perform.*4
In conclusion, something must be said about so-called indirect speech-
acts*. It has been proposed by Gordon and Lakoff (1971), on the basis
of suggestions made by Searle (1969; cf. also 1975), that mands may be
issued, not only directly by uttering a jussive sentence (e.g., Open the
24 References to the literature relating to the performative analysis, especially as
it was developed by Ross (1970), are to be found in several of the articles in
Cole & Morgan (1975). There is a full critique in Gazdar (1976).
 16.5. The performative analysts of sentences 785
door), but also indirectly, by (i) asserting that a speaker-based sincerity
condition holds (e.g., [ want you to open the door) or (ii) questioning
whether an addressee-based sincerity condition holds (e.g., Can you
open the door?). 'This proposal has been generalized and extended, with
certain modifications, to other kinds of illocutionary acts by Heringer
(1972). As Heringer points out, one may indirectly perform an illocu-
tionary act, not only by questioning, but also by asserting, an addressee-
based condition: so that, not only Can you open the door?, but also You
can open the door, may be used as a mand.
For the purposes of the present section, there are just two points to
be made in connexion with the notion of indirect illocutionary acts. The
first is that it blurs the distinction between the meaning of a sentence
and the illocutionary force of an utterance; or rather, it introduces the
possibility that an utterance may have two kinds of illocutionary force,
which we may refer to as its actual and its incidental illocutionary
force. For example, I want you to do it may be meant incidentally as a
statement, but actually as a request — “‘a request made by way of
making a statement”’ (cf. Searle, 1975: 59). The incidental illocutionary
force of an utterance is directly determined by the grammatical structure
of the sentence that is used in making the utterance; the actual illocu-
tionary force of an utterance is derivable from the meaning of the sen-
tence and its incidental illocutionary force, according to the principles
discussed by Gordon and Lakoff (1971), Heringer (1972) and Searle
(1975). For example, Can you tell me the time? and Do you know what
time it 1s? are perhaps most commonly uttered in order to make a
request; and this, their actual illocutionary force, is explicable in
terms of the principles which govern the performance of indirect
illocutionary acts. We do not have to say that, when the sentence Can you
tell me the time? is used to make a request, it no longer has its literal
meaning. We can say instead that the sentence may be used, without any
change of meaning, either directly to ask a question or indirectly to make
a request; and, if it 1s used indirectly to make a request, it has two kinds
of illocutionary force. It is because it can always be understood, at least
incidentally, as a question, that it can also be held, in context and in
terms of what Grice (1975) calls conversational implicatures (cf. 14.3),
to imply, or implicate, a particular request.
The second, and more important, point is that the sincerity conditions
that are asserted or questioned in the performance of indirect illocu-
tionary acts all have to do with the knowledge, beliefs, will and abilities |
of the participants; and these, as we shall see, are the factors which are |
786 Mood and illocutionary force
involved in epistemic* and deontic* modality (cf. 17.2, 17.4). It is for
this reason that mands, in particular, may be indirectly performed in
English by means of sentences beginning Will you... , Can you... , Isit
possible for you to... , Would you mind... , I want youto..., Id like
you to... , etc.; and it is for the same reason that the so-called tags that
may be attached to imperative sentences in English contain a modal verb
(Open the door, will you?, Stop making that awful noise, can’t you?).
Granted that the use of so-called indirect speech-acts can be ex-
plained along these lines, it must be recognized, of course, that certain
expressions are more readily usable than others in the performance of
indirect speech-acts. For example, Can you close the door? is a more nor-
mal request than Are you able to close the door? (cf. Sadock, 1974: 78).
Furthermore, the conventionalization of such formulae as Can you . . .?
Would you mind . . .? may reach the point where it is no longer reasonable
to treat the utterances in which they occur as being, even incidentally,
anything other than requests.
 I7
Modality

17.1. Necessity and possibility

As we saw in a previous chapter, logicians generally draw a distinction
between propositions that are contingently true or false, synthetic*
propositions, and propositions that are either necessarily true, analytic*
propositions or tautologies*, or necessarily false, contradictions* (cf.
6.5). To say that a proposition is contingently true is to imply that,
although it is in fact true of the world, or of the state of the world, that
is being described, there are other possible worlds, or states of the
world, of which it is, or might be, false. A necessarily true, or analytic,
proposition, on the other hand, is one whose truth is not simply a matter
of the way the world happens to be at some particular time. Analytic
propositions, as Leibniz put it, are true in all possible worlds. Their
truth is established, or guaranteed, by the meaning of the sentences
which express them; and our knowledge, or belief, that they are true is
non-empirical, in the sense that it is not grounded in, and cannot be
modified by, experience.
Necessity and possibility are the central notions of traditional modal
logic; and they are related, like universal and existential quantification
(cf. 6.3), in terms of negation. If p is necessarily true, then its negation,
~p, cannot possibly be true; and if p is possibly true, then its negation
is not necessarily true. These relationships may be expressed, in terms
of an intensional system with two modal operators (cf. 6.5), as follows:
(1) nec p = ~poss ~p
(2) poss p = ~nec ~p
Either necessity or possibility is therefore eliminable in favour of the
other under a double application of negation: once to the modal operator
and once to the unmodalized constituent proposition. These two kinds
of negation are commonly referred to as external and internal negation,
respectively (cf. 16.4); and in modal logic they must obviously be dis-
788 Modality
tinguished in terms of their scope. ‘“‘It is not possible that p is true”’
does not mean the same as “‘It is possible that not—p is true’’.
Various entailments, or implications, can be defined as holding
between logically necessary or logically possible propositions and the
unmodalized propositions embedded in them: notably
(3) nec p > p
and
(4) p — poss p
Of these (3) expresses the fact that, if we know or are told that p is
necessarily true, we can without more ado legitimately assert p or use it
as a premiss in argument. How we come to know, or believe, that p 1s
necessarily true is of course the question that is at the very heart of the
philosophical controversies surrounding the analytic and the a priori. It
is undeniable, however, that we constantly operate with inferences such
as (3). We assert or assume that certain propositions are in fact true,
because we know, or believe, that they must be true; and this know-
ledge, or belief, is built into the semantic structure of the languages we
use to construct our state-descriptions of the universe (cf. 6.5). For
example, ‘‘If Alfred is a bachelor, he must be unmarried’? may be
formalized as
(5) nec (B(a) > ~ M(a))
where B is a predicate meaning “‘bachelor’’ and M a predicate meaning
‘married’ (cf. 6.3).1 Now, (5) implies
(6) Bia) > ~ M(Q),
“Tf Alfred is a bachelor, he is not married”; and (6) is a formalization
of the kind of inference that we would naturally draw, though we might
not make it explicit, from the statement
(7) Alfred ts a bachelor.
But (6) is derivable from the universally quantified proposition
(8) (x) (B(x) > ~ M(x);
and, as native speakers of English, we might be prepared to assert, not
only that (8) is in fact true in the world that we inhabit, but that it could

1'The term ‘predicate’ is here being used in the sense in which it 1s custo-
marily employed by logicians (cf. 6.2, 12.2).
 17.1. Necessity and possibility 789
not but be true (i.e. it is true in all possible worlds) by virtue of the
sense of ‘bachelor’ and ‘married’. In which case, we can say that the
truth of (8), and therefore of (6), is guaranteed by the truth of
(9) nec ((x) B(x) > ~ M(z)).
The difference between (8) and (9) is that (8) states it as being simply a
matter of fact that all bachelors are unmarried, whereas (g) makes it a
matter of logical necessity.
Either (8) or (9), it should be emphasized, is sufficiently strong to
validate (6); and we need not be concerned here with the problem of
deciding whether the relationship between ‘bachelor’ and ‘married’
that is built into the lexical structure of English is such that we would be
justified in treating the proposition expressed by the sentence ‘All
bachelors are unmarried’ as analytic. It is to avoid, or circumvent, such
problems that Carnap (1956) casts all his meaning-postulates* in the
form of non-modal universally quantified propositions like (8).
What we want to explicate here is the meaning of ‘must’ in sentences
like

(10) Alfred must be unmarried,

interpreted in the sense in which the proposition that it expresses is a
natural and legitimate inference from the proposition expressed by (7).
Whether we account for this in terms of (8) or (g) is, for our present
purpose, irrelevant. In either case, under this interpretation of (10), the
English verb ‘must’ has the same function as the modal operator of
logical necessity in (5).
The fact that sentences like (10) are rarely used in everyday discourse
to express logically modalized propositions of the form
(11) nec (~ M(a))
does not in any way detract from the applicability or utility of the notion
of logical necessity in descriptive semantics. Our everyday use of lan-
guage, which includes the ability to draw inferences and to paraphrase
one sentence by means of another, clearly depends upon our intuitive
understanding of the operation of logical necessity. Moreover, as we
shall see presently, there are other interpretations of (10) in which
‘must’ has a sense which is analogous to, though distinguishable from,
the sense that it bears when it is being used to express the kind of logical
necessity with which we are at present concerned.
Let us now introduce a symbolic representation of the notion of actual
790 Modality
and possible worlds, in the sense in which the term ‘world’ is now
commonly used in modal logic (cf. 6.6). We will interpret the formula
(12) wi(p)
as meaning “‘p is true in world w,”’; and
(13) (w) (w(p))
as meaning ‘‘p is true in all worlds”. This is tantamount to saying,
(12), that p is included in a particular state-description of the universe
or, (13), in all state-descriptions of the universe (cf. 6.5). Now, if p is
given the value (B(a) - ~ M(a)) in (13), we have a formal representa-
tion of logical necessity in terms of possible worlds:
(14) (2) (w(B(a)) > ~ M(a)),
which may be interpreted as meaning “‘in all worlds it holds true that, if
Alfred is a bachelor, Alfred is not married”’. And (14) may be taken as
equivalent to (5). In order to apply this notion of possible worlds, we
must of course be able to operate with some concept of trans-world
identification of individuals. But this is a concept that we do in fact
operate with in our everyday affairs. We assume that the person we met
yesterday or last year, for example, is the same person as the person we
met this morning, although the world in which we met him previously
is in various respects different from the world in which we met him, and
re-identified him, this morning. In this case, a relation of temporal
sequence holds between the two worlds; and this relation can be used,
as we shall see later, to provide an interpretation of tense (cf. 17.3). We
can also carry out the psychological process of trans-world identification
across real and imaginary worlds of various kinds. We can identify our-
selves and others in our dreams; we can create hypothetical situations
involving real persons and then talk about these situations in much the
same way as we talk about things that are actually happening or have
actually happened. We can reason conditionally that, if a certain event
had taken place, or were to take place, in world uw; at time ti, then a
certain state-of-affairs, describable by a particular proposition or set of
propositions, would obtain in another hypothetical world w; at time tj.
The notion of possible worlds, which, when we first encounter it, may
seem rather fanciful is therefore one that we can be said to use con-
stantly in everyday discourse. Modern modal logic takes this intuitive
notion and employs it to formalize a variety of phenomena with which
we shall be concerned, somewhat informally, in this section.
 17.1. Necessity and possibility 791
The interpretation that we have given to the sentence “Alfred must be
unmarried’ is not of course the only interpretation that it can have. There
are at least two others, which may be identified by means of the following
glosses:
(15) “I (confidently) infer that Alfred is unmarried”’
and
(16) “Alfred is obliged to be unmarried”’.
Of these, (15) is closer than (16) is to the interpretation that was forma-
lized above as (11). But (15) 1s clearly not identical with what was
formalized as (11).
It has long been recognized that most of the sentences containing such
modal* verbs as ‘must’ and ‘may’ are ambiguous; and furthermore that
their ambiguity is not satisfactorily accounted for by saying that each of
the modal verbs happens to have two or more meanings. There is an
intuitively obvious connexion, on the one hand, between the notions of
necessity and obligation, which are relevant to the semantic analysis of
sentences containing ‘must’, and, on the other, between the notions of
possibility and permission, which are relevant to the semantic analysis
of sentences containing ‘may’. Moreover, the ambiguity found in
English sentences containing ‘must’ and ‘may’ is also found, in com-
parable sentences, in other languages.
Until recently, modal logic has been concerned almost exclusively,
with what is nowadays referred to as alethic* modality: i.e. with the
necessary or contingent truth of propositions. (The term ‘alethic’ 1s
derived from the Greek word meaning “true’’.) What are traditionally
described as necessary truths (i.e. propositions which, according to
Leibniz, are true in all logically possible worlds) may now be referred to
as alethically necessary propositions. Similarly, propositions that are not
necessarily false (1.e. propositions that are true in at least one logically
possible world) may be described as alethically possible. All alethically
necessary propositions are alethically possible, but not conversely. An
alternative traditional term for ‘alethically necessary’ is apodeictic*. |
Nowadays, two other kinds of necessity and possibility are recognized
and formalized, in various ways, by logicians: epistemic* and deontic*.
The significance of these terms will be explained presently. But here it
may be simply mentioned that it is epistemic necessity that is involved in
(15) and deontic necessity in (16). Epistemic necessity is intuitively
closer to alethic necessity than deontic necessity is.
792 Modality
Various terms have been used by linguists to distinguish (15) from
(16). Kurylowicz (1964) would say that (15) involves subjectivity (i.e.
the expression of the speaker’s attitude), whereas (16) does not. Halliday
(1970a) says that (16) involves what he calls modulation, rather than
modality (in a different sense of ‘modulation’ from the sense in which
we are using this term: cf. 3.3); and he too takes (15) to be more sub-
jective than (16). Generative grammarians working in the Chomskyan
tradition frequently distinguish between (15) and (16) in terms of a
difference between epistemic modals and root-modals; and it has been
further proposed that this difference can be accounted for in terms of
the intransitive or transitive use of the same underlying modal verb of
necessity (cf. Ross, 196ga).? In what follows, we shall make no attempt to
synthesize or compare these several views of what we shall refer to as
epistemic and deontic modality.
However, one very general point may be made, which would be highly
relevant to such a synthesis or comparison. This is that linguists have
generally taken a rather different view of epistemic modality than that
taken by logicians; and (15) is a linguist’s, rather than a logician’s, gloss.
The logician would probably favour
(17) “In the light of what is known, it is necessarily the case that
Alfred is unmarried”’,

in which there is no reference to the speaker or to the actual drawing of

inferences, but only to the evidence which determines the epistemic
necessity of the proposition in question; and this evidence would be
treated as something objective. The subjectivity of epistemic modality,
which has seemed to certain linguists, including Kurylowicz, to be one
of the features which distinguishes it most sharply from deontic modality,
is not represented at all in standard systems of epistemic logic. We will
try to account for this difference between the typically linguistic and the
typically logical view of epistemic modality, in the distinction that we
will draw between subjective and objective epistemic modality.
When we take up the discussion of deontic modality, we shall see that
a similar distinction can also be drawn between subjective and objective

2 The so-called root modals are treated by Ross (1969a), G. Lakoff (1972) and
others who have adopted the auxiliaries-as-main-verbs analysis as being
trivalent rather than bivalent (i.e. as being di-transitive, rather than transi-
tive: cf. 12.4) in the case of deontic utterances that are given a performative,
rather than constative, interpretation. On this and related points, cf. Huddles-
ton (1974).
 17.2. Epistemic modality and factivity 793
deontic modality (cf. 17.4). Not only is (16) a deontic interpretation of
‘Alfred must be unmarried’. So too is
(18) “I (hereby) oblige Alfred to be unmarried”’,
which is performative, rather than constative, in the original sense of
‘performative’ (cf. 16.1). The fact that both epistemic and deontic
modality can be interpreted either subjectively or objectively means that
Kurylowicz’s account of the distinction between epistemic and deontic
modality cannot be correct, as it stands. Nor can the proposed analysis
in terms of the difference between an intransitive (epistemic) and a
transitive (deontic) use of modal verbs. Why, then, has it seemed
plausible to so many linguists to think of epistemic modality as being
more subjective than deontic modality? This is a question that requires
an answer.

17.2. Epistemic modality and factivity

Epistemic* modality, as the term is used by philosophers, is less easy to
characterize non-technically than alethic modality (cf. 17.1); and there
is some discrepancy, in fact, between the sense in which philosophers
employ the term and the sense in which it has come to be used in
linguistic semantics. The term ‘epistemic’, like ‘epistemology’, is
derived from the Greek word meaning “‘knowledge’’. Whereas episte-
mology is concerned with the nature and source of knowledge, epistemic
logic deals with the logical structure of statements which assert or imply
that a particular proposition, or set of propositions, is known or be-
lieved. Epistemic logic, in the opinion of some authorities, also lends
itself, as does alethic logic, to formalization in terms of the notion of
possible worlds. As we saw in an earlier chapter, the semantic content
of a proposition is the set of state-descriptions (or, equivalently, possible
worlds) that it excludes (cf. 6.5). Knowing what a proposition means
implies knowing under what conditions (i.e. in which possible worlds)
it is true; and knowing what someone knows or believes implies knowing
the semantic content of the propositions that he subscribes to, or takes
to be true.
The distinction between knowledge and correct belief, if indeed there
is one from an epistemological or psychological point of view, has been
one of the enduring concerns of Western philosophy since Plato. But we
do not need to resolve this question in order to establish at least some
of the conditions under which it is appropriate to use the words ‘know’
and ‘believe’, and other words denoting what philosophers commonly
794 Modality
describe as propositional attitudes* (‘doubt’, ‘think’, ‘imagine’, etc.).
It is generally agreed that ‘X knows that p” implies “‘X believes that
p” (i.e. that X takes p to be true); and furthermore that the assertion of
‘““X knows that p’’ commits the speaker to the belief that p is true,
whereas the assertion of “‘X believes that p’’ obviously does not. If X 1s
the speaker himself, he is committed by the assertion of either “X
knows that p” or ““X believes that p”’ (i.e. by the utterance of I know
that p or I believe that p) to the truth of p. But if he employs ‘know’,
rather than ‘believe’, he is making a stronger commitment: he is claim-
ing that his belief in the truth of p is well-grounded and in his judgement
at least unassailable, and that by virtue of this fact, which he should be
able to substantiate, if called upon to do so, by providing the evidence,
he has the right to assert p and to authorize others to subscribe to
its truth. Without going into the principles of epistemic logic, let us
merely say that the proposition “X knows that p”’ can be formalized
as

(1) Kx(p),
where Kx is the operator of epistemic necessity; and that (1) is related
to p in terms of epistemic necessity as nec (p) 1s related to p in terms of
alethic necessity. If Kx(p) is true, then p must be true, in some sense of
‘must’ that it is the task of epistemic logic to explicate.
Many philosophers would say that the relationship between Kx(p)
and p is one of presupposition* and not simply implication, since the
truth of p is a necessary condition, not only of the truth of “X knows
that p’’, but also of the truth of its negation, “X does not know that
p”. In saying He does not know that Edinburgh 1s the capital of Scotland
(provided that it is intended as a context-free statement, rather than a
denial: cf. 16.4), the speaker 1s committed to the truth of the proposition
‘Edinburgh is the capital of Scotland’’, as he is committed to the truth
of the same proposition by saying (as an assertion) He knows that
Edinburgh is the capital of Scotland.®? Any predicator (verb, adjective,
etc.), which behaves like ‘know’ in this respect can be said to have the
property of factivity*, or to be a factive* predicator. These terms were
introduced into linguistics by Kiparsky & Kiparsky (1971), who pointed
out that there is a wide range of factive predicators in English in addi-

3 As was noted in 14.5 and 16.4, there is an additional interpretation possible

for negative sentences that contain ‘know’ with first-person subject (e.g., ‘I
don’t know that the earth is round’). We are not concerned with this inter-
pretation here.
 17.2. Epistemic modality and factivity 795
tion to those that are used to assert that someone knows or does not
know that p.
For example, anyone who says
(2) It 1s amazing that they survived
is committed to the truth of the proposition expressed by the statement
(3) They survived.
The proposition expressed by (3) is not of course asserted by the person
uttering (2); and whether we say that the proposition expressed by (3)
is presupposed by the proposition expressed by (2) or by the person
uttering (2) is of little consequence in the present connexion (cf. 14.3).
The point is that ‘amazing’ is a factive predicator.
No less interesting or important than factivity is what we will call
non-factivity and contra-factivity. The use of a non-factive* predicator,
like ‘believe’ or ‘think’, commits the speaker to neither the truth nor
the falsity of the proposition expressed by its complement clause in such
statements as He belteves/thinks that Edinburgh 1s the capital of Scotland.
A contra-factive* utterance, on the other hand, commits the speaker, not
to the truth, but to the falsity, of the proposition, or propositions, ex-
pressed by one or more of its constituent clauses; and the property of
contra-factivity may be assigned more specifically to whatever com-
ponent of the utterance it is that indicates the speaker’s commitment to
the falsity of the proposition or propositions expressed.
The most obvious examples of contra-factive utterances (and they are
recognized as such in traditional grammar) are wishes* and so-called
unreal* (or counter-factual*) conditionals (with past-time reference),
such as
(4) L wish he had been to Paris
and
(5) If he had been to Paris, he would have visited Montmartre,
respectively, which, in so far as they are contra-factive, can be uttered
appropriately only if the speaker believes that the proposition expressed
by
(6) He went to Paris
or
(7) He has been to Paris
796 Modality
is false (where ‘he’ of course must have the same referent in (6) or (7)
as it has in (4) or (5)).*
Unreal conditional statements, like (5), contrast with so-called real*
conditional statements, such as
(3) If he went to Paris, he visited Montmartre,
which (construed in the sense in which it is not equivalent to Whenever
he went to Paris, he visited Montmartre) is neither factive nor contra-
factive, but non-factive*: its utterance commits the speaker to neither
the truth nor the falsity of the proposition expressed by (6).
Open questions, like
(9) Did he go to Paris?
it should be noted, are also non-factive. So too are statements like |
(10) I don’t know whether he went to Paris;
and utterances like
(11) He may have gone to Paris
(12) Perhaps he went to Parts
(13) It’s possible that he went to Paris.
(11), (12) or (13), which contain a modal verb (‘may’), a modal adverb
(‘perhaps’) or a modal adjective (‘possible’) are typical of the utter-
ances that linguists discuss in terms of the notion of epistemic modality;
and we shall henceforth be concerned mainly with utterances of this
kind.
4 Recent works on conditionals include Isard (1974), Lewis (1973) and Stal-
naker (1968), all of which operate with the notion of possible worlds. Of
particular interest in Isard’s procedural approach is the use he makes of the
opposition between the functions PRESENT and REMOTE, which serve as indices
to possible worlds and may be envisaged as computerizable instructions that
take us into an actual (PRESENT) or non-actual (REMOTE) situation. In many
languages, including English, so-called past-tense verb-forms are used to
describe, not only situations that have, or may have, existed or occurred in
the past, but also, in counter-factual conditionals like (5), purely hypothetical
situations that have never existed or occurred and are no longer actualizable.
Both uses of the so-called past tense are naturally accounted for in terms of
the more general notion of remoteness, or non-actuality (cf. 17.3). It may be
noted at this point that, if (5) is taken to be non-factive, rather than contra-
factive the function of the past-tense marker in the first clause will be tem-
poral rather than modal. What will be at issue in this case is not whether the
person being referred to went or has been to Paris at time ti, but whether he
had been there at the more remote time tj (t; < t).
 17.2. Epistemic modality and factivity 797
Straightforward statements of fact (i.e. categorical assertions) may be
described as epistemically non-modal. The speaker, in uttering an un-
qualified assertion, is committing himself to the truth of what he asserts,
by virtue of the felicity-conditions which govern the illocutionary act
of assertion, but he is not explicitly laying claim to knowledge in the
utterance itself: he is not asserting the epistemically modalized propo-
sition ‘‘I know that p”’; he is saying, without qualification of either the
I-say-so component or the it-is-so component of his utterance, that (it
is the case that) p is true (of the world that he 1s describing). Any utter-
ance in which the speaker explicitly qualifies his commitment to the
truth of the proposition expressed by the sentence he utters, whether this
qualification is made explicit in the verbal component, as in (11), (12)
and (13), or in the prosodic or paralinguistic component, is an epis-
temically modal, or modalized, utterance.
In principle, two kinds of epistemic modality can be distinguished:
objective* and subjective*. This is not a distinction that can be drawn
sharply in the everyday use of language; and its epistemological justifi-
cation is, to say the least, uncertain. It is also difficult to draw a sharp
distinction between what we are calling objective epistemic modality and
alethic modality; both of them are subsumed by Carnap’s notion of
logical probability, to which reference was made in an earlier chapter
(6.5). It is nonetheless of some theoretical interest to draw the distinc-
tion between objective and subjective epistemic modality. In order to do
so, we will briefly compare the statement
(14) Alfred may be unmarried
with
(15) Alfred must be unmarried,
which was used earlier to illustrate the difference between alethic,
epistemic and deontic necessity (cf. 17.1).
Under one interpretation of (14) the speaker may be understood as
subjectively qualifying his commitment to the possibility of Alfred’s
being unmarried in terms of his own uncertainty. If this is what the
speaker meant, he might appropriately have added to (14) some such
clause as
(16) but I doubt it
or
(17) and I’m inclined to think that he 1s,
which clearly indicate the subjectivity of the speaker’s commitment.
798 Modality
Under this interpretation, which is probably the most obvious, (14) is
more or less equivalent to
(18) Perhaps Alfred 1s unmarried.
But now consider the following situation. There is a community of
ninety people; one of them is Alfred; and we know that thirty of these
people are unmarried, without however knowing which of them are
unmarried and which are not. In this situation, we can say that the
possibility of Alfred’s being unmarried is presentable, should the speaker
wish so to present it, as an objective fact. The speaker might reasonably
say that he knows, and does not merely think or believe, that there is a
possibility (and in this case a quantifiable possibility) of Alfred’s being
unmarried; and, if he is irrational, his own subjective commitment to
the truth or falsity of the proposition ‘Alfred is not married”’ might be
quite unrelated to his knowledge of the objective possibility, or degree
of probability (1/3), of its truth, in the way that a gambler’s subjective
commitment to the probability of a particular number coming up in
roulette might be quite unrelated to the objective probabilities. (18),
unlike (14), cannot be used to express objective epistemic modality.
Let us now look at (15). Suppose that, having established the marital
status of every member of the community except Alfred, we find that
we have identified exactly twenty-nine persons that are unmarried. In
these circumstances, it would be appropriate to utter (15); and we should
be justified in claiming to know that the proposition ‘Alfred is un-
married’? must be true in some sense of ‘must’ which, as we have
already mentioned, Carnap (1962) relates to the sense that ‘must’ bears
in statements of alethic necessity. But there are other situations, and in
the everyday use of language they are of more frequent occurrence, in
which it would be more natural to interpret (15) in terms of subjective
epistemic modality.
Granted that the distinction between subjective and objective epis-
temic modality is theoretically defensible (and we have already pointed
out that objective epistemic modality, if it is a viable notion, lies between
alethic modality, on the one hand, and subjective epistemic modality, on
the other, and might be assimilated. to either), the question now arises
how we should account for this distinction in terms of the tripartite
analysis of utterances developed in the previous chapter. Categorical
assertions, it will be recalled, were described as having an I-say-so
neustic and an it-is-so trophic; and their illocutionary force was accoun-
ted for as the product of both of these components. Now, it seems clear
 17.2. Epistemic modality and factivity 799
that the main difference between subjectively and objectively modalized
utterances 1s that the latter, but not the former, contain an unqualified,
or categorical, I-say-so component. The speaker is committed by the
utterance of an objectively modalized utterance to the factuality of the
information that he is giving to the addressee: he is performing an act
of telling. What he states to be the case can be denied or questioned
(That’s not true; Is that so?; I don’t believe you; etc.); it can be accepted

married), |
as a fact by the addressee (J agree; Yes, I know, etc.); it can be hypothe-
sized in a real conditional statement; and it can be referred to by the
complement of a factive predicator (1 knew that Alfred must be un-

In all these respects, objective modalization differs from subjective

modalization, the very essence of which is to express the speaker’s
reservations about giving an unqualified, or categorical, ‘‘I-say-so”’ to
the factuality of the proposition embedded in his utterance. Subjectively
modalized statements (if indeed they can be properly called statements)
are statements of opinion, or hearsay, or tentative inference, rather than
statements of fact; and they are reported as such. For example,
(19) ‘He said that it might be raining in London’
could probably be used to report
(20) It may be raining in London
under either of the two interpretations of (20) that concern us at this
point. But
(21) ‘He told me that it might be raining in London’
would be appropriately used to report a token of (20), only if the person
uttering (21) had construed (20) as having objective epistemic modality:
if it had been uttered, for instance, by a meteorologist. If (20) is inter-
preted as a subjectively modalized utterance, it is more appropriately
reported by means of some such statement as
(22) He told me that he thought it might be raining in London
) or
(23) He expressed the opinion that it might be raining in London.
This suggests that subjectively modalized utterances, unlike categorical
assertions and objectively modalized statements, are not acts of telling;
and that their illocutionary force is in this respect similar to that of
800 Modality
questions, which are also non-factive: cf. (9). There is of course an
obvious difference between putting forward the opinion, with greater or
less confidence and authority, that such-and-such may be, or must be,
so and asking (or wondering) whether such-and-such is in fact so. In
both cases, however, there is an overt indication of the speaker’s un-
willingness or inability to endorse, or subscribe to, the factuality of

doubt. }
the proposition expressed in his utterance; and both of them may
well originate, ontogenetically, in the same psychological state of

Subjective epistemic modality can be accounted for, as we have seen,

in terms of the speaker’s qualification of the I-say-so component of his
utterance. Objectively modalized utterances (whether their modality is
alethic or epistemic) can be described as having an unqualified I-say-so
component, but an it-is-so component that is qualified with respect to a
certain degree of probability, which, if quantifiable, ranges between 1
and o. If the factuality of an epistemically modalized proposition (as it is
presented by the speaker) is of degree 1 it is epistemically necessary; if
its factuality is of degree o it 1s epistemically impossible. In everyday
discourse we do not normally quantify the factuality of the propositions
expressed in our utterances by means of a numerical variable. But we
can express at least three different degrees of factuality in English by
selecting one modal adverb rather than another from a set which includes
‘certainly’, ‘probably’ and ‘possibly’; and the difference between
‘probably’ and ‘possibly’, when they are used in objectively modalized
statements, would seem to correlate, at least roughly, with the difference
between a degree of factuality that is greater than and one that is less
than o-5. But this admittedly rough correlation with the numerical
quantification of factuality need not be pressed. For there is no reason
to believe that either subjective or objective epistemic modality, in
non-scientific discourse, is grounded 1n a mathematically precise calcu-
lation of probabilities.
The important point is that objective epistemic modality is in prin-
ciple quantifiable on a scale whose extremes are necessity and impossi-
bility; and different language-systems may well grammaticalize or
lexicalize distinctions along this scale in terms of more or fewer degrees.
As we have already seen, either necessity or possibility is definable in
terms of the other, together with external negation (cf. 16.4). The ques-
tion that now faces us is whether there are any linguistic reasons for
preferring one alternative to the other. Is the grammaticalization or
lexicalization of epistemic modality in all languages, or certain languages,
 17.2. Epistemic modality and factivity 801
such that they may be more appropriately described as being possibility-
based or necessity-based?
There is some evidence to suggest that, in English at least, epistemic
modality is possibility-based. The following utterances containing
‘may’ or ‘can’ are all acceptable as having either subjective or objective
epistemic modality:
(24) It may be raining
(25) It may not be raining
(26) It can’t be raining
(27) It can’t not be raining (or It cannot but be raining).
(25) and (26) differ from one another in that the former means “‘It is
possible that it is not raining”’ and the latter “‘It is not possible that it is
raining’’. (27) means “It is not possible that it is not raining”. The
following utterances are also acceptable:
(28) It must be raining |
(29) It mustn’t be raining.
(28) can be related to (24) in terms of external and internal negation;
and, if it is given the appropriate stress pattern, it is equivalent to (27),
which, according to whether the stress falls on not or on can’t, is the
denial of either (25) or (26). Similarly, (29), uttered with the appropriate
stress patterns, is equivalent to (26).
What we do not find in English, however, are utterances containing
‘must’ and two negations which are as acceptable and as natural as (27)
and which can be construed as being equivalent to (24). The following
(30) It mustn’t not be raining
can be interpreted, in an appropriate context, as containing a deontically
modalized proposition. But it is at best doubtful whether it could ever be
uttered with the meaning “‘(I say that) it is not necessarily the case that
it is not raining’’. In order to express this with a modal verb of necessity,
rather than possibility, we must use ‘need’, as in
(31) It needn’t not be raining;
and (31) is hardly as natural, as a subjectively or objectively modalized
utterance, as (27). Furthermore, ‘need’ is more restricted in the range
of constructions in which it can be used than ‘may’, ‘can’, ‘must’,
802 Modakty
‘will’ and ‘shall’; and it is generally agreed that these five verbs are the
principal modal auxiliaries in English.
The modal auxiliary verbs occupy a more central position in the
grammatical structure of English than do modal adjectives (‘ possible’,
‘necessary’, etc.) or adverbs (‘possibly’, ‘necessarily’, etc.). It would
therefore seem to follow from what has just been said about the natural-
ness of It can’t not be raining, in contrast with the unnaturalness of It
mustn't not be raining (under an epistemic interpretation), that, although
there may be no difference in terms of naturalness or acceptability
between
(32) It’s not possible that it’s not raining
and
(33) It’s not necessary that it’s not raining,
in English at least, possibility, rather than necessity, should be taken as
primitive in the analysis of epistemic modality. Whether this holds for
all languages is an empirical question that is probably unanswerable on
the basis of the evidence available at present. Grammarians have not
generally concerned themselves with points of this kind; and such points
are difficult enough to establish with respect to the interpretation of
utterances in one’s native language. It may be noted, however, for what
it is worth, that traditional treatments of the category of mood by
grammarians tend to invoke the notion of epistemic possibility more
commonly than they do the notion of epistemic necessity. 'The use of a
distinctive mood (e.g., the subjunctive) represents a further degree of
grammaticalization than does the use of a special set of auxiliary modal
verbs or modal particles.
At this point it will be convenient to introduce a notational system
within which we can conveniently distinguish modal and non-modal
utterances of various kinds, Categorical assertions will be represented
symbolically as
(34) --?
The first full-stop stands for the unqualified neustic, and may be read
‘‘T-say-so”’; the second full-stop stands for the unqualified tropic, and
may be read ‘“‘it-is-so”’. As we have seen (16.4), all three components
of the utterance may be negated; so that
(35) ~--?
will mean “‘I-don’t-say that it is the case that p”’ (non-commitment);
 17.2. Epistemic modality and factivity 803
(36) . ~-p
will mean “‘I say that it is not the case that p”’ (denial); and
(37) .. ~P
will mean “I say that it is the case that not—p”’ (context-free assertion
of a negative proposition). We may also have any two, or all three, kinds
of negation within the same utterance. In contrast with the full-stop
symbolizing ‘‘I say so”’ in the neustic position, we will use a question-
mark to represent the I-wonder or I-can’t-say-so component of ques-
tions:
(38) °.p
will therefore mean “‘I wonder whether it 1s the case that p” or “Is it
the case that p?’’. In contrast with the full-stop symbolizing ‘‘it 1s so”’
in the tropic position, we will use an exclamation-mark to represent
the so-be-it component of commands, demands and deliberative ques-
tions:
(39) - !p
will represent the structure of commands and demands;
(40). ~!p
will represent the structure of prohibitions;
(41) ?!p
will represent the structure of deliberative questions. This system of
notation obviously needs to be extended in order to represent the struc-
ture of x-questions and requests. But this extension is irrelevant here;
and we will leave further discussion of the so-be-it component (!) for a
later section.
Since there are some languages at least for which it seems to be more
appropriate to analyse epistemic modality in terms of possibility rather
than necessity, we will opt for this alternative in our symbolic represen-
tation of the difference between categorical assertions and subjectively or
objectively modalized assertions.
The difference between subjective and objective epistemic modality
can now be represented by substituting the modal operator of possibility
for the first and the second full-stop, respectively, in the formulae for
categorical assertions:
(42) poss . p
804 | Modahty
and

(43) . poss p
which may be read as (42) “‘Possibly/perhaps, it is the case that p”’ and
(43) “I say that it is possibly the case that p”’. Each of these operators
can be negated; so can the propositional variable and either the un- -
qualified neustic or the unqualified phrastic.
Under this system of representation, possibility is treated as a two-
valued variable, which in one position alternates with interrogation as a
qualifier of the I-say-so component and in the other position operates as
a qualifier of the it-is-so component.
The system also allows for utterances with both subjective and objec-
tive modality:
(44) poss poss p
which may be read as “‘Possibly/perhaps, it is possibly the case that p”’
and is exemplified by one interpretation of
(45) Perhaps it may be raining.
It therefore allows for a total of twenty-four notationally distinct and
semantically non-equivalent modal formulae (in addition to eight non-
modal formulae) each containing the same propositional variable; and
each of these can in fact be realized in English, as the reader can verify
for himself by constructing appropriate situations-of-utterance. This
number can be increased if possibility in either or both positions is
treated as a scale along which a greater or less number of degrees are
recognized; and English provides the means for indefinitely finer and
finer gradations of subjective and objective epistemic modality. We can
not only say, as has been noted, I¢ 1s probably raining, but also It is very
probably raining, It ts almost certainly raining, and so on.
Of the twenty-four formulae that can be constructed in terms of
possibility and negation within a framework which distinguishes the
neustic from the tropic, only four find their equivalents in standard
possibility-based systems of modal logic:
(46) . poss p
(47) . ~ poss p
(48) . poss ~ p
(49) . ~ poss ~p
 17.2. Epistemic modality and factivity 805
The reason is that in standard systems of non-epistemic and non-deontic
modal logic all statements are construed, in our terms, as having pre-
fixed to them an unqualified I-say-so component. Subjective epistemic
modality, as was mentioned in the previous section, is something that
logicians have not been concerned with. It is of such importance,
however, in the ordinary use of language that it should be explicitly
recognized in any formalization of linguistic semantics and distin-
guished, in principle, from objective modality.
At the same time, the relationship between subjective and objective
epistemic modality is such that it seems appropriate to handle both in
terms of the same modal operator, assigning to it two places of occur-
rence. Its function in the neustic position is to express different degrees
of commitment to factuality; and in this respect it qualifies the illocu-
tionary act in much the same way that a performative verb parenthetic-
ally qualifies, or modulates, the utterance of which it is a constituent in
an explicitly performative utterance or a primary performative with a
performative clause tagged on to it. Looked at from this point of view,
It may be raining (construed as a subjectively modalized utterance)
stands in the same relationship to It’s raining, I think or I think it’s
raining, as Is it raining? does to Is it raining, I wonder? or I wonder
whether it’s raining and It’s raining to It’s raining, I tell you or I tell
you it’s raining. Relationships of this kind are quite naturally expressed
within a system which treats interrogation and subjective modality as
qualifications of the I-say-so, or performative, component of the utter-
ance.
As epistemic possibility seems to be more basic, in English at least,
than epistemic necessity, so subjective epistemic modality seems to be
more basic than objective epistemic modality. This is the most natural
conclusion to be drawn from the fact, mentioned in the previous section,
that few linguists have even considered the possibility that epistemic
modality could be anything other than a matter of the speaker’s attitude
towards the propositional content of his utterance; and most discussions
of mood and modality in linguistics seem to take it for granted that
epistemic modality is subjective, in this sense. Hence the common, but
strictly speaking false, statement that the modal verbs cannot occur with
epistemic function in conditional sentences in English. Provided that an
utterance like
(50) If it may be raining, you should take your umbrella
is taken to express objective, rather than subjective, epistemic modality,
806 Modality
it is interpretable and fully acceptable. What is excluded from condi-
tional clauses is the expression of subjective epistemic modality; and
this is for the obvious reason that everything that comes within the scope
of the conditional operator (“if p, then q’’) must be part of the propo-
sitional content.
Utterances like (50) are undoubtedly rare in English; and the reason
is that it is much more natural to use modal verbs for subjective, than
for objective, epistemic modality. Far more natural than (50) is :
(51) If tt ts possible that it will rain, you should take your umbrella;
or, better still,
(52) If there ts a possibility of rain you should take your umbrella.
The constructions used in (51) and (52) are more appropriate than is the
construction used in (50) for the expression of objective epistemic
modality; and it is consistent with their status that, unlike the construc-
tion used in (50), they can contain a tense other than the simple present
and they can be recursively combined without limit. The objectification*
of epistemic modality is a precondition of our being able to talk about
past or future possibilities (cf. Yesterday there was some possibility that it
would rain today) and of there being one epistemic modality within the

formulae ,
scope of another (cf. It is possible that it is necessary that .. .).
What might be proposed, therefore, is that of the following two

(53) poss. p
and
(54) . poss p
(53) is more basic than (54), as far as the everyday use of language is
concerned; and that (54) can be thought of as being derived from (53)
by a process of objectification.® Indeed, we can go one step further, from

§ Halliday (1970a: 337) uses the term ‘objectification’ in connexion with

examples like 7t was certain that this gazebo had been built by Wren until the
discovery of the title deeds. Since he restricts the application of the term
‘modality’ to what is here called subjective epistemic modality, he takes
objectification to be a process whereby the expression of such notions as
certainty is ‘‘removed from the realm of modality’’ and made part of what
he calls the thesis (i.e. the propositional content). Although I have preferred
to extend the term ‘modality’, as logicians do, to objective possibility and
necessity, I am in agreement with Halliday in that I take the objectification
of epistemic modality to be of secondary development in the ordinary use of
language.
 17.2. Epistemic modality and factivity 807
the partial objectification of (54), in which the modal operator is assigned
to the tropic, to the complete objectification of modality.represented by
(55) - - (poss p)
in which the modal operator is combined with p to make a modalized
proposition. As we have already seen, in standard systems of alethic
modal logic, not only are p and ~p propositions; so also are poss p,
~poss p, poss ~ p and ~poss ~ p (cf. 6.5). And this is obviously a
necessary condition of the recursion of modality: i.e. as poss p is a
proposition which results from the application of poss to p, so poss
poss p is a proposition which 1s derived by applying poss to poss p; and
so ad infinitum. Instead of (46)-(49), therefore, we should perhaps think
of the following as being the equivalents, in the notational system being
used here, of formulae containing modal operators in standard systems:

(57) . . (~poss p) :
(56) . . (poss p)

(58) .. (poss ~ p)
(59) .-(~poss ~ p)
The difference between .possp and ..(possp) is the difference
between “I say so —it is possibly so—that p” and “‘I say so — it is
so — that possibly p’’.
In most dialects of English not more than one modal verb can occur
within the same clause. But both a modal verb and a modal adverb may
be combined. When this happens a distinction is to be drawn between
modally harmonic* and modally non-harmonic* combinations. For
example, ‘possibly’ and ‘may’, if each is being used epistemically, are ,
harmonic, in that they both express the same degree of modality,
whereas ‘certainly’ and ‘may’ are, in this sense, modally non-harmonic.
It has been pointed out by Halliday (1970a: 331) that the adverb and
the modal verb may, and normally do, ‘reinforce each other’ in a
modally harmonic combination; so that, in an utterance like
(60) He may possibly have forgotten
which, under the interpretation that concerns us here, differs but little
from either
(61) He may have forgotten
or
(62) He has possibly forgotten,
808 Modality
there is a kind of concord running through the clause, which results in
the double realization of a single modality. In contrast with (60), how-
ever,
(63) He may certainly have forgotten
and
(64) Certainly he may have forgotten
cannot be construed as having but a single modality running right
through the clause. Since the adverb and the modal verb are non-
harmonic they cannot but be independent; and one must be within the
scope of the other.
As far as (64) is concerned, ‘may’ is obviously within the scope of
‘certainly’, since (64) means “‘It is certainly the case that he may have
forgotten’’, rather than ‘‘It may be the case that he has certainly for-
gotten’’. The interpretation of (63) is rather less clear cut. What is
clear, however, is that, whether it is interpreted as “It is certainly the
case that he may have forgotten”’ or as “It may be the case that he has
certainly forgotten’’, no more than one of the two modal expressions can
express subjective epistemic modality (though they may both express
objective epistemic modality) and it is the one which expresses subjec-
tive epistemic modality that has the wider scope. These two principles
are very important and we shall draw upon them later: (i) subjective
modality always has wider scope than objective modality; (ii) no simple
utterance may contain more than a single subjective epistemic modality
(though this single modality may be expressed, as in (60), in two or more
places). Both of these principles are explicable in terms of the fact that
subjective epistemic modality has to do with the qualification of the
performative component of the utterance.
We shall return to the question of epistemic modality in various
places in later sections of this chapter. But there will not be space to
deal with it fully. One point should be borne in mind throughout.
Although it might appear that a statement is strengthened by putting
the proposition that it expresses within the scope of the operator of
epistemic necessity, this 1s not so, as far as the everyday use of language
is concerned. It would be generally agreed that the speaker is more
strongly committed to the factuality of “It be raining” by saying Jt zs
raining than he is by saying It must be raining. It is a general principle,
to which we are expected to conform, that we should always make the
strongest commitment for which we have epistemic warrant*. If there is
 17.3. Tense as a modality 809
no explicit mention of the source of our information and no explicit
qualification of our commitment to its factuality, it will be assumed that
we have full epistemic warrant for what we say. But the very fact of
introducing ‘must’, ‘necessarily’, ‘certainly’, etc., into the utterance
has the effect of making our commitment to the factuality of the propo-
sition explicitly dependent upon our perhaps limited knowledge. There
is no epistemically stronger statement than a categorical assertion, sym-
bolized here as .. p.

17.3. Tense as a modality

The deictic basis of the grammatical category of tense has been dis-
| cussed in a previous chapter (15.4). Something must now be said about
tense, and more generally about temporal reference, in relation to
modality. For simplicity, we will operate with the traditional notions of
past, present and future. It was pointed out earlier that there is no rea-
son, in principle, why language, even if it has tense, should gram-
maticalize deictic temporal reference in terms of the traditional
tripartite system; and that most languages that have tense treat the
distinction of past and non-past as being of greater importance than
the distinction of present and non-present or future and non-future.
Nothing that is said in this section about tense and modality should be
understood as implying that the tripartite distinction of past, present
and future is inherent in the structure of language.
Throughout this chapter and the previous chapter we have tacitly
assumed that tense, unlike modality, is part of the propositional content
of an utterance. We have already seen, however, that propositions are
generally treated by logicians as being tenseless and that the term
‘tenseless’, or ‘timeless’, can be interpreted in several different ways
(cf. 15.4). We can incorporate within the propositions expressed by
sentences like ‘It is raining’ some reference to a point or period of time
that is fixed in relation to an absolute or external system of temporal
co-ordinates; and then treat these propositions as being eternally true or
false. Alternatively, we can take the truth or falsity of a proposition as
being relative to the world, or world-state, that it purports to describe;
saying, for example, that the proposition expressed by ‘It is raining’,
It was raining’, ‘It will be raining’, ‘It has been raining’, etc., is true
in world 2; (i.e. in the world at time t), but false in world w; and that
the same proposition — ‘It be raining” — is involved in each case. It is
this second alternative that has been adopted in certain systems of
tense-logic.
810 Modality
Tense-logic is a branch of modal logic which extends the proposi-
tional calculus by introducing into the system a set of tense-operators
whose function and interpretation is similar to that of the operators of
logical necessity and possibility in standard modal logic (cf. 6.3). For
example, the past-tense operator of P or the future-tense operator of F
may be prefixed to a propositional variable, p, g, 7, etc., to yield
(1) P(p)
and
(2) F(d)
These formulae may be read as meaning ‘“‘It was/has been the case (in
the past) that p”’ and “It will be the case (in the future) that p”’. Under
this analysis, the relationship between the propositions expressed by the
sentences ‘It was raining’ and ‘It is raining’ is taken to be the same
kind of relationship as that which holds between the propositions
expressed by ‘It must be raining’ and ‘It is raining’. Such principles as
the following are generally accepted; and they are intuitively obvious:
(i) If p is always (i.e. omnitemporally) true, then p is sometimes true.
This is analogous to nec p — poss p and to (x) fx — (Ax)fx.
(ii) If it is always the case that p, then it is never the case that not—p:
cf. nec p > ~ poss ~p and (x)fx > ~ (Ax) ~ fx.
(iii) If it is always the case that p, then it is now the case that p (Le.
p is actual): cf. nec p — p.
We will not go into the details of tense-logic as such, or the more
particular relationships that alternative systems establish between the
tense-operators and the operators of logical or epistemic necessity (cf.
Rescher & Urquhart, 1971). What we are concerned with here is the
general conception of tense that is formalized in tense-logic; and this
can be discussed in terms of the notion of possible worlds, or world-
states.
Every utterance establishes its own spatiotemporal point of reference
~— the zero-point of the deictic system — in relation to which the entities,
events and states-of-affairs referred to by the speaker may be identified
(cf. 15.1). We will here neglect all but the temporal component of the
zero-point of reference; and we will symbolize it as to (“‘time-zero”’).
This temporal point of reference can now be used to identify one of the
possible world-states, wo, to which the speaker can refer in the utterance
and to which he can relate other world-states by means of tense and
 17.3. Tense as a modality Sir
modality. wo is the state of the world (as it 1s conceived by the speaker)
at fo: it is the present state of the actual world.
The present state of the actual world can be related in terms of past
tense to other temporally distinct states of the actual world preceding
Wo. A statement like

(3) lt was raining

can be interpreted as meaning “‘I say here and now (at to) — it was the
case (at tj < fo)—it be raining”, where “it be raining’’ is a tenseless
proposition which the speaker asserts as being true in w;-—1.e. in the
actual world as it was at ti < to (t% being a point or period of time
preceding Zo: cf. 15.4). Similarly
(4) It will be raining
can be interpreted as meaning “I say here and now (at fo) — it will be
the case (at tj > to) — it be raining’’ (where % is a point, or period, of
time following fo).
Under this analysis of the meaning of (3) and (4), we have assigned
the temporal reference of past and future tense, like objective epistemic
modality, to the tropic, rather than the phrastic (cf. 17.2). The distinc-
tion that we are operating with (which we have taken from Hare and, in
part, reinterpreted: cf. 16.2) is one that is not drawn in standard systems
of logic.
It is arguable that the speaker, in uttering (3) and (4), 1s putting it
forward as a fact that holds in we that it was or will be raining in wi.
St Augustine, in his meditations upon the nature of time and eternity in
Book x1 of the Confessions, was exercised by the problem of how the past
and the future can be said to exist: ‘There might be people who would
maintain that there are not three divisions of time but only one, the
present, because the other two do not exist. Another view might be that
the past and future do exist . . . Otherwise, how do prophets see the
future, if there is not yet a future to be seen? It is impossible to see what
does not exist. In the same way people who describe the past could not
describe it correctly unless they saw it in their minds, and if the past
did not exist it would be impossible for them to see it at all’’ (Confesszons,
XI: 17, translated by Pine-Coffin, 1961). Augustine reconciles these two
views by saying that the past and the future exist only in so far as they are
present in the mind: that there is “‘a present of past things”’ in memory,
‘present of present things’’ in direct perception, and ‘‘a present of
future things’’ in expectation (Confessions, XI: 20).
812 Modality
Augustine’s doleful question Quid est ergo tempus? (“ What, then, is
time?”’’) was taken by Wittgenstein (1953: 42ff) to be a classic instance
of the kind of question that philosophers should not ask. It has nonethe-
less seemed to be a genuine and important question, not only to genera-
tions of philosophers in earlier centuries, but also to many philosophers
and logicians of our own day (cf. Gale, 1968b). Furthermore, Augustine’s
attempt to answer the question has been taken seriously by scholars like
Prior (1967, 1968), who has probably done more than anyone else in
recent times to make precise, and to formalize, the logic of time and
tense. What we will, for convenience, refer to as the Augustinian view of
tense is obviously based more closely upon the experiential, than upon
the historical, mode of description (cf. 15.4). Time for Augustine 1s
something essentially subjective or psychological. We will come back to
this point.
What Augustine calls the “present of past things”’ can be described,
in terms that linguists have often used in their discussions of tense, as a
past embedded in a present. The semantic plausibility of this inter-
pretation of tense is revealed in such utterances as the following:
(3') It is a fact (in w;) that it was raining (in w))
(4’) It ts a fact (in w;) that it will be raining (in w)).
These sentences may not be exactly equivalent to ‘It was raining’ and
‘It will be raining’, but they do hold constant the present and the past or
future point of reference; and they help us to see what is meant by “a
present of past things” and ‘‘a present of future things’’.
The necessity of assigning two further points of temporal reference to
each utterance, #; and t;, in addition to the zero-point, to, established by
the act of utterance itself, becomes even clearer as soon as we compare
It was raining (yesterday) and It will be raining (tomorrow) with such
utterances as
(5) It had been raining (on the previous day)
(6) It would be raining (on the following day)
(7) It may have been raining (yesterday)
(8) It may be raining (tomorrow)
(9) It might have been raining (on the previous day) |
(10) It might be raining (on the following day).
 17.3. Tense as a modality 813
These may be paraphrased as follows:
(5°) It was a fact (in w;) that it had been raining (in a)
(6°) It was a fact (in wz) that it would be raining (in wy)
(7’) It is possible (in w;) that it was raining (in w,)
(8’) It is possible (in w;) that it will be raining (in wy)
(9’) It was possible (in w;) that it had been raining (in wy)
(10’) It was possible (in wz) that it would be raining (in wy).
Each of the modalized utterances (7)-(10), can be interpreted, on par-
ticular occasions of utterance, in terms of either subjective or objective
epistemic modality. For example, (7), and its paraphrase (7’), can be
construed as meaning either “Possibly (at fo) —1it is so (at %)—it be
raining at (¢;) or “I say (at to) — it is possibly so (at t) — it be raining (at
t;)"’; and in either case, tp = t > ty. (9) differs from (7) in that
to > ty > tj; and this 1s precisely the way in which (5) differs from (3).
Similarly, (10) differs from (8) in the same way that (6) differs from (4):
to >t < ty vs. to = hi < ty.
It follows from our consideration of these utterances that both the
tropic and the phrastic should be assigned a temporal index.® Extending
our symbolism to allow for this, we can represent the logical structure of
(3)-(6) as
(11) (to.) (tc .) (4D) |
and of (7)-(10) as
either
(12) (to poss) (t .) (47)
or
(13) (to .) (t« poss) (4p)
according to whether they are subjectively or objectively modalized.
We can now use the temporal indices to distinguish the worlds in
which propositions are true from the worlds of which propositions are
true (cf. 6.6). For example, when (5) is uttered to make a statement, it
8 Huddleston (1969) also argues, within the more orthodox framework of
transformational grammar, that utterances like (5)—(10) contain two tenses,
one associated with the auxiliary verb and the other with what at a deeper
syntactic level is the verb of its complement-clause.
814 Modality
can be construed as meaning “I assert here and now (in wp») that it was
true in w; that the proposition “it be raining”’ was true of w;’’. In other
words, its having rained (on the previous day) is a fact which belongs to,
or is part of w:, some previous state of the actual world; and ‘“‘it be
raining’’ is a proposition which describes, or is true of, another state of
the world, w;, prior to w;. Similarly, (8), when it is uttered to make an
objectively modalized statement, can be construed as meaning “I assert
(in wo) that it is possible in w; (= wo) that “it be raining”’ will be true
of w;(> Wo)’.
The world of fact and objective possibility (w:) is in principle a
different world both from the world described by the propositional con-
tent, or phrastic, of an utterance (w,;) and from the world in which the
utterance is made (w»); and these three worlds may be related temporally
in all sorts of ways. Given the right situation, even such sentences as
“Yesterday it was possible that 1t would rain tomorrow’ are perhaps
interpretable as being objectively modalized, the possibility of a future
event or state-of-affairs in w; > wo being asserted as a fact that was true
in some past state of the actual world, w; < wo, but is no longer neces-
sarily true in w». Sentences like ‘John was coming tomorrow’ can also
be interpreted in the same way: “I assert (in wo) that it was a fact in
wi (< we) that ‘John be coming”’ is true of w; (> we)’’.?
More commonly, however, in utterances in which the reference of t;
is past in relation to fo and the reference of t; is future in relation to Zo,
w; will be interpreted intensionally, rather than extensionally: not as a
world of facts and objective possibilities, but as a world that is composed
of subjective expectations, predictions and intentions. Fohn was coming
tomorrow, under an intensional interpretation of w:, might be held to
have roughly the same meaning (as an utterance, though not as a sen-
tence) as fohn said that he was/would be coming tomorrow, John intended
coming tomorrow, I thought/was told that Fohn was|would be coming
tomorrow, etc.
This distinction between an intensional and an extensional interpre-
tation of futurity can also be drawn in relation to sentences like (4), ‘It
will be raining (tomorrow)’. Ever since Aristotle first raised the ques-
tion, the factuality of statements descriptive, or predictive, of future

7 Palmer (1974: 38) says of sentences like ‘Yesterday John was coming to-
morrow’: ‘‘ The past tense . . . is epistemic — there was a time at which the
statements that John is coming tomorrow or John comes tomorrow were
valid’’. This notion of validity is captured in the analysis presented here by
the relativization of factuality to the intensional world indexed by wi.
 17.3. Tense as a modality 815
events, or states-of-affairs, has been philosophically controversial (cf.
Prior, 1955: 240); and many philosophers would deny that we can make
statements about the future at all, on the grounds that we cannot have
knowledge, but only beliefs, about future world-states. What purports to
be a statement describing a future event or state-of-affairs is therefore,
of necessity, a subjectively modalized utterance: a prediction rather than
a statement. Looked at from this point of view, the difference between
It will be raining (tomorrow) and It may be raining (tomorrow) depends
upon the speaker’s subjective evaluation of the probability of ‘‘it be
raining’”’ being true of w;. But we have already seen that It may be raining
(tomorrow), like It may be raining (now), can in principle be interpreted
as being either subjectively or objectively modalized; and the same is
true of It will be raining (tomorrow). The speaker can treat the future as
known, as a fact that belongs to w; (= wo), whether he is epistemologic-
ally justified in doing so or not. He can say, without doing violence to the
structure of English, I know that it will rain tomorrow (or I know that it
is going to rain tomorrow); and he can also embed a future-tense clause as
the complement of the factive verb ‘know’ in attributing knowledge of
the future to another, as in He knows that it will rain tomorrow. As we
have seen, the use of a factive verb in a context-free assertion commits the
speaker to the truth of the proposition expressed by the complement
clause (cf. 17.2). We must therefore allow for the possibility of seman-
tically well-formed descriptive statements with the logical structure
(14) (to) (te) (tp)|to = ts <
Utterances of this kind have the illocutionary force of acts of telling (cf.
He told me that it would be raining); and they are to be distinguished, in
principle, from predictions, wishes, promises, etc., all of which are
modalized in one way or another.
It is nonetheless a linguistically important fact that, unlike the
prophets referred to by Augustine, we are seldom in a position to lay
claim to knowledge of the future; and it is no doubt for this reason that
reference to future world-states is grammaticalized in the category of
mood, rather than tense, in many languages. The so-called future tenses
of the Indo-European languages, as we have already seen, are of secon-
dary development (cf. 15.4). There was no future tense in Proto-Indo-
European. But there was probably a fairly rich system of moods,
including, in addition to the indicative and the imperative, a subjunctive, .
an optative, and a desiderative. The desiderative*, as the term implies,
was used for the expression of wants, needs and desire. The optative*,
 816 Modality
however, which was preserved as a formally distinct mood in Greek and
Sanskrit was never, as far as we know, restricted to wishes (although the
term ‘optative’ might suggest that it was). It was the mood of contra-
factivity and remote possibility, as the subjunctive was the mood of
non-factivity and less remote possibility and the indicative the mood of
factivity and straightforward assertion. What appears to be the same
triple distinction of contra-factivity, non-factivity and factivity is gram-
maticalized in certain American Indian languages (e.g., Hopi) which,
unlike the Indo-European languages, lack the category of tense entirely
(cf. Whorf, 1950); and it has often been suggested that such languages,
which grammaticalize modal, rather than either deictic or non-deictic
temporal, distinctions are in some sense more subjective than languages
that grammaticalize tense. Interesting though this question is, we will
not pursue it here. Our main purpose is to emphasize the connexion
between subjective epistemic modality and reference to the future.
That there is a connexion is perhaps obvious enough, independently
of any arguments that might be based upon the grammatical structure of
particular languages. But there is, in fact, ample linguistic evidence.
What is conventionally regarded as the future tense (in languages that
are said to have a future tense) is rarely, if ever, used solely for making
statements or predictions, or posing and asking factual questions, about
the future. It is also used in a wider or narrower range of non-factive
utterances, involving supposition, inference, wish, intention and desire.
Furthermore, the future vs. non-future distinction is frequently neutral-
ized in subordinate or negative clauses, in participles and nominaliza-
tions, in association with all moods other than the indicative, and in
various other constructions: this fact would suggest that the opposition
of the future to the present is less central in the structure of the lan-
guages than is the opposition of the past to the present (which Is, in any
case, an opposition that is more widely grammaticalized throughout the
languages of the world). So too would the fact that the so-called future
tense is in many languages constructed according to a different pattern
of formation than is the past or the present tense.
There is also a good deal of diachronic evidence to support the view
that reference to the future, unlike reference to the past, is as much a
matter of modality as it is of purely temporal reference. ‘Throughout the
history of the Indo-European languages what are traditionally described
as future tenses have invariably been created, independently in different
languages, from word-forms or phrases that were originally used to
express, not futurity as such, but various kinds of non-factivity. We
 17.3. Tense as a modality 817
cannot go into the details of the historical development of the forms that
came to be used for reference to the future in different languages. It is
sufficient to point out that the most common sources of the future
tenses in the Indo-European languages are, on the one hand, the sub-
junctive, which we have classified as the mood of generalized non-
factivity, and, on the other, word-forms or phrases expressing the more
specific non-factive notions of intention and desire. Future tenses out-
side the Indo-European languages, in so far as their historical develop-
ments can be traced, commonly appear to have had a similar origin. The
fact that the so-called future tense in English is formed by means of an
auxiliary verb, ‘will’, which originally expressed intention, is by no
means an isolated historical accident. Nor is the fact that English sen-
tences containing the so-called future tense may be used to express
various kinds of subjective epistemic modality including inference,
supposition and prediction. There is a demonstrable historical con-
nexion between reference to the future and non-factivity in too many
languages for it to be regarded as a matter of accident that languages
should rarely, if ever, distinguish systematically between statements of
fact about the future and subjectively modalized predictions.
There is also ample synchronic and diachronic evidence for the con-
nexion between prediction and deontic* modality, which we shall be
discussing in a later section. It was pointed out earlier (16.2) that there
is a sense in which directives necessarily refer to the future; and also
that in many languages the functions of the subjunctive merge with those
of the imperative. The relationship between the subjunctive and the
imperative is traditionally accounted for in terms of the notion of will.
But there is no formal difference in the Indo-European languages
between the two kinds of subjunctive that are traditionally distinguished
by grammarians as the subjunctive of will and the subjunctive of likeli-
hood (or possibility); and it is arguable that there is no difference of
meaning between them either.
In many languages throughout the world, the notions of possibility
and obligation are associated with the same non-factive, or subjunctive,
mood; and this is commonly also the mood of prediction, supposition,
intention, and desire, as it is in many of the Indo-European languages.
The so-called subjunctive of will is related to the imperative and to
deontic modality in traditional treatments of the Indo-European system
of moods and tenses; and the so-called subjunctive of likelihood to the
future tense and to prediction. But, as we have just seen, the future tense
in many languages, within and outside the Indo-European family, has
} 818 Modality
developed from forms that originally expressed desire and intention; and
desire and intention can hardly be separated from will. The notions of
will and likelihood, as these notions are employed in standard treatments
of the Indo-European system of mood, can therefore both be subsumed
under the more general notion of non-factivity.
It was because the subjunctive was the mood of non-factivity that it
could be used, in contrast with the indicative, for the expression of either
likelihood or will: i.e. in either epistemic or deontic utterances. We do
not need to say that the Proto-Indo-European subjunctive as such had
two distinct meanings, in order to account for the possibility of its being
used in predictions, like the future, or in directives, like the imperative.
The distinction between a prediction and a directive, like the distinc-
tion between a prediction and a promise is a matter of illocutionary force;
and this is not necessarily grammaticalized in the structure of primary
performatives (cf. 16.1). The sentence ‘You will be here at three
o'clock’ can reasonably be said to have the same meaning whether it is
uttered as a prediction or a directive, just as the sentence ‘I will be here
at three o’clock’ can be said to have the same meaning whether it is
uttered as a prediction or a promise.
So far we have tacitly assumed that all non-actual world-states must
be either past or future in relation to wo, the world as it actually is at fo.
But this is not so. We can make contra-factive assertions about states of
the world as it might have been at to. We can say, for example,
(15) If he hadn’t missed the plane, he would now be in London.
The reason for mentioning such utterances in this section is that in
many languages, including English, the grammatical category of past
tense is regularly used to convert a non-factive into a contra-factive
utterance.® Unlike (15), the following is non-factive, rather than contra-
factive:
(16) If he hadn’t missed the plane, he will now be in London.
Similarly in Latin, which, unlike Greek and Sanskrit, did not preserve a
distinctive mood of contra-factivity and remote possibility (the opta-
tive), the past tense of the subjunctive could be employed in contrast
with the present tense of the subjunctive to distinguish between non-
factive and contra-factive statements. ®

8 There is also a non-factive interpretation of (15): cf. p. 796, n. 4 above.

® On Latin, cf. Handford (1947), Householder (1954), Palmer (1954); and on
mood in the Indo-European languages in general, cf. Gonda (1956), Kurylo-
 17.3. Tense as a modality 819
It seems reasonable to interpret this use of the past tense as one
which, when it combines with epistemic modality, represents a possible,
but non-actual, state of the world at tp as being more remote than what
is or may be the actual state of the world at t,. In other words, the
deictic distinction of ‘‘then”’ vs. “‘now’’, which is the temporal correlate
of “‘there’’ vs. ““here’’, is used to express, not only the temporal remote-
ness of w,; from wo = wi (and, more particularly, its being in past time
in relation to wo = w;), but also what might be called its modal remote-
ness. This would seem to be the explanation for the fact that the past
tense form of the modal verb is used in English sentences like

(17) We could be in Africa

which ‘‘can be meaningfully used, say, on a fine night in Scotland”’
(Boyd & Thorne, 1969: 73). The possible world, w;, of which the
proposition expressed by (17) is true and the intensional world, wi, in
which this proposition is true are envisaged as being contemporaneous
with, but modally remote from, wo.
It is worth noting, in this connexion, that (17) is not necessarily
contra-factive. The person uttering (17) might equally well add to his
utterance either but we’re not or but I doubt it. The admittedly rather
vague and intuitively based notion of modal remoteness is intended to
cover both interpretations; and it cuts across the distinction between
contra-factivity and non-factivity drawn in the preceding section (17.2).
Contra-factivity is simply a special case of subjectively modalized
remote possibility; and, as we have just seen, this, like tense, is connec-
ted with deixis. What is grammaticalized in Latin and English (and
many other languages) in terms of tense, within non-factive modality, is
very similar to, if not identical with, the distinction that was gram-
maticalized in Greek and Sanskrit in terms of mood in the opposition of
the optative and the subjunctive.
It might even be argued that what is customarily treated as being
primarily an opposition of tense — past vs. non-past —in English and
other languages, should be more properly regarded as a particular case
of the distinction, remote vs. non-remote (“‘then”’ ws. ““now” being a

wicz (1964). Mood in Latin and Ancient Greek is treated from the viewpoint
of generative semantics (and on the basis of postulated underlying abstract
performative verbs: cf. 10.5, 16.5) by Lakoff (1968) and Lightfoot (1975)
respectively. On the non-factivity of the Spanish subjunctive, in certain con-
texts, cf. Rivero (1971). .
820 Modality
particular case of “‘there’’ vs. “here’’).1° Under this interpretation, tense
would bea specific kind of modality; and modality would be more closely
related to deixis. What is conventionally described as the present tense
would be the product of non-remoteness (“now’’) and factivity; past-
ness and futurity would not be defined directly in terms of temporal
indices (to = t; > tj; and tp = t; <t,;), but in terms of remoteness
(“then’’) and either factivity or non-factivity, the so-called past tense
being the product of remoteness and factivity, and the so-called future
tense being the product of non-remoteness and non-factivity; and
contra-factivity would be the product of remoteness and non-factivity.
There are certain obvious advantages in an analysis of this kind. It
would more directly reflect the difference in the epistemic status of the
past and the future. This difference is often expressed in the philoso-
phical literature by saying that, whereas the past is closed, the future is
open; and that, as a consequence, whereas statements about the past are
either true or false at the time of their utterance, statements about the
future (or, at least, about the occurrence of contingent events in the
future) are neither true nor false, but indeterminate in truth-value, at
the time of their utterance (cf. Gale, 1968b: 169ff). Regardless of
whether we say that statements about so-called future contingents have a
determinate truth-value or not (and this is a philosophically contro-
versial question with which we need not be concerned), we can hardly
deny that there is a difference in the epistemic status of the past and the
future; and, as we have seen, there are languages in which it is the
difference in the epistemic status of the situations being described,
rather than their temporal location in relation to the zero-point of
utterance that is of primary importance in determining the selection of
one so-called tense rather than another. For such languages (e.g., Hopi),
if not for all, what is often described as tense (cf. Hockett, 1958: 237) is
surely more properly regarded as the grammaticalization of epistemic
modality (i.e. as mood).
We will not pursue this question any further. What has been said will
be sufficient to show what is meant by saying that tense is a kind of
modality; and it is at least conceivable that different language-systems
should attach greater or less weight to temporal, rather than epistemic,
considerations in the grammaticalization of what is traditionally de-
scribed in terms of past, present and future. Certain systems of tense-
logic give recognition to the’epistemic difference between the past and the
10 This is the view that Joos (1964) takes of the past vs. non-past opposition in
English.
 17.3. Tense as a modality 821
future, but they all seem to take the temporal notions of presentness and
precedence to be basic in the logical definition of tense (cf. Rescher &
Urquhart, 1971: 52ff). It may very well be that different kinds of tense-
logic are appropriate for the formalization of the tensed propositions
that are expressible in different languages and that, for certain languages
at least, tense-logic and epistemic logic will be indistinguishable. We
shall continue to operate with the temporal indices ¢o, ¢; and tj and with
the notions of pastness and futurity that are definable in terms of them.
In doing so, however, we recognize that something other than temporal
considerations may be involved.
The notational system developed, in this chapter, for the representa-
tion of tense is well adapted for the representation of what we described
earlier as the Augustinian view of tense: the view that past, present and
future are all located (in memory, observation or anticipation) in the
experiential present. The normal condition would therefore be t; = to:
i.e. the intensional world (w;) from which, as it were, we are asked to
look at the extensional world (w;) is the one that is defined as being
identical, temporally, with the speaker’s world. Deviation from this
norm would involve deictic projection into a past (t < to) or a future
intensional world (cf. 15.4). This would account for the use of so-called
secondary tenses in such texts as: Fohn was in a quandary (to = ti > ty)
—it was raining (t, >t; =t,;)—he had caught a cold on the previous
occasion (t, > t; > t;)- he would see her (anyway) on the following day
(to > t; < t;). Here, it will be noted, everything after John was in a
quandary (which has primary tense, defined by the normal condition
to = t;) involves secondary tense (cf. to > ti).
Of particular interest, in this connexion, is the analysis of
(18) It was raining
as
(19) (fo .) (te .) (tip)/to > te = by
(i.e. aS meaning, as it were, “It was a fact that it is raining’’) rather
than as
(20) (to.) (ta -) (stp)/to = te > by
(i.e. as “It is a fact that it was raining’’). It so happens that in English,
though not in all languages, the past tense can be used either as a
primary or a secondary tense: the difference between (19) and (20)
would come out quite strikingly in a language like Latin, which would
use a so-called indirect discourse construction for (19).
 822 Modality
Granted that the notational system developed here for the representa-
tion of tense accounts neatly enough for the Augustinian view of tense
(and for deictic projection, or empathetic deixis, in so far as this deter-
mines the use of the so-called secondary tenses), it may be seen as an
objection to the system that it makes truth and factuality relative to some
temporally indexed intensional world. But truth, according to our
everyday view of the matter, 1s timeless (i.e. eternal); and, as was pointed
out above, for anyone who takes this view, “‘It is the case that p”’ will be
a timeless proposition, regardless of whether p itself is time-bound or
timeless (cf. 15.4).
There is no difficulty in representing the notion of objective and
eternal truth within the notational system that we have devised here. All
that is required is the restriction of the value of t; to timelessness: let us
symbolize this as tx. Tense would then be simply a matter of the
relation between ¢o and ¢;; and (18) would be analysed as
(21) (to.) (t¢-) (4p) /to > te; ty = be
(i.e. “It is timelessly true that at t; < fo it be raining”’).
The condition t; = ta would have exactly the same meaning (whatever
that might be) as the traditional phrase ‘sub specie aeternitatis’ (‘‘from
the viewpoint of eternity’’).
The problem lies in trying to make sense, philosophically, of a
formula like (21). This is a problem that faces anyone who tries to
reconcile the objectivity and eternity of truth with the subjectivity of
temporal becoming. It is more acute, of course, for a theologian who,
like St Augustine, wishes to reconcile the omniscience of God with the
openness of the future and with free will. But it is a problem which
arises, independently of all such considerations, in any serious analysis
of our everyday assumptions.
In so far as we lay claim to factual knowledge of the past and the
future, we assume that the factual status of what we know at any point
in time is unchanging and eternal, even though some of the propositions
that we know are tensed propositions (cf. 15.4). If it makes sense to say
_ _I know that it was raining yesterday and I know that it will rain tomorrow,
we must allow that both 4 = ¢) and t= to are assumed to hold
simultaneously by the person making either of these utterances. It is
part of what some philosophers would refer to as the logic of the verb
‘know’ that anyone using it commits himself to the objectivity and
eternity of what he claims to know. If, for any reason, he subsequently
 17.4. Deontic modality 823
comes to the view that what at f, he claimed to know is not true, he will,
of course, say that he was wrong in what he said earlier, rather than that
what was true in wij = wo has now become false. It is by making this
kind of practical adjustment that we resolve, in everyday matters, the
problem of reconciling the eternity of truth with the temporality and
subjectivity of the viewpoint from which we assert that something is
true. We make the same kind of adjustment as we switch from the
experiential to the historical mode of discourse and from the dynamic to
the static conception of time (cf. 15.4).
The notational system that has been introduced in this section
neither resolves, nor does it beg without resolving, the philosophical
controversy between the temporalistic and the timeless theories of truth,
which “has raged for over two thousand years, with able philosophers
lining up on different sides’’ (cf. Gale, 1968a: 137). What the system
does, and is intended to do, is to make possible the treatment of tense
from an epistemic point of view. It allows us to relativize truth to some
time-bound intensional world (t; 4 t). But it does not force us to do
this: we can always hold to the notion of absolute truth (4; = tx), if we
wish. There are considerable advantages, however, as far as the analysis
of tense in natural languages is concerned, in allowing for the maximum
degree of flexibility in the assignment of values to t.

17.4. Deontic modality

The term deontic* (from the Greek ‘deon’: “what is binding’’) is now
quite widely used by philosophers to refer to a particular branch or |
extension of modal logic: the logic of obligation and permission (cf. Von
Wright, 1951). There are certain obvious differences between alethic
and epistemic necessity, on the one hand, and what we might call deontic
necessity (i.e. obligation), on the other. Logical and epistemic necessity,
as we have seen, have to do with the truth of propositions; deontic
modality is concerned with the necessity or possibility of acts performed
by morally responsible agents. When we impose upon someone the
obligation to perform or to refrain from performing a particular act, we
are clearly not describing either his present or future performance of
that act. There is a sense in which the sentence we utter can be said to
express a proposition; but it is not a proposition which describes the act
itself. What it describes is the state-of-affairs that will obtain if the act
in question is performed; and we have already seen that directives can
be analysed, along these lines, as utterances which impose upon someone
the obligation to make a proposition true (or to refrain from making it
824 Modality
true) by bringing about (or refraining from bringing about) in some
future world the state-of-affairs that is described by the proposition. The
notion of truth is not therefore irrelevant to the analysis of directives (the
only class of deontically modalized utterances that we have so far been
concerned with), but it applies less directly than it does in the analysis of
subjectively or objectively modalized statements.
A second difference between deontic and logical or epistemic modality,
which was mentioned in the previous section, is that there is an intrinsic
connexion between deontic modality and futurity. The truth-value of a
deontically modalized proposition is determined relative to some state of
the world (w;) later than the world-state (wi) in which the obligation
holds; and the world-state in which the obligation holds cannot precede,
though it may be simultaneous with, the world-staté“(w.) in which the
obligation is imposed. This does not mean that we cannot assert at time
to that someone was under an obligation at t; < to to perform a particular
act at tj > tj. We can say, for example:
(1) You should have gone to the meeting yesterday.
But in uttering (1) we are not imposing upon the addressee the obliga-
tion to go to yesterday’s meeting; we are asserting that he was at t; < 4,
under the obligation to go. We are making a statement, rather than
issuing a directive. The analysis of utterances like (1) — deontic state-
ments — will be taken up later in this section.
A third characteristic of deontically modalized utterances which
differentiates them, or appears to differentiate them, from logically and
epistemically modalized utterances is that deontic necessity typically
proceeds, or derives, from some source or cause. If X recognizes that he
is obliged to perform some act, then there is usually someone or some-
thing that he will acknowledge as responsible for his being under the
obligation to act in this way. It may be some person or institution to
whose authority he submits; it may be some more or less explicitly
formulated body of moral or legal principles; it may be no more than
some inner compulsion, that he would be hard put to identify and make
precise. Philosophers, in their discussion of deontic modality, have been
mainly concerned with the notions of moral obligation, duty and right
conduct. But it seems preferable for the linguist to take a maximally
inclusive view of what constitutes obligation, drawing no distinction,
in the first instance at least, between morality, legality and physical
necessity. In the analysis of the lexical structure of particular languages,
distinctions will need to be drawn between various kinds of obligation;
 , 17.4. Deontic modality 825
but many of these will be culture-dependent, if not purely language-
dependent, and will have to be correlated with institutionalized beliefs
and norms of conduct. We will here assume that there is some univer-
sally valid notion of obligation which may be variously categorized and
differentiated in terms of its courses and sanctions in different cultures;
and we will further assume that such distinctions as are lexicalized in
words like ‘right’ and ‘wrong’ (and, at a lower level of specificity,
‘unconstitutional’, ‘illegal’, ‘improper’, ‘immoral’, ‘blasphemous’,
‘taboo’, ‘unjust’, etc.) are naturally regarded as deriving from the
modification of a single very general deontic operator, comparable with
the operators of alethic and epistemic modality. Different kinds of
deontic modality can then be distinguished by specifying the source or
cause of the obligation.
The logical structure of directives was accounted for in an earlier
section in terms of a component meaning “‘so be it”, symbolized by an
exclamation mark, in the tropic position (cf. 16.2, 17.2). We must now ,

as
relate this so-be-it component to the modal notions of necessity and
possibility, bearing in mind the fact that deontic modality, like predic-
tion, involves a reference to a future world-state and that it is connected
in some way with intention, desire and will. If we assign tense-operators
to the neustic, tropic and phrastic of directives, we can represent the
logical structure of a command like
(2) Open the door

(3) to . ty! tip

and we can make it a condition upon the interpretability of (3) that zo
should precede or be simultaneous with # and that t; should precede
ty: Le. to < ti < ty. It is at least partly by virtue of these temporal
relationships that mental or illocutionary acts involving intention and
will differ from mental or illocutionary acts involving wishing. We can
want it to have been true in w < w, that a certain proposition p (e.g.,
“Tt be raining’’) be true of w; < w;: cf. the non-factive
(4) I want it to have rained on the previous day
and the contra-factive
(5) L wish it had rained on the previous day.
But, as we saw earlier, we cannot rationally will or intend something to
826 Modality
happen, or to have happened, in the past; nor can we impose upon some
agent the obligation in w; to bring about in w; the state-of-affairs
described by p unless fo < t < tj. It is only if these temporal condi-
tions are satisfied that an utterance can be ambiguous or indeterminate
in respect of the distinction between logical or epistemic modality and
deontic modality.
The origin of deontic modality, it has often been suggested, is to be
sought in the desiderative and instrumental function of language: that
is to say, in the use of language, on the one hand, to express or indicate
wants and desires and, on the other, to get things done by imposing
one’s will on other agents. It seems clear that these two functions are
ontogenetically basic, in the sense that they are associated with language
from the very earliest stage of its development in the child. It is equally
clear that they are very closely connected. It is a small step from a
desiderative utterance meaning ‘‘I want the book”’ to an instrumental
utterance meaning “‘Give me the book’’; and parents will commonly
interpret the child’s early desiderative utterances as mands, thereby
reinforcing, if not actually creating, the child’s developing awareness
that he can use language in order to get others to satisfy his wants and
desires. The so-be-it component that we have introduced into our
representation of the logical structure of directives is intended to have an
instrumental meaning, but it may be assumed to have developed out of,
and to include, an ontogenetically prior desiderative. To issue a directive
that one does not want to be carried out is to be insincere in the per-
formance of one’s illocutionary act: Open the door, but I don’t want you to
is anomalous in the same way that The door is open, but I don’t believe it
is. It is not surprising therefore to find that there are languages in which
no sharp distinction can be drawn between, in primary performatives at
least, between desiderative and instrumental utterances: between ‘I
want X to bring it about that p”’ and “‘ Let X bring it about that p’’. The
so-called subjunctive of will in the Indo-European languages can be
interpreted in either way; and the imperative mood, whose characteristic
function we have taken to be instrumental, is regularly employed also
for the expression of wishes, hopes and desires (cf. Have a good timel,
Get well soon!, Give us this day our daily bread!, etc.).
Language, as the child soon realizes, is not only an instrument that he
can use in order to impose his will on others and get them to do what he
wants. It is an instrument that others can employ to impose their will on
him, regulating his behaviour by command and prohibition. The child’s
compliance with the directives that are addressed to him may depend, in
 17.4. Deontic modality 827
. part, upon the maturation of an innate sense of right and wrong or a
propensity to social conformity. Alternatively, as a behaviourist might
claim, it may be explicable wholly in terms of sanctions and rewards
(cf. 5.2). This is a question upon which we need take no stand. What
concerns us here is the fact that commands and prohibitions do not of
themselves compel obedience: the addressee, if he understands their
illocutionary force, will know that the person issuing them wants him to
act or to refrain from acting in a certain way. But the addressee must
have other grounds for complying with a directive over and above his
recognition that it is the speaker’s will that he should act or refrain from
acting in this way. He must acknowledge that the speaker has the
authority or power to impose his will upon him; and whatever it is that
is acknowledged as establishing the speaker’s authority to issue the
directive is what was described above as the source or cause of the
obligation or deontic necessity in the particular instance. The speaker
may back up his commands and prohibitions with threats or explana-
tions; but he need not do so. If it is clear that he has the authority to
issue directives, his commitment to “‘so be it”? by means of an un-
qualified and unexplained “‘I say so”’ will suffice.
In all that has been said so far about directives, it has been assumed
that they are necessarily addressed to other agents (normally other
human beings) whose compliance or non-compliance depends upon the
exercise of their own free will; and we shall continue to limit our con-
sideration of directives and other kinds of deontic utterances to those
which can reasonably be said to satisfy this condition. It should not be
forgotten, however, that there are many cultures in which certain
language-utterances (provided that they are produced by an authorized
person in the appropriate circumstances) are held to have a magical or
sacramental effect; and such utterances may take the form of directives
addressed to either animate or inanimate entities. What is especially
interesting about magical or sacramental directives in the present con-
nexion is that (according to those who believe in their efficacy) their
“so be it (in w,) that p (in w;)” automatically guarantees the truth of
‘it is so (in zw) that p (in w;)”’: their perlocutionary effect is an automatic
consequence of their illocutionary force. Utterances such as Austin’s
I name this ship ‘Liberté’ (cf. 16.1) are not sharply to be distinguished
from utterances which we would more readily describe as magical or
sacramental. Furthermore, even the most rational and secularized
member of a modern industrialized society is wont, at times, to issue
directives, threats or pleas to the machines and gadgets with which he is
 828 Modality
surrounded. He may not believe that his utterances will have any direct
effect upon their addressee. But the fact that he can produce them at all
and that others will accept them as normal means that we cannot,
strictly speaking, make it a condition upon the utterance of directives
that they should be addressed to other agents.
The range of utterances whereby parents and those in authority over
him regulate the behaviour of the child and inculcate in him (if they are
successful) a set of moral and social norms and patterns of behaviour
includes not only directives, but also statements. The parent can say, not
only
(6) Don’t tell lies,
| but also,
(7) It’s wrong to tell hes.
The difference between (6) and (7), as they would normally be under-
stood, is that (6) is (or may be) a directive, the utterance of which
creates, or brings into existence, a certain obligation, whereas (7) is a
deontic statement* to the effect that the obligation exists; and the word
‘wrong’ in (7) specifies, in a fairly general way, the nature of this
obligation. The problem that now confronts us — and it is one of the
central problems in the formalization of deontic logic — is to say pre-
cisely how directives are related to deontic statements. Granted that the
logical structure of (6) may be represented as
(8) to. tg~! tip
(“I say in wo — let it not be so in w;— that p holds of w;’’), what is the
logical structure of (7), construed as a deontic statement?
It should be noted first of all that both (6) and (7) can be understood
as referring to either a restricted or an unrestricted obligation. We
might wish to say, for example, that the prohibition on telling lies is
unrestricted*, or absolute, in the sense that it holds at all times and 1n all
possible worlds: that the validity of this obligation is analogous with the
truth-value of logically necessary propositions. Whether we should take
this view of any particular obligation is of course a philosophical ques-
tion. The point that concerns us here is that it is possible in principle to
distinguish between unrestricted and restricted obligations; and this
distinction can be accounted for by quantifying over the worlds in
which the obligation holds. If we believe that there are certain unrestric-
ted obligations which are created by a directive uttered at a particular
 17.4. Deontic modality 829
point in time, fo, then we can say that the obligation exists in all world-
states simultaneous with and subsequent to wo. Many people might take
this view with respect to the ‘Ten Commandments which were issued,
through Moses, on Mount Sinai, saying that they had been issued by
God to man, or to the chosen people, and were henceforth in existence
for all time. An alternative view of those obligations which are held to
be of unrestricted or absolute validity in a particular society or a par-
ticular culture is that they derive their validity, not from their having
been created by the directives issued, at some point in time, by some
sovereign human or supernatural agency, but from omnitemporally or
eternally applicable principles governing social behaviour. According to
this view, such principles would be analogous to the so-called laws of
nature; and deontic necessity, in so far as it is grounded in these
principles, would be analogous to logical necessity. In most societies the
generally accepted moral and legal obligations governing the behaviour
of their members are associated in the traditions and myths of these
societies with some identifiable divine, heroic or secular authority which
created the obligations by means of a directive; and parents, priests,
judges and others who establish and maintain the norms of conduct
within the society by means of deontic statements, such as (7), see their
function as that of bringing to the attention of those over whom they
exercise some institutionally recognized authority the existence of
obligations which have been created in the past by directives, such as
(6), issued by some higher, and ultimately sovereign, authority. In this
sense, therefore, it might be argued that, whatever view we take of the
logical status of unrestricted obligations, our understanding of their
force is based, at least analogically, upon our understanding of the way
in which directives function in everyday social interaction.
Most of the directives that are issued in everyday social interaction
do not have as their function the creation of an obligation to behave in a
certain way or to refrain from behaving in a certain way for all time or
on all occasions. The obligation will be explicitly or implicitly restricted
in one respect or another. One way of restricting an obligation explicitly
is by means of a conditional clause attached to the clause which contains
the so-be-it component: cf. If you get home before I do, turn the thermo-
stat up. In connexion with conditional directives of this kind, it should
be noted that the condition normally applies, not to the existence of the
obligation, but to the propositional content of the directive. They are to
be construed as meaning “‘I hereby (in wo) impose upon you the obliga-
tion (in w; 2 wo) to bring it about — if g holds (in wx < wy 2 wi) — that
 830 Modality
p (in w;)””. It is the fulfilment of the obligation, rather than the obliga-
tion itself, that is restricted. Corresponding deontic statements can,
however, be ambiguous with respect to this distinction: there is a
logically important difference between “It is obligatory (in wz) that, if ¢
holds (in we < wy > wi), then p should be made true (of a)” and “If g
holds (in wx < wi), then it is obligatory (in w;) that p should be made
true (of w; > w:)’’. The former, as a whole, expresses a proposition
which is true or false, depending on whether there does in fact exist an
obligation to act in such-and-such a way under certain conditions: it is a
statement of conditioned obligation. The latter is a conditional state-
ment of obligation: it says that the existence of the obligation is depen-
dent upon the satisfaction of certain conditions in some world (wz) prior
to, or simultaneous with, the world (w;) in which the obligation would
hold.
An obligation may also be restricted explicitly by means of an adverb
or an adverbial clause or phrase of time; and this adverb or adverbial
expression may have either singular or generic reference, and it may
refer to either a point or period of time. For example, in Turn up the
thermostat, when you get home the expression ‘when you get home’ may
refer to some single individual event of home-coming; and it would
normally, though not necessarily, be understood in this way, if it were
supplemented with a phrase like ‘this evening’. But the same expression
might also be employed with generic reference; in which case it would
be more or less equivalent to ‘whenever you get home’. In certain
languages this difference between singular and generic temporal
reference will be grammaticalized in terms of the category of aspect*
(cf. 15.6). In English, however, the imperative mood is aspectually un-
marked and the same is true, as we have just seen, of certain adverbs and
adverbial expressions of time with which obligations may be explicitly
restricted. It follows that the same sentences may be used to impose
either a restricted or an unrestricted obligation; and even when the
obligation is explicitly restricted by means of some temporal expression,
it may still be unclear from the form of the utterance itself precisely what
the temporal restriction is.
Generally speaking, however, it seems to be the case that, unless there
| is something in the utterance or in the context-of-utterance which makes
it clear that it is not to be so interpreted, a directive which imposes an .
explicitly unrestricted obligation will be understood to bring this —
obligation into existence with effect from the time of utterance: that is
to say, the world-state (w;) in which the obligation holds will follow
 17.4. Deontic modality 831
immediately upon the world-state (wo) in which the utterance is made
and may be regarded, for practical purposes, as simultaneous with it. If
the obligation is such that its fulfilment would consist in the perfor-
mance of a single act (e.g., that of opening the door), it will remain in
force (unless it is cancelled in the meantime by the issuing authority
or by some higher authority) until the act in question has been duly
performed or circumstances beyond the control of the person upon
whom the obligation is imposed render its performance impossible or
irrelevant. If the obligation is such that its fulfilment would consist in
the performance of a generic act (e.g., that of telling the truth or coming
to work at some prescribed time every day), it will remain in force as
long as the authority imposing the obligation endures and is acknow-
ledged.
So much for the temporal relationship between wo» and w; in direc-
tives. What now of the temporal relationship between w; and w;? We
have already seen that it is a general condition upon directives that
ti < tj. If the obligation that is created by means of the directive is
explicitly restricted, the interval between ¢ and #; can be of any length
whatever: cf. Come and see me now|next week/in twenty years. If the
obligation is explicitly unrestricted, it would seem to be the case that,
just as ¢; is understood to follow immediately upon fo, so t; is under-
stood to follow immediately upon z# — ‘immediately’ in this case being
interpreted as meaning “‘as soon as is practicable or reasonable”’. We
cannot, and need not, be more precise than that in our specification of
what ‘immediately’ means. For this will be determined, not only by the
nature of the act or course of action that 1s imposed by the directive, but
also by a variety of social conventions governing interpersonal relations
and behaviour. The important point is that both the person imposing
the obligation and the person upon whom it is imposed will normally be
aware of what the implicit “immediately”’ means on particular occa-
sions; and that is sufficient. Henceforth, we shall be concerned with
directives which impose upon those to whom they are addressed the
obligation to carry out immediately the act or course of action that is
prescribed; and we will refer to them, whether they are conditional or
unconditional, as temporally unrestricted directives. The vast majority
of directives (and requests) that are issued in everyday language be-
haviour are temporally unrestricted, unless they contain an adverbial
expression of time (‘soon’, ‘next week’, ‘when you have a moment’,
etc.) which explicitly, and more or less precisely, specifies the world-
state in which they are to be complied with.
832 Modality
17.5. Obligation, permission, prohibition and exemption
We have so far discussed deontic utterances, both directives and state-
ments, in terms of the notion of obligation; and we have seen that there
is at least an intuitive relationship, which we have yet to explicate,
between utterances such as
(1) Open the door
and
(2) Don’t open the door,
construed as commands and prohibitions respectively, and statements to
the effect that there exists an obligation to act or to refrain from acting
in a certain way. But we must now bring in the notion of permission,
which is related to possibility in the same way that obligation is related
to necessity. If X is not obliged to do a (where a is either an individual
or a generic act), he is permitted not to do a; and if he 1s obliged to do a,
he is not permitted not to do a: cf. ~ nec p — poss ~ p and nec p >
~ poss ~ p. Also, if X 1s permitted to do a, then he is not obliged not
to do a; and if X is not permitted to do a he is obliged not to do a: cf.
poss p -> ~ nec p and ~ poss p - nec ~ p. This parallelism between
necessity/possibility and obligation/permission is exploited in all the
standard systems of deontic logic (cf. Hilpinen, 1971; Rescher, 1966).
Some of these take permission as primitive and define obligation in
terms of it; others take obligation to be the most basic notion. We shall
return to this question presently.
But first let us consider how (1) and (2) differ from, and how they are
related to, the following:
(3) You must open the door
(4) You mustn’t open the door
(5) You may open the door
(6) You needn’t/don’t have to open the door.
One important difference 1s that, whereas (1) and (2), would normally be
uttered and taken as directives, (3)~-(6) may be interpreted either as
directives (e.g., as meaning “I hereby impose upon you the obligation
to open the door’’) or as statements (e.g. as meaning “I hereby assert
that you are obliged (by some unspecified authority) to open the
door’’). In either case it is an unqualified act of telling that is involved;
 17.5. Obligation, permission, prohibition and exemption 833
and this is true also of (1) and (2). We cannot therefore account for the
difference between the two interpretations in terms of the qualification
of the I-say-so component. Furthermore, since (3)~(6) can be used as
statements we must, for this interpretation at least, allow for their
having an unqualified it-is-so tropic. Otherwise, we could not account
for the fact that the addressee can not only refuse to comply with the
obligation; he can deny that any obligation exists (That’s not true: I don’t
have to) or question the existence of the obligation by querying its
source (Who says so?). Nor could we account for the fact that the state-
ment made by (3) may be either subjectively or objectively modalized:
(7) You may have to open the door
(‘Perhaps you have to open the door”’ or “‘I tell you that it is possible
that you have to/will have to open the door”’). The only way of account-
ing for the use of (3) as a statement, it would seem, is by treating its
deontic modality as being part of the propositional content or phrastic.
But this must be done without destroying the intuitive relationship that
holds between (3) and (7).
One alternative, and possibly the most satisfactory, is to treat (3) as
being equivalent to “I say so — it is so — (that) X obligates your opening
the door”’, where X stands for the unidentified source of the obligation
and the verb ‘obligate’ is a two-place predicator taking as one of its
arguments an expression referring to the source of the obligation and
the other a so-be-it modalized proposition (“‘that the door be opened’”’).
The proposition that is asserted in uttering (3) would be true of wy, (at
t;) if, and only if, X has performed the act of obligating at tj: i.e. if X
has successfully created the obligation by giving his authoritative ‘I say
so’’ to “‘so be it” at ty.
This analysis of the meaning of (3) is satisfactory enough as far as it
goes. But it fails to account for the fact that in making a deontic state-
ment we commit ourselves to the existence of the obligation in w;. Under
our analysis, You must open the door, construed as a deontic statement,
would mean something like “It 1s now the case (at # = t,) that X
puts/put/will put you under the obligation (at ¢) to open the door (by
saying, at t;, Open the door!)’’. It is a simple enough matter to impose the
requirement that # > ¢. But this will only partly solve the problem.
What is wrong with our analysis is that we have interpreted the predica-
tor ‘obligate’ as a verb which denotes an act, and ¢; as the point of time
at which this act is performed. The interpretation that is required, how-
ever, is one under which ‘obligate’ is taken to denote the state resulting
834. Modality
from, or initiated by, some previous event or act. Provided that we take
‘obligate’ in this sense, we can interpret You must open the door as
meaning “I say so (at ¢o)—it is so (at t; = ¢o)-—that X obligates (at
t; = t;) that p be so”’; and this is the kind of interpretation that we
require for most present-tense sentences when they are used to make
deontic statements.
But how do we come to understand the meaning of the two-place
deontic operator ‘obligate’ and its connexion with mands? At this point,
it should be observed that declarative sentences like (3)-(6) may be
used, not only to issue mands and to assert the existence of an obliga-
tion, but also to report or transmit mands without commitment to the
existence of the obligation. If X has issued a command or demand of the
form Open the door! or Let Y open the door!, his illocutionary act may be
reported to Y by Z in various ways: X told you to open the door, X said
that you should open the door, etc. It may also be transmitted or reported
by means of the sentence You must open the door. In this case, Z’s
utterance can be understood to mean “‘X tells you to open the door”’.
Z is not necessarily committed to the existence of the obligation; but
he is committed by his selection of the non-past, rather than the past,
tense of ‘tell’ to the fact that the mand is still in force (in w; = wo).
X tells you to open the door differs in meaning from X told you to open
the door, even though they may both be used to report or transmit the
content of a mand that was issued in the past. Now, if Z implicitly
acknowledges the authority that X exercises over Y, he is logically com-
mitted by his utterance of either X tells you to open the door or You must
open the door (or You are to open the door) to a belief in the existence of
the particular obligation created by the mand that he is transmitting or
reporting. His utterance will be understood to imply, though it is not
exactly equivalent to, “You are under the obligation (imposed by X at
some time prior to ¢ and still in force) to open the door’’. In practice,
therefore, statements that transmit or report mands that are still in
force are not sharply distinguishable from statements which assert the
existence of the obligations created by these mands, even though there
is an obvious difference in the truth-conditions of the propositions they
express.
The fact that in many languages, and not only in English, the same
sentences may be used either to report mands or to assert that an obliga-
tion exists is of considerable importance. It lends support to the view,
put forward above, that our understanding of deontic statements is
based upon an ontogenetically prior understanding of the illocutionary
 17.5. Obligation, permission, prohibition and exemption 835
force of mands. As we saw earlier, one of the ways in which we can
naturally challenge or query the validity of a deontic statement is by
saying Who says (so)? Now this question is not generally interpreted to
mean “‘Who asserts that the obligation exists?’’, but rather ‘‘Who has
created the obligation?’’. This is readily accounted for if we assume that
the person to whom the deontic statement is addressed takes it to mean
‘‘T say so —it is so — (that) X says/has said so be it that p”. Who says
(so)? 1s of course ambiguous in relation to the utterance to which it is a
response. But the identity of the person saying “it is so” is not nor-
mally in doubt; the person who has created the obligation by saying
“so be it”? to the innermost proposition is the person whose identity is
being questioned.
It is worth noting also that commands and demands may be reported
or transmitted by means of imperative sentences with an explicit
acknowledgement of their source: cf.
(8) Come in and have your bath: Mummy says so.
In the situations in which (8) would normally be uttered, it would be
more or less equivalent to
(9) You've got to come in and have your bath: Mummy says so
or
(10) Mummy says you've got to come in and have your bath.
And in all three cases the person making the utterance would normally
be understood to be transmitting a mand from someone whose authority
to create an obligation he himself acknowledges. Situations of this kind
we can think of as being models for the child’s understanding of the
meaning of deontic statements.
What we are suggesting, then, is that creating an obligation should be
understood, ontogenetically at least, in terms of authoritative acts of
saying “‘so be it’’; and that deontic statements originate, ontogenetic-
ally, in utterances which report or transmit the content of such acts.
This does not mean that we cannot come later to an appreciation of the
distinction between indirect commands (in the traditional sense of this
term) and deontic statements. The distinction may not be clear cut in
many everyday situations, but it is one that most philosophers and
logicians will wish to draw. For anyone who is committed to the
possibility of there being obligatory norms of conduct that have not
been established by human or divine fiat must be able to refer to and
836 Modality
discuss such norms without, explicitly or implicitly, basing them upon
an underlying illocutionary act of commanding. Whether all human
languages provide the means for talking about the existence of moral
and social norms in abstraction from their source is an interesting
philosophical and linguistic question, which we will not go into here. It
suffices for our present purpose that we should be able to account for the
underlying logical structure of abstract deontic statements in those
languages in which such statements can be made.
The two-place deontic predicate ‘obligate’ can be related to the
one-place deontic predicate ‘obligatory’ by means of a very general
causative and inchoative relationship which is lexicalized or gram-
maticalized in various ways in different languages. This relationship is
lexicalized, for example, in the English words ‘kill’ and ‘die’. As “X
kill Y at t;”’ entails ““Y be dead from t;”’, so “X obligate Y at t; to do a
at t; (>t)” entails ““Y be obligated from 4; to do a at t > tj” or,
equivalently, “It be obligatory from t; that Y do a at t,’’. Now, just as
we can refer to the state-of-affairs described by “Y is dead’’ without
implying that this state-of-affairs has been caused by any external
agency, so we can refer, in certain languages at least, to the state-of-
affairs described by “It is obligatory that Y do a (at t,)’’ without neces-
sarily committing ourselves to the prior creation of this obligation by
some X. We can interpret ‘that Y do a (at t,)’ as an expression which
refers, not to a proposition, but to a third-order entity of a somewhat
different category: an obligation.
If this way of looking at deontic statements is correct, the predicate
‘obligatory’ adds nothing to the meaning of the existential proposition
in which it occurs. Everything having to do with the notion of obligation
is captured in the structure of the referring expression. The existential
proposition “It is obligatory that Y do a (at ¢,)”’ will be analysable in
terms of the logical structure suggested by the gloss “‘that Y do a (at tx)
exists (in some world-state)”’. To assert that an obligation holds is to
perform the same kind of act as we perform when we assert that a
proposition is true. In the one case we say “‘it is so”’ to the proposition
‘the obligation / p exists’; in the other to the proposition “the propo-
sition p exists’. Similarly, to assert that a prohibition holds is to say
‘“ ~! p exists’? (where ‘~ / p’ is a referring expression).
Permission, as we saw above, is related to obligation in standard
systems of deontic logic in the same way that possibility is related to
necessity. But a distinction can be drawn, intuitively at least, between a
passive, or weaker, sense of ‘permission’ and an active, or stronger,
 17.5. Obligation, permission, prohibition and exemption 837
sense. A course of action is permissible in the weak sense if, and only if,
it is not explicitly prohibited. Under this interpretation of permission,
every possible course of action is either permissible or prohibited, and
permission is interdefinable with obligation. In practice, however, we
do not always operate with the principle that the non-existence of a
- prohibition implies the existence of a permission. There are certain
codes of conduct or sets of regulations that are interpreted in terms of a
stronger sense of both ‘permission’ and ‘prohibition’, according to
which any course of action that is not explicitly permitted or prohibited
by the regulations is deemed to be deontically indeterminate and subject
to further legislation. This question has been fully discussed by writers
on deontic logic (cf. Von Wright, 1968: 85ff); and we need not go into
it here. Whether we are, as human beings or as members of a particular
society, implicitly permitted to do whatever we are not expressly pro-
hibited from doing is hardly a question for the semanticist. We will
simply note that the difference between the stronger and the weaker
sense of ‘permission’ can be captured in the logical structure of deontic
statements by invoking the distinction between nexus-negation and
predicate-negation (if this distinction is valid: cf. 16.4). “It is not-
prohibited to do a”’ is the contrary of ‘It is prohibited to do a”’, and it
is equivalent to ‘‘It is permissible to do a’’. On the other hand, “‘It is
not (the case that it is) prohibited to do a’’ is the contradictory of “‘It is
prohibited to do a’’; though it is not logically equivalent to “It is
permissible to do a’’, it will often be taken to imply this proposition.
As positive commands and demands are related to prohibitions, so
permissions are related to what we will call exemptions*. When we issue
a permission, by means of a permission-granting utterance, e.g., You
may open the door, we either cancel a pre-existing prohibition or deter-
mine the deontic status of some action whose deontic status was
previously undetermined. When we issue an exemption-granting utter-
ance, for example, You needn’t open the door, we cancel a pre-existing
obligation or determine, by fiat, a deontically undetermined action.
Given that mands, for example, Open the door! and Don’t open the door!,
have the structure . / p or . ~/ p (with the appropriate tense-operators)
and that they bring into existence in some world, w;, obligations to do
or not to do something we must now face the problem of providing an
analysis for permission-granting and exemption-granting utterances.
The first point that must be made in this connexion is that, although
imperative sentences are characteristically used to issue mands, they
may also be employed, in certain situations, to grant permission. For
 838 Modalhty
example, when we say Come in! in response to a knock on the door, we
are not normally understood to be issuing a command (or even a re-
quest), but to be granting to the person who has knocked permission to
enter the room; and his knock is, by convention, taken to be equivalent
to a request for our permission: i.e. as meaning “May I come in?”’. |
Similarly, the instruction Go! or Cross now! associated with a green
light (unlike Stop! or Don’t cross the road now! associated with a red
light) does not impose upon the motorist or pedestrian the obligation to
behave or not to behave in a certain way: it creates or maintains a
permissible course of action by removing a prohibition or determining
as deontically possible what was deontically undetermined. There is of
course a relationship of inclusion or implication between mands and
permissions, based upon the principle that, as necessity implies possi-
bility, so obligation implies permission. When we impose an obligation,
we implicitly cancel, or commit ourselves to the non-existence of, any
conflicting prohibitions, and we thereby commit ourselves not to pre-
vent the person upon whom we impose the obligation from complying
with it: implicitly, we permit him (in the stronger sense of ‘ permission’)
to do whatever he is obliged to do. Furthermore, when we issue a mand,
we commit ourselves (with our ‘‘so be it’’) to the desirability of the
proposed course of action. Permission-granting utterances, however, do
| not create any obligations; nor do they commit the speaker to the
desirability of what they permit. They are neutral too with respect to
the speaker’s assumptions about the addressee’s wishes or intentions. In
practice, however, we do not normally grant people permission to do
something which we have no reason to believe that they wish to do, or
might wish to do.
The fact that imperative sentences may be used to grant permission
does not imply that there is not some intrinsic connexion between
imperative sentences and mands. Since we do not normally command or
request people to do what we expect they will do without being told or
asked, an utterance which is characteristically used to issue a mand can
be interpreted as a permission-granting utterance in situations in which
it is clear that the addressee wishes to carry out some course of action;
and it would seem to be the case that it is only in such situations that
imperative sentences (with or without the addition of a conditional
clause, like zf you wish or tf you will) are used to make permission-
granting utterances. Similarly, what is characteristically a permission-
granting utterance, or a statement to the effect that something is per-
mitted, may be used and understood as a mand, if the situation is such
 17.5. Obligation, permission, prohibition and exemption 839
that the addressee may not be assumed to wish or intend to carry out the
proposed course of action. This being so, we can account for the fact
that Open the door! and You may open the door are characteristically
distinct, as far as their basic illocutionary force is concerned, though
each may be used in particular circumstances to issue a mand or to grant
permission.
Having established this point, we can now concentrate upon such
sentences as You may open the door and You needn’t open the door (or
You don’t have to open the door). English is like many languages, and
possibly all languages, in that it does not have a distinct permissive
mood. In view of the interdefinability of obligation and permission, it
would be easy enough to construct a permission-based, rather than an
obligation-based, system of deontic modality. Suppose, for example,
that what is traditionally described as the imperative mood was charac-
teristically associated not with mands, but with permissions and
exemptions, so that Open the door! was roughly equivalent, in its most
characteristic use, to the explicitly performative utterance I (hereby)
permit you to open the door and Don’t open the door to I (hereby) permit
you not to open the door. Suppose, further, that the sentences containing
modal verbs
(11) You must open the door
(12) You may open the door
(13) You must not open the door
(14) You need not open the door
still had the same meaning that they have in present-day English and
that there was no way of issuing mands other than by uttering sentences
like (11) and (13). If this were the case, there would be good grounds for
analysing obligation in terms of a more primitive notion of permission,
as we analysed epistemic necessity in terms of epistemic possibility in an
earlier section of this chapter (17.2). As things stand, however, the reverse
is the case; and there can be little doubt that it is obligation that is dis-
tinctively grammaticalized in the structure of English. Apart from sen-
tences containing explicitly performative verbs of permission and
exemption, any sentence that may be used to grant a permission or an
exemption may also be used to assert the existence of a permission or
exemption: e.g., (12) and (14). But, in addition to sentences with modal
verbs or modal adjectives of obligation and prohibition, which may be
used either to issue mands or to make existential statements, we also
840 Modality
have simple imperative sentences, which are used primarily to issue
mands and only secondarily, in transmitting or reporting the content of
mands, to make statements to the effect that /p or ~/ 9 exists.
That deontic modality, unlike epistemic modality, should be
necessity-based rather than possibility-based is not surprising, if it
derives ontogenetically, as we have suggested, from the desiderative and
instrumental function of language. For there is a so-be-it component in
the tropic of permission-granting and exemption-granting utterances;
and it is this component that is commonly grammaticalized in the im-
perative and the so-called subjunctive of will.
If permission is analysed in terms of the absence of prohibition and
exemption in terms of the absence of obligation, we can provide what
seems to be a plausible account of the meaning of (11)-(14). Let us
consider first of all situations-of-utterance in which they are to be
construed as existential statements. In such situations, as we saw earlier,
(11) will mean “I say so — it is so—that / p exists” and (13) will mean
“TI say so—it is so—that ~/p exists’. Given that permission is
derived from obligation, (12) and (14) will mean “I say so — it is so —
that ~/ does not exist” and “I say so — it is so—that /p does not
exist’’, respectively. If this analysis of permission-granting and exemp-
tion-granting utterances is acceptable, their analysis is straightforward.
We simply substitute “‘so be it’’ for “it is so” in the tropic position.
You may open the door, interpreted as “I (hereby) permit you to open
the door’ will be analysed as “I say so—so be it—that ~/p not
exist’? and You needn’t open the door as “I say so—so be it — that / p
not exist’’. Provided that the person making these utterances has the
authority to grant the permission or exemption, his unqualified “‘I say
so”’ to “‘so be it”’ will of itself create a world in which the prohibition or
exemption does not exist.
At this point, the reader will naturally wonder whether (11) and (13)
should not be analysed, as mands, in the same way: i.e. not as “I say
so—so be it—that p” and “I say so-—let it not be —that p’’, but as
“TI say so —so be it — that /p exist” and “I say so — so be it—that ~/ p
exist’’. There is every reason to take this view, since it gives explicit
recognition to the fact that (11) and (13) are structurally parallel to (12)
and (14). Indeed, by analysing all four sentences in this way we can
avoid saying that any of them is ambiguous. The difference between “I
permit you to do a”’ or “I oblige you to do a”’ and “‘ You are permitted
(by X) to do a”’ and “You are obliged (by X) to do a”’ cannot be drawn
within sentences containing the modal. verbs in English, except by
 17.6. A tentative synthesis SAI
adding parenthetical clauses like J say so or X says so. One might argue
therefore that the alleged ambiguity, in utterance, between the use of
such sentences to make existential statements and their use to create
obligations, prohibitions, permissions and exemptions is not a lin-
guistically based ambiguity at all. On this view of the matter, (11)—(14)
would always have the primary illocutionary force of statements; and the
propositions they express would be true or false according to the exis-
tential status of the obligation or prohibition at the time of utterance.
The speaker’s creation of the obligation or permission, if he is in fact
its source, would be logically separable from his assertion of its exis-
tence.
We will not discuss this question further. Whether (11)-(14) are
ambiguous as utterances, as our analysis in terms of the distinction
between ‘‘it is so”’ and ‘‘so be it’’ assumes, may be open to doubt. For
any declarative sentence whatsoever may be used in the same way,
under the appropriate non-linguistic conventions, to bring about a
change in the intensional world of facts and beliefs. ‘There are many
magical and sacramental uses of language in which what appears to be
the assertion that something is so 1s held, by believers, to have the effect
of making it so. In all such cases we are faced with the same alternatives.
We can either say that a statement is being used to report a state-of-
affairs which has been brought about independently of the utterance
(though the utterance may be an integral and essential part of the ritual
itself) or we can say that the utterance 1s not in fact a descriptive state-
ment, but something which has an instrumental function, symbolized
in our analysis by the so-be-it component. The problem 1s therefore a
very general one; and it is unclear that it admits of any non-arbitrary
solution.

17.6. A tentative synthests

It now remains for us to bring together the threads of the somewhat
lengthy, and at times tortuous, argument that has been presented 1n this :
chapter and, in doing so, to relate deontic modality to subjective and
epistemic modality, on the one hand, and to the grammatical category of
mood, on the other. It is a widely held view among linguists that there 1s
a fundamental difference between the epistemic and the deontic uses of
‘may’ (or ‘can’) and ‘must’ in English; and this difference has been
accounted for in some recent transformational analyses by classifying
the epistemic modal verbs as intransitive and the deontic modals as
transitive. The so-called intransitive modals, under this analysis of their
 842 Modality
meaning and function, would have a nominalized sentential subject; and
the transitive modals would have as their subject an expression referring
to the person who is the source of the obligation or permission and as
their object a nominalized sentence referring to what is obligatory or
permissible. Roughly speaking, under this analysis, John may come in
would mean ‘‘ That John will come in is possible”’ with the intransitive
‘may’ and, under the permission-granting interpretation, ‘I make-it-
possible that John will come in”’ with the transitive ‘may’.
There are several points that may be made in relation to this analysis.
First of all, like most treatments of epistemic modality it fails to account
for the difference between a subjective and an objective interpretation of
epistemic modality. By comparing an objectively modalized epistemic
interpretation with a permission-granting deontic interpretation it makes
the transitive—intransitive distinction look semantically more plausible
than it is. As soon as we gloss the allegedly intransitive sentence as ‘I
think-it-possible that John will come in’”’ we see that there can be, in
principle, a transitive analysis of epistemic modality. The transitive
analysis of John may come in, interpreted as a deontic statement, is on
the other hand highly implausible: “‘John makes-it-possible that John
will come in”’ is semantically inappropriate, and such analyses as “ For
John it is possible that John will come in’”’, even if they are interpretable
in the appropriate way, depend upon some rather questionable syntactic
processes and an eccentric notion of what it means for an expression to
be the subject of a transitive verb. It would seem, therefore, that the
transitive—intransitive distinction does not adequately capture and
formalize the difference between epistemic and deontic modality. There
is, however, as we shall see presently, a different way of applying the
distinction, so that both epistemically and deontically modalized
sentences are derivable, under certain assumptions about a deeper
level of grammatical or semantic analysis, from underlying transitive
structures.
A second general criticism that can be made of the transformational
analysis of epistemic and deontic sentences outlined above 1s that it says
nothing about the difference in the way in which we have to interpret the
nominalized sentences in the two different syntactic environments. But
if they are referring expressions — and it is natural to interpret all
nominalized sentential structures functioning as subjects or objects in
larger sentences as referring expressions — we must ask ourselves what
kind of entities they refer to. The expression ‘that John will come in’ in
It is possible that Fohn will come in is readily interpretable as referring
 17.6. A tentative synthesis 843
either to a proposition (a third-order entity) or to the event or action
that would be described by the proposition expressed by the sentence
John will come in (a second-order entity). We need not decide between
these alternatives in the present connexion. The point is that neither of
them seems appropriate for the allegedly embedded sentential object of
the modal verb ‘may’ under the deontic interpretation of Fohn may come
in. Deontic statements do not assert the potential existence of proposi-
tions in a set of intensional worlds (or state-descriptions). Nor do they
assert the potential occurrence of events, acts or states-of-affairs in some
past, present or future state of the actual world. What they assert is the
actual existence of permissions and obligations in some particular ex-
tensional world or, alternatively, of propositions which describe the
content of these permissions and obligations. In other words, the notion
of potentiality (or necessity) in a deontically interpreted sentence is to be
associated, not with the existence of the referent of the nominalized
sentential complement of the modal verb, but rather is to be incorporated
within the complement itself. The most obvious way of achieving this
effect within a transformational framework is by embedding as the
complements of the deontically interpretable modal verbs and adjec-
tives modalized subject—predicate structures identical with those which
underlie sentences characteristically used to issue mands, to grant per-
missions and exemptions, and to express wishes, exhortations, etc. If we
do this, we simultaneously account for the tense restrictions that hold
_ within deontic sentences. For they are the same restrictions as hold
within simple non-indicative sentences. John may come yesterday con-
strued as a permission-granting utterance is semantically anomalous for
the same reason that Come yesterday, John! is anomalous.
One of the points that was made at the beginning of this section was
that deontic necessity and possibility are usually understood to originate
in some causal source: 1.e. if someone is obliged or permitted to carry
out some course of action, it is generally, though not necessarily,
assumed that some person or institution has created the obligation or
permission (cf. 17.4). Let us now refer to this person or institution as the
deontic source*. In our discussion of alethic and epistemic modality in
an earlier section we made no mention of there being any comparable
source for logical and epistemic necessity or possibility. It is obvious,
however, that just as we can account for our understanding of the
one-place predicate ‘obligatory’ in terms of our prior understanding of
the two-place predicate ‘obligate’, so we can account for our under-
standing of the one-place predicates ‘necessary’ and ‘possible’ in their
844 Modality
objective interpretation, in terms of the two-place predicates ‘make-
necessary’ and ‘make-possible’. Fohn may open the door, meaning “It is
(objectively) possible that John will open the door” could, in principle,
be derived from an underlying transitive structure of the kind that
underlies the sentence X makes/has made it possible that Fohn will open
the door. It is not being suggested that epistemically and logically
modalized sentences should, in fact, be derived in this way in a grammar
of English. The point is simply that an underlying transitive structure
is no less appropriate for epistemically modalized sentences than it is
for deontically modalized sentences. The difference between them is
readily accounted for in terms of the complement of ‘make-possible’,
and it is arguable that our understanding of what is meant by a state-of-
affairs being possible is based intuitively upon our understanding of
what is involved in making a state-of-affairs possible.
If we look at the matter in this way, we can see why the same modal
verbs and adjectives are used, in many languages, in what appear to be,
and are often classified as, different senses: the ‘may’ of permission, the
‘may’ of alethic possibility etc. Making possible a state-of-affairs can be
conceptualized, dynamically, in terms of permitting it to exist; and
making it necessary, in terms of obliging it to exist. Theological and
mythical accounts of the structure of the world are often cast in such
terms; so too are certain philosophical analyses of alethic modality. This
fact of itself would suggest that it is natural for us to think of objective
epistemic modality and alethic modality (in so far as they can be dis-
tinguished: cf. 17.2) in terms of a conceptual model within which the
possibility or necessity of something being so is understood by analogy
with our understanding of the deontic notions of granting permission
and imposing obligations. We are not committed, of course, by the
structure of the languages we speak to the view that physical possibilities
and necessities are created by acts of will performed by agents external
or internal to the physical system within which these possibilities and
necessities operate. But the continued use of such expressions as ‘the
laws of nature’ or ‘the law of the excluded middle’ suggests that we find
the analogy of the external law-giver useful and natural, even though we
may not believe that there is, or was at any time, any actual legislator to
determine by fiat (‘‘so be it’’) the behaviour of physical entities or the
principles of valid reasoning. However that may be, it is certainly the
case that, not only in English, but in many languages, the same modal
predicates are used for the expression of alethic, epistemic and deontic
modality; and it is easier to account for this by assuming that the concept
 17.6. A tentative synthesis 845
of deontic necessity serves as the analogical model for alethic and
objective epistemic necessity than it is by making the contrary assump-
tion.
How can the assumption that objective epistemic necessity is model-
led upon deontic necessity be reconciled with the view that epistemic
necessity is to be defined in terms of epistemic possibility and that this
notion derives, in the child’s acquisition of his native language, from un-
certainty and doubt (cf. 17.2)? The answer may well be that the ob-
jectification of epistemic necessity and possibility is a rather sophistica-
ted and impersonal process which plays little part in ordinary non-
scientific discourse; and languages, it should always be remembered, are
semiotic systems that are used primarily for non-scientific discourse.
The more basic notions, it would seem, are subjective epistemic possi-
bility, on the one hand, and the issuing of mands, on the other. Not only
objectively modalized epistemic statements, but also existential deontic
statements, are to be regarded as secondary. The one involves the
objectification of the probability of such-and-such a situation obtaining,
occurring or coming into being; the other involves the objectification
and depersonalization of the content of obligation-creating mands. Once
they are objectified, there is no longer any primacy of possibility over
necessity or of necessity over possibility: either may be defined in terms
of the other. It is only in relation to what we have taken to be the more
basic structures
(1) poss. p
and
(2) ./p
that we have argued for the primacy of epistemic possibility, on the one
hand, and of deontic necessity, on the other.
As we have seen, there is a dynamic interpretation of objective
epistemic modality which construes it by analogy with the dynamic
interpretation of deontic modality: i.e. as being the result of authorita-
tive acts of saying “‘so be it’’ to p. This comes about by virtue of the
instrumental function of language. Provided that the person who says
“‘so be it”’ to p has the authority to do so and the power to make his
utterance efficacious, his saying of ‘‘so be it”’ to p (i.e. his saying . ! p)
will result, directly or indirectly, in the truth of p at t; > to. It follows that
epistemic and deontic necessity are often difficult to separate in the case
of utterances whose propositional content relates to some future world-
846 Modality
state. If the person predicting that p will be so at ty > fo is in a position
to bring it about that p is so at t; > to by saying “‘so be it”’ to p at fo, no
clear distinction can be drawn between predicting and promising. The
limiting case in which epistemic and deontic modality merge com-
pletely is in respect of utterances by an omnipotent and omniscient
being, which, whether we are theists or not, may be taken as a standard
to which utterances by non-omnipotent and non-omniscient beings can
be seen as approximating. From this point of view, so-called statements
of future requirement, like
(3) The successful candidate will be a woman in her mid-thirties of
demonstrated ability,
no matter how mundane they might be, are comparable with that
archetype of the creative illocutionary act in Genesis “‘Let there be
light’’, Utterances like (3) may be analysed as demands. They may also
be analysed as statements of either epistemic or deontic necessity; and,
in cases like this, the two kinds of modality are hardly distinguishable.
There is much that has not been dealt with in our treatment of
modality. Our main concern has been to show, in a general way, both
the similarities and the differences between epistemic and deontic
modality and, furthermore, to make a case for the view that statements
of objective modality, of the kind that logicians tend to be concerned
with, are plausibly regarded as being of secondary development. Nothing
has been said about physical necessity and possibility, which, though it is
not normally judged to fall within the province of modality, is obviously
relatable to, and at times may be indistinguishable from, objective
deontic modality. Nothing has been said, either, about such differences,
within deontic modality, as are manifest, in English, in such utterances
as
(4) He ought to go
(5) He should go
(6) He must go.
The fact that but he won’t may be appropriately appended to (4) and (5),
but not to (6), shows that there can be other kinds of interaction between
epistemic and deontic modality." There are also various interactions
between modality and tense (or aspect) that we have not dealt with.
Limitations of space have prevented this.
11 On these and other related points, cf. Boyd & Thorne (1969), Leech (1971),
Palmer (1974).
 17.6. A tentative synthests 847
We have been operating with the following two basic patterns for
utterances (or utterance-units: cf. 14.6): poss. p and . / p (cf. (1) and
(2)). What is symbolized here by means of the modal operator poss in
the neustic position may be realized in the utterance-signal in various
ways: by prosodic and paralinguistic modulation, by the use of a
particular grammatical mood, by the use of one of a set of modal verbs
or adjectives, by the use of a parenthetical word-form like perhaps or a
parenthetical clause like J think in English. The speaker may subjectively
qualify his commitment to the truth-value of a proposition that he 1s
more or less confidently putting forward in any of these functionally
equivalent ways.
If the language-system in question provides a grammatical mood
whose sole or basic function is that of expressing subjective epistemic
possibility, this mood would be appropriately described as the potential
mood, It might be in opposition with, not only an indicative and an
imperative, but also a dubitative, a conditional, a presumptive, a con-
cessive or an inferential mood. All of the notions that are implied by the
terms ‘potential’, ‘dubitative’, ‘conditional’, ‘presumptive’, ‘con-
cessive’, ‘inferential’, etc., though distinguishable, are obviously con-
nected; and it is doubtful whether any language distinguishes more than
two or three of them in the category of mood.” In this connexion, it is
important to emphasize that, at the present stage of linguistic theory and
descriptive practice, it is impossible to formulate any very clear notion
of the distinctions that are grammaticalized, within the category of mood,
throughout the languages of the world. The labels that are used in
standard descriptions of particular languages are often misleading in
that they imply that the functions of the moods are narrower or more
specific than they really are. This is true, for example, of the term
‘conditional’ as it is used with respect to French, or the ‘inferential’
as it is applied to Turkish. In general, we cannot be sure that, because
the same term is used in relation to two different languages, the moods
that the term refers to have exactly the same function in the two lan-
guages. Nor can we be sure that, because two different terms are used,
two different functions are involved.
Both poss in (1) and / (‘‘so be it”’) in (2) represent different facets of
the expressive, or indexical, function of language (cf. 2.4, 3.4). In the
one case it is an indication of the speaker’s opinion or judgement that 1s
involved; in the other it is an indication of the speaker’s will or desire
12 For a comprehensive list of terms that linguists have used in their analysis of
mood in different languages, cf. Jespersen (1929).
848 Modality
that something should be done. As we saw in an earlier section, these
two kinds of speaker-involvement are not necessarily distinguished in
the grammatical systems of particular languages. For example, what is
traditionally referred to as the subjunctive mood in the older Indo-
European languages grammaticalized the notion of non-factivity (coupled
with the deictically based notion of non-remoteness): its function could
be, not only potential or predictive (in which latter function it was
interchangeable in early Latin with the future tense), but also obligative,
hortative or desiderative. (The term ‘subjunctive’ derives from the view
that the verb-forms in question were used, characteristically if not ex-
clusively, in subordinate clauses. It says nothing about their use in
independent clauses.) It is possible then for both poss in the neustic
position and / in the tropic position to be realized by one and the same
grammatical mood, just as it is possible for finer distinctions to be
grammaticalized within either epistemic or deontic modality.
According to the view of mood and modality that has been taken in
this chapter, mood is a grammatical category that 1s to be found in some,
but not all, languages. It cannot be identified with either modality or
illocutionary force as such, any more than either one of these can be
identified with the neustic or the tropic utterances to the exclusion of
the other. A rather different view is taken by such scholars as House-
holder (1971: 81ff) and Halliday (1970 a,b), who, drawing a distinction
between mood and modality, relate the former to illocutionary force and
the communicative role that the speaker is performing and relate the
latter, as we have done, to the expression of necessity and possibility.
The difference between these two points of view is largely terminologi-
cal. In traditional usage, ‘mood’ is applied to such subsets of inflected
forms of verbs as are distinguished one from another by means of the
terms ‘indicative’, ‘imperative’, ‘subjunctive’, etc.; and we have chosen
to respect this usage. One advantage of doing so is that it enables us to
draw a distinction, not only between utterances and sentences (between
statements and declarative sentences, between questions and interroga-
tive sentences, etc.), but also between sentences that are subclassified as
declarative, interrogative, jussive, etc., in terms of their characteristic
use and sentences that are subclassified as indicative, dubitative, im-
perative, etc., in terms of the mood of the main verb (cf. 16.2). Up to a
point, this too is in conformity with traditional usage, according to which
‘imperative’ contrasts with ‘indicative’ and ‘subjunctive’, quite
differently from the way in which ‘declarative’ contrasts with ‘in-
terrogative’.
 17.6. A tentative synthesis 849
Setting aside such purely terminological questions, we can emphasize
the more general point that has been made in this and the preceding
chapter. This is that, although there are methodological advantages in
restricting one’s attention to what we earlier referred to as micro-
linguistic semantics — the semantic analysis of maximally decontextuali-
zed system-sentences (cf. 14.2, 14.6), there are limits to this process of
depragmatization, as far as the representation of modality is concerned
(cf. Bar-Hillel, 1970: 219). Languages are learned and used in contexts
which are in part determined by the variable assumptions and pre-
suppositions of the people who use them; and these assumptions and
presuppositions are not necessarily representable in terms of a set of
determinable propositions (cf. 14.3). The objectification of both episte-
mic and deontic modality is something that we have here taken to be
secondary in the acquisition of language; and it may very well be that
not all languages, but only those that have been long used in literate
societies for the specialized purpose of academic discussion, provide the
means for this kind of objectivization. However that may be, it has been
argued here that modality, as it operates in a good deal of everyday
language-behaviour, cannot be understood, or properly analysed, other-
wise than in terms of the indexical and instrumental functions of
language, to which its descriptive function is, at times if not always,
subordinate (cf. 2.4).
 Bibliography

Abraham, W. (ed.) (1971). Kasustheorte. Frankfurt: Athenaium.

Adams, V. (1973). An Introduction to English Word-Formation. London:
Longman.
Agricola, E. (1968). Syntaktische Mehrdeutigkeit (Polysyntaktizitat) bet der
Analyse des Deutschen und des Englischen. Berlin: Akademie Verlag. __.
Al, B. P. F. (1975). La Notion de Grammaticalité en Grammaire
Générative-Transformationelle. Leyde: Presse Universitaire de Leyde.
Allan, K. (1976). ‘Collectivizing’. Archivum Lingutsticum 7. 99-117.
Allan, K. (1977). ‘Classifiers’. Language 53. 285-311.
Allen, W. S. (1964). ‘Transitivity and possession’. Language 40. 337-53.
Allerton, D. J. (1969). ‘The sentence as a linguistic unit’. Lingua 22. 27-46.
Allerton, D. J. (1975). ‘Deletion and proform reduction’. Journal of
Lingutstics 11. 213-37.
Allerton, D. J. & Cruttenden, A. (1974). ‘English sentence adverbials:
their syntax and their intonation in British English’. Lingua 27. 1-29.
Altham, J. E. J. & Tennant, N. W. (1975). ‘Sortal quantification’. In Keenan
(1975: 46-58).
Anderson, J. M. (1971). The Grammar of Case: Towards a Localistic Theory.
London: Cambridge University Press.
Anderson, J. M. (1973a). ‘Maximi Planudis in memoriam’. In Kiefer &
Ruwet (1973: 20-47).
Anderson, J. M. (1973b). An Essay Concerning Aspect. The Hague: Mouton.
Anderson, J. M. (1975). ‘La grammaire casuelle’. In Anderson &
Dubois-Charlier (1975: 18-64).
Anderson, J. M. & Dubois-Charlier, F. (eds.) (1975). La Grammaire des Cas.
(Languages 38.) Paris: Didier-Larousse.
Antinucci, F. (1974). ‘Sulla deissi’. Lingua e Stile 11. 223-47.
Apresjan, Ju. D. (1973). Principles and Methods of Contemporary Structural
Linguistics (translated from the Russian by D. B. Crockett). The Hague:
Mouton.
Apresjan, Ju. D. (1974). Leksiteskaja Semantika: Synonimiteskie Sredstva
Fazyka. Moskva: ‘Nauka’.
Aaqvist, Lennart (1965). A New Approach to the Logic of Questions.
(Philosophical Studies, University of Uppsala.)

Tavistock Publications. |
Ardener, E. (ed.) (1971). Social Anthropology and Language. London:
 Bibhography 851
Argyle, M. (1972). ‘Non-verbal communication in human social interaction’.
In Hinde (1972: 243-69).
Asher, R. E. (1968). ‘Existential, possessive, locative and copulative sentences
in Malayalam’. In Part 2 (1968) of Verhaar (1967-73). Pp. 88—111.
Ashton, E. O. (1947). Swahili Grammar, 2nd ed. London: Longman.
Atkinson, M. A. & Griffiths, P. G. (1973). ‘Here’s here’s, there’s here and
there’. In Edinburgh Working Papers in Linguistics, 3. 29-73.
Austin, J. L. (1962). How To Do Things With Words. Oxford: Clarendon Press.
Austin, J. L. (1970). Philosophical Papers, 2nd ed. London & New York:
Oxford University Press. (First edition, 1961.)
Ayer, A. J. (1936). Language, Truth and Logic. London: Gollancz. (Second
edition, revised, 1946.)
Babcock, 5S. (1972). “Periphrastic causatives’. Foundation of Language 8. 30-43.
Bach, E. (1968). ‘Nouns and noun-phrases’. In Bach & Harms (1968: 91-122.)
Bach, E. (1970). ‘Problominalization’. Linguistic Inquiry. 1. 121-2.
Bach, E. (1971). ‘Questions’. Linguistic Inquiry. 2. 153-66.
Bach, E. & Harms, R. (eds.) (1968). Universals in Linguistic Theory. New
York: Holt, Rinehart & Winston.
Bacon, John (1973). ‘The semantics of generic THE’. Journal of
Philosophical Logic 2. 323-39.
Baker, C. L. (1970). ‘Notes on the description of English questions’.
Foundations of Language 6. 197-219.
Baker, C. L. (1975). ‘The role of the part-of-speech distinctions in
generative grammar’. Theoretical Linguistics 2. 113-31.
Banfield, A. (1973). ‘Narrative style and the grammar of direct and indirect
discourse’. Foundations of Language 10. 1-39.
Bar-Hillel, Y. (1964). Language and Information. Reading, Mass. :
Addison-Wesley.
Bar-Hillel, Y. (1967a). ‘Dictionaries and meaning rules’, Foundations of
Language 3. 409-14. (Reprinted in Bar-Hillel, 1970: 347-53.)
Bar-Hillel, Y. (1967b). Review of Fodor & Katz (1964). In Language 43.
526-50. (Reprinted in Bar-Hillel, 1970: 243-69.)
Bar-Hillel, Y. (1970). Aspects of Language. Jerusalem: Magnes.
Bar-Hillel, Y. (ed.) (1971). Pragmatics of Natural Language.
Dordrecht-Holland: Reidel.
Barthes, R. (1970). S/Z. Paris: Seuil.
Bartsch, R. (1972). Adverbialsemanttk. (Linguistische Forschungen, 6.)
Frankfurt: Athenium.
Bartsch, R. & Vennemann, T. (1972). Semantic Structures. Frankfurt:
Athendum.
Bates, E. (1976). Language and Context: the Acquisition of Pragmatics. New
York: Academic Press.
Bauman, R. & Sherzer, J. (eds.) (1974). Explorations in the Ethnography of
Speaking. London & New York: Cambridge University Press.
Bazell, C. E. (1964). ‘Three misconceptions of ‘‘grammaticalness’’’.
Monograph Series on Language and Linguistics 17. 3-9. (Reprinted in
O’Brien, 1968: 25-31.)
 852 Bibliography
Bazell, C. E. et al. (eds.) (1966). In Memory of 7. R. Firth. London:
Longman.
Beekman, J. & Callow, J. (1974). Translating the Word of God. Grand
Rapids, Michigan: Zondervan.
Bellert, I. (1970). ‘On the semantic interpretation of subject-predicate
relations in sentences of particular reference’. In Bierwisch & Heidolph
(1970: 9-26).
Bellert, I. (1972). On the Logico-Semantic Structure of Utterances. Wroclaw:
Ossolineum.
Bene3, E. & Vachek, J. (eds.) (1971). Stilisttk und Soziolinguisttk. Miinchen: List.
Bennett, D. C. (1972). ‘Some observations concerning the locative directional
distinction’, Semiotica 5. 109-27.
Bennett, D. C. (1975). Spatial and Temporal Uses of English Prepositions.
London: Longman.
Bennett, J. (1973). ‘The meaning-nominalist strategy’. Foundations of
Language 10. 141-68.
Bennett, J. (1976). Linguistic Behaviour. Cambridge: Cambridge University
Press
Benveniste, E. (1946). ‘Structure des relations de personne dans le verbe’.
Bulletin de la Société de Linguistique 126. 1-12. (Reprinted in Benveniste,
1966: 225-36.)
Benveniste, E. (1956). ‘La nature des pronoms’. In M. Halle et al. (eds.)
For Roman Jakobson. The Hague: Mouton. (Reprinted in Benveniste,
1966: 251-7.)
Benveniste, E. (1958a). ‘De la subjectivité dans le langage’ (from Journal de
Psychologie). In Benveniste (1966: 258-66). |
Benveniste, E. (1958b). ‘Les verbes délocutifs’ (from A. G. Hatcher & K. L.
Selig (eds.) Mélanges Spitzer. Bern: Francke). (In Benveniste, 1966: 277-85.)
Benveniste, E. (1966). Problémes de Linguistique Générale. Paris: Gallimard.
(English translation, Problems in General Lingutstics. Coral Gables, Fla.:
University of Miami Press, 1971.)
Benveniste, E. (1967). Le Vocabulaire des Institutions Indo-Européennes.
Paris: Minuit. (English translation, Indo-European Language and Society.
London: Faber & Faber, 1973.)
Berlin, B. (1968). Tzeltal Numeral Classifiers: A Study in Ethnographic
Semantics. The Hague: Mouton.
Berlin, B., Breedlove, D. E. & Raven, P. M. (1966). ‘ Folk taxonomies and
biological classification’. Science 154. 273-5. (Reprinted in Tyler, 1969.)
Berlin, B., Breedlove, D. E. & Raven, P. H. (1974). Principles of Tzeltal
Plant Classification. New York & London: Academic Press. }
Bierwisch, M. (1967). ‘Some semantic universals of German adjectivals’.
Foundations of Language 3. 1-36.
Bierwisch, M. (1970). ‘Semantics’. In Lyons (1970: 166-84).
Bierwisch, M. & Heidolph, K. E. (1970). Progress in Linguistics. The Hague:
. Mouton.
Biggs, C. (1975). ‘Quantifiers, definite descriptions and reference’. In
Keenan (1975: 112-20).
 Bibliography 853
Bloomfield, L. (1931). Review of Ries (1931). In Language 7. 204-9.
Bloomfield, L. (1935). Language. London: Allen & Unwin. (American
edition, 1933.)
Boadi, L. A. (1971). ‘Existential sentences in Akan’. Foundations of Language
7. 19-29.
Boas, F. (1911). Introduction to the Handbook of American Indian Languages.
Washington D.C. (Reprinted, Washington, D.C.: Georgetown University
Press, 1965. Also, Lincoln, Neb.: University of Nebraska Press, 1965.)
Bolinger, D. L. (1952). ‘Linear modification’. Publications of the Modern
Language Association of America 67. 1117-44. (Excerpts reprinted in
Householder, 1972: 31-50.)
Bolinger, D. L. (1965). ‘The atomization of meaning’. Language 41. 555-73.
Bolinger, D. L. (1967a). ‘The imperative in English’. In To Honor Roman
Jakobson. The Hague: Mouton.
Bolinger, D. L. (1967b). ‘Adjectives in English’. Lingua 18. 1-34.
Bolinger, D. L. (1968). ‘Judgements of grammaticality’. Lingua 21. 34-40.
Bolinger, D. L. (1971). The Phrasal Verb in English. Cambridge, Mass. :
Harvard University Press.
Bonomi, A. & Usberti, B. (1971). Sintasst e Semantica nella Grammatica
Trasformazionale. Milano: Saggiatore.
Botha, R. P. (1968). The Function of the Lexicon in Transformational
Grammar. The Hague: Mouton.
Botha, R. P. (1973). The Justification of Linguistic Hypotheses. The Hague:
Mouton.
Bowers, J. S. (1975). ‘Adjectives and adverbs in English’. Foundations of
Language 13. 529-62.
Boyd, J. & Thorne, J. P. (1969). ‘The semantics of modal verbs’. Fournal of
Linguistics §. 57-74.
Braine, M. (1963). ‘The ontogeny of English phrase structure’. Language 39.
I-14.
Bréal, M. (1897). Essai de Sémantique. Paris. (English translation, Semantics.
London: Holt, 1900. Republished, New York: Dover, 1964.)
Brekle, H. E. (1970). Generative Satzsemantik und Transformationelle Syntax
im System der Englischen Nominalkomposition. Miinchen: Fink.
Bresnan, J. (1970). ‘On complementizers: towards a syntactic theory of
complement types’. Foundations of Language 6. 297-321.
Bright, W. (ed.) (1966). Sociolinguistics. The Hague: Mouton.
Brooke-Rose, C. (1958). 4 Grammar of Metaphor. London: Secker &
Warburg.
Brown, R. (1958). Words and Things. Glencoe, IIl.: Free Press.
Brown, R. (1973). A First Language. Cambridge, Mass.: Harvard University
Press.
Brown, R. & Ford, M. (1961). ‘Address in American English’. Journal of
Abnormal and Social Psychology 62. 375-85. (Reprinted in Hymes, 1964:
234-44.)
Brown, R. & Gilman, A. (1960). ‘The pronouns of power and solidarity’.
In Sebeok (1960: 253-76). (Reprinted in Giglioli, 1972: 252-82.)
854 Bibliography
Browne, W. (1970). ‘Noun phrase definiteness in relatives: evidence from
Macedonian’ & ‘More on definiteness markers: interrogatives in Persian’.
Linguistic Inquiry 1. 267-70 & 359-63.
Bruner, J. S. (1974/5). ‘From communication to language’. Cognition 3.
255-87.
Bruner, J. S. (1975). ‘The ontogenesis of speech acts’. Journal of Child
Language 2. 1-19.
Bihler, K. (1934). Sprachtheorie. Jena: Fischer. (Reprinted Stuttgart:
Fischer, 1965.)
Bull, W. E. (1963). Time, Tense and the Verb. Berkeley & Los Angeles:
University of California Press.
Burling, R. (1965). ‘How to choose a Burmese classifier’. In M. Spiro (ed.)
Context and Meaning in Cultural Anthropology. New York: Free Press.
Pp. 243-65.
Burton-Roberts, N. (1975). ‘Nominal apposition’. Foundations of Language
13. 391-419.
Callow, K. (1974). Discourse Considerations in Translating the Word of God.
Grand Rapids, Michigan: Zondervan.
Carnap, R. (1956). Meaning and Necessity, 2nd ed. Chicago: Chicago
University Press.
Carter, R. M. (1976). ‘Chipewyan classificatory verbs’. International Journal
of American Linguistics 42. 24-30.
Castafieda, H. (1968). ‘On the logic of attributing self-knowledge to others’.
Fournal of Philosophy 65. 439-56.
Caton, C. E. (ed.) (1963). Philosophy and Ordinary Language. Urbana, HL:
University of Illinois Press.
Caton, C. E. (1966). ‘Epistemic qualification of things said in English’.
Foundations of Language 2. 37-66.
Cedergren, H. J. & Sankoff, D. (1974). ‘Variable rules: performance as a
statistical reflection of competence’. Language 50. 333-55.
Chafe, W. L. (1968). ‘Idiomaticity as an anomaly in the Chomskyan
paradigm’. Foundations of Language 4. 109-27.
Chafe, W. L. (1970). 4 Semantically Based Sketch of Onondaga. (Indiana
University Publications in Anthropology and Linguistics, 25. 1.)
Chafe, W. L. (1971). Meaning and the Structure of Language. Chicago &
London: University of Chicago Press.
Chao, Y. R. (1968). A Grammar of Spoken Chinese. Berkeley & Los Angeles:
University of California Press.
Chapin, P. (1967). ‘On affixation in English’. In Bierwisch & Heidolph
(1970: 49-63).
Chatman, S. (ed.) (1971). Literary Style: A Symposium. London: Oxford
University Press.
Chomsky, N. (1957). Syntactic Structures. The Hague: Mouton.
Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, Mass.:
MIT Press.
Chomsky, N. (1968). Language and Mind. New York: Harcourt, Brace &
World. (Enlarged edition, 1972.)
 Bibhography 855
Chomsky, N. (1969). ‘Deep structure, surface structure and semantic
interpretation’. In Steinberg & Jakobovits (1971: 183-216). Reprinted in
Chomsky (1972: 62-119).
Chomsky, N. (1970). ‘Remarks on nominalization’. In Jacobs & Rosenbaum
(1970: 184-221). (Reprinted in Chomsky, 1972: 11-61.)
Chomsky, N. (1972). Srudies on Semantics in Generative Grammar. The
Hague: Mouton.
Chomsky, N. (1976a). The Logical Structure of Linguistic Theory. New York
& London: Plenum. (Partly revised version of unpublished 1955
manuscript, with a new Introduction.)
Chomsky, N. (1976b). Reflections on Language. London: Temple Smith.
Christie, J. J. (1970). ‘Locative, possessive and existential in Swahili’.
Foundations of Language 6. 166-77.
Christophersen, P. (1939). The Articles: A Study of their Theory and Use in
English. Copenhagen: Munksgaard & London: Oxford University Press.
Cicourel, A. V. (1973). Cognitive Sociology. Language and Meaning in Social
Interaction. Harmondsworth: Penguin.
Clark, E. V. (1973). ‘What’s in a word’. In Moore (1973: 28-63).
Clark, E. V. (1974). ‘Normal states and evaluative viewpoints’. Language 50.
316-32.
Clark, E. V. (1977). ‘From gesture to word: the natural history of deixis’. In
J.S. Bruner & A. Garton (eds.) Human Growth and Development. London:
Oxford University Press. .
Clark, E. V. & Garnica, O. K. (1974). ‘Is he coming or going?’. fournal of
Verbal Learning and Verbal Behavior 13. 556-72.
Clark, H. H. (1973). ‘Space, time, semantics and the child’. In T. E. Moore
(1973: 28-64).
Cohen, L. J. & Margalit, A. (1970). ‘The role of inductive reasoning in the
interpretation of metaphor’. Synthese 21. 483-500. (Reprinted in
Davidson & Harman, 1972: 722~—740.)
Cole, P. (1974). ‘Indefiniteness and anaphoricity’. Language 50. 665-74.
Cole, P. & Morgan, J. L. (eds.) (1975). Syntax and Semantics, 3: Speech
Acts. New York & London: Academic Press.
Closs, E. (1967). ‘Some copula construction in Swahili’, fournal of
African Linguistics 6. 105-31. .
Collinson, W. E. (1937). Indication: A Study of Demonstratives, Articles and
Other ‘ Indicaters’. (Language, Monograph 17.) Baltimore: Linguistic
Society of America.
Comrie, B. (1976). Aspect. London, New York & Melbourne: Cambridge
University Press. |
Conklin, H. C. (1972). Folk Classification: A Topically Arranged Bibliography
of Contemporary and Background References Through 1971. New Haven:
Department of Anthropology, Yale University.
Cooper, David E. (1974). Presupposition. The Hague & Paris: Mouton.
Coseriu, E. (1970). Sprache: Strukturen und Funktionen. 'Tibingen: Gunter
Narr.
Cresswell, M. J. (1974). ‘Adverbs and events’. Synthese 28. 455-81.
826 Bibliography ,
Cruse, D. A. (1973). ‘Some thoughts on agentivity’. Journal of Linguistics
Q. 11-23.
Crymes, R. (1968). Some Systems of Substitution Correlations in Modern
American English. The Hague: Mouton.
Crystal, D. & Davy, D. (1969). Investigating English Style. London: Longman.
Culler, J. (1975). Structuralist Poetics. London: Routledge & Kegan Paul.
Dahl, O. (1969). Topic and Comment: A Study in Russian and General
Transformational Grammar, (Slavica Gothenburgensia, 4.) Stockholm:
Almaqvist & Wiksell.
Dahl, O. (1970). ‘Some notes on indefinites’. Language 46. 33-41.
Dahl, O. (1971). ‘Tenses and world-states’. Gothenburg Papers in Theoretical
Linguistics, 6.
Dane, F. (1968). ‘Some thoughts on the semantic structure of the sentence’.
Lingua 21. §5-9.
Dascal, M. & Margalit, A. (1974). ‘A new ‘“‘revolution”’ in linguistics? —
‘“Text-grammars’”’ vs. ‘‘Sentence-grammars’”’’. Theoretical Linguistics 1.
195-213.
Davidson, D. (1967). ‘The logical form of action sentences’. In N. Rescher
(ed.) The Logic of Decision and Action. Pittsburgh: University of
Pittsburgh Press. Pp. 81~120.
Davidson, D. & Harman, G. (eds.) (1972). Semantics of Natural Language.
Dordrecht-Holland: Reidel.
Davidson, D. & Harman, G. (1975). The Logic of Grammar. Encino,
California: Dickenson.
Davy, D. & Quirk, R. (1969). ‘An acceptability experiment with spoken
output’. fournal of Linguistics 5. 109-20.
De Lemos, C. G. (1975). The Use of ‘Ser’ and ‘Estar’ in Brazilian
Portuguese with Particular Reference to Child Language Acquisition.
University of Edinburgh, Ph.D. Dissertation.
Dik, S. (1968). Co-ordination. Amsterdam: North-Holland.
Dik, S. (1973). ‘Crossing co-reference again’. Foundations of Language 9.
306-26.
Dik, S. (1975). ‘The semantic representation of manner adverbials’. In
A. Kraak (ed.) Linguistics in the Netherlands. Assen & Amsterdam:
Vangorevm.
Dixon, R. M. W. (1968). “Noun-classes’. Lingua 21. 104-25.

205-23). :
Dixon, R. M. W. (1971). ‘A method of semantic description’. In Steinberg
& Jakobovits (1971: 436-71).
Dixon, R. M. W. (1973). ‘The semantics of giving’. In Gross et al. (1973:

Dougherty, R. (1969). ‘An interpretive theory of pronominal reference’.

Foundations of Language 5. 488-519.
Dougherty, R. (1975). ‘Harris and Chomsky at the syntax-semantics
boundary’. In Hockney et al. (1975).
Dowty, D. R. (1972a). Studies in the Logic of Verb Aspect and Time Reference
in English. (Studies in Linguistics, 1.) Department of Linguistics, :
University of Texas at Austin.
 Bibhography 857
Dowty, D. R. (1972b). ‘On the syntax and semantics of the atomic predicate
CAUSE’. In Papers from the Eighth Regional Meeting of the Chicago
Linguistic Circle. Chicago: Chicago Linguistic Circle.
Drange, T. (1966). Type Crossings. The Hague: Mouton.
Dressler, W. (1972). Einfiihrung in die Textlinguistik. ‘Tubingen: Niemeyer.
Dressler, W. & Schmidt, S. J. (1973). Textlinguistik: Kommentierte
Bibliographie. Miinchen: Fink.
Dubois-Charlier, F. & Galmiche, M. (1972). La Semantique Générative.
(Languages, 27.) Paris: Didier/Larousse.
Dubois, J. (1962). Etude sur la Dérivation Suffixale en Francais Moderne et
Contemporain. Paris: Larousse.
Dubois, J. & Dubois, C. (1971). Introduction a la Lexicographie: Le
Dictionnaire. Paris: Larousse.
Ducrot, O. (1972). Dire et Ne Pas Dire. Paris: Hermann.
Ducrot, O. (1973). La Preuve et le Dire. Paris: Mame.
Ducrot, O. & Todorov, T. (1972). Dictionnaire Encyclopédique des Sctences du
Langage. Paris: Seuil.
Dummett, M. (1973). Frege: Philosophy of Language. London: Duckworth.
Ellis, J. (1966). ‘On contextual meaning’. In Bazell et al. (1966: 79-95).
Ellis, J. & Boadie, L. (1969). ‘‘‘To be” in Twi’. In Part 4 (1969) of Verhaar
(1967-73). Pp. 1-71.
Emeneau, M. B. (1951). Studies in Vietnamese (Annamese) Grammar. Berkeley
& Los Angeles: University of California Press.
Empson, William (1953). Seven Types of Ambiguity, 3rd edition. London:
Chatto. (Reprinted, Harmondsworth: Penguin, 1961.) (First edition, 1930.)
Ervin-Tripp, S. (1964). ‘An analysis of the interaction of language, topic and
listener’. American Anthropologist 64. 86-102.
Ervin-Tripp, S. M. (1969). ‘Sociolinguistics’. In L. Berkovitz (ed.) Advances
in Experimental Social Psychology, vol. 4. New York: Academic Press.
(Reprinted, abridged, as ‘Sociolinguistic rules of address’, in Pride &
Holmes, 1972: 225-40.)
Feerch, C. (1975). ‘Deictic NPs and generative pragmatics’. Foundations of
Language 13. 319-48.
Fann, K. T. (ed.) (1969). Symposium on }. L. Austin. London: Routledge &
Kegan Paul.
Fauconnier, Gilles (1974). La Coréference: Syntaxe ou Sémantique. Paris:
Seuil.
Ferguson, C. A. (1959). ‘Diglossia’. Word 15. 325-40. (Reprinted in Hymes,
1964: 429-39; Giglioli, 1972: 232-51.)
Ferguson, C. A. (1971). Language Structure and Language Use. Stanford:
Stanford University Press.
Ferreiro, E. (1971). Les Relations Temporelles dans le Langage de ? Enfant.
Geneve: Droz.
Fillenbaum, S. & Rapoport, A. (1971). Structures in the Subjective Lexicon.
New York & London: Academic Press.
Fillmore, C. J. (1966). ‘Deictic categories in the semantics of ‘come’’.
Foundations of Language 2. 219-27.
 858 Bibliography
Fillmore, C. J. (1968). ‘The case for case’. In Bach & Harms (1968: 1-88).
Fillmore, C. J. (1970). ‘Subjects, speakers and roles’, Synthese 21. 251-74.
(Reprinted in Davidson & Harman, 1972: 1-24.)
Fillmore, C. J. (19714). ‘Types of lexical information’. In Steinberg &
Jakobovits (1971: 3'70-92).
Fillmore, C. J. (1971b). ‘Verbs of judging: an exercise in semantic
description’. In Fillmore & Langendoen (1971: 273-89).
Fillmore, C. J. (1972). ‘On generativity’. In Peters (1972: 1-19).
Fillmore, C. J. & Langendoen, D. T. (eds.) (1971). Studies in Linguistic
Semantics. New York: Holt.
Firbas, J. (1964). ‘On defining the theme in functional sentence analysis’.
Travaux Linguistiques de Prague 1. 267-80.
Firbas, J. (1972). ‘On the interplay of prosodic and non-prosodic means of
functional sentence perspective’. In Fried (1972: 77-94).
Firth, J. R. (1935). ‘The technique of semantics’. In Transactions of the
Philological Society 36~72. (Reprinted in Firth, 19572.)
Firth, J. R. (1957a). Papers in Linguistics, 1934-1951. London: Oxford
University Press.
Firth, J. R. (1957b). ‘Ethnographic analysis and language with reference to
Malinowski’s views’. In R. Firth (ed.) Man and Culture. London:
Routledge. Pp. 93-118. (Reprinted in Palmer, 1968: 137-67.)
Firth, J. R. (1968). Cf. Palmer (1968). |
Fishman, J. A. (1965). ‘Who speaks what language to whom and when?’.
La Linguistique 2. 67-88. (Revised version: Fishman, 1972¢c.)
Fishman, J. A. (1968). Readings in the Sociology of Language. ‘The Hague:
Mouton.
Fishman, J. A. (1969). ‘Sociolinguistics’. In K. W. Back, Social Psychology.
New York: Wiley. (French version: J. A. Fishman, Sociolinguistique.
Bruxelles: Labor & Paris: Nathan, 1971.)
Fishman, J. A. (ed.) (1971). Advances in the Sociology of Language I: Basic
Concepts, Theories and Problems. "The Hague: Mouton.
Fishman, J. A. (ed.) (1972a). Advances in the Sociology of Language IT;
Selected Studies and Applications. ‘The Hague: Mouton.
Fishman, J. A. (1972b). The Sociology of Language. Rowley, Mass.: Newbury
| House.
Fishman, J. A. (1972c). ‘The relationship between micro- and macro-
sociolinguistics in the study of who speaks what language to whom and
when’. In Pride & Holmes (1972: 15-51).
Fodor, J. A. (1970). ‘Three reasons for not deriving ‘kill’ from ‘cause to
die’’. Linguistic Inquiry 1. 429-38.
Fodor, J. A. & Katz, J. J. (1964). The Structure of Language: Readings in the
Philosophy of Language. Englewood Cliffs, N.J.: Prentice-Hall.
Fodor, J. D. (1977). Semantics. New York: Crowell.
Fodor, J. D., Fodor, J. A. & Garrett, M. F. (1975). ‘The psychological
unreality of semantic representations’. Linguistic Inquiry 6. 515-31.
Follesdal, D. & Hilpinen, R. (1971). ‘Deontic logic: an introduction’. In
Hilpinen (1971: 1-35).
 Bibliography 859
Franck, D. & Petéfi, J. (eds.) (1973). Prdsuppositionen in der Linguistik und
der Philosophie. Frankfurt: Athenéum.
Fraser, B. (1970). ‘Idioms within a transformational grammar’. Foundations
of Language 6. 22-42.
Fraser, B. (1971). ‘An analysis of ‘even’ in English’. In Fillmore &
Langendoen (1971: 151-80).
Fraser, B. (1974). ‘A partial analysis of vernacular performative verbs’. In
R. Shuy & C. J. Bailey (eds.) Toward Tomorrow's Linguistics. Washington,
D.C.: Georgetown University Press.
Fraser, B. (1975). ‘Hedged performatives’. In Cole & Morgan (1975:
187-210).
Fraser, J. T. (ed.) (1968). The Voices of Time. London: Lane.
Frege, G. (1892). ‘Uber Sinn und Bedeutung’. Zeitschr. f. Philosophie und
philosoph. Kritik 100. 25-50. (English translation, ‘On sense and reference’
in Geach & Black, 1960: 56-78. Reprinted in Zabeeh et al., 1974:
118-140; Feigl & Sellars, 1949: 82—102.)
Frei, H. (1944). ‘Systéme de deictiques’. Acta Lingutstica 4. 111-29.
Fried, V. (ed.) (1972). The Prague School of Linguistics and Language
Teaching. London: Oxford University Press. .
Friedrich, P. (1966). ‘Structural implications of Russian pronominal usage’.
In W. Bright (ed.), Sociolinguistics. The Hague: Mouton. Pp. 214-53.
Friedrich, P. (1969). On the Meaning of the Tarascan Suffixes of Space.
(Indiana University Publications in Anthropology and Linguistics;
Memoir 23 of International Journal of American Linguistics.) Bloomington:
Indiana University.
Friedrich, P. (1970). ‘Shape in grammar’. Language 46. 379-407.
Friedrich, P. (1974). On Aspect Theory and Homeric Aspect. (Indiana
University Publications in Anthropology and Linguistics; Memoir 28 of
International Journal of American Linguistics.) Bloomington: Indiana
University.
Friedrich, P. (1975). Proto-Indo-European Syntax. (journal of Indo-European
Studies, Monograph 1.) Batte, Montana.
Fries, C. C. (1952). The Structure of English. New York: Harcourt Brace.
(London: Longman, 1957.)
Fromkin, V. (1971). ‘The non-anomalous nature of anomalous utterances’.
Language 47. 25-52.
Gale, R. M. (1968a). The Language of Time. London: Routledge & Kegan Paul.
Gale, R. M. (ed.) (1968b). The Philosophy of Time. London: Macmillan.
Galmiche, M. (1975). -Sémantique Générative. Paris: Larousse.
Garner, R. (1971). ‘‘Presupposition’ in philosophy and linguistics’,
In Fillmore & Langendoen (1971: 22-42).
Gazdar, G. (1976). Formal Pragmatics for Natural Language Implicature,
Presupposition and Logical Form. Ph.D. dissertation, University of Reading.
Geach, P. T. (1962). Reference and Generality. Ithaca, N.Y.: Cornell
University Press. |
Geckeler, H. (1971). Strukturelle Semantik und Wortfeldtheorie. Miinchen:
Wilhelm Fink.
860 Bibliography
Giglioli, P. P. (ed.) (1972). Language and Social Context. Harmondsworth:
Penguin.
Givon, 'T.. (1970). ‘Notes on the semantic structure of English adjectives’.
Language 46. 816~37.
Givon, T. (1973). ‘The time-axis phenomenon’. Language 49. 890-925.
Givoén, T. (1975). ‘Cause and control’. In Kimball (1975: 59-89).
Gladwin, T. & Sturtevant, W. C. (eds.) (1962). Anthropology and Human
Behavior. Washington, D.C.: Anthropological Society of Washington.
Gleason, H. A. (1962). ‘The relation of lexicon and grammar’. In
Householder & Saporta (1962: 103-10). |
Gleitman, L. R. & Gleitman, H. (1970). Phrase and Paraphrase. New York:
W. W. Norton.
Goftman, E. (1964). ‘The neglected situation’. American Anthropologist 66.
133-6. (Reprinted in Giglioli, 1972: 61-6.)
Gonda, J. (1956). The Character of the Indo-European Moods, Wiesbaden:
Harrassowitz.
Goodenough, W-H. (1956). ‘Componential analysis and the study of
meaning’. Language 32. 195-216.
Goodman, N. (1951). The Structure of Appearance. Cambridge, Mass. :
Harvard University Press.
Gordon, D. & Lakoff, G. (1971). ‘Conversational postulates’. In Papers
from the Seventh Regional Meeting of the Chicago Linguistic Circle.
Chicago: Department of Linguistics, University of Chicago. Pp. 63-84.
(Reprinted in Cole & Morgan, 1975: 83-106.)
Green, G. M. (1970). Review of R. Lakoff (1968) in Language 46. 149-67.
Green, G. M. (1974). Semantics and Syntactic Irregularity. Bloomington,
Ind.: Indiana University Press.
Green, G. M. (1975). ‘How to get people to do things with words’. In Cole
& Morgan (1975: 107-41).
Greenbaum, S. (1969). Studies in English Adverbial Usage. London:
Longman.
Greenbaum, S. (1970). Verb-Intensifier Collocations in English. The Hague:
Mouton.
Greenberg, J. H. (ed.) (1963). Universals of Language. Cambridge, Mass.: MIT ~
Press.
Greenberg, J. H. (1972). ‘Numeral classifiers and substantival number:
problems in the genesis of a linguistic type’. Working Papers in Language
Universals 9. 1-39. Stanford, Calif.: Stanford University.
Greenfield, P. & Smith, J. (1974). Communication and the Beginnings of
Language: The Development of Semantic Structure in One-Word Speech and
Beyond. New York: Academic Press.
Gregersen, E. A. (1967). Prefix and Pronoun in Bantu. (Indiana University
Publications in Anthropology and Linguistics; Memoir 21 of International
Journal of American Linguistics.) Bloomington: Indiana University.
Greimas, A. J. (1966). La Sémantique Structurale. Paris: Larousse.
Grewendorf, G. (1972). ‘Sprache ohne Kontext. Zur Kritik der
performativen Analyse’. In Wunderlich (1972: 144-82).
 Bibliography 861
Grice, H. P. (1957). ‘Meaning’. Philosophical Review 66. 377-88. (Reprinted
in P, F. Strawson (ed.) Philosophical Logic. Oxford: Oxford University
Press, 1971. Also in Steinberg & Jakobovits, 1971.)
Grice, H. P. (1968). ‘ Utterer’s meaning, sentence-meaning, and word-
meaning’. Foundations of Language 4. 1-18. (Reprinted in Searle, 1971:
54-70.)
Grice, H. P. (1975). ‘Logic and conversation’. In Cole & Morgan (1975:
41-58). (Part of the William James Lectures, delivered in 1967. 'To be
published in full by Harvard University Press.)
Groenendijk, J. & Stokhof, M. (1975). ‘Modality and conversational
information’. Theoretical Linguistics 2. 61-112.
Grootaers, W. A. (1952). ‘Différences entre langage masculin et féminin’. ,
Orbis 1. 84-85.
Gross, M. (1975). ‘On the relations between syntax and semantics’. In
Keenan (1975: 389-405).
Gross, M., Halle, M. & Schurzenberger, M-P. (eds.) (1973). The Formal
Analysis of Natural Languages. ‘The Hague: Mouton.
Gruber, J. S. (1967). ‘Topicalization in child language’. Foundations of
Language 3. 37-65. (Reprinted in Reibel & Schane, 1969: 422-47.)
Gruber, J. S. (1975). ‘‘* Topicalization”’ revisited’. Foundations of Language
13. 57-72.
Guilbert, L. (1975). La Créativité Lexicale. Paris: Larousse.
Guillaume, G. (1929). Temps et Verbe. Paris: Champion
Gumperz, J. (1968). ‘The speech community’. In International Encyclopaedia
of the Social Sciences. London & New York: Macmillan. (Reprinted in
Giglioli, 1972: 219-31.)
Gumperz, J. J. (1971). Language in Social Groups. Stanford University
Press.
Gumperz, J. H. & Hymes, D. H. (eds.) (1971). Directions in Sociolinguistics.
New York: Holt, Rinehart & Winston.
Haas, M. R. (1942). ‘The use of numeral classifiers in Thai’. Language 18.
201-6.
Haas, M. R. (1944). ‘Men’s and women’s speech in Koasati’. Language 20.
142-9. (Reprinted in Hymes, 1964: 228-33.)
Haas, M. R. (1967). ‘Language and taxonomy in Northern California’.
American Anthropologist 69. 358-62.
Haas, W. (1957). ‘Zero in linguistic description’. In Studies in Lingutstic
Analysis. Oxford: Blackwell.
Haas, W. (1973a). ‘Meaning and rules’. Proceedings of the Aristotelian
Society, 126-55.
Haas, W. (1973b). ‘Rivalry among deep structures’. Language 49. 282-93.
Habermas, J. (1972). ‘Vorbereitende Bemerkungen zu einer Theorie der
Kommunikativen Kompetenz’. In Holzer & Steinbacher (1972: 208-36).
Halle, M. (1972). ‘Prolegomena to a theory of word-formation’. Lingutstic
Inquiry 4. 3-16.
Halliday, M. A. K. (1961). ‘Categories of the theory of grammar’. Word 17.
241-92.
862 Bibliography
Halliday, M. A. K. (1966). ‘Lexis as a linguistic level’. In Bazell (1966:
148-62).
Halliday, M. A. K. (1967a). ‘Notes on transitivity and theme in English:
Part 1’. Journal of Linguistics 3. 37-81.
Halliday, M. A. K. (1967b). ‘Notes on transitivity and theme in English:
Part 2°. Journal of Linguistics 3. 199-244.
Halliday, M. A. K. (1970a). ‘Functional diversity in language as seen from a
consideration of modality and mood in English’. Foundations of Language
6. 322-65.
Halliday, M. A. K. (1970b). ‘ Language structure and language function’. In
Lyons (1970: 140-65).
Halliday, M. A. K. (1973). Explorations in the Functions of Language. London:
Arnold.
Halliday, M. A. K. (1975). Learning How To Mean. London: Arnold.
Halliday, M. A. K. & Hasan, R. (1976). Cohesion in English. London:
Longman.
Halliday, M. A. K., McIntosh, A. & Strevens, P. (1964). The Linguistic
Sciences and Language Teaching. London: Longman.
Hamblin, C. L. (1973). ‘Questions in Montague English’. Foundations of
Language 10. 41-53.
Hamp, E. P., Householder, F. W. & Austerlitz, R. (eds.) (1966). Readings in
Linguistics II, Chicago & London: University of Chicago Press.
Handford, S. A. (1947). The Latin Subjunctive. London: Methuen.
Harada, S. I. (1975). ‘Honorifics’. In Shibatani (1975: 499-61).
Harder, P. & Kock, C. (1976). The Theory of Presupposition Failure.

Akademisk Forlag.
(Travaux du Cercle Linguistique de Copenhague, 17.) Copenhague: —

Hare, R. M. (1949). ‘Imperative sentences’. Mind 58. 21-39. (Reprinted in

Hare, 1971: 1-21.)
Hare, R. M. (1952). The Language of Morals. Oxford: Clarendon Press.
Hare, R. M. (1970). ‘Meaning and speech acts’. Philosophical Review 79.
(Reprinted in Hare, 1971: 74-93.)
Hare, R. M. (1971). Practical Inferences. London: Macmillan.
Harman, G. (1970). ‘Deep structure as logical form’. Synthese 21. 2'75—97.
Harris, Z. (1951). Methods in Structural Linguistics. Chicago: University of
Chicago Press. (Reprinted as Structural Linguistics, 1961.)
Harris, Z. (1968). Mathematical Structures of Language. New York: Wiley.
Harris, Z. (1970). Papers in Structural and Transformational Linguistics.
Dordrecht: Reidel.
Harris, Z. (1976). Notes du Cours de Syntaxe. Paris: Seuil.
Hartvigson, H. H. (1969). On the Intonation and Position of the So-Called
Sentence Modifiers in Present-Day English. Odense, Denmark: Odense
University Press.
Harweg, R. (1968). Pronomina und Textkonstitution. Miinchen: Fink.
Hasan, R. (1971). ‘Syntax and Semantics’, In Morton (1971: 131-57).
Hasan, R. (1972). ‘The verb “‘be”’ in Urdu’. In Part 5 (1972) of Verhaar
(1967-73). Pp. 1-63.
 Bibliography 863
Hasegawa, K. (1972). ‘Transformations and semantic interpretation’.
Linguistic Inquiry 3. 141-59.
Haugen, E. (1953). The Norwegian Language in America, Philadelphia:
University of Pennsylvania Press.
Heger, K. (1963). Die Bezetchnung temporal-detktischer Begriffskategorien
im franzésischen und spanischen Konjugationssystem. (Zeitschr. f. Roman.
Philologie, 104.) Tiibingen: Niemeyer.
Heger, K. (1971). Monem, Wort und Satz. Tubingen: Niemeyer.
Heider, F. (1958). The Psychology of Interpersonal Relations. New York:
Wiley.
Heinrichs, H. M. (1954). Studien zum Bestimmten Artikel in den
Germanischen Sprachen. Giessen: Schmidt.
Helbig, G. (ed.) (1971). Bettrdge zur Valenztheorie. The Hague: Mouton.
Heringer, J. (1972). ‘Some grammatical correlates of felicity conditions’.
Working Papers in Linguistics 11. 1-110. Columbus, Ohio: Ohio State
University.
Hill, A. A. (1961). ‘Grammaticality’. Word 17. 1-10.
Hilpinen, R. (ed.) (1971). Deontic Logic: Introductory and Systematic Readings.
Dordrecht: Reidel.
Hinde, R. A. (ed.) (1972). Non-Verbal Communication. London & New York:
Cambridge University Press.
Hintikka, J]. (1962). Knowledge and Belief. Ithaca: Cornell University Press.
Hintikka, J. (1969). ‘Semantics for propositional attitudes’. In J. W. Davis
et al, (eds.) Philosophical Logic. Dordrecht: Reidel. (Reprinted in Linsky,
1971: 145-67.)
Hintikka, J. (1971). ‘Some main problems of deontic logic’. In Hilpinen
(1971: 59-104).
Hintikka, J. (1973). Logic, Language-Games and Information. London:
Oxford University Press.
Hintikka, J. (1974a). ‘Quantifiers vs. quantification theory’. Linguistic
Inquiry 5. 153-77.
Hintikka, J. (1974b). ‘Questions about questions’. In Munitz & Unger
(1974).
Hintikka, J., Moravcsik, J. M. E. & Suppes, E. (eds.) (1973). Approaches to
Natural Language. Dordrecht: Reidel.
Hirtle, W. H. (1967). The Simple and Progressive Forms: An Analytical
Approach. Quebec: Presses de |’Université Laval.
Hirtle, W. H. (1975). Time, Aspect and the Verb. Quebec: Presses de
Université Laval.
Hjelmslev, L. (1928). Principes de Grammaire Générale. Copenhague.
Hjelmslev, L. (1937). ‘La nature du pronom’. In Mélanges .. . Facques van
Ginneken. Paris: Klincksieck. (Reprinted in Hjelmslev, 1959.)
Hjelmslev, L. (1957). ‘Pour une semantique structurale’. (Reprinted in
Hjelmslev, 1959: 96-113.)
Hjelmslev, L. (1959). Essais Linguistiques. (Travaux du Cercle Linguistique
de Copenhague, 12.) Copenhague: Akademisk Forlag.
Hjelmslev, L. (1972). ‘Structural analysis of language’. In Malmberg (1972).
864 Bibliography
Hockett, C. F. (1958). A Course in Modern Linguistics. New York:
Macmillan.
Hockney, D., Harper, W. & Freed, B. (eds.) (1975). Contemporary Research
in Philosophical Logic and Linguistic Semantics. Dordrecht: Reidel.
Hoepelman, J. P. (1974). ‘Tense-logic and the semantics of the Russian
aspects’. Theoretical Lingutstics 1. 158-80.
Hoijer, H. (1945). ‘Classificatory verb stems in the Apachean languages’.
International Journal of American Linguistics 11. 13-23.
Hoijer, H. (ed.) (1954). Language in Culture. Chicago: Chicago University
Press.
Holt, J. (1943). Etudes d’ Aspect. (Acta Jutlandica, 15.) Copenhague &
Aarhus.
Holzer, H. & Steinbacher, K. (ed.) (1972). Sprache und Gesellschaft.
Hamburg: Hoffmann & Campe.
Hooper, J. B. (1975). ‘On assertive predicates’. In Kimball (1975: 91-124).
Householder, F. W. (1954). Review of E. A. Hahn, Subjunctive and Optative,
in Language 30. 389-99.
Householder, F. W. (1959). ‘On linguistic primes’. Word 15. 231-39.
Householder, F. W. (1962). ‘On the uniqueness of semantic mapping’. Word
18. 173-85.
Householder, F. W. (1971). Linguistic Speculations. London & New York:
Cambridge University Press.
Householder, F. W. (ed.) (1972). Syntactic Theory I: Structuralist.
Harmondsworth. Penguin.
Householder, F. W. (1973). ‘On arguments from asterisks’. Foundations of
Language 10. 365-76.
Householder, F. W. & Saporta, S. (eds.) (1962). Problems in Lexicography.
(Publications of Indiana University Research Center in Anthropology,
Folklore and Linguistics, 21.) (Supplement to International Fournal of
American Linguistics 28.) Baltimore: Waverley Press.
Huddleston, R. (1965). ‘Rank and depth’. Language 41. 474-86.
Huddleston, R. (1969). ‘Some observations on tense and deixis in English’.
Language 45. 777-806.
Huddleston, R. (1970). ‘Some remarks on case-grammar’. Linguistic Inquiry
I. SOI-11.
Huddleston, R. (1971). The Sentence in Written English: A Syntactic Study
Based on the Analysis of Scientific Texts. London & New York:
Cambridge University Press.
Huddleston, R. (1974). ‘Further remarks on auxiliaries as main verbs’.
Foundations of Language 11. 215-29.
Huddleston, R. (1976). An Introduction to English Transformational Syntax.
London: Longman.
Huddleston, R. & Uren, O. (1969). ‘Declarative, interrogative and imperative
in French’. Lingua 22. 1-26.
Hudson, R. A. (1973). ‘Tense and time reference in reduced relative clauses’.
Linguistic Inquiry 4. 251-6.
Hudson, R. A. (1975). ‘The meaning of questions’. Language 51. 1-31.
 Bibhography 865
Hudson, R. A. (1976). ‘Lexical insertion in a transformational grammar’.
Foundations of Language 14. 89-107.
Hull, R. (1975). ‘A semantics for superficial and embedded questions in
natural language’. In Keenan (1975: 35-45).
Hurford, J. (1973). The Linguistic Theory of Numerals. London & New York:
Cambridge University Press.
Huxley, R. (1970). ‘The development of the correct use of subject personal
pronouns in two children’. In G. B. F. Flores d’Arcais & W. Levelt (eds.)
Advances in Psycholinguistics. Amsterdam: North-Holland. Pp. 141-6s.
Huxley, R. & Ingram, E. (eds.) (1972). Language Acquisition: Models and
Methods. New York: Academic Press.
Hymes, D. (1962). ‘The ethnography of communication’. In Gladwin &
Sturtevant (1962: 13-53).
Hymes, D. (ed.) (1964). Language in Culture and Society. New York:
Harper & Row.
Hymes, D. (1967). ‘Models of the interaction of language and social setting’.
Journal of Social Issues 23. 8-28.
Hymes, D. (1971). On Communicative Competence. Philadelphia: University
of Pennsylvania Press. (Reprinted in an abridged form in Pride & Holmes,
1972: 269-93.)
Hymes, D. (1974). Foundations in Sociolinguistics. Philadelphia: University of
Pennsylvania Press.
Ikegami, Y. (1970). The Semological Structure of the English Verbs of Motion.
Tokyo: Sanseido.
Inoue, K. (1975). ‘Reflexivization in Japanese: an interpretive approach’.
In Shibatani (1975: 117-200).
Isard, S. (1974). ‘What would you have done if . . .?’. Theoretical
Lingutstics 1. 233-55.
Isard, S. (1975). ‘Changing the context’. In Keenan (1975: 287-96).
Jackendoff, R. S. (1968). ‘Quantifiers in English’. Foundations of Language.
4. 422-42.
Jackendoff, R. S. (1969). ‘An interpretive theory of negation’. Foundations of
Language §. 218-41.
Jackendoff, R. S. (1972). Semantic Interpretation in Generative Grammar.
Cambridge, Mass.: MIT Press. |
Jackendoff, R. S. (1975). ‘Morphological and semantic regularities in the
lexicon’. Language 51. 639-71.
Jacobs, R. & Rosenbaum, P. S. (eds.) (1970). Readings in English
Transformational Grammar. Waltham, Mass.: Ginn.
Jakobson, R. (1957). Shifters, Verbal Categories, and the Russian Verb.
Cambridge, Mass.: Harvard University Press. (Reprinted in Jakobson,
1971: 130-47.)
Jakobson, R. (1971). Selected Writings, vol. 2. The Hague: Mouton.
Jenkins, L. (1975). The English Existential. Tubingen: Niemeyer.
Jespersen, O. (1917). Negation in English and Other Languages. Copenhagen.
Jespersen, O. (1929). The Philosophy of Grammar. London: Allen & Unwin.
Jesperson, O. (1933). Essentials of English Grammar. London: Allen & Unwin.
 866 Bibliography
Jespersen, O. (1937). Analytic Syntax. Copenhagen: Munksgaard. (Reprinted,
New York: Holt, Rinehart & Winston, 1969.)
Jessen, M. (1974). A Semantic Study of Spatial and Temporal Expressions.
. Ph.D. Dissertation, University of Edinburgh.
Johanson, L.. (1971). Aspekt im Tirkischen. Uppsala: Universitetsbiblioteket.
Joos, M. (ed.) (1957). Readings in Linguistics. Washington, D.C. (Republished
as Readings in Linguistics I. Chicago & London: Chicago University Press.)
Joos, M. (1962). The Five Clocks. (Publications of the Indiana University
Research Center in Anthropology, Folklore and Linguistics, 22.)
Bloomington, Ind.: Indiana University & The Hague: Mouton.
Joos, M. (1964). The English Verb. Madison & Milwaukee: University of
Wisconsin Press.
Kachru, Y. (1968). ‘The copula in Hindi’. In Part 2 (1968) of Verhaar
(1967-73). Pp. 35-59.
Kahn, C. H. (1973). The Verb ‘be’ in Ancient Greek. (Part 6 of J. W. M.
Verhaar (ed.) The Verb ‘be’ and its Synonyms = Foundations of Language
Supplementary Series, 16.) Dordrecht: Reidel.
Karttunen, L. (1972). ‘Possible and must’. In Kimball (1972: 1-20).
} Karttunen, L. (1973). ‘Presuppositions of compound sentences’. Linguistic
Inquiry 4. 169-93.
Karttunen, L. (1974). ‘Presupposition and linguistic context’. Theoretical
Tanguistics 1. 181-94.
Kasher, A. (1972). ‘Sentences and utterances reconsidered’. Foundations of
Language 8. 312-45.
Kasher, A. (1974). ‘Mood implicatures: A logical way of doing pragmatics’.
Theoretical Lingutstics 1. 6-38.
Kastovsky, D. (1973). ‘Causatives’. Foundations of Language 10. 255-315.
Katz, J. J. (1964). ‘Semi-sentences’. In Fodor & Katz (1964: 400-16).
Katz, J. J. (1970). ‘Interpretative semantics vs. generative semantics’.
Foundations of Language 6. 220-59.
Katz, J. J. (1971). ‘Generative semantics is interpretive semantics’. Linguistic
Inquiry 2. 313-31.
Katz, J. J. (1972). Semantic Theory. New York: Harper & Row.
Katz, J. J. (1973). ‘On defining ‘presupposition’’. Linguistic Inquiry 4.
256-60.
Katz, J. J. & Fodor, J. A. (1963). ‘The structure of a semantic theory’.
Language 39. 170-210. (Reprinted in Fodor & Katz, 1964: 479-518.)
Katz, J. J. & Langendoen, D. T. (1976). ‘Pragmatics and presupposition’.
Language 52. 1-17.
Katz, J. J. & Nagel, R. I. (1974). ‘Meaning postulates and semantic theory’.
Foundations of Language 11. 311-40.
Katz, J. J. & Postal, P. M. (1964). An Integrated Theory of Linguistic
Descriptions. Cambridge, Mass.: MIT Press.
Kay, P. (1975). ‘Constants and variables in English kinship semantics’.
Language and Society 4. 257-70.
Kazazis, K. (1968). ‘The Modern Greek verbs of ‘‘being’’’. In Part 2 (1968)
of Verhaar (1967-73). Pp. 71-87.
 Bibliography 867
Keenan, E. L. (1971). ‘Two kinds of presupposition’. In Fillmore &
Langendoen (1971: 45-54).
Keenan, E. L. (1972). ‘On semantically based grammar’. Linguistic Inquiry 3.
413-61.
Keenan, E. L. (ed.) (1975). Formal Semantics of Natural Language. London
& New York: Cambridge University Press.
Keenan, E. L. & Hull, R. (1973). ‘The logical presuppositions of questions’.
In Franck & Petéfi (1973: 348-71).
Kempson, R. M. (1975). Presupposition and the Delimitation of Semantics.
London & New York: Cambridge University Press.
Kempson, R. M. (1977). Semantic Theory. London, New York & Melbourne:
Cambridge University Press.
Kenny, A. (1963). Action, Emotion and Will. London: Routledge.
Kholodovié, A. A. (ed.) (1969). Tipologya Kausatiunykh Konstrukci.
Leningrad: Izd. ‘ Nauka’.
Kholodovié, A. A. (ed.) (1974). Tipologija Passiunykh Konstrukcit. Leningrad:
Izd. ‘Nauk’.
Kiefer, F. (1968). ‘A transformational approach to the verb ‘van’ (‘‘to be’’)
in Hungarian’. In Part 3 (1968) of Verhaar (1967-73). Pp. 53-85.
Kiefer, F. (ed.) (1969). Studies in Syntax and Semantics. Dordrecht: Reidel.
Kiefer, F. (1973a). ‘Morphology in generative grammar’. In Gross et al.
(1973: 265-80).
Kiefer, F. (1973b). Generative Morphologie des Neufranzdésischen. 'Tiibingen:
Niemeyer.
Kiefer, F. & Ruwet, N. (eds.) (1973). Generative Grammar in Europe.
Dordrecht: Reidel.
Kimball, J. P. (ed.) (1972). Syntax and Semantics, vol. 1. New York &
- London: Seminar Press.
Kimball, J. P. (ed.) (1973). Syntax and Semantics, vol. 2. New York &
London: Seminar Press.
Kimball, J. P. (ed.) (1975). Syntax and Semantics, vol. 4. New York &
London: Academic Press.
Kiparsky, P. (1968). ‘Tense and mood in Indo-European syntax’.
Foundations of Language 4. 30-57.
Kiparsky, P. & Kiparsky, C. (1970). ‘Fact’. In Bierwisch & Heidolph (1970:
143-73).
Kirkwood, H. (1969). ‘Aspects of word order and its communicative function
in English and German’. Fournal of Linguistics 5. 85-107.
Kirkwood, H. (1970). ‘On the thematic function and syntactic meanings of
the grammatical subject in English’. Lingutstische Berichte 9. 34-46.
Klein, H. G. (1974). Tempus, Aspekt, Aktionsart. Tiibingen: Niemeyer.
Klima, E. (1964). ‘ Negation in English’. In Katz & Fodor (1964: 246-323).
Kénig, E. (1971). Adjectival Constructions in English and German. Heidelberg:
Julius Gross.
Kooij, J. G. (1971). Ambiguity in Natural Language. Amsterdam:
North-Holland.
Kraak, A. (1966). Negatieve Zinnen. Hilversum: W. de Haan.
868 Bibhography —
Kramsky, J. (1972). The Article and the Concept of Definiteness in Language.
Mouton: The Hague.
Kratochvil, P. (1968). The Chinese Language Today. London: Hutchinson.
Krause, M. E. (1969). On the Classification in the Athapascan, Eyak, and the
Tlingit Verb. (Indiana University Publications in Anthropology and
Linguistics; Memoir 24 of International Fournal of American Linguistics.)
Bloomington: Indiana University Press. ,
Kristeva, J. (1969). Semiottké. Paris: Seuil.
Kuno, S. (1971). ‘The position of locatives in existential sentences’.
Lingutstic Inquiry 2. 333-78.
Kuno, S. (1972a). ‘Pronominalization, reflexivization and direct discourse’.
Linguistic Inquiry 3. 161-95.
Kuno, 8. (1972b). ‘Functional sentence perspective: a case study for
Japanese and English’. Lingutstic Inquiry 3. 269-320. :
Kuno, S. (1973). ‘Constraints on internal clauses and sentential subjects’.
Linguistic Inquiry 4. 363-85.
Kurytowicz, J. (1936). ‘ Dérivation lexicale et dérivation syntaxique’. Bulletin
de la Société de Linguistique de Paris 37. 79-92. (Reprinted in Hamp et al.,
1966: 42~50.)
Kurylowicz, J. (1964). The Inflexional Categories of Indo-European.
Heidelberg: Winter.
Kurylowicz, J. (1972). ‘The role of deictic elements in linguistic evolution’.
Semiotica 5. 174-83. |
Labov, W. (1966). The Social Stratification of English in New York City.
Washington, D.C.: Center for Applied Linguistics.
Labov, W. (1970). ‘The study of language in its social context’. Studium
Generale 23. 66-84. (Reprinted in Giglioli, 1972: 283-307.)
Labov, W. (1972). Sociolinguistic Patterns. Philadelphia: University of
Pennsylvania Press.
Lakoff, G. (1971a). ‘On generative semantics’. In Steinberg & Jakobovits
(1971: 232-96).
Lakoff, G. (1971b). ‘Presupposition and relative grammaticality’. In
Steinberg & Jakobovits (1971: 329-40).
Lakoff, G. (1971c). ‘A note on vagueness and ambiguity’. Linguistic Inquiry
I. 357-9.
Lakoff, G. (1972). ‘ Linguistics and natural logic’. In Davidson & Harman
(1972 : 545-665).
Lakoff, G. (1975). ‘Pragmatics in natural logic’. In Keenan (1975 : 253-86).
Lakoff, G. & Ross, J. R. (1972). ‘A note on anaphoric islands and causatives’.
Linguistic Inquiry 3. 121-5.
Lakoff, R. (1968). Abstract Syntax and Latin Complementation. Cambridge,
Mass.: MIT Press.
Lakoff, R. (1969). ‘Some reasons why there can’t be any some-any rule’.
Language 45. 608-15.
Lakoff, R. (1970). ‘Tense and its relations to participants’. Language 46. 838-49.
Lakoff, R. (1971). ‘If’s and but’s about conjunction’. In Fillmore &
Langendoen (1971: 232-96).
 Bibliography 869
Lakoff, R. (1972). ‘Language in context’. Language 48. 907-27.
Lakoff, R. (1973). ‘The language of politeness’. In Papers from the Ninth
Regional Meeting of the Chicago Linguistic Society. Chicago: Department
of Linguistics, University of Chicago. Pp. 292-305.
Langacker, R. W. (196s). ‘French interrogatives: a transformational
description’. Language 41. 587-600.
Langacker, R. W. (1969). ‘In pronominalization and the chain of command’.
In Reibel & Schane (1969: 160-86).
Langacker, R. W. (1974). ‘The question of Q’. Foundations of Language 11.
I-37.
Langendoen, D. T. (1971). ‘The projection problem for presuppositions’. In
Fillmore & Langendoen (1971: 55-62).
Laver, J. (1970). ‘The production of speech’. In Lyons (1970: 53-75).
Leblane, H. (ed.) (1973). Truth, Syntax and Modality. Amsterdam:
North-Holland.
Leech, G. N. (1966). English in Advertising: A Linguistic Study of
Advertising in Great Britain. London: Longman.
Leech, G. N. (1969). Towards a Semantic Description of English. London:
Longman.
Leech, G. N. (1971). Meaning and the English Verb. London: Longman.
Leech, G. N. (1974). Semantics. Harmondsworth: Penguin.
Lees, R. B. (1960). The Grammar of English Nominalizations. (Publications of
the Indiana University Research Center in Anthropology, Folklore &
Linguistics, 12.) Bloomington, Ind.: Indiana University & The Hague:
Mouton.
Lees, R. B. (1970). ‘Problems in the grammatical analysis of English
nominal compounds’. In Bierwisch & Heidolph (1970: 174-86).
Lees, R. B. & Klima, E. S. (1963). ‘Rules for English pronominalization’.
Language 39. 17-28.
Lehiste, I. (1969). ‘‘‘ Being’”’ and ‘‘Having”’ in Estonian’. Foundations of
Language §. 324-41.
Lehrer, A. (1974). ‘Homonymy and polysemy: measuring similarity of
meaning’. Language Sciences 3. 33-9.
Lehrer, A. (1975). ‘Interpreting certain adverbs: semantics or pragmatics?’.
Journal of Linguistics 11. 239-48. |
Leisi, E. (1953). Der Wortinhalt. Heidelberg: Winter. (3rd Ed., 1967.)
Leist, A. (1972). ‘Zur Intentionalitat von Sprechhandlungen’. In
Wunderlich (1972: 59-98).
Lemmon, E. J. (1966). ‘Sentences, statements and propositions’. In Williams
& Montefiore (1966: 87-107).
Leontjev, A. A. (ed.) (1971). Semantiteskaja Struktura Slova. Moskva:
‘Nauka’.
Levelt, W. J. M. (1974). Formal Grammars in Linguistics and Psycholinguistics,
3 vols. The Hague: Mouton.
Lewis, D. (1969). Convention. Cambridge, Mass.: Harvard University Press.
Lewis, D. (1972). ‘General semantics’. In Davidson & Harman (1972:
169-218).
 870 Bibliography
Lewis, D. (1973). Counterfactuals. Oxford: Blackwell.
Lewis, D. (1975). ‘Adverbs of quantification’. In Keenan (1975: 3-15).
Li, Ying-Che (1972). ‘Sentences with be, exist and have in Chinese’.
Language 48. 573-83.
Lieberson, S. (ed.) (1967). Explorations in Sociolinguistics. The Hague:
Mouton.
Lightfoot, D. (1975). Natural Logic and the Greek Moods. 'The Hague &
Paris: Mouton.
Linsky, L. (ed.) (1971). Reference and Modality. London: Oxford University
Press.
Lipka, L. (1971). ‘Grammatical categories, lexical items and word-formation’.
Foundations of Language 7. 211-38.
Lipka, L. (1972). Semantic Structure of Word-Formation: Verb-Particle
Constructions in Contemporary English. Munich: Fink.
Ljung, M. (1970). English Denominal Adjectives. Lund: Acta Universitatis
Gothoburgensis.
Loewenberg, I. (1975). ‘Identifying metaphors’. Foundations of Language 12.
315-38.
Lounsbury, F. G. (1956). ‘A semantic analysis of the Pawnee Kinship usage’.
Language 32. 158-94.
Lounsbury, F. G. (1963). ‘The structural analysis of kinship semantics’. In
H. Lunt (ed.) Proceedings of the Ninth International Congress of
Linguistics. The Hague: Mouton. Pp. 1073-93.
Lyons, J. (1966). ‘Firth’s theory of meaning’. In Bazell et al. (1966:
288-302).
Lyons, J. (1967). ‘A note on possessive, existential and locative sentences’.
| Foundations of Language 3. 390-96.
Lyons, J. (1968). Introduction to Theoretical Linguistics. London & New
York: Cambridge University Press.
Lyons, J. (ed.) (1970). New Horizons in Linguistics. Harmondsworth:
Penguin Books.
Lyons, J. (1972). ‘Human Language’. In Hinde (1972: 49~85).
Lyons, J. (1975). ‘Deixis as the source of reference’. In Keenan (1975:
61-83).
Lyons, J. (1977). ‘Deixis and anaphora’. In 'T’, Myers (ed.) The Development
of Conversation and Discourse. Edinburgh: Edinburgh University Press.
McCawley, J. D. (1968). ‘The role of semantics in a grammar’. In Bach &
Harms (1968: 125-69).
McCawley, J. D. (1969). ‘Where do noun-phrases come from?’. In Jacobs &
Rosenbaum (1969: 166-83).
McCawley, J. D. (19714). ‘Tense and time-reference in English’. In Fillmore
& Langendoen (1971: 96-113).
McCawley, J. D. (1971b). ‘Prelexical syntax’. In R. J. O’Brien (ed.) Report
of the Twenty-Second Annual Round Table Meeting on Linguistics and
Language Studies. Washington, D.C.: Georgetown University. (Reprinted
in Seuren, 1974: 29-42.)
McCawley, J. D. (1973). Grammar and Meaning. Tokyo: 'Taishukan.
 Bibliography 871
McCawley, N. A. (1975). ‘Reflexivization in Japanese: A transformational
approach’. In Shibatani (1975: 51-116).
McIntosh, A. (1961). ‘Patterns and ranges’. Language 37. 325-37.
(Reprinted in McIntosh & Halliday, 1966: 183-99.)
McIntosh, A. & Halliday, M. A. K. (1966). Patterns of Language. London:
Longman.
Macnamara, J. (1971). ‘Parsimony and the lexicon’. Language 47. 359-73.
Makkai, A. (1972). Idiom Structure in English. The Hague: Mouton.
Malinowski, B. (1930). ‘The problem of meaning in primitive languages’. In
second and subsequent editions of Ogden & Richards, The Meaning of
Meaning. London: Routledge & Kegan Paul.
Malinowski, B. (1935). Coral Gardens and Their Magic, vol. 2. London: Allen
& Unwin. (Reprinted, Bloomington, Ind.: Indiana University Press, 1965.)
Malkiel, Y. (1968). Essays on Linguistic Themes. Oxford: Blackwell.
Malmberg, B. (1972). Readings in Modern Linguistics. The Hague: Mouton.
Marchand, H. (1969). The Categories and Types of Present-Day English
Word-Formation, 2nd ed. Munich: Beck.
Martin, S. (1964). ‘Speech levels in Japan and Korea’. In Hymes (1964:
407-15).
Matthews, P. H. (1967). ‘Latin’. Lingua 17. 153-81. (Special Volume
entitled Word Classes. Amsterdam: North-Holland.)
Matthews, P. H. (1972). Inflectional Morphology. London & New York:
Cambridge University Press.
Matthews, P. H. (1974). Morphology. London: Cambridge University Press.
Meljéuk, I. A. (1970). ‘Towards a functional model of language’. In
Bierwisch & Heidolph (1970: 198-207).
Meljéuk, I. A. (1974). Opyt Teorit Lingvisticeskikh Modelej ‘‘ Smysl <> Tekst’.
Moskva: Nauka.
Miller, G. A. (ed.) (1970). The Psychology of Communication.
Harmondsworth: Penguin.
Miller, G. A. & Johnson-Laird, P. N. (1976). Language and Perception.
Cambridge, Mass.: Harvard University Press & London: Cambridge
University Press.
Miller, J. E. (1972). ‘Towards a generative account of aspect in Russian’.
Journal of Linguistics 8. 217-36.
Mitchell, T. F. (1975). Principles of Firthian Linguistics. London: Longman.
Montague, R. (1974). Formal Philosophy: Selected Papers of Richard
Montague, edited with an introduction by Richmond Thomason. New
Haven, Conn.: Yale University Press.
Moore, T. (1973). ‘Focus, presupposition and deep structure’. In Gross et al.
(1973: 88-99).
Moore, T. E. (ed.) (1973). Cognitive Development and the Acquisition of
Language. New York & London: Academic Press.
Morton, J. (ed.) (1971). Biological and Social Factors in Psycholinguistics.
London: Logos.
Munitz, M. K. & Unger, P. K. (eds.) (1974). Semantics and Philosophy.
New York: New York University Press.
872 Bibliography
Nelson, K. (1974). ‘Concept, word and sentence’. Psychology Review 81.
267-85.
Newmeyer, F. J. (1971). ‘The source of derived nominals in English’.
Language 47. 786-96.
Nida, E. A. & Taber, C. R. (1969). The Theory and Practice of Translation.
Leiden: Brill.
Nilsen, D. L. F. (1972). Towards a Semantic Specification of Deep Case.
The Hague: Mouton.
Nilsen, D. L. F. (1973). The Instrumental Case in English. The Hague:
Mouton.
O’Brien, R. J. O. (ed.) (1968). Georgetown University Round Table Selected
Papers in Linguistics. Washington, D.C.: Georgetown University Press.
Olshewsky, T. M. (ed.) (1969). Problems in the Philosophy of Language.
New York: Holt, Rinehart & Winston.
Ota, A. (1963). Tense and Aspect in Present-Day American English. Tokyo:
Kenkyusha.
Paduéeva, E. V. (1970). ‘Anaphoric relations and their manifestations in the
text’. In Bierwisch & Heidolph (1970: 224-32).
Palek, B. (1968). Cross-Reference: A Study from Hyper-Syntax. (‘Travaux
Linguistiques de Prague, 3.) Prague.
Palmer, L. R. (1954). The Latin Language. London: Faber. |
Palmer, F. R. (ed.) (1968). Selected Papers of }. R. Firth, 1952-59. London:
Longman & Bloomington, Ind.: Indiana University Press.
Palmer, F. R. (1974). A Linguistic Study of the English Verb, and ed.
London: Longman.
Palmer, F. R. (1976). Semantics: A New Outline. Cambridge: Cambridge
University Press.
Parsons, T. (1970). ‘The analysis of mass terms and amount terms’.
Foundations of Language 6. 362-88.
Parsons, T. (1972). ‘Some problems concerning the logic of grammatical
modifiers’. In Davidson & Harman (1972: 127-41).
Partee, B. H. (1970). ‘Opacity, reference and pronouns’, Synthese 21.
359-85. (Reprinted in Davidson & Harman, 1972: 415-41.)
Partee, B. H. (1971). ‘On the requirement that transformations preserve
meaning’. In Fillmore & Langendoen (1971: 1-21).
Partee, B. H. (1972). ‘On the semantics of belief sentences’. In Hintikka,
Moravesik & Suppes (1972).
Partee, B. H. (1975a). ‘Deletion and variable binding’. In Keenan (1975:
16-34).
Partee, B. H. (1975b). ‘Montague grammar and transformational grammar’.
Linguistic Inquiry 6. 203-300.
Perlmutter, D. M. (1970). ‘On the article in English’. In Bierwisch &
Heidolph (1970: 233~48).
Peters, S. (ed.) (1972). Goals of Linguistic Theory. Englewood Cliffs, N.J.:
Prentice-Hall.
Petofi, J. S. & Rieser, H. (eds.) (1973). Studies in Text Grammar. Dordrecht:
Reidel.
 } Bibliography 873
Pettersson, T. (1972). On Russian Predicates: A Theory of Case and Aspect.
(Slavica Gothoburgensia, 5.) Gothenburg: Elander & Stockholm:
Almavist & Wiksell.
Phillips Griffiths, A. (ed.) (1967). Knowledge and Belief. London: Oxford
University Press.
Piaget, J. & Inhelder, B. (1969). The Psychology of the Child. London:
Routledge & Kegan Paul.
Pine-Coffin, R. S. (1961). St. Augustine: Confessions (English translation with
introduction). Harmondsworth: Penguin Books.
Postal, P. M. (1967). ‘On so-called ‘‘Pronouns”’ in English’. Monograph
Series on Language and Linguistics 19. 177-206. (Reprinted in Reibel &
Schane, 1969: 201-24; Jacobs & Rosenbaum, 1970: 56-82.)
Postal, P. M. (1971). Crossover Phenomena. New York: Holt, Rinehart &
Winston.
Postal, P. M. (1974). On Raising: One Rule of English Grammar and its
Theoretical Implications, Cambridge, Mass.: MIT Press.
Pottier, B. (1962). Systématique des Eléments de Relation. (Bibliothéque
Francaise et Romane. Série A: Manuels et Etudes Linguistiques, 2.)
Paris: Klincksieck.
Pottier, B. (1974). Linguistique Générale, Paris: Klincksieck.
Pride, J. (1970). Social Meaning and Language. London: Oxford University
Press.
Pride, J. & Holmes, J. (ed.) (1972). Sociolinguistics: Selected Readings.
Harmondsworth: Penguin.
Prior, A. N. (1967). Past, Present and Future. Oxford: Clarendon Press.
Prior, A. N. (1968). Time and Tense. Oxford: Clarendon Press.
Prior, A. N. & Prior, M. (1955). ‘Erotetic logic’. Philosophical Review 64. 43-59.
Puhvel, J. (ed.) (1969). Structure and Substance of Language. Berkeley &
Los Angeles: University of California Press.
Putnam, H. (1970). ‘Is semantics possible?’. In H. Kiefer & M. Munitz
(eds.) Languages, Belief and Metaphysics. New York: State University of
New York Press, 1970. (Reprinted in Putnam, 1975: 139-52.)
Putnam, H. (1975). Mind, Language & Reality. London & New York:
Cambridge University Press.
Quine, W. V. (1951). ‘Two dogmas of empiricism’. Philosophical Review 60.
20-43. (Reprinted in Quine, 1953: 20-46; Olshewski, 1969: 398-417;
ZLabeeh et al., 1974: 584-610.)
Quine, W. V. (1953). From a Logical Point of View. Cambridge, Mass.:
Harvard University Press.
Quirk, R. (1968). The Use of English, 2nd revised ed. London: Longman.
(First edition, 1962.)
Quirk, R. et al. (1972). A Grammar of Contemporary English. London:
Longman.
Quirk, R. & Svartvik, J. (1966). Investigating Linguistic Acceptability. 'The
Hague: Mouton.
Reibel, D. A. & Schane, S. A. (eds.) (1969). Modern Studies in English.
Englewood Cliffs, N.J.: Prentice-Hall.
 874 Bibhography
Reichenbach, H. (1947). Elements of Symbolic Logic. London & New York:
Macmillan.
Rescher, N. (1966). The Logic of Commands. London: Routledge & Kegan
Paul.
Rescher, N. (ed.) (1967). The Logic of Decision and Action. Hertford:
University of Pittsburgh Press.
Rescher, N. (ed.) (1968). Studies in Logical Theory. Oxford: Blackwell.
Rescher, N. & Urquhart, A. (1971). Temporal Logic. Vienna & New York:
Springer.
Rey, A. (ed.) (1970). La Lexicologie: Lectures. Paris: Minuit.
Richards, B. (1971). ‘Searle on meaning and speech acts’. Foundations of
Language 7. 519-38.
Richards, B. (1976). ‘Adverbs: from a logical point of view’. Synthese 32.
329-72.
Richards, I. A. (1925). Principles of Literary Criticism. London: Routledge &
Kegan Paul.
Ries, J. (1931). Was ist Ein Satz? Prague.
Rivero, M-L. (1971). ‘Mood and presupposition in Spanish’. Foundations of
Language 7. 305-36.
Rivero, M-L. (1972). ‘Remarks on operators and modalities’. Foundations of
Language 9. 209-41.
Robbins, B. L. (1968). The Definite Article in English Transformations. The
Hague: Mouton.
Robins, R. H. (1967). A Short History of Linguistics. London: Longman.

109~26). |
Robins, R. H. (1971). General Linguistics, 2nd ed. London: Longman.
Rohrer, C. (1971). ‘Zur Theorie der Fragesatze’. In Wunderlich (1971:

Rohrer, C. (1973). ‘On the relation between disjunction and existential

quantification’. In Gross et al. (1973: 224-32).
Rosetti, A. (1947). Le mot, 2nd ed. Copenhague & Bucharest. (Reprinted in
A. Rosetti, Linguistica. The Hague: Mouton, 1965.)
Ross, A. (1968). Directives and Norms. London: Routledge & Kegan Paul.
Ross, J. R. (1969a). ‘Auxiliaries as main verbs’. In W. Todd (ed.) Studies in
Philosophical Linguistics 1. Evanston, Ill.: Great Expectations Press.
Ross, J. R. (r969b). ‘On the cyclic nature of English pronominalization’. In
Reibel & Schane (1969: 187-200).

(1970: 222—72). ,
Ross, J. R. (1970). ‘On declarative sentences’. In Jacobs & Rosenbaum

Russell, B. (1905). ‘On denoting’. Mind 14. 479-93. :

Russell, B. & Whitehead, A. N. (1910). Principia Mathematica. London.
Ryle, G. (1949). The Concept of Mind. London: Hutchinson.
| _ Sadock, J. (1974). Towards a Linguistic Theory of Speech Acts. New York:
Academic Press.
Sampson, G. (1973). ‘The concept “‘semantic representation’’’. Semiotica 7.
97-134.
Sampson, G. (1975). The Form of Language. London: Weidenfeld &
Nicolson.
 Bibliography 875
Sandmann, M. (1954). Subject and Predicate. Edinburgh: Edinburgh
University Press.
Sankoff, G. (1972). ‘Language use in multilingual societies’. In Pride &
Holmes (1972: 33-51).
Sapir, E. (1921). Language. New York: Harcourt, Brace & World.
Sapir, E. (1949). Selected Writings in Language, Culture and Personality,
edited by D. G. Mandelbaum. Berkeley: University of California Press.
Schiffer, S. (1973). Meaning. Oxford: Oxford University Press.
Schopf, A. (ed.) (1974). Der Englische Aspekt. Darmstadt: Wissenschaftliche
Buchgesellschaft.
Searle, J. R. (1969). Speech Acts. London & New York: Cambridge
University Press.
Searle, J. R. (ed.) (1971). The Philosophy of Language. London: Oxford
University Press.
Searle, J. R. (1975). ‘Indirect speech acts’. In Cole & Morgan (1975: 59-82).
Sebeok, T’. A. (ed.) (1960). Style in Language. Cambridge, Mass.: MIT
Press.
Seuren, P. (1972). ‘Autonomous versus semantic syntax’. Foundations of
Language 8. 237-65. (Reprinted in Seuren, 1974: 96-122.)
Seuren, P. (ed.) (1974). Semantic Syntax. London: Oxford University
Press.
Sgall, P. et al. (1969). A Functional Approach to Syntax. Amsterdam:
Elsevier.
Segall, P. et al. (1973). Topic, Focus and Generative Semantics. Kronberg:
Scriptor.
Shaumjan, S. K. (1965). Strukturnaja Lingvistika. Moskva: ‘Nauka’.
Shaumjan, S. K. (1974). Applikativnaja Grammatika Kak Semanticeskaja
Teorija Estestvennykh Jazykov. Moskva: ‘Nauka’.
Shibatani, M. (1972). ‘Three reasons for not deriving ‘‘kill’’ from ‘‘cause to
die’’ in Japanese’. In Kimball (1972: 125-38).
Shibatani, M. (1973). ‘Lexical versus periphrastic causatives in Korean’.
Journal of Linguistics 9. 281-97.
Shibatani, Masayoshi (ed.) (1975). Syntax and Semantics, Volume 5: |
Japanese Generative Grammar. New York & London: Academic Press.
Shopen, T. (1973). ‘Ellipsis as grammatical indeterminacy’. Foundations of
Language 10. 65-77.
Sinclair, H. (1972). ‘Sensorimotor action patterns as a condition for the
acquisition of syntax’. In Huxley & Ingram (1972: 121-30). ]
Sinclair, H. (1973). ‘Language acquisition and cognitive development’. In
T. E. Moore (1973: 9-25).
Sinclair, J. McH. (1966). ‘Beginning the study of lexis’. In Bazell et al.
(1966: 410~30).
Sinclair, J. McH. (1972). A Course in Spoken English: Grammar. London:
Oxford University Press.
Slama-Cazacu, T. (1961). Langage et Contexte. La Haye: Mouton.
Sommerstein, A. H. (1972). ‘On the so-called definite article in English’. .
Lingutstic Inquiry 3. 197-209.
 876 Bibliography
Spang-Hanssen, H. (1954). Recent Theories of the Nature of the Language
Sign. (Travaux du Cercle Linguistique de Copenhague, 9.) Copenhagen;
Akademisk Forlag.
Sparck-Jones, J. (1964). Synonymy and Semantic Classification. Cambridge:
Cambridge Language Research Unit.
Spiro, M. (ed.) (1965). Context and Meaning in Cultural Anthropology.
New York: Free Press.
Stalnaker, R. C. (1968). ‘A theory of conditionals’. In Rescher (1968:
98-112).
Stalnaker, R. C. (1972). ‘Pragmatics’. In Davidson & Harman (1972:
380-97).
Stalnaker, R. (1974). ‘Pragmatic presuppositions’. In Munitz & Unger
(1974: 197-214).
Stampe, Dennis W. (1975). ‘Meaning and truth in the theory of speech
acts’. In Cole & Morgan (1975).
Steinberg, D. D. & Jakobovits, L. A. (eds.) (1971). Semantics. London &
New York: Cambridge University Press.
Steiner, George (1975). After Babel. London, New York & Toronto: Oxford
University Press.
Steinitz, R. (1969). Adverbial-Syntax. Berlin: Akademie-Verlag.
Stenius, E. (1967). ‘Mood and language-game’. Synthese 17. 254-74.
Stevenson, C. L. (1944). Ethics and Language. New Haven, Conn.: Yale
University Press.
Stockwell, R. P., Schachter, P. & Hall Partee, B. (1973). The Major
Syntactic Structures of English. New York: Holt, Rinehart & Winston.
Strang, B. M. H. (1968). Modern English Structure, 2nd ed. London:
Arnold. (First edition, 1962.)
Strawson, P. F. (1950). ‘On referring’. Mind 59. 320-44. (Reprinted in Caton,
1963: 162-93; Strawson, 1971; Olshewsky, 1969; Zabeeh et al., 1974.)
Strawson, P. F. (1952). Introduction to Logical Theory. London: Methuen.
Strawson, P. F. (1959). Individuals. London: Methuen.
Strawson, P. F. (1964a). ‘Intention and convention in speech acts’.
Philosophical Review 73. 439-60. (Reprinted in Searle, 1971: 23-38;
Strawson, 1971: 149-69.)
Strawson, P. F. (1964b). ‘Identifying reference and truth-values’. Theoria 30.
96-118. (Reprinted in Steinberg & Jakobovits, 1971: 88-99; Strawson,
1971: 73-95.)
Strawson, P. F. (1971). Logico-Lingutstic Papers. London: Methuen.
Strawson, P. F. (1974). Subject and Predicate in Logic and Grammar.
London: Methuen.
Stroll, A. (1954). The Emotive Theory of Ethics. Berkeley, Calif.: University
of California Press.
Sturtevant, E. H. (1917). Linguistic Change. Chicago: University of Chicago
Press. (Reprinted with a new Introduction by E. P. Hamp, Chicago, 1961.)
| Sudnow, D. (1972). Studies in Social Interaction. New York: Free Press.
Sussex, R. (1974). ‘The deep structure of adjectives in noun phrases’, In
Fournal of Linguistics 10. 111-31. .
 Bibliography 877
Swadesh, M. (1939). ‘ Nootka internal syntax’. International Fournal of
American Linguistics 9. 77-102.
Talmy, L. (1975). ‘Semantics and syntax of motion’. In Kimball (1975:
181-238).
Tarski, A. (1944). ‘The semantic conception of truth’. Philosophy and
Phenomenological Research 4. 341-75. (Reprinted in Tarski, 1956;
Olshewsky, 1969; Zabeeh et al., 1974.)

University Press. -
Tarski, A. (1956). Logic, Semantics, Metamathematics. London: Oxford

Teller, P. (1969). ‘Some discussion and extension of Manfred Bierwisch’s

work on German adjectivals’. Foundations of Language 5. 185-216.
Tesniére, L. (1959). Eléments de Syntaxe Structurale. Paris: Klincksieck.
Thomas, E. W. (1969). The Syntax of Spoken Brazilian Portuguese.
Nashville: Vanderbilt University Press.
Thomason, R. H. (1972). ‘A semantic theory of sortal incorrectness’.
Fournal of Philosophical Logic 1. 209-58.
Thomason, R. H. & Stalnaker, R. C. (1973). ‘A semantic theory of adverbs’.
Linguistic Inquiry 4. 195-220.
Thorne, J. P. (1972). ‘On the notion ‘‘definite’’’. Foundations of Language 8.
562-68.
Thurneysen, R. (1946). 4 Grammar of Old Irish. Dublin: Institute for
Advanced Study.
Todorov, T. (1966). Recherches Sémantiques. (Languages, 1.) Paris: Larousse.
Traugott, E. (1975). ‘Spatial expressions of tense and temporal sequencing’.
Semiotica 15. 207-30.
Turner, R. (1974). Ethnomethodology. Harmondsworth: Penguin.
Twaddell, W. F. (1960). The English Verb Auxiliaries, 2nd ed. Providence, ,
Rhode Island: Brown University Press.

& Winston.
Tyler, S. A. (ed.) (1969). Cognitive Anthropology. New York: Holt, Rinehart

Ullmann, S. (1957). Principles of Semantics, 2nd ed. Glasgow: Jackson &

Oxford: Blackwell.
Ullmann, S. (1962). Semantics. Oxford: Blackwell & New York: Barnes &
Noble.
Urdang, L. (1968). The Random House Dictionary of the English Language.
New York: Random House.
Urmson, J. O. (1952). ‘Parenthetical verbs’. Mind 61. 480-96. (Reprinted in
A. Flew (ed.) Essays in Conceptual Analysis, London: Macmillan, 1956;
C. E. Caton (ed.) Philosophy and Ordinary Language. Urbana, IIL:
University of Illinois Press, 1963.)
Vachek, J. (ed.) (1964). A Prague School Reader in Lingutstics.
Bloomington, Ind.: Indiana University Press.
Vachek, J. (1966). The Linguistic School of Prague. Bloomington, Ind.:
Indiana University Press.
Van Dijk, T. A. (1972). Some Aspects of Text Grammars. 'The Hague: Mouton.
Vendler, Z. (1957). ‘Verbs and times’. Philosophical Review 66. 143-60.
(Revised version in Vendler, 1967: 97-121. Also in Schopf, 1974: 213-34.)
878 Bibliography
Vendler, Z. (1967). Linguistics in Philosophy. Ithaca, N.Y.: Cornell
University Press.
Vendler, Z. (1968). Adjectives and Nominalizations. The Hague: Mouton.
Vendler, Z. (1972). Res Cogitans. Ithaca, N.Y. & London: Cornell
University Press.
Vennemann, T. (1973). ‘Explanation in syntax’. In Kimball (1973:
I-50).
Verhaar, J. W. M. (ed.) (1967-73). The Verb ‘Be’ and its Synonyms:
Philosophical and Grammatical Studies, Parts 1-6. (Foundations of
Language, Supplementary Series, vols, 1, 6, 8, 9, 14, 16.) Dordrecht:
Reidel.
Verkuyl, H. J. (1972). On the Compositional Nature of the Aspects.
Dordrecht: Reidel.
Voegelin, C. F. & Voegelin, F. M. (1957). Hopi Domains. (Indiana
University Publications in Anthropology and Linguistics, 14.)
Bloomington: Indiana University.
Von Wright, G. H. (1951). ‘Deontic logic’. Mind 61. (Reprinted in Von ©
Wright, 1956.)
Von Wright, G. H. (1956). Logical Studies. London: Kegan Paul.
Von Wright, G. H. (1963). Norm and Action. London: Routledge & Kegan Paul.
Von Wright, G. H. (1968). An Essay in Deontic Logic and the General Theory
of Action. (Acta Philosophica Fennica, 21.) Amsterdam: North-Holland.
Von Wright, G. H. (1971). ‘A new system of deontic logic’. In Hilpinen
(1971: 105-20).
Wall, R. (1972). Introduction to Mathematical Linguistics. Englewood Cliffs,
N.J.: Prentice-Hall. :
Waterhouse, V. (1963). ‘Independent and dependent sentences’. International
Fournal of American Linguistics 29. 45-54. (Reprinted in Householder,
1972: 65-81.)
Weinreich, U. (1953). Languages in Contact. (Special publication of the
Linguistic Circle of New York, Supplement to Word, 9.) (Second edition,
revised and enlarged, The Hague: Mouton, 1962.)
Weinreich, U.(1962). ‘Lexicographic description in descriptive semantics’.
In Householder & Saporta (1962: 25-44).
Weinreich, U. (1966). ‘Explorations in semantic theory’. In T. A. Sebeok
(ed.) Current Trends in Linguistics, vol. 3. The Hague: Mouton, 1966.
Pp. 395-477. (Published separately, Explorations in Semantic Theory. The
Hague: Mauton, 1972.)
Weinreich, U. (1969). ‘Problems in the analysis of idioms’. In Puhvel (1969:
23-81).
Weinrich, H. (1964). Tempus: Besprochene und Erzahlte Welt. Stuttgart:
Kohlhammer. (French translation, Le Temps. Paris: Seuil, 1973.)
Whorf, B. L. (1950). ‘An American Indian model of the universe’.
International Journal of American Linguistics, 16. (Reprinted in Whorf
1956: 57-64. Also in Gale, 1968b.)
Whorf, B. L. (1956). Language, Thought and Reality: Selected Writings of
Benjamin Lee Whorf, edited by J. B. Carroll. New York: Wiley.
 Bibliography 879
Wienold, H. (1967). Genus und Semantik. Meisenheim-am-Glan: Hain.
Wierzbicka, A. (1972). Semantic Primitives. Frankfurt: Athenéum.
Wierzbicka, A. (1975). ‘Why ‘kill’ does not mean ‘‘cause to die’’’.
Foundations of Language 13. 491-528.
Williams, B. & Montefiore, A. (eds.) (1966). British Analytical Philosophy.
London: Routledge & Kegan Paul.
Wilmet, M. (1972). Gustave Guillaume et son Ecole Linguistique. (Langues et
Culture, 12.) Paris: Nathan & Bruxelles: Labov.
Wilson, D. (1975). Presupposition and Non-Truth-Conditional Semantics.
London & New York: Academic Press.
Wise, H. (1975). A Transformational Grammar of Spoken Egyptian Arabic.
(Publications of the Philological Society, 26.) Oxford: Blackwell.
Wittgenstein, L. (1953). Philosophical Investigations. Oxford: Blackwell &
New York: Macmillan.
Wunderlich, D< (1970). Tempus und Zeitreferenz im Deutschen. Miinchen:
- Hueber.
Wunderlich, D. (ed.) (1971). Probleme und Fortschritte der Transformations-
grammatik. Miinchen: Fink.
Wunderlich, D. (ed.) (19724). Linguistische Pragmatik. Frankfurt: Athenadum.
Wunderlich, D. (1972b). ‘Zur Konventionalitit von Sprachhendlungen’. In
Wunderlich (1972: 11-58).
Wurzel, W. W. (1970). Studien zur Deutschen Lautstruktur. (Studia
Grammatica, 8.) Berlin: Akademie-Verlag.
Youngren, W. H. (1972). Semantics, Linguistics and Criticism. New York:
Random House.
Zabeeh, F., Klemke, E. D. & Jacobson, A. (eds.) (1974). Readings in
Semantics. Urbana, Chicago & London: University of Illinois Press.

Leontjev (1971: 94-106). .

Zhurinskij, A. H. (1971). ‘O semantiteskoj strukture prilagateljnykh’. In

Ziff, P. (1960). Semantic Analysis. Ithaca, N.Y.: Cornell University Press.

Ziff, P. (1964). ‘About ungrammaticalness’. Mind 73. 204-14.
Zimmer, K. E. (1964). Affixal Negation in English and Other Languages.
(Supplement to Word 20: 2, Monograph 5.) New York: Linguistic Circle
of New York.
Zuber, R. (1972). Structure Présuppositionnelle du Langage. Paris: Dunod.
Zwicky, A. (1971). ‘On reported speech’. In Fillmore & Langendoen
(1971: 73-7).
Zwicky, A. M. & Geis, M. (1971). ‘On invited inference’. Linguistic Inquiry
2. 561-6.
Zwicky, A. M. & Sadock, J. M. (1975). ‘Ambiguity tests and how to fail
them’. In Kimball (1975: 1-36).
BLANK PAGE
 Index of subjects

Asterisks indicate technical terms; bold type denotes the page number
where a technical term ts introduced.

absolute homography, 563 allative* case, 694

absolute homonymy, 563, 564, 565, ambiguity, 397, 409, 791
566, 567, 569 ambiguous*, 396
absolute homophony, 563 ambivalent*, 492
abstract deontic statements, 836. American accent, 608
abstract location, 719 American English, 548, 576, 617
accent*, 587, 616, 617 American Indian languages, 816
acceptability, 379 analytic* propositions, 787
accomplishments*, 711 anaphora*, 637, 659, 660, 667, 668,
achievements*, 712 670, 675
acquisition of language by children, anaphoric expressions, 670
449, 482, 566, 656 anaphoric* function, 637
act*, 483 anaphoric reference, 659, 660, 661
action*, 442, 483, 706 Ancient Greek, 737n, 819n
action-denoting expressions, 449 Ancient Greek aorist, 713
action-denoting verbs, 450 antecedent*, 659
action-schemata, 695 anthropocentrism, 690
activity*, 483 anthropomorphism, 690
address, 516 anticipatory* anaphoric reference,
adjectivalizations*, 396, 425 659
adjectivalized deictic adverbial, 655 antinomies, 781
adjectivalized interrogative, 761 aphorisms, 681
adjective, 448, 528, 530, 532 apodeictic*, 791
adjoined*, 388 apposition*, 472
adjuncts*, 435, 470 arbitrariness*, 3°75
adverbials, 496 Aristotelian logic, 434
adverbials of manner, 721 Aristotelian tradition 429
adverbs, 450 article*, 636
adverbs of manner, 451 ascribed*, 500
agency, 484 ascriptive, 437
agglutinating*, 376 ascriptive sentences, 471, 472
Aktionsart*, 705, 706 aspect*, 485, 682, 687, 703, 705, 708,
alethic logic, 793 713, 714, 719, 830
alethic* modality, 791, 793, 797, 798, assertion, 769
843, 844 Athapaskan, 466
alethic necessity, 794, 823, 845 Augustinian view of tense, 821, 822
882 Index of subjects
base*, 411 cohesion*, 631
base-rules*, 394, 395, 411 collocation*, 612, 613
behaviouristic analysis of meaning, command, 748, 749, 751, 752, 767,
745 768, 803, 826, 827, 832, 834, 835,
659 comment*, 501
Bloomfieldian tradition, 392, 629, 837, 838
border-crossings*, 720 commitment*, 734
bound* form, 533 communication, 730
Brazilian Portuguese, 717 communication in face-to-face
British English, 535, 537, 544, 546, interaction, 637, 638
548 communicative competence*, 573
communicative intention, 732, 733
canonical encounter*, 691 communicative role, 848
703 complement, 769
canonically* orientated, 698, 699, competence’, 398, 572, 586

638 534
canonical situation of utterance, 637, | complex* lexemes, 521, 523, 524,

case, 426, 488 complex sentence, 633, 661

case-grammar*, 436, 437, 476, 495 complex utterance-units, 634
cataphora*, 659 componential analysis, 414, 422, 553
categorial grammars, 394 compound*, 534
categorical assertion, 745, 768, 7098, compound lexeme*, 424, 521, 534,
809 535, 536, 537, 539, 549, 541, 542,
causality, 493 543, 547, 548
causativity*, 490, 511, 721 compound* sentence, 626, 633, 661
centrality of the verb, 678 conative function of language, 609
channel*, 581 concessive mood, 847
channel-noise, 398 concord*, 382, 435, 808
character*, 706 conditional in French, 847
checking tags*, 764, 765 conditional mood, 847
Chinese, 448, 462, 678, 679, 687 conditioned obligation, 830
Chomskyan generative grammar, Confessions, 811
410, 515n confirm*, 771
Chomskyan transformational- confrontation*, 698, 699
generative grammar, 410 confronting*, 691
Chomskyan transformational conjoined*; 388
grammar, 674, 775 constative*, 511n, 793
circumstantial* roles, 497 constative* utterances, 726, 727, 729,
citation-form*, 376, 378, 513 445
class-denoting expressions, 449 context, 397, 398, 545, 572, 573, 611
Classical Arabic, 462 conext-bound* statements, 771%
Classical Chinese, 383, 568 context-dependence, 589
Classical Greek, 688, 759, 761 context-dependent proposition, 683
Classical Hebrew, 687 context-free assertion, 803, 815
Classical Latin, 579 context-free negative assertions, 774
classifiers*, 453, 460, 461, 675 context-free* statements, 7°71
classifier-languages, 461, 464 context-independent proposition, 683
clausal parataxis, 627 context-of-culture, 609
cleft-sentence construction, 598 context-of-situation*, 607, 609, 615,
code-switching*, 580, 581, 585, 617, 634
617 context-of-utterance*, 572, 643, 644,
cognitive response, 732 672, 830
 Index of subjects 883
contextual*, 572, 607 deictic simultaneity*, 685
contextualize*, 404, 574, 588, 610, deictic temporal reference, 679
611, 625 deixis*, 575, 636, 637, 638, 654, 656,
contextual theories, 735 659, 660, 667, 673, 677, 689, 819,
contextual theory of meaning, 607, 820
611 deliberative questions, 803
contextual variables, 574 delocutive* verbs, 739, 740, 742
contradictions*, 416, 417, 787 demand*, 751, 752, 803, 834, 835,
contradictory*, 772 837
contradictory statements, 774 demonstrative, 637, 654, 658
contra-factive statements, 818 demonstrative adjectives, 453, 646,
contra-factive* utterance, 795 652
contra-factivity, 795, 796, 816, 810, demonstrative adverbs, 646
825 demonstrative pronoun, 646, 647
contrary*, 772 denial*, 770, 771, 774, 777, 803
contrary statements, 774 denotation, 610
conventionalized pseudo-descriptions, deontic*, 823
645 deontically modalized utterances, 824
conversational implicatures*, 574, deontic logic, 836
785 deontic* modality, 739, 773, 786,
conversion*, 522, 684 792, 793, 817, 824, 825, 826, 833,
co-ordination, 405, 406, 782 839, 840, 841, 844, 845, 846, 849
co-ordinative* phrases, 391 deontic necessity, 797, 823, 829, 845
copulative* verb, 433 deontic* necessity and possibility, 791
copy tags*, 764 deontic source*, 843
co-reference, 667, 675, 676 deontic statement*, 828, 830, 833,
co-referential*, 660 834, 835, 836, 837, 843
correct belief, 793 deontic utterances, 827, 832
corrigibility*, 379, 380, 381, 384, dependency grammars, 394
385, 665 derivation*, 521, 565, 706
counter-factual* conditionals, 795 derivational rules, 524
creativity, 549 derived lexemes, 534, 536
Czech, 506, 717 descriptive function of language, 591,
609, 618
dative* case, 694, 721 descriptive meaning, 615
declarative sentence, 729, 748, 749, desentential transform*, 395
834, 841 desiderative*, 815
decontextualization*, 588 desiderative utterances, 826
deep structure*, 394, 395, 404, 412, determinable propositions, 849
413, 414, 663, 782 determiner*, 452, 453, 454, 465
definite article, 453, 646, 647, 650, diachronic linguistics, 620
654, 655, 656 diachronic semantics, 620
definite descriptions, 641, 643, 645 diachronic* variation, 616, 619, 620,
definite noun-phrases, 640 . 622
definiteness of reference, 454 diachrony-in-synchrony, 621, 622
deictic distinctions, 670 dialect*, 587, 616, 617, 619
deictic elements, 591 dialectal* variation, 616, 619, 620,
deictic function, 637, 661 622
deictic point-of-observation, 700 diglossia*, 5°76, 580, 617
deictic projection, 579, 822 dimensionality*, 697, 699
deictic reference, 661 directional* relations, 694
deictic roles, 574 directional* schemata, 495
884 Index of subjects
directive*, 745, 746, 826, 827, 828, 681, 682, 683, 687, 689, 694, 695,
829, 831, 832 696, 701, 703, 705, 706, 707, 708,
direct object, 488 7 712, 713, 715, 716, 718, 719, 722,
discovery procedures, 571 728, 736, 739, 749, 750, 751, 753,
displacement*, 657 754, 757, 758, 759, 760, 762, 763,
distal demonstrative, 669 764, 765, 766, 774, 777, 778, 779,
distribution*, 556 780, 788, 789, 790, 794, 800, 801,
distributional theory of meaning, 613 802, 804, 805, 806, 807, 817, 818,
domain*, 584, 585 819, 820n, 821, 830, 834, 839, 841,
duality*, 375 844, 846, 847
dubitative mood, 847, 848 English perfect, 715, 716
durative situations, 711 English progressive, ‘709, 719
dynamic relations, 694 entailment, 592, 788
dynamic situation, 483, 485, 707, 710 entity, 462, 702
dynamic verbs, 706 epistemic logic, 793, 794, 821
epistemic modality*, 682, 739, 763,
egocentric*, 638 786, 792, 793, 797, 798, 800, 801,
Finsteinian theory of relativity, 685 802, 803, 805, 806, 808, 819, 820,
elements, 386 824, 840, 841, 842, 843, 844, 849
elision*, 399 epistemic* necessity, 791, 792, 794,
ellipsis, 589 797, 805, 823
embedded*, 388 epistemic possibility, 791, 805, 845
emotive theory of ethics, 727 epistemic warrant*, 808
emotive* utterances, 727 epistemology, 793
empathetic* deixis, 677, 822 equative sentences, 471
encapsulated*, 618 erotetic* logic, 762
encapsulation, 652 essential conditions, 733, 734, 737. -
endocentricity*, 391, 392, 393 "etymology, 516, 551
538 event*, 483, 712
endocentric noun expressions, 537, European languages, 708, 710
English, 373, 375, 377, 3780, 379, exemptions*, 773, 837, 839, 840
381n, 382, 385, 386, 389, 390, 391, exemption-granting utterances, 840
396, 397, 399, 401, 402, 404, 407, — existence, 722
408, 409, 421, 425, 426, 428, 430, existential causatives*, 723
432, 433, 436, 437, 438, 440, 441, exocentric*, 392
443n, 444, 446, 447, 448, 450, 451, experiential description, 688
452, 453, 455, 456, 458, 460, 461, — experiential mode, 689
462, 464, 465, 468, 470, 471, 473, | experiential mode of description, 812
474, 476, 478, 479, 482, 485, 487, explicitly performative command, 730
488, 489, 490, 492, 494, 497, 501, explicitly performative demand, 751
503, 504, 506, 507, 508, 509, 510, explicitly performative element, 744
512, 513, §15, 518, §23, §524n, 525, explicitly performative question, 729
527, 528, 532, 533, 534, 535, 536, explicitly performative statement, 729
542, $43, 545, 547, 550, 551, 556, explicitly performative utterances,
558, 561, 562, 563, 567, 568, 560, 778, 779, 781
575, 576, 579, 581, 585, 587, 588, explicit performatives, 728, 736, 743
598, 599, 600, 614, 616, 618, 619, expression-class*, 423, 424, 428, 438,
623, 625, 630, 631, 632, 634, 639, 448, 449,714
640, 641, 642, 643, 645, 646, 647, expressions, 392, 658, 675
648, 650, 651, 653, 654, 655, 656, expressive*, 615
658, 660, 661, 664, 668, 669, 670, expressive* function of language,
671, 672, 675, 676, 678, 679, 680, 583
 Index of subjects . 885
expressive* information, 725 generic* propositions, 680
extension, 701 generic* reference, 454
extensional interpretation of futurity, | generic states, 717
814 German, 453, 488, 506, 551, 576,
669, 703, 704, 705, 706, 707, 726n,
factitive schemata, 493 753, 760, 763
factitive* verb, 491, 815 Germanic languages, 619, 646
factive* predicator, 794, 795 given*, 508
factive* verb, 599, 772 glossematic*, 411
factivity*, 794, 795, 796, 816, 820 gnomic*, 681
factuality, 805 gnomic propositions, 681
factual questions, 773 gnomic utterances, 681
falling intonation, 759 goal* of locomotion, 694
felicity conditions*, 605, 733, 735; government*, 435
737, 738, 752 Graeco-Roman grammatical tradi-
field-theory*, 583 : tion, 704
figurative expressions, 567 _ grammar, 378, 423, 727
figurative extension, 552 grammatical acceptability, 379
445 402, 404, 406
first-order entities*, 442, 443, 444, grammatical ambiguity*, 396, 400,
first-order nominals*, 446 grammaticality, 379, 383, 384, 385,
first-order nouns, 447 418, 632
first-person imperatives, 747 grammaticalization, 706
first-person pronoun, 638, 645, 667 grammatically well-formed, 385, 578
Firthian theory of meaning, 607 grammatical meaning, 608
focus*, 503, 509, 598 grammatical morphemes, 384
formal identity, 562 grammatical structure, 634
formality*, 580 grammatical subject, 504, 505, 506
forms, 452, 453, 658, 675 grammatical unacceptability, 370,
form-class*, 375, 423, 424 380, 381, 385
fossilized compounds, 547, 548 grammatical utterances, 384, 387
frame-of-reference, 684 Greek, 446, 507, 515, 636, 646, 659,
free* form, 533n 660, 704, 707, 709, 758, 791, 816,
French, 399, 404, 453, 488n, 515, 818, 819 |
535, 542, 548, 576, 579, 643, 648,
669, 688, 703, 704, 705, 707, 708, habituality, 716
709, 710, 715, 716, 722, 739, 756, head*, 391
758, 763, 766 head-noun, 538
French simple past, 713 head-word*, 512, 513, 515
functional equivalence, 634 heavy stress, 662
functionalism, 506 higher predicate calculus, 683
functional sentence perspective, 509 historical* description, 688
fusional*, 376 historical mode of description, 812
future tense, 816 homography*, 558, 559, 562
futurity, 824 homonymous lexemes, 550, 583
homonymy*, 398, 401, 403, 512, 543,
gender*, 386 545, 550, 551, 553, 555, $57, 560,
generality, 409 561, 562, 563, 564, 565, 567, 568,
generative grammar*, 386, 387, 392, 569
396, 397, 404, 526, 662, 665, 753, + homophones*, 558, 563
792 homophony’*, 558, 559, 562
generative semantics, 409, 413, 422 honorific expressions, 642, 643
 886 Index of subjects
hyponymy, 537, 547 interrogation, 804
hypostatization*, 445, 657 interrogative sentences, 747, 748,
hypostatized*, 693 749, 758
interrogative tags*, 764
iconic*, 511n intertextuality*, 672
identity-symbol, 546 intonation, 374, 598, 762
illocutionary* act, 730, 731, 732, 733, intonation contour, 626, 627, 661
734, 735, 736, 737, 740, 769, 780, intonation pattern, 754
781, 805, 826, 836 intransitive, 793
illocutionary force*, 634, 644, 731, invited reference, 7’70n
733, 734, 735, 736, 750, 751, 755, irony, 578
764, 778, 815, 827, 834, 835, 848 isolating*, 376
illocutionary uptake*, 731 isolating* languages, 517
imperative mood, 839, 847, 848 Italian, 453, 478, 579, 707
imperative sentences, 747, 748, 835,
837, 838, 840 Japanese, 508n, 575, 576, 641
imperfective, 705, 714 journey*, ‘720
implicata*, 593 junctural* phenomena, 399, 400
implicated*, 593 jussive* sentence, 748, 749
implicature*, 592, 593, 605, 606
incidental* illocutionary force, 785 kernel-sentence*, 467, 468, 497
indefinite declarative sentence, 761 Korean, 576, 641
indexical*, 615, 637
indexical* information, 587, 725 language behaviour, 617
indexical and instrumental functions language-competence, 589
of language, 849 language-games, 727, 728
indicate*, 615 language-process, 571 |
indicative mood, 847, 848 language-system, 409, 415, 416, 419,
indicative sentences, 752 421, 422, 477, 529, 538, 548, 552,
indirect illocutionary acts, 785 555, 560, 561, 633, 642, 847
indirect interrogative, 756 language-utterances, 827
indirect object, 488 Latin, 446, 447, 454, 488, 494, 507,
indirect speech-acts*, 784, 786 515, 524n, 551, 568, 636, 639, 648,
Indo-European languages, 436, 460, 659, 660, 669, 670, 671, 688, 704,
462, 518, 519, 535, 646, 677, 703, 707, 716, 739, 742, 746, 751, 753,
746, 747, 758, 760, 767, 815, 816, 757n, 758, 759, 760, 818, 819, 821,
817, 818, 826, 848 848
inferential mood, 847 levels*, 373
inflexion, 706 lexeme, 381, 383, 393, 437, 441, 452,
inflexional rules, 524 468, 471, 513, 514, 515, 517, 519,
information-focus*, 774, 775 520, 530, 535, 542, 543, 544, 549,
information structure*, 508, 509 551, 552, 553, 555, 556, 558, 561,
institutionalization, 543 564, 569, 619, 647, 657, 658, 696,
instrumental adverbials, 721 714
instrumental function of language, lexemic distinctness, 560, 561
485, 826 lexemic identity, 565, 568
| intensional interpretation of futurity, lexical entries*, 512
814 lexicalization*, 414, 549, 706
intentionality, 733 lexically ambiguous, 406
interpersonal function of language, lexical morphemes, 384
sor. lexicon*, 412, 514, 536
interpretive* semantics, 411, 413 liaison*, 399
 Index of subjects 887
linguistic ambiguity, 398 metaphysics, 727
linguistic competence, 566, 573 metonymy*, 548
linguistic relevance, 571 microlinguistics*, 585
linguistic semantics, 590, 591 microlinguistic semantics, 590, 591,
locale*, 584 635, 660, 849
localism*, 718, 721, 724 mimicry, 741
localist*, 496 modal adjectives, 802, 839
localist hypothesis, 718 modal adverbs, 802, 807
localist theory of truth, 723, 724 modal auxiliary verbs, 802
localization*, 669 modal expressions, 808
location, 516 modal formulae, 804
locative* adverbs, 392 modal logic, 686, 769, 791, 804, 805,
locative* case, 694 810
locative complement, 473 modality*, 452, 677, 724, 725, 735,
locative copula, 723 809, 820, 848, 849
locative expressions, 699, 700 modalized proposition, 833
locative subjects, 475, 723 modally harmonic* combinations,
locutionary* acts, 730 807
logical modality, 824, 344 modal negation*, 771
logical necessity, 823, 829 modal operator, 769, 807
logical positivists, 727, 728 modal remoteness, 719
logical semantics, 730 modal utterances, 802
logical structure of directives, 825 modal* verb, 791, 807, 808, 839
510 modifier,* 391
logical subject, 502, 504, 505, 506, modes* of description, 688
logic of tense, 686 modifying noun, 538
lower predicate calculus, 443 mood*, 386, 436, 682, 725, 735, 746,
841, 848
Macedonian, 763 morphemes*, 383
machine translation, 583 morphological criteria, 427
Malay, 678, 679, 687 morphosyntactic criteria, 427
mand, 745, 746, 748, 751, 753, 754, morphosyntactic* words, 377, 383,
755, 766, 768, 784, 826, 834, 835, 555, 557, 558, 564, 565
837, 838, 839, 840 moveable entities, 698
Mandarin Chinese, 461
marked*, 503, 647 natural kinds*, 447
markedness, 708. natural semiotic system, 639
maximal extension, 702 necessity, 802, 803, 836, 838, 848
maximally decontextualized utterance- negation, 724, 768, 774, 776, 777
signals, 625 negative jussive sentences, 774
meaning, 415, 542, 544, 592, 608, negative particles, 777
609, 622, 733 negative sentences, 777
meaningfulness, 735 Neogrammarians (Junggrammatiker),
meaning of utterances, 778 620
meaning-postulates*, 789 neustic*, 749, 750, 752, 767, 770,
medium*, 58x 802, 825, 848
medium-transferability, 624, 638 new* information, 508
mensural* classifier, 463, 464 nexus*, 773
metalanguage, 682 nexus negation*, 773, 837
metalinguistically, 417 noise*, 586
metaphor*, 548, 550 nominal expressions, 425
metaphorical extension, 549, 552 nominalizations*, 396, 425
388 / Index of subjects
nominals, 425, 426, 432, 496, 637, omnitemporal* proposition, 680
657 one-place deontic predicate, 836
non-accomplishments, 711, 712 ontological, 441
non-commitment, 802 open questions, 765
non-constative utterances, 745 open yes—no questions, 762
non-deictic point-of-observation, 700 operative* verb, 491
non-deictic reference, 610 operators, 454
non-deictic temporal reference, 679 optative*, 815, 816
non-deontic modal logic, 805 orientation*, 697, 699, 701
non-descriptive meaning, 615 ostension*, 637
non-epistemic modal logic, 805
non-existence, 777 paralinguistic features, 374, 624
non-factive*, 796, 825 paralinguistic modulation*, 738, 754
non-factive mood, 817 paralinguistic phenomena, 576
non-factive* predicator, ‘795 parataxis*, 782
non-factive questions, 800 parenthetical modulation, 782
non-factive statements, 818 parenthetical verbs*, 738, 772n
non-factivity, 795, 816, 817, 848 partial homonymy, 566, 569
non-generative, 411 participant roles*, 497, 641, 642, 643
non-isolating languages, 517 participants*, 570
non-kernel* sentence, 395, 492 part-of-speech*, 423, 425, 428, 438,
non-modal formulae, 804 440, 448, 440, 535, 555, 565, 714
non-modal utterances, 802 past, 678
non-open* yes—no questions, 768 past perfect, 704
non-past tense, 678 path*, 699, 720
non-proximal*, 650, 651 perfect, 712
non-proximity, 647, 668 perfect in English, 689
non-remoteness, 848 perfective, 705, 712, 714
non-restrictive* relative clauses, 760 performance*, 398, 572, 586
non-transformational*, 394 performance phenomena, 586, 587
Nootka, 430, 432, 433, 441 performative*, 511n, 729, 793
normative grammar, 380 performative analysis, 778, 779, 780,
noun, 425, 426, 441, 448, 637, 657 781, 784
noun expressions, 425 performative formula, 742, 743
nucleus*, 430 performative negation*, 770, 776
number*, 386, 426 performative* utterances, 726, 727,
numeral classifier, 461 : "143, 744
performative verb, 729, 736, 738, 740,
object*, 414 : 742, 769, 805
objectification*, 806 perlocutionary* acts, 730
objective* epistemic modality, 797, perlocutionary effect*, 731, 732, 827
811, 844 840, 843
798, 799, 800, 801, 803, 805, 808, = permission*, 773, 836, 837, 838, 839,
objective epistemic necessity, 845 permission-granting utterances, 838,
objectively modalized statement, 814, 840, 843
833 permissive mood, 839
objectively modalized utterance, 799 person*, 575, 638
objective modality, 804, 805 personal pronouns, 641, 642, 643
objective possibility, 814 person-deixis, 643, 645, 646
objective probabilities, 798 petrification, 536, 543, $47
obligation, 829, 830, 831, 832, 833, phase*, 710
834, 836, 838, 843 phatic, 740
 Index of subjects 889
phonemes, 373 primary illocutionary force, 785,
phonetic meaning, 608 841
phonic medium, 726 primary performatives, 728, 736, 743,
phonological representation, 409 778, 779, 781
phonological rules, 412 process*, 483
phonological stress, 654 progressive aspect, 712
phonological structure, 373 progressivity*, 707, 719
phonology, 374, 411 prohibition, 766, 774, 803, 826, 827,
phrase-marker*, 390 832, 837, 838, 840
phrase-structure*, 389, 390, 401 projection-rules*, 410
phrase-structure grammars, 394 pro-locatives*, 675, 676
phrastic*, 749, 750, 752, 811, 813, pronominal, 659
814, 825, 833 pronominalization*, 662
place-referring expressions, 676 pronominal reference, 673
pluperfect in English, 689 pro-nominals*, 658, 675
point-of-observation, 701 pronoun, 637, 640, 657, 659, 664
point-of-reference*, 671, 700, 703 pro-noun*, 658, 675
polarity, 695 pronouns-as-deictics analysis, 675
polysemous lexemes, 553, 583, 620 pronouns-as-variables* analysis, 673,
polysemy*, 398, 401, 403, 543, 551, 674
552, 553, 554, 555, 561, 563, 564, pronouns-of-laziness*, 674, 676
565, 566, 567, 569 proper names, 640, 641, 645
Portuguese, 551, 717 property, 439, 442
posing*, 755 property of a place, 693
positional adjectives, 701 proposition, 644, 668, 723, 724, 771,
positional* relations, 694 772, 791, 793, 795, 843
positive jussive sentences, 774 propositional negation*, 771, 773
possession, 722 prosodic contour, 373
possessive complement, 473 prosodic features, 399, 624
possessive constructions, 722, 723 prosodic modulation*, 738
potential mood, 847 prosodic patterns, 399
pragmatics*, 591, 603 prosodic signal-information, 398
Prague school, 475, 506, 507, 509, prosodic structure, 634
621 pro-temporals*, 676
650, 772 pro-verb, 659 :
predicate*, 414, 428, 470, 500, 649,. Proto-Indo-European, 815, 818
predicate-negation, 837 pro-verbal, 659
predicating* expressions, 500 proverbs, 681
predication, 393, 438 province*, 582, 584, 585
predicative locative structures, 701 proximal*, 650, 651
predicative negation*, 773 proximal demonstrative, 669
predicative structures, 469 proximity, 647, 668, 669
predicator*, 434 psychological subject, 507, 508
prediction, 817 punctuated*, 624
preparatory conditions, 733, 737 pure textual deixis, 668
present perfect, 704
present imperfect, 704 qualifies*, 391
present tense, 704 qualitative adjectives, 447
presumptive mood, 847 quality-denoting expressions, 449
presupposed*, 503 quantification, 724
presupposition*, 757, 509, 762, 598, quantifier*, 453, 454, 455, 458, 464,
600, 603, 604, 605, 606, 794 465
 890 Index of subjects
Quasi-English, 461, 462, 640, 641, semantic change, 620
642, 643, 648, 651, 686, 761 semantic identity, 413
questions, 748, 755, 767 semantic information, 594
semantic interpretations, 422
rank*, 439 semantic marking, 651
real* conditional statements, 796 semantic meaning, 608
recursion*, 388, 389 semantic parallelism, 407
redundancy*, 586 semantic relatedness, 553
reference, 393, 438, 645 semantic representation, 409
reference to the future, 817 semantic rules, 412
referential ambiguity, 398 semantics, 410, 411, 413, 423, 549,
referential function, 641 550, 591, 605, 608, 610, 614, 616
referential identity, 663 semiotic functions, 590
referential indices*, 663 sensorimotor intelligence, 482
referring expression, 642, 648 sentence-adverbs, 451
reflectiveness*, 657 sentence-fragments*, 589, 625, 649
reflexive pronouns, 665, 667 sentence-meaning, 643
register*, 584 sentences, 387, 397, 590, 623, 629,
regularization*, 586, 587 750
relatedness of meaning, 552 sentence-schemata, 469, 475
relativism, 423 signal-information, 744
replication*, ‘741 simple* lexemes, 521
request, 748, 749, 751, 766, 767, 768, simple proposition, 769
838 simple sentence, 467, 633
respond*, 762 simple utterance-units, 635
restricted obligation, 828 sincerity condition, 733, 734, 737;
restricted X-question*, 762 785
restrictive* relative clauses, 760, 761 single modality, 808
rheme*, 506, 598, 740 single speech-acts*, 635
rising intonation, 759 single text-sentences, 635
role*, 574, 576 situation*, 483
role-relations, 584 situation-of-utterance, 780, 804
Romance languages, 577, 647 Slavonic languages, 577, 705, 709
root*, 556 social* information, 725
rules, 549 social roles, 574, 576, 641
Russian, 404, 427, 454, 455, 488, sociolinguistics*, 585
508n, 547, 575, 688, 689, 705, 706, _—sortal* classifier, 463
707, 708, 709, 714, 716, 717, 758, sound-laws (Lautgesetze), 620
763, 777 source, 720
Russian imperfective, 709 source* of locomotion, 694
Russian perfective, 713 Spanish, 478, 581, 669, 707, 717, 748,
819n
Sanskrit, 816, 818, 819 spatial deixis, 669, 718
| Saussurean view, 620 spatial location, 719
Scots English, 617, 618 spatial organization, 718
second-order entities*, 443, 444, 445, | spatiotemporal co-ordinates, 570
843 specificity of reference, 457
second-order nominals*, 446, 493 speech-acts*, 484, 590, 725, 726, 728,
second-person pronouns, 638 735, 736, 737
selection restrictions*, 414, 421, 539 spoken texts, 624
semantically based*, 414 Sprechakt, 726n
semantically well-formed, 578 standardization*, 587, 588, 615, 742n
 Index of subjects 891
state*, 483 syntactic equivalence, 555, 560, 562
state-description*, 687 syntactic parallelism, 407
statements*, 725, 729, 746, 748, 751, syntax*, 374, 375, 410
767, 773, 828, 832 synthetic* propositions, 787
statements of future requirement, system-sentence*, 387, 388, 419, 588,
846 591, 622, 624, 625, 626, 629, 630,
static relations, 694 631, 632, 633, 634, 635,643, 663,779
static situation, 707, 713
stative non-perfect, 715 tag*, 749, 766
stative perfect, 715 tautologies*, 416, 417, 787
stative verbs, 449, 706 temporal deixis, 718
stativity*, 706, 707, 719 temporal indices, 683, 813
status*, 574, 576 temporal location, 719
stem*, 513n, 556 temporally restricted propositions,
strategies, 549 683
| strings*, 387 680 |
stress, 374, 535, 598 temporally-unrestricted propositions*,
structuralism, 620 temporally-unrestricted directives,
structure of utterances, 778 831
style*, 580, 581, 613, 614 tense*, 386, 436, 638, 677, 678, 682,
stylistics*, 549, 585, 613, 614 687, 705, 718, 790, 809, 820, 821
stylistic variation, 412, 622 tensed*, 679, 682
subject*, 414, 500, 649, 650 tense-distinctions, 719
subjective epistemic modality, 776, tensed proposition, 684, 686
797, 798, 800, 801, 803, 805, 806, tenseless*, 679, 682, 686
808, 816 tenselessness of propositions, 679
subjective epistemic possibility, 847 tenselessness of sentences, 679
subjectively modalized remote tense-logic, 810, 821
possibility, 819 tense-systems, 719
subjective modality, 644, 804, 841 text*, 611, 622, 623, 631
subjective modalization, 799 text-sentence*, 385, 387, 388, 590,
subjectivity*, 739 622, 623, 624, 625, 626, 627, 628,
subject-matter, 581 629, 630, 632, 633, 634, 643, 649
subjunctive, 817, 848 textual deixis*, 667, 668
subjunctive of will, 840 Thai, 577
subjunctive mood, 817 thematic* structure, 500, 508, 509,
subjunctive sentences, 752 601
subordination, 782 thematic subject, 504, 510
subordinative*, 39% thematization*, 500n, 507
substitution, 660 theme*, 506, 598
suppletion*, 377, 567 thesaurus*, 583
suprasegmental* features, 398 third-order entities*, 443, 444, 445,
surface structure*, 394, 403, 412, 782 843
surface-structure ambiguity, 403 third-order nominals*, 446
Swahili, 716, 717 third-person imperative, 747, 751
synchronic* analysis of languages, 551 ~=third-person pronouns, 638, 639,
synchronic dialectal variation, 621 646, 654, 658
synchronic linguistics, 620 three-dimensional first-order entities,
syntactically based*, 414 ' 693
syntactically well-formed, 376 time-bound* proposition, 680, 682,
542 timeless, 682
syntactic compounds’, 535, 536, 539, 687
892 Index of subjects
timeless* proposition, 680, 686 utterance-signal, 387, 633, 643
token-reflexivity, 781, 782 utterance-token, 570
topic*, 50r utterance-type, 570, 731
topicalization*, soon utterance-units*, 633
transferred negation*, 7°75
transformational ambiguity, 402, 403 valency*, 434, 481, 486, 695, 707
transformational* grammar, 394, 415, valency roles, 497
467, 615 valency-schemata*, 491, 492, 493,
transformational rules, 411, 412 494, §11, 720
transitive—intransitive distinction, valency-sets*, 487
793, 842 verb, 425, 426, 432, 448
transitivity, 511, 721 verbal expressions, 425
translation, 643 verbals, 426, 432
tree-diagrams*, 390 verb-expressions, 425
tropic*, 749, 750, 751, 752, 771, 773, | verbs of evaluation, 532
802, 811, 813, 825, 833, 840, 848 verbs-of-motion, 494
truth, 724, 822 verticality, 702, 703
truth-conditional* theories, 597, 611 visual perception, ‘700
truth-value, 644 vocative expressions, 642
truth-value gaps*, 600, 602 vocative* function, 575, 641
Turkish, 568, 6609, 670, 688, 689, voice-quality, 617, 741
705, 716, 763
two-place deontic operator, 834 well-formedness*, 3°79
two-place deontic predicate, 836 well-formed sentences, 391, 520
type-token identity, 509, 512, 559 well-formed texts, 630
Tzeltal, 460, 461, 462, 463 western grammatical tradition, 501
western philosophy, 793
understanding, 733 Whorfian hypothesis, 609
ungrammatical utterances, 384 wishes*, 795
uniqueness of reference, 655 word, 426
universalism, 423 word-formation*, 522, 565
universal quantification, 717 word-forms, 398, 555, 567
universe-of-discourse*, 508, 510, word-order, 598
670, 675, 676 Worter-und-Sachen movement, 620
unmarked*, 503, 647
unreal* conditionals, 795, 796 X~questions*, 757
unrestricted* prohibition, 828
utterance, 387, 397, 570, 633, 635, yes—no* questions, 747, 754
643, 779, 830
utterance-act, 668 zero-derivation*, 522
utterance-meaning, 643 zeugma*, 405, 407
 Index of personal names

Abraham, 488n Bates, 482n, 648n

Adams, 534n Bauman & Sherzer, 574n
Agricola, 409 Bazell, 380n, 386n, 549n
Ahlqvist, 696n Bellert, 755
Ajdukiewicz, 439n Bennett, D. C., 694, 700, 707n, 720
Al, 386n Bennett, J., 735n
Allan, 465 Benveniste, 636n, 646, 688, 739
Allen, 722 Berlin, 460
Allerton, 589n, 629n Berlin & Kay, 440
Allerton & Cruttenden, 452n Bierwisch, 420, 422n, 690, 703
Anderson, 488n, 695, 715, 718, 719, Bloomfield, 389, 449, 514n, 533n,
721, 722, 723 629, 630, 648, 659
721 Boas, 433
Anderson & Dubois-Charlier, 488, Boadi, 722
Antinucci, 636n Bolinger, 386n, 422n, 450n, 506n,
Aqvist, 753n, 754, 762 766
Apollonius Dyscolus, 636 Bonomi & Usberti, 422n
Apresjan, 486 Botha, 422n, 515n
Argyle, 587 Bowers, 450n, 452n
Aristotle, 704, 711, 814 Boyd & Thorne, 751, 778, 819, 846n
Asher, 471n, 722 Braine, 435
Ashton, 716 Bréal, 619
Augustine, St, 811, 812, 815 Brekle, 534n
Austin, J. L., 725, 726, 727, 728, 729, Bright, 574n
73°, 7325 733) 734, 735,739, 737,739, Brooke-Rose, 548n
74°, 743, 745, 746, 779, 780, 827 Brown, 482n, 495, 777
Ayer, 727 Brown & Browne, Ford,763576
Babcock, 489 Bruner, 648
Bach, 508n, 667n, 753n Bihler, 636n, 659, 672, 726n :
Baker, 753n Bull, 683, 688, 717n
Banfield, 741n Burton-Roberts, 653n
Bar-Hillel, 386n, 439n, 583, 611, 849
Barthes, 672 Callow, 510
Bartsch, 452n Carnap, 572, 797, 798
Bartsch & Vennemann, 422n Carter, 466
894. Index of personal names
Castafieda, 645n, 667 Fann, 745
Caton, 594 Fauconnier, 667n
Chafe, 422n, 550n, 664n Ferguson, 580
Chao, 461 Ferreiro, 705 }
Chapin, 534n Fillmore, 422n, 488n, 500n, 603, 604,
Chatman, 614 636n
Chomsky, 385n, 386, 386n, 387, 388, Firbas, 506, 509
390, 392, 394, 395, 402, 403, 409, Firth, 438n, 573, 607, 608, 609, 610,
410, 412, 413, 414, 415, 416, 422n, 612, 735
428, 467, 500n, 503, 504, 505, 506n, Fishman, §74n, 580, 581, 582, 584,
526, 534, 571, 585, 586, 598, 610, 585
614, 615, 629, 662, 663, 716 Fodor, J. A., 483n, 489
Christie, 4’771n, 722 Fodor, J. D., 422n
Christophersen, 64’7n Franck & Petéfi, 606n
Clark, E. V., 648n Fraser, B., 550n, 737n
Clark, H. H., 648n, 690, 691 Frege, 596, 750
Close, 716 Frei, 636
Cohen & Margalit, 548n Friedrich, 462, 465, 511n, 576, 703,
Cole & Morgan, 745n, 784n 707, 709, 714, 717N
Collinson, 636n Fries, 438n, 629n
Comrie, 704, 706, 707, 709, 710, 711, Fromkin, 386n
714, 715, 717, 719
Cooper, 596, 602, 605, 606n Gale, 682, 812, 820, 823
Cresswell, 452n Galmiche, 422n
Cruse, 483n Garner, 596, 606n
Crymes, 659, 672 Gazdar, 594n, 606n
Crystal & Davy, 580, 582, 584, 614 Geach, 666, 674
Culler, 673 Giglioli, 574n, 745
Givon, 450n, 483n
Dane, 506 , Gleason, 514n
Davidson, 483n Goffman, 574n
De Lemos, 717n Gonda, 818n
Dik, 452n Gordon & Lakoff, 748, 784, 785
Dionysius Thrax, 636 Greenbaum, 452
Dixon, 466 Greenberg, J. H., 461, 464, 465, 511n
Dougherty, 422n, 667n Greimas, 703
Dowty, 489, 712 Grice, 574, 592, 593, 594, 606, 608,
Drange, 422n 732, 735, 785
Dressler, 630 Grootaers, 577
Dubois, 534n Gross, 561
Dubois & Dubois, 512n Gruber, 5oon, 511n
Dubois-Charlier & Galmiche, 422n Guilbert, 534n
Ducrot, 594, 606n, 739, 742, '745n Gumperz & Hymes, 574n, 580 7
Ducrot & Todorov, 673, 688
Dummett, 750 Haas, M. R., 466, 577
Haas, W., 386n, 523, 549n
Ellis, 607n Habermas, 745n
Ellis & Boadie, 471n Halle, 524n
Emeneau, 461 Halliday, 439n, 493, 497, 500, 506n,
Empson, 409 507, 508, 607n, 611, 632, 747N, 792,
Ervin-Tripp, 576 : 806n, 807, 848
Faerch, 648n Halliday & Hasan, 623, 631, 632
 Index of personal names 895
584 416, 421, 573
Halliday, McIntosh & Strevens, Katz & Fodor, 410, 411, 412, 415,
Hamblin, 753n Katz & Postal, 410, 412, 415, 416,
Handford, 818n 614, 615, 757
Harada, 577 Keenan, 600, 772
Hare, 730, 749, 750, 753, 755, 769, Keenan & Hull, 598n, 754n
S11 Kempson, R. M., 422n, 596, 601,
633 Kenny, 483n, 711
Harris, 395, 410, 438n, 467, 505, 612, 606n
Hartvigson, 452n Kholodovié, 483n, 486, 487n
Hasan, 529 Kiefer, 471n
Hasegawa, 422n Kiparsky & Kiparsky, 599, 794
Haugen, 582 Kirkwood, 586n
Heger, 486, 717n Klein, 717n
Heinrichs, 647n K6nig, 450n
Helbig, 486 Koolj, 409
Heringer, 748, 785 Kramsky, 647n
Hill, 386n Kratochvil, 448
Hilpinen, 832 Krause, 466
’ Hintikka, 456, 754 Kristeva, 672
Hirtle, 717n Kuno, 475, 506, 508n, 665n, 667n
Hjelmslev, 411, 439, 636n Kurylowicz, 534n, 636n, 792, 818,
Hockett,
Hoier, 466
501, 820 819
Householder, 375, 386n, 422, §34n, Labov, 621
74’7n, 754, 778, 818n, 848 Lakoff, G., 386n, 413, 422n, 603,
Householder & Saporta, §12n 792Nn
Huddleston, 386n, 439n, 488n, 506n, Lakoff, R., 457, 594, 778, 819n
~g2n, 813n Lakoff & Ross, 489
Hudson, 515n, 554, 754n, 765 Langacker, 667n
Hull, 598n, 754n, 762 Langendoen, 606n
Hymes, 386n, 574n, 580, 581, 586, Laver, 587
745n Leech, 420, 422n, 536, 580, 584, 694,
696, 698, 700, 706, 846n
Inoue, 665n Lees, §34n .
Isard, 671, 673, 796n Lees & Klima, 662, 667n
Lehiste, 471n
Jackendoff, 411, 421, 422n, 452, 464, Lehrer, 452n, 550n
500n, 66an, 667n, 775 Leibniz, 787, 791
Jakobson, 636n Lemmon, 725
Jenkins, 481 Legniewski, 439n
Jespersen, 439, 450, 610, 757, 773; Levelt, 386n
847n Lewis, 439n, 732n, 744, 753, 778,
Jessen, 690n, 720 796n
Johanson, 717n Li, 471n, 722
Joos, 438n, 580, 710, 820n Lightfoot, 819n
Linsky, 645n
Kahn, 446, 471n, 473 Lipka, 534n
Karttunen, 604, 606n, 673 Ljung, 450n, 534n
Kasher, 629n, 747n Loewenberg, 548n
Kastovsky, 483n, 489 Lyons, 439, 471n, 502, 607n, 648n,
Katz, 386n, 422n, 606n, 757, 758 653, 673n, 707Nn, 722, 723, 747N
 896 Index of personal names
Makkai, 5son Ross, J. R., 667n, 738, 778, 784n, 792
Malinowski, 607, 609, 735 Russell, 472, 473, 596, 600, 601
Marchand, 534n Russell & Whitehead, 750
Martin, 577 Ryle, 711
Matthews, 376n, 524, 564
McCawley, J. D., 386n, 422n, 554, Sadock, 778, 781, 786
603, 663, 667n, 716 Sampson, 386n, 422n
McCawley, N., 665 Sandmann, 429n, 506, 511
McIntosh, 548n, 607n Sankoff, 581
Miller & Johnson-Laird, 438n, 648n, Sapir, 429, 430, 432, 449
690, 710, 718, 720, 724 Saussure, 619
Mitchell, 607 ._ Schopf, 717n
Montague, 439n Searle, 726, 732, 733, 735, 736, 737,
784, 785
Nelson, 482n Sgall, 506
Newmeyer, 435n Shaumjan, 450
Nida & Taber, 510 Shibatani, 489
Nilsen, 488n Shopen, 589n
Sinclair, H., 482n

846n Skinner, 745

Palmer, F. R., 706, 710, 716, 814n, Sinclair, J. McH., 607, 764
Palmer, L. R., 818n Sommerstein, 64’7n
Parsons, 451, 464 Spang-Hanssen, 411
741n Stampe, 744
Partee, 412n, 666, 667n, 673, 674, Stalnaker, 591, 796n
Peirce, 637, 654 Steinitz, 452n
Pettersson, 714 Stenius, 74’7n
Piaget, 482 , Stevenson, 727
Pine-Coffin, 811 Strang, 584
Plato, 793 Strawson, 429n, 443n, 447, 502, 596,
Postal, 422n, 647n, 667n 597, 602, 649n, 680, 732, 737
Pottier, 719, 722 Sussex, 450n
Pride, 574n Swadesh,
Pride & Holmes, 574n, 580
432, 433
Prior, 812, 815 | Tarski, 685
. Thomas, 717
Prior & Prior, 754n, 762 Teller, 696
Putnam, 447, 465 Tesniére, 486
Quirk, 443n, 453, 454, 523n, 714, Thomason & Stalnaker, 451
762, 775 Thorne, 655 Thurneysen, 696n
Reichenbach, 451, 667n Traugott, 682, 718
Rescher, 832
Rescher & Urquhart, 810, 821 Ullmann, 614, 620
Rey, 512n Urdang, 425, 448
Richards, B., 451 Urmson, 738, 739
Richards, I. A., 727
Ries, 629n
Rivero, 819nVachek,
Van 506,630
Dijk, 621
Robins, 376n, 607n, ‘704 Vendler, 445, 450n, 460, 711, 712,
Rohrer, 754n 720, 740, 741
Ross, A., 745 Verhaar, 471n
 Index of personal names 897
Verkuyl, 717n Wilson, 596, 601, 606
Von Wright, 823, 837 Wittgenstein, 727, 735, 812
Wunderlich, 717n, 745n .
Wall, 386n
Waterhouse, 630 Ziff, 386n
Weinreich, U., 548n, 554, 582, 658 Zuber, 606n
Whorf, 816 Zwicky, 741n
| Wienold, 519n Zwicky & Geis, 7’70n
Wierzbicka, 422n, 483n Zwicky & Sadock, 404, 409
 BLANK PAGE

27 October 2010 ma WMA

202385LV00003B/3/A 9 8780521 291866" —