Typological approaches to modality

Ferdinand de Haan

1. Introduction

This chapter surveys the typological literature on modality.1 Studies within

the typological tradition place an emphasis on explaining language structures
through analyzing their function (hence typology is part of the functionalist
approach to linguistics).2 Typological analyses involve cross-linguistic
comparisons and generalizations, for which explanations are sought. Ex-
planations for cross-linguistic generalizations very often are extra-linguistic
in nature. That is, explanations can be drawn from language use, cognition,
and from sociological factors. In addition, explanations may be drawn from
the history of the language (the diachronic dimension). Diachronic explana-
tions play an important role in typological approaches to modality (known
as grammaticalization studies). However, since they are treated in detail
elsewhere (Traugott, this volume), the area of grammaticalization will re-
ceive little attention here.
Most of the typological literature on modality is concerned with its se-
mantic aspects. This is in accord with the general philosophy that linguistic
form can best be explained from language use. Thus, an analysis of the
modal must in sentence (1) below could involve taking into account the
context in which it occurs, the mode of language (for instance, spoken or
written), a comparison with cognate verbs in related languages, a compari-
son with must in earlier stages of English, and possibly even the sociologi-
cal data of the speaker:

(1) He told MSPs that a thorough investigation of the cause was needed
and lessons must be learned for the future. (BBC online, May 13, 2004)

This does not mean that such detail is always required or even rigorously
followed where it would be appropriate. It is merely meant to illustrate the
range of phenomena that can be taken into account when one takes a typo-
logical approach to linguistics.
28 Ferdinand de Haan

Within typology, the area of modality has not received the same level of at-
tention that categories like ergativity or causativity have enjoyed. Although
there are some good typological studies of modality that are older than 20
years, Palmer (1986) is the first book-length work that takes a typological
outlook. Palmer (1986) and (2001) are standard reference works on modality
in a typological perspective. Even though Palmer (2001) is nominally the
second edition of Palmer (1986), there are enough differences in theoretical
outlook between the two books to consider them separate works. There are
works on modality in functional-typological frameworks, such as Functional
Grammar (e.g., Dik 1997) and Role and Reference Grammar (e.g., Foley
and Van Valin 1985), but for reasons of space these will not be discussed in
any great detail. Other typological surveys are Chung and Timberlake
(1985) and Givón (1984). A very good recent work on epistemic modality
in a functional-cognitive framework is Nuyts (2001). There are of course
numerous studies and monographs of modality in a single language or lan-
guage family, and they will be mentioned as is warranted in the general
discussion. Much typological work on modality is currently being done at
the University of Antwerp by Johan van der Auwera’s group, with several
papers on modal categories in the World Atlas of Language Structures
(Haspelmath et al. 2005), e.g., a chapter on Imperative – Hortative systems
(Van der Auwera et al. 2005).
Because of the relative youth of typological studies on modality, there is
as yet no consensus on the proper terminology for modal meanings. For this
reason, a section of the paper is devoted to various proposals for developing
a consistent and cross-linguistically valid set of terminology.
The next section is devoted to a survey of the ways in which modality
can be expressed. As is the norm for typological studies, an element is con-
sidered modal if it has modal meanings (like obligation, permission and
prohibition). This means that there are quite a number of formal modal
elements besides the familiar (from English) modal verb.
The sections thereafter deal with various topics that are currently in
fashion, and that also have relevance for other theoretical frameworks.
They are a discussion on the status of the irrealis, and interactions of mo-
dality and tense and modality and negation.
Another recent development within typology is that of semantic maps,
in which the semantic inventory of a given feature in a given language is
mapped on an abstract representation of that feature. This has been done for
several features (perfect, evidentiality, indefiniteness) and also for modality.
 Typological approaches to modality 29

The chapter closes with some remarks on evidentiality, a category which is

very often considered to be modal. It is treated as such in Palmer (1986),
for instance.

2. Terminology

Part of the typological literature on modality is concerned with the proper

terminology of modality. Over the last couple of decades several different
sets of terminology have been proposed in the literature, often with subtle
differences in meaning. This section surveys the most important termino-
logical debates in the typological literature. It is not meant to be an exhaus-
tive listing of all terms used in modality, of which there are many. (See
Nuyts this volume).
Following the logicians (from von Wright 1951 on), the original divi-
sion in modality is between epistemic and deontic. Epistemic modality, as
in John must have been at home, refers to the degree of certainty the
speaker has that what s/he is saying is true. Deontic modality, as in John
must go to school, deals with the degree of force exerted on the subject of
the sentence to perform an action. This force can come from the speaker
but also from an unspecified third source. This division is used in such
works as Lyons (1977), Palmer (1990 [1979], 2001 [1986]), Frawley (1992),
de Haan (1997), van der Auwera and Plungian (1998, but see below),
Traugott and Dasher (2002), as well as in various grammatical studies of
single languages. Generally, this division requires a separate modality, of-
ten referred to as dynamic modality, to encode ability (and, depending on
the author, often volition as well). An example is the sentence John can
swim, in which one reading of the modal verb can denotes the subject’s
ability to swim.3
While the status of epistemic modality is not in doubt, scholars have
proposed new terminology for the deontic side of the spectrum. The most
influential proposal is that of root modality. This term makes reference to
root (or main) clauses and it is somewhat of a misnomer, since it does not
only occur in root clauses. The term has been around since the late 1960s
(no doubt owing to the rise of generative grammar and its emphasis on syn-
tax), but the first influential study to employ the notion of root modality
appears to be Coates’ (1983) corpus study on the English modals. She re-
jects the term deontic on the grounds that this term primarily refers to the
logical notions of obligation and permission, while modals such as must
30 Ferdinand de Haan

and may have other interpretations as well (1983: 20–21). The term root
covers both deontic and dynamic modality, as defined above. In Coates’
model of modality, modal meanings are gradual, without “arbitrary cut-off
points,” as she says. Her view is that modals have core and peripheral
meanings, and the terms deontic and dynamic refer only to the core mean-
ings, hence her use of a neutral term, root modality (but see Palmer 1986:
103–104 for arguments against the term root).
There is a clear difference between the terms deontic and root modality.
Unfortunately, this difference is somewhat subtle and in many studies the
two terms are used interchangeably. However, in coining and using the
term root modality, linguists can show that there are aspects of modality
that lie outside the traditional domain of modality in logic and lend them-
selves poorly to descriptions in terms of modal logic. The use of a term
such as root modality highlights this aspect of modality (Sweetser 1990).
An influential set of terminology is given by Bybee, Perkins and Pagliuca
(1994), whose work is based on the premise that, in order to understand the
range of modal meanings in a language, one must understand the diachronic
developments of modal elements. They propose, therefore, the following
division of modality (1994: 177):

– Epistemic
– Subordinating
– Agent-oriented
– Speaker-oriented.

The use of the term epistemic is relatively straightforward, since it includes

possibility and probability among the meanings. Another epistemic category
is inferred certainty, which is used when the speaker has good reasons to
believe that the statement is true (an example is There must be some way to
get from New York to San Francisco for less than $600).
Subordinating moods refer to the use of modality in subordinate clauses,
such as concessive (although …) and purposive (so that …) clauses. One
exponent of subordinating moods is the subjunctive (see below).
Agent-oriented modality refers to those cases in which the agent of a
clause is influenced in some way in performing the action described in the
clause: “Agent-oriented modality reports the existence of internal and ex-
ternal conditions on the agent with respect to the completion of the action
expressed in the main predicate.” (1994: 177). Some types of agent-oriented
modality are obligation (there exist external factors that compel the agent to
 Typological approaches to modality 31

complete the action, as in All students must obtain written permission from
the Dean …), necessity (there exist physical conditions, as in I need to hear
a good loud alarm in the morning to wake up), ability (there exist agent-
internal enabling conditions – see above), and desire (there are internal
volitional conditions). A very important type of agent-oriented modality is
root possibility, which is related to ability, but also takes external factors
into account. An example of root possibility is I actually couldn’t finish it
because the chap whose shoulder I was reading the book over got out at
Leicester Square (1994: 178, from Coates 1983: 114). The use of the modal
couldn’t does not denote an internal inability, but rather an inability caused
by the external factor of someone else’s leaving.
Speaker-oriented modality refers to those cases in which the speaker is
the “enabling condition,” i.e., those cases in which the speaker gives some-
one an order or gives someone permission. This type of modality includes
directives (a term from Lyons 1977), imperatives (the command mood, see
below), prohibitions, optatives (see below), admonitions (warnings), and
permissions.
In this framework, agent-oriented and speaker-oriented modality roughly
divide the area of root modality, or deontic/dynamic modality. The deciding
factor in Bybee et al.’s framework is the enabling factor. If it is the speaker,
then we are dealing with speaker-oriented modality; otherwise it is an in-
stance of agent-oriented modality.
Some scholars have sought to refine this framework while keeping the
basic structure intact. Hengeveld (2004) uses the term participant-oriented
modality instead of agent-oriented modality. This is done to include those
cases in which the subject of the sentence is not actually an agent (as in
John needs to be left in peace today, in which the subject, John, has the
thematic role of patient).
Van der Auwera and Plungian (1998: 80–86) make a distinction be-
tween participant-internal and participant-external modality. Participant-
internal modality is more or less identical with dynamic modality as it deals
with ability and need (as in John needs a book). Participant-external modal-
ity is again divided into deontic and non-deontic participant-external mo-
dality. In this view, deontic modality is a subtype of participant-external
modality. It encompasses permission and obligation (either from the
speaker or another source). Non-deontic modality deals with possibility and
necessity. It refers to circumstances wholly external to the situation. An
example is the sentence To get to the station, you can take Bus 66 (1998:
80). In this view, there is no need for a special category of subject-oriented
32 Ferdinand de Haan

modality because it is either subsumed under deontic modality or, in the

case of imperatives, optatives, etc., because it is not considered to be part of
modality. In addition, volition is in their view not part of modality either.
As the authors themselves admit (1998:84), the term participant-oriented is
perhaps too vague, given that a sentence usually has more than one partici-
pant. They consider the term subject-oriented as an alternative but reject it
on the grounds that it is probably not valid cross-linguistically. They cite
the case of the experiencer in Kannada, a Dravidian language, which is not
a subject (based on Bhat 1991).
Another distinction frequently made in typological studies is that be-
tween real and unreal events, or a realis-irrealis distinction. It has been
claimed that there are languages that encode modality this way rather than
in a deontic-epistemic way. This is an important observation and a separate
section is devoted to this distinction.

3. Expressions of modality

In the typological tradition, categories tend to be defined semantically. Thus,

a morpheme is classified as modal if it has a modal meaning (epistemic,
deontic, etc.). This means that modal meanings are expressed by various
morphological, syntactic, and lexical categories. This section is a survey of
the ways in which modality is expressed across languages.

3.1. Modal auxiliary verbs

The best-known means of expression for speakers of English is doubtless

the modal verb:

(2) a. John must go to school.

b. John must be at school.

The modal verb must is used to denote necessity on the subject (deontic
modality) and strong conviction of the speaker (epistemic modality). These
are examples of strong modality (Palmer 1986). English also has expres-
sions for weak modality:

(3) a. John may go to school

b. John may be at school.
 Typological approaches to modality 33

The modal verb may denotes permission (deontic modality) or possibility

(epistemic modality).4
Thus, modal verbs such as must and may are ambiguous between epis-
temic and deontic modality. This is a very frequent cross-linguistic
phenomenon. Ambiguity of modal verbs is found, among others, in the
Germanic, Slavic, and Romance language families in Europe, as well as in
certain languages outside Europe. An example of an African language with
modal verbs is Yoruba, in which the verb lè is used for weak epistemic and
deontic modality, and the verbs gb.d. and ní láti for strong epistemic and
deontic modality (Adewole 1990: 80):

(4) Ó gb` d´ wà nílé.

‘He must be at home.’ or ‘He must be in.’

3.2. Mood

The category of mood is here defined as a morphological verbal category

which expresses the modal value of the sentence. Mood is therefore the
grammaticalized expression of modality, just as, say, tense is the grammati-
calized expression of time. Mood is therefore an obligatory category in
those languages that have it.
The most common moods cross-linguistically appear to be the indicative
and the subjunctive. These moods are found in, among others, the classical
languages (e.g., Sanskrit, Classical Greek, and Latin) but also in the modern
descendants of Latin, in the Slavic languages, and in certain Germanic lan-
guages (notably Icelandic and German). It has also been described for
Bantu languages like Swahili, Native American languages (for instance,
languages from the Algonquian language family), and Australian languages
like Gooniyandi and Mangarayi.
As a first approximation, we can say that the indicative is used to de-
scribe real, factual events, while the subjunctive is used for unreal, hypo-
thetical events (but see below). This can be illustrated with data from Latin
(Hale and Buck 1903). Example (5) shows the use of the Subjunctive5 in
main clauses. Example (5a) shows the potential use of the Subjunctive,
while (5b) shows an example of a wish, the optative use (Hale and Buck
1903: 273, 269):
34 Ferdinand de Haan

(5) a. cuneo hoc agmen disici-as

wedge-ABL this line split-2SG.SUBJ.PRES
‘With a wedge, one could split this (military) line.’
b. sint beati
be-3PL.SUBJ.PRES happy-PL
‘May they be happy!’

Other uses of the Subjunctive in main clauses in Latin are obligations, horta-
tives, yes/no questions, and hypotheses.
In subordinate clauses, the Subjunctive is used after main verbs of hoping,
fearing, volition, and surprise, among others. Examples are shown in (6),
from Palmer (2001: 133 with glosses slightly adjusted):

(6) a. Time-o ne laborem auge-am

fear-1SG.IND.PRES COMP work.ACC increase-1SG.SUBJ.PRES
‘I am afraid that I shall increase my work.’
b. Ut mihi aedis aliquas conduc-at volo
COMP I.DAT house some buy-3SG.SUBJ.PRES wish-1SG.PRES.IND
‘I want him to rent a house for me.’

The Indicative in Latin in used for facts (again from Hale and Buck 1903:
293–294):

(7) a. quid tac-es

why be.silent-2SG.IND.PRES
‘Why are you silent?’
b. quoad potu-it, resist-it
as.long.as can-3SG.IND.PERF resist-3SG.IND.PERF
‘As long as he could, he resisted.’

Given the data from Latin, it is tempting to equate the indicative-subjunc-

tive distinction with the realis-irrealis distinction alluded to in the section
on terminology. This is a frequently held view, but there are good reasons
for keeping the two sets of terminology distinct. As this problem has at-
tracted some attention in the typological literature, a separate section will
be devoted to it.
One objection that can be raised here is the fact that there are languages
which distinguish other moods besides indicative and subjunctive. One
such mood is the optative, which is frequently used to express the semantic
 Typological approaches to modality 35

categories of wishing and hoping. A separate optative mood has been re-
constructed for Proto-Indo-European, and is found in such languages like
Classical Greek and Sanskrit. In Latin, the old Optative and Subjunctive
merged into one Subjunctive mood (Buck 1933: 298–301). An example
from Classical Greek can be found in Sophocles, Ajax 550, cited in Palmer
(2001: 205):

(8) ó: pai génoio patrós eutuxésteros

o child become-2SG.OPT.AOR father-GEN luckier
‘O child, may you be luckier than your father.’

Other languages that have a separate optative category are the Athabaskan
languages. In Navajo, for instance, it is formed by a verbal prefix ó- plus
optionally a sentence-final particle, such as laanaa ‘would that’ (Young
and Morgan 1987: 162). In Slave, the verbal prefix for the Optative is ghu-
(Rice 1987: 548). However, the optative has other uses besides wishing.
The Navajo Optative has potentiality as a possible interpretation (similar to
the Latin example (6a) above). In Slave, the Optative can have additional
meanings, such as necessity and even futurity (Rice 2000: 249–51). This
suggests that we are dealing with a mood in Athabaskan that is not unlike
the Latin Subjunctive and that calling it an Optative may not be accurate. It
shows that one cannot always judge the content of a given category by its
given label.
A further mood which is found in many, indeed most, languages is the
imperative, or the mood which is used to give a direct command. Thus, in
Russian, the Imperative of the verb rabota-t’ work-INF is rabota-j ‘work-
IMP.2SG or, in the plural, rabota-jte work-IMP.2PL.
The imperative differs somewhat from the other moods discussed in that
it is limited to a certain speech situation, namely when the speaker ad-
dresses the hearer directly and gives a direct command. 6 Therefore, it is per-
formative in nature because by uttering the imperative, the speaker gives the
command. It is quite different from deontic modality in this respect. While
it may seem as though the sentence You must go to school is identical in all
respects to Go to school!, they are in fact different. In the construction with
the modal verb the command can come from other sources beside the
speaker (e.g., a person not present in the discourse or even abstract objects,
in this case the law), and can be used on non-second persons (as in He must
go to school). With an imperative the “commander” is the speaker and the
“commandee” the hearer(s). There are syntactic differences as well which
to a degree depend on the language in question. In English, an Imperative
36 Ferdinand de Haan

cannot be in a subordinate clause (*I said that go to school!) but a corre-

sponding sentence with a modal verb can (I said that you must go to
school.). The fact that imperatives are performative has led some scholars
to treat them as distinct from modality proper.
Related to imperatives are jussives and hortatives. When the subject of a
command is not the addressee, but someone not participating in the speech
situation (i.e., 3rd person) or a group to which the speaker belongs (1st per-
son), then we are dealing with jussives or hortatives. This can be illustrated
by the English form let’s, but can be found as part of the verbal paradigm in
many languages (Palmer 1986: 109–11; Van der Auwera et al. 2005).

3.3. Modal affixes

In many languages modality is marked by means of affixes on the verb.

This is, for instance, the case in Turkic languages, Greenlandic Eskimo,
Dravidian languages like Tamil, and many Native American languages. In
(9), some examples from various languages are shown:

(9) a. Tamil (Dravidian: Asher 1979: 170; -laam permission)

avan peeca-laam
3SG speak-PERM
‘He is allowed to speak.’
b. Koasati (Muskogean: Kimball 1991: 200; -sahá:wa probability)
ó:la-fon a í:ya-:sahá:w-ok …
town-ALL go-PROB-SS.FOC
‘She must have gone to town [also possibility].’
c. Turkish (Turkic: Lewis 1967: 125–7; -meli necessitative)
gel-me-meli-siniz
come-NEG-OBLIG-2PL
‘You ought not to come.’

The difference between mood and modal affixes is that mood is an obligatory
category. That is, a speaker of a language like Italian must choose between
the Indicative and Subjunctive, while a speaker of, say, Tamil can choose
not to use a modal affix. This is similar to English, where speakers always
have the option to use a modal auxiliary or not.
 Typological approaches to modality 37

It is not always easy to tell if we are dealing with a mood or with a modal
affix. Many grammars do not make a distinction between the two categories
and indeed many works on modality do not do so either. Nevertheless, the
distinction is worth making because a language can have both moods and
modal affixes. Such a language is Turkish, which besides the Necessitative
morpheme meli also has a Subjunctive mood (Lewis 1967: 132). One way
of telling moods and modal affixes apart might be to consider the degree of
cohesion. If the morpheme can easily be separated from the rest of the verb
(as is the case with the morphemes shown in example (9) above), then we
are dealing with an affix. If we are dealing with an inseparable part of the
verb (as is the case in the Latin examples (7) above), then we are dealing
with a mood. This area needs more research, and it quite likely involves
degrees of grammaticalization.

3.4. Lexical means

There are also less grammaticalized (and more lexical) means of expressing
modality. These means can, for the most part, be found in English, though
some types are better known from other languages.

3.5. Modal adverbs and adjectives

Modality can be expressed by means of adverbs. In English, typical examples

are probably, possibly, necessarily, and maybe. This list is by no means
exhaustive, obviously. Some examples are:

(10) a. John is probably at home.

b. Maybe John is at home.
c. John is supposedly at home.

As mentioned in Perkins (1983: 89), modal adverbs in English are primarily

epistemic in nature.
In many languages these forms have become more grammaticalized, and
sometimes they are the only way to express modal notions. The main ways
of expressing strong modality in Russian, for instance, are with adjectives
(dolen) and adverbs (nado or nuno). The former has a subject in the
Nominative and declines according on the gender of the subject, while the
latter takes a Dative subject and is invariant. An example is shown in (11):
38 Ferdinand de Haan

(11) a. ja dol en idti v voksal

I-NOM must-MASC go-INF to station-ACC
‘I must go to the station.’
b. mne nado idti v voksal
I-DAT must go-INF to station-ACC
‘I must go to the station.’

3.6. Modal tags

Epistemic modality, the expression of speaker’s confidence, can be ex-

pressed with such tags as I think, I guess, and I believe (Thompson and Mu-
lac 1991). An example is in (12) (1991: 313):

(12) It’s just your point of view you know what you like to do in your
spare time I think.

While tags like I think are derived from pure matrix clauses, they behave
more like modal adverbs and they show signs of grammaticalization. Corpus
research shows that the complementizer that is often omitted (in about 90%
of the cases). They can occur at various places in the sentence (initially,
medially, and finally). A further step on this grammaticalization path would
be the use of tags like I think as a pure adverb. This seems to have happened
in certain Creole languages, such as Tok Pisin (Keesing 1988), where this
tag has been simplified to ating with the meaning ‘maybe.’
Modal tags have not received very much attention in the literature, but it
would seem that they are also mostly epistemic in nature. An exception
might be volition verbs, like want. Although English is not a good example
(the verb want has a different clause structure), something like this seems to
have happened in the history of Greek. The Future particle tha in Modern
Greek derives ultimately from the construction thelo na ‘want-1SG that.’
Volition is a dynamic modal category, not an epistemic one, but this example
shows that matrix clauses with a 1SG subject have a tendency to become
tags, and then adverbs or particles.
 Typological approaches to modality 39

3.7. Modal particles

Another common means of expression is the modal particle. This is a

method not very common in English. Modal particles are rarely found in
British English but seem to become more popular in American English. An
example is (13), where the words too and so function as “contrary to what
you said/thought” modal particles (from the WWW):

(13) a. Affective is too a word!

b. There is so a Santa Claus!

Modal particles are well known from Germanic languages like Dutch and
German (Abraham 1991), where they are ubiquitous. Two examples from
German are shown in (14). As can be seen from the translations, it is not
always easy to give an adequate rendering of modal particles in languages
that lack them. Sometimes the best solution is to just leave them untrans-
lated, as in (14b), where the particle doch can best be translated with a
slightly irritated intonation (Abraham 1991: 335, 340):

(14) a. Kommt er denn

come-3SG.PRES he MP
‘Will he really come?’ or ‘Will he come after all?’

b. Gib mir doch den Löffel

give-IMP me-DAT MP the-ACC spoon
‘Give me the spoon!’

One problem is that it is not always clear whether we are dealing with an
adverb or a particle. This is a poorly researched area and it is likely that any
distinction between the two must be made on a language-by-language basis.
For instance, it has been claimed that modal particles in Dutch can never
receive stress, nor can they occupy the first position in a sentence (Geerts et
al. 1984: 891). Modal adverbs can have stress and occupy sentence-initial po-
sition. However, given the English data shown in (13), in which the particles
do have stress, this may not be a cross-linguistically useful diagnostic.
The nebulous status of modal particles is at least partly due to their ori-
gins. German and Dutch modal particles can derive from a number of
sources: Abraham (1991: 332) lists modal particles that come from adverbs,
adjuncts, scalar particles, adjectives and interjections. The typological con-
40 Ferdinand de Haan

cern addressed by Abraham is why languages like German and Dutch have
a multitude of modal particles while closely related English has none (or at
least very few). Abraham’s (1991) view is that the peculiar syntax of German
and Dutch is responsible. Modal particles occur mainly between the verbal
elements of the sentence; the verbs bracket the part of the sentence where
modal particles typically occur. English (as well as the Romance languages)
lack such a sentence part and hence lack the possibility of developing
modal particles. There are problems with this analysis, however. Languages
like Russian have a number of modal particles, yet word order is very free.
The same goes for the older Indo-European languages (like Classical
Greek), which teem with modal particles but whose word order is very free,
or at least lack a German-type sentence structure.
Because modal particles have the entire sentence in its scope, they are
often found at clause boundaries, and very often clause-finally. Cantonese
(Matthews and Yip 1994: chapter 18) is an example of a language with a
plethora of sentence-final particles, many of which are modal. An example
is given in (15), in which the particle marks a polite request (Matthews and
Yip 1994: 351):

(15) Léih béi d d sìhgaan ngóh l

you give more some time me PRT
‘Give me a bit longer, won’t you?’

3.8. Modal case

The Tangkic language family of Northern Australia provides us with a typo-

logically unusual device for marking modality. In Lardil, Yangkaal and
Kayardild, modality can be marked on the Noun as a case marker.7 In
Kayardild, (Dench and Evans 1988; Evans 1995, 2003) a (non-subject)
noun phrase can have an optional case morpheme, besides a regular case
marker, which denotes mood. An example is shown in (16), from Evans
(2003: 208):

(16) dangka-a burldi-ju yarbuth-u thabuju-karra-ngun u

man-NOM hit-POT bird-M.PROP brother-GEN-INSTR-M.PROP
wangal-ngun-u
boomerang-INSTR-M.PROP
‘The man will/can hit the bird with brother’s boomerang.’
 Typological approaches to modality 41

The morpheme -u, which occurs on all non-subject NPs, is called the Modal
Proprietive and is used to denote future and potential meanings. This mor-
pheme must co-occur with the Potential suffix -ju which is found on the verb.
Besides the modal Proprietive, there are several other modal case mor-
phemes, not all of them used for modal categories as we understand the
term here; some have tense meanings (like the modal Ablative which, used
together with the Past tense morpheme, denotes anteriority). Evans (1995:
chapter 10) is a full discussion of modal case and its development. 8

4. Realis and irrealis

We now turn to some recent developments in the typological literature on

modality. We will start with the realis-irrealis distinction. This distinction
divides the world into real and unreal events or situations. It has been
claimed that there are languages which encode modality in precisely this
way, i.e., there are languages with irrealis morphemes, which mark an action
or situation as unreal.
While languages with irrealis morphemes can be found on every conti-
nent, they have most prominently been described for New Guinea (Roberts
1990; Bugenhagen 1994). Other languages with irrealis morphemes can be
found in North America (Chafe 1995; Mithun 1995). A recent typological
study is Elliott (2000).
One of the major problems in discussing irrealis issues is the fact that
the term irrealis is very vague and can refer to a number of different cir-
cumstances (see Palmer 2001: 149). Furthermore, the semantic content of
irrealis morphemes differs from language to language (even between lan-
guages that are closely related). This makes defining a cross-linguistic
category of irrealis very hard.
One example is the future. It can be argued that future is a prototypical
irrealis category because it refers to events that have not yet happened and
are therefore unreal. In languages like Amele and Muyuw, the future is
indeed an irrealis category. However, in others it is treated as a realis cate-
gory. One such language is the Native American language Caddo (Chafe
1995: 358), shown in (17). The Future morpheme -/ a/ occurs not with the
Irrealis prefix t’a-/t’i- but with the Realis prefix ci-:

(17) cííbáw-/a/ / ci-yi=bahw-/a/

1SG.AG.REAL-see-FUT
‘I will look at it.’
42 Ferdinand de Haan

In yet other languages, the future can be used with either realis or irrealis,
depending on the speaker’s judgement of likelihood that the event described
will actually occur. One such language is Central Pomo, a Californian lan-
guage (Mithun 1995: 378–80). The same goes for other categories that can
be considered part of irrealis, including categories like negation, hypothesis,
and imperative.
Given these facts, it appears that irrealis is a term that is not comparable
from language to language and, consequently, it may be asked whether it is
a useful or even valid object for typological research. This problem is ex-
amined in Bybee et al. (1994: 236–40) and Bybee (1998). These studies
reach the conclusion that the term irrealis is too broad to be of real use. This
is illustrated with data from the Australian language Maung (Capell and
Hinch 1970). In Maung, Irrealis and Realis have the following categories
under them:

(18) Realis and Irrealis in Maung

REALIS PRESENT: Indicative present, future
IMPERATIVE: Negative only
PAST: Simple and complete past, imperfect
IRREALIS PRESENT: Potential, negative present and future
PAST: Negative past, Conditional, Imperative

This can be illustrated with the verb -udba ‘to put’ (Capell and Hinch 1970:
67): 9

(19) REALIS
i-udba ‘I put (pres.)’
i-wan-udba ‘I shall put’
juwunji g-udba ‘don’t put it!’
i-udba- ‘I put it (perfect)’
i-udba--u ‘I was putting’
IRREALIS
ni-udba-ji ‘I can put’
marig ni-udba-ji ‘I don’/won’t put’
da i-udba-nji ‘if I put’
marig i-udba-nji ‘I didn’t put’
g-udba-nji ‘put it!’
 Typological approaches to modality 43

The two problematic cases for treating the morpheme -ji as an Irrealis mor-
pheme are the Future and the Imperative. The positive Future is a Realis
category while the negative Future is an Irrealis one (it is identical to the
negative Present). The positive Imperative is Irrealis while the negative
Imperative (or Prohibitive) is a Realis one, at least in the analysis of Capell
and Hinch (1970). The important point for the present discussion is that, as
is the case with the Latin Subjunctive, there is no correspondence between
Realis-Irrealis on the one hand, and real and unreal situations on the other.
As a further complication, it is very possible for closely related lan-
guages to differ in their treatment of irrealis and realis category. Examples
can be found in various Papuan New Guinean languages (Roberts 1990;
Bugenhagen 1994). It is cases such as these that have led Bybee (1998) to
conclude that these irrealis morphemes actually encompass a range of modal
notions, and there is little, if any, consistency between irrealis morphemes
in different languages. Morphemes that are analyzed as irrealis morphemes
are in fact morphemes that only encode part of the irrealis spectrum. Conse-
quently, the term irrealis should not be used, in this view.
According to Palmer (2001: 145) there are two ways in which irrealis can
manifest itself. Palmer refers to these types as joint and non-joint systems.
In one type of language, the joint type, an irrealis morpheme co-occurs with
another morpheme which encodes the actual type of irrealis. An example is
(20) from the New Guinean language Amele (Roberts 1990: 372). The Dif-
ferent Subject morpheme -eb is used when the main verb has some marker
of irrealis modality, in this case the Future morpheme -an:

(20) ho bu-busal-eb age qo-qag-an

pig SIM-run.out-3SG.DS.IRR 3PL hit-3PL-FUT
‘They will kill the pig as it runs out.’

It might be premature to call this joint-marking because the Irrealis mor-

pheme occurs as part of a Different Subject morpheme, which would have
to be present in any case. A difference between realis and irrealis DS mor-
phemes only shows up in examples like (20) above, where the action in
both clauses is simultaneous. When the action is not simultaneous or when
the subjects in both clauses are identical, there is no difference in realis-
irrealis marking on these morphemes.
The second type, called non-joint, has irrealis morphemes that do not
need other morphemes but function all by themselves. An example is (21),
from Muyuw (Bugenhagen 1994: 18). The Irrealis morpheme b(i)- is used
44 Ferdinand de Haan

here to denote future. It contrasts with the Realis morpheme n(i)-, which is
used for realis events. Note that it is not necessary to choose between either
morpheme. It is perfectly grammatical to omit the Realis and Irrealis mor-
phemes. In that case the modality of the sentence is determined by other
means, for instance by using the particle bo, which denotes certainty.

(21) yey b-a-n Lae nubweig

I IRR-1SG-go Lae tomorrow
‘I will go to Lae tomorrow.’

This distinction is a useful one if one keeps in mind that they are not abso-
lute. The discussion on Muyuw shows that neither the Realis nor the Irrealis
morpheme is obligatory; they can be omitted without impacting the modal
status of the sentence. It is also not possible to classify a language as either
joint or non-joint since both can occur in one and the same language, as
Palmer himself notes (2001: 146).
Given the foregoing, it might be useful to ask again if the Latin Sub-
junctive is a marker of the category irrealis. It has already been mentioned
that there is a correspondence between the two categories, but no absolute
one-to-one match. There are cases in which there is a mismatch between
realis-irrealis and Indicative-Subjunctive in Latin. The Indicative is used
for Future events, for instance, while the Subjunctive does not even have a
Future paradigm (Hale and Buck 1903: 304):

(22) ded-emus ergo Hannibal-em?

give.up-1PL.INDIC.FUT then Hannibal-ACC
‘Shall we then give up, Hannibal?’

Although the future can be considered to be an irrealis category, as it refers

to events that have not (yet) taken place, it is expressed with the Indicative
mood in Latin. The same is true for sentences which contain a negation. As
argued below, negation is also an irrealis category, yet it is very much com-
patible with the Indicative in Latin. There is then a problem in mapping the
term irrealis onto the Latin Subjunctive. While it is appropriate in certain
circumstances, it fails to apply in others.
Should we then give up on the term irrealis, as advocated by Bybee
(1998)? While it may be tempting to do so, it must also be acknowledged
that it fulfills a useful role in linking linguistic categories to the cognitive
domain. There is no doubt that irrealis tends to be used as a vague term for
anything that is even remotely modal. But it is widely used in grammatical
 Typological approaches to modality 45

descriptions. In the literature on pidgins and creoles, it is the standard way

of referring to modal distinctions (cf. Bickerton 1975; Holm 1988: 164–6).
It may be that the difficulties sketched in this section are insurmountable
and that we will have to give up realis and irrealis as linguistic categories. It
may also be possible to retain the terms, and to account for variation within
these categories. One way of doing so might be with the use of semantic
maps, to which we now turn.

5. Semantic map of modality

One of the most powerful tools to analyze the complex interactions of modal
meanings in the world’s languages is a representation that is the sum total of
the semantic possibilities of the category under investigation, what Anderson
(1982, 1986) calls the mental space. An exponent of this category can then
be mapped onto this representation and thus be compared to similar means
of expression in other languages.10
Semantic maps also can be used for both synchronic and diachronic
purposes, i.e., they can be used to chart possible changes in meaning. The
latter is put to good use by van der Auwera and Plungian (1998), who apply
semantic maps to chart meanings and changes in meaning in modal elements.
Their investigation starts with the grammaticization paths of Bybee et al.
(1994), which can be considered to be diachronic semantic maps. One such
path, the path of development from ‘ability,’ is shown in (23 = Fig. 6.3. in
Bybee et al. 1994: 240):

(23) Semantic map of path of ‘ability’

protasis

ability root possibility epistemic possibility comp to think

permission concessive

This path shows the development of a morpheme meaning ‘ability’ (in Bybee
et al.’s terms, this is an ability gram). First, it develops into a marker of
root possibility and then into a gram meaning ‘permission’ or ‘epistemic
possibility’ (or both, as in English may). After this, the morpheme may take
on additional meanings. This path is taken to be universal and unidirec-
tional: a morpheme with an ability meaning will follow the path in (23),
and does not skip a step (i.e., an ability morpheme does not change directly
46 Ferdinand de Haan

into an epistemic possibility gram). This does not mean, of course, that an
ability gram necessarily will take on these additional meanings. Language
is not deterministic and an ability gram can quite happily remain an ability
gram. But if it is going to take on additional meanings, the first one will be
that of root possibility.
Bybee et al. provide other grammaticalization paths for mood and mo-
dality. Their conclusion (1994: 241) is that the overall developmental path
is from agent-oriented modality to the other kinds of modality, with subor-
dinate modality at the end of the path. This corresponds to a development
from less to more grammaticalization. The agent-oriented modal grams are
typically least likely to be bound, while subordinate modal grams are most
likely to be bound. Still, there are problems with this analysis, as the authors
freely admit. For instance, it should be predicted that imperative, a speaker-
oriented modality, is less frequently bound than subjunctive, a subordinate
modality. However, the opposite is the case (1994: 242). Nevertheless, there
is a general correlation between the various modal categories and their
boundedness.
Van der Auwera and Plungian (1998) take these grammaticalization
paths and develop them into a full-blown semantic map. (24 = Fig. 5 in Van
der Auwera and Plungian 1998: 91) shows their representation of Bybee et
al.’s ability path, otherwise in (23), above:

(24) Semantic map of ‘possibility’

 Typological approaches to modality 47

The semantic map is divided in three parts. The main part is the large en-
closed rectangle, which shows the modal domain. To the left side of the
rectangle is the premodal domain, graphically represented by dotted blocks.
This contains the lexical sources for the modal domain. To the right, the
postmodal domain contains further grammaticalization paths for modal
grams, which van der Auwera and Plungian do not consider part of the modal
domain.
Within the modal domain, individual meanings are represented by ovals,
and arrows mark the pathways from one meaning to another. In this repre-
sentation, meanings can be part of other meanings. As mentioned above,
deontic possibility is part of participant-external possibility in this model.
The arrow wholly within the oval for participant-external possibility shows
the pathway to deontic possibility.
By unifying Bybee et al.’s grammaticization paths, plus adding their
own material, van der Auwera and Plungian end up with the following map
for modality (25 =Fig. 19 in van der Auwera and Plungian 1998: 111):

(25) Semantic map of modality

The basic representation is similar to the one shown in (24). Represented

are, from left to right, the premodal, modal, and postmodal domains. The
premodal and postmodal domains are represented abstractly. As before,
grammaticalization paths are denoted by arrows, with the arrows leading
48 Ferdinand de Haan

from premodal to modal and those from modal to postmodal being abstract
representations.
As is readily visible, some meanings are contained in larger enclosures.
For instance, the notions of participant-external possibility and participant-
external necessity are grouped together into a larger meaning block. This
represents the fact that in some languages, these individual meanings are
expressed by one and the same morpheme. An example of this is Swedish
få ‘get,’ which is vague between both types of participant-external modality
(van der Auwera and Plungian 1998: 103, citing Wagner 1976: 56):

(26) Lasse får köra bil

Lasse gets drive car
‘Lasse gets to drive the car.’ ‘Lasse may/must drive the car.’

Similarly, there are languages in which all types of participant modality can
be expressed with one vague construction. Van der Auwera and Plungian
cite the German modal infinitive (sometimes called the modal passive,
which is similar to English constructions like We are to meet there at
seven). It is said to be vague between all types of non-epistemic modality.
By combining diachronic changes and synchronic states, one can get a
good representation of what is a possible interpretation of a given modal
element and what is universally ruled out. The semantic map shown in (24)
is certainly not the last word on the matter: several modal meanings are not
included in this map, such as imperatives. But the semantic map does go a
long way toward a unified picture of modal meanings and forms. 11

6. Modality and tense

In the typological literature tense is commonly defined as the grammatical-

ized expression of location in time (Comrie 1985: 9). There are clear inter-
actions between tense and modality. An obvious candidate for such interac-
tion is the future. Since events in the future have not (yet) happened, it is
easy to see that there is a certain amount of uncertainty surrounding it.
Hence there is a connection between future and epistemic modality. The
future is sometimes classified as a realis category, and sometimes as an
irrealis category.
There is a similar connection between modality and the past tense. This
seems at first to be somewhat counterintuitive since the events happened in
 Typological approaches to modality 49

the past and one can be certain (or, at least, modally neutral) about past
events. Nevertheless, the past tense is routinely used to express notions of
modality.

6.1. Modality and future tense

The fact that future events can be described both temporally and modally is
an observation which goes back a long time. Discussions on the relation
between the two areas can be found in Comrie (1985: 43–46), Dahl (1985:
103, 2000b), Bybee (1988), Bybee et al. (1991, 1994) and Palmer (1986,
2001), among others. These studies currently represent the state of the art in
the typological thinking on the future.12
Future is quite often a part of so-called irrealis morphemes, as discussed
above. Nevertheless, as argued previously, there is no one-to-one correla-
tion between the two categories since there are many languages in which
the future is a realis category (Maung, Caddo, Latin). Ontologically, future
events have not come to pass, and, epistemologically, the speaker cannot
know for certain that the event will occur. Despite these logical uncertain-
ties, there are languages in which these considerations either do not play a
role or in which there is a choice between various future tense forms to
denote various shades of certainty. A language in which the future has been
analyzed as a pure tense (that is, it refers to events occurring subsequent to
the moment of speech without conveying a modal meaning as well) is the
Tibeto-Burman language Manipuri (Bhat 1999: 18–19). In Manipuri, there
is a basic future/non-future distinction. That is, there is one morpheme to
denote future and one to denote present and past tense. In (27), the mor-
pheme -i is used for Non-future (a,b), and -k.ni is used for Future (c).

(27) a. si no m-i

today rain cloudy-NFUT
‘It is cloudy today.’
b. ra no m-i
yesterday rain cloudy-NFUT
‘It was cloudy yesterday.’
c. nudawayrmd no m-gni
evening.LOC rain cloudy-FUT
‘It will be cloudy in the evening.’
50 Ferdinand de Haan

There are languages in which a speaker has a choice between different future
morphemes to denote various shades of certainty. In Bybee et al. (1994:
247–248) several languages are listed in which there are two or more future
morphemes with various levels of confidence. An example is Southern
Agaw, a Cushitic language (Bybee et al. 1994: 248, data cited from Hetzron
1969). The Future certainty morpheme -áGá is used when the speaker is
certain that the action will occur, while -e is the Future possibility mor-
pheme.13

(28) a. táq-áGá
know-2SG-FUT.CERT
‘You will [certainly] know [it].’
b. dngéta a des-é
perhaps tomorrow study-FUT.POSS
‘Perhaps tomorrow I shall study.’

Another connection between modality and future is the fact that quite often
future morphemes develop from modal (deontic) forms. This has happened
in English, of course, where the modal auxiliaries will and shall were origi-
nally modal verbs (see Traugott this volume). The connection between ob-
ligation/volition and future is clear: one can oblige someone to do some-
thing only in the future. Thus, a sentence such as you must go to school
means that the action of going to school is necessarily subsequent to the
moment at which the obligation was uttered. This is a widespread develop-
ment, accounting for the vast majority of cases in Bybee et al. (1994). Simi-
larly, Fleischman (1982) discusses the French verb devoir ‘must, ought to,’
which functions in many respects like a marker of future tense rather than
obligation (1982: 146):

(29) Je dois dîner avec Joseph la semaine prochaine

I must dine with Joseph ART week next
‘I must/will have dinner with Joseph next week.’
‘I am to have dinner with Joseph next week.’

Fleischman comments that the modal verb can be replaced by other future
forms (the go-future or the synthetic future) without much change in mean-
ing.
 Typological approaches to modality 51

6.2. Modality and past tense

While the past tense is usually taken as a pure temporal category, because it
refers to events that are immutable and known, there is nevertheless a con-
nection between the past tense and modality. This can be demonstrated
even with English. The past tense morpheme -ed is usually a pure tense
morpheme, yet in certain environments it can mark various modal mean-
ings, as in the following examples, from Comrie (1985:19):

(30) a. If you did this, I would be very happy.

b. If John was /were here.
c. I just wanted to ask you if you could lend me a pound.

Sentences (30a) and (30b) are counterfactuals and hypotheticals, which are
usually considered to be modal in nature (they are typical irrealis categories).
Sentence (30c) is a polite request and refers to a non-actual event as well; a
request is a type of wish. Is this an isolated fact of English (in which case
examples (30a) and (30b) could be explained away as an instance of homo-
phony: the irrealis forms in (30a) and (30b) are homophonous with the
regular past tense forms), or is this a cross-linguistic feature (in which case
we can look for a principled account of the phenomenon)?
Steele (1975) discusses the relation between past tense and modality
through a reconstruction of part of the tense-aspect-modality system of
Proto-Uto-Aztecan. She reconstructs two morphemes: *ta- as a general
Irrealis morpheme, and *ta- as a Past tense morpheme. She then goes on to
state that both are actually the same morpheme and that there is one ab-
stract feature that underlies both categories. This feature is called dissocia-
tive, as past tense is dissociated from the present, and irrealis is dissociated
from reality. Steele suggests that this observation is valid cross-linguisti-
cally. This view of past tense as a remoteness device has been echoed in
other works, such as James (1982) and Fleischman (1989). Palmer (1986:
211; 2001: 210) considers this line of reasoning circular but does not pro-
vide any real alternative. Bybee (1995: 513–516), which is a paper con-
cerned with the development of the past tense forms of the modals should
and would, rejects the notion that it is the past tense alone that is responsi-
ble for the modal interpretation. In her view, it is the combination of past
tense plus some other element, such as a modal verb, the subjunctive or
(per Fleischman 1995) the imperfective aspect.
52 Ferdinand de Haan

7. Modality and negation

By itself, negation has been considered by some scholars to be part of mo-

dality. The reason for this is that by talking about nonexistent events or
states, we are talking about events or states that are not real. Hence, negation
can be considered an instance of an irrealis category and it is so classified
in some languages as was discussed above. Example (31), from the Native
American language Caddo, is illustrative (Chafe 1995: 355):

(31) kúyt’áybah / kúy-t’a-yi=bahw

NEG-1.AG.IRR-see
‘I don’t see him.’

When the negative prefix kúy- is present, the Irrealis form of the pronominal
prefixes must be used.
However, in other languages, the presence of a negation has no influence
on the choice of realis/irrealis. Latin has already been mentioned in this
regard, and we can add Central Pomo (Mithun 1995: 380–382) and Amele
(Roberts 1990) to this list. This distinction between negation and irrealis
can occur in very closely related languages. Mithun (1995: 383–384) makes
mention of two dialects of the Yuman language Diegueño, Mesa Grande
and Jamul, where the former has an obligatory irrealis morpheme whenever
a negation is present, but in the latter, negation plays no role in the choice
of irrealis or realis.
This divergence of irrealis and negation is not to deny additional close
relations between modality and other semantic factors associated with ne-
gation. Scope relations, for example, between modality and negation have
been known for at least 2500 years. Ever since Aristotle, philosophers and
logicians have been concerned with this issue. It has attracted the attention
of linguists as well, but only comparatively recently.

7.1. Modals, logic and typology

While modal linguistic issues are different from the issues in modal logic,
the representation system and some of the logical analyses shed light on the
natural language problems (Karttunen 1972). (32) shows the basic symbols
mapped on sample sentences from English (see also Kaufmann, Condo-
ravdi, and Harizanov this volume):
 Typological approaches to modality 53

(32) a. John must be a bachelor p

b. John may be a bachelor p
c. John must not be a bachelor ¬p
d. John need not be a bachelor ¬p
e. John may not be a bachelor ¬p (or ¬p)

Two questions concerning the logical structure of modality have been ad-
dressed in the recent typological literature:

– How are modal meanings mapped onto the logical possibilities?

– How do languages disambiguate possible scope ambiguities?

7.2. Mapping of modals onto logical structures

What types of modals are there and can there be? We know there are mo-
dals that express the notions  and , but are there modals that specifically
and uniquely express the notion ¬ or ¬?
In two important papers van der Auwera has studied the connection be-
tween modality and negation. Van der Auwera (1996) investigates the inad-
equacy of the well-known square of oppositions for dealing with modality
in natural language (see, e.g., Horn 1989 on this issue). Van der Auwera
(2001) is an investigation into which combinations of modality and nega-
tion receive a specialized modal element.
In van der Auwera (1996), a distinction is made between two logical
types of possibility. One, known from logic dating back to antiquity, is the
familiar , used in such formulae as (33):

(33) a. p = ¬¬p (‘it is possible that p = it is not necessary that not-p’)

b. p p (‘it is necessary that p’ entails ‘it is possible that p’)

The logical operator  is not always directly translatable into real language
use because it fails to differentiate the uses of the modal may in (34a). This
sentence suggests the equivalency shown in (34b), which is absurd from a
logical point of view:

(34) a. John may be there and he may not be there.

b. p = ¬p
54 Ferdinand de Haan

Example (34a) illustrates what is called the contingent use of the modal
may and not the possibility use. This was first argued by Hintikka (1960)
and also noted by Horn (1989) among others. Hence a new symbol is nec-
essary to distinguish between the two uses. Van der Auwera suggests a new
operator
, which is used for cases like (34a) – the relation between  and

is one of implicature: when something is possible, it is taken for granted
that it is also contingent:

(35) p conversationally implicates
p

There appear to be no languages in which a lexical distinction is made be-

tween weak modals that express possibility and contingency.
Are there specialized modals for every possible logical combination of
modality and negation? This question is addressed in van der Auwera
(2001) and, to an extent, in de Haan (1997). Van der Auwera surveys a
number of languages, mostly European ones, and provides the following
cases (Kashmiri added):

(36) modality language element

¬p English need
Dutch hoeven
Kashmiri lagun
¬p Bengali nei
¬p Russian nel’zja
¬p English mightn’t

That is, every possible combination is attested in at least one language, with
the possible exception of ¬p. The combination form mightn’t is very rare
and not accepted by many speakers (its equivalent form mayn’t is even
rarer and even less accepted). The reason for the rarity of specialized ¬p
forms is possibly the peripheral status of ¬p itself.
A further issue is the status of double modals. It is well-known that
strong modality can be expressed by means of double negation and a weak
modal, to form the logical equivalency of p = ¬¬p. This can be seen in
English constructions like (37):

(37) John can’t not go to school = John must go to school.

 Typological approaches to modality 55

There are languages in which strong modality is expressed only by means

of a weak modal and a double negation. That is, there is no separate strong
modal in the language, and ¬¬p is the only way to express p:

(38) a. Malagasy
tsy main-tsy
NEG able-NEG
‘Can’t not.’ (Horn 1989: 220)
b. Classical Tibetan
mI-V mthu mI-V
NEG-V ability NEG-V
‘Can’t not.’ (Beyer 1992: 247–8)
c. Japanese
Rekishi no hon o yoma-na-kereba nari-masen
History GEN book ACC read-NEG-PROV work.out-NEG
‘I have to read history books.’ (Han 1983: 341)

There is also a logical equivalence between ¬¬p and , but there seem to
be no languages that have a weak modal that is made up of a strong modal
and a double negation.

7.3. Resolving scope ambiguities

How do languages differentiate between, say, ¬p and ¬p and resolve
scope ambiguity in modality and negation? The basic problem can be illus-
trated by considering the sentences in (39). In (39a) the negation is in the
scope of the modal, while the modal is in the scope of the negation in (39b).
The first instance is a case of narrow scope of negation and the second one
of wide scope of negation:

(39) a. John must not go to school. ¬p

b. John need not go to school. ¬p

As is clear from the example, English uses two different modal verbs to
show the difference in scope. The modal verb must is used for narrow
scope, and need is used for wide scope. De Haan (1997) calls this the Modal
Suppletion Strategy for disambiguation.
56 Ferdinand de Haan

Another strategy can be exemplified by a language like Russian, as shown

in (40). The modal verb mo ‘may/can’ interacts with the negation ne in the
following way:

(40) a. Ivan ne mo et rabotat’ ¬p

Ivan NEG can.3SG work.INF
‘Ivan is not allowed/able to work.’
b. Ivan mo et ne rabotat’ ¬p
Ivan can.3SG NEG work.INF
‘Ivan is allowed/able not to work.’

In Russian, the modal verb stays the same but the place of the negation
changes. This strategy is referred to in de Haan (1997) as the Negation
Placement Strategy. It iconically shows the scope relation: if the negation
precedes the modal, it has wide scope, but if it immediately precedes the
main verb, it has narrow scope.
English uses primarily the first strategy, but there are some verbs that do
not have suppletion. As seen in (32e) above, the verb may is ambiguous
when a negation is present (and also the verb can), although there are pro-
sodic differences depending on scope. Another complication is the fact that
languages that use the Negation Placement strategy also tend to have the
phenomenon of NEG-raising, which muddies the iconic relation.14 Since
Russian is one of these languages, sentence (41) can be interpreted with a
narrow scope reading:

(41) Ma a ne dol na rabotat’ ¬p / ¬p

Masha NEG must.FEM work.INF
‘Masha mustn’t/needn’t work.’

8. Evidentiality

The category of evidentiality deals with the source of evidence a speaker

has for his or her statement.15 Evidential morphemes mark whether a speaker
has been a direct witness to the action he/she is describing or whether s/he
has received the information about the action or event from another source.
Thus evidentiality can be divided into two subcategories, direct evidential-
ity, which marks that the speaker was a witness to the action, and indirect
evidentiality, which marks that the speaker was not a witness but obtained
 Typological approaches to modality 57

knowledge about the action from another person (hearsay or quotative evi-
dentiality) or through inference (inferential evidentiality). This is exemplified
in (42), from Evenki (Tungusic; Nedjalkov 1996: 239), where the suffix -re
in (42a) is used for direct evidence, and -che in (42b) for indirect evidence.
The Turkish data in (43), from Slobin and Aksu (1982) and DeLancey
(1997: 37), show the evidential use of the two past tenses, -di to show direct
evidence, -mi indirect evidence (see also (46) below for other uses of -mi):

(42) a. eni eme-re-n

mother come-NFUT-3SG
‘Mother came (direct evidence).’
b. eni eme-che-n
mother come-PST-3SG
‘Mother came (no direct evidence).’

(43) a. Kemal gel-di

Kemal come- PAST
‘Kemal came (direct).’
b. Kemal gel-mi
Kemal come- PAST
‘Kemal came (indirect).’

Evidentiality has long been considered an exotic category, associated with

Native American languages primarily. However, it is a category that mani-
fests itself in languages on every continent, including in some well-studied
languages of Europe, such as German. English lacks a grammatical category
of evidentiality, but it can be expressed with lexical means like evidently,
ostensibly, and with verbs like seem.
Evidentiality is typically considered to be a modal category in the typo-
logical literature. Palmer (1986: 51) distinguishes four kinds of epistemic
modality; one is the degree of confidence in the truth discussed above and
the other three types are evidential categories. In Palmer (2001) evidentiality
and epistemic modality are themselves categories of the hypercategory of
propositional modality. This point of view is found in other typological
studies of modality or evidentiality, such as Bybee (1985), Bybee et al.
(1994), Willett (1988), and Frawley (1992).
The reasoning for treating evidentiality as a modal category is the belief
that one is inherently less certain about actions one has not witnessed than
about those one has witnessed. This belief is anchored in the fact that when
58 Ferdinand de Haan

one sees something with one’s own eyes, one tends to accept that sight as a
true representation of the world while a secondhand report is viewed with
more suspicion. Hence, indirect evidentiality presents the action as less
certain than does direct evidentiality.
While this is an appealing belief, there are some problems with it. First,
it is risky to compare grammatical categories across languages, as the typo-
logical literature has amply shown. This is especially true for notional cate-
gories which are strongly rooted in a subjective environment, like eviden-
tiality and epistemic modality. It is hard to compare modal verbs in related
languages, and comparing English to, say, Evenki, might be extremely
hard. To illustrate this, example (44) shows a typical Dutch sentence with
the modal verb moeten, which is cognate with English must, and yet a
straightforward comparison is not always possible:

(44) Het moet een goede film zijn

It must a good movie be
‘It is bound to be a good movie (epistemic).’
‘It is required to be a good movie (deontic).’
‘It seems to be a good movie (evidential).’

Sentence (44) has three possible interpretations (as always, out of context),
only two of which have must as a possible English translation, (44a) and
(44b). In its evidential reading, must is not appropriate. Thus, even in two
closely related languages there is not always a one-to-one correspondence.
Consequently, how can one assess the status of a sentence like (45), from
Tuyuca, an Eastern Tucanoan language from the Vaupés River region of
Western Amazon (Barnes 1984: 257)?

(45) diíga apé-yi

soccer play-INFER
‘He played soccer (I have seen evidence that he played, but I have not
seen him play).’

A typical analysis is to equate the Inferential in Tuyuca with the English

category of probability (possibly by translating the above sentence as he
must have played soccer), and lump the two together and claim that the two
are synonymous. But this way, the epistemic modality has been introduced
through the back door. Still, at no stage of the game has it been shown that
the Inferential in Tuyuca is a modal category. Indeed, there is good reason to
assume that it is not. Not only does Barnes never use the term (epistemic)
 Typological approaches to modality 59

modal in her description, but there are also true epistemic modals in the
language. By translating an evidential with the modal must, the analysis is
prejudiced at best, wrong at worst.
A recent proposal is to analyze evidentiality not as a modal, but as a
deictic category (de Haan 2001, 2005a). Frawley (1992) already made an
attempt in this direction but called it modal deixis, so still a modal category.
De Haan (2005a) notes the connection between spatial deictic elements
such as demonstratives, temporal deictic elements such as tense, and evi-
dential elements. In all cases, the morphemes in question denote the dis-
tance between the speaker and an object (spatial), time (temporal), or the
entire proposition (evidential). Hence de Haan (2005a) proposes the term
propositional deixis for evidentiality.

9. Mirativity

An offshoot from the research into evidentiality is mirativity, which refers

to the fact that the speaker has received his information from an unexpected
source (DeLancey 1997). The connection with evidentiality is that the two
categories are usually expressed by the same morphemes in those lan-
guages that have both categories. An example is (46), from Turkish (De-
Lancey 1997: 37, citing Slobin and Aksu 1982, but my translations). It has
already been mentioned above that there are two past tenses in Turkish, one
for direct information, one for indirect. The Indirect Past tense -mi can
also be used to mark unexpected information. The context for (46) is that
Turkish Prime Minister Ecevit resigned unexpectedly, whereas President
Nixon’s resignation was widely expected. This warrants the use of the -mi
Past in (46a) and the -di Past in (46b). Note that this parallels the use of the
so-called Hot News Perfect in English (McCawley 1971):

(46) a. Ecevit istifa et-mi

Ecevit resignation make- PAST.MIR
‘Ecevit has resigned!’
b. Nixon istifa et-ti
Nixon resignation make- PAST.DIR
‘Nixon resigned.’

DeLancey notes (1997) that there is a widespread correspondence between

mirative and (indirect) evidential morphemes. He cites examples from Hare,
60 Ferdinand de Haan

Slave, several Tibeto-Burman and Dardic languages, and mentions several

more, from all parts of the world.
Although a thorough analysis of the connection between evidentiality
and mirativity is still to be done, as indeed an analysis of evidentiality and
its proper place in grammar, it is clear that this particular area is currently
one of the more exciting ones in all of linguistics. There is every reason to
think that a convincing analysis of evidentiality and related areas will prove
important for all areas of linguistics.

Abbreviations

ABL ablative MIR mirative

ACC accusative MP modal particle
AG agent M.PROP modal proprietive
ALL allative NEG negation
AOR aorist NFUT non-future
ART article NOM nominative
CERT certainty OBLIG obligation
COMP complementizer OPT optative
DAT dative PAST past tense
DS different subject PERF perfect
DIR direct evidential PERM permissive
FEM feminine PL plural
FOC focus POSS possibility
FUT future POT potential
GEN genitive PRES present tense
IMP imperative PROB probability
IND indicative mood PRT particle
INF infinitive REAL realis
INFER inferential evidential SG singular
INSTR instrumental SIM simultaneous
IRR irrealis SS same subject
MASC masculine SUBJ subjunctive mood
 Typological approaches to modality 61

Notes

1. This paper has benefited from advice by Johan van der Auwera and Sheila
Dooley. Neither of them is responsible for the final product.
2. An excellent recent book on typology is Croft (2003).
3. Many languages make a distinction between physical and mental ability. For
instance, German uses the verb kennen for mental ability and können for
physical ability.
4. In recent years, the field of corpus linguistics has produced several important
studies that deal with modal verbs, ranging from studies that deal with a single
modal verb to those that deal with the entire modal system. English is the lan-
guage best represented here, as should come as no surprise. An early study is
Ehrman (1966), which probably has the honor of being the first corpus linguis-
tic study overall. Other significant studies are Palmer ([1979], 1990) and
Coates (1983), which are still cited. More recent studies are Westney (1995)
and Krug (2000), the latter of which deals mostly with semi-modals. A study
of modality as a whole is Perkins (1983). Other languages are lagging behind
but are catching up. For German, there is Diewald (1999).
5. In this paper I will follow the typological tradition of making a typographical
distinction between language-specific categories (capitalized) and typological
categories (not capitalized). An example is: the English Past Tense morpheme
does not always indicate past tense.
6. The best typological study on the imperative and related areas is Xrakovskij
and Volodin (1986), in Russian.
7. Modality can be marked on the verb as well.
8. Although marking modality on nouns is typologically rare, it does seem to
occur elsewhere. Guy (1974: 35–36) describes the case of the Oceanic lan-
guage Sakao, where (some) nouns may be marked with an irrealis prefix.
Verbs are marked for modality as well.
9. Maung has several verb classes, each with its own allomorphs. The above table
is therefore not typical for all verbs.
10. For a full discussion on the usefulness of semantic maps in typology see Croft
(2003: 133). An easy introduction is Haspelmath (2003). This technique has
been applied to various categories. A small sample of (non-modal) studies is:
the perfect (Anderson 1982), evidentiality (Anderson 1986), voice (Kemmer
1993), case (Croft 1991), coming and going (Lichtenberk 1991), and indefinite
pronouns (Haspelmath 1997). In addition, semantic maps play a prominent
role in Radical Construction Grammar (Croft 2001).
11. Hansen (2001) takes the findings of van der Auwera and Plungian (1998) and
applies them quite successfully to modality in the Slavic language family.
12. There are several important studies on the use and development of the future in
individual languages and language families. Palmer (1990) and Coates (1983)
62 Ferdinand de Haan

include sections on the modal verbs with future reference and Fleischman
(1982) is an in-depth study on the future in Romance.
13. Hetzron calls these two morphemes the Imperfect Definite and Imperfect In-
definite, respectively. The morpheme -áGá has only the one function exempli-
fied in (28a) but -é has other meanings besides future possibility.
14. NEG-Raising refers to the process in which the negation associated with an
embedded verb is moved (“raised”) to a verb higher in the sentence. An exam-
ple from English is the sentence I don’t think that he will come in which the
negation has been raised from the original I think that he will not come. See
Horn (1989) for details of this process.
15. Evidentiality currently enjoys an upsurge in popularity, with several important
studies being released. An early study (in German) is Haarmann (1970). The
classic in the field is Chafe and Nichols (1986), which is a collection of articles
on evidentiality in all parts of the world. More recent books are Johanson and
Utas (2000) and Aikhenvald and Dixon (2003), which are both collections of
articles in the vein of Chafe and Nichols (1986). There is a special issue of the
Journal of Pragmatics devoted to evidentiality (Dendale and Tasmowski
2001). An early typological study is Givón (1982). Willett (1988) is in need of
updating but still useful. A recent book-length treatment is Aikhenvald (2004).
Other important articles are de Haan (1999, 2005a), which explore the limits of
evidentiality and epistemic modality. De Haan (2005b, c) are studies for the
World Atlas of Language Structures (Haspelmath et al., 2005) on the existence
of evidential morphemes in over 400 languages. Barnes (1984) is an article on
the evidential system of the Tucanoan language Tuyuca, which continues to
feature prominently in the field. In addition, there are many individual studies
on “evidentiality in language X” in the literature, which do not always deal
with evidentiality proper.

References

Abraham, Werner
1991 The grammaticalization of the German modal particles. In Ap-
proaches to Grammaticalization, F. Traugott and B. Heine (eds.),
331–380. Amsterdam: Benjamins.
Adéwolé, Fémi
1990 Gbódò ‘must’: Analysis of a Yoruba modal verb. Journal of West
African Languages 20: 73–82.
Aikhenvald, A. Y.
2004 Evidentiality. Oxford: Oxford University Press.
Aikhenvald, A. and R. M. W. Dixon
2003 Studies in Evidentiality. Amsterdam: Benjamins.
 Typological approaches to modality 63

Anderson, Lloyd B.
1982 The ‘perfect’ as a universal and as a language-particular category. In
Tense-aspect: Between Semantics and Pragmatics, P. Harper (ed.),
227–264. Amsterdam: Benjamins.
1986 Evidentials, paths of change, and mental maps: Typologically regular
asymmetries. In Evidentiality, W. Chafe and J. Nichols (eds.), 273–
312. Norwood: Ablex.
Asher, N.
1979 Tamil. Amsterdam: North Holland.
Barnes, Janet
1984 Evidentials in the Tuyuca verb. International Journal of American
Linguistics 50: 255–271.
Beyer, Stephan V.
1992 The Classical Tibetan Language. Albany: SUNY Press.
Bhat, D. N. S.
1991 Grammatical Relations: The Evidence against their Necessity and
Universality. London: Routledge.
1999 The Prominence of Tense, Aspect, and Mood. Amsterdam: John
Benjamins.
Bickerton, Derek
1975 Dynamics of a Creole System. Cambridge: Cambridge University
Press.
Buck, Carl D.
1933 Comparative Grammar of Greek and Latin. Chicago: University of
Chicago Press.
Bugenhagen, Robert D.
1994 The semantics of irrealis in the Austronesian languages of Papua
New Guinea. In Topics in Descriptive Austronesian Linguistics, Ger
P. Reesink (ed.) 1–39. Leiden: Rijksuniversiteit Leiden.
Bybee, Joan L.
1985 Morphology: A Study of the Relation between Meaning and Form.
Amsterdam: Benjamins.
1988 The diachronic dimension in explanations. In Explaining Language
Universals, John A. Hawkins (ed.). 350–379. Oxford: Blackwell.
1998 ‘Irrealis’ as a grammatical category. Anthropological Linguistics 40:
257–271.
Bybee, Joan, William Pagliuca and Revere Perkins
1991 Back to the future. In Approaches to Grammaticalization, E. Traugott
and B. Heine (eds.), 17–58. Amsterdam: John Benjamins.
Bybee, Joan, Revere Perkins, and William Pagliuca
1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Lan-
guages of the World. Chicago: University of Chicago Press.
64 Ferdinand de Haan

Capell, A. and H. E. Hinch

1970 Maung Grammar: Texts and Vocabulary. The Hague: Mouton.
Chafe, Wallace
1995 The Realis – Irrealis Distinction in Caddo, the Northern Iroquoian
languages, and English. In Modality in Grammar and Discourse, J.
Bybee and S. Fleischman (eds.), 349–366. Amsterdam: John Benja-
mins.
Chafe, Wallace and Johanna Nichols (eds.)
1986 Evidentiality: The Linguistic Coding of Epistemology. Norwood, NJ:
Ablex.
Chung, Sandra and Alan Timberlake
1985 Tense, aspect and mood. In Language Typology and Syntactic De-
scription, vol. III. Tim Shopen (ed.), 202–258. Cambridge: Cam-
bridge University Press.
Coates, Jennifer
1983 The Semantics of the Modal Auxiliaries. London: Croom Helm.
Comrie, Bernard
1985 Tense. Cambridge: Cambridge University Press.
Croft, William
1991 Syntactic Categories and Grammatical Relations. Chicago: Univer-
sity of Chicago Press.
2001 Radical Construction Grammar: Syntactic Theory in Typological
Perspective. Oxford: Oxford University Press.
2003 Typology and Universals (2nd edition). Cambridge: Cambridge
University Press.
Dahl, Östen
1985 Tense and Aspect Systems. Oxford: Blackwell.
2000 The grammar of future time reference in European languages. In
Tense and Aspect in the Languages of Europe, O. Dahl (ed.), 309–
328. Berlin /New York: Mouton de Gruyter.
Dahl, Östen (ed.)
2000 Tense and Aspect in the Languages of Europe. Berlin /New York:
Mouton de Gruyter.
de Haan, Ferdinand
1997 The Interaction of Modality and Negation: A Typological Study.
New York: Garland.
1999 Evidentiality and epistemic modality: Setting boundaries. Southwest
Journal of Linguistics 18: 83–101.
2001 The place of inference within the evidential system. International
Journal of American Linguistics 67: 193–219.
2005a Encoding speaker perspective: Evidentials. In Linguistic Diversity
and Language Theories, Z. Frajzyngier, D. Rood and A. Hodges
(eds.), 379–397. Amsterdam: John Benjamins.
 Typological approaches to modality 65

2005b Semantic distinctions of evidentiality. In World Atlas of Language

Structures, M. Haspelmath et al, (eds.), 314–317. Oxford: Oxford
University Press.
2005c Coding of evidentiality. In World Atlas of Language Structures, M.
Haspelmath et al, (eds.), 318–321. Oxford: Oxford University Press.
DeLancey, Scott
1997 Mirativity: The grammatical marking of unexpected information.
Linguistic Typology 1: 33–52.
Dench, Alan and Nicholas Evans
1988 Multiple case-marking in Australian languages. Australian Journal
of Linguistics 8: 1– 47.
Dendale, P. and L. Tasmowski (eds.)
2001 Evidentiality. Special issue of Journal of Pragmatics 33 (4).
Diewald, Gabriele
1999 Die Modalverben im Deutschen: Grammatikalisierung und Poly-
funktionalität. Tübingen: Niemeyer.
Dik, Simon C.
1997 The Theory of Functional Grammar (2 volumes). Berlin /New York:
Mouton de Gruyter.
Ehrman, Madeleine
1966 The Meaning of the Modals in Present-Day English. The Hague:
Mouton.
Elliott, Jennifer R.
2000 Realis and Irrealis: Forms and Concepts of the grammaticalization of
reality. Linguistic Typology 4: 55–90.
Evans, Nicholas D.
1995 A Grammar of Kayardild, with Historical-Comparative Notes on
Tangkic. Berlin/New York: Mouton de Gruyter.
2003 Typologies of Agreement: Some problems from Kayardild. Transac-
tions of the Philological Society 101: 203 –34.
Fleischman, Suzanne
1982 The Future in Thought and Language. Cambridge: Cambridge Uni-
versity Press.
1989 Temporal distance: A basic linguistic metaphor. Studies in Language
13: 1–51.
1995 Imperfective and Irrealis. In Modality in Grammar and Discourse,
J. Bybee and S. Fleischman (eds.), 519–551. Amsterdam: John Ben-
jamins.
Foley, William A. and Robert D. Van Valin
1984 Functional Syntax and Universal Grammar. Cambridge: Cambridge
University Press.
Frawley, William
1992 Linguistic Semantics. Hillsdale, NJ: Lawrence Erlbaum.
66 Ferdinand de Haan

Geerts. C. et al.
1984 Algemene Nederlandse Spraakkunst. Groningen: Wolters Noordhoff.
Givón, Talmy
1982 Evidentiality and epistemic space. Studies in Language 6: 23–49.
1984 Syntax, vol. 1. Amsterdam: John Benjamins.
Guy, J. B. M.
1974 A Grammar of the Northern Dialect of Sakao. Pacific Linguistics
Series B-33. Canberra: Australian National University.
Haarmann, Harald
1970 Die indirekte Erlebnisform als grammatische Kategorie. Wiesbaden:
Harrassowitz.
Hale, William G. and Carl D. Buck
1903 A Latin Grammar. Boston /London: Ginn & Co.
Han, Mieko S.
1983 Modern Japanese (2nd edition). Los Angeles: Institute for Inter-
cultural Studies Press.
Hansen, Björn
2001 Das slavische Modalauxiliar. Semantik und Grammatikalisierung im
Russischen, Polnischen, Serbischen /Kroatischen und Altkirchen-
slavischen. München: Verlag Otto Sagner.
Haspelmath, Martin
1997 Indefinite Pronouns. Oxford: Oxford University Press.
2003 The geometry of grammatical meaning: semantic maps and cross-
linguistic comparison. In The New Psychology of Language: Cogni-
tive and Functional Approaches to Language Structure, vol. 2, M.
Tomasello (ed.), 211–242. Mahwah, NJ: Lawrence Erlbaum.
Haspelmath, Martin, Matthew Dryer, David Gil and Bernard Comrie (eds.)
2005 World Atlas of Language Structures. Oxford: Oxford University
Press.
Hengeveld, K.
2004 Mood and modality. In Morphology: A Handbook on Inflection and
Word-Formation, G. Booij, C. Lehmann and J. Mugdan (eds.).
1190 –1202. Berlin /New York: Mouton de Gruyter.
Hetzron, Robert
1969 The Verbal System of Southern Agaw. Berkeley: University of Cali-
fornia Press.
Hintikka, J.
1960 Modality and quantification. Theoria 27: 119 –128.
Holm, John
1988 Pidgins and Creole, vol. I: Theory and Structure. Cambridge: Cam-
bridge University Press.
Horn, Laurence R.
1989 A Natural History of Negation. Chicago: University of Chicago Press.
 Typological approaches to modality 67

James, Deborah
1982 Past tense and the hypothetical: a cross-linguistic study. Studies in
Language 6: 375–403.
Johanson, Lars and Bo Utas (eds.)
2000 Evidentials. Berlin /New York: Mouton de Gruyter.
Karttunen, Lauri
1972 Possible and must. In Syntax and Semantics, vol. 1, John P. Kimball
(ed.), 1–20. New York: Academic Press.
Keesing, Roger
1988 Melanesian Pidgin and the Oceanic Substrate. Stanford: Stanford
University Press.
Kemmer, Suzanne
1993 The Middle Voice. Amsterdam: John Benjamins.
Kimball, Geoffrey D.
1991 Koasati Grammar. Lincoln: University of Nebraska Press.
Krug, Manfred G.
2000 Emerging English Modals: A Corpus-Based Study of Grammaticali-
zation. Berlin /New York: Mouton de Gruyter.
Lewis, G. L.
1967 Turkish Grammar. Oxford: Oxford University Press.
Lichtenberk, F.
1991 Semantic change and heterosemy in grammaticalization. Language
67: 475–509.
Lyons, John
1977 Semantics. Cambridge: Cambridge University Press.
Matthews, Steven and Virginia Yip
1994 Cantonese: A Comprehensive Grammar. London: Routledge.
McCawley, James D.
1971 Tense and time reference in English. In Studies in Linguistic Seman-
tics, Charles Fillmore and D. Terence Langendoen (eds.), 96 –113.
New York: Holt, Rinehart and Winston.
Mithun, Marianne
1995 On the relativity of irreality. In Modality in Grammar and Discourse,
J. Bybee and S. Fleischman (eds.), 367–88. Amsterdam: John Ben-
jamins.
Nedjalkov, Igor
1996 Evenki. London: Routledge.
Nuyts, Jan
2001 Epistemic Modality, Language, and Conceptualization: A Cognitive-
Pragmatic Perspective. Amsterdam: John Benjamins.
Palmer, Frank R.
1990 [1979] Modality and the English Modals (2nd edition). London:
Longmans.
68 Ferdinand de Haan

Palmer, Frank R.
1986 Mood and Modality. Cambridge: Cambridge University Press.
1995 Negation and the modals of possibility and necessity. In Modality in
Grammar and Discourse to Grammaticalization, J. Bybee and S.
Fleischman (eds.), 453–72. Amsterdam: John Benjamins.
2001 Mood and Modality (2nd edition). Cambridge: Cambridge University
Press.
Perkins, Michael R.
1983 Modal Expressions in English. London: Frances Pinter.
Rice, Keren D.
1987 A Grammar of Slave. Berlin /New York: Mouton de Gruyter.
2000 Morpheme Order and Semantic Scope: Word Formation in the
Athapaskan Verb. Cambridge: Cambridge University Press.
Roberts, John R.
1990 Modality in Amele and other Papuan languages. Journal of Linguis-
tics: 26: 363–401.
Slobin, Dan I. and Ayhan Aksu
1982 Tense, aspect and modality in the use of the Turkish evidential. In
Tense-Aspect: Between Semantics and Pragmatics, P. Hopper (ed.),
185–200. Amsterdam: John Benjamins.
Steele, Susan
1975 Past and irrealis: Just what does it all mean? International Journal of
American Linguistics 41: 200–217.
Sweetser, Eve E.
1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects
of Semantic Structure. Cambridge: Cambridge University Press.

Thompson, Sandra A. and Anthony Mulac

1991 A quantitative perspective on the grammaticization of epistemic
parentheticals in English. In Approaches to Grammaticalization, E.
Traugott and B. Heine (eds.), 313–329. Amsterdam: John Benjamins.
Traugott, Elizabeth Closs and Richard B. Dasher
2002 Regularity in Semantic Change. Cambridge: Cambridge University
Press.
van der Auwera, Johan
1996 Modality: The Three-layered Square. Journal of Semantics 13: 181–
195.
2001 On the typology of negative modals. In Perspectives on Negation
and Polarity Items, J. Hoeksema, H. Rullmann, V. Sánchez-Valencia
and T. van der Wouden (eds.), 23–48. Amsterdam: John Benjamins,
van der Auwera, Johan and Vladimir Plungian
1998 Modality’s semantic map. Linguistic Typology 2: 79 –124.
 Typological approaches to modality 69

van der Auwera, Nina Dobrushina and Valentin Goussev

2005 Imperative-Hortative Systems. In World Atlas of Language Struc-
tures, M. Haspelmath et al. (eds.), 294 –297. Oxford: Oxford Univer-
sity Press.
von Wright, E. H.
1951 An Essay in Modal Logic. Amsterdam: North Holland.
Wagner, Johannes
1976 Eine kontrastive Analyse von Modalverben des Deutschen und
Schwedischen. International Review of Applied Linguistics in Lan-
guage Teaching 14: 49–66.
Westney, Paul
1995 Modals and Periphrastics in English: An Investigation into Semantic
Correspondence between Certain English Modal Verbs and their
Periphrastic Equivalents. Tübingen: Niemeyer.
Willett, Thomas L.
1988 A cross-linguistic survey of the grammaticization of evidentiality.
Studies in Language 12: 51–97.
Xrakovskij, V. S. and A. P. Volodin.
1986 Semantika i tipologija imperativa. Leningrad: Nauka.
Young, Robert and William Morgan
1987 The Navajo Language: Grammar and Dictionary. Albuquerque:
University of New Mexico Press.