In this paper it will be shown that modality may be defined along six parameters. Languages occasionally make further distinctions within these categories, yet since these labels represent grammaticalised 'speaker attitudes' to propositions, not all languages grammaticalise every category. It will also be demonstrated that the relation between irrealis marking and modality in a language follows a hierarchy, and the marking of a particular modality as irrealis may be predicted.

0. Introduction

This paper is concerned with the interaction of irrealis marking and modality. Data from approximately sixty languages were examined to determine what modalities existed in the language, whether there was a category of irrealis and, if so, what modal categories were obligatorily marked as irrealis. The sample included languages from Australia, North America, Polynesia, Europe, North Africa and Asia. Neither sample size nor sample composition were ideal; the size was constrained by time and the composition was determined largely by the adequacy of the language materials available. Nonetheless, it seems that preliminary conclusions such as these may be drawn from the sample, with the warning that many languages are yet to be examined and many exceptions may yet be uncovered.

Previous definitions of modality and irrealis will be discussed (1.1 and 1.2). Section 2 concerns an explication of an alternative definition and delineation of modality, arising from the most frequently occurring grammaticalisations of speaker attitudes to events. Section 3 is devoted to the interaction between modality and irrealis. For every aspect of modality and irrealis included, there have been many which were omitted. Discussion of irrealis, for example, is confined only to modality; no comment can be made on the interaction of irrealis and other grammatical systems.

1.1 Modality

Modality is an area of grammar where there are almost as many definitions of the subject as there are linguists who have worked on it. Modern definitions begin with Jesperson (1955), who divided the uses of subjunctive, imperative and indicative into those meanings which contain an element of 'will' and those which do not. This latter parameter - the element of will - is the basis for later categorisations. E H von Wright (1951), for example, divides modality into the epistemic and the deontic, where deontic refers to commands, wishes and the like, and epistemic to ideas relating to doubt, potentiality and so on. Lyons (1977), who uses von Wright's terminology, says that epistemic modality is 'concerned with opinion rather than fact ... [and] matters of knowledge and belief.' Within the epistemic and deontic modalities several authors give lists of subtypes, such as jussive, advisive, obligative, assertive, presumptive, potential, concessional, apodictive, bouletic, and so on. Others (such as F. de Haan (1996), based on Siewierska (1991)) use a continuum of 'strong' and 'weak' modality within the deontic/epistemic framework.

There are, however, a number of serious problems with these definitions. First, the attempted exhaustive definitions of modalities look rather like a list of Latinate terms for Germanic auxiliaries, and not particularly like a cross-linguistic categorisation of modality. There is not a sufficient distinction between the basic modalities which are found in many languages and the extensions from these basic meanings, which are found very seldom. Evidence from cross-linguistic research would indicate that base meanings and extensions can be quite clearly identified. Moreover, the definition of modality in terms of epistemic and deontic parameters is misleading. As Roberts (1990;368) has found, 'no natural language appears to mark this distinction of epistemic versus deontic modality unambivalently', and the data collected for this survey agree with this view. Such a categorisation may appear convenient, but it does not reflect linguistic reality.

Finally, form is not often sufficiently separated from function when modality is compared cross-linguistically. This leads often to a great deal of confusion. For example, it is often said that Ancient Greek had two modalities (the subjunctive and the optative) where Latin had only one (the subjunctive). Morphologically this is true; however, functionally, it is not, for the same modal functions are expressed in each language. Grounding a definition of modality in function (not form) provides a solution to those who argue that modals, such as the 'subjunctive', cannot be compared cross-linguistically because they mean different things. Comparing the Ngiyambaa potential and the Latin subjunctive leads nowhere; comparing the grammaticalised modal meanings, however, has yielded some interesting results.

1.2 Irrealis

Morphological definitions of irrealis, like those of modality, suffer from a lack of uniformity. Semantically, irrealis refers to 'non-actualised events or states' (Mithun (1995:382)). The irrealis may contrast formally with the realis (such as in Central Pomo (Mithun 1995;368 ff)); in such languages every sentence is marked obligatorily for the category of 'reality' (see note 1). In other languages, however, irrealis cannot be said to contrast with realis. Here irrealis is usually a property of the tense/aspect system (such as in Ngiyambaa (Donaldson (1980)). Like modality, irreality may be signalled by verbal inflection, but this is not necessarily the case.

The signalling of a proposition as realis or irrealis is not confined only to modal aspects of grammar. Irrealis spreads across a number of grammatical categories, and these will be briefly surveyed. In Caddo, for example, interrogatives are irrealis:

1.     sáyybàwnah
       sah?-yi=bahw-nah
       2.AG.IRREALIS-see-PERFECT
       'Have you seen him? '
                         (Chafe (1995;354))

A further example of non-modal irrealis marking comes from the Nyulnyulan languages of Western Australia, where negated clauses are obligatorily irrealis:

2.     arri   i-li-jid-an          bur-ung 
       not    3sgNOM-IRR-go-PERF   camp-ALL

       i-ngkudal
       3sgNOM-got.lost.

       'He didn't go to his camp; he got lost.'
                          (McGregor (1996;26))

Other areas of grammar where irrealis can occur include in subordinate clauses, in reported speech and in the aspectual system, where habitual actions are frequently obligatorily irrealis (as in, for example, Iraqw (Mous (1993)) and Gooniyandi (McGregor (1990)). Such irrealis marking is beyond the scope of this paper and will not be futher discussed.

2. Modality and its limits.

Modality is the semantic category which shows the speaker's attitude to a particular proposition. This category contains a number of semantic types which are frequently grammaticalised. While the number of 'speaker attitudes' which can be expressed are, if not infinite, at least practically uncountable, the number which are grammaticalised are considerable fewer. Reality, however, has been defined here as the formal classification of a proposition or grammatical category as either actual (realis) or unrealised (irrealis), a category which cross-cuts many areas of grammar.

It might be thought that this definition of modality tacitly implies that all modal sentences are irrealis, since modality describes the speaker's attitude to the proposition and not the proposition itself. Indeed, there are systems in which all modality is irrealis (for example, the Nyulnyulan languages mentioned above). Yet not all modalities are marked as irrealis in all languages. So, this investigation into modality and irrealis is a study of what areas of modality are classed as 'actual' and what are 'non-actual'.

The following table provides a summary of this author's modal categories.

Table 1: Modal Categories

          Positive                  Negative
          Potential                      Counterfactual
          Imperative/Obligative          Prohibitive
          Desiderative                   Apprehensive/Aversive

The labels for modalities used in this table (such as prohibitive) are all to be found in previous literature, so no definitions are given here. All the modal meanings of all the languages surveyed can be grouped into these categories. Of course, there are languages with futher distinctions in these categories, and others in which certain modalities are not grammaticalised. An example of the former is the Australian language Yindjibarndi, which has two verbal suffixes in the potential modality, depending on how definite the speaker views the event. Wordick's (1982) glosses are retained. (3) shows the more definite suffix, whereas (4) is less definite.

3.     tharrwayi   ngurrayi    ngarrii  thurnungka.
       enter-POT   ground-OBJ   lie-POT  inside.
       'She enters the ground and lies inside.'
                    (Wordick (1982:226))

4.     Witypanha parniyaamu    yirtiyala?
       Jerry-PN  sit-OPT-ANA    street-LOC.
       'Could Jerry have been sitting by the street before?'
                    (Wordick (1982;103))

Another example of this would be the different levels of imperatives that one finds in some languages. Awa Pit (Colombia), for example, has two sets of imperative affixes; one functionally unmarked which differentiates singular and plural, and the other a polite suffix unmarked for number (Tim Curnow, pc). There are, of course, many further examples.

The modality types in table 1 have been divided into positive and negative. These labels refer to the meanings of the categories, not to their morphological marking, although these may coincide. Thus the prohibitive has the meaning of a negative imperative. In English, it also has the form of a negated imperative - compare "do this" as opposed to "don't do this". In some other languages, however, the two are marked separately. In Yindjibarndi, for example, the imperative and the prohibitive have different verbal suffixes.

5.     nyinta    karrima    murrirni   ngayhala!
       you.NOM   stand-IMP   behind     1ST.LOC.
       'Stand behind me.'

6.     mirta  pangkarrii   mirnayhui!
       not    go-POT       while-DET.
       'Don't go for a while.'
                    (from Wordick (1982:167))

The same applies to the remaining modal categories in table 1; the negative modality might always be the formally negated version of the positive, but is not always so.

More than half the languages sampled marked prohibitives differently from the morphological equivalent of a negated imperative. In the literature, prohibition is not often defined as a modality distinct from obligation; the prevalence of the morphological separation of these modal meanings in language, however, implies that such a treatment is justified.

Not all languages have separate marking for all these categories; nor are all categories represented as irrealis in all languages. For example, in the Nyulnyulan languages all six categories are encoded using the same irrealis suffix, yet in Ao (Naga) each is marked with a distinct suffix (Alec Coupe, pc). In Awa Pit five categories are encoded; the apprehensive meaning, however, is formed lexically and cannot therefore be said to be part of the modality system in this language (Tim Curnow, pc). Since modality is grammaticalised speaker attitudes it is necessary to distinguish between grammatical and lexical realisation.

3.1 Irrealis and Modality and their interaction.

When reality and modality interact, the interaction is not random. The marking of modalities are irrealis follows a certain hierarchy. Two hypotheses can be formulated regarding the marking of modality as irrealis. The first is valid for all languages investigated; the second could be said to be a strong tendency, for an exception has been discovered (although it is an exception for very interesting historical reasons). Modal categories are those defined in table 1.

7.     If a negative modal category is marked by the irrealis,
       then the corresponding positive modal category is also
       marked by the irrealis.


8.     If a category on the hierarchy is obligatorily marked by
       the irrealis in a particular language, then all other
       modal categories which are marked in the language and
       occur higher on the hierarchy are likewise marked by the
       irrealis.

No exceptions have been found to (7). If, for example, the language has an apprehensive/aversive use of the irrealis, then there will also be a desiderative use. This is found in the Nyulnyulan languages of the Kimberley, for example (compare the results in Wagner (1997)). A further example involves the prohibitive: if the prohibitive is irrealis, so will be the imperative or obligative. This may be illustrated from Latin, where both prohibitives and obligatives/imperatives may be expressed using the indicative or the subjunctive. The following sentences are all found in inscriptions on earthenware:

9.a  noli me tanger(e). M. sum.
     do.not.wish-IMP  me touch(INF). of M. I.am.
     'Do not touch me. I belong to Marcus.'

b.   cape me. tua sum
     take-IMP me. yours I.am
     'Take me. I'm yours.'

c.   ne atigas. non tua sum
     no take-2ndp.SUBJ. not yours(FEM.SG) I am.
     'Don't take me. I don't belong to you.'

d.   capas me.
     take-2ndp.SUBJ me.
     'Take me.'
     

The second hypothesis of hierarchical ordering of modality meanings (as expressed in (8)) is summarised in figure 1.

Figure 1.
     Potential    <- Counterfactual
        ^
        |
      Jussive       <- Prohibitive
        ^
        |
     Desiderative   <- Apprehensional

The arrows may be read as 'implies'. So, if a language marks a particular modality for irrealis then it will mark every positive category further up the hierarchy for irrealis also. If a language marks volition with the irrealis, then it will also mark potential events and commands for irrealis. An example of such a language is Yapese (Jensen (1977)).

It does not follow, however, that if a language marks volitive for irrealis it must also have irrealis prohibitions; nor that if there are irrealis prohibitions there are irrealis frustratives or counterfactuals. The negated modality types could be thought of as off-shoots from the main modality hierarchy. A negated type implies not only the corresponding positive modality type, but also the positive modalities further up the hierarchy. It does not, however, imply the negated types above it in the hierarchy. Thus in Ancient Greek volition, prohibition, commands and potential are all expressed by modality markers (either the subjunctive or the optative), but counterfactual statements must be expressed with the indicative, or realis, set of verbal endings:

9.     ei hoi barbaroi epethento,
       if ART.MASC.NOM.PLU Persians.NOM attack.3PL.AOR.INDIC

       machesthai he:mi:n an parekeleusato.
       fight.INFIN 1PL.DAT exhort.3SG.AOR.INDIC.

       'If the Persians had attacked, he would have exhorted
       us to fight (but they didn't so he didn't)'.

10.    ho pais so:phronein manthanoi.
       ART.MASC.NOM.SG child be.wise.INFIN learn.3SG.PRES.OPT.
       'Would that the child would learn to be wise.'

3.2 The exception.

Now, thus far one demonstrable exception to this hierarchy has been discovered, although there are a number of languages where lack of data has prevented conclusions from being drawn, and many, many languages which are yet to be examined. This language is Caddo. Caddo has two sets of pronominal prefixes, one of which is used for interrogatives, negatives, prohibitions, obligations and conditions (Chafe (1995)). This set is irrealis. The other prefixes are used elsewhere, and this elsewhere includes imperatives, future tense and potential modality. This clearly contravenes the hierarchy, since we expect that if there is irrealis marking of modality, if will occur in the potential category if it occurs anywhere.

The explanation for this seems to lie in the historical origins of the marking. Caddo is quite different from the other languages of the same family; it has been heavily influenced by the neighbouring Muskogean languages. It is possible that Caddo could have borrowed (or innovated on a Muskogean model) an alternative set of pronominal prefixes for use in interrogative clauses. These pronominal prefixes would then have spread to other areas of the grammar and acquired 'irrealis' meaning (Wallace Chafe pc). There is, however, no way to fit Caddo into the hierarchy; it would appear to be a clear counter example. Nonetheless, it is the only example found so far. More work needs to be done, therefore, in investigating other possible exceptions.

4. Conclusions.

This definition of modality has many similarities to those proposed by authors such as Lyons, Palmer and de Haan. Where it differs, however, is in the combination of the two scales 'epistemic' and 'deontic', which, given the way many languages treat modality, appears better. The treatment of modal categories as semantically 'positive' or 'negative' is also peculiar to this framework.

Modality has been defined in terms of seven grammaticalised speaker attitudes. Languages may make further distinctions within these categories, not every category may have identical denotation in every language, and not every category may be grammaticalised. The marking of modality as irrealis is not random. A hierarchy has been developed which allows the prediction of certain modality types as being considered irrealis on a language specific basis. While an exception to this hierarchy has been discovered, the exception is explicable because of the historical development of the language.

Notes

* This research was undertaken at the Research Centre for Linguistic Typology at the Australian National University over the summer break 1997-98. I am very grateful to the Centre and its Directors, Professors R M W Dixon and A Y Aikhenvald, for providing the Summer Student Fellowship which funded this research. Many thanks also to Keira Ballantyne, the other Summer Research Fellow, who provided much data and contributed many valuable ideas. Information on modality in the languages which formed the sample was collected from published sources and through questionnaires distributed to members of the Linguistics Department at the Australian National University and to the LINGUIST electronic mailing list. I am very grateful to all those who provided information. Many thanks to Prof. Aikhenvald for commenting on an earlier draft of this paper.

1. Included here are languages in which one or other of the paradigm is zero marked. In Tagalog, for example, irrealis is formally (though not functionally) unmarked; realis is realised by a verbal infix which was historically the marker of perfective aspect (Carl Rubino, pc).

REFERENCES

• Bybee, J and S Fleischman (Eds, 1995); Modality in Grammar and Discourse. Typological Studies in Language 32, John Benjamins.
• Bybee, J R Perkins and W Pagliuca (1994); The Evolution of Grammar: Tense, aspect and modality in the languages of the world. Chicago, University of Chicago Press.
• Chafe (1995); "The Realis-Irrealis Distinction in Caddo, the northern Iroquoian Languages, and English" in Bybee, J and S Fleischman (Eds); Modality in Grammar and Discourse. Typological Studies in Language: 32, John Benjamins
• Donaldson, T (1980); Ngiyambaa, the Language of the Wangaaybuwan. Cambridge, Cambridge University Press.
• Evans, N (1995); A Grammar of Kayardild. Berlin, Mouton de Gruyter.
• de Haan, Ferdinand (1996); The Interaction of Modality and Negation New York, Garland Publishing
• Jensen (1977); Yapese Reference Grammar. Honolulu, University of Hawai'i Press.
• Jespersen (1924); The Philosophy of Grammar. New York, George Allen and Unwin Ltd.
• Lyons (1977); Semantics. Cambridge, Cambridge University Press
• McGregor, W (1990); A Functional Grammar of Gooniyandi. John Benjamins
• - (1996); Nyulnyul. Munchen, Lincom Europa.
• Mithun (1995); "On the relativity of Irreality" in Bybee, J and S Fleischman (Eds); Modality in Grammar and Discourse. Typological Studies in Language 32, John Benjamins.
• Mous, M (1993); A Grammar of Iraqw (Kuschitische Sprachstudien Band 9) Hamburg, Helmut Buske Verlag
• Palmer, F R (1986); Mood and Modality. Cambridge, Cambridge University Press.
• Roberts (1990); "Modality in Amele and other Papuan Languages" Journal of Linguistics 26, 363 - 401.
• Siewierska, A (1991); Functional Grammar. Routledge, New York.
• Wagner (1997); On the Category of Irrealis Mood in Nyulnyulan Languages. Honours thesis, University of Melbourne.
• Wordick, F J F (1982); The Yindjibarndi Language. Canberra, Pacific Linguistics
• von Wright (1951); An essay in modal logic. Amsterdam, North Holland.