The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

42
The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages McGregor, William, 1952- Wagner, Tamsin. Oceanic Linguistics, Volume 45, Number 2, December 2006, pp. 339-379 (Article) Published by University of Hawai'i Press DOI: 10.1353/ol.2007.0005 For additional information about this article Access Provided by University of South Australia at 07/18/12 6:25PM GMT http://muse.jhu.edu/journals/ol/summary/v045/45.2mcgregor.html

Transcript of The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

Page 1: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

The Semantics and Pragmatics of Irrealis Mood in NyulnyulanLanguages

McGregor, William, 1952-Wagner, Tamsin.

Oceanic Linguistics, Volume 45, Number 2, December 2006, pp. 339-379(Article)

Published by University of Hawai'i PressDOI: 10.1353/ol.2007.0005

For additional information about this article

Access Provided by University of South Australia at 07/18/12 6:25PM GMT

http://muse.jhu.edu/journals/ol/summary/v045/45.2mcgregor.html

Page 2: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

Oceanic Linguistics, Volume 45, no. 2 (December 2006)© by University of Hawai�i Press. All rights reserved.

The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

William B. McGregor and Tamsin Wagner

The languages of the small Nyulnyulan family of the far northwest of West-ern Australia all exhibit a grammatical category traditionally dubbed irrealis.In this paper we describe the grammatical expression of this category, and itsrange of meanings and uses. It is argued that these can be accounted for ascontextual senses or pragmatic inferences based on a single encoded coremeaning, that the referent situation is construed by the speaker as unrealized.This semantic component remains invariant across all uses of the category,and is not defeasible. Contra claims by some investigators, the realis-irrealismood contrast is fundamental, and encapsulates a viable conceptual contrastbetween real and unreal events; epistemic and deontic notions of probability,necessity, desirability, and the like are secondary pragmatic inferences. Theirrealis is thus a modal category that can grammaticalize in human lan-guages; indeed, it is a communicatively useful category. We explicate thenature of the conceptual contrast between the construed real and unreal. It isfurther argued that the notion of scope is essential to an understanding of theirrealis, and its interaction with other mode-like categories. Finally, we situ-ate the Nyulnyulan irrealis in the wider cross-linguistic context of irrealis.

1. INTRODUCTION.1 In this paper we investigate irrealis mood in the Nyulnyulanlanguages of the Dampier Land peninsula and adjoining parts of the Kimberley regionin the far northwest of Western Australia. As in many non�Pama-Nyungan languages,in all Nyulnyulan languages�with the possible exception of Yawuru�tense distinc-tions are made in the irrealis. Correspondingly, irrealis categories are marked by thecombination of pre²xes indicating mood and a suf²x indicating tense (see Verstraete2005, 2006; Lazard 2006).2 This is illustrated by examples (1) and (2), where the irrea-lis pre²x -la- cooccurs with the past tense suf²x -na, as well as with a marked form ofthe nominative pronominal pre²x.

1. We are grateful to Claire Bowern, Jan Rijkhoff, Alan Rumsey, and Jean-Christophe Verstraete fordiscussion of the ideas presented in this paper and/or detailed comments on earlier drafts. Nyul-nyulanists Gedda Aklif, Claire Bowern, Komei Hosokawa, and Bronwyn Stokes are thanked forsharing data and insights with us. The usual disclaimers apply. Our greatest debts are, of course, tothe speakers of Nyulnyulan languages, in particular Mary Carmel Charles (Nyulnyul) and MaudieLennard and Freddy Marker (Warrwa), who generously shared their languages with us.

2. The terms �irrealis� and �irrealis category� are used in this paper in reference to the generalcategory marked by the irrealis pre²x, where relevant in combination with the pronominalpre²x. The temporal subcategories of the irrealis�such as past irrealis, that are typicallymarked by temporal suf²xes in addition�are referred to as �irrealis categories.�

Page 3: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

340 oceanic linguistics, vol. 45, no. 2

(1) Marlu wi-la-rr-arli-na kinya mayi, marlu.3 warrwanot 3:nom:irr-irr-aug-eat-pst this food not

�They didn�t eat it.� (McGregor 1994:25)

(2) Milarra oo-la-rli-na-ngayoo joorroo-nim. bardialmost 3:min:nom:fut/irr-irr-eat-pst-1:min:acc snake-erg

�The snake almost bit me.� (Aklif 1991a)

The Nyulnyulan languages constitute a separate family of non�Pama-Nyungan lan-guages, within which approximately ten languages are distinguished (McGregor 1988,2004:40�42). These fall into two groups (Stokes and McGregor 2003): Western Nyul-nyulan (Nyulnyul, Bardi, Jawi, Jabirrjabirr, Nimanburru, and perhaps Ngumbarl) locatedon the Dampier Land peninsula, to the north of Broome, in the far northwest of WesternAustralia, and Eastern Nyulnyulan (Warrwa, Nyikina, Yawuru, and Jukun) spoken in acrescent surrounding the peninsula to the south and east, and extending into the Kimber-ley region. The genetic distance within and between the two groups is small.

Not only are the Nyulnyulan languages closely related genetically, but they are alsotypologically quite similar. Like most other non�Pama-Nyungan languages, they are�pre²xing�: that is, they have both pre²xes and suf²xes�Pama-Nyungan languages,by contrast, are almost exclusively suf²xing (Capell 1940). Nominal morphology isrelatively simple. All languages show a small number of nominal derivational suf²xes,and a set of around a dozen bound postpositions, phrase-level enclitics that are attachedto the ²rst word of an NP, and mostly mark case relations. Verb morphology by con-trast is fairly complex, as will be seen in 2.1.

All Nyulnyulan languages are in precarious states; most are either extinct or mori-bund. Bardi and Nyikina are among the strongest, with around thirty ³uent speakerseach; Yawuru has slightly fewer, while Warrwa has a single ³uent speaker. To the bestof our knowledge, no ³uent speakers remain of the other languages.

The following are the main sources of data for this investigation: McGregor�s²eldnotes for Nyulnyul and Warrwa (see also his sketch grammars, McGregor 1996,1994); Stokes (1982) for Nyikina; Hosokawa (1991) for Yawuru; Metcalfe (1975),Aklif (1991a, 1991b), and Bowern (2004) for Bardi; Nekes and Worms (1953) (repub-lished as Nekes and Worms 2006) for Nyulnyul, Bardi, Jabirrjabirr, Nimanburru,3. Words in Nyulnyulan languages are transcribed as far as possible in the orthography adopted

by the speech community; for extinct languages and languages for which no system has beenadopted to our knowledge by the community of speakers or their descendants we haveadopted the erstwhile standard system for Nyulnyulan languages (McGregor 2004:23�26),with the addition of initial capitals and terminal punctuation in the style of this journal. Whereexamples are cited from previous sources, these are retranscribed in the accepted modernorthography; this sometimes involves a considerable amount of interpretation, and for thisreason the original representation is given in the second interlinear line. The exceptions areexamples cited from Nekes and Worms (2006), which are represented as in the source, whichis generally fairly reliable. Glosses and morpheme divisions follow the sources as closely aspossible, although we do interfere in instances where they are plainly inadequate. The follow-ing additional abbreviations, not included in the Leipzig Glossing Rules, are used: aug, aug-mented number; cm, conjugation marker; emph, emphatic; en, epenthetic nasal; exp,expectation; hab, habitual; inter, interrogative; IV, in³ecting verb; min, minimal number;num, number; pro, pronominal; quant, quanti²er; ref, re³exive/reciprocal; and sub, subor-dinate. IV roots are cited in capitals. Commas indicate the boundaries of intonation contours;clitic and affix boundaries are not distinguished, both being indicated by a hyphen (-).

Page 4: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 341

Yawuru, and Nyikina. None of the sources, however, provides comprehensive infor-mation on or discussion of the irrealis in any Nyulnyulan language, and there are manygaps in our knowledge and understanding of the category. Virtually no information isavailable on the irrealis in Jukun, Jawi, or Ngumbarl, and these languages are barelymentioned in the paper; based on their apparent similarity to nearby languages, onewould guess that they possessed an irrealis category substantially comparable to that ofthe better known Nyulnyulan languages.

This paper has three main aims. The ²rst is to describe the range of uses of the irre-alis category, and to show that these can be accounted for as contextual senses or prag-matic inferences based on a single encoded core meaning, namely that the referentsituation is construed as unreal.4 What this means, roughly, is that the situation is repre-sented as belonging to a possible world, though not what the speaker considers to bethe realm of real or actual world of past, present, or future events�it lies off the real-world time/event line construed by the speaker.5 Thus we argue that in Nyulnyulan lan-guages there is indeed (contra Bybee, Perkins, et al. 1994; Bybee 1998) a grammaticalopposition that encapsulates a conceptual contrast real vs. unreal situations. Weattempt to re²ne the concept of the unreal in sections 3 and 4, and distinguish it fromcompeting notions such as negation and potentiality. Second, it is argued that thenotion of scope is essential to an understanding of the uses of the irrealis, and its inter-action with other mode-like categories. Third, we situate the Nyulnyulan irrealis withinthe wider grammatical phenomenon of mood. We attempt to situate the Nyulnyulanirrealis in a typological context of irrealis and related moods, and to explore the typo-logical and theoretical implications of our ²ndings. Before getting down to business,however, it is essential to provide some background information on the verbal con-struction in Nyulnyulan languages.

2. GRAMMATICAL PRELIMINARIES

2.1 OUTLINE OF NYULNYULAN VERB MORPHOLOGY. Two types ofverbal construction are distinguished in Nyulnyulan languages, SIMPLE and COM-POUND. Simple verbal constructions consist of a single in³ecting verbal lexeme,referred to here as an in³ecting verb (IV). Order class formula (3) illustrates the struc-ture of ²nite IVs in Warrwa. This is fairly typical, as can be seen by a comparison withthe order class formula (4) for Bardi (adapted from Metcalfe 1979:204; Nicolas 1998;and Bowern 2004:101). See also Stokes (1982:237, 293) for a description of theNyikina IV; Hosokawa (1991:114) for the Yawuru IV; and McGregor (1996:38) forthe Nyulnyul IV. Illustrations are provided in the example sentences.

4. We use the term �situation� in reference to the phenomenon in a real or imaginary world that isdenoted by an entire clause, minus the elements that provide modal modi²cation. The term�event� is used in a narrower sense, in reference to just the happening, that which is designatedby the verb.

5. It should be borne in mind that the irrealis category expresses a construal by the speaker of asituation in the real world or some possible world. To avoid excessively prolix and pedanticexpression we sometimes omit the framing quali²cation �in the speaker�s construal/opinion/representation�; it should be consistently read as present.

Page 5: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

342 oceanic linguistics, vol. 45, no. 2

(3) nom pro + (cm) + mood/tense + (num + [cm]) + (ref1) + (en) + root

+ (ref2) + (tense) + (aspect) + (relator) + ( pro) + (du)

warrwa

(4) nom pro + (cm) + tense/mood + (num + [cm]) + (ref1) + root + (ref2) + (appl1) + (tense/aspect1) + (tense/aspect2) + (appl2) + (tense/aspect) + (simultaneous marker) + (sentential clitic) + (obl1

pro) + (obl2 pro) + (quant) + (acc pro) bardi

Although there are differences in the ranges of morphemes available in each languageand the position they occupy, there are striking regularities. In the initial position is a pro-nominal pre²x, obligatory in all languages,6 that cross-references the Actor (roughly, sub-ject) of the clause.7 Different allomorphs of the pronominal pre²xes are selectedaccording to tense, mood, and conjugation class. The pronominal pre²x may be followedby a tense or mood marker, a number marker,8 a conjugation marker, and/or a re³exive/reciprocal marker; the morphemes usually occur in that order, though there is some vari-ation depending on whether the number marker is present. Immediately before theroot�obligatory in all languages�an optional epenthetic nasal (en) may occur in cer-tain environments. Following the root is a range of optional items, including a re³exive/reciprocal suf²x, tense and aspect suf²xes, an accusative and/or oblique pronominalenclitic,9 and (in Eastern languages) a dual marker.

Most Nyulnyulan languages distinguish two primary conjugation classes, to whichthe majority of IVs belong (McGregor 2002:214�19). Glossing over many complexi-ties not germane to the present paper, these are de²ned by the allomorph sets of thepronominal pre²xes, and by the absence or presence of the conjugation marker. Thusthe two classes are called Ø and na respectively, na- being the most common allomorphof the conjugation marker. There is a good correlation between the two conjugationclasses and valency of the IV: most Ø-class IVs are monovalent, while na-class IVs aregenerally bivalent (McGregor 2002:218�19).

Various dependencies and cooccurrence restrictions obtain between the orderclasses, and the morphemes that may ²ll them. Thus, in most languages, the accusative

6. This is true for ²nite IVs in all Nyulnyulan languages. Many of the languages, however, alsoshow an in²nitival verbal construction in which an invariant pre²x, usually ma-, replaces thepronominal pre²x. This construction shows fewer potential order classes than ²nite IVs; inparticular, it does not admit speci²cation of tense or mood. Hence this construction is irrele-vant to the present paper.

7. The bound pronominals in the IV show three or four different sets of forms according to thegrammatical roles they cross-reference, and for this reason are labeled as separate case forms.

8. This indicates augmented number�that is, one or more items in addition to the minimumnumber consistent with the category (Conklin 1962; McGregor 2004:113). Free pronouns andpronominal enclitics to the verb show an Ilocano type system distinguishing four person cate-gories (1, 1+2, 2, 3) and two numbers (minimal and augmented). However, the contrastbetween 1 augmented and 1+2 augmented is neutralized in well de²ned environments (thatdiffer from language to language) in the system of person and number pre²xes, where theyshow an Assiniboyne system (Greenberg 1988).

9. Only in Bardi is it possible for an IV to show both an accusative and an oblique enclitic pronomi-nal, and this may be a recent development: according to Claire Bowern (pers. comm.) combina-tions are less common in Gerhardt Laves�s corpora from the late 1920s than in modern recordings.In all other Nyulnyulan languages the two case forms are in complementary distribution.

ACCOBL

⎧⎨⎩

⎫⎬⎭

Page 6: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 343

and oblique pronominal enclitics are in complementary distribution (as shown in [3]).And in some languages (e.g., Nyulnyul) the irrealis mood pre²x appears to be in com-plementary distribution with the conjugation pre²x (see 2.2). On the other hand, inmost languages the re³exive/reciprocal pre²x and suf²x usually cooccur, though occa-sionally only the suf²x is found (McGregor 2000).

The compound verb construction consists of an IV together with an almost invariantverbal particle. This almost always precedes the IV, and accordingly is referred to here asa preverb. (5) and (6) are examples. (See McGregor 2002 for further discussion.)

(5) Marlu ngij nga-la-ama-na. warrwanot smile 1:min:nom-irr-laugh-pst

�I didn�t smile.�

(6) Doodool i-n-joo-na-na. bardiknocking:sound 3:min:nom-cm-say-dur-pst

�It kept on making a knocking noise.� (Metcalfe 1975:58)

2.2 MORPHOLOGY OF THE IRREALIS

2.2.1 Morphological structure of irrealis categories. As remarked in section 1,the irrealis categories in each Nyulnyulan language are formally composite. They areusually marked not by a single morpheme, but by combinations of morphemes fromtwo or more order classes�we know these are separate morphemes (rather than a sin-gle circum²xal morpheme) because they do not always cooccur. There is always apre²x marking irrealis mood (see 2.2.2 for allomorphy), which almost always occursin second position, directly following the obligatory subject (nominative) pronominalpre²x, and preceding the number pre²x (for augmented numbers). This mood pre²xtypically cooccurs with a suf²x specifying tense, as illustrated by the IVs in Warrwaexamples (1) and (5) and Bardi example (2).

Three complications to this general pattern are notable. First, in some languagesone irrealis category lacks a tense suf²x�for instance, in Nyulnyul it is the nonpast,while in Warrwa it is the future. Second, in perhaps all Nyulnyulan languages the pro-nominal pre²x allomorph sets in the irrealis show some differences from the realisallomorphs. Third, in Eastern Nyulnyulan languages the irrealis pre²x itself shows dif-ferent forms according to tense.

The composite nature of the formal expression of the irrealis categories, as sug-gested by Verstraete (2005, 2006), is diagrammatically iconic of the semantic compo-sitionality of the categories. Although this iconicity breaks down in places, thesemantic compositionality of the irrealis categories shines through in all Nyulnyulanlanguages, and underlies the approach of the present paper.

2.2.2 Allomorphy of the irrealis pre²x. This section outlines the major charac-teristics of the morphology of the irrealis pre²x in Nyulnyulan languages. No attempt hasbeen made to be exhaustive, and a number of complexities not germane to the paper areglossed over (see references cited in section 1 for more comprehensive coverage).

In Western Nyulnyulan the irrealis pre²x has allomorphs la- ~ li- ~ lu ~ l-. It is notentirely clear what conditions the choice between these allomorphs, though it does

Page 7: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

344 oceanic linguistics, vol. 45, no. 2

seem that both phonological and morphological factors are relevant. In Nyulnyul la- isusually found following a of the 1 and 1+2 person minimal (e.g., nga-la-kalak �I mightcome�, ya-la-barnj �we [me and you] might exchange�), while li- usually follows the iof the 2 person (e.g., mi-li-jid �you might go�, mi-li-janb �you might kick�). The li- allo-morph is usually selected in the augmented, that is, preceding the number pre²x rr-, asin ya-li-rr-kanyj �we all might forget�, ku-li-rr-kanyj �you lot might forget�, and i-li-rr-ngurid �they might paint (it)�. In the 2 person augmented, however, the allomorph lu-sometimes occurs, evidently conditioned by the quality of the preceding vowel, invari-ably u: ku-lu-rr-karrmar �you lot might break it�. This seems to be the only environ-ment in which the lu- allomorph occurs. Similar conditioning factors may apply inJabirrjabirr and Nimanburru (Nekes and Worms 2006:254),10 though these authorsindicate that the vowel e (their symbol for the high front vowel) occurs instead of aeverywhere except following the 1+2 augmented ya.

The ²fth allomorph, l-, is found in the minimal number forms of the Nyulnyul, Jabirr-jabirr, and Nimanburru IV -dam �hit� (e.g., the Nyulnyul nga-l-dam-an �I might have hithim/her/it [but didn�t]�).11 This is one of the very few IVs with an initial apical stop, andthe only one for which irrealis forms are available, so it is not known whether lexical orphonological features motivate allomorph choice. (Other apicals�nasals and laterals�do not seem to condition this allomorph.) It appears that this monoconsonantal form isthe elsewhere allomorph in Bardi (Metcalfe 1975:96�97; Bowern 2004:11, 106).

The Eastern Nyulnyulan languages show additional allomorphs, conditioned by tenseand other features including the conjugation class of the IV (basically indicated by thepresence or absence of the conjugation pre²x). In Nyikina and Warrwa the la- ~ lu- ~ l-(etc.) allomorph set is restricted to nonfuture tenses, with la- the elsewhere variant. InWarrwa l- occurs immediately preceding a stop-initial IV (the stop sometimes lenites to aglide), and la- almost everywhere else (see McGregor 1994:41 for further details). InNyikina the vowel is lost before certain IVs including those with initial n and rr, the initialsegment of which assimilates to the lateral (Stokes 1982:278�79).

According to Hosokawa (1991:138), Yawuru does not show the l(V)- allomorph setat all, and indeed, lacks a distinct nonfuture irrealis category, suggesting that examplesinvolving these allomorphs �are not only rejected by the informants but are also absentin natural speech.� Most other sources, by contrast, exemplify the la- allomorphs,including Worms (1942), Nekes and Worms (2006), Stokes (1979), and Yawuru Lan-guage Team (1998). Hosokawa himself recorded a few instances of la- irrealis IVs. Heobserves, however, that these were provided by a speaker bilingual in Nyikina, andproposes that they represent transference from that language (Hosokawa 1991:138).This explanation could be extended to the other attestations of the l(V)- allomorph set,

10. Nekes and Worms (2006) consider the lateral to be the marker of the irrealis, and the follow-ing vowel, if there is one, to be an augment, added in for phonological reasons (�euphony�). Itseems that they regard this mood marker as an �indicative form� rather than a morpheme(Nekes and Worms 2006:150).

11. According to Nekes and Worms (2006:161), the allomorph l- frequently disappears entirely,especially in the second person singular, as in arri mi-dam �do not hit�, instead of the regulararri mil-dam. Such elisions do not occur in Nyulnyul.

Page 8: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 345

given that all speakers of Yawuru are�and presumably traditionally were�³uent sec-ond language speakers of Nyikina.12

Different allomorphs are found in the future. In Nyikina ya- occurs when the sub-ject is augmented in number, and rra- when it is minimal (Stokes 1982:278)�com-pare ya-ya-rra-ma (1:nom-irr-aug-go) �we might go� with nga-rra-ma (1:min:nom-irr-go) �I might go�. In Warrwa, ya- is found in future forms of Ø-class IVs, while forna-class verbs ya- occurs only in the augmented number, as in wi-ya-rr-a �they mightcarry it�; in the minimal number the erstwhile conjugation marker na- occurs instead ofthe irrealis pre²x, as in nga-na-ra �I might spear it� and wi-na-ra �s/he might spear it�.13

The same situation obtains in Yawuru future irrealis.

2.2.3 Pronominal pre²xes. The paradigms of pronominal pre²xes in many Nyul-nyulan languages differ somewhat according to tense and/or mood. In Bardi, as shownin table 1, there are two such paradigms, one for nonfuture, the other for future andirrealis. The forms are syncretic except in the second and third person minimal catego-ries. Notice that in the pronominal pre²xes the contrast between 1 augmented and 1+2augmented�found in the free pronouns and the oblique pronominal enclitics�is notmaintained in Western Nyulnyulan languages (note 8). (See also the paradigms givenin chapter 5 of Nekes and Worms 2006, especially pp.153�54, 253�54.)

There is one unusual wrinkle: according to Nekes and Worms (2006:252�58),Jabirrjabirr has a distinct third person minimal �irrealis� pronominal pre²x ngarr-(¥ar-) in addition to the expected pronominal and irrealis pre²xes, i- and -lV-. How-ever, neither their discussion nor the data they provide (also Nekes and Worms1953:790) are suf²cient to determine whether the forms are in free variation, and if not,what conditions the choice between them.

12. Within the Eastern Nyulnyulan group there is evidence of grammatical in³uence from Nyikina,presently the most viable of the languages. This includes among other things borrowing of anergative allomorph into Warrwa (McGregor 2006).

13. There is no way of deciding whether -na- should be treated as the conjugation marker with a Ø(or no) irrealis allomorph, as the irrealis marker without the conjugation marker, or as a port-manteau of mood and conjugation marker.

TABLE 1. PRONOMINAL PREFIXES TO BARDI IVs†

� After Bowern 2004:180.

minimal augmented�

� The system has been laid out as a four-person and two-number system for expositoryconvenience, and to permit comparisons with forms in other Nyulnyulan languages.We take the view that the person-number system is genuinely distinct from the systemfound in the free pronouns, not the same system with accidental homophony.

1 Non-fut nga- a- � arr-*

* The second discontinuous morpheme is the augmented number pre²x.

fut, irr nga- a- � arr-1+2 Non-fut a- a- � arr-

fut, irr a- a- � arr-2 Non-fut mi- goo- � arr-

fut, irr a- ~ nga- goo- � arr-3 Non-fut i- i- � arr-

fut, irr oo- oo- � arr-

Page 9: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

346 oceanic linguistics, vol. 45, no. 2

2.3 MARKEDNESS AND NEUTRALIZATIONS. As already remarked (2.1),cooccurrence constraints exist among the ²llers of some order-class pairs in Nyul-nyulan IVs. Notable among these are constraints on the morphemes that can cooccurwith the irrealis. The result is that the number of cross-cutting categories is never morein the irrealis than in the realis, as expected given that irrealis is the more marked cate-gory both functionally and formally. We turn now to three circumstances in which dis-tinctions maintained in realis mood are in some circumstances neutralized in theirrealis. These are person-number categories, tense, and conjugation classes.

As mentioned already, in many Nyulnyulan languages pronominal pre²x para-digms differ according to tense and mood. In Eastern Nyulnyulan languages we ²nd aparadigm speci²c to the irrealis mood, the forms of which are partly syncretic with therealis forms. This can be seen from the Warrwa pronominal pre²xes shown in table 2.In Warrwa (McGregor 1994:41) and Nyikina (Stokes 1982:277) an Ilocano system isfound in the realis, while in the irrealis the system is Assiniboyne: the contrast between1 augmented and 1+2 augmented is neutralized, though a different form is employedfor the 1+2 minimal (i.e., speaker-hearer dyad). In Yawuru the situation is a little morecomplicated, the distinction between 1 augmented and 1+2 augmented in the irrealisbeing sometimes maintained, sometimes not (Hosokawa 1991:116�18; 137�38).

All Nyulnyulan languages distinguish at least three tenses, past, present, and future,where these are de²ned by the form of the pronominal pre²x to in³ecting verbs, atense pre²x, and/or tense (or aspect) suf²xes. Some languages make just this three-way contrast; Nyulnyul is such a language (McGregor 1996:43). Other languagesreputedly make ²ner distinctions. Thus Nyikina, according to Stokes (1982:296), dis-tinguishes a general from a recent past. For Bardi, Metcalfe (1975:67) distinguishes nofewer than seven tenses. Not all of these appear to be pure tenses, however. It seemsthat the three-way tense distinction is marked by pre²xes, while cooccurring suf²xescan be used to make further tense/aspect distinctions (Bowern 2004:210, 217).

TABLE 2. PRONOMINAL PREFIXES TO WARRWA IVs

minimal augmented

1 Non-fut nga- nga- � rra-

fut ka- ka- � rra-

irr nga- ya- � rra-

1+2 Non-fut ya- ya- � rra-

fut ya- ya- � rra-

irr ya- ya- � rra-

2 Non-fut mi- ku- � rra-

fut wa- (na-class); nga- (Ø-class) wa- � rra-

irr mi- ku- � rra-

3 Non-fut Ø- (na-class); Ø- ~ i- ~ ngi- (Ø-class) ngi- � rra- ~ i- � rra-

fut Ø- (na-class); ku- (Ø-class) ku- � rra-

irr wi- wi- � rra-

Page 10: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 347

At most three tense distinctions are made in the irrealis in any Nyulnyulan lan-guage. The Eastern Nyulnyulan languages Nyikina and Warrwa maintain a three-waytense distinction in the irrealis; the contrast between general and recent past in Nyikinais not maintained, and only the former occurs. The only aspectual category distin-guished in the irrealis in these two languages is the habitual.14 By contrast, the irrealisin Yawuru is restricted to the future, according to Hosokawa (1991:138).

In Western Nyulnyulan languages, the story gets more complex. To begin with,there is no tense-conditioned allomorphy in the irrealis pre²x, so that the pronominaland mood pre²x collocations are identical throughout the irrealis paradigm (see alsoBowern 2004:11, 184). Temporal distinctions in the irrealis are made exclusively bysuf²xes. In Nyulnyul a two-way distinction is maintained between past and nonpastirrealis, the former being marked by the suf²x -an pst, the latter by absence of a suf²x(McGregor in preparation: section 7.6). The contrast between present and future isneutralized in the irrealis in Nyulnyul. In Bardi it appears that a three-way distinction ismaintained between past (marked by the general past suf²x -na), present (marked bythe continuous aspect suf²x -n), and future (marked by the future suf²x -a) (Bowern2004:106�7).15 The ²ner tense/aspect distinctions are neutralized. Thus there is no con-trast between remote and recent past, as in the realis (Bowern 2004:217). It is impossi-ble to be certain what the situation is in the other Western Nyulnyulan languages, dueto paucity of data. From examples cited in the text and the dictionary of Nekes andWorms (2006), however, the situation in Jabirrjabirr and Nimanburru appears to be asreported for Nyulnyul, that is, a contrast is maintained between past (realized by thesuf²x -an) and nonpast (no suf²x) irrealis.16

Conjugation class distinctions are not always maintained in the irrealis category inNyulnyulan languages: that is to say, the in³ection of regular IVs in the irrealis can beidentical across the conjugation classes, at least in certain environments. First, pronom-inal pre²xes do not show allomorphy conditioned by conjugation class (see, forinstance, the Warrwa forms in table 2). The only exception is Nyikina, in which thetwo allomorphs of the second person minimal pre²x are conditioned by conjugationclass. Second, the conjugation marker is usually in complementary distribution withthe irrealis mood pre²x. As a result, in Nyulnyul the conjugation marker never occursin the irrealis, and verbs of both conjugation classes show identical in³ection (McGre-gor 1996:42). In Bardi, the distinction between the two classes disappears in minimalbut not augmented number (Bowern 2004:83�84, 184�85).

14. Note that in these two languages habitual appears not to be a possible sense of the irrealis, asin some languages; rather, habitual aspect is marked by a suf²x (see example [61]) that alsooccurs in realis contexts.

15. According to Bowern (2004:225) one of the three tense suf²xes must occur in irrealis categories.However, at least one example she cites, (7.10c), shows no suf²x (and none is indicated in thegloss line).

16. Nekes and Worms (2006:252�58) do not explicitly remark on any temporal distinctions in theirrealis, and give examples only in the nonpast (i.e., without suf²xes). One might infer from theirstatement that the irrealis marker is �inserted into the realis forms� that the same tense distinctionsexist in both moods. However, elsewhere (e.g., Nekes and Worms 2006:154�55) they produce,albeit without discussion, distinct paradigms for Jabirrjabirr past and nonpast irrealis. In still otherplaces they provide in their paradigms for Eastern Nyulnyulan language two alternating forms forthe irrealis pre²xes, presumably to be interpreted as future and nonfuture sets.

Page 11: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

348 oceanic linguistics, vol. 45, no. 2

In Warrwa the situation is slightly different (McGregor 1994:41�42). In the pastand present irrealis the conjugation classes are distinguished in augmented but notminimal numbers. But in the future, the conjugations are maintained: Ø-class verbsemploy the allomorph -ya- of the irrealis pre²x, while na-class IVs employ the -ya-allomorph in augmented number, while -na- (the erstwhile conjugation pre²x) appearsin minimal number. The situation for the Yawuru irrealis (restricted to future, as will berecalled), is similar, except in the Julbayi dialect where -na- in the minimal number ofna-class IVs tends to be replaced by -ya- there too (Hosokawa 1991:136�37), indicat-ing progressive shift toward neutralization of the conjugational distinction. In Nyikinathe two conjugation classes select almost the same allomorphs of the irrealis, la- in thenonfuture, and in the future, rra- for minimal numbers and ya- for augmented num-bers. However, the ²rst two forms show slightly different behaviors according to con-jugation class, so that for IVs with initial n or rr slightly different paradigms are found(Stokes 1982:278). In minimal numbers there are just a few differences in the respec-tive paradigms. In nonminimal numbers, the conjugation distinction seems to belargely neutralized across all tenses (Stokes 1982:253, 278).

The neutralizations discussed above are summarized in table 3.Category neutralizations in the irrealis are expected given the markedness of the

irrealis with respect to the realis. That neutralizations should occur in the domain oftense/aspect marking is not surprising. It seems intuitively reasonable to expect that it isless important to make temporal distinctions for events belonging to the realm of theunreal, and in many languages irrealis admits no such distinctions.

The (partial) neutralizations of the conjugation distinctions are also explicable. Thisfollows from the strong correlation between conjugation class and transitivityremarked on in 2.1. The neutralizations in the conjugation class contrasts re³ect neu-

TABLE 3. OVERVIEW OF NEUTRALIZATIONSIN THE IRREALIS CATEGORIES IN NYULNYULAN LANGUAGES

category type of neutralization

person/number Neutralization of contrast between 1+2 aug and 1 aug in Warrwa and Nyikina throughout irrealis paradigm (maintained in realis); partial neutralization in Yawuru.This contrast is not maintained in the pronominal pre²xes in Western Nyulnyulan languages.

tense/aspect Reduction in number of tense/aspect distinctions available in the irrealis:� neutralization of present/future contrast in Nyulnyul;� loss of temporal contrasts in the past in Bardi and Nyikina;� irrealis available only in the future in Yawuru.

conjugation class Loss or reduction in conjugation class conditioned allomorphy in pronominal pre²xes in the irrealis.Conjugation classes show more syncretisms in the paradigms of pre²xes than in the realis, and the distinction between them is lost in some grammatical contexts; this tendency is more developed in Western Nyulnyulan than Eastern Nyulnyulan languages.

Page 12: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 349

tralizations in explicit transitivity of IVs;17 for the few IVs that are ambicategorial, thisneutralization results in the loss of an overt contrast between intransitive and transitiveverb forms. Thus, whereas in Nyikina the na-class IV -ma �make, put� and the Ø-class-ma �go� in³ect differently in each person-number form in the past realis, in the pastirrealis they in³ect identically, save for the second person minimal.

Irrealis IVs show a tendency to be in³ected according to the morphologically lessmarked intransitive Ø-conjugation. This is consistent with the correlation observed byHopper and Thompson (1980:252) between transitivity and mood, whereby irrealiscorrelates with low transitivity, realis with high transitivity. Moreover, future irrealisevents allow some margin of possibility, which does not exist for past irrealis (see fur-ther section 3). Thus we might say that future irrealis is notionally less unreal thanpast or present irrealis. Thus the maintenance of transitivity in the future in Warrwa isalso explicable.18

To conclude, it is observed that the status of realis as a single grammatical categoryin minimal contrast with the irrealis is somewhat dubious. Except for a partially dis-tinct set of realis pronominal pre²xes in some languages, there is little in favor of thecategory.19 There is no overt second-position pre²x marking realis mood in any Nyul-nyulan language; instead, tense pre²xes occur there, in complementary distributionwith the irrealis. This suggests that rather than a single realis category we should iden-tify a number of realis tense categories. Furthermore, whereas the irrealis may be pre-sumed to be speci²ed positively as +unreal, to the extent it is distinguishable, realis ispresumably unspeci²ed, that is, ±unreal. As will be seen in section 4, there is reason tobelieve that this is so.

3. USES OF THE IRREALIS MOOD IN NYULNYULAN LANGUAGES.This section outlines the range of uses and etic senses of irrealis mood. Two compo-nents of meaning are recurrent. One is the notion of nonoccurrence or nonactualiza-tion: the referent situation belongs to the realm of the unreal: basically, it has notoccurred, is not occurring, or will not occur. The other is the modal notion of potential-ity, the idea that the situation could occur, or could have occurred: it is not an impossi-bility. The situation thus belongs to a possible, but not actual, world. Both of thesecomponents are speaker construals, and need not necessarily correspond with reality.

To orient the reader, we provide diagrammatic representation of our conceptualiza-tion in ²gure 1. This represents possible and actual worlds on a temporal grid, repre-sented as usual on the horizontal axis, progressing from left to right. Actual-world

17. Nekes and Worms (2006:253) agree, saying �It [irrealis mood] has the same forms in transi-tive and intransitive verbs.�

18. Not all neutralizations are so readily explained functionally, and some may well result from his-torical phonological or morphophonemic processes. Thus the loss of the conjugation distinctionin the minimal but not augmented number in Bardi is unexpected in our story. One possibility isthat this loss has a morphophonemic motivation: the cluster n-l that arises in na-class minimalirrealis forms is reduced to l (Bowern 2004:83; Claire Bowern, pers. comm.), resulting in thesame form as in the ø-class minimal irrealis. This process would not apply in the augmentedforms where the n-l cluster does not appear, and the conjugation distinction is maintained.

19. Because it is useful for expository purposes, we will continue to use the term, though withoutthe presumption that it delimits an emic grammatical category.

Page 13: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

350 oceanic linguistics, vol. 45, no. 2

situations fall on the central unbroken line, while possible-world situations fall on theregion off this line, on broken lines indicating possible worlds. (Notice that somefuture situations are evaluated as �real,� and belong to the actual-world; we return tothis point in section 4.) Intuitively, the distance of a broken line from the central unbro-ken line represents the closeness of the match of the possible world to the real world, ifyou will, moving from a probability of one (actual world) to zero (impossible worlds).

Figure 1 also affords an understanding of tense in relation to realis and irrealis. Thecentral line represents the familiar Reichenbachian time line, with a zero point at thehere-now as the reference point for temporal deixis. Possible worlds are aligned tem-porally with the actual world, permitting use of the same reference point of the speechsituation to locate possible world situations. To forestall misapprehensions, it isobserved that the temporal reference point for tense marking in Nyulnyulan languagesis never shifted to a secondary point: there are no relative tenses. The only environmentin which shifting of the temporal reference point occurs is in quotation.

FIGURE 1. REAL AND POSSIBLE WORLDS LOCATED IN TIME

Page 14: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 351

3.1 IRREALIS IN INDEPENDENT CLAUSES

3.1.1 Negative clauses. The environment in which the irrealis is most commonlyencountered is negative clauses, where it seems to be obligatory, coselected with thenegative particle, arra ~ arri (Western Nyulnyulan), and marlu (Eastern Nyulnyulan),when it has clausal scope.20 This is true for the Western Nyulnyulan languages Bardi(Metcalfe 1975:97�98; Bowern 2004:55), Nyulnyul (McGregor 1996:26), Jabirrjabirr(Nekes and Worms 2006:252), Nimanburru (Nekes and Worms 2006:252), and theEastern Nyulnyulan languages Nyikina (Stokes 1982:280, 376) and Warrwa (McGre-gor 1994:42). Examples (7�9) are illustrative; (8) and (9) suggest that the irrealis isprobably also obligatory in negatives in the two very poorly documented languagesJawi and Jukun.

(7) Arri nga-li-j-an wal. jabirrjabirrarri na-li-j-an walnot 1:nom-irr-get-pst ²sh

�I couldn�t get any ²sh.� (Peile n.d.:38)

(8) Juy arra nga-la-mur. jawijou arra nu-la-moor2:min:nom not 1:min:nom-irr-throw

�I did not throw [it to you].� (Bird 1910:455)

(9) Marlu nga-la-jala wula. jukun21

maaloo ngalla woola jallanot 1:min:nom-irr-see water

�Water, I cannot ²nd.� (Bates n.d.:45)

Yawuru is an exception. According to Hosokawa (1991:143), negative clauses invari-ably select realis mood, as illustrated by the following example. It will be recalled thatHosokawa argues that irrealis is restricted to future tense in Yawuru; other sources,however, give examples of the irrealis in negative clauses in past tense�e.g., Nekesand Worms (2006:252); Stokes (1979:14). According to Hosokawa such examplesshow in³uence from neighboring Nyikina.

(10) Marlu nga-na-ga-bilka-dyuyu ngurru. yawurunot 1:nom-cm-fut-hit-2:min:acc more

�I won�t hit you again.� (Hosokawa 1991:367)

20. In some Nyulnyulan languages, the negative particle�and/or negative and privative particlesderived from it�can be used in NPs for nominal constituent negation (as in, e.g., �I wentwithout my things�). In these cases the irrealis is not selected (except, of course, when theclause is also negated, as in �He can�t see without his glasses�). Negative particles cannot,however, be used to specifically negate a finite verbal construction serving as the main predi-cate of a clause. The closest we can get to this is through employing a clausal negative particlein an irrealis clause, with specific focus on the verb (as in the Warrwa example Nga-n-jala-nykinya jurru marlu-yangarri nga-l-nka-na. [1:nom-cm-see-pst this snake not-emp 1:nom-irr-hit-pst] �I saw the snake, but I didn�t kill it.�).

21. The provenance of this example is not entirely certain, and it could be either Jukun or Ngum-barl; we favor the ²rst possibility from the form of the negative particle. Note that the originalversion of this sentence appears to have morphemes in the wrong order (see second line); wehave reconstituted the example in line with our reanalysis.

Page 15: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

352 oceanic linguistics, vol. 45, no. 2

Negated clauses, of course, code the meaning that the proposition expressed is false,and the referent event did not occur, is not occurring, or will not occur. They carryadditional nuances of meaning, and are pragmatically marked with respect to positivepolarity clauses, as observed by Strawson (1952:7), Givón (1984:323�24), McGregor(1997:226�27), Miestamo (2003:171�72), and Israel (2004:706), among others. Theyare in a sense context-bound, and invoke and deny the corresponding positives. Forexample, as speech acts, (7�9) would seem to deny the presuppositions that �I gotsome ²sh�, �I threw [something to you]�, and �I found water�, respectively. It might beobjected that this is reading too much into decontextualized examples in poorly knownlanguages. But evidence from better documented Nyulnyul supports the viability ofthe claim. Thus the ²rst clause of (11) would normally occur in the context in whichthe object was broken, and the speaker appeared to be responsible. Following along thelines of the discussion of Givón (1984), if this were uttered out of the blue, theaddressee might well respond that they had not accused the speaker of having brokenthe object. That is to say, either the presupposition is reasonably presumed present inthe hearer�s mind, or else it is invoked by the speech act itself.

(11) Arri nga-la-arrmar-an i-ny-jalk; nyulnyulnot 1:min:nom-irr-break 3:min:nom-en-fall

murrul-murrul nga-mal-uk.little-little 1:min-arm-loc

�I didn�t try to break it; it fell to pieces in my hand.�

This observation is perhaps even clearer for negative polarity in the nonpast or future.As implied by the discussion of 2.1, Nyulnyulan languages do not distinguish impera-tive mood. While positive commands are prototypically expressed by the plain future,negative commands and prohibitives are expressed by the combination of negative par-ticle and nonpast or future irrealis mood, as in (12). This invokes a presumption that thehearer might spill the water, just as (13) invokes the possibility that women might visitthe place in question.

(12) Arra mi-la-moor-a oola. bardinot 2:min:nom-irr-spill-fut water

�Don�t spill the water!� (Aklif 1991a)

(13) Kinyingk bur arri urany-in ya-li-rr-jid nyulnyulthis place not woman-erg 1:nom-irr-aug-go

ngurlangurl bur.sacred place

�We women shouldn�t go to this place.�

These few remarks hardly do justice to the complexities of negative polarity in Nyulnyulanlanguages, and we return to the topic in 4.2 where we contextualize it in relation to scope.

3.1.2 Positive clauses

3.1.2.1 Past irrealis. In positive clauses the past irrealis expresses counterfactual-ity, the speaker�s assessment that a situation did not happen, although it might have. It

Page 16: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 353

evaluates the referent situation as nonactualized, and situated at some point of timeprior to the time of the speech situation (see ²gure 1). No doubt as to the reality-statusof the situation is admitted: it is construed by the speaker as not having occurred. Thepast irrealis also indicates potentiality of the situation: it might have occurred. In posi-tive clauses, that is, past irrealis is used not just of any situation that didn�t occur, butone that was not outside the range of possibility. The situation is presented as belong-ing to a possible world, though not to the actual world. Both these senses are present inthe range of contextual senses identi²ed below.

(a) The situation, according to the speaker, almost or nearly happened, although itdidn�t actually occur, as in examples (2) and (14�16).

(14) Miliyarri dumarra nga-l(a)-andi-na. nyikinaalmost take:off 1:min:nom-irr-pick:up-pst

�I nearly took off.� (Stokes 1982:281)

(15) Ngurrngurr yi-la-w-an i-ny-jalk-uk wurl-uk. nyulnyuldrown 3:nom-irr-give-pst 3:min:nom-en-fall-loc water-loc

�He nearly drowned when he fell into the water.�

(16) Miliyarri nga-l-janba-na kinya juurru ngayu-na. warrwalong.ago 1:min:nom-irr-step-pst this snake I-erg

�I nearly stepped on the snake.�

(b) The Actor tried unsuccessfully to perform the event, as in (17) and (18). (Thelatter example is an instance of what Hosokawa [1991] refers to as the idiolectalusage of the -la- irr pre²x in Yawuru.)

(17) Nga-l-dam-an-jii jurnk mi-nyu way. nyulnyul1:min:nom-irr-hit-pst-2:min:acc run 2:min:nom-catch away

�I tried to hit you, but you ran away.�

(18) Wi-la-ra-ngayu nga-mirdibi-and-jina. yawuru3:min:nom-irr-spear-1:min:acc 1:min:nom-run.away-pst-3:min:obl

�He tried to spear me and I ran away.� (Hosokawa 1991:139)

(c) The Actor wanted or intended the situation to occur, but in the end it didn�t.

(19) Mi-la-r-an karrkuj. nyulnyul2:min:nom-irr-kill-pst dead

�You wanted to kill him.�

(20) Nga-la-ma-na-ji miliya marlu nga-la-ma-na. nyikina1:min:nom-irr-go-pst-exp now not 1:min:nom-irr-go-pst

�I was going to go this morning but I didn�t go.�

(d) The speaker presents the situation as one that ought to have obtained, or shouldhave obtained, but didn�t, due to unforeseen or unexpected circumstances. Itcould be that the evaluation is epistemically based as in (21), from a Nyikinabukarri (Dreamtime) story concerning a cheeky boy who became the willywagtail. A deontic basis for the �ought� is also possible, as shown by (22),where the event of going to Derby was an obligation on the addressee.

Page 17: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

354 oceanic linguistics, vol. 45, no. 2

(21) Kina wa-la-ni-na jayida marlu ngarri. nyikinasame 3:min:nom-irr-sit-pst 1+2:aug:obl not then

�He should have been the same as us (humans), but he isn�t�.

(22) Mi-li-jid-an Derby-ung karnambird. nyulnyul2:min:nom-irr-go-pst Derby-all other:day

�You were supposed to go to Derby the other day (though you didn�t).�

3.1.2.2 Nonpast irrealis. In Nyulnyul�and probably Jabirrjabirr and Niman-burru (see 2.2.2)�nonpast irrealis indicates unrealized situations belonging to possibleworlds that are temporally located either in the present or at some future time. Whichof the two temporal readings applies is left unspeci²ed, and can only be determinedfrom context, if at all. In contrast to the present realis, the situation is asserted as non-actualized; in contrast to the future, the speaker evaluates its occurrence as less likely�or perhaps more accurately, they do not commit themselves as strongly to its occur-rence, and refrain from pinning it down to any particular time. Thus nonpast irrealismay convey the contextual sense �might happen (at some unspeci²ed future time)�, asin (23) and (24).

(23) Nga-la-marr-karr jungk-in. nyulnyul1:min:nom-irr-burn-sub ²re-erg

�I might get burnt by the ²re.�

(24) Ngay a juy ya-li-rr-jird Derby-ung. nyulnyulI and you 1+2-irr-aug-go Derby-all

�You and I might go to Derby.�

Commands are prototypically expressed by the future realis in Nyulnyulan languages.The nonpast irrealis is not used in commands; however, it is used to convey the some-what weaker sense of �obligation�, reminding the addressee of a duty that they ought toful²ll. With second person subject, the utterance is less face-threatening than the realisfuture in that the event is placed in the realm of the possible or potential, allowing theaddressee a ready �out.� (25) is an example; notice that it admits construal as a requestrather than a command.

(25) Juy-in mi-la-w-ngay banangkarr arri jan.2:min-erg 2:min:nom-irr-give-1:min:acc today not 1:min:obl

�You should give me (money) today as I have none.� Nyulnyul

However, it need not be the addressee to whom the obligation applies.

(26) I-la-m kinyingk-in. Nyulnyul3:nom-irr-put this-erg

�He should put it.�

Examples (25) and (26) show that the referent situation may be positively, rather thannegatively, evaluated, as is not infrequently the case (see [23] and [27]). The malefac-tive sense is thus not a part of the meaning encoded by the future irrealis. Nonetheless,other things being equal, speakers tend to be more committed to the occurrence of a

Page 18: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 355

desirable event than an undesirable one, and the former tends to be represented by theplain future tense, the latter by the nonpast irrealis.

The same two senses are available to the future irrealis in languages that distinguishthe category, as illustrated by (27) and (28).

(27) Boowa-nim oo-la-rr-a-rl-a-goorr. bardiinsect-erg 3:nom-irr-aug-cm-eat-fut-2:aug:acc

�The insects might bite you.� (Aklif 1991a)

(28) Anggi arra arr mi-l-i Broome-ngan? bardiwhat/why not go 2:min:nom-irr-say Broome-all

Birrii mi-l-al-a-jiy.mother 2:min:nom-irr-see-fut-2:min:obl

�Why don�t you go to Broome? You should see your mother.�(Aklif 1991b:5)

The obligative sense of the future irrealis category is quite poorly attested in EasternNyulnyulan languages. (29) is one of the few examples available that admit this inter-pretation�it is more likely that the speaker would be questioning their obligation to gorather than their desire.

(29) Ngadyi-gardu nga-ya-rnda? yawuruinter-still 1:min:nom-irr-go

�Should I go?� (Stokes 1979:8)

Like Bardi, all Eastern Nyulnyulan languages maintain a contrast between realis andirrealis in the future. Opinions differ as to the basis of the contrast, which may, ofcourse, also differ from language to language. Thus, according to Hosokawa(1991:143�44), in Yawuru, irrealis future emphasizes the undesirability of the event, incontrast to realis future which indicates its desirability or the intention of the Actor tocarry it out. For Nyikina, Stokes (1982:282) suggests that the irrealis conveys a greaterdegree of uncertainty than the realis concerning the occurrence of a future event. How-ever, the examples she cites all admit a malefactive interpretation, as illustrated by (30).The same typically holds in Warrwa, as illustrated by (31); however (as in the Westerngroup) this is not an invariable interpretation, as shown by (32).

(30) Juwa nyi-rra-ni ma-kula-n-junu � nyikina2:min 2:min:nom-fut/irr-sit inf-tie-prs-abl

wa-na-rra-kula-yi linyju-ni.3:min:nom-cm-irr-tie-2:min:acc policeman-erg

�You might get tied up � the policeman might tie you up.�

(31) Kinya yila bilimu-jun wi-na-mungka-ngayu. warrwathis dog savage-abl 3:min:nom-irr-bite-1:min:acc

�This dog is savage; it might bite me.�

(32) Ngayu ngurranya-n-ngurndany nga-ya-wula. warrwaI tomorrow-loc-indf 1:min:nom-irr-come

�Maybe tomorrow I might come.�

Page 19: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

356 oceanic linguistics, vol. 45, no. 2

Present irrealis in Nyikina and Warrwa is almost exclusively found in negative clauses,in reference to situations asserted as nonoccurring as of the time of speaking. (33) is arare example of present irrealis in a positive clause. As in the past irrealis, it seems thata presupposition or expectation of the situation�s occurrence is invoked, against whichthe nonoccurrence is contrasted. This leads readily to the deontic modal sense apparentin the example.

(33) Mi-la-lungga kinya nganka. warrwa2:min:nom-irr-understand this word

�You should understand that word.�(Speaker�s gloss: �You gotta understand that word.�)

The (perhaps frustrated) desiderative sense is invoked in the following Bardi example.

(34) Gaadiliny nga-l-arli-n laalboo-yoon. bardimonkey.²sh 1:min:nom-irr-eat-prs earth.oven-abl

�I would like to eat the monkey ²sh from an earth oven.�(Bowern 2004:215)

3.1.2.3 Overview of senses of irrealis in positive clauses. Table 4 overviewsthe senses of the irrealis in positive clauses identi²ed in the preceding subsections. Asfar as we have been able to tell, the senses of nonpast irrealis in languages that makethe two-way tense contrast cover the domain covered by present and future irrealis in

TABLE 4. SYSTEMATIZATION OF THE MAIN CONTEXTUAL SENSESOF THE IRREALIS IN POSITIVE CLAUSES IN DIFFERENT TENSES

IN NYULNYULAN LANGUAGES†

� Greyed cell indicates only sense available in Yawuru, according to Hosokawa 1991.

past nonpast

present future

irrealis realis

almost, nearlyhappened (prevented at last moment)

occurrence of event iscontingent oncircumstances

no quali²cations on occurrence of event; not contingent

thwarted attempt (desired and attempted)

uninformed expectation�speaker uncommitted to occurrence of situation

informed evaluation�speaker committed to occurrence of situation

desired (notnecessarily attempted)

desired (perhapsno opportunityto achieve)

occurrence of eventis undesirable

occurrence of eventis desirable

ought to havehappened (desirable; evaluated as expected, or a social obligation)

ought to do or be doing (obligationon Actor; situationis expected)

should happen later (social obligation on Actor)

must happen (strong obligation)

polite request command, demand

Page 20: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 357

languages that make a three-way contrast. Table 4 also contrasts the future irrealis withfuture realis, elaborating on remarks in the previous subsection. For the other realis cat-egories, we merely observe that the senses of the corresponding irrealis tenses (desire,attempt, and so on) may or may not be invoked.

3.1.3 Irrealis and other modal particles. We have already seen that, as a clausalnegator, the negative particle demands the irrealis mood in Nyulnyulan languages. Thisis not the only modal particle that occurs with the irrealis, however. Nekes and Worms(2006:255�58) list a number of modal particles in various Nyulnyulan languages,including (apparent) epistemic modals of probability and the apprehensional particle�lest� (see 3.2). Epistemic modals are restricted to the nonpast or future irrealis,depending on the language; they are incompatible with past irrealis (see 4.2 for anexplanation). The following illustrate use of epistemic modal particles.

(35) Gonad garbor ¥a-l-e-djed bindan mo¥-o¥. nimanburru¥or-edj garor ¥a-l-ed bindan mo¥-¥an. barditomorrow perhaps I-irr-e-go bush honey-for

�Perhaps I will go into the bush for honey tomorrow.�(Nekes and Worms 2006:256)

(36) Ni¥ara-do wa-r-a-bula wandal. nyikinalikely it-irr-a-come boat

�It is likely that the boat will arrive, the boat might arrive.�(Nekes and Worms 2006:257)

One might question the accuracy of the above translations, and hypothesize that thespeaker is expressing doubt about the occurrence of the situation, that the translationsevaluate as more likely the nonoccurrence of the situation. However, we suspect thatNekes and Worms (2006) are fundamentally correct in their translations: the inclusionof the epistemic particle seems to make explicit the basis for the evaluation of the situa-tion as unrealized, ruling out other contextual interpretations (see 3.1.2.2); this interpre-tation is further supported by example (6.42) in Bowern (2004:202) and examplescited in Aklif (1991a). We return to this point in 4.2 below.

At least one of these particles, namely the Bardi particle gorror �perhaps, if� (Aklif1991a gives the second gloss) can occur with both realis and irrealis moods in thefuture tense (we are grateful to Claire Bowern [pers. comm.] for bringing its correctphonological form and examples to our attention). How these choices contrast inmeaning is uncertain. One other epistemic particle Nekes and Worms (2006) exem-plify, nyanangkarr �perhaps� (Nyulnyul) and cognates in Bardi, Jabirrjabirr, andYawuru, is also attested with both realis and irrealis futures. Although in this case againthe meaning difference is uncertain, the few examples available lead us to suspect thatthe difference lies in the degree of commitment expressed in the occurrence of the situ-ation�more for the realis than the irrealis.

As already stated, epistemic modal particles are precluded from clauses in the pastand present irrealis. But for these categories at least one particle other than the negativeis permissible, namely the evasive particle �nearly, almost (but didn�t)�:

Page 21: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

358 oceanic linguistics, vol. 45, no. 2

(37) Miliyarri22 nga-l-yanba-na kinya juurru ngayu-na. warrwanearly 1:min:nom-irr-step:on-pst this snake I-erg

�I nearly stepped on the snake.�

(38) Djolog i-n-ar ¥ambal bada¥g-en, jabirrjabirrkick 3:nom-cm-spear my:foot tree-erg

bana¥gar ¥a-le-djalg.nearly 1:min:nom-irr-fall-pst�I stumbled over a stump, I nearly fell.� (Nekes and Worms 1953:351)

(39) ¥a-n-djudar badag ¥am-bal-a¥, bardi1:min:nom-cm-kick stump my-foot-ins

milar ¥a-l-alg-on.nearly 1:min:nom-irr-fall-pst

�I stumbled over a stump so that I nearly fell.�(Nekes and Worms 1953:515�16)

3.2 THE IRREALIS IN COMPLEX SENTENCE CONSTRUCTIONS.In Nyulnyulan languages the irrealis mood is often found in complex sentences,though given the discussion of 3.1 it cannot be regarded as a marker of subordination.In the following subsections we look at the attested range of complex sentences involv-ing the irrealis, distinguishing three main types�apprehensional, conditional, andcomplement constructions�and a few minor or marginal types. Although we refer tothese complex sentence types as �constructions,� it must be stressed that we use theterm loosely. There is no implication that each represents a distinct and separate lin-guistic sign or construction in the sense of Construction Grammar (Goldberg 1995). Inmost available instances apprehensional constructions are paratactic, while conditionalconstructions are hypotactic.23 But they are unlikely to be grammatically distinct fromother paratactic and hypotactic constructions. The complement construction wouldseem to be grammatically distinct from both, and is neither paratactic nor hypotactic.The reader is cautioned that our knowledge of complex sentences in Nyulnyulan lan-guages is far from adequate, and virtually nothing is known about prosodic features.Moreover, few sources discuss the topic in any depth.

3.2.1 Apprehensional constructions. Apprehensional or �lest� constructionsdescribe a potential but undesirable situation, and the means by which its occurrencecan be prevented�the evasive action that could be taken. Whereas some Australianlanguages have a verbal in³ection that speci²cally marks the subordinate clause in an

22. This evasive particle is in Warrwa and Nyikina identical in form to a temporal adverb meaning�long ago�, while in Jabirrjabirr it is identical with the word for �now, today� (according to Nekesand Worms 1953:351 �nearly� is one of the senses of this word in Nyulnyul, Nimanburru, andBardi as well, though this has not been independently con²rmed). Although these perhaps repre-sent a single lexeme in each language, this lexeme has two quite different uses, as a temporaladverb and as a particle. The Bardi form milarr(a) (see example [39]) is cognate with Warrwamiliyarri (the same form occurs in Nyikina), deriving from *miliyarri by regular historical pho-nological processes. (We are grateful to Claire Bowern [pers. comm.] for pointing this out to us.)

23. In Nyikina the evidence suggests both types are (or may be) hypotactic, and involve differentlinking enclitics. We suspect that in other Nyulnyulan languages the association betweendependency type and semantic type is probabilistic.

Page 22: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 359

apprehensional construction (Dixon 1980:380�81), in Nyulnyulan languages this isone of the functions of the irrealis. The ²rst clause usually denotes a preemptive actionto be undertaken in order to avert the undesirable situation referred to by the followingirrealis clause. The verb describing the preemptive event is often in future tense; inmost examples the Actor is, or includes, the addressee, and the construction indicatesaction that should be undertaken to avoid the undesirable consequences, as in (40) and(41). The evasive action can be a situation to be avoided, and is thus denoted by a neg-ative irrealis clause, as in (42).

(40) Doogoo a-n-j-a baawa, bardiwipe 2:min:nom-cm-say-fut child

inkoorr-nim oo-l-iiding-a.cold-erg 3:min:nom-irr-touch-fut

�Wipe the kid dry lest he catch a cold.� (Aklif 1991a)

(41) Wai! Ya-¥ge-djaredjar, ya-mor ya-le-bed jabirrjabirraway 1+2:min:nom-fut-rise 1+2:min-backs they-irr-burn

dju¥g-en.²re-erg

�Away! We have to stand up, lest our backs get burnt.�(Nekes and Worms 2006:255)

(42) Malo mi-n-era-ma wandal ga ¥anal-an, nyikinaAre mi-l-am wilal dje bindj-en, jabirrjabirrnot 2:min:nom-irr-put coolamon that edge-loc

wa-ra-djalge.i-le-djalg.3:nom-irr-fall

�Do not put the coolamon at the corner, it might fall.�(Nekes and Worms 2006:255)

The ²rst clause may denote an event performed so that the undesirable situation did notoccur. In this case the construction is limited to circumstances in which the undesirableevent did not occur. As far as we can tell, the tense of the irrealis verb consistentlyemploys the reference point of the speech situation; it does not shift to the time of the ref-erent evasive action. Thus in (43) the undesirable event is in past irrealis, the situationbelonging to the realm of past time.

(43) Gajoordoo ing-algi-na, bardiash 3:min:nom-hide-pst

arr-amb oo-l-ala-na ngaarri-nim.not-rel 3:min:nom-irr-see-pst devil-erg

�He hid in the ashes (so) the devil wouldn�t see him.� (Aklif 1991b:12)

There is one set of exceptions to this pattern. In Nyikina an apparently hypotacticapprehensional construction exists in which the second clause, the dependent clause, ismarked by the �expectation� marker -ji, and is in the future irrealis, as shown by (44)

Page 23: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

360 oceanic linguistics, vol. 45, no. 2

and (45).24 It would seem that in such examples the time of the situation to be avoidedis according to the reference point of the ²rst situation.25

(44) Mundu ya-rr-a-ma-n buru-ngany war nyikinabury 1:aug:nom-aug-cm-put-prs sand-ins burn

wa-rra-ni-ji.3:nom-fut:irr-sit-exp

�We cover (it) with sand for fear a bush²re will start.�(Stokes 1982:329)

(45) Jin yi-n-mi-ny-janu kurrak wa-n-a-rr-i-ji. nyikinasignal 3:nom-cm-give-pst-1:min:obl away 3:nom-cm-fut-irr-say-exp

�He made a signal to me lest (the emu) run away.� (Stokes 1982:276)

As illustrated by (43) Nyulnyulan languages show a range of particle-like words thatcan occur in the irrealis clause in the apprehensional construction. Their exact sensesand usage are not understood well.

3.2.2 Conditionals. Conditional constructions are of two main types: counterfac-tual and hypothetical. In both cases, the consequent situation (apodosis) is unrealized,and represented by a clause with verb in the irrealis. The verb of the protasis clause isalso in the irrealis, and may be marked by an enclitic indicating the subordinate statusof that clause; alternatively, a particle such as Bardi gorror �perhaps, if� may occur inthe protasis clause. Neither enclitic nor particle is obligatory, and in their absence itseems that the construction is paratactic.

Counterfactual conditionals have both clauses in the irrealis. The apodosis is in thepast irrealis and speci²es a nonoccurring event the occurrence of which allegedly guar-antees the occurrence of the consequent event.26

(46) Jimara wi-l-wani-na-yarri now, wi-la-manda-na-yadirr. warrwaclose 3:nom-irr-sit-pst-sub now 3:nom-irr-saturate-pst-1+2:aug:acc

�If it had come up close, it would have wet us.�

(47) Nga-li-jal-an-karr-ji kalb nyulnyul1:min:nom-irr-see-pst-sub-2:min:acc up

24. The status of -ji exp, which indicates that the event is expected (though it is unspecific as to whoseexpectation), is uncertain. According to Stokes (1982:327) its formal identity with the dat postpo-sition -ji is accidental: the postposition, though not the verbal enclitic, shows lenition of the initialstop to the corresponding glide. This need not, however, argue in favor of homophony of the twoforms�the boundary type could condition application of the lenition rule. Another possibility isthat -ji is cognate with the simultaneous action verbal suffix in Bardi, as observed by Claire Bow-ern (pers. comm.), although in Nyikina, by contrast, it specifies that the event is subsequent to thereference event. Given this uncertainty, we have opted for Stokes�s analysis.

25. Evidence for this claim is not watertight. It is equally possible that in Nyikina as in other Nyul-nyulan languages the evaluation of the second situation is in accordance with the reference pointof the speech situation. Thus in both examples the situation could well be evaluated as futureirrealis with respect to the time of speaking. Moreover, consistent with this interpretation, the²rst clauses are either in present or what Stokes (1982) refers to as the �recent past� tense.

26. In Yawuru, where the irrealis is restricted to the future, it seems that counterfactual condition-als involve realis forms for the antecedent, with the verb either in past perfective or future(Hosokawa 1991:474�75).

Page 24: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 361

nga-li-m-an-ji mudikard-uk.1:min:nom-irr-put-pst-2:min:acc car-loc

�If I had seen you, I�d have picked you up in the car.�

While the Warrwa and Nyulnyul corpora show this pattern consistently, in Nyikina thefollowing example (the only instance of a counterfactual conditional we have found inthat language) shows the antecedent in realis:

(48) Yim-bula-ny-jarri nga-la-ba-na. nyikina3:min:nom-come-pst-sub 1:min:nom-irr-see-pst

�If he had come, I would have seen (him).� (Stokes 1982:321)

In Warrwa, if the antecedent is a condition that is hypothesized not to have occurred, con-trary to reality, it is expressed by means of an ordinary negative clause, as in (49). Thismode of expression may seem (as pointed out to us by Søren Friis, pers. comm.) to con-tradict the story developed in this paper for the irrealis, in as much as it is used in refer-ence to an event that actually did occur; we return to this point in section 4 where wesuggest an explanation. No examples of this type of counterfactual have come to lightfrom other Nyulnyulan languages, so we do not know how this sense is expressed.

(49) Marlu ngayak nga-la-ndi-na marlu wi-la-nga-ngay. warrwanot ask 1:min:nom-irr-say-pst not 3:nom-irr-give-1:min:acc

�If I hadn�t asked him, he wouldn�t have given it to me.�

In future conditionals the antecedent and consequence are both in the future or nonpastirrealis (depending on the language), and the antecedent is optionally marked by the subor-dinating enclitic. Examples are (50) and (51). In Warrwa, as in counterfactual conditionals,the antecedent can be a hypothetically nonoccurring situation, in which case it is repre-sented by a negative clause, as in (52). The future conditional contrasts with complex sen-tences in which the verbs in both clauses are in the plain future, in which case a futuretemporal reading (�when�) is invoked, or the events are construed as less hypothetical.

(50) Juwa-na mi-n-ka-yarri minya kinya-na wi-n-ka-yu.2:min-erg 2:min:nom-irr-hit-sub ?? this-erg 3:nom-irr-hit-2:min:acc�If you hit him, he�ll hit you back.� warrwa

(51) Mi-li-jid-ikarr kinyingk-ung bur i-li-rr-ar-juy.2:min:nom-irr-go-sub this-all camp 3:nom-irr-aug-spear-2:min:nom�If you go into that country, they might spear you.� nyulnyul

(52) Marlu mi-na-ma-yarri-yina baalu kud wi-ya-nda.not 2:min:nom-irr-put-sub-3:min:obl wood dead 3:nom-irr-go

�If you don�t put the wood on the ²re, it could die.� warrwa

3.2.3 Complement constructions. Mistaken thoughts regarding the past and doubt-ful or uncertain thoughts about the future can be represented by �thought� complementsin which the complement clause is in the irrealis. The complement clause thus representsthe cognisized situation from the perspective of the speech situation, and the constructionshows the hallmarks of indirect rather than direct thought reporting.

Page 25: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

362 oceanic linguistics, vol. 45, no. 2

In Nyulnyul, mistaken thoughts are framed by a verbless clause involving thepre²x-taking -mungk �believe�. Although -mungk �believe� often behaves like a modalparticle, it can belong to a separate clause with other grammatical roles ²lled out, as in(53). (Examples exist with an ergatively marked NP denoting the cognisant.)

(53) Angk-ij nyi-mungk nga-la-w-an juy kumbarrwhat-dat 2:min-believe 1:min:nom-irr-give-pst you money

nyi-mungk arri nga-la-bakad-an. nyulnyul2:min-believe not 1:min:nom-irr-have-pst

�Why did you think I would give you money when you knew I had none?�

Doubtful or uncertain thoughts about the future can be expressed in some Nyulnyulanlanguages by a complement clause in the future or nonpast irrealis, the matrix clauseinvolving the generic �say�, �do�, �think� verb. The only examples of this sort availableare a few cited in Nekes and Worms (2006).

(54) Waber ¥an-den ni¥ara-do wandal wa-ra-bula njinagara.doubt 1:min:nom-say likely boat 3:nom-irr-arrive this.time

�I doubt whether the boat will arrive this time.� nyikina(Nekes and Worms 2006:257)

(55) Wabar ¥an-den djer wamborinj ¥adjegad dar e-le-r-aÉ.doubt 1:min:nom-say them people perhaps arrive 3:nom-irr-aug-spear

�I doubt whether the people will come.� nyulnyul(Nekes and Worms 2006:257)

Both complement types express meanings that can also be expressed by particles (see3.1.4), albeit often with less precision�for instance, in (54) and (55) with just the par-ticle it would remain inexplicit who is responsible for the doubting. McGregor (1997,in press) argues that the grammatical relation between matrix and complement clauseis the same as that between a modal particle and the rest of the clause.

3.2.4 Other uses in complex sentences. A range of other interpretations areavailable for complex sentences in which an irrealis clause is juxtaposed to anotherclause; these interpretations seem to be engendered by the context, implicated ratherthan encoded (at least in the absence of prosodic information). In some instances it isthe second clause that speci²es a condition under which the ²rst holds, or the range ofits application�how it is to be interpreted, as in (56).

(56) Marlu mawu nga-la-ngula-na jinal. warrwanot good 1:min:nom-irr-throw-pst spear

�It�s not good if I had thrown a spear.�(Elicited in response to �It�s not good to spear people.�)

A dependent clause marked by the subordinate enclitic can follow rather than precedethe main clause, and can be interpreted either temporally (i.e., �when still unrealized�)or conditionally:

Page 26: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 363

(57) Nga-n-andi-ny badkurru wa-rra-bul(a)-arri kinya. nyikina1:min:nom-cm-pick.up-imp kangaroo 3:nom-irr-come-sub this

�I got a kangaroo before it came.� (Stokes 1982:282)

(58) War nga-rra-ni jungk-an-ka bilika nga-rra-n(i)-arri. nyikinaburn 1:min:nom-irr-sit ²re-loc-emph close 1:min:nom-irr-sit-sub

�I might get burned if I stop close to the ²re.� (Stokes 1982:283)

The clause following an irrealis clause can provide an explanation or excuse for theevent�s nonoccurrence, or a restatement or elaboration of the irrealis clause, as in thefollowing two examples, respectively.

(59) Nga-li-jid-an Mowanjumbirdi yubul nga-ng-in. nyulnyul1:min:nom-irr-go-pst Mowanjum yesterday sick 1:min:nom-be-imp

�I might have gone to Mowanjum yesterday, but I was sick.�

(60) Nga-la-ma-na-ji miliya marlu nga-la-ma-na. nyikina1:min:nom-irr-go-pst-exp now not 1:min:nom-irr-go-pst

�I was going to go this morning but I didn�t.� (Stokes 1982:330)

The few examples discussed here illustrate the wide range of possible interpretationsof complex sentences involving the irrealis. We could continue identifying further and²ner distinctions ad nauseam, though with vanishing returns for the effort.

4. ACCOUNTING FOR THE MEANING OF THE NYULNYULAN IRREALIS

4.1 SEMANTICS AND PRAGMATICS OF IRREALIS IN NYULNYULANLANGUAGES. We have seen that two components of meaning are recurrent acrossthe range of uses of the irrealis mood in Nyulnyulan languages: that the event belongs, inthe speaker�s view, to the realm of the unreal or unrealized, and that it might haveoccurred or might occur, that it was or is a potentiality. The latter modal aspect of themeaning is what gives rise to the various contextual interpretations such as �try but fail�,�almost happen�, �want to�, and so on; the former component of meaning excludes thepossibility that the situation went on to occur. The question we now address is whethereither or both components of meaning are encoded by the irrealis, and represent the coremeaning of the category, which remains invariant across all uses.

We make the following two proposals. First, the meaning component �unrealized�(speci²cally in the speaker�s construal) is a semantic invariant of the irrealis, and repre-sents the core meaning of the category. Second, the sense of potential associated withthe irrealis is not a semantic invariant associated with the category itself, but is associ-ated with, and encoded by, the use of an independent clause in the irrealis�that is tosay, it is coded by the speech act performed by the clause. We consider these two prop-ositions in order in the following subsections.

4.1.1 The semantic component +unrealized. We have claimed that the irrealiscategory invariably expresses the meaning component +unrealized. This is demon-strated by the examples discussed in section 3, and the observation that we have foundno convincing counterexamples to the claim. In ²eldwork on Nyulnyul and Warrwawe attempted by every means we could think of to elicit examples in which the +unre-

Page 27: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

364 oceanic linguistics, vol. 45, no. 2

alized sense does not obtain; these attempts were singularly unsuccessful. This senseappears to be ever-present, and nondefeasible.

We are aware of a single potential counterexample:

(61) Wa-la-(rr)-di-na-da idany-barri makarra. nyikina3:nom-irr-(aug)-sit-pst-hab long-com tail

�They used to have long tails (but they don�t now).� (Stokes 1982:281)

This example concerns certain creatures that had long tails in the Dreamtime, that nowhave short tails. Stokes (1982:281) suggests that this usage is motivated by thespeaker�s recognition that the situation in the Dreamtime con³icts with present realityrather than with past reality, as is normally the case. This usage would then represent ameans of highlighting such contrasts. According to Stokes�s story, what is shared by alluses of the past irrealis in Nyikina is not the feature +unrealized, but the counterfactualsense, which admits shifting from past to present.

The problem is that while this might account for the counterfactuality of the past irre-alis, it cannot be an invariant of meaning common to all uses of the category. It is incom-patible with the fact that future irrealis is noncounterfactual, and thus the counterfactualitysense is not shared by all uses of the category. Unfortunately, Stokes (1982) gives very lit-tle information on the context of the example, and it is impossible to be certain about howsuch examples should be accounted for. In any event, it is possible to interpret (61) inaccordance with our proposals. It is possible that this example invokes the desiderative orobligative sense of the irrealis (senses [c] or [d] in 3.1.2.1), and that the desired orexpected situation is not the possession of tails, but the situation of habitually havingtails�thus by implication keeping them. That is to say, the key to this example lies in theaspectual choice, which represents a part of the situation and lies within the scope of theirrealis. This implies that a more accurate interpretation of (61) would be �they wanted tokeep long tails (though something prevented them)� or �they ought to have kept long tails(but something they did resulted in their losing them)�.

We conclude that there are no irrefutable counterexamples to our claim, and that the+unrealized sense cannot be canceled pragmatically in Nyulnyulan languages. This is atvariance with proposals by Jean-Christophe Verstraete, who argues that in a range ofnon�Pama-Nyungan languages this meaning is an implicature (Verstraete 2005, 2006).Verstraete also adduces compelling arguments in favor of construing the counterfactualsense of the past irrealis as an implicature, rather than encoded semantics. On this matterwe are in full agreement; where we disagree is which aspect of meaning is encoded in theirrealis, and how the counterfactual implicature arises (see next subsection).

According to Verstraete (2005, 2006), the counterfactuality of the past irrealis arises viaa Gricean implicature, a Generalized Conversational Implicature (GCI). His argumentgoes as follows. First, the irrealis category encodes the potential sense, which is consis-tently associated with the irrealis. Second, clauses in the (plain) realis past and the irrealispast form a Horn scale, on which the realis clause is epistemically stronger than the irrealis.Moreover, assertion of the realis past constitutes an epistemic maximum, namely that thesituation occurred�it admits no room for doubt. By the ²rst Gricean maxim of Quantity(Grice 1989), or Levinson�s Q-principle (Levinson 2000), it follows that use of the

Page 28: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 365

epistemically weaker expression implicates the negation of the stronger expression.27 Thuswe can conclude that the event did not occur, and thus we have the counterfactual sense.

While Verstraete�s arguments may hold water in some non�Pama-Nyungan lan-guages (e.g., Rembarrnga and Jaminjung), we dispute their applicability in Nyulnyulan.There are two major dif²culties as far as Nyulnyulan languages are concerned. One isthat particles expressing probability such as nyanangkarr �perhaps, maybe� (Nyulnyul)occur in the past realis, as demonstrated by (62).

(62) Nyanangkarr kinyingk-in i-na-ngurl kinyingk walangk.perhaps this-erg 3:nom-cm-throw this spear

�Maybe he threw the spear.� nyulnyul

Applying the above reasoning, one would conclude by the Q-principle that the mandid not throw the spear. This is not so. (62) is neutral in regard to truth or falsity of theproposition �he threw the spear�. There is no reason to expect different behavior for themodal particles of probability and the irrealis in Verstraete�s scenario, yet they are quitedistinct in terms of their effects in the past tense.

The second problem is that there are no compelling examples of cancellation of the+unrealized sense of the past irrealis, and thus it seems that the sense is encoded rather thanimplicated. Verstraete (2005:241�42) admits that this is so in some non�Pama-Nyunganlanguages, but disputes the relevance of this fact, proposing that nondefeasibility of thesense does not argue against his proposal. Here he recontextualizes his arguments asdiachronic rather than synchronic, and observes that an erstwhile implicature can becomepart of the semantics, as observed by Traugott and Dasher (2001), for example. We are infull agreement. However, this observation yields precisely our point: nonactualization is apotentially codable feature, and is actually encoded by the irrealis in Nyulnyulan languages.

4.1.2 The meaning component of potentiality. In contrast to Verstraete (2005,2006), we argue that potentiality is not a part of the meaning coded by the irrealis cate-gory itself, but rather is expressed by speech-acts enacted by independent clauses in theirrealis. The idea behind this proposal is that, like use of negative polarity, use of irrealismood is pragmatically marked, and highly context-dependent. But whereas a negatedclause invokes as a presupposition the corresponding positive (see 3.1.1), which is thendenied, a positive clause in the irrealis invokes the potential occurrence of the corre-sponding realis situation, which is then stated as unrealized. In our view, these presup-positions are in each instance encoded by the speech-act itself, rather than beingmerely implicated by it. That is to say, the presuppositions are invoked�coded�bythe negative and irrealis speech-acts, be they statements, questions, commands,requests, or whatever. They are not inferences drawn by GCIs, and are not defeasible.By contrast, speech-acts expressed via independent realis clauses encode neither typeof presupposition, and are unmarked with respect to the potential sense.

27. We also disagree on the status of the �epistemic maximum� associated with the assertion of aclause in the realis: we suggest that it is no more than a pragmatic implicature, invoked inusual circumstances. It can be defeated by addition of epistemic particles, and in questions.Moreover, as any historian knows, the past is as much a site of contention as the future is, andone�s knowledge of it is no more certain than one�s knowledge of the future.

Page 29: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

366 oceanic linguistics, vol. 45, no. 2

This semantic feature permits an explanation of the counterfactual sense of the irrea-lis, which (as already stated) is not an invariant of meaning associated with the category.Speci²cally, the irrealis mood codes the meaning +unrealized, while the use of the clausein the irrealis as speech-act retroactively presupposes the potentiality of the situation. Sit-uating this in the past or present, it follows that the situation did not occur or cannot beoccurring now, while at the same time it potentially could have occurred or could beoccurring. This gives rise to the counterfactual sense. This sense does not arise with thefuture irrealis. The modal category itself speci²es nonrealization of the situation, while itsuse in a speech-act codes the situation�s potentiality. There is no con³ict between non-realization and potentiality; the counterfactual sense does not arise.

Both positive clauses in irrealis mood and negative clauses (invariably in the irrea-lis) encode the meaning that the situation did not occur, the feature +unrealized beingassociated with the irrealis category in each. But the two modes of expression are notsynonymous, and an account of the contrast in meaning between them is afforded bythe semantic features we have suggested to be encoded by the speech-acts correspond-ing respectively to negative clauses and positive irrealis clauses. In particular, accord-ing to our proposal, not only is there a crucial asymmetry between a negative clauseand the corresponding positive clause (see 3.1.1), but also between a negative clauseand the corresponding positive irrealis clause. This falls out as a consequence of theloci of the usage values in the speech-act, rather than in the irrealis category itself. Wedo not need to invoke for the negative speech-act the presuppositions associated withthe speech-act corresponding to the positive irrealis clause: negative speech-acts do notinvolve assertion of positive irrealis clauses.

This explains why and how negative clauses and positive irrealis clauses invoke dif-ferent types of presupposition, and therefore contextual senses. The negative clauseinvokes presuppositions concerning propositions entertained by the speech participants,while the positive irrealis clause invokes expectations about the occurrence of situations.These not infrequently coincide, due to the implicational links between truth value ofpropositions and occurrence of situations. But not always: negative clauses like (11)above invoke none of the presuppositions of the corresponding positive irrealis�that theevent almost happened; that an attempt was made by the Actor; that the Actor desired theevent; or that the event ought to have occurred. Similarly in the nonpast, as illustrated in(63). Here again, the senses invoked by the corresponding positive irrealis do not arise.The speaker is not suggesting the contingency of the situation�its potentiality, its unde-sirability, that it should happen�or making a polite request. Rather, in both instances thenegative clause invokes and denies propositional expectations. The negative clauses andtheir corresponding positive irrealis clauses (which do invoke these situation-related con-textualizations) are not synonymous in these instances.

(63) Ngay-in arri nga-li-j-jii. nyulnyuli-erg not 1:min:nom-irr-say-2:min:obl

�I won�t tell you.�

One remaining problem is dispelled by this theory. This is the problematic case of Warr-wa negative conditionals in complex sentences, such as the counterfactual (49) and pro-

Page 30: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 367

jected (52). In these instances, the negative clause does not denote a situation that, in thespeaker�s opinion, did not happen or will not happen, but rather one that, according to thespeaker�s evaluation, did (or will) happen. Thus the speaker did ask the person for moneyin (49), and the addressee presumably will (now, as a result of the speech-act) put woodon the ²re in (52). The key to understanding these examples lies in the observation thatthe speech-act performed by the utterance of the dependent negative clause is not thesame as in the utterance of independent negative clauses. The speech-act is not one ofassertion (used here as a cover term for statement, question, command, etc.), whereby thespeaker attests to the truth value of the proposition: the speaker does not assert the apodo-sis, but rather hypothesizes it, puts it forward as a notion to entertain. This meaning, thatthe proposition is entertained or hypothesized, it should be noted, is not coded by the irre-alis�as we have seen, the irrealis is perfectly consistent with assertive speech acts. (Wesuspect�though this remains to be con²rmed�that the difference between the speech-acts of assertion and guessing is conveyed by intonation.)

The hypothesis expressed by the apodosis is contrary to known or believed fact�itcon³icts with the actual world. This arises as a scalar implicature: if a speaker hypothe-sizes a proposition, this is for a reason: it con³icts with known, believed, or expected real-ity (otherwise, one would simply assert it). Thus the negative polarity clause represents acounterfactual situation in the past, and one that goes against expectations in the future.Here it is the speaker�s construal of the apodosis situation as unrealized (didn�t happen)that is crucial, not the actual realization-value, realized (did happen). In keeping with thisinterpretation, we observe that if the nonassertion of the apodosis is interpreted instead notas a guess, but as a presumption, we obtain a temporal interpretation, and the apodosis isnot counterfactual (�when not�). These alternative interpretations appear to be availablefor examples such as (49) and (52): �when I didn�t ask him (for money), he didn�t give itto me�, and �when you don�t put wood on the ²re, it will go out�. In other words, it seemsthat the counterfactual and counter-expectation interpretations of examples like these arebut one interpretation, and are defeasible. This contrasts with the positive conditional, inwhich the apodosis always represents a counterfactual situation, or one not expected, thecorresponding �factual� or expected situations being represented by a realis clause.

4.2 SCOPE. The notion of scope is crucial, we would argue, to an understanding ofthe irrealis mood in Nyulnyulan languages. The two components we have identi²edfor the irrealis categories, namely the modal and tense components, both hold the situ-ation (the event along with its participants and circumstances) in their scope. However,they modify the situation independently, each providing a different �slant� on it. Ratherthan one having scope over the other, they operate orthogonally. Speci²cally, we pro-pose that: (a) the irrealis mood pre²x speci²es that the referent situation was unreal-ized; and simultaneously (b) the tense component grounds the situation, giving rise to aproposition or possible fact, something that can be true or false and that can be arguedabout (see, e.g., Halliday 1985:72; Dik 1989:202; Van Valin 1993:8; McGregor1997:238).

Evidence for the narrow situation�rather than proposition�scope of the irrealis isthat the contextual senses always concern the occurrence of situation, singling out fac-

Page 31: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

368 oceanic linguistics, vol. 45, no. 2

tors relating to this (that an attempt was made, that it is intended, that it is undesirable,etc.). It does not modalize the proposition expressed by the grounded tensed clause,presenting the speaker�s stance on its truth value or knowledge status�or rather, itdoes so inconsistently and indirectly via implications and implicatures.

This argues that, in Nyulnyulan languages, the irrealis does not hold tense in its scope.Perhaps tense holds the irrealis in its scope? One argument against this possibility comesfrom negation. Clausal negative particles in Nyulnyulan hold the proposition�the tensedsituation�in their scope, specifying that it is false. But they do not hold the irrealis intheir scope. For if they did, the meaning expressed would be �it is not the case that [S notoccur]� (where S is the referent situation), which logically implies that S did occur, con-trary to the meaning expressed. Since the irrealis falls outside of the scope of the negativeparticle, while tense falls within it, it follows that irrealis cannot fall within the scope ofthe negative. Irrealis and tense thus scope the situation orthogonally, as depicted in²gure 2, which shows the structure of example (1) according to this proposal.

This approach also accounts for the uses of modal particles. First, if the probabilityparticles hold the situation grounded in the future in their scope, this indicates an evalua-tion of the future situation as possible; the irrealis orthogonally expresses the meaning thatthe event is unrealized. So the net effect in examples such as (35) and (36) above is thatthe free translations given by Nekes and Worms (2006) are indeed correct�the meaningis �perhaps/probably the situation will occur�, not �perhaps/probably the situation will notoccur�. The effect of using the particle is to specify the motivation for the irrealis evalua-tion, precluding other interpretations such as that the situation is undesirable.

Second, consider what happens in the past and present. If tense had scope over theirrealis, then the probability particles ought to be able to scope over them both, toexpress the meaning �probably not�. But this does not happen.28 Assuming, on theother hand, that the tense and irrealis independently scope over the situation, and theprobability scopes only over tense, we have �probably/possibly [S occurred/is occur-ring]�, which contradicts �S is unrealized� (i.e., IRR[S]). Thus we are able to accountfor the incompatibility of the particles of probability with the past and present irrealis,and their compatibility with the future irrealis.

28. The notion �probably not� is expressed by use of the two particles, �probably� and �not�, inthat order; the verb is in irrealis mood, being selected by the negative particle.

FIGURE 2. SCOPAL RELATIONS IN EXAMPLE 1

Page 32: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 369

Third, as we have seen, the only particles consistent with the past and present irrea-lis are the negative and evasive particles. We have already dealt with the former. Theevasive particles specify that the situation almost occurred, but didn�t, and hence arecompatible with just past and present tenses. Assuming they also hold the situation intheir scope independently of the irrealis, it is clear that no contradiction arises, and theeffect is as in the case of probability particles with the future: the evasive particle expli-cates the basis for the irrealis speci²cation.

To conclude this section, it is stressed that we have been specifically concernedwith issues of scope in relation to the irrealis on the one hand and finite clause negationon the other. Negation in Nyulnyulan languages is a wider phenomenon, including NPnegation (see note 20) and, in some languages, negation of nonfinite verbs (includingboth preverbs and infinitival forms of IVs). These phenomena are not directly relevantto the concerns of the present paper, which is concerned primarily with the irrealis, andwith negation only in as far as it interacts with the irrealis; thus they are not discussed indetail. Suffice it to remark that negation in Nyulnyulan languages (with the exceptionof Yawuru) is consistently associated with either marked mood (finite clause negation)or no mood (NP negation, nonfinite clause/verb negation). A second thing we havedealt with inadequately is the focus of negation in finite clauses. The salient observa-tion is that the scope of both irrealis and clausal negation is always over the entire situ-ation represented by the clause, and not just the verb or verb plus a selection ofarguments and/or circumstantial roles.

5. THE NYULNYULAN IRREALIS IN A WIDER PERSPECTIVE

5.1 THE REALIS OF IRREALIS. The reality of irrealis mood as a possiblegrammatical category in human languages has been questioned by some investigators.Thus Palmer (1986:26) cautions against the use of the term, while Joan Bybee and col-leagues have in various publications (e.g., Bybee and Dahl 1989; Bybee, Perkins, andPagliuca 1994:236�40; Bybee 1998) argued vehemently against the viability and use-fulness of the irrealis as a grammatical category. Both Elliott (2000) and Verstraete(2005) challenge the rejection of the irrealis on the grounds that at least some lan-guages appear to exhibit a grammatical category in which reality status is a central con-cern. Givón (1995:168) presents a telling critique of Bybee�s arguments, which heboils down to the proposition: �Only cognitive-communicative categories that aremarked uniformly by a single language, or are grouped in the same way by most lan-guages, have mental reality.� Most linguists (including the present authors) wouldreject this extreme functionalism/mentalism. We do not propose to develop any ofthese counterarguments further: they are adequately dispelled in the three referencescited. Rather, we select ²ve points from the arguments of Bybee (1998) that on the onehand bring aspects of the Nyulnyulan irrealis into focus, and on the other highlight dif-ferences in our theoretical approaches.

First, Bybee rejects (1998:261) the essentialist Jakobsonian approach to grammati-cal categories, proposing prototype theory in its place, according to which experientialcategories are perceived in terms of central and peripheral members, as per Rosch(1973). It follows that not all senses of a grammatical morpheme are on equal footing,

Page 33: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

370 oceanic linguistics, vol. 45, no. 2

some being better exemplars of the category than others. Bybee�s discussion is set at ahigh level of generalization, as a rejection on theoretical grounds of essentialistclassi²cation via shared recurrent components of meaning. It does not address theissue of the existence or not of a recurrent invariant of meaning in the irrealis categoryin any particular language. The proposition that there is an invariant core meaningassociated with the Nyulnyulan irrealis category (as argued in section 4) remainsunchallenged. Nor, on the other hand, is it clear that the senses of the irrealis in Nyul-nyulan languages are organized in the Roschian fashion: they may be, but no evidencewe are aware of points to an organization in terms of center and periphery. Bybee(1998:267) in fact observes that this seems to be the case cross-linguistically, that noviable proposal has been made in the literature for characterizing senses of irrealis intocore and periphery, and doubts its likelihood. This can only be an argument against theviability of the irrealis as a grammatical category if all grammatical categories are orga-nized in a prototype way. There is reason to doubt this. The category of person, forinstance, would seem to be a counterexample, as Claire Bowern (pers. comm.) hasdrawn to our attention. So also is the irrealis category in Gooniyandi (McGregor1990:524�27) and Nyulnyulan (this paper): contra Bybee (1998:267) it is possible toidentify invariant core meanings for the irrealis in these languages.29

Second, according to Bybee (1998:265), �a Jakobsonian theory of oppositionsleads us to expect contrasting grams to be maximally different in meaning (as theoppositional terms �realis� and �irrealis� suggest).� While Jakobson may have operatedunder this simplistic view, it has been known to be inadequate since Trubetzkoy�s workon phonological systems in the 1930s (Trubetzkoy 1969). Few today hold to the pre-sumption that features in binary opposition must contrast maximally as positive (+) tonegative (�): the existence of members of an opposition that are unspeci²ed (±) for afeature are well known. In fact, there is no reason to presume from the labels �realis�and �irrealis� anything about the nature of the opposition. What we have proposed inthis paper is that in Nyulnyulan languages the irrealis can be characterized as showinga + value for the semantic feature, which we have labeled, for want of a better term,unrealized. But, as we have made clear, the irrealis is not in maximal contrast with rea-lis, to the extent to which the latter is recognizable as a category. Indeed, realis isunspeci²ed for this feature: it does not specify anything about the actuality status of thesituation. In the past and present realis, this is evident from the fact that modal particlesof probability can be added to specify lack of certainty, consistent with either realiza-tion or nonrealization of the situation. The epistemic maximum associated with anunmodalized clause in the past or present is a pragmatic implicature. The irrealis vs.irrealis contrast in Nyulnyulan is a contrast of + vs. ± feature values.

Bybee may well be correct that �there is not one case in which a grammatical distinc-tion corresponds directly to the notional distinction between real and unreal situations�

29. There is not necessarily any con³ict between the Jakobsonian and Roschian approaches: they areconcerned with different aspects of categorization. What Jakobson identi²es as the core meaningof a grammatical category is not an instantiated sense, but rather an abstraction from the range ofinstantiations. It is on a totally different footing from instantiated senses. It is the latter phenom-ena that are�or may be�organized in terms of center and periphery. Their organization in thisfashion does not imply that there is no abstract shared component of meaning among them.

Page 34: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 371

(Bybee 1998:266). But this argues for the rejection of the grammatical distinction of rea-lis and irrealis if and only if we assume features can have just + or � values.

This brings us to the third problem, namely that features should not be interpretednotionally, in accordance with the lexical meaning of the feature label. It is in the natureof grammatical (emic) categories that they do not correspond directly to nonlinguisticnotional (etic) distinctions; it is doubtful whether any instance of a direct correspon-dence exists. If this were used as a basis to refute grammatical distinctions, all gram-matical categories could be rejected. The meaning of the feature [unrealized] is not tobe derived from the meaning of English lexeme unrealized, or from everyday notionsof unreal situations. Rather, it must be understood in a language-particular way (whichshould correspond reasonably well with the sense of the English term, or otherwise abetter term is warranted). The task of the linguist is to attempt to come to grips with theemic semantics of the category, via investigation of the senses associated with uses ofthe category�not via logical consideration of the label�s semantics. How to explainthe meaning of the feature is a dif²cult problem on which the opinions of semanticistsdiffer. We have contented ourselves here with a very informal explication; otherswould argue the need for much more precise explications (e.g., Wierzbicka 1996;Goddard 1998). To attempt more precise formulations would require another investi-gation, and adherence to a semantic theory, on which we remain agnostic.

Fourth, Bybee (1998:268�69) observes that the existence of pragmatic meaningsfor the irrealis makes it dif²cult to isolate invariants of meaning associated with the cat-egory. This is indeed so: it is undeniably a very dif²cult task to separate that which iscoded from that which is implicated. However, the distinction is there�as much as aperson has ears, even though it remains dif²cult to draw a precise line between the earand the scalp. Indeed, the very process of drawing pragmatic inferences via Griceanmaxims and the like is dependent on the existence of coded meaning.

Implicit (or at best partly explicit) to Bybee�s critique is the presumption that unlessa necessary connection can be motivated from notional categories to use of the gram-matical category, the coding relation cannot exist (see Givón�s critique above). Thisputs the cart before the horse. Use of a sign is never predictable from the need or desireto express a notion: to move from prelinguistic notions to necessary expression by par-ticular linguistic forms is plainly impossible. If we accepted the dictum of Vidal andKlein (1998:182) that irrealis is only coded if the category can be predicted from the�function,� we could indeed conclude that there is no irrealis category in any language.We could also conclude that there is no category of any type in any language.

Finally, as a parting shot, Bybee (1998:267) suggests that the irrealis is a highlygeneral notion, that is probably too abstract to be of any communicative use. We agreecompletely with the ²rst proposition: the irrealis is indeed an abstract category. Butcontrary to Bybee, this is what makes it useful communicatively as a grammatical cat-egory. Concrete grammatical categories in our view would be far too restricted to be ofany use in communication. In our opinion the Nyulnyulan irrealis category is highlyuseful communicatively precisely for this reason; it encapsulates a wide range of phe-nomena under a single umbrella that would otherwise require separate treatment�thatis, additional grammatical resources. Moreover, it does so in a fashion that appears to

Page 35: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

372 oceanic linguistics, vol. 45, no. 2

be no less effective than the modal categories of familiar SAE languages, which are nomore concrete or semantically and pragmatically �pure�: deontic and epistemicmodalities are as nicely mixed in a modal porridge as they are in Nyulnyulan. Indeed,it seems to us that there are better grounds for viewing deontic and epistemic moods asproducts of �logical� or philosophical thought, and for questioning their reality as pos-sible grammaticalizable emic categories in human languages.

We conclude that the Nyulnyulan irrealis is a genuine grammatical category, andthus that the irrealis is one of the categories that can be grammaticalized in human lan-guage. This brings us to the next issue, namely the position of the Nyulnyulan irrealisin the grammatical space of irrealis categories.

5.2 NYULNYULAN IRREALIS IN A TYPOLOGICAL PERSPECTIVE.In sections 3 and 4 of this paper we identi²ed a range of properties relevant to thedescription of the irrealis category in Nyulnyulan languages. We suggest that theseproperties provide handles on the typological parameters according to which irrealiscan vary cross-linguistically. In our view (as in the views of Palmer 1986 and Bybee1998; see also de Haan 2006:41�45), the term irrealis carves out neither a single cate-gory nor a single grammatical domain. Different values may be assigned to the proper-ties, and by giving them different assignments we can characterize at least some of therange of cross-linguistic variation of the category. Contra Palmer and Bybee, the rangeof variation does not argue against identi²cation of irrealis as a category any more thandoes the vast variability among case-marking systems (as is revealed when one looksclosely at the range of facts in particular languages rather than a handful of carefullyselected examples) argue against identi²cation of nominative and accusative, or erga-tive and absolutive case systems. Putting things in a slightly different way, we are sug-gesting that the search for a uni²ed and universal cross-linguistic de²nition of theirrealis is a misguided exercise. Rather, it is preferable to conceptualize the irrealis ascarving out a domain of variation encompassing language-particular categories show-ing more or less resemblance to one another, and perhaps prototype effects.30 In ourview, the grammatical categories of a language are characterizable in essentialist terms;prototype effects may be discernible in the instantiation (tokens) of use. We do not pre-sume universal grammatical categories, but rather take the view that grammatical cate-gories are language-particular, and must be described in their own terms for eachlanguage. This does not mean that they vary limitlessly from language to language. Itis clear that they do not, that there are strong family resemblances among them; this iswhat makes linguistic typology doable, and interesting.

The main parameters of variation that our investigation of Nyulnyulan suggest are:(a) the contrast between realis and irrealis; (b) the semantics of the categories in oppo-sition; (c) scope; and (d) what other categories can be simultaneously chosen with theirrealis. A ²fth parameter, not immediately apparent, is (e) the locus of the category ingrammatical space. These parameters are not entirely independent, but to explore theirmutual dependencies is beyond the scope of this paper. We now look brie³y at the ²veparameters in turn.30. Strangely, while advocating rejection of essentialist semantics of the irrealis within a lan-

guage, Bybee (1998) appears to demand it for the grammatical category of irrealis.

Page 36: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 373

(a) nature of the modal contrast. We have suggested that the Nyulnyulanirrealis contrasts with realis as + (positive) to ± (vague). This is just one of a range of pos-sibilities, assuming the contrast to be binary, and the three values: +, �, and ±. Two ques-tions arise. First, do all of the logical possibilities actually occur? Our guess is probablynot: an unmarked category contrasting with a vague one (� vs. ±) and two vague catego-ries (± vs. ) would seem rather unlikely. A marked realis contrasting with an unmarked(either � or ±) irrealis category would seem less likely than unmarked realis and markedrealis (Elliott 2000), consistent with the cross-linguistically normal situation of markednegative against unmarked positive. But it is not outside the bounds of possibility, thoughwe are aware of no clear examples. Unfortunately grammatical descriptions are rarelysuf²ciently detailed to decide on the nature of the contrast.

The second question is: to what extent do the �functional� contrasts correlate withformal contrasts? That is to say, are there correlations between �functional� marked-ness and formal markedness? For instance, are equipollent oppositions (neither cate-gory functionally marked) restricted to cases where both categories have a formalmark, as in Caddo (Chafe 1995) and Muyuw (Bugenhagen 1994:18)?

Aside from this, it is questionable whether irrealis always enters into a binary para-digmatic contrast with realis. We have questioned this for Nyulnyulan languages.Another possibility is that irrealis is a member of a larger set of paradigmatic contrasts.For instance, in Valley Zapotec irrealis appears to contrast with subjunctive andde²nite moods (Munro 2006:175). And Gooniyandi has two marked moods, labeledsubjunctive and factive in McGregor (1990), along with an unmarked indicative. Thesubjunctive and possibly factive could be regarded as a type of irrealis, the indicative asa realis, giving a three-way contrast. (See also under [c].)

(b) semantics of the modal categories. This factor is obviously closely con-nected to (a), the semantic feature characterizing the irrealis being determined in partby the nature of the contrast with the realis. Thus if the irrealis is negatively speci²edwith respect to the realis, the semantics will not be the same as it is when the irrealis ispositively speci²ed: realized will be the operative feature. We believe, contrary to theoft-repeated dictum of Saussurean structuralism, that there is more to the semantics ofa category than its paradigmatic oppositions with other categories: it has genuine inher-ent content. Thus in languages such as Rembarrnga and Jaminjung where the irrealiscategory arguably encodes the potentiality rather than unrealized status of the event(Verstraete 2005, 2006), the irrealis should show characteristics not associated with theNyulnyulan irrealis. The paradigmatic contrasts give the value of the category, itsmeaning invariants only indirectly and partially. Scope is also relevant to the semanticsof the category, as we will now see.

(c) scope. We have argued in this paper that the irrealis in Nyulnyulan languageshas scope over situations, these being construed by the experiential component ofclause grammar, and concerns the occurrence of the situation. Another possibility isthat the scope of irrealis is the proposition. Givón (1995:114�23) construes the irrealisin this way, as one of the modes of proposition assertion: epistemic uncertainty is in hisview the crucial semantic feature. The irrealis of Caddo is a fairly convincing exampleof such a category, as suggested by the fact that polar interrogatives are in the irrealis,

±

Page 37: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

374 oceanic linguistics, vol. 45, no. 2

while content interrogatives are in the realis (they presuppose the proposition for somevalue of the variable in focus) (Chafe 1995). So also is irrealis of Central Pomo, whichexpresses �actions or states that the speaker portrays as purely within the realm ofthought� (Mithun 1995:378�80)�that is to say, propositions that are entertained (irre-alis) as against propositions asserted (realis).

We mentioned under (a) the case of Gooniyandi modal categories that could be con-strued as expressing reality status. If we adopt this categorization, the irrealis has proposi-tional scope (McGregor 1990:556). The proposition is shifted beyond considerations oftruth and falsity: it is hypothesized or entertained rather than asserted. The semantics ofthe Gooniyandi irrealis category, the meaning it encodes, is thus rather different from thesemantics of the Nyulnyulan irrealis, and this shows up in the very different (though notdisjoint) range of contextual senses and pragmatic implicatures of the categories. Thus,compare the following senses of the Gooniyandi future subjunctive irrealis with those ofthe corresponding Nyulnyulan category: predictions, guesses, hypotheses, certainties,claims, desires, attempts, and abilities (McGregor 1990:545�47). Both Gooniyandi andNyulnyulan irrealis categories differ from the irrealis as construed by Givón (1995:123),which codes epistemic uncertainty. In the Gooniyandi case epistemic uncertainty is irrel-evant (the proposition being taken from the range of questions of truth or falsity), while inNyulnyulan no epistemic uncertainty is associated with the irrealis.

(d) simultaneous grammatical categories. In Nyulnyulan languages, atleast some tense categories are maintained with the irrealis, although neutralizations dooccur. There are languages in which temporal contrasts are impossible in the irrealis,and languages in which a realis/irrealis contrast applies instead of a tense contrast (e.g.,Manam [Lichtenberk 1983], Pilagá [Vidal and Klein 1998:183]). The (partial) neutral-ization of the transitivity contrast in irrealis IVs in many Nyulnyulan languages maywell be an uncommon neutralization: although we have no direct information on thedistribution of this phenomenon across the world�s languages, it seems from remarksin Miestamo (2003:113�14) that it may be rare.

As we have seen, in all Nyulnyulan languages, with the exception of Yawuru, therealis/irrealis contrast is neutralized in negated finite clauses, where just the irrealis ispermissible. This seems to be a somewhat unusual situation cross-linguistically. InMiestamo�s (2003) representative sample of two hundred and forty languages, onlythirty-two (13 per cent) show obligatory use of the irrealis in clausal negation. Mies-tamo (2003:191) suggests that such languages �have grammaticalized the fact thatnegation belongs to the realm of the non-realized.� We are not convinced that the paral-lelism of negation and nonrealisation is what has been grammaticalized, or driven thegrammaticalization. Rather, we believe that the neutralization correlates instead withgrammatical characteristics of the irrealis (and negation), including (a�c) above, in par-ticular with the scope of the two categories. One wonders whether the rarity of lan-guages with this neutralization may be due to an instability inherent in systems such asthe Standard Average Nyulnyulan, whereby clausal negators show a tendency to�attract� modal categories into their scope. If this is so, this may have been where thechanges in Yawuru began, the loss of nonfuture irrealis following as a consequence.

Page 38: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 375

(e) locus of the irrealis. This can be interpreted in two ways. One is in termsof the morphological locus, conceived of purely formally: where is the irrealis real-ized? For instance, is it an in³ectional category of the verb (as in many dictionaryde²nitions), or is it realized by a free marker such as a particle (e.g., Burmese�Cornyn and Roop 1987:43) or deictic classi²er (as in Pilagá and Toba�Vidal andKlein 1998), or is it realized by a variety of morphological devices (as in Mocho�Martin 1998)? Two questions arise here. Does the difference always correspond to adifference in grammatical vs. nongrammatical irrealis, and what differences follow forthe semantics of the respective irrealises? Second, although we have seen that the mor-phological locus of the irrealis can be at variance with its semantic locus, one wonderswhether there might be positive correlations in a representative cross-linguistic sample.

Different values of variables (a�e) will give rather different types of irrealis categories,showing different semantic, pragmatic, and formal characteristics. Testable predictions aregenerated, and a rack of questions arise concerning the interactions among the variables.We submit that a fruitful though underexplored domain of investigation concerns theranges of contextual and pragmatic senses of irrealis categories, and how they vary accord-ing to the typological position of the irrealis category. While we believe these are relevantdimensions for a typology of the irrealis, we do not maintain their comprehensiveness.

6. CONCLUSION. Bybee (1998:266) has suggested that broad terms like irrealis havedistracted linguists from their real task of providing in-depth semantic analyses of the lan-guage-particular categories. While we are in full agreement with her sentiment, we do notagree that the blame goes to the terms themselves. The blame lies squarely with linguistswho seem to be too often satis²ed with labeling categories, thereby implying that theirmeanings are obvious. A viable program of descriptive and theoretical linguistics shouldplace semantics and pragmatics in a primary position (e.g., McGregor 1990, 1997), andwe consider it to be one of the most fundamental components of our job as linguists.

We have attempted in this paper to provide as comprehensive as possible a descriptionof the meanings and uses of the irrealis in Nyulnyulan languages. This does not, ofcourse, quite meet the goal of depth description of particular languages. We expanded thescope of the paper to this small family of languages because of the lack of depth treat-ments of the irrealis in any particular languages, and the less-than-comprehensive corporaavailable to us on any language, a result of their highly endangered states. Thus, while itwould be ideal to focus on a particular language, we simply do not have adequate data forthe level of detail that we would like. Since the languages are overall quite similar typo-logically, and appear to show fundamental similarities on a number of grammaticalparameters, we believe that�with exercise of caution�it is reasonable to extrapolatefrom the cross-linguistic semantic and pragmatic variability shown by the irrealis cate-gory to intra-language variability. Admittedly there are disjunctures, as in the case ofYawuru, which diverges from Standard Average Nyulnyulan. But otherwise the evi-dence, such as it is, seems to suggest a high degree of comparability across the languages.

We have also presented the basis of a theoretical framework to account for thedescriptive facts of the Nyulnyulan irrealis, one in which we distinguish between thesemantics of the irrealis category and the semantics of its speech-act usage. We have

Page 39: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

376 oceanic linguistics, vol. 45, no. 2

argued that for both of these it is possible to isolate a semantic core that remains invari-ant across all uses. Moreover, we submit, the range of senses of clauses in the irrealisidenti²ed in section 3 can be accounted for as contextualizations and pragmatic infer-ences. Limitation of space precludes justi²cation of this claim for each of the sensesidenti²ed. Our proposals are deliberately language-speci²c, and we do not imply thatthey extend to other languages, including other Australian languages. On the otherhand, we have identi²ed a set of ²ve features of variation that might be useful in typol-ogizing the irrealis, and explaining limits on its variation.

As in many languages, the Nyulnyulan irrealis does not have simple expression, butis realized through the combination of a range of morphemes some of which markother things as well. One morpheme, the irr pre²x is characteristic, and singles out thegeneral category. Cooccurring morphemes serve to narrow down to temporal sub-types, the semantics of which are compositional, allowing us to separately assignscope to the irrealis and tense markers.

Irrealis is generally conceived of as a modal category and in our opinion the Nyul-nyulan irrealis is such a category. It presents the speaker�s line on the occurrence of asituation, and is thus an interpersonal category in the sense of Halliday (1985) andMcGregor (1997). More particularly, it expresses a type of rhetorical modi²cation ofthe clause, presenting the referent situation in such a way as to integrate it into theframework of knowledge, beliefs, expectations, and presuppositions of the speechinteraction (McGregor 1997:222�23).

To wind up the paper we raise two issues for further investigation. One is the his-tory of the irrealis in Nyulnyulan languages. As Bowern (2004:215�16) observes, it isreasonable to reconstruct the irrealis to Proto-Nyulnyulan, where the form was proba-bly *-l(a)-. The Eastern Nyulnyulan languages appear to have restructured the futureirrealis: thus we ²nd an irregular l in the second person minimal future pronominalpre²x, which is presumably a relic of the irrealis morpheme in Proto-Nyulnyulan. The-ya- irrealis pre²x of Eastern Nyulnyulan would appear to be an innovation. As Bow-ern (2004) remarks, the source of the innovation is not known. One possibility is that itis a borrowing from the Gooniyandi subjunctive -ja- ~ -ya-. (Gooniyandi and Nyikinaare adjacent languages.)

Second, we have by and large considered the irrealis as an isolated category inNyulnyulan, linking it to other grammatical categories here and there as helpful to ourstory. But, of course, the category is not isolated, and must be seen in the context ofother modal categories and means of expressing related meanings such as desiderativeconstructions and the like. We suggest that a fruitful direction of research to explore inthe future lies in adopting such a paradigmatic perspective on the irrealis. Of particularinterest here is the story of the historical development of Yawuru: what changesoccurred in the language simultaneous with (driving or being driven by) the phasing-out of the irrealis in the nonfuture?

Page 40: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 377

REFERENCES

Aklif, Gedda. 1991a. Bardi material. Electronic archive ²les.���. 1991b. Fieldnotes on the Bardi language. Unpublished manuscript.Bates, Daisy M. n.d. Native vocabularies�Broome Magisterial District. Typescript.Bird, William. 1910. Some remarks on the grammatical construction of the Chowie

language, as spoken by the Buccaneer Islanders, North-Western Australia. Anthro-pos 5. 454�56.

Bowern, Claire L. 2004. Bardi verb morphology in historical perspective. PhD thesis,Harvard University.

Bugenhagen, Robert G. 1994. The semantics of irrealis in the Austronesian languagesof Papua New Guinea. In Topics in descriptive Austronesian linguistics, ed. by GerP. Reesink, 1�39. Leiden: Rijksuniversiteit Leiden.

Bybee, Joan. 1998. �Irrealis� as a grammatical category. Anthropological Linguistics40:257�71.

Bybee, Joan L., and Östen Dahl. 1989. The creation of tense and aspect systems in thelanguages of the world. Studies in Language 13:51�103.

Bybee, Joan, Revere Perkins, and William Pagliuca. 1994. The evolution of grammar:tense, aspect, and modality in the languages of the world. Chicago: University ofChicago Press.

Capell, Arthur. 1940. The classi²cation of languages in north and north-west Australia.Oceania 10:241�72, 404�33.

Chafe, Wallace. 1995. The realis-irrealis distinction in Caddo, the Northern Iroquoianlanguages, and English. In Modality in grammar and discourse, ed. by Joan Bybeeand Suzanne Fleischman, 349�66. Amsterdam: John Benjamins.

Conklin, Harold C. 1962. Lexicographical treatment of folk taxonomies. In Problemsin lexicography, ed. by Fred Householder and Sol Saporta, 119�41. Bloomington:Indiana University Research Center in Anthropology, Folklore, and Linguistics.

Cornyn, William S., and D. Haigh Roop. 1987. Beginning Burmese. Honolulu: Univer-sity of Hawai�i at Mânoa.

de Haan, Ferdinand. 2006. Typological approaches to modality. In The expression ofmodality, ed. by William Frawley, 27�69. Berlin: Mouton de Gruyter.

Dik, Simon C. 1989. The theory of functional grammar, part 1, The structure of theclause. Dordrecht: Foris Publications.

Dixon, R. M. W. 1980. The languages of Australia. Cambridge: Cambridge UniversityPress.

Elliott, Jennifer. 2000. Realis and irrealis: Forms and concepts of the grammaticalisa-tion of reality. Linguistic Typology 4:55�90.

Givón, Talmy. 1984. Syntax: A functional-typological introduction. Vol. 1. Amsterdam:John Benjamins.

���. 1995. Functionalism and grammar. Amsterdam: John Benjamins.Goddard, Cliff. 1998. Semantic analysis: A practical introduction. Oxford: Oxford Uni-

versity Press.Goldberg, Adele E. 1995. Constructions: A construction grammar approach to argu-

ment structure. Chicago: University of Chicago Press.Greenberg, Joseph. 1988. The ²rst person inclusive dual as an ambiguous category.

Studies in Language 12:1�18.Grice, Paul. 1989. Studies in the way of words. Cambridge, Mass.: Harvard University

Press.Halliday, Michael A. K. 1985. An introduction to functional grammar. London: Edward

Arnold.

Page 41: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

378 oceanic linguistics, vol. 45, no. 2

Hopper, Paul, and Sandra Thompson. 1980. Transitivity in grammar and discourse.Language 56:251�99.

Hosokawa, Komei. 1991. The Yawuru language of West Kimberley: A meaning-baseddescription. PhD thesis, Australian National University.

Israel, Michael. 2004. The pragmatics of polarity. In The handbook of pragmatics, ed.by Laurence R. Horn and Gregory Ward, 701�23. Oxford: Blackwell.

Lazard, Gilbert. 2006. More on counterfactuality, and on categories in general. Lin-guistic Typology 10:61�66.

Levinson, Stephen C. 2000. Presumptive meanings: The theory of generalized conversa-tional implicature. Cambridge, Mass.: The MIT Press.

Lichtenberk, Frantisek. 1983. A grammar of Manam. Honolulu: University of Hawai�iPress.

Martin, Laura. 1998. Irrealis constructions in Mocho (Mayan). Anthropological Lin-guistics 40:198�213.

McGregor, William B. 1988. Handbook of Kimberley languages, vol. 1: General infor-mation. Canberra: Paci²c Linguistics.

���. 1990. A functional grammar of Gooniyandi. Amsterdam: John Benjamins.���. 1994. Warrwa. München & Newcastle: Lincom Europa.���. 1996. Nyulnyul. München & Newcastle: Lincom Europa.���. 1997. Semiotic grammar. Oxford: Clarendon Press.���. 2000. Re³exive and reciprocal constructions in Nyulnyulan languages. In

Reciprocals: Form and function, ed. by Zygmunt Frajzyngier and Traci S. Curl,85�122. Amsterdam: John Benjamins.

���. 2002. Verb classi²cation in Australian languages. Berlin: Mouton de Gruyter.���. 2004. The languages of the Kimberley, Western Australia. London: Routledge-

Curzon.���. 2006. Focal and optional ergative marking in Warrwa (Kimberley, Western

Australia). Lingua 116:393�423.���. In press. Complementation as interpersonal grammar. To appear in Word.���. In preparation. A grammar of Nyulnyul, Dampier Land, Western Australia.Metcalfe, Christopher D. 1975. Bardi verb morphology. Canberra: Paci²c Linguistics.���. 1979. Some aspects of the Bardi language: A non-technical description. In

Aborigines of the West: Their past and their present, ed. by Ronald M. Berndt andCatherine H. Berndt, 197�213. Perth: University of Western Australia Press.

Miestamo, Matti. 2003. Clause negation: A typological study. PhD thesis, Universityof Helsinki.

Mithun, Marianne. 1995. On the relativity of irreality. In Modality in grammar and dis-course, ed. by Joan Bybee and Suzanne Fleischman, 367�88. Amsterdam: JohnBenjamins.

Munro, Pamela. 2006. Modal expression in Valley Zapotec. In The expression of modal-ity, ed. by William Frawley, 173�205. Berlin: Mouton de Gruyter.

Nekes, Hermann, and Ernest A. Worms. 1953. Australian languages. Fribourg: AnthroposInstitut.

���. 2006. Australian languages. Ed. by William B. McGregor. Berlin: Mouton deGruyter.

Nicolas, Edith. 1998. Étude du système verbal du bardi, langue du nord-ouest aus-tralien, avec une présentation contrastive du système bunuba. PhD thesis, Univer-sité de Paris VII � Denis Diderot.

Palmer, Frank R. 1986. Mood and modality. Cambridge: Cambridge University Press.Peile, Anthony Rex. n.d. Fieldnotes on Warayungari. Unpublished manuscript.Rosch, Elanor. 1973. On the internal structure of perceputal and semantic categories.

In Cognitive development and the acquisition of language, ed. by Terence E. Moore,111�44. New York: Academic Press.

Page 42: The Semantics and Pragmatics of Irrealis Mood in Nyulnyulan Languages

semantics and pragmatics of irrealis mood 379

Stokes, Bronwyn. 1979. Fieldnotes on the Yawuru language. Unpublished manuscript.���. 1982. A description of Nyigina: A language of the West Kimberley, Western

Australia. PhD thesis, Australian National University.Stokes, Bronwyn, and William B. McGregor. 2003. Classi²cation and subclassi²cation

of the Nyulnyulan languages. In The non�Pama-Nyungan languages of northernAustralia: Comparative studies of the continent�s most linguistically complex region,ed. by Nicholas Evans, 29�74. Canberra: Paci²c Linguistics.

Strawson, P. F. 1952. Introduction to logical theory. London: Methuen.Traugott, Elizabeth Closs, and Richard Dasher. 2001. Regularity in semantic change.

Cambridge: Cambridge University Press.Trubetzkoy, Nikolai S. 1969. Principles of phonology. Translated by C. A. M. Baltaxe.

Berkeley: University of California Press.Van Valin, Robert D. 1993. A synopsis of role and reference grammar. In Advances in

role and reference grammar, ed. by Robert D. Van Valin, 1�164. Amsterdam: JohnBenjamins.

Verstraete, Jean-Christophe. 2005. The semantics and pragmatics of composite moodmarking: The non�Pama-Nyungan languages of northern Australia. LinguisticTypology 9:223�68.

���. 2006. The nature of irreality in the past domain: Evidence from past inten-tional constructions in Australian languages. Australian Journal of Linguistics26:59�79.

Vidal, Alejandra, and Harriet E. Manelis Klein. 1998. Irrealis in Pilagá and Toba? Syn-tactic versus pragmatic coding. Anthropological Linguistics 40:175�97.

Wierzbicka, Anna. 1996. Semantics: Primes and universals. Oxford: Oxford UniversityPress.

Worms, Ernest A. 1942. Sense of smell of the Australian Aborigines: A psychologicaland linguistic study of the natives of the Kimberley division. Oceania 13:107�30.

Yawuru Language Team. 1998. Yawuru Ngan-ga: A phrasebook of the Yawuru language.Broome: Magabala Books.

Afdeling for LingvistikInstitut for Antropologi, Arkæologi og LingvistikAarhus UniversitetBuilding 1410Ndr. RinggadeDK-8000 Århus [email protected]

ScribePO Box 523North CarltonVIC [email protected]