mood as verbal definiteness in a 'tenseless'...

This article argues that the mood morphemes found on punctual verbs in Mohawkare to be analyzed semantically as markers of verbal definiteness/specificity. In par-ticular, the so-called future marker is an indefinite morpheme, indicating that theevent argument of the verb undergoes Heim’s (1982) rule of Quantifier Indexing. Incontrast, the seeming past marker is a marker of definiteness/specificity, indicating thatthe event argument is immune to Quantifier Indexing. This explains many apparentpeculiarities of the Mohawk verbal system, including: the use of “future” as a pasthabitual form, the use of mood prefixes in conditionals, free relatives, and complementclauses, and the incompatibility of “past” with negation. The relationship betweenindefinite mood and future events, where it exists, is explicated in terms of thebranching theory of time proposed by Dowty (1979) and Kamp and Reyle (1993),which is grounded in a fundamental asymmetry in how humans conceive of the futureversus the past.

MARK BAKER AND LISA TRAVIS

Natural Language Semantics

5: 213–269, 1997. 1997 Kluwer Academic Publishers. Printed in the Netherlands.

MOOD AS VERBAL DEFINITENESS IN A

“TENSELESS” LANGUAGE*

* The research for this paper was supported by the Social Sciences and Humanities ResearchCouncil of Canada, grant 410-95-0979 to the first author and grant 410-93-0897 to the secondauthor, and Fonds pour la Formation de Chercheurs et l’Aide à la Recherche de Québec,grant 94ER0578, whose support we appreciate. We have also benefited from the opportu-nity to present parts of this work at colloquia at Cornell University, the University of Calgary,and MIT, and thank the audiences there for their input and suggestions; particular thanksto Brendan Gillon, Nigel Duffield, Jennifer Ormston, Ben Shaer, Mürvet Enç, Veena Dwivedi,Betsy Ritter, Karin Michelson, and Claire Lefebvre. It is no secret that our training andexperience is more focused on “natural language” than on “semantics”; we thank RichardLarson and two anonymous NALS reviewers for their detailed comments, corrections, andhelp with background references, and editors Angelika Kratzer and Irene Heim for theirguidance in sorting through these materials. No doubt there are still problems and mistakes;these are our responsibility.

Mohawk is a Northern Iroquoian language spoken in Quebec, Ontario, and New York.Typologically, it is a polysynthetic, nonconfigurational head-marking language. Generalinformation about the language can be found in Deering and Delisle (1976) and Baker (1996),among other sources. In addition to published sources, new data for this article come fromfieldwork done at Kahnawake, Quebec, by the first author and his students between 1990and 1995. Several consultants have been used, but in particular Ms. Carolee Jacobs.

One comment about the title. Mohawk is not really a tenseless language: it has tensemorphemes -kwe’ or -hne’ that attach only to verbs in the habitual and stative aspects;-kwe’ can be found in some examples below. However, these tense morphemes never appearwith the punctual verbs we focus on here, so this part of the Mohawk paradigm is truly “tense-less.” For discussion of why mood appears with punctual aspect whereas tense appears withhabitual and stative aspects, see Baker and Travis (1996).

1. I N T R O D U C T I O N

For a linguist with comparative interests, a large part of the fun of doinglinguistic research is searching out all the fascinating, deep, and intricatedifferences in how languages work. Indeed, the only thing that gives acomparable thrill is discovering the deep and fascinating ways in which theyare all the same.

The tension between these two joys can be seen in a close look at theverbal inflection systems of many Native American languages. For example,in the Mohawk language, eventive verbs in the punctual aspect appear inthe following three forms:1

(1) a. wa- ha- rast- e’???-MsS-draw-punc

‘he drew it’

b. v- ha- rast- e’???-MsS-draw-punc

‘he will draw it’

214 MARK BAKER AND LISA TRAVIS

1 The glosses of the Mohawk examples include the following abbreviations: fact, factualmood; punc, punctual aspect; plur, plural; rev, reversive; stat, stative aspect; srfl, semi-reflexive; dup, duplicative; cis, cislocative; fut, future mood; sim, simultaneous; opt, optativemood; part, partitive; past, past tense; neg, negative; Q, question particle; trans, transloca-tive; hab, habitual aspect; iter, iterative; loc, locative; refl, reflexive; purp, purposive; caus,causative. Glosses of agreement morphemes include indication of person/gender (1, 2, M,F, Z(oic), N), number (s, d or p), and series (S (roughly subject), O (roughly object), or P(possessor)). Ne is a particle of uncertain significance and so remains unglossed. Abbreviationsfor sources from which textual examples have been taken are: D&D (Deering and Delisle1976), KO (Williams 1976), OK (Scott 1991).

We consider only the punctual forms of eventive verbs in this article, because those arethe simplest forms in which the mood prefixes can appear. Mood prefixes cannot attachdirectly to the habitual or stative forms of the Mohawk verb, nor to inherently stative (adjec-tival) verbs. We believe that the reason for this is that the event position in the argumentstructure of the verb is already bound by habitual and stative morphology, and hence is notavailable to be modified by a higher mood morpheme. Stative/adjectival verbs, on the otherhand, cannot take mood prefixes because they do not have an event position as an inherentlexical property.

(i) (*wa’-/*v-/*a-)-ka-rakv-(‘)fact/fut/opt- NsS-be.white-punc

‘It is (was, will be, should be) white.’

This is a result of independent semantic interest, because it suggests that Parsons (1990)and others are wrong to generalize the Davidsonian theory of event roles from eventivepredicates to (all) stative predicates. However, space considerations do not allow us to discussthis matter here. See Baker and Travis (1996) for some discussion.

c. a- ha- rast- e’ ???-MsS-draw-punc

‘he should draw it’

Based on a casual inspection of the English translations of these forms, itis very tempting to conclude that wa’- is a past tense morpheme in Mohawk,whereas v- is a future tense morpheme. However, many Amerindianists overthe years have argued that this temptation should be resisted, that whatlook at first like ordinary tense morphemes actually do not directly expresstense or time reference at all. When one gets to know the languages better,one discovers that forms like (1a) and (1b) are also used in contexts wherethere is no sense of pastness or futureness intended. Conversely, there areEnglish sentences that have ordinary past or future tenses but their normaltranslations in Mohawk do not have wa’ - or v-. While the most brutishof the Eurocentrics might (and have) attribute this to the illogicalness ofthe aboriginal mind, the standard conclusion is that the inflectional systemof Mohawk has a logic of its own, different from that of most Indo-Europeaninflectional systems. Researchers are often rather vaguer on the questionof what these morphemes actually are than on what they are not, but amongthe many language-particular names that are used for them, the terms ‘mood’or ‘modality’ come up frequently. Perhaps the most famous example ofthis kind is Benjamin Whorf’s (1956) discussion of Hopi – the non-Indo-European language he studied most thoroughly – which he claimed to beso devoid of tense and time expressions, and so structured by its ownlogic, that he could infer that Hopis had no concept of time – somethingthat someone influenced by European language and culture finds almostinconceivable.

Nevertheless, it seems clear that Whorf and others like him went toofar in this. Not only do Mohawks (and Hopis) have a sense of time, but theycan easily use their language to express it. In practice, speakers do not sensethe kind of “radical translation” problem when they are asked to render asimple Mohawk verb in English or vice versa, as one might expect if thetwo systems were really incommensurate. On the contrary, the Englishtranslations in (1) really are the right translations for the Mohawk formsin many contexts. Thus, while it may be necessary to resist the tempta-tion of saying that wa’- and v- are tense morphemes in Mohawk, thecorrect theory should also account for why this is so tempting. Temptationsare like lies in that in order to be powerful they must contain a large measureof truth. Thus, there are two complementary challenges posed by materialof this kind: one must do justice to the logic of the Mohawk system in itsown terms, but one must also account for the fact that the output of that

MOOD AS VERBAL DEFINITENESS 215

system is individual forms that are comparable to English ones on a token-by-token basis. In other words, one must explain how it can be that Englishand Mohawk cut the semantic pie quite differently, without undermining thefact that it is largely the same pie that is being cut.

Our goal in this paper is to develop an analysis of mood morphemesin Mohawk that satisfies both these conditions. Our leading idea is thatthe mood prefixes are not analogous to tense morphemes in English; rather,they are analogous to the article system in English noun phrases, in thatthey mark the verbal equivalent of definiteness and indefiniteness. Once thisperspective is adopted, the peculiar-looking patterns of use of these mor-phemes can be explained in terms of the influential theory of definitenessinitiated by Kamp (1981) and Heim (1982). On the other hand, the equiv-alences between the mood-based system and the tense-based system comefrom peculiarities in how humans conceive of time (universally, as far aswe can tell). These peculiarities imply that there is only one natural wayto associate the basic mood distinctions with time distinctions. If this basicidea is correct, it shows that both English and Mohawk are in fact usingthe same formal mechanisms of (say) Quantifier Indexing and ExistentialClosure in their representation systems. However, the languages differ inthe domains of grammar in which these mechanisms are used most promi-nently: in the NP system in English but the clausal system of Mohawk.This then would be another striking example of the notion that the morelanguages differ, the more they are the same.

2 . S O M E IN A D E Q U A C I E S O F A TE N S E -BA S E D A N A LY S I S

Foster (1985, 1986) gives a brief history of how the earliest work onIroquoian languages took the prefixes illustrated in (1) to be tense mor-phemes, but how there has been a gradual shift, beginning with Lounsbury(1953) and accelerating with the work of Chafe,2 that recognizes that thesemorphemes are actually markers of mood more than tense. In other words,they are used not so much to locate events in time as to describe theactuality of these events (see Chung and Timberlake (1985) for a standardcharacterization of the difference in a crosslinguistic context). For theso-called optative prefix a- illustrated in (1c), this is obvious and uncon-troversial; (1c) asserts that a certain kind of event takes place in all desirableworlds, not that it has or will take place in the actual world. For the prefixeswa’- and v-, the distinction between ordinary tense and mood is much


2 See especially Chafe (1970).

more subtle. However, Foster points out that there are at least two envi-ronments in which wa’- does not mean past. In some situations, a Mohawkverb with this morpheme is properly translated as present, rather than past.Two examples from texts are:

(2) a. Ka’ wa-hs-e-’? Ka-nat-a-ku wa’-k-e-’.where fact-2sS-go-punc Ns-town-Ø-in fact-1sS-go-punc

(D&D, 91)

‘Where are you going? I’m going to town.’

b. Wa-hts-kv-’ kv thi rukwe . . . ? (D&D, 474)fact-2sS/MsO-see-punc Q that man

‘Do you see that man (with the fur coat and the straw hat)?’

For some reason, this reading seems to be found only with certain verbsin Mohawk – particularly verbs of motion – although the exact lexicalrestrictions are unclear. (Apparently present meanings are more freelyavailable in other Northern Iroquoian languages.) Second, Foster observesthat the wa’- prefix is always used on verbs used in performative situa-tions, where English would use a simple present. An example is (3), as itmight be used in a christening ceremony:

(3) Sak wa’-ku-hsvn-u-’.Sak fact-1sS/2sO-name-give-punc

‘I (hereby) give you the name Sak.’

Apparently, then, wa’- is really neutral between present and past; the factthat the vast majority of examples are understood as past follows fromthe well-known fact that there are severe restrictions on when one can usea perfective verb with present reference (see Comrie 1976, pp. 66–71).Presumably, these facts could be treated by analyzing wa’- as a nonfuturetense. However, Iroquoianists have the intuition that this is missing the point;the real reason behind this range of meanings is that wa’- says that anevent of the type in question has definitely taken place in the real world.As such, it contrasts with optative a-, and the older term ‘aorist tense’has been replaced with the term ‘factual mode.’ (Compare also Whorf’s(1956) discussion of the ‘reportative assertion’ category in Hopi.) We willsee further empirical support for this intuition below.

The prefix v- is more subtle still. For this morpheme, Iroquoianists havegenerally maintained the label “future,” and Foster (1985) describes this(together with -kwe’, a past tense morpheme that appears with habitualaspect verbs) as the most tense-like morpheme in the Iroquoian languages.


When he tentatively suggests changing the name of this morpheme to ‘pre-dictive mood,’ his reasoning is based not on a new insight into the meaningor use of v-, but rather on the desire to capture the fact that v- alternatesparadigmatically with a- and wa’-. However, there are clear cases inMohawk in which the “future” morpheme is used without any kind ofreference to future time; indeed, we claim that these examples provide auseful key to understanding how the mood system as a whole works.

2.1. Reminiscence Texts

The most striking nonfuture use of the “future” verbs appears in texts ofthe “reminiscence” genre. (4) is a typical example, drawn from a story aboutthe Kahnawake reservation’s first fire brigade, of which the teller had beena member (OK, 14). The immediate context was describing their old firetruck, which had no place to hook the firehose. Therefore:

(4) O’nvk tsi ki tehniyahse ohnakvnecessary that this two-people behind

v-t-hy-atyv-’ tanu v-hni-yena-’ ne ohurota.fut-dup-MdS-sit-punc and fut-MdS-hold-punc NE hose

‘Two men would have to sit behind and hold the hose.’

It is clear that there is no reference to the absolute future here; the firetruck in question was scrapped decades before. The context also showsthat this sentence is not understood as relative future, describing an eventthat will happen after some salient moment in the past, such as the plansof the main character of the story or foreshadowings (“flash-forwards”)by the narrator. Verbs with v- do contrast systematically with verbs withwa’- in texts of this kind, but the contrast has nothing to do with timereference or the sequencing of events. Thus, (5) is an example from laterin the same story, also describing actions performed by the fire brigadeon the firehose.

(5) Tsi kana’tseratatye tho nu:we the canal there

y-a’-akwa-hurot-ohw-e’trans-fact-1pS-hose-put.in.water-punc

‘We put the hose into the canal.’

The contrast is that factual verbs are used to describe unique, specific events,whereas “future” verbs are used to describe general practices. Thus, sitting


in the back of the fire truck and carrying the hose was a standard oper-ating procedure for the Kahnawake fire brigade. Putting the hose into thecanal was not; rather, the immediate context for (5) was “One time therewas a fire in Lachine, and we were called to assist. I parked the fire truckon Sixth Avenue and then . . .” Thus, future verbs occurring in “reminis-cence” contexts are nearly equivalent to past habitual verbs (for whichMohawk has distinct morphology, consisting of the habitual morpheme -ha’or -s, followed by the past morpheme -kwe’). Other examples from textsin which future-punctual verbs receive a past habitual reading are givenin (6) and (7). (6) is from near the beginning of a story about the tradi-tional Mohawk way to prepare corn flour; (7) is from a story about howthe Mohawks used to make Lacrosse balls from part of a sturgeon’s throat.

(6) Kíkv okára ne tsi ni-ye-yer-hah-kwe’this story is how part-FsS-do-hab-past

ak-sótha nónv v-ye-nvhst-ohare-’ tánu’my-grand.mother when fut-FsS-corn-clean-punc and

v-ye-the’ser-uni-’ ohvtu tsi niyóre kana’taro-k-húwefut-FsS-flour-make-punc before bread-Ø-real

v-ye-na’tar-ísa-’.fut-FsS-bread-finish-punc

‘This is the story of what my grandmother used to do whenshe would clean the corn and make the flour before she wouldmake (traditional) corn bread.’ (KO, 174)

(7) Ne sé’s yákv’ thi tsi-kúhs-es kvtsu that then it’s.said there sim-face-be.long fish

v-kuwa-yéna-’ tanu’ v-kuwa-nya’t-ó’as-e’.fut-3S/ZsO-catch-punc and fut-3S/ZsO-throat-slit-punc

‘It is said that they would catch a sturgeon and slit its throat.’(KO, 185)

(6) is particularly interesting in that it uses an explicitly past habitual verb,niyeyerhahkwe’, in parallel with a sequence of future-punctual verbs, illus-trating the near-equivalence of the two. Indeed, something similar can beseen in English with the use of the modal would in the glosses we havegiven of (4), (6), and (7); in these contexts would catch a sturgeon is thenear-equivalent of the explicitly past habitual form used to catch a sturgeon.


2.2. If and Whenever Constructions

A second environment in which Mohawk uses “future” verbs without anyreference to future times is in conditionals. Verbs with v- are common inboth clauses of a conditional sentence, and such sentences have a varietyof temporal interpretations.

(8) Toka v-ke-nvsko-’ akaret, v-yuk-hrewaht-e’If fut-1sS-steal-punc cookie fut-FsS/1sO-punish-punc

ake-nistvha.my-mother.

‘If I steal a cookie, my mother will punish me.’or ‘If I steal a cookie, my mother punishes me.’or ‘If I stole a cookie, my mother punished (would punish)me.’

(8) can be interpreted as a future prediction, but it is also the standardway of expressing timeless, lawlike conditionals, which take the simplepresent in English. Even more strikingly, it can receive a purely past inter-pretation, as for example when spoken by an elderly person talking abouther childhood. Factual verbs can (rarely) be used in conditional sentencesas well, but with a different interpretation:

(9) Toka s-a-ha-ahtvti-’ ne Sak,if iter-fact-MsS-leave-punc NE Sak

tsh-a-yu-[a]htvti-’ ki ni ne’e ne Uwari.sim/trans-fact-FsS-leave-punc also NE Mary

‘If Sak left, then Mary left too (at the same time).’

(8) expresses that there is a systematic relationship between events of onetype (cookie-stealings) and events of another type (punishings of me). Incontrast, (9) only refers to an ad hoc relationship between two particularevents. It has an epistemic quality: the speaker did not actually see Maryleave, but infers that she did from the fact that Sak did. Here too the contrastbetween v- and wa’- has nothing to do with time or the sequencing of events,but rather depends on whether the sentence is a general statement about acertain kind of event or a statement about particular events. Other examplesof future verbs in timeless, lawlike conditions are the following:


(10) Tóka’ v-té-hs-ya’k-e’ sa-núhkwis nónvif fut-dup-2sS-cut-punc 2sP-hair when

o-ráhkw-ase svha yohsnóre’NsO-moon-fresh more quickly

v-se-w-ate-hyáru-’.fut-iter-NsS-srfl-grow-punc

‘If you cut your hair during the new moon, it grows back muchfaster.’ (KO, 119)

(11) Wa’-[e]-íru-’ tóka ohvtu ne vtyefact-FsS-say-punc if before NE noon

v-t-yakw-e-’ athwawenékta v-keni-yótv-’fut-cis-1pS-go-punc diaper fut-FdS-hang-punc

kwah tsi yeyothete . . .prt where the.edge

‘She said that if we would come before noon, then the two ofthem would hang a diaper near the edge (of the clothesline).’

(OK, 8)

(Example (11) comes from a text in which the narrator describes a par-ticular signaling system that she and her sister worked out before there weretelephones in Kahnawake.) Not surprisingly, the same kinds of patternsare found with the Mohawk equivalents of when and whenever clauses.Future forms can be used in both the matrix clause and the temporal adver-bial clause to express a systematic relationship between classes of events,regardless of whether those events are located in the past, present, or future,as in (12).

(12) Nonvwhen

t-v-yo-raku-tser-o:yak-e’ v-ka-rakv-’-ne’dup-fut-NsO-sun-nom-strike-punc fut-NsS-white-inch-punc

‘When the sun shines on it, it turns white.’or ‘When the sun shined on it, it would turn white.’or ‘When the sun shines on it, it will turn white.’

In contrast, factual forms are used with temporal adjunct clauses whenand only when there was a temporal link between two specific events inthe past, as in (13).


(13) Sak uk-yeshu-’ shu-t-a-ha-ataweya’t-e’Sak fact/1sO-laugh-punc sim-trans-fact-MsS-enter-punc

‘I laughed when Sak came in.’

These parallels are not surprising given that when-clauses and if-clausesreceive parallel treatments in the literature on donkey anaphora (see, e.g.,Heim 1982; Partee 1984).

From these facts we can conclude that the morpheme v- does not havefuture time reference as an inherent part of its meaning. Indeed, if it hasa consistent meaning at all, it must be a rather abstract one to allow forits presence in sentences with superficially quite different interpretations,interpretations that seem to vary with the discourse environment in whichthe sentence appears. Nor does this range of uses seem to be the result ofa purely accidental homophony in Mohawk. Ultan (1978, pp. 102–104)observes that future morphemes often have the “secondary function” ofmarking general truths and customary or habitual events, as well as hypo-thetical conditions and consequences. “Past tenses,” on the other hand, donot seem to show this kind of variability and tendency to take on “secondaryfunctions.”

3 . TO WA R D A D E F I N I T E N E S S A N A LY S I S O F M O O D

So far we have shown that v- and wa’- are not tense operators in the sensefamiliar from studies of English; rather, they seem to be part of a moodsystem. However, it is equally important to realize that v- and wa’- arealso not in themselves modal operators in the familiar sense.3 In Englishit is standard to assume that different modals generated in Infl bear differentmodal forces, where this can be understood as different kinds of quantifi-cation over suitable possible worlds. Thus, may is typically treated as a kindof existential quantification over possible (future) worlds; it contrasts withmust, which expresses a kind of universal quantification over possibleworlds. Significantly, these semantic distinctions do not show as differencesin the “mood” prefix in Mohawk. Rather, sentences that are translatedwith will, may, or must in English all contain the same “future” prefix v-.A selection of simple examples is given in (14).


3 We thank an anonymous reviewer for NALS for pressing us to clarify this point.

(14) a. V-s-ha-[a]htv

′ti-’.fut-iter-MsS-move-punc

‘He will go home.’

b. O′nek tsi v-s-ha-[a]htv′ti-’. (D&D, 334)must that fut-iter-MsS-move-punc

‘He must go home.’

c. Vwátu tsi n-v-tewa-[a]te-rohró’k-ha-’.possible that NE-fut-1pS-srfl-watch-purp-punc

(D&D, 310, 333–334)

‘We may go to the movies./It is possible that we will go to themovies.’

Thus, the “future” itself does not have any fixed modal force; rather theparticular modal force of the clause is picked up from preverbal particlessuch as onek and vwatu.4 Again, we see that if v- has a consistent meaningin all its various uses, that meaning must be a very abstract one.

This chamelionlike behavior of the future recalls certain well-known factsabout the interpretation of indefinite NPs in English. While the Russelliantradition treats indefinite NPs essentially as existentially quantified NPs,Lewis (1975) pointed out that the understood quantificational force of anindefinite NP varies with its context. Thus, the indefinite NP in the if-clausein (15a) is nearly equivalent to one of the explicitly quantified NPs in (15b),depending on which quantificational adverb appears in the matrix clause.

(15) a. Always/usually/sometimes/rarely, if a person falls from thefifth floor of a building, he survives.

b. All/most/some/few people who fall from the fifth floor of abuilding survive.


4 There are no doubt many properties of these constructions that stand in need of dis-covery and explication. For example, vwatu is actually a verbal form meaning somethinglike ‘it will be possible’, which is itself in “future” mood (v-w-atu-’ fut-NsS-be.possible-punc.)Onek, on the other hand, is an unanalyzable particle synchronically, although the comple-mentizer tsi that follows it suggests that the rest of the clause may be its CP complement.Thus, both constructions might ultimately be related to the general CP complementationconstructions discussed briefly in section 4.3. From the semantic perspective, there aremany uninvestigated questions about when these constructions have root and epistemicmeanings, and how they relate to other periphrastic expressions of modality in the language.(There is another verbal construction for necessity, for example; it selects for a CPcomplement in the optative mood.) However, it is not our purpose here to pursue theseparticular issues.

In particular, when an indefinite NP is used in a suitable context it canreceive a generic interpretation. Thus, (16a) is easily understood as makinga claim about most or all whales; it means ‘Most whales eat 1,000 poundsof food per day.’

(16) a. A whale (generally) eats 1,000 pounds of food per day.b. This whale (generally) eats 1,000 pounds of food per day.

Definite NPs typically resist such an interpretation, however; they continueto refer to a unique individual regardless of their environment. Thus, (16b)makes a claim only about a single whale, which may or may not be typicalin regard to its eating habits.5 Similarly, if one replaces a person with thatman in (15), the sentence is understood as talking about a particularindividual – an accomplished stuntman, perhaps.

Thus, there is a clear analogy between the semantics of NPs in Englishand verbs with mood prefixes in Mohawk: indefinite NPs and future verbsboth pick up a generic meaning when they occur in a suitable genericcontext, whereas definite NPs and factual verbs refer to unique individ-uals or events in any context. Thus, we claim that the so-called future markerv- in Mohawk is actually an indefinite marker – the verbal equivalent ofan indefinite article. Similarly, the factual prefix wa’- is a definite marker– the verbal equivalent of a definite article. This basic intuition has a gooddeal of precedent, particularly in the descriptive and typological litera-ture. Thus, the factual is sometimes called the “definite mode” in theIroquoianist literature (Deering and Delisle 1976, p. 149), while the optative(which is closely related to the future, as we will see) is called the “indef-inite tense” by Lounsbury (1953) and those who follow him. In a broadercomparative context, Ultan (1978, p. 101) makes the following observation:

At least the connection appears to be more than fortuitious, and it would be of interest tolook into the relationships that exist between future tense and other semantically indefinite(or prone to indefinitenss) categories such as interrogatives, evidentials, certain determiners,etc.

However, as far as we know this intuition has not yet been developed intoa formal theory of mood in Mohawk or similar languages.

The natural starting point for such a theory is the theory of definite-ness in Kamp (1981) and Heim (1982), which is explicitly developed to


5 We are oversimplifying somewhat here. In particular, definite NPs with determiner thecan sometimes receive generic readings, as in one reading of The whale eats 1,000 poundsof food per day. The Heim/Kamp approach must treat this case rather differently from theexamples in (16). Perhaps the whale also has a ‘kind’ reading, in which it means (instancesof?) the whale kind (compare Carlson 1977).

account for Lewis’s observation above. In this approach, definite andindefinite articles do not have any distinctive semantics of their own; rather,they indicate that the argument of the NP they are attached to is to be under-stood as a free variable. The difference between the two is then that theyindicate how this free variable interacts with other elements in the envi-ronment, such as adverbs of quantification. In Heim’s terminology, indefiniteexpressions but not definite ones undergo an obligatory rule of QuantifierIndexing, which is stated as follows:

(17) Quantifier Indexing: Copy the referential index of every indef-inite NP6 as a selection index onto the lowest c-commandingquantifier. (Heim 1982, p. 146)

(In this paper, we primarily use Heim’s terminology and representations,while recognizing that her and Kamp’s systems are thought to be largelyequivalent.) Sentences like (16a) are, then, treated roughly as follows. Thequantificational adverb or the present tense of the verb (or both together)causes a generic operator to be introduced into the semantic analysis, withscope over the rest of the sentence. This operator needs both a restrictiveclause and a nuclear scope. Following the “tree-splitting” algorithm ofKratzer (1989) and Diesing (1992), we can assume that the material inthe VP of the example is interpreted as the nuclear scope (and hence isthe domain of existential closure), while material outside the VP – includingthe subject NP – is interpreted as the restriction. The generic operator thenacts like an unselective binder that can bind more than one variable insidethe restrictive material. When the subject is indefinite, the variable asso-ciated with it is coindexed with the quasi-universal generic adverb, givinga representation such as the one in (18).


6 Later in her discussion, Heim (1982, pp. 224–225) considers the possibility that specificindefinites differ from other indefinites in that they, like definite NPs, do not undergoQuantifier Indexing. The relevance of this to the analysis of Mohawk will be discussed insection 4. Until then, we take it for granted that definiteness is a simple binary distinction,to simplify the discussion.

This makes the structure nearly equivalent to one in which the subject NPhas its own universal quantifier; thus (16a) is approximately synonymousto “Almost every whale eats 1,000 pounds of food per day.” The repre-sentation of (16b) is geometrically very similar. The only difference wouldbe that the subject NP is marked as definite. Hence, it is immune toQuantifier Indexing and remains a free variable. Heim shows that this isinterpreted as referring to an individual already known to the hearer.

3.1. If and Whenever Constructions

Suppose, then, that we generalize these techniques to clauses in Mohawk,starting with conditionals and ‘whenever’ constructions like those in (12),because here the parallelism between quantification over times and/or eventsand the donkey anaphora problems that originally motivated the Kamp/Heimanalysis is already familiar (see especially Partee 1984). The leading ideais simply that future v- marks the verbal equivalent of indefiniteness, whilefactual wa’- marks the verbal equivalent of definiteness. More specifi-cally, following Davidson (1967), Higginbotham (1985), Parsons (1990) andothers, we assume that verbs are predicates that take event arguments,symbolized as ‘e’ in the argument structure of the verb. In the syntax, wetake the punctual morpheme to be a functional head of category E (tosuggest “event”), which selects for a VP complement containing an undis-charged event role (see Travis 1994, forthcoming).7 v- and wa’- are


(18)

NP

whale(xi)

In generali

NP

1,000 pounds of food

V

eat

NP

xi

Exists

Infl

IP

I′

VP

V′

IP

7 The exact label of this functional category is not crucial for current purposes; readers

definiteness (or specificity) markers adjoined to this E node; their primarysemantic function is to indicate whether the event argument of the associ-ated VP undergoes Quantifier Indexing or not. The relevant aspects of themorphology of the verb can easily be derived from this structure. We assumethat the head of the verb phrase undergoes X0 movement, adjoining to theleft of the minimal head of the dominating phrase, in accordance with theprinciples of head movement in polysynthetic languages defined in Baker(1996). This yields the surface morpheme order [mood prefix–verb–punctualsuffix], which is essentially correct.8

Consider, then, how these basic assumptions interact with the standardHeim/Kamp approach to conditionals. Following Kratzer, Heim (1982)analyzes if-clauses as providing the restriction on a phonetically null uni-versal-like modal necessity operator; when-clauses may act similarly (cf.Partee 1984, p. 269).9 The main clause then gives the nuclear scope forthis operator. Generalizing slightly on the proposal in Rothstein (1995),we also assume that the if- or when-clause binds a variable in the com-plement position of a null-headed PP inside the VP of the main clause.This is reasonable syntactically, given the well-known distributional simi-larities between adverbial PPs and adverbial clauses (Emonds 1985). Hence,a sentence like (8) would have the representation in (19).


may substitute their favorite Infl-type functional category as they wish. See Baker andTravis (1996) for a more complete discussion of the various verb inflections in Mohawkand their relationships to functional categories.8 See Baker (1996, ch. 5) for discussion of the syntax and morphology of the agreementprefix, which is the fourth morpheme in simple verb forms, coming between the verb rootand the mood prefix.9 In other examples, the quantificational force comes from linguistically overt material; inMohawk, such material can include a habitual morpheme on the matrix verb, perhaps inconjunction with an adverb of quantification; see (23b) below and Chamorro (1992, pp. 53–60)for examples.

Now, the crucial step is how Quantifier Indexing applies to this structure.Since the first clause is marked with v-, the event position of its verb isindefinite; therefore it is coindexed with the closest c-commanding operator,‘in general’. Hence all (or almost all) events of me cookie-stealing areconsidered. Similarly, the event position of the main verb of the secondclause is coindexed with the existential quantifier immediately above it.Thus, the sentence asserts the existence of certain punishing events, asdesired.

However, if the analysis stopped at this point, the structure would beruled out as a violation of the ban on vacuous quantification, because thegeneric quantifier does not bind a variable in the nuclear scope.10 What ismissing is the assertion of a systematic connection between the cookie-stealing events and the punishing events. This is provided by thephonologically null P, motivated in Rothstein (1995). This P takes two eventarguments: one of them is identified with the event argument of the matrixverb, as in the standard neo-Davidsonian treatment of PP adverbials (seeHigginbotham 1985; Parsons 1990); the second is a trace bound by thefronted adverbial clause and hence coindexed with its event argument.This P is then interpreted as a surjective ‘matching function’ (M) which


(19)

V

steal(x, z, ei)

In generali, k EP

S

NP

x NP

cookie(zk)

V′

VP

E

punc

v-

E

V

punish(y, x, en)

NP

x

NP

y

P

Ø(en, ei)

ei

E

punc

v-

E

PP

V′

VP

VP

VPExistsn

EP

10 We thank nearly every semanticist who read the first version of this article for pointingout this technical mistake in our initial analysis, where we tacitly assumed that a purelypragmatic connection between the two classes of events was sufficient. We also thank oneNALS reviewer and the editors for pointing out the relevance of Rothstein (1995) to ourconcerns.

is true of a pair of events ⟨e1, e2⟩ if e2 is the unique event that correspondsto e1. The exact content of Rothstein’s matching function is determinedpragmatically; often it is a temporal relation of happening at the samereference time (particularly in when(ever)-clauses; cf. Partee 1984) or arelation of causality, but Rothstein shows that neither of these ingredientsis present in all examples. In the case at hand, the most plausible inter-pretation is that each punishing event is matched with the cookie-stealingevent that it seeks to negatively reinforce – even though there may be asignificant time delay between the two. Therefore, the LF in (19) corre-sponds to the logical formula in (20):

(20) Gene, z[steal(x, z, e) & cookie(z)] Existse′[punish(y, x, e′) &M(e′) = e]

In short, then, (8) means that for (almost) every event that is a stealing ofa cookie by me there is an event that is a punishing of me by a particularperson (my mother) which corresponds to that stealing event – which iscorrect.11 Note that there is no time reference here; hence it is left entirelyto contextual factors and pragmatics to determine whether the relevantperiod during which cookie-stealings are considered is limited to a present,future, or past interval, if it is limited at all. This explains why (8) couldbe translated with any tense in English. Parallel analyses can be easily givenfor (10)–(12).

Suppose now that the verb in the subordinate clause of (8) had theprefix wa’- rather than v-. Since wa-’ is a marker of definiteness, in thatcase, the verb’s event argument would not be indexed with the genericoperator by the rule of Quantifier Indexing. This means that the genericquantifier would not bind any variable in both the restrictive clause andthe nuclear scope; therefore the structure is ruled out as a violation of theban on vacuous quantification. Thus, a factual verb is only possible in thesubordinate clause of this kind of structure if there is no habitual morphologyor quantificational adverb in the matrix clause, and if no modal necessityoperator is posited. This is what we find in a sentence like (13), whichhas the LF representation in (21).


11 Note that punishments correspond to cookie-stealing events, but not directly to stolencookies. Thus it is possible that stealing more than one cookie as a group or in a connectedseries could be considered a single cookie-stealing event, and hence provoke a singlepunishment. The sentence would still be true in this case. See Rothstein (1995) for adiscussion of similar examples.

Here there is no quantification over events, but merely an assertion thatthere was a single event of Sak entering, an event of me laughing, andthat there was a correspondence between those events – in this case, theyhappened at the same reference time.12 This is a proper expression of whatsentences like (13) actually mean. Note also that the factual morphologyimplies that the event of arriving happened in the past; we return to whythis is so below in section 6. However, given that it is so, and that thetwo events happened at the same reference time, it follows that the eventreferred to by the main clause also happened in the past. Hence, the matrixverb must bear factual morphology as well (or it is in some aspect otherthan punctual). Similar patterns hold in conditional sentences: if theantecedent clause is factual, then the consequent clause is also factual, asin (9), and the conditional as a whole gets a reading that does not involvequantification over events. In particular, the conditional has only an epis-temic reading, in accordance with Kratzer’s (1989, pp. 10–11) conclusionthat epistemic modals are not quantifiers and may not bind variables.

Consider next the effect of putting a factual verb in place of the futureverb in the matrix clause of a structure like (19). We have just seen thatthis is possible if the subordinate clause verb is also factual and the wholeconstruction receives a nonquantificational reading. However, it is impos-sible for the matrix clause to be factual when the subordinate clause is a“future” form acting as the restrictive clause of a quantification structure.


12 That the correspondence is in this case a temporal one may in fact be guaranteed bythe fact that the subordinating element is the so-called “simultaneous” morpheme. However,we have not investigated whether this morpheme can have nontemporal uses as well and leavethis open.

(21)

V′

V

arrive(x, ei)

EPi

S

NP

x

VP

E

punc

wa’-

E

NP

y

P

Ø(en, ei)

ei

E

punc

wa’-

E

PPVP

VP

EP

V′

V

laugh(y, en)

At first glance, it seems that this pattern should also be ruled out as acase of vacuous quantification, but this turns out not to be quite right tech-nically. The LF representation will be the same as (19), except that the eventvariable of the main clause verb will not be coindexed with the existen-tial closure operator, and will be left as a free variable, as in (22). Herethe existential quantifier does not bind a variable, but it is not requiredto, whereas the generic operator still binds the variable e in both the restric-tive clause and the nuclear scope. Thus, there is no violation of the banon vacuous quantification.

(22) Gene, z[steal(x, z, e) & cookie(z)] Exists[punish(y, x, e′) &M(e′) = e]

There is a related problem, however. So far, we have not been explicit abouthow such free event variables are interpreted (see the following section).However, the intent is clearly to say that there is one specific punishingevent that is being discussed. (22) then says of this particular punishingevent that for all cookie-stealing events, this punishing event was thecorresponding response to it. The problem with this hinges on the matchingfunction M. Recall that Rothstein (1995) crucially argues that M is notjust any relation, but must be a surjective function. In other words, foreach event in the domain of M (each punishment) there must be exactlyone event in its range (a cookie-stealing) that it corresponds to, and allthe cookie-stealing events must be in the range of M. These stipulationsare not incidental to Rothstein’s analysis; on the contrary, they are her wayof accounting for the fact that sentences like (8) imply that there were atleast as many punishing events as stealing events. But if this is so, then asentence meaning (22) could only be truthfully asserted if the speaker knewthat in fact there was only one stealing event, to which the punishing eventwas a response. In that case, the use of a generic quantification structureimplied by the future form in the subordinate clause is anomalous. Thus,(22) is inappropriate given the meaning of M, and we explain why mixedfuture-factual conditionals and when-constructions are impossible.

In the data considered so far, a correct generalization is that the verbsin the matrix and subordinate clauses must match in inflection. However,this simple rule breaks down in an interesting way when one broadens thedomain of inquiry to include verbs in the so-called “habitual aspect.” Suchforms alternate rather freely with future forms in the matrix clause of theconstruction, as shown in (23a,b); however, habitual forms do not alter-nate freely with future forms in the subordinate clause, as shown in (23c,d).13


13 We are not sure that examples like (23c,d) are completely out. It might be that in some

(23) a. Nonv t-v-yo-raku-tsher-oyak-e’when dup-fut-NsO-sun-nom-strike-punc

v-ka-hutsi-’-ne’.fut-NsS-black-inch-punc

‘When the sun strikes it, it will turn black/turns black.’

b. Nonv t-v-yo-raku-tsher-oyak-e’, (tyotku)when dup-fut-NsO-sun-nom-strike-punc always

ka-hutsi-’-s.NsS-black-inch-hab

‘When the sun strikes it, it (always) turns black.’

c.# Nonv (tyotku) t-yo-raku-tsher-oyak-s,when always dup-NsO-sun-nom-strike-hab

v-ka-hutsi-’-ne’.fut-NsS-black-inch-punc

‘When the sun (always) strikes it, it will turn black/turns black.’

d.#Nonv (tyotku) t-yo-raku-tsher-oyak-s,when always dup-NsO-sun-nom-strike-hab

ka-hutsi-’-s.NsS-black-inch-hab

‘When the sun (always) strikes it, it will turn black/turns black.’

(23a) is another example of the kind we have already discussed. We analyzethe Mohawk habitual morpheme found in (23b) (perhaps modified by aquantificational adverb) as a kind of universal/generic element that quan-tifies over the event argument of the verb, saying in essence that there isan event of the relevant kind that occurs at (almost) every suitable oppor-tunity (see Baker and Travis (1996) for some details). The indefinite/future-mood clause in (23a) then turns out to be essentially equivalent tothe habitual clause in (23b) whenever the quantificational force that itpicks up from its environment is comparable to the quantificational forcethat is associated lexically with the habitual morpheme -s. In particular,(23a) is equivalent to (23b) when a universal/generic modal necessity


cases they are possible with a special epistemic reading (i.e., ‘If the sun regularly shineson it, then (by inference) it will turn black’). Indeed, such readings are predicted by our theory.Also, the “habitual” inflection -s has a progressive reading with some verbs which may appearin the antecedent of a conditional. However, the claim at hand is that there is no event-matching reading when the habitual is in the subordinate clause, although there is one whenit is in the matrix clause.

operator is understood with it. On our analysis, then, the semantic similaritybetween (23a) and (23b) recalls the similarity between Lewis’s sentencesin (15a) and (15b): an indefinite expression becomes the equivalent of anexplicitly quantified expression when embedded in the right environment.However, in another context, (23a) could also have the meaning of a simplefuture prediction, whereas (23b) could not; as an inherently quantifiedexpression, its meaning is not variable in the same way that indefiniteexpressions are.

Now consider the effect of using a habitual verb form in the subordi-nate clause, as in (23c,d). This would produce the LF in (24), where thewide scope quantification comes either from the habitual morpheme onthe matrix clause (in (23d)) or from the covert modal necessity operator(in (23c)).

(24) differs from previous structures primarily in that it contains a secondquantificational structure inside the restrictive clause of the first, inducedby the habitual morphology on the verb in the when-clause. (The restric-tive clause of this quantification is pragmatically recovered from contextsomehow, in a way that does not concern us here; we have included thepredicate ‘o is a suitable occasion’ as a place-holder.) Now the crucialfact is that the event variable of the subordinate verb is bound internallyto this quantification; hence it cannot be bound by the higher quantifier‘in general’. Indeed, there is no variable in the restrictive clause that ‘ingeneral’ can bind, and the structure is ruled out as a case of vacuous quan-tification. The fact that the event variable of the embedded verb is bound


(24)

V

strike(x, z, ek)

In general

VP

S

NP

x NP

z

V′

VPNP

z

P

Ø(en, e)

e

E

punchab

(v-)

E

PPVP

VP

EPExistsn

P

Ø(ek, oi)

oi{ }V′

V

turn-black(z, en)

PP

Alwaysi suitable(oi)

Existsk E

hab

EP

S

by an operator that has scope only over the restrictive clause also meansthat this variable is inaccessible to the complement of the PP in the matrixclause; hence this structure cannot assert a correspondence between strikingevents and turning-black events. Therefore, (23c,d), if possible at all, cannothave readings similar to (23a,b). (It is possible that sentence (23c) couldbe interpretable with a nonquantificational structure in which case it wouldassert a causal link between a “habit” of being struck by the sun and a singleevent of turning black; see fn. 13.) Therefore, the Mohawk contrast in(23) is exactly parallel to the English contrast in (25), discussed by Partee(1984).

(25) a. If Sheila walks into the room, Peter (always) wakes up.b.*If Sheila always walks into the room, Peter wakes up.

This in turn is parallel to the famous contrast between indefinite NPs andexplicitly quantified NPs found in donkey-type sentences, as discussed atlength by Partee.

(26) a. If a man owns a donkey, he beats it.b.*If every man owns a donkey, he beats it.

In fact, there is all told a three-way contrast, with definite expressionsdiffering across the board from indefinites and quantified expressions. Thecorresponding definite sentences would be:

(27) a. If Pedro owns a donkey, he beats it.

b. When Sheila walked into the room, Peter woke up.

c. Sh-a’-t-yo-raku-tsher-oyak-e’,sim-fact-dup-NsO-sun-nom-strike-punc

wa’-ka-hutsi-’-ne’.fact-NsS-black-inch-punc

‘When the sun struck it, it turned black.’

In these cases, an argument in the subordinate clause can be the antecedentof an element in the matrix clause (the agent in (27a); the event in (27b,c));in this way definites are like indefinites and unlike overtly quantifiedexpressions. However, definites must refer to unique individuals rather thanbeing quantified over; in this they differ from indefinites. This replicatesin Mohawk Partee’s basic argument for extending the theory of donkeyanaphora to the domain of verbs and events. In addition, Mohawk makesthe following striking contribution: whether the event argument of theverb is understood as definite, indefinite, or quantified is systematically rep-resented in the verbal morphology as “future,” factual, or habitual. This


is like the English determiner system, but unlike the English system ofverbal inflection, where the same simple tensed form can have all threeinterpretations.14

3.2. Reminiscence Texts

The next task is to see how this account can be extended to explain thedifferences in how “future” verbs and factual verbs are interpreted whenthey appear in reminiscence texts, as in (4) and (5). Suppose that part ofwhat it can mean to be a reminiscence text is to be a series of sentencesthat are inside the nuclear scope of a past generic operator of some kind.We have not investigated in detail exactly what linguistic and contextualfactors can be responsible for introducing such an operator into the repre-sentation of the text, but typical factors include explicit frame adverbialslike wahunise’ ‘in the old days’ or shikeksa’a ‘when I was a child’. Underthis assumption, the LF representation of the second conjunct of (4) isroughly (28).


14 Note, for example, that always can be added to the matrix clause of (27b) to give thepast tense equivalent of (25a), but always cannot be added to the subordinate clause of(27b), just as it cannot in (25c). So here we see the three-way paradigm with no change atall in the verb morphology in English.

(28)

Existsi

NP

x

P

Ø(ei, o)

o

E

punc

v-

E

PPVP

VP

Text

. . . EP . . .

V

hold(x, y, ei)

NP

hose(y)

V′

TextIn generalo

Text

relevant(o) &o is in tr

Here again, the restriction on the quantification is a rather minimal onederived from the broader context: it says that only relevant occasions withinthe reference time frame (tr) need to be considered, where the reference timeis set explicitly by an earlier adverb (shutayakwatáhsawv tsi shakwa-tenvrú:ni ne ruswátha ‘when we started the fire brigade’) and what countsas a relevant occasion is only inferred pragmatically.15 The nuclear scopecontains the content of the clause in question, together with several adjacentclauses in this part of the text. An existential quantifier is adjoined to thisnuclear scope by Heim’s rule of Existential Closure. Now, because themorpheme v- is present, the existential quantifier is coindexed with theevent argument of the head of the VP by the rule of Quantifier Indexing.Finally, individual events of hose-holding are related to the various occa-sions that are being quantified over by Rothstein’s matching function, asbefore. (Here the only difference is that there is no overt syntactic binderof the PP, and hence no (apparent) evidence for the presence of the PP inthe syntax. We leave open whether the PP is still present in the syntax,perhaps bound by a Chinese-style null topic, or whether the equation M(e)=ois simply added by the interpretive component.) This representation hasthe desired interpretation: that on most suitable occasions in the early daysof the fire brigade – specifically, those occasions on which the fire truckwas called out – there was an event of two men holding the hose that cor-responded to the occasion.

Consider next (5), a factual clause from the same text. The internal struc-ture of this clause is largely the same, except that wa’- indicates that theevent argument of the verb does not undergo Quantifier Indexing, and henceremains free from every operator that c-commands it. If all the clauses inthe text were also factual, then the text-level past generic quantifier wouldbe vacuous. Hence, the interpreter would not posit the quantifier for thisparticular text, and the sentence would be understood as saying that oneevent of putting the hose in the canal happened sometime within the ref-erence time period tr. Matters are slightly more complex if the factual clauseis both preceded and followed by “future” verbs with past habitual inter-pretations, as is the case with (5). Here there are two possibilities. First, onecould say that such a text actually contains two (or perhaps more) distinct


15 For current purposes, we want to leave open just what “occasions” are, and therefore whattheir precise relationship to events is. Possibilities include reference times, which events couldoccur within (see Partee 1984, pp. 272–273), or some kind of large scale events, with whichsmaller events would be in part-whole relationships. A NALS reviewer suggests that ouroccasions could be identified with situations in a situation-based semantics, and suggeststhat we review our proposal in light of this possibility. Unfortunately, we do not find our-selves in a position to do so satisfactorily at this time.

generic operators: one with scope over the first part of the text and onewith scope over the last part, but neither with the factual clause in itsscope. This then reduces to the simpler case just discussed. The secondpossibility is that there is only one generic operator with scope over thewhole text. Then the future-mood clauses end up being quantified over,so that there is one such event for every occasion; however, the factual moodclauses resist quantification, their event variables being left free, so thatit is only claimed that there is one such event in the whole time period inquestion. Either approach gives essentially the right interpretation for theseclauses, so far as we can see, pace some clarification of exactly how freeevent variables are to be understood (see next section).

In this way, the Kamp-Heim theory of definiteness extends rather natu-rally to explain the use of punctual verbs in reminiscence texts in Mohawk.In particular, we do not have to say that v- is ambiguous between futureand habitual (as Ultan (1978) does, for example). Rather, we say that“future” verbs are actually indefinites that sometimes acquire universal-like generic force from their environment.16

Once again, it is instructive to explicitly compare the use of futureverbs and factual verbs with that of true (past) habitual verbs in Mohawk.We have already seen that while both future verbs and habitual verbs mayend up having very much the same LF representations, those representa-tions are arrived at in different ways: the habitual verb carries its ownquantificational force, whereas the future-punctual verb picks up its quan-tificational force from the environment. From a sentence-level perspective,the result is very much the same. However, our analysis correctly predictsthat these two forms are used somewhat differently in texts. Consider firstsequences of future-punctual clauses in the middle of a reminiscence text.Such sequences are very commonly used to describe a standard sequenceof events that all fall within a single, prototypical episode. For example,(58) gives a three-sentence fragment from a text describing how people usedto help one another in Kahnawake during the Great Depression.


16 Some reviewers see our use of a text-level generic operator in reminiscence texts asundermotivated. They suggest that these cases could instead be instances of modal subordi-nation of one clause under a preceding one that has a sentence-level operator (see below).We have nothing against this suggestion in general. Indeed, it is presumably correct at leastfor cases where the text containing “future” verbs is introduced by a clause with a past habitualverb, the morphology of which presumably has inherent scope only over its own clause(see (6) for an example). However, it is not clear to us why positing covert sentence-leveloperators and covert subordination relationships is in general more restrictive or less abstractthan positing a covert text-level operator in the first place. The net result seems essentiallythe same in many cases.

(29) V-hati-yvt-a-kó-ha-’, s-ha-yá’t-afut-MpS-wood-Ø-pick-purp-punc iter-MsS-body-one

ra-uhwvts-á-yv okh ne s-ha-yá’t-aMsO-land-Ø-have and NE iter-MsS-body-one

ro-náhskw-a-yv nv’ né’e akohsátvhs.MsO-animal-Ø-have NE horse

‘Two men would go to cut wood, one a land-owner and the othera horse-owner.’

Onv t-v-hy-atat-snyénv-’.then dup-fut-MdS-refl-help-punc

‘Then they would help each other.’

Tékeni y-v-ka-yv-ht-e’ v-hni-karéni-’ netwo trans-fut-NsS-lie-caus-punc fut-MdS-bring-punc NE

óyvte’.wood

‘They would bring (back) two loads of wood (i.e. one foreach).’

Clearly, these three actions (going to cut, helping, bringing two loads)were a typical sequence of events that was habitually repeated as a unitin those days. In contrast, later in the same story one finds sequences ofhabitual verbs, such as the following:

(30) Yáhtv nowv’tu úhka te-yako-[a]tuhkaryá’k-u ne tsinot ever anyone neg-FsO-hunger-stat because

akwéku svhs ye-yvtho-s-kwe’.all past FsS-plant-hab-past

‘No one ever starved because everyone used to plant (crops).’

Kitkit óni svhs rati-nahskw-a-yv-s kweskweschicken also past MpS-animal-Ø-put-hab pig

tanu tyuhnhúhskwaru.and cow

‘They also used to raise chickens, pigs, and cows.’

In this case, planting and raising animals do not constitute a typical sequenceof events; rather, they are two activities that went on quasi-independentlyduring the period in question. This difference in how sequences of futureverbs and sequences of habitual verbs are used seems to be general, and


follows from our theory. Since future verbs are treated as indefinites, theirevent arguments are all coindexed with the same generic operator that hasscope over the whole stretch of text:

(31) Geno[suitable(o) & o in tr] Existsi, k, n

[. . . gather-wood(x, ei) & M(ei) = o& help(x, ek) & M(ek) = o& bring-wood(x, en) & M(en) = o . . .]

This means that for each choice of an occasion, there was an event of wood-gathering, an event of helping, and an event of wood-bringing, all associ-ated with that same occasion. Habitual verbs, on the other hand, each havetheir own quantificational force. Therefore, the event positions of the twoverbs in (30) are each bound by their own quantifier:

(32) Geno[suitable(o) & o in tr] Existsi[plant(x, ei) & M(ei) = o]& Geno′[suitable(o′) & o′ in tr] Existsk[raise-animals (x, ek) &M(ek) = o′]

Here no connection is made between individual events of planting andindividual events of animal-raising; indeed, occasions which are suitablefor one may not be suitable for the other, given the rhythms of the agri-cultural year. The only thing that ties these two kinds of events togetheris that both happened regularly during the same contextually defined periodtr (the Depression). This illustrates in a different way the three-way dis-tinction between verbs with inherent quantificational force, verbs that canpick up quantificational force from their environment (indefinites), and verbsthat cannot be quantified over (definites). The fact that such subtle differ-ences in how Mohawk discourses are structured follow immediately fromour theory confirms that it is on the right track.

4 . M O R E O N FA C T U A L V E R B S : DE F I N I T E O R S P E C I F I C ?

So far our discussion has centered on the semantics of “future” indefiniteverb forms in Mohawk, with factual verb forms being presented primarilyas foils to the future forms. Our fundamental assumption has been thatthe two prefixes mark the definiteness of the event argument of the verbin a way that is parallel to the way determiners mark the definiteness of NPsin English according to Kamp (1981) and Heim (1982). More specifically,the key parallel concerns the rule of Quantifier Indexing: indefinite clauses,like indefinite NPs, pick up quantificational force from their environment,while definite NPs and clauses do not. Thus, the rule of Quantifier Indexingleaves the event argument of a factual verb as a free variable. The chal-


lenge, then, is to be more precise about how such free variables are inter-preted.17

In order to answer the parallel question in the domain of NPs, Heim(1982) posits two other, interrelated differences between definites and indef-inites in addition to the difference with respect to Quantifier Indexing. First,definite NPs must refer to something that is already present in the textor in the shared background assumptions of the speaker and hearer, whileindefinite NPs must refer to something that is new in the discourse (theNovelty/Familiarity Condition). Second, definite NPs presuppose theirdescriptive content, using this content as a way of indicating which familiarreferent is intended, whereas indefinite NPs assert their descriptive content.In this way, Heim looks to the context of a clause containing a definiteNP to provide the value for the free variable introduced by that NP.

Do these aspects of the Kamp/Heim system carry over to the verbalsystem of Mohawk? The answer seems to be no. Consider, for example,the following factual clause from near the beginning of a narrative text (KO,210).

(33) 1893 sh-yo-hser-ótv wa-huwatí-nha’-ne’ ne1893 sim-NsO-year-be.kind fact-3S/MpO-hire-punc NE

Hudson Bay Co., rak-sótha tánuHudson Bay Co. my-grandfather and

áhsv ni-ho-tate-’kv′-shv. . . three part-MsO-refl-brother-plur

‘In the year 1893, the Hudson Bay Company hired my grand-father and his three brothers [to guide surveyors in the NorthwestTerritory].’

This sentence describes an event that is not referred to previously in the textand is not recoverable from the nonlinguistic context; nor does the narratorassume that this event is familiar to her audience. Moreover, the contentof the clause is clearly asserted, not presupposed. Thus, if we take factiveclauses to mark definite event variables in Mohawk and carry over Heim’stheory without modifications, the majority of Mohawk examples violatethe Familiarity Condition. It seems, then, that factive clauses are immuneto Quantifier Indexing, but lack the other characteristic properties of definiteexpressions.


17 We thank both NALS reviewers and Richard Larson for pressing us to clarify this point,and the former for suggesting that an account in terms of specificity/wide-scope existentialquantification be considered.

In fact, there is a class of noun phrases in English that has this same“mixed” behavior: namely specific indefinites introduced with words likea certain. At one point in her discussion, Heim (1982, pp. 224–225) con-siders some evidence suggesting that these specific indefinites are likedefinite NPs in that they do not undergo Quantifier Indexing, whereasthey are like other indefinite NPs in that they introduce novel variablesand do not presuppose their descriptive content. Heim then notes that thismight imply that definiteness is not a single binary feature after all. She ten-tatively proposes that the variables introduced by specific indefinites aresubject to the Novelty Condition, but then are bound by widest-scopeexistential quantification rather than by the nearest c-commanding quanti-fier; see also Kamp and Reyle (1993, pp. 288–293) for a similar suggestion.18

This is what we seem to need for Mohawk as well, at least to a firstapproximation. Thus, it seems that we should revise our proposal tosaying the v- in Mohawk marks the event variable of the verb as[–definite, –specific], while wa’- marks the event variable of the verb as[–definite, +specific]. Our analysis of the examples in the previous sectionremains unchanged, apart from this clarification of how the event vari-ables left free in that section are interpreted.

However, this proposal is not quite right either. Taken literally, it wouldimply that factive clauses in Mohawk never refer to events known to thehearer or presuppose their descriptive content. That seems much to strong.Rather, Mohawk factive clauses can be presupposed in roughly the samekinds of environments in which English clauses are presupposed: as com-plements to factive verbs, relative clauses, certain temporal adverbialclauses, and so on. Consider, for example, the simple question-answer pairin (34) (D&D, 189).

(34) A: Uhka wa’-e-hsa-’ ne s-atya’tawi?who fact-FsS-finish-punc NE 2sP-dress

‘Who made your dress?’

B: Istv′ha wa’-é-hsa-’.mother fact-FsS-finish-punc

‘My mother made it.’


18 Kamp and Reyle (1993) also suggest that indefinite NPs can take other scopes, as wellas narrowest and widest, contra a well-known claim of Fodor and Sag (1982); see also Aubasch(1994). These and other issues surrounding specific indefinites are complex, controversial,and as far as we can see not crucial to our basic line of argument. In any case, we do notpursue the details of specificity beyond the simple remarks in the text.

Here both speakers presuppose that there was an event of dress-making,the exact nature of which they are discussing. Moreover, speaker B’s answerclearly purports to be about the very same event of dress-making as speakerA’s question. Thus, speaker B neither introduces a new event into the sharedrepresentation of the discourse nor asserts for the first time that the eventunder discussion was an event of making. Thus, e need to say that theevent argument of the verb in B’s utterance acts like a definite element.Nevertheless, it has exactly the same factual morphology as the event-introducing clause in (33). From this we conclude that wa’- in Mohawkis actually unspecified for definiteness: its lexical features are [+specific,u definite], with the value for definiteness filed in in accordance with thecontext of use. In general, then, the event variable of a factual clause caneither get a value from context or be bound by a wide scope existentialquantifier.

We see then that the morphology of verbs in Mohawk is not quite assimilar to that of NPs in English as we first suggested: in English, theprimary difference that is encoded is +definite vs. –definite, whereas inMohawk it is +specific vs. –specific. Is this a coincidence? We suspectthat it may not be. Rather, we conjecture that this difference between(English) NPs and (Mohawk) clauses is closely related to the mysteriousfundamental difference between verbs and nouns, between events andentities. Entities are assumed to be relatively stable over time and cantake part in may different events, so it is an important question whetheran entity of a particular type that is involved in a given event is the sameas or different from a previously mentioned entity. Events, on the other hand,are transitory and unique; hence the question of whether a given event isthe same or different from another event of the same type generally doesnot arise in the same way. In short, the definiteness distinction may havemore functional utility than the specificity distinction in the NP domain,whereas the pragmatics reverses in the clausal domain.

An interesting comparison can be made at this point with Haitian andFongbe, as described by Lefebvre (to appear). These languages have overtdefinite determiners that appear with nominals.

(35) M manje krab la (Haitian)N Dú àsón ó (Fongbe)I eat crab det

‘I ate the crab.’

These determiners have the expected interpretation; thus, Lefebvre writes:

As has been extensively discussed in the literature on Haitian and Fongbe, the presence ofthe determiner with a noun phrase necessarily identifies old or known information; it entails


that the information conveyed by the noun phrase is part of the shared knowledge of theparticipants. . . . The semantics of the Haitian and Fongbe determiners suggest that theyare minimally specified for the feature [+definite] (following Heim’s (1982) analysis of definiteand indefinite noun phrases).

Interestingly, the same lexical items can also be used with a clause, as shownin (36):

(36) Jan rive a. (Haitian)Jan wá ó (Fongbe)John arrive det

‘Actually, John arrived.’

Lefebvre presents various arguments to show that the final element in (36)is in fact the same as the one in (35), including the fact that they showthe same phonologically conditioned allomorphy. Nevertheless, in (36) thedefinite determiner functions as an assertion marker: “In one of its func-tions, the Haitian and Fongbe determiner may be used to assert the contentof the proposition . . .” (Lefebvre, to appear).19 Thus, these languagesshow the same asymmetry as Mohawk: definite NPs presuppose theirdescriptive content, as in English, whereas “definite” clauses may asserttheir descriptive content, as in Mohawk. Moreover, the facts of Haitianand Fongbe strongly suggest that this difference is not primarily the resultof a fundamental lexical semantic difference in the definiteness markers(since they are the very same morpheme), but rather a fundamental dif-ference between how nominals and clauses are used.

5 . O T H E R C O N T R A S T S B E T W E E N D E F I N I T E A N D I N D E F I N I T E

M O O D

An important test of any theory is whether it sheds light on phenomena otherthan those that motivated it in the first place. In this section, we showthat this is true of our theory of Mohawk mood prefixes in terms of defi-niteness/specificity. In particular, we show that the idea of treating wa’-as a marker of verbal specificity and v- as a marker of verbal indefinite-ness helps to explain differences in the use of these morphemes in freerelatives, in negated sentences, and in complement clauses. In the courseof this section, we will also have cause to integrate into the analysis theso-called optative prefix – the third and last member of the Mohawk mood


19 In fact, this is only one of several uses of the definite determiner inside a clause in Haitianand Fongbe, according to Lefebvre. We cannot do justice to the full richness of the factsshe uncovers in a paper of this scope.

paradigm. We will argue that this morpheme is essentially a variant of thefuture prefix v-.

5.1. Mood and Free Relatives

The most straightforward extension of the theory is to the use of the moodprefixes in free relatives – those relative clauses that contain an interrog-ative-type word but no other head. Either factual or future morphologycan appear in such a clause, as shown in the minimal pair in (37).

(37) a. Khe-nuhwe’-s uhka v-khe-kv-’.1sS/3O-like-hab who fut-1sS/FsO-see-punc

‘I like whoever I see.’

b. Khe-nuhwe’-s uhka wa’-khe-kv-’1sS/3O-like-hab who fact-1sS/FsO-see-punc

‘I like (all the people) who I saw.’

Again, the semantic contrast is not one of time reference; rather, (37b) isunderstood as saying that there was a well-defined event in which thespeaker saw people, and she likes all the people seen in that event. (37a)is more general; it says that for any seeing event, the speaker likes the peoplehe saw in that event.

While we do not intend to give a full account of free relatives, there isa clear similarity between (37) and the other examples we have seen: wa’-is used to refer to a particular event, whereas v- is used to define a classof events for purposes of quantification. For concreteness, we may supposethat a universal quantifier is inserted in these structures (somehow), and thatthe relative clause functions as the restriction on that quantifier. The mainclause, in contrast, constitutes the nuclear scope of the quantification. Theargument position within the relative clause that is associated with theinterrogative word (which we take to be an indefinite NP) is then coindexedwith the universal quantifier by Quantifier Indexing. Thus, the logicalform of both sentences in (37) is something like (38).

(38) Allx, (e)[see(I, x, e)]] Existse′[like(I, x, e′)]

The crucial difference between the two is that the event position of theverb in the relative clause is also coindexed with the universal quantifierif and only if the verb is marked as indefinite by the prefix v-. In thatcase, the interpretation is that for all pairs of a person and an event suchthat the event is a seeing of the person by me, I like the person of the


pair. This is a suitable interpretation for (37a). In (37b), on the other hand,the e-role of the relative verb is marked as specific by wa’-; hence it isnot copied as a selection index on the quantifier, but is left as a free variable.Therefore, in (37b) there is no universal/generic quantification over seeingevents, but only over people who are seen in a single event. This too cor-responds with the intuitions of Mohawk speakers. Note also that in this casethere is no syntactic motivation and no semantic need to include Rothstein’smatching function; the NP variables created by free relative formation andNP raising already prevent the quantification from being vacuous. Thus,in (37a) there need not be a temporal or causal relationship between theseeing events and the liking events; rather, the events are matched up byvirtue of the fact that they share the same themes.

A similar elicited contrast is given in (39), except that inanimate ‘what’is used instead of animate ‘who’.

(39) a. Akweku shvs wak-eka’-s tsi nahotv ake-nistvhaall past 1sO-like-hab what my-mother

v-ye-khuni-’.fut-FsS-cook-punc

‘[When I was a child] I used to like whatever my mothercooked.’

b. Uk-eka’w-e’ tsi nahotv ake-nistvha fact/1sO-like-punc what my-mother

wa’-e-khuni-’ thetvre’.fact-FsS-cook-punc yesterday.

‘I liked what my mother cooked yesterday.’

We have not happened to collect many examples of free relatives frompublished texts, but (40) is one that we have noticed. It comes from astory in which the narrator is telling how her grandmother survived withher four small children during a time when her husband was away and givenup for dead.20


20 This example is also interesting in that the wh-word tsi nahótv ‘whatever’ is separatedfrom its head kayo’tvshera’ ‘bead’. This kind of discontinuous dependency is found inMohawk relative clauses of all kinds; see Baker (1996, sec. 4.3) for examples and analysis.

(40) Tsi nahótv v-yé-hsa-’ kayo’tvhshera’ what fut-FsS-finish-punc bead.work

t-v-yu-[a]tát-u-’, atvnátshera’ tanu’ . . .dup-fut-FsS-refl-give-punc food and

‘She would exchange whatever beadwork she finished (wouldfinish) for food and [fuel].’

In this sentence both the matrix verb (atatu ‘exchange’) and the verb inthe relative clause (hsa ‘finish making’) are future, although the circum-stances being discussed are strictly in the past. As expected, the inter-pretation is that there was no unique event of finishing beadwork or ofexchanging it. Rather, the events are understood quantificationally: (40) canbe glossed as saying that for all pieces of beadwork such that there wasan event of the narrator’s finishing that beadwork, there was also an eventof her exchanging that beadwork for food and fuel. The fact that “future”morphology should be used in both the main clause and the embedded clauseof this sentence is similar to what we saw in conditional sentences suchas (8) in section 2.2. Indeed, given the Heim/Kamp approach this is notsurprising, since if-then-clauses and sentences with quantified NPs con-taining a relative clause are treated in parallel fashion in that framework,these being the two main contexts for donkey anaphora. Therefore, it isexpected that the same restrictions on mood morphology should be foundin these two contexts as well.

5.2. Mood and Negation

A somewhat different source of evidence for the definiteness theory of moodcomes from certain cooccurence restrictions that hold between mood andclausal negation in Mohawk. Clausal negation is expressed by the combi-nation of a preverbal particle yah(tv) ‘not’ and a prefix on the verb te- orth-. Interestingly, clausal negation is incompatible with the factualmorpheme; rather, the negation of a past event must be expressed by averb in the stative aspect:21


21 The reader might notice that (41a) and (41b) differ not only in their aspect suffixes,but also in their agreement markers: (41a) has a subject agreement prefix, whereas (41b)has an object prefix. This shift from subject agreement to object agreement is conditionedby the stative aspect, and has nothing to do with the grammar of negation per se; seeOrmston (1993) and Baker and Travis (1996) for an analysis.

(41) a.* Uwari yahtv th-a’-u-[a]ther-a-hninu-’Mary not neg-fact-FsS-basket-Ø-buy-punc

‘Mary did not buy a basket.’

b. Uwari yahtv te-yako-[a]ther-a-hninu-ØMary not neg-FsO-basket-Ø-buy-stat

‘Mary has not bought a basket.’

Why should this be? Notice that if we take factual-punctual verbs to bethe equivalent of simple past verbs in English, no deep explanation of thisfact is possible, since simple past clearly is compatible with clausal negation.Starting from this view, one would have to introduce a stipulative rule ofthe morphology to replace factual-punctual features with stative featuresin the context of negation, as Chafe (1970, p. 44) does to account for cognatepatterns in Onondaga. Alternatively, within a realizational approach toinflectional morphology, one could stipulate that the negative prefix and thefactual prefix cannot cooccur, even though the morphemes in question wouldnot be realized in the same position class in traditional structuralist terms(see Lounsbury 1953). This is in effect what Stump (1992) does forsomewhat similar nonlocal dependencies between tense and negation inSwahili.

However, once we realize that wa’- is not a tense marker but rather amood marker, a more principled and elegant line of analysis becomes apossibility. In the Heim/Kamp system of NP interpretation, negation isalso an operator that induces existential closure (Heim 1982, pp. 142–144).Thus, the representation of a sentence like (41a) would be something like(42).


(42)

V

buy(e, x, y)

Not

NP

basket(y)

V′NP

x

wa’

E

E

punc

VP

EP

EP

EP

Exists

In this structure the wa’- prefix indicates that the e-role of buy is notbound by the existential quantifier introduced by negation. Rather, the eventargument is (or at least may be) understood as being bound by a widest-scope existential quantifier. Thus, the LF in (42) corresponds to the formulain (43).

(43) Existse[Not [Existx[buy(e, x, y) & basket(y)]]]

This structure could be used to assert that there is an event (presumablyin the past) such that that event was not a buying of a basket by Mary.However, notice that this is an extremely weak assertion, which is alwaystrue in any nontrivial possible world. Thus, a sentence like (41a) willnever be informative, and therefore is ruled out as deviant, or at leastunusable. In this way, we can begin to explain the incompatibility ofnegation and factual mood on general semantic grounds, without recourseto bald morphological stipulations.22

This line of analysis predicts that an example like (41a) should begrammatical if wa’- is replaced by the indefinite marker v-. v- indicatesthat the e-role of the verb does undergo Quantifier Indexing, so it wouldbe coindexed with the existential quantifier in a structure like (42). Theresult is a representation that denies that there exists an event which is anevent of Mary buying a basket – either in general or within a time perioddeducible from context. This is a meaningful and standard interpretation.Surprisingly, however, the future prefix does not occur with negation either.


22 People have pointed out to us two serious gaps in this explanation, however. First, RichardLarson reminds us that just because a sentence is “uninformative” does not in general meanthat it is ungrammatical or even unusable. For instance, one can perfectly well utter tautologieslike Every man is a man. Thus, we do not yet know why (41a) seems to be ungrammatical,rather than just rare, weird, or boring.

Second, an anonymous NALS reviewer points out that the text account may work forthe specific indefinite reading of the verb’s event variable, but it is not clear that itgeneralizes to the definite reading of the event variable. Imagine that two people are startledby a loud noise in the middle of the night, and one of them asks the other “What was that?”A priori, it seems reasonable to interpret this as a question about a contextually salientevent, one which inquires about the nature of that event. Thus it is reasonable for the secondperson to say something like “I don’t know, but I’m sure that it wasn’t Mary coming home.”If the assumptions are right, our theory then predicts that the italicized part of the responsein Mohawk could be a negated factual clause, with a representation like (42). We do not knowwhy this is impossible. Perhaps wa’- in Mohawk really is [+specific, –definite] after all,and the definite interpretation does not arise, contra the discussion in section 4. Or perhapsthere is something else about how humans talk about events that bars this kind of discoursefor independent reasons.

(44) * Uwári yáhtv th-v-yu-[a]ther-a-hnínu-’Mary not neg-fut-FsS-basket-Ø-buy-punc

‘Mary will not buy a basket.’

Instead, the future prefix is replaced by the optative prefix in this context;the result is understood as a future negation:

(45) Uwári yáhtv th-a-yu-[a]ther-a-hnínu-’Mary not neg-opt-FsS-basket-Ø-buy-punc

‘Mary will not buy a basket.’

This too has traditionally been analyzed as a kind of low-level morpho-logical replacement operation (see Chafe 1970; Deering and Delisle 1976).This time, however, we believe that the traditional view is more or lesscorrect. More precisely, this paradigm suggests that we think of optativea- as being a “negative polarity” alternate of v-; both are nonspecific indef-inites, but a- replaces v- in certain semantically defined contexts. Thusthe shift from future to optative in (44)–(45) is analogous to the switch fromsome to any in (46) in English.

(46) a. Mary bought something.b. Mary didn’t buy anything.

The complete comparison between the English determiner system and theMohawk mood system is shown in (47) (here we suppress the question ofwhether wa’- is primarily a marker of definiteness or specificity, raised insection 4).

(47) definite indefinite polarity indefinitenominal system

(English) the a, some anyverbal system

(Mohawk) wa’- v- a-

Like v-, a- indicates that the event argument of the verb it attaches toundergoes Quantifier Indexing, but it also indicates that the minimallyc-commanding quantifier must be of a particular type. Still to be explainedis where (45) gets its future force; we return to this in a more general contextbelow.

In general, then, we see that the quirks that hold between “tense” andnegation in Mohawk become more understandable once tense is reana-lyzed as mood in the way we have suggested. Interactions of this typeseem to be reasonably common across languages; we conjecture that our


analysis will generalize to explain a range of similar phenomena.23 If so,negation will provide a useful diagnostic for distinguishing true tenses frommoods: tenses will not change under negation, whereas moods may.

Confirming evidence that optative is essentially a variant of the futurecomes from conditionals. Above we saw that the future can appear in bothclauses of a conditional sentence, but that the factual does not unless theconditional has a purely epistemic reading. Significantly, optatives also mayappear in both clauses of a conditional; (48) is an example from a text(KO, 2).

(48) Tóka au-sa-k-atkáhtho-’ au-s-a-k-yvtere’-ne’ ne if opt-iter-1sS-see-punc opt-iter-1sS-know-past? NE

ókwire.tree

‘If I saw it again, I would remember the tree [the one whichthe narrator’s grandmother used the bark of to make cataractmedicine].’

(48) is like (8) in that it draws a connection between events of one type(seeings) and events of another type (recognizings). Therefore, a- causesthe event roles of the verbs to which it attaches to be coindexed with theoperators involved in the conditional. However, (48) differs from (8) inthat it is counterfactual: there is a presupposition that there will be no eventsof either type in the actual world, because the specific tree in questiondied or was cut down years before. Once again, a- acts like a kind ofnegative polarity variant of v-. This is comparable to English, in whichNPs with negative polarity determiners are licensed in counterfactualconditionals as well as in the context of clausal negation:

(49) a. If I see something on my way to town, I will pick it up.(it is reasonably likely that I will)

b. If I had seen anything on my way to town, I would have pickedit up.(in fact, I didn’t)


23 Languages which this should work in include Mayali (Evans 1991) and other Gunwinjguanlanguages, and probably Swahili (cf. Stump 1992), although there are some complications.

Within the Northern Iroquoian languages, Onondaga has the same patterns as Mohawk(Chafe 1970), while those of Seneca are very similar although perhaps not identical (Chafe1967). Curiously, Lounsbury (1953) reports that negation can appear with any “tense” inOneida, which is generally the language closest to Mohawk. We have no explanation forthis.

In the next section, we will see another way in which future and optativebehave in parallel in Mohawk, different from factual; see also Baker andTravis (1996) for evidence that they show the same pattern of interactionswith aspectual morphology. Moreover, we have the impression that thereare other languages that are generally like Mohawk in having a primarydistinction between definite and indefinite mood, but that lack the secondarydistinction between positive and negative polarity indefinite mood.24 Forexample, the Kiowa language has a single mood affix (called “future”)which is used in a set of situations that is approximately the union of thesituations in which Mohawk uses a- and the situations where Mohawkuses v- (Watkins 1984, pp. 170–172). In particular, the Kiowa affix is usedin both simple future statements and deontic statements, and in both actualand counterfactual conditionals. It is also compatible with negation, atleast when the future is understood as an imperative.

5.3. Mood and Complementation

A fifth area on which this analysis of mood sheds light is the area ofclausal complementation. Some Mohawk verbs allow their CP complementsto be in any of the three moods; these include verbs of saying, thinking,and knowing, for example. Other verbs take primarily optative complements;a few of these also allow future complements. Two examples of this secondtype are:

(50) Wa-shako-rharatstv-’ tsifact-MsS/FsO-promise-punc that

{v-/a-/*wa}-ha-kúrek-e’{fut-/opt-/*fact}-MsS-hit-punc

‘He promised her to hit it.’


24 The other obvious place that negative polarity determiners are licensed in English is inyes/no questions, such as Did you see anything? Following the parallels through, one mightexpect that factual would be impossible with yes/no questions in Mohawk while futurewould be replaced by optative. This is false: yes/no questions in Mohawk are expressed byadding a second position clitic kv to the clause; this clitic can co-occur with a punctualverb in any mood. We conjecture that these questions in Mohawk are semantically morelike tag questions in English, which presuppose a positive answer and ask for confirmation,rather than true yes/no questions. Crucially, tag questions in English do not license negativepolarity items:

(i) You bought something/*anything, didn’t you?

This conjecture accords with Chafe’s (1970, p. 45) impressions of Onondaga, which hesays has “confirmation questions” but not true “alternative questions.”

(51) Te-ho-[a]tvhutsoni-Ø (*wa’-)t-{a/*v}-ha-nunyahkw-e’dup-MsO-want-stat (*fact-)dup-{opt/*fut}-MsS-dance-punc

‘He wants to dance.’

Mohawk has no known verbs that require factual complements, however.These patterns make sense when analyzed as follows. The verbs that

require the optative and/or future are essentially verbs with modal meaningsor nonfactive propositional attitude meanings of some kind: verbs like ‘towant’, ‘to be necessary,’ ‘to try,’ and so on. Indeed, these verbs are roughlythe same as the class of verbs that take infinitive or subjunctive comple-ments in Indo-European languages. Semantically these verbs can beanalyzed as operators that quantify over events of the kind defined bytheir complement in suitable possible worlds. Thus, (51) means somethinglike ‘For every world w such that w is compatible with what x wants,there is an event of x dancing in w.’ Similarly, ‘He promised her to hit it’means that ‘For all worlds in which x fulfils his obligations to y, there isan event such that x hits z in that world.’ Since they are quantificational,these verbs trigger Existential Closure over their domain, being similar inthis respect to negation and adverbs of quantification (Heim 1982, p. 258).A- and v- indicate that the event argument of the complex verb is boundby the inserted existential operator, giving the intended interpretation. TheLF of (51) is roughly (52) (here we ignore the aspect-mood morphologyof the matrix clause).


(52)

NP

x

VP

V

want

V′

C

Ø

CP

Existi

NP

x

V′

V

dance(x, ei)

a- E

punc

EVP

EP

EP

Once the quantificational force of the lexical verb ‘want’ is spelled out,25

this representation corresponds to the logical formula in (53), where thepredicate ‘want’ is considered a function from an individual to a set ofpossible worlds.

(53) Allw [w is in want(x)] Existse[dance(x, e) & e is in w]

This is the desired interpretation.Now suppose that the a- adjoined to the lower E node were replaced

by wa’-. Then the event variable of the lower verb would not undergoQuantifier Indexing, but would be left free of the existential closuretriggered by want. The result is (or entails) a wide scope existential inter-pretation of the embedded verb’s event argument. The correspondingformula would be:

(54) Existse[Allw[w is in want(x)] Exists [dance(x, e) & e is in w]]

Now we need to ask under what conditions this representation constitutesa coherent interpretation. In fact, the question hinges on the relationshipbetween the base world wo – which is said to contain an event of the typein question – and the set of worlds that are quantified over. If wo is disjointfrom want(x), then by hypothesis there could be no event that is both inwo and in all the worlds in want(x), and (54) would necessarily be false.On the other hand, if wo can be included in the set of worlds quantified over,formulas like (54) have a chance of being true. Consider the English verbwant in this regard, and imagine the following discourse:

(55) Mary did something in this room yesterday. Nobody knows whatit was yet. Soon the videotapes will be available for publicinspection. #John wants her to dance.

The first three sentences of this discourse set up an environment where aspecific event in the “real world” is being discussed. The last sentencethen attempts to say of this event that in all worlds that John desires, thatevent will have been a dancing event; this is what the representation in(54) would come to (apart from the irrelevant difference that the subjectof the embedded predicate is not controlled by the subject of the matrixone). This is logically coherent, but it is not something that the Englishsentence in question can mean. However, if the last sentence contained adifferent propositional attitude verb, the discourse could be perfect: forexample, John thinks that she danced, or even the minimally differentJohn hopes that she danced. Intuitively, what is going on here is that when


25 Like Heim, we leave open how exactly how this is done.

one says “John wants her to dance,” only future events of her dancingwill satisfy him, not past events. This is part of the lexical meaning of want,in contrast to other verbs like hope. As an implementation of this intu-ition, we can say that want(x, t) contains not whole worlds, but “future worldparts” that consist of sequences of events and states that follow the eval-uation time t. However, in (55) the event under discussion is clearly apast event; hence it cannot be in ‘want(John, t).’ Thus, the last sentencemust be false, even without our knowing anything about John or the eventunder discussion. Now, given that Mohawk atehutsoni is like English wantin this respect, we begin to explain why sentences like (51) cannot havefactual mood in the complement clause: such sentences would always befalse, given the lexical meaning of the verb together with the propertiesof wa’-.26

More generally, this analysis predicts that the verbs that require optativeor future mood in Mohawk will be what we might call “anti-factive verbs”:verbs that create a quantification over worlds that do not include theevaluation world. Verbs that do not have this property, on the other hand,should allow factive complements. This seems to be true. Thus, both factiveverbs like ‘know’ and “neutral” verbs like ‘think/believe’ can take factual-mood complements as well as future or optative mood complements:

(56) a. Ro-[a]teyvtare tsi (wa’-/v-/a-)k-hninu-’ ne kahi.MsO-know/stat that fact-/fut-/opt-1sS-buy-punc NE fruit

‘He knows that I bought (will buy/should buy) the fruit.’

b. I-hr-ehr-e’ tóka (wa’)-t-(v/a)-ye-nunyahkw-e’.Ø-MsS-think-impf maybe fact-dup-fut/opt-FsS-dance-punc

‘He thinks that maybe she danced (will dance/should dance).’

On the other hand, the Mohawk equivalent of wish takes only optativecomplements, as shown in (57).

(57) a. Ra-[a]skanek-s a/(*wa’/*v)-ha-[a]torat-e’MsS-wish-hab opt/fact/fut-MsS-hunt-punc

‘He wishes to hunt.’


26 Again, however, there is a leap here from saying that a sentence is necessarily alwaysfalse and saying that it is ungrammatical/unusable; compare fn. 22.

Note that the discourse in (55) becomes perfect if the first sentences are put in future tenseinstead, consistent with the analysis. However, this kind of change would do the Mohawksentence in (51) no good, because wa’- clauses can only assert the existence of events inthe past, for reasons discussed in the next section.

b. Ra-[a]skanek-s a-ho-[a]torat-uMsS-wish-hab opt-MsO-hunt-stat

‘He wishes he had hunted.’

Here the problem is clearly not that ‘wish’ must necessarily be “futureoriented,” as we said about ‘want’; it clearly can be past oriented, as shownin (57b). Given this, time reference alone cannot create the problem whenthe complement in (57b) is in the factual mood. However, askanek ‘wish’does presuppose that the worlds that correspond to Sak’s desires do notinclude the real world. A wide scope existential quantification over the eventinduced by wa’- then conflicts with this presupposition, although a narrowscope existential quantification induced by a- does not.27

5.4. Modal Subordination in Mohawk

For completeness, we include a few remarks on so-called ‘modal subordi-nation’ in Mohawk, since this topic is closely related to our concerns inthis article (as pointed out by both anonymous reviewers). Modal subor-dination can be characterized roughly as a discourse phenomenon in whichone sentence in a nonfactual mood is understood as dependent on a previoussentence for various purposes involving quantification and anaphora, eventhough the second sentence is not syntactically subordinate to the first(see Roberts 1989). This phenomenon seems to be as prevalent in Mohawkas it is in English, and one of the important functions of v- in discourseis to show that the clause in question is modally subordinate to somethingin the previous discourse.

To take a concrete example, in section 2.1 we showed that “future”verbs in Mohawk could have the effect of past habituals in various kindsof reminiscence texts. In section 3, we analyzed such texts as having atacit generic operator quantifying over typical past occasions with scopeover the text as a whole. However, in some cases this quantification seemsto be indicated overtly, in the form of explicit (past) habitual morphology


27 In the text the focus has been on why some verbs allow optative complements but notfactual ones. However, the fact that these verbs have a presupposition that the worlds inquestion do not include the evaluation world may also explain why they typically takecomplements in the optative mood rather than the future mood. This effect would be similarto the one seen in counterfactual conditionals such as (48), where the presupposition ofcounterfactuality is sufficient to license the ‘polarity’ form of the indefinite mood. We donot offer a formal analysis of this effect, however. Note that the analogy with negative polarityitems in English is not perfect, since verbs like want and wish do not license negativepolarity items in English (I wish I had bought something/*anything at the store).

on the main verb of a text-introducing clause. Indeed, the text about howthe Kahnawake residents used to help one another during the Depressionpresented in (29) has this property; (58) gives the introductory sentence,along with the first of the subsequent string of future clauses.

(58) Nónv svhs akohser-á-’ke kíkv ron-úkwe when past depression-Ø-loc this MpS-person

te-hu-[a]tat-snyénv-hs-kwe’.dup-MpS-refl-help-hab-past

‘During the Depression the people used to help each other.’

V-hati-yvt-a-kó-ha-’, s-ha-yá’t-afut-MpS-wood-Ø-pick-purp-punc iter-MsS-body-one

ra-uhwvts-á-yv okh ne s-ha-yá’t-aMsO-land-Ø-have and NE iter-MsS-body-one

ro-náhskw-a-yv nv’ né’e akohsátvhs . . .MsO-animal-Ø-have NE horse

‘Two men would go to cut wood, one a land-owner and the othera horse-owner.’

Now, we already know that the habitual inflection -s counts as a genericquantifier in its own right. As a verbal inflection, it is natural to say thatit has scope only over its own clause in the first instance. Thus, the firstsentence in (58) will correspond roughly to the logical formula in (59).

(59) Geno[in-Depression(o)] Existe[help-each-other(e, x) & M(e) = o]

The next clause is in the nonfactual, indefinite mood. As such it has noquantificational force of its own, but is prone to picking up quantifica-tional force from its environment. It can then be interpreted by assigningits content to the nuclear scope of the quantificational structure in (59)and applying Quantifier Indexing to its event variable in the usual way:

(60) Geno[in-Depression(o)] Existe, e′

[help-each-other(e, x) & M(e) = o &[cut-wood(e′, y) & . . . & M′(e′) = o]

Indeed, the nuclear scope can be further extended with other indefinite-moodverbs. Thus the same kind of representation we analyzed previously ascoming from a sequence of clauses under a text-level operator can also arisevia a sentence-level operator plus modal subordination. This reduces thedependence of our analysis on covert text-level operators significantly.


Notice also that while it is common for past-habitual verbs to come before“future” verbs with a past-habitual interpretation in a text, the reverseorder would be atypical. This confirms in another way our proposal thathabitual verbs have inherent quantificational force but “future” verbs do not.

In the preceding paragraph, we assumed what Roberts (1989) calls the“insertion approach” to modal subordination, in which the representationof a nonfactual clause is simply added to the nuclear scope of a previousclause. However, Roberts (1989) argues that this approach is not suffi-cient for the full range of modal subordination cases. Thus, she discussesat length the following example, attributed to Fred Landman:

(61) a. A thief might break into the house.b. He would take the silver.

Here the second sentence is clearly dependent on the first; the first sentenceprovides the antecedent for the pronoun in the second, for example. Nowthe first sentence is understood as having a future possibility quantifica-tion: it can be paraphrased informally as “There exists some possible futurein which a thief breaks into the house.” If we applied the insertion approachto this case, the discourse should mean “There exists some possible futurein which a thief breaks into the house and takes the silver.” However,Roberts points out that this is not what (61) means: rather it means some-thing stronger, roughly “It’s possible that a thief will break into the house,and if he does, he will undoubtedly take the silver.” Apparently, then, thesecond sentence in (61) has its own inherent modal force, different from thefirst clause – the force of necessity, rather than possibility. However, it isnecessity relative to the possibility laid out in the first sentence. For thisreason, Roberts develops a different approach, which she calls “accom-modation of the missing antecedent.” Without going into details, shebasically suggests that (61b) should be interpreted as a kind of condi-tional, in which the antecedent is reconstructed from the preceding context.

How does Roberts’s theory interact with our claims about mood mor-phology in Mohawk? At first glance, it seems that we predict that caseslike (61) should not arise, because our central claim is that Mohawk v- neverhas quantificational force of its own. Hence it should never be in conflictwith the quantificational force of a previous sentence, and the insertionapproach should always give an adequate reading. However, this simple-minded prediction is not borne out. Consider, for example, the text in (62);this is part of a disagreement between a married couple about what theyshould do for their summer vacation.


(62) a. Ne ki’ ni:’ a-ke-rako-’ ne kanyatara’-kowa-hneNE PRT I opt-1sS-prefer-punc NE water-big-loc

y-a-yetew-e-’. (D&D, 235)trans-opt-1pin-go-punc

‘Well, I’d prefer that we go to the ocean.

b. Eso kvtsu v-tewa-k-e’a lot fish fut-1pinS-eat-punc

‘We’d eat lots of fish.’

Here the first sentence contains a verb that requires an optative complement,similar to those discussed in the previous section: it means roughly thatin all future worlds that the speaker values highly, there is an event ofhim and his family going to the ocean in that world. The question then ishow the second sentence is understood. If it were not modally subordi-nated to the first sentence, then it would have the status of a simple futureprediction. This is clearly not the case: the speaker’s wife has already saidshe doesn’t want to go to the ocean, and he knows this plan is very muchin doubt. On the other hand, if (62b) were interpreted by simple inser-tion, then the text should mean that in all future worlds that the speakervalues highly, there are events of them going to the ocean and of them eatingfish. This is not quite right either; rather, (62) is a conditional prediction:if they do in fact go to the ocean (which is unlikely), they will definitelyeat fish. Thus, (62b) seems to require the “accommodation of a missingantecedent” approach of Roberts, parallel to (61). This in turn seems toconflict with our claim that v- in Mohawk has no quantificational forceof its own.

In fact, this is not a problem when one looks more carefully. It is not acontradiction to say that the verb in (62b) has no inherent quantificationalforce and yet does not pick up quantificational force from the precedingsentence; what is needed is simply to identify a third potential source forthe observed quantificational force. Now, on Roberts’s analysis (62b) isinterpreted as a conditional structure, the restrictive clause of which isrecovered from context. The natural claim, then, is that (62b) should getits quantificational force in the same way as more obvious conditionals inMohawk. What way is that? In section 3, we showed that event-quanti-fying conditionals typically have future morphology, and get theirquantificational force from an invisible necessity operator, following Heim’s(1982) analysis of English (based on earlier work by A. Kratzer). Giventhis, it costs nothing to say that the same invisible necessity operator ispresent in the representation of (62b). Therefore, the patterns of modal


subordination do not undermine our claims about the nature of the moodprefixes in Mohawk.

6 . O N T H E C O R R E S P O N D E N C E B E T W E E N M O O D A N D TE N S E

Summarizing so far, we have developed a theory of the mood prefixes onpunctual verbs in Mohawk that has significant explanatory value. In par-ticular, it explains the usage of these morphemes in a variety of contextswhere time reference is not relevant, as well as their interactions withsentence-level logical operators of various kinds. The key to this analysisis that a- and v- pick up quantificational force from their environment, whilewa’- is immune to influence from quantifiers. This leaves one importanttask undone, however: we must still explain why v- is associated with futureinterpretations in most “neutral” contexts. Similarly, we must explain whywa’- is generally associated with past interpretations. Putting the problemanother way, given that Mohawk has a mood-based inflectional system, whydoes it come out looking so much like a tense-based system in most cases?

The natural first point of comparison is with the way that definite andindefinite NPs are interpreted in the Heim/Kamp system for English. Heim(1982) assumes that in addition to the rule of Existential Closure that appliesto the nuclear scope of operators, there is a second rule of Existential Closurethat applies to texts as a whole. Thus, any indefinite NP in the discoursethat is not c-commanded by a more specific operator will have its selec-tion index copied onto this text-level operator by the rule of QuantifierIndexing. This accounts for the fact that the default reading of indefiniteNPs is an existential reading. (See however Kratzer (1989), Diesing (1992,pp. 56–58) and references cited there for some criticisms of this par-ticular aspect of Heim’s approach.) The equivalent of this within the verbaldomain would thus be to say that in Mohawk there is some kind of covertFuture operator that has scope over any text. If an event argument is markedas indefinite, it will automatically be bound by this covert future operator,unless it is within the domain of some other operator of the kind we havediscussed above. This expresses formally the fact that the default inter-pretation of a verb marked with v- is a future interpretation. This would giverepresentations such as (63) for a simple example like vharaste’ ‘he willdraw.’


While this may work formally, it is not very intuitive by itself: whyshould what is generally considered the most marked tense function asthe default value in the system? Moreover, this is presumably not just apeculiarity of Mohawk. If our proposals are on the right track, then futuremust be the default interpretation of an indefinite event in other languagesas well. Indeed, typological surveys point to a general relationship betweenfuture tense and indefinite, irrealis moods (Chung and Timberlake 1985;Ultan 1978).

Some insight into why this should be so can be gleaned from com-monplace observations about how humans conceptualize time. In particular,we think of the future in quite a different way from the past. Kamp andReyle (1993, p. 534) express the basic intuition as follows:

The tense logical systems we discussed in Section 5.1 treat past tense and future tense asmirror images of each other. We already noted in passing that this is one of the points onwhich those systems are open to criticism. For our use of the future tense differs in manyways from the uses we make of the past tense. In fact, that the two tenses should be useddifferently is hardly surprising, given that our attitude toward the future is so very differentfrom our attitude towards the past. It is part of our conception of ourselves and of our rolein the world in which we live that the future is “open” while the past is “closed.” What thefuture will be like is to a significant degree undetermined, and we ourselves are amongthose who can help to shape it. As to the past, nothing we do can make any difference.

In the immediately following discussion, they briefly discuss certain lin-guistic reflexives of this fact, such as the close relationship between futuretense and modals in English, and the fact that the modal particles arefuture oriented. Then they comment on the implications of this for semantictheory as follows (pp. 534–55):


(63)

. . . . .

NP

z

V′

V

draw(z, ei)

v- E

punc

EVP

EP

CP

Ø

Existi

Text

TextIn the future

To describe the differences between them [the various modals] model-theoretically, we wouldneed models with a more complex temporal structure than those discussed in Section 5.1.At each point in time, the future, as seen from that point, would have to be open in that severalpossible futures issue from it, whereas in the opposite direction there is only one past. In otherwords, our models should not be based on a time structure which can be represented as astraight line but rather on a structure part of which (the part that is relevant from theperspective of the time t) looks something like this:

(of course, there would be such ramifications at each time point on each branch!)

However, they do not go on to implement these intuitions in their formaltheory, but rather put aside modals and concentrate on simple uses of thefuture for which this complex time structure can be ignored. (See, however,Dowty (1979, pp. 150–153) for an earlier and more formal discussion ofthis view of time, together with some comments on how it compares to morestandard views.28)

We believe that this general conception of time sheds light on the problemof how mood distinctions get associated with temporal distinctions. Onthis conception, it is generally impossible to refer directly to a futureevent, simply because there is more than one future with respect to anygiven point of time. The closest one could come would be to say that ineach of the relevant futures associated with that point in time, there is anevent of a certain type. Thus, the logical form of a sentence like (64a) shouldbe something like (64b), rather than (64c).

(64) a. Mary will buy a basket tomorrow.b. In all worlds w such that w is a future of n(ow), there exists

an event e such that e is a buying of a basket by Mary.c. There exists an event e such that e is a buying of a basket by

Mary and e is in all the worlds w which are futures of n.

Thus, we judge that (64a) can be true in a situation in which Mary hasdetermined to buy a basket, but depending on her husband’s travel plans


t

28 We thank an anonymous reviewer for bringing this reference to our attention. Note thatwe do not need to interpret this theory as a making any metaphysical claims about the truenature of time. Time need not actually have this kind of future-past asymmetry for our analysisto work; all that is necessary is that people think it does. Indeed, an even weaker claim suffices:it need not be the case that humans (necessarily) have this view of time, but only that thisbe the way time is used in the language faculty. If the human mind is modular, than thescience-forming faculty (or the theological faculty, or whatever) could have conceptions oftime that are quite different.

she might be in Montreal or Toronto tomorrow. (64b) allows for this straight-forwardly, while (64c) does not: events have spatial location (Parsons 1990),so if one event (in one possible future) takes place in Toronto and another(in another possible future) takes place in Montreal, they cannot be the sameevent.29 In short, we see that quantificational techniques are inherentlynecessary in order to talk about the future. Now we have seen that thereis no verb form in Mohawk that has future quantificational force built intoit inherently.30 Since talking about the future inherently involves quantifi-cation, it follows that one must use indefinite verbal expressions put insuitable environments to speak about the future. In contrast, there is onlyone past with respect to any given point in time. Thus, events in the pastcan be referred to directly, without the use of quantificational techniques.From this perspective, it begins to make sense that factual verbs, whichare immune to Quantifier Indexing and hence are left as free variables,are generally understood as referring to past events. It is clearly no accident,then, that indefinite mood often comes out as equivalent to future tenseand definite mood to past tense, not vice versa.

The major remaining question, then, is what textual factors trigger theintroduction of a covert future operator. Some insight into this matter comesfrom the interpretation of sentences like (65).

(65) Thetvre’ Sak wa-hak-hrori-’ tsi Tyer uwayesterday Sak fact-MsS/1sO-tell-punc that Tyer today

v-ha-torat-e’.fut-MsS-hunt-punc

‘Yesterday Sak told me that Tyer would hunt today.’

In this sentence the lower clause receives a future interpretation, not ageneric or habitual interpretation of some kind. Moreover, it clearly receives


29 An anonymous reviewer points out that this assumption touches on some deep philo-sophical issues about when events in different possible worlds can be identified as the sameevent. It may be too simple to say that such identification is never possible; the reviewercites examples like (i) as instances where there does seem to be reference to the same event(a particular battle) in different possible worlds.

(i) The battle might have taken place in Leesburg, but it actually took place inManassas.

We do not take up this issue here, assuming that even if the text account is too strong,something like it is sufficient to ground the indefinite-future vs. definite/specific-pastassociations.30 This is a particular property of Mohawk. For example, there is a future morpheme withinherent quantificational force in Hopi (see Baker, in prep.).

a relative future interpretation, not an absolute one: given the temporaladverbs, Tyer’s hunting could very well occur before the utterance timeof the sentence as a whole, or even at the same time as it. What (65) saysis that Tyer’s hunting was in the future with respect to the event of Sak’stelling me about it. Here, then, we must have a covert future operator thatdoes not have scope over the entire text, but only over a subpart of it –the complement of hrori ‘tell’.

Here we assume that Fut in the restrictive clause of the quantification is afunction that maps events or times onto sets of parts of possible worlds.In particular, w is in Fut(e) if and only if w′ is a possible world that isidentical to wo (the evaluation world) before the time of e and w is thepart of w′ that is after e. Intuitively, these are the futures that were possibleat (say) the time of the telling. The normal future operator is, then, a near-universal generic quantification over these world parts. This fits with theintuition, reported in the grammars, that Mohawk future verbs in isolationmake quite a strong claim: “The future prefix [v-] is only used if some-thing is definitely going to happen” (Deering and Delisle 1976, p. 335). Theevent argument of the verb, marked as indefinite by the prefix v-, is then


(66)

NP

x V

tell(x, y, z, e′)

V′

Existi

NP

u

V′

V

hunt(u, ei)

v- E

punc

EVP

EP

CP

CP

NP

y

tsi

In generalw

CPz

w in Fut(e′)

VP

coindexed with the existential closure operator in the usual way. Finally, thenuclear scope of this quantification also needs a statement such as ‘ei isin w’ to prevent the quantification from being vacuous; we leave openexactly how this expression comes to be in the representation. The resultis the desired interpretation for (65).

(65) shows, then, that one possible source of future operators is verbsof speaking and other logophoric verbs that introduce a perspective,including a temporal perspective. Where, then, does the “default” futureoperator that has scope over the entire clause come from? It is a shortstep to say that this operator is automatically introduced by the temporalperspective implicit in the speech act itself. One way to think of this wouldbe in terms of the view, characteristic of Generative Semantics, that at thehead of any text is a speech act verb such as ‘I say’ that does not get realizedphonetically. This verb could then introduce a future operator and anexistential closure over its nuclear scope in exactly the same way as ‘tell’in (65) does. However, there is no real need to go quite this far: we can keepthe future operator and the existential closure without literally including‘I say’ in the representation at any level. On these assumptions, a moreaccurate representation for vharaste’ ‘he will draw’ would be (67).

Consider next a simple factual verb uttered in relative isolation, suchas wa-ha-rast-e’ ‘he drew’. With such a verb, the factual morphologyindicates that the event argument of the verb does not undergo QuantifierIndexing; instead, it receives something like a widest-scope existentialreading. As such, a factual clause asserts that there is in the evaluation worldwo an event of the kind described; however, there need be only one suchevent for the sentence to be true, regardless of the context. Now the crucial


(67)

Existi

NP

x

V′

V

draw(x, ei)

v- E

punc

EVP

EP

EP(I now assert that) In generalw

S

w in Fut(now)

question is, exactly how does one understand wo? Following the intuitionsof many Amerindianists and in keeping with the branching conception oftime, we suggest that wo must be understood as consisting only of statesand events that are prior to the utterance time of the clause (or, perhaps,some other salient time of evaluation). Here the idea is that since the pastis unique with respect to any point in time, but the future is not, simpledirect reference is possible to past events only. Given this, a factual clausecannot be under the scope of a future operator like the one in (67). If itwere, the representation would correspond to the formula in (68).

(68) Existe(e in wo) [Genw [w in Fut(now)] [draw(x, e) & e is in w]]

If the italicized condition is omitted, the structure violates the ban onvacuous quantification, because the generic operator does not bind anythingin its nuclear scope. On the other hand, if the condition is included, thestructure is contradictory, because it asserts that the particular event inquestion is both in wo – the relative past – and in all the relative futures,which is impossible by hypothesis. Therefore, there must not be a futureoperator with scope over a factual clause, and such clauses will have therepresentation in (69).

(69) Existe(e in wo) [draw(x, e)]

So far this discussion has focussed on fundamental differences betweenthe past and the future; where then does the present fit in? In fact, wementioned that factual verbs can sometimes be understood as referring topresent events instead of past events (see (2)). This empirical observationfollows naturally from the analysis given here. We have said that withrespect to any moment of time there is a unique past, but many possiblefutures. Notice, however, that there is a unique present as well: one cannotchange the present any more than one can change the past. Therefore,present events can also be referred to directly, without using quantificationaltechniques. This explains why wa’- can translate into English present aswell as past. Technically, we can say that wo includes the present momentalso. The fact that a present tense interpretation is so rare, at least inMohawk, presumably follows from the fact that punctual verbs refer tocomplete, bounded events, and it is rare for such an event to exactly coincidewith the moment of speech. The one case in which the two would sys-tematically coincide is in performative utterances, in which the event of(say) giving a baby a name is identical to the event of uttering a certainsentence. And indeed, Mohawk productively uses factual-punctual verbsin this context, as pointed out by Foster (1985, 1986) (see (3)). More gen-erally, we conjecture that many of those languages in which the fundamental


distinction in the “tense” system is between future and past-present(“retrospective languages” in the terminology of Ultan 1978) are actuallymood-based systems, while languages like English, in which the funda-mental distinction is between past and present-future (Ultan’s “prospectivelanguages”), are probably true tense-based systems. In this we givesomewhat more formal grounding to the intuitions of many Amerindianists(including Whorf ) that such “homophonies” indicate that quite a differentsystem is at work.

The last simple clauses to consider are those that contain an optative-punctual verb. In fact, optative verbs are not found in simple matrix clausesvery often; by far the most common uses of the optative in spontaneoustexts are as the complement of a modal or propositional attitude verb, orin the context of negation. In these environments, the event argumentreceives its quantificational force from the matrix verb or the negation, inthe way we have already discussed. When an optative form does appearas a matrix verb, it is understood as having deontic modal force, and isusually translated into English as ‘should’:

(70) Kak nu: y-a-yétew-e-’ n-v-yó-karahw-e’ someplace trans-opt-1pinS-go-punc NE-fut-NsO-get.dark-punc

(D&D, 310)

‘We (inclusive) should go someplace tonight.’

Thus, while a verb marked v- picks up a force like “In all likely futureworlds, there is an event . . . ,” verbs marked a- pick up a force like “Inall desirable (or proper) future worlds, there is an event . . .” We there-fore must say that a- in Mohawk is not just a negative polarity item, buta negative/deontic polarity item; it appears whenever an indefinite clauseis bound by a negative operator (as in clausal negation or counterfactualconditionals) or a deontic operator (under verbs like ‘want’, in purposiveclauses, in matrix contexts like (70)). It would clearly be desirable to identifysomething that negation and deontic modalities have in common to justifythis treatment, but we have no specific proposal to make in this regard.Finally, while a covert future operator is available in all texts due to theirqualities as speech acts, a covert deontic operator seems to be more limited.Perhaps a closer study of Mohawk discourse would reveal that specificlinguistic elements are required to introduce this operator; for example,at least one of our consultants does not accept optative matrix verbs inisolation unless a particle such as wahi is present – although no such restric-tion is mentioned in the grammars.


7. C O N C L U S I O N

In this article, we have developed an analysis of the mood-based inflectionalsystem of Mohawk that tries to do justice to both its similarities withtense-based systems and its differences. Once the mood morphemes areanalyzed as definiteness markers for the clause rather than as tenses, variousotherwise mysterious properties of their use fall into place, including theuse of “future” morphology in past habitual contexts, in nomic conditionalcontexts, and in free relatives, as well as the incompatibility of “past”morphology with negation and its inability to appear as the complementof certain matrix verbs. Each of these properties follows from a general-ization of the Kamp-Heim system of NP interpretation to the clausal domain.It turns out, however, that the parallelism is not complete: while Heim’srule of Quantifier Indexing finds many uses in the clausal domain, herNovelty/Familiarity condition does not seem to be relevant. In essence,this means that the distinction between specific definites and specific indef-inites collapses in the verbal domain. We conjectured that this is ultimatelyrelated to the mysterious but fundamental difference between nouns andverbs in human language. Then in order to account for why Mohawk moodsentences turn out so much like English tensed sentences in simple contexts,we invoked the fact that humans conceive of time as unidirectional, withmany possible futures but only one past at any given point. This helps toexplicate the easily observed connections between definiteness and pastness,on the one hand, and indefiniteness and futureness, on the other hand.

As a final remark, we have assumed throughout that the Mohawk systemis fundamentally different from the English system. In fact, it is not soclear that this is so. As a number of people have pointed out to us, mostof the uses of v- in Mohawk can be replicated in English, as long as oneabstracts away from the will/would distinction. In particular, would canbe used to express past habits, in lawlike conditionals, and as a relativefuture under past verbs of speech.

(71) a. When I was a child, my father would take me hiking in themountains.

b. If you caught a dodo bird, it would squeal like crazy.c. John said that Peter would go hunting today.

If one said that would is simply a morphological variant of will that appearsin past tense contexts because of some kind of “sequence of tense” rule,then this English element would have almost the same distribution asv- in Mohawk; in particular, the claims about quantificational variabilitymight well apply to it also. Moreover, it is well known that certain formally


tenseless constructions in English have future-oriented interpretations,including all the modal auxiliaries (Kamp and Reyle 1993) and many to-infinitives (Stowell 1982). It may be that these can be profitably thoughtof as indefinite constructions that get future readings in the same way thatMohawk v- does.31 Exploring these conjectures would go well beyond ourcurrent space and competence; however, it is worth pointing out that ifour proposals are found to be valid for Mohawk and similar languages,this could lead to a re-evaluation of how tense, mood, and definiteness relateto each other even in more familiar Indo-European languages.

R E F E R E N C E S

Aubasch, D: 1994, ‘The Scope of Indefinites’, Natural Language Semantics 2, 83–137.Baker, Mark: 1996, The Polysynthesis Parameter, Oxford University Press, New York.Baker, Mark and Lisa Travis: 1996, ‘Events, States, and Mohawk Verb Inflection’, unpub-

lished ms., McGill University.Carlson, Greg: 1977, Reference to Kinds in English, Ph.D. dissertation, University of

Massachusetts, Amherst.Chafe, Wallace: 1967, Seneca Morphology and Dictionary, Smithsonian Contributions to

Anthropology, vol. 4, Washington, D.C., Smithsonian Press.Chafe, Wallace: 1970, A Semantically Based Sketch of Onondaga, Waverly Press, Baltimore.Chamorro, Adriana: 1992, On Free Word Order in Mohawk, M.A. thesis, McGill University.Chung, Sandra and Alan Timberlake: 1985, ‘Tense, Aspect, and Mood’, in T. Shopen (ed.),

Language Typology and Syntactic Description, Cambridge University Press, Cambridge,pp. 202–258.

Comrie, Bernard: 1976, Aspect, Cambridge University Press, Cambridge.Davidson, Donald: 1967, ‘The Logical Form of Action Sentences’, in N. Rescher (ed.), The

Logic of Decision and Action, University of Pittsburgh Press, Pittsburgh, pp. 81–95.Deering, Nora and Helga Delisle: 1976, Mohawk: A Teaching Grammar, Thunderbird Press,

Kahnawake, Quebec.Diesing, Molly: 1992, Indefinites, MIT Press, Cambridge, Mass. Dowty, David: 1979, Word Meaning and Montague Grammar, Reidel, Dordrecht.Emonds, Joseph: 1985, A Unified Theory of Syntactic Categories, Foris, Dordrecht.Evans, Nicholas: 1991, ‘A Draft Grammar of Mayali’, unpublished ms., University of

Melbourne. Fodor, Jerold and Ivan Sag: 1982, ‘Referential and Quantificational Indefinites’, Linguistics

and Philosophy 5, 355–398.Foster, Michael: 1985, ‘The Language of Tense, Mood, and Aspect in Northern Iroquoian

Descriptions’, International Journal of American Linguistics 51, 403–405.Foster, Michael: 1986, ‘Updating the Terminology of Tense, Mood, and Aspect in Northern

Iroquoian Descriptions’, International Journal of American Linguistics 52, 65–72.Heim, Irene: 1982, The Semantics of Definite and Indefinite Noun Phrases, Ph. D. disserta-

tion, University of Massachusetts, Amherst.


31 On the other hand, the differences between simple past in English and factual in Mohawkare more striking. For example, simple past verbs in English can have habitual interpretations,can appear in nonepistemic conditionals, and are compatible with negation, unlike the factualin Mohawk.

Higginbotham, James: 1985, ‘On Semantics’, Linguistic Inquiry 16(4), 547–594.Kamp, Hans: 1981, ‘A Theory of Truth and Semantic Representation’, in J. Groenendijk,

T. Janssen, and M. Stokhof (eds.) Formal Methods in the Study of Language, MathematicalCentre, Amsterdam, pp. 277–322.

Kamp, Hans and Uwe Reyle: 1983, From Discourse to Logic, Kluwer, Dordrecht.Kratzer, Angelika: 1989, ‘Stage Level and Individual Level Predicates’, unpublished ms.,

University of Massachusetts, Amherst.Lefebvre, Claire: to appear, ‘Multifunctionality and Variation among Grammars: The Case

of Determiners in Haitian and in Fongbe’, Journal of Pidgin and Creole Languages.Lewis, David: 1975, ‘Adverbs of Quantification’, in E. Keenan (ed.), Formal Semantics of

Natural Language, Cambridge University Press, Cambridge, pp. 3–15.Lounsbury, Floyd: 1953, Oneida Verb Morphology, Yale University Press, New Haven.Ormston, Jennifer: 1993, Some Aspects of Mohawk: The System of Verbal Inflectional

Categories, M.A. thesis, McGill University, Montreal, Quebec.Parsons, Terence: 1990, Events in the Semantics of English: A Study in Subatomic Semantics,

MIT Press, Cambridge, Mass.Partee, Barbara: 1984, ‘Nominal and Temporal Anaphora’, Linguistics and Philosophy 7,

243–286.Roberts, Craige: 1989, ‘Modal Subordination and Pronominal Anaphora in Discourse’,

Linguistics and Philosophy 12, 683–721.Rothstein, Susan: 1995, ‘Adverbial Quantification over Events’, Natural Language Semantics

3(1), 1–33.Scott, K. Shirley (ed.), 1991: Old Kahnawake, Kanie’kehaka Raotitiohkwa Press, Kahnawake,

Quebec.Stowell, Timothy: 1982, ‘The Tense of Infinitives’, Linguistic Inquiry 13, 561–570.Stump, Gregory: 1992, ‘On the Theoretical Status of Position Class Restrictions on Inflectional

Affixes’, Yearbook of Morphology 1991, 211–242.Travis, Lisa deMena: 1994, ‘Event Phrase and a Theory of Functional Categories’, in P.

Koskinen (ed.), Toronto Working Papers in Linguistics, Canadian Linguistics Association,pp. 559–570.

Travis, Lisa deMena: forthcoming, Inner Aspect, Kluwer, Dordrecht.Ultan, Russell: 1978, ‘The Nature of Future Tenses’, in J. Greenberg (ed.), Universals of

Human Language, Stanford University Press, Stanford, Cal., pp. 83–123.Watkins, Laurel J.: 1984, A Grammar of Kiowa, University of Nebraska Press, Lincoln.Whorf, Benjamin: 1956, Language, Thought and Reality: Selected Writings of Benjamin L.

Whorf. John B. Carroll (ed.), MIT Press, Cambridge, Mass.Williams, Marianne (ed.): 1976, Kanien’kéha’ Okara’shón:’a (Mohawk Stories), New York

State Museum, Albany.

Department of LinguisticsMcGill University1001 Sherbrooke Street WestMontreal, Québec H3A 1G5CanadaE-mail: [email protected]: [email protected]


mood as verbal definiteness in a 'tenseless'...

Documents