Journal of Semantics 22: 339–387 doi:10.1093/jos/ffh029 Advance Access publication August 24, 2005
Future Discourse in a Tenseless Language MARIA BITTNER Rutgers University
Abstract
Is it possible for a language to have no tense marking at all? And if so, how would one talk about the past, present and future? Typological surveys of tense systems generally report some languages as tenseless—for instance, Lakhota and Chamorro (Chung & Timberlake 1985), or Burmese and Dyirbal (Comrie 1985). But surveys do not offer any detailed evidence and until recently, neither did modeltheoretic semantic literature. By now, however, we have a substantial body of evidence that the following languages may be tenseless: Yukatek Maya (Bohnemeyer 2002), West Greenlandic (Shaer 2003), and possibly Mohawk (Baker & Travis 1997). In addition, the syntactic work of Ritter and Wiltschko (2004) suggests that Blackfoot (Algonquian) and Halkomelem (Salish) may be tenseless as well. At this point, a sceptic may still consider this evidence inconclusive. In this report I aim to provide conclusive evidence for West Greenlandic, which reference grammars describe as tensed, unlike Shaer (2003). I prefer to call this language by its official name, Kalaallisut, to highlight its political status as the language of a country, Kalaallit Nunaat, which includes all of Greenland, not just the west coast. The true name Kalaallisut also highlights the linguistic relation to Inuktitut and other Eskimo languages. Ó The Author 2005. Published by Oxford University Press. All rights reserved.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The Eskimo language Kalaallisut (alias West Greenlandic) has traditionally been described as having a rich tense system, with three future tenses (Kleinschmidt 1851; Bergsland 1955; Fortescue 1984) and possibly four past tenses (Fortescue 1984). Recently, however, Shaer (2003) has challenged these traditional claims, arguing that Kalaallisut is, in fact, tenseless. This paper settles the debate, in favour of Shaer, based on text studies examining how the English future auxiliaries will/would and is/was going to are rendered in Kalaallisut translations of five books: Harry Potter, The Old Man and the Sea, Pippi Longstocking (translated from the Swedish), The Blind Colt, and Black Star, Bright Dawn. The results of these five text studies are reported here in detail and in theory-neutral terms. They conclusively show that Kalaallisut is truly tenseless, but has an alternative system that conveys temporal information, even about the future, as precisely as the English tenses.
340 Future Discourse in a Tenseless Language
1 LIBERAL CRITERIA FOR TENSE Claims of tenselessness are of most theoretical interest if they assume maximally inclusive criteria for what counts as a tense marker. But to be meaningful, the criteria must also reflect the actual use of the term tense in descriptive work. The following quotes from a typological survey of tense, aspect, and mood systems, and from a monograph on tense in English, offer intuitively clear and consistent criteria, which I shall endeavour to apply: d
Tense, aspect, and mood are all categories that further specify [. . .] the basic predication, which can be referred to as the event. Tense
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
My goal here is to report the relevant facts of Kalaallisut in theoryneutral terms and in enough detail to demonstrate two claims. One, that Kalaallisut is truly tenseless, and two, that nonetheless, it conveys temporal information as precisely as the English tenses, but by means of a very different system. This demonstration will include a detailed description of the Kalaallisut alternative to a tense system. I devote special attention to discourses about the future, because it is particularly difficult to imagine how one could talk about the future without any tense. The paper is organized as follows. The claim that Kalaallisut is tenseless is of most theoretical interest if we adopt maximally liberal criteria for what counts as a tense marker. The criteria I adopt are spelled out in section 1. By these criteria, the English tense system includes not just tense inflections, but also two future auxiliaries: will/ would and is/was going to. Though otherwise very different, Kalaallisut is also an inflecting language. Its full inflectional system is described in section 2. This system is clearly tenseless, but it still predictably locates non-future eventualities in time based on aspect and context. Section 3 then turns to a long-standing controversy whether certain preinflectional suffixes in Kalaallisut are semantically equivalent to Germanic future auxiliaries. In section 4 I give a negative answer based on the results of five text studies examining how the English future auxiliaries are rendered in the Kalaallisut translations of five books: Harry Potter, The Old Man and the Sea, Pippi Longstocking (translated from the Swedish), Blind Colt, and Black Star, Bright Dawn. It transpires that the Kalaallisut translations of will/would and is/was going to comprise some thirty morphemes. While morphologically diverse, they form a natural semantic class which I characterize by means of a PROSPECTIVITY THESIS. The claims of this thesis are then supported, point for point, with detailed text-based evidence in section 5. Finally, section 6 presents my conclusions.
Maria Bittner 341
d
By these criteria, all of the italicized elements in English (1) and ChiBemba (2) arguably qualify as tense markers—that is, grammatical markers that locate eventualities in time. (1) English Past TENSE inflections He was asleep. He slept until noon. Present TENSE.ASPECT inflections He is asleep. He sleeps until noon. (habit) Future TENSE.ASPECT.MOOD auxiliaries He will sleep until noon. He would sleep until noon. (habit or hypothesis) Immediate future TENSE.ASPECT auxiliaries He is going to sleep until noon. He was going to sleep until noon. (2) ChiBemba (Givon 1972, Chung & Timberlake 1985: 208, 227–28) Past TENSE.ASPECT inflections ba-a`lı´-bomb-ele ‘they worked (before yesterday)’ ba-a`le´e´-bomba ‘they were working/kept on working/worked repeatedly (before yesterday)’ ba-a`lı´ı´-bomba ‘they worked (yesterday)’ ba-a`cı´-bomba ‘they worked (today)’
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
locates the event in time. Aspect characterizes the internal temporal structure of the event. Mood describes the actuality of the event in terms such as possibility, necessity, or desirability. [. . .] an event occurs on an interval of time, the event frame [MB: reference time]. Tense locates the event in time by comparing the position of the frame with respect to the tense locus [MB: now or then]. (Chung & Timberlake 1985:202–3) There are essentially two kinds of linguistic devices that allow the speaker to express temporal reference. The first one is TENSE, that is, a systematic grammatical marking of the verb (by affixes, vowel alternations, auxiliaries, sometimes particles); . . . The other devices are adverbials of different types (yesterday, next week, when the saints go marching in, etc) (Klein 1994:16)
342 Future Discourse in a Tenseless Language ba-a`-bomba ‘they worked (within the last three hours)’ Present TENSE.ASPECT inflections ba-le´e´-bomba ‘they are working’ ba-la`-bomba ‘they work’ Future TENSE.ASPECT inflections ba-a`la´a´-bomba ‘they’ll work (soon, within three hours)’ ´ ´ ba-lee-bomba ‘they’ll work (later today)’ ba-ka`-bomba ‘they’ll work (tomorrow)’ ba-ka´-bomba ‘they’ll work (after tomorrow)’
2 KALAALLISUT INFLECTIONS Kalaallisut is a polysynthetic language (in the sense of Sapir 1922) with rich inflection. There are two classes of inflected words: nouns and verbs. Most nouns and verbs are built according to parallel templates:2 (3) Noun template: root-(derivational suffix*)-NAGR-CASE(¼clitic*) (4) Verb template: root-(derivational suffix*)-MOOD-VAGR(¼clitic*) That is, a root may combine with any number of derivational suffixes. The result is a base. The last morpheme of the base selects the form of the inflection, nominal or verbal. Nouns inflect for nominal 1
I am not committed to Klein’s view: arguably, is/was going to is better analysed as an instance of the present/past tense plus an aspectual auxiliary be going to. My point is that even if one allows tense auxiliaries to be complex, one still will not find any in Kalaallisut. 2 In addition to nouns and verbs, there are particles (e.g., irniinnaq ‘soon’ in (7#)), which do not inflect but can combine with clitics. There are also a few derivational clitics, which attach to fully inflected words or particles and derive new bases (e.g., ¼it- ‘be’ in (27#), ¼kar- ‘go’ in (33#)).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
It is not required that a tense marker only mark temporal location: it may also mark, e.g., aspect or mood. As (1) and (2) suggest, and Chung and Timberlake (1985) confirm, such fusion is common. A tense marker may also be morphologically complex—like the auxiliary is going to, which Klein (1994: 19) counts as a future tense auxiliary, along with will.1 A tense system is a grammatical paradigm such that exactly one term of this paradigm—one tense marker—is required to form a grammatical finite clause. This distinguishes, e.g. the metrical tense system of ChiBemba from the temporal adverbs of English. The grammar of ChiBemba requires the tense affixes italicized in (2). In contrast, the grammar of English does not require yesterday or any other adverb to form a finite clause. What it requires is a present or past tense inflection or a future tense auxiliary.
Maria Bittner 343
agreement and case, as in (5), whereas verbs inflect for mood and verbal agreement, as in (6).3 Finally, either inflection can be followed by any number of clitics. (5)
Direct N inflection: NAGR-CASE (rn-paradigm) Nominative (iv subject, Ergative (tv subject, possessor) tv object) qimmi-kka qimmi-ma dog-1s.pl dog-1s.pl.ERG ‘my dogs’ ‘(of ) my dogs’
(6)
3
(rn-paradigm) Ablative (source) qimmi-n-nit dog-1s.pl-ABL ‘from my dogs’
NAGR-CASE
Locative (location) qimmi-n-ni dog-1s.pl-LOC ‘among my dogs’
Vialis (path) qimmi-k-kut dog-1s.pl-VIA ‘via my dogs’
Modalis (modifier) qimmi-n-nik dog-1s.pl-MOD e.g. ‘with my dogs’
Equative (standard) qimmi-t-tut dog-1s.pl-EQU ‘as my dogs’
Matrix V inflection: Indicative (assertion) Sinip-pu-q. sleep-IND.IV-3s ‘He is/was asleep.’
(iv-paradigm) Interrogative (question) Sinip-pa? sleep-QUE.3s ‘Is/was he asleep?’
MOOD.CENTERING-VAGR
Irrealis (negation) Sini-nngi-la-q. sleep-not-IRR-3s ‘He is/was not asleep.’
Sini-nngi-la? sleep-not-IRR.3s ‘Is/was he not asleep?’
Optative (wish) Sinig-(niar)-li. sleep-(please)-OPT.3s ‘(Please) let him sleep.’
Imperative (request) Sinig-(niar)-it! sleep-(please)-IMP.2s ‘(Please) go to sleep!’
I use the standard Kalaallisut orthography, minus the allophones (e, o, f ) of i, u, v.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Oblique N inflection: Dative (goal) qimmi-n-nut dog-1s.pl-DAT ‘to my dogs’
344 Future Discourse in a Tenseless Language Dependent V inflection: MOOD.ASP.CENTERING-VAGR (iv-paradigm) Topic factive Background factive sinik-ka-mi sini-mm-at sleep-FCT>-3s> sleep-FCT?-3s? ‘because/when he? ‘because/when is/was asleep’ he> is/was asleep’ (‘-not’ + Background elab.) sini-nngit-su-q sleep-not-ELA?.IV-3s? ‘while he? is/was not asleep’
Topic hypothetical sinik-ku-ni sleep-HYP>-3s> ‘if/when he> is/was asleep’
Background hypothetical sini-pp-at sini-HYP?-3s? ‘if/when he? is/was asleep’
Topic habitual sinik-kaanga-mi sleep-HAB>-3s> ‘whenever he> is/was asleep’
Background habitual sinik-kaang-at sleep-HAB?-3s? ‘whenever he? is/was asleep’
Topic elaborating sinil-lu-ni sleep-ELA>-3s> ‘being asleep’
Background elaborating sinit-tu-q sleep-ELA?.IV-3s? ‘while he? is/was asleep’
The proper analysis of the system of verbal inflection in Kalaallisut is a matter of some controversy, summarized in Table 1. In several cases, what I analyse as a difference in meaning—mood, aspect, or centering—has traditionally been viewed as morphological suppletion. But all analyses agree that this inflectional system does not include any tense markers. Unlike the English or ChiBemba tenses, verb inflections in Kalaallisut do not mark temporal location. For instance, consider the factual moods, which introduce (IND), presuppose (FCT), or inquire (QUE) about facts. All of these moods in Kalaallisut allow eventualities to be either past or present, as the discourses in (7#) through (10#) illustrate. But as these discourses also illustrate, no temporal ambiguity arises in actual use, because the relevant temporal interpretation is predictable based on the aspect and the context. Example (7#) illustrates these points for event-predicates. The base of each verb ends in an event-predicate: -si ‘become’, angirlar- ‘come
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Topic non-factual sinin-na-ni sleep-NON>-3s> ‘not/without sleeping’
Maria Bittner 345 Table 1 Three analyses of Kalaallisut verbal inflections Form(s)
Kleinschmidt 1851
Bergsland 1955
This work
-vu -va -va/-vi -la -la/-li -la/-li -gi/Ø -ga
Indicativ Indicativ Interrogativ
Indicative iv form Indicative tv form Interrogative Indicative neg. form Interrogative neg. form Optative Imperative Causative 4th person form Causative 3rd person form Contemporative neg. form Conditional 4th person form Conditional 3rd person form Caus. habitual 4th pers. form Caus. habitual 3rd pers. form Contemporative Participial iv form Participial tv form
IND.IV
Optativ/Imperativ Optativ/Imperativ Conjunctiv
-na -gu
Subjunctiv
-pp -gaanga -gaang -llu -su -ga/gi
Infinitiv Particip Particip
QUE IRR IRR OPT IMP FCT> FCT? NON> HYP>
HYP? HAB> HAB? ELA> ELA?.IV ELA?.TV
home’, or -lir ‘begin’. Since the mood is factual (FCT or IND), all of these events are reported as facts. In Kalaallisut an event can be reported as a fact only if it has already happened. So (7#) is equivalent to the English past tense (7). (7) I came home when it got dark. Juuna soon fell asleep. angirlar-pu-nga. (7#) Taar-si-mm-at be.dark-become-FCT?-3s? come.home-IND.IV-1s Juuna Juuna
irniinnaq soon
sini-lir-pu-q. sleep-begin-IND.IV-3s
Other aspectual classes are understood to be current at the topic time—i.e. now, the discourse-initial default, or the explicitly
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
-mm
IND.TV
346 Future Discourse in a Tenseless Language
Current states (now or then) (8) a. Juuna is sleeping peacefully. b. I came home after dark. Juuna was sleeping peacefully. (8#) (Taar-si-riir-m-at (be.dark-become-be.already-FCT?-3s? Juuna Juuna
iqqissi-sima-llu-ni calm.down-prf-ELA>-3s>
angirlar-pu-nga.) come.home-IND.IV-1s)
sinip-pu-q. sleep-IND.IV-3s
In (9#) and (10#) analogous temporal shifts are illustrated for processes and habits, and for topic time updates by prior dependents (a particle in (9#), a dependent verb in (10#)), instead of prior discourse. Current processes (now or then) (9)
a. Juuna> and his> father are playing chess. b. Juuna> and his> father were playing chess a moment ago.
(9#) Juuna ataata-ni¼lu (aatsaannguaq) skakkir-pu-t. Juuna father-3s>.sg¼and (moment ago) play.chess-IND.IV-3p Current habits (now or then) (10) a. I come home when it starts getting dark. Juuna soon falls asleep. b. Last winter I would come home when it would start getting dark. Juuna would soon fall asleep.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
determined then. This is illustrated in (8#)–(10#), for states, processes, and habits. First, consider state-predicates. If we leave out the first sentence in (8#), then the result state (-sima ‘prf ’) of calming down and sleeping holds at the default topic time—i.e., now. So this version of (8#) translates into the present tense (8a). But with the parenthesized topic time update, the state of peaceful sleep instead holds at the new topic time—(today) after dark, when the speaker came home. So this version of (8#) is again unambiguous, but now equivalent to the past tense (8b). In Kalaallisut this temporal shift involves no change in the form of the verb, just a change in the context.
Maria Bittner 347
(10#) (Ukiur-m-at) (be.winter-FCT?-3s?) taar-si-lir-aang-at be.dark-become-begin-HAB?-3s?
angirlar-tar-pu-nga. come.home-habit-IND.IV-1s
Juuna irniinnaq sini-lir-tar-pu-q. Juuna soon sleep-begin-habit-IND.IV-3s
Current states (now or then) of attitudes to future prospects (11) a. I will come home at dusk. Juuna will soon fall soundly asleep. b. (That morning I made a plan.) I would come home at dusk. Juuna would soon fall soundly asleep. (11#)(Ullaaq taanna pilirsaaru-siur-pu-nga.) (morning that plan-make-IND.IV-1s) Taar-si-lir-p-at be.dark-become-begin-HYP?-3s? Juuna Juuna
angirlar-niar-pu-nga. come.home-intend-IND.IV-1s
irniinnaq iti-ssa-na-ni soon wake.up-be.expected-NON>-3s>
sini-lir-umaar-pu-q. sleep-begin-be.hoped-IND.IV-3s For now, I just draw the two conclusions boxed below. The first conclusion is negative and generally agreed on (by Kleinschmidt 1851, Bergsland 1955, and Fortescue 1984, who on this point follows Bergsland): Kalaallisut inflectional system is tenseless. Even by liberal criteria, it does not contain any tense inflection, be it pure or fused, simple or complex. The second conclusion depends on more controversial details of my analysis of the Kalaallisut mood system. By the end of this
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Future discourse in Kalaallisut is exemplified in (11#). In a nutshell, what I will argue is that this is just another instance of the same temporal pattern—more precisely, of the stative variant (8#).
348 Future Discourse in a Tenseless Language report, the reader may have enough evidence to evaluate this proposal: Kalaallisut verbal inflections form a mood system, fused with aspect and centering, which contrasts facts, nonfacts, prospects, and circumstances: d d d
d
Factual moods: indicative (IND), interrogative (QUE), factive (FCT) Nonfactual moods: irrealis (IRR), non-factual (NON) Prospective moods: optative (OPT), imperative (IMP), hypothetical (HYP) Circumstantial moods: elaborating (ELA), habitual (HAB)
3.1 Reference grammars I propose to subsume future discourse of type (11#) under stative discourse of type (8#) by analysing -niar, -ssa and -jumaar as statepredicates that refer to mental states of future-oriented attitudes—intent, expectation/desire, or hope/dread. This analysis constitutes a major departure from the traditional view of these suffixes—as bound future tense auxiliaries. Kleinschmidt (1851: 148) lists -ssa and -jumaar as v\v suffixes ‘mit hu¨lfsverben-bedeutung’. According to him, -jumaar ‘entspricht . . . unserm Futurum’, while -ssa ‘dru¨ckt ebenfalls ha¨ufig das Futurum aus, schliesst aber immer zugleich den begriff des sollens in sich.’ Bergsland (1955: 115) agrees, adding that -jumaar indicates ‘especially nonimmediate future’. Fortescue (1984: 274–5) has more to add: 2.1.3.2.1.4. Future. There are three common affixes indicating future tense ssa, niar, and jumaar. The first of these also has the modal value ‘should/ shall’, covering in fact most of the senses of Danish ‘skal’; as a future marker it is the nearest to an absolute or pure future, as in: tuqu-ssa-atit, die-future-2s.indic, ‘You will die’ (e.g. if you drink the poison) [. . .] niar is more common as a pure future in north-west Greenland, but is used especially in an intended or inevitable future sense in the central dialect area, as in atuarniarpara ‘I’m going to read it/will try reading it’ and siallirniarpuq ‘it’s going to rain’. As a verb-extending affix it means ‘try to’ [. . .]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
3 CONTROVERSY ABOUT KALAALLISUT ‘TENSES’
Maria Bittner 349
jumaar indicates vague indefinite future (cf. related affix -juma ‘want to’) as in the parting formula takuqqikkumaarpugut ‘we’ll see each other again’.
3.2 Shaer (2003) Fortescue’s (1984) claim that Kalaallisut has three future and four past tenses has recently been challenged by Shaer (2003), who conjectures that Kalaallisut is, in fact, tenseless. Shaer speculates that all of Fortescue’s ‘tenses’ should be analysed as ‘bound adverbs’. But he does not offer any evidence in favour of this idea over the traditional view of these suffixes as bound auxiliaries. So I shall only address the evidence that Shaer does present—that whatever these suffixes might be, they are not tenses. Shaer argues at length that Fortescue’s ‘tense’ suffixes precede the inflection. That is correct—and uncontroversial. It is also irrelevant, according to the more liberal criteria for tense adopted here (pace Klein 1994). Tenses are often morphologically realized as inflections. But they can also be auxiliary verbs, so pre-inflectional affixes do not seem inconceivable. In my view, Shaer’s strongest argument is that the putative ‘past’ and ‘future’ suffixes can co-occur—as -sima and -ssa in (12): (12) Atur-sima-ssa-va-a. use-sima-ssa-IND.TV-3s.3s ‘He must have used it.’
(Fortescue 1980: 267)
According to Fortescue (1980), -sima-ssa is a ‘semilexicalization’. But its use in Kalaallisut is similar to epistemic will have in English, e.g., in (He stole this, but we can’t return it.) He will have used it already. The English will have is not lexicalized, so perhaps the Kalaallisut -sima-ssa is not lexicalized either. Assuming that, -sima and -ssa cannot both be tenses.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
According to Fortescue (1984: 272–3), Kalaallisut also distinguishes four ‘past tenses’: -sima, -nikuu, -qqamir, -riikatag. The language would thus appear to have a rich tense system. However, older grammars list -sima not as a past tense, but as a perfect aspect (Kleinschmidt 1851: 148; Bergsland 1955: 116). That is also how all of Fortescue’s ‘past tenses’ are glossed in the dictionaries: -sima ‘have . . . ed’, -nikuu ‘have once . . . ed’, -qqamir ‘have just . . . ed’, -riikatag ‘have . . . ed long ago’. One may therefore begin to wonder about the accuracy of Fortescue’s claims.
350 Future Discourse in a Tenseless Language
4 EVIDENCE FROM TEXTS: PROSPECTIVITY THESIS According to the grammarians, the Kalaallisut suffixes -ssa, -niar, and -jumaar are used like Germanic future auxiliaries. That is a testable claim. To test it, I took the following five pairs of English and Kalaallisut texts (see Appendix A for full references) and systematically examined the Kalaallisut equivalents of the English future auxiliaries will/would and is/was going to. [D] [D#] [H] [H#] [L] [L#] [P] [P#] [R] [R#]
O’Dell, S. 1988. Black Star, Bright Dawn. O’Dell. S. 1994. Milalinnguaq Sikkersorlu. Hemingway, E. 1952. The Old Man and the Sea. Hemingway, E. 1991. Angutitoqaq imarlu. Lindgren, A. 1997. Pippi Longstocking. Lindgren, A. 2000. Pippi Langstrømpe—ikinngutaalu. Rounds, G. 1941. The Blind Colt. Rounds, G. 1988. Hesti piaraq tappiitsoq. Rowling, J. K. 1997. Harry Potter and the Philosopher’s Stone. Rowling, J. K. 2002. Harry Potter ujaraallu inuunartoq.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
But note that one of them still could be. In fact, if we accept what the dictionaries and the older grammars say—that -sima marks perfect aspect—then we might even take (12) to support the traditional view that the Kalaallisut -ssa corresponds to the English future auxiliary will/ would. Unfortunately, most of Shaer’s examples are similarly inconclusive. This is fatal for his argument that Kalaallisut is tenseless. For given Shaer’s evidence, a sceptic could still maintain that -ssa, at least, looks like a future tense. So Kalaallisut would still have a tense system, albeit less rich than Fortescue’s and of a different type. It would be the type that distinguishes future (-ssa, -niar, -jumaar) from nonfuture (unmarked)—like, for example, Hua (Haiman 1980, Comrie 1985: 46). The key examples of apparently non-future -ssa in Shaer’s paper are attributed to ‘Bittner (2002)’—my 2002 text study posted online (now at http://www.rci.rutgers.edu/;mbittner/kal.html) and formally reported in Bittner (2003). But since that text study was not specifically designed to determine the status of the Kalaallisut ‘future tenses’, I shall not rehearse this—in my view, still inconclusive—evidence here. Instead, I shall report on five other text studies that, to my mind, leave no room for doubt.
Maria Bittner 351
I considered only clearly future uses (relative to the current topic time, now or then). Any habitual or counterfactual uses of will/would were ignored. For the future uses of both will/would and is/was going to, I found essentially the same list of Kalaallisut translations. The combined list is presented in Table 2. Table 2 Kalaallisut translations of English future auxiliaries Proposed English gloss
A1. A2. A3. A4. A5. A6. A7.
ilimagi- (tv) ilimanar- (iv) -gunar (v\v) -juma (v\v) -jumaar (v\v) -navianngi (v\v) -niar (v\v)
A8. A9. A10.
-qina (v\v) -qqajaa (v\v) -qqu (v\tv)
A11.
-riaannaa (v\v)
A12. A13. A14. A15. A16. A17. A18. A19. A20. A21. A22. A23. A24.
-rusuk (v\v) sapir- (v) sapirnar- (iv) -sariaqar (v\v) -sinnaa (v\v) -ssa (v\v) -ssaq (n\n) -ssamaar (v\v) -ssanga (v\v) -ssangatip (v\tv) -ssaqqaar (v\v) -ssaqqip (n\iv) ..-tsir, -tsii (v\tv) . -lir .. (v\v) . -li/-la (OPT) -niar-li (please-OPT) -niar (please-IMP)
expect be likely be likely to want to, be willing to (be) hope(d), (be) dread(ed) be very unlikely to (be) intend(ed), be about to, (process use: try) be liable to be just about to, be ready to want ? to, enable ? to, (event use: order) be well able to, be easily possible for > to wish to, feel like be unable to, not dare to be impossible, be hard to do need to, be necessary for > to be able to, be possible for > to (be) expect(ed), (be) desire(d) expected, desired plan to, be about to expect to expect ? to be confidently expected be suitable for ..wait for ? to . begin .. . let >. . . please let >. . . please do. . . , please let us. . .
B1. C1. C2. C3.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Kalaallisut item
352 Future Discourse in a Tenseless Language
A. Prospective statives (13) They’ll all forget this in a few weeks. (13#) Sapaati-t Sunday-pl.ERG
akunnir-i space-3p?.pl
marlussuit a.few
[R: 265] qaangiu-pp-ata pass-HYP?-3p?
puigur-unar-pa-at. forget-be.likely-IND.TV-3p.3s
[R#: 302]
(13$) They’re likely to forget this in a few weeks. (14) [Muggles] don’t know we’re not allowed to use magic at home. I’m going to have a lot of fun with Dudley this summer. [R: 332] (14#) Ilimaga-a-ra aasaq manna Dudley expect-IND.TV-1s.3s summer this Dudley quianar-tur-si-vvigi-ssa-llu-gu. be.fun-iv\cn-get-from-expect-ELA>-3s?
[R#: 380]
(14$) I expect to get some fun out of Dudley this summer. B. Prospective inchoatives (15) Yessir! I’ll bring him right over!
[P: VIII]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
It transpires that the ‘future tense auxiliaries’ in Kalaallisut comprise not just -ssa, -niar, and -jumaar, but nearly thirty morphemes: nineteen verb-extending suffixes (sixteen transitivity preserving v\v, three transitive-deriving v\tv), four verbal roots (tv, iv, one complex predicate forming v), one noun-extending suffix (n\n), one de-nominal verb-forming suffix (n\iv), and three mood inflections. That is not a natural morphological class. Semantically, on the other hand, this is clearly a natural class. It falls into three subclasses—A, B, and C—exemplified below. The examples are presented as follows: first the English version (e.g. (13)), usually with some context; then the published Kalaallisut translation (13#); and finally my own attempt to render the Kalaallisut, as literally as possible, back into English (13$). To get a sense of the strictly literal English translation—usually intelligible, albeit not fully grammatical—the reader is advised to read the Kalallisut glosses from the end, adding each suffix to the entire base.
Maria Bittner 353
(15#) Aap, akkaa. yes uncle
Aggiuti-lir-pa-ra! bring-begin-IND.TV-1s.3s
[P#: 53]
(15$) Yes, Uncle. I’m coming with him! C. Prospective matrix moods (16) We were running twenty-first and twenty-second in the race. Oteg said, ‘We will feed the dogs now and rest until night comes.’ [D: 44] (16#)
Qimmi-t dog-pl
nirukkar-niar-tigik. feed-please-IMP.1p.3p
[D#: 55]
The English glosses in Table 2 are justified in Appendix B, by examples from the same five texts, plus three dictionaries (examples (a)). Based on this evidence, I take the English glosses in Table 2 to be correct. Assuming that, the main translation strategy involves class A items, which I dub prospective statives. These are glossed as stative predicates that evoke future-oriented mental states—expectation (e.g. (14#)), desire, hope, dread, intent, plan, considering a prospect likely (e.g. (13#)), unlikely, liable to happen, or the like. If the predicate is impersonal (e.g., -gunar in (13#)) then the experiencer of the mental state is grammatically unspecified (as in the English gloss ‘be likely’). But the context often suggests a plausible experiencer—usually, the speaker (here, Ron) or the topic. This translation strategy involves paraphrasing the English future auxiliary away in favour of a contextually plausible prospective stative. The mental state—here, considering a certain mishap likely to be forgotten in a few weeks, or expecting to have fun with Dudley this summer—holds at the current topic time. In (13#) and (14#) this means now, the discourse-initial default. The current mental state is reported as a fact, in a factual mood (IND). For indeed, the current state of expectation is a fact, even if the expected prospect should fail to become a reality. Intuitively, (15) is a plan for the immediate future—equivalent to the present progressive paraphrase in (15$). In Kalaallisut such plans can be rendered by class B items, which I dub prospective inchoatives. In (15#) aggiut- ‘bring-’ is interpreted as an expected process. The prospective inchoative -lir ‘begin’ evokes the first stage, and the factual mood (IND) reports this start event as a done deed. Normally, events
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(16$) Let us feed the dogs, ok?
354 Future Discourse in a Tenseless Language
PROSPECTIVITY THESIS Kalallisut translations of future auxiliaries comprise three related classes: A. prospective statives evoking (current) attitude states to de se prospects, B. prospective inchoatives evoking (realized) starts of expected processes, C. prospective matrix moods marking the speech act as a request or wish. Class B and C items are clearly not tenses, so I focus on class A. The remaining task is to clarify and support my description of this class. 5 PROSPECTIVE STATIVES: EVIDENCE IN DETAIL
5.1 Prospective statives are not tenses Before I elaborate on what prospective statives are, let me elaborate on what they are not. Reference grammars list three of these items, -ssa, -niar, and -jumaar, as ‘future tenses’ (section 3.1). But already a glance at Table 2 makes this dubious. For the putative ‘future tenses’ are listed in Table 2 along with over twenty similar items. A tense system with three future tenses is typologically plausible—ChiBemba has four, English, arguably two. But twenty four future tenses, and still counting?! That is unheard of. Turning doubt into certainty, prospective statives in Kalaallisut do not form a grammatical paradigm. Shaer (2003: 10) claims that ‘-ssa 4 Kalallisut translators often disregard sentence boundaries, as in (16#). But they respect paragraph boundaries. I take it that the paragraph is the intuitive unit of equivalence.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
are understood to occur during the topical period. But when the topic time is an instant—such as the default now in (15#)—then events are instead required have a current result state. So (15#) literally means: I have (already) started to bring him. The remaining future auxiliaries in the studied texts express requests or wishes for the future—e.g. (16). These translate into prospective moods—class C—here, the imperative (16#),4 with the literal meaning (16$). Table 2 probably lists most, but not all of the translation options for the English future auxiliaries. Further text studies may add, e.g. -junnar(si) ‘be likely, be hoped for’ and -lirsaar ‘plan to’ to class A, -jartur ‘go to’ to class B, and -gi/Ø ‘IMP’ to class C. But the proposed descriptive generalizations should hold. In other words, I propose the following thesis:
Maria Bittner 355
(17) a. John {will, is going to} be going to Paris soon. b. *John will be going to be going to Paris soon. In contrast, prospective statives in Kalaallisut commonly co-occur. Thus, in (18#) -niar and -(g)unar combine in a clearly compositional way to yield the meaning predicted by their glosses, ‘intend’ and ‘be likely’: (18)
When she heard the dogs bark, the cow [moose] took to the stream and disappeared. But the bull stood sideways on the bridge and did not
move. (18#) (18$)
[D: 53]
. . . nikin-niar-unar-a-ni¼lu . . . budge-intend-be.likely-NON>-3s>¼ and
[D#: 65]
. . . and was not likely to intend to budge.
In (19#) two of the putative ‘future affixes’, -niar and -ssa, combine via eventuality anaphora: -niar evokes a state of intent, and -ssa adds that it is also a desire (19$). This compositional strategy instantiates a very general anaphoric phenomenon, which I dub verb doubling (see section 5.2). (19)
Eat [the bait] so that the point of the hook goes into your heart and kills you, he thought. [H: 44]
(19#) . . . qarsursaq uummati-n-nut . . . hook heart-2s.sg-DAT
apissigul-lu-gu have.?.go.all.way.in-ELA>-3s?
tuqqutigi-nia-ssa-va-t die.from-be.intended-be.desired-IND.TV-2s.3s
[H#: 29]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
appears to be (virtually) obligatory in sentences that refer to the future’, but this claim is misleading. Consider English: if one wants to talk about dogs, one has to use the item dog or equivalent. But the grammar of English does not require any such item to form a finite clause. What it requires is a tense inflection or auxiliary. Similarly in Kalaallisut, if one wants to talk about the future, one has to use -ssa or equivalent. But the grammar of Kalaallisut does not require any such item to form a finite clause. What it requires is a mood inflection and verbal agreement. It is the grammar that is relevant for tenses and, as far as the grammar goes, prospective statives are optional. Further evidence that prospective statives in Kalaallisut do not form a grammatical paradigm comes from co-occurrence tests: in any finite clause there cannot be more than one tense marker. If we apply this test to the English will/would and is/was/be going to, they behave like true tense auxiliaries. That is, they cannot co-occur, as (17) attests:
356 Future Discourse in a Tenseless Language (19$) . . . [I] intend and desire that you die from having the hook go all the way into your heart.
(20) a. He {is willing to, intends to, will, is going to} help me. b. His {being willing to, intent to, *willing, *being going to} help me pleased her. In Kalaallisut the nominalization test confirms the diagnosis of the co-occurrence test: prospective statives are predicates, not tenses. In particular, this is the diagnosis for the putative ‘future tenses’ -ssa, -niar, and -jumaar. All three survive nominalization, as (21#), (22#), and (23#) attest. (21) [Whitey] pictured himself riding up in front of the store where all the loafers were gathered, and guessed he’d teach the colt to kneel to let him get off like he’d heard the Arabs or some such fellers did. [P: VII] (21#) . . . isuma-liur-pu-r¼lu . . . idea-make-IND.IV-3s¼and histi piaraq ilinniar-tin-niar-lu-gu horse young learn-cause-intend-ELA>-3s? aqqa-lir-aanga-mi get.down-begin-HAB>-3s> aqupit-tar-ni-ssa-a-nik. kneel-habit-v\n-desired-3s?.sg-MOD
[P#: 47]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Fortescue (1984:64) is aware that -nia-ssa is a possible sequence. But he still maintains that -niar and -ssa are ‘tenses’ because for him, -niassa is lexicalized as a single ‘affix’ (like -sima-ssa in (12)). But his lexical account lacks empirical support: the dictionaries do not list any ‘affix’ -niassa, only -niar and -ssa. I take it that native speakers view the sequence -nia-ssa as compositional. Also, the putative ‘affix’ -niasssa is supposed to mean ‘in order to’. But that is not what -nia-ssa means in (19#). The indicative mood marks (19#) as a main clause and a factual report. Fortescue’s gloss ‘in order to’ does not fit the bill; the compositional paraphrase in (19$) does. Finally, consider evidence from nominalization. Recall that tense markers are not part of the main predicate, but only locate it in time (section 1). Accordingly, tense markers—including tense auxiliaries—are lost under nominalization, whereas otherwise similar raising and control predicates survive. The minimal pairs in (20) illustrate this diagnostic test for English. Like the co-occurrence test, it distinguishes tenses from predicates.
Maria Bittner 357
(21$) . . . and he> had an idea: he> intended to teach the colt a nice trick of kneeling whenever he> started to get down. (22) The whole crowd were on their feet, watching, terrified, as the Weasleys flew up to try and pull Harry safely on to one of their brooms, but it was no good—every time they got near him, the broom would jump higher still. They dropped lower and circled beneath him . . . [R: 207]
(22#). . . qatannguti-gii-t brother-rcp.group-pl.ERG
Weasley-kku-t Weasley-group-pl.ERG
Harry-p tunga-a-nut Harry-ERG direction-3s?.sg-DAT
saniguti-min-nut nuu-tin-niar-lu-gu. broom-3p>.pl-DAT move-cause-intend-ELA>-3s?
. . . Annaa-ssi-nia-t . . . rescue-apass-intending-pl
apparsar-pu-t . . . drop.lower-IND.IV-3p
[R#: 236]
(22$) . . .and they watched in horror the Weasley brothers’ flight toward Harry, with the intent to move Harry on to one of their brooms. [. . .] The would-be rescuers dropped lower . . . (23) I had heard about the Iditarod for years . . . but . . . I had never dreamed in my wildest dreams that someday I would march in a band playing my silver horn . . . and welcome racers of the famous Iditarod. [D: 21] (23#) Sinnatturi-sima-nngisaannar-pa-ra¼li dream.of-prf-never-IND.TV-1s.3s¼but sukka-niut-tu-nik go.fast-compete-iv\cn-pl.MOD tikilluaqqu-si-qata-u-jumaar-ni-ssa-ra welcome-apass-fellow-be-hope-v\n-desired-1s.sg nipilirsu-qatigiin-ni make.music-coll.group-pl.LOC trumpiti-mik trumpet-MOD
qarlurtaati-qar-lu-nga wind.instr.-have-ELA>-1s
[D#: 30]
(23$)But I had never dreamed of realizing my wish to have a chance to participate in welcoming these racers, as a trumpet player in a band of musicians . . .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Harry Harry
timmi-nir-at fly-v\n-3p?.sg
358 Future Discourse in a Tenseless Language In summary, the Kalaallisut suffixes -ssa, -niar, and -jumaar are grammatically optional, they can co-occur, and they survive nominalization. On every diagnostic test these putative ‘future tenses’, as well as other prospective statives in Kalaallisut, behave like predicates, not tenses. That is correctly predicted by the Prospectivity Thesis, which thus receives initial empirical support. I now proceed to demonstrate this thesis point for point. Specifically, I show that prospective statives refer to states (section 5.2); that these states are current (section 5.3); that they are attitude states (section 5.4); and finally, that the attitudes concern the experiencer’s future prospects viewed from his own perspective (section 5.5).
In Kalaallisut basic verbal meanings fall into four aspectual classes: (24) State-predicates: sinig- ‘sleep’, -sima ‘prf ’, ¼it- ‘be’, . . . Event-predicates: tikit- ‘come’, -lir ‘begin’, ¼guuq ‘x say’, . . . Process-predicates: aggir- ‘approach’, -liur ‘make’, -tur ‘use’, . . . Habit-predicates: iliqqur- ‘habit.of ’, -tar ‘habit’, -gajut ‘do.often’, ... Reliable diagnostic tests include temporal anaphora (discussed in section 2 and further in section 5.3; see also Bittner 2005) and a species of eventuality anaphora I dub verb doubling (recall (19#)). In verb doubling a discourse referent (Karttunen 1976) for an eventuality is introduced and then referred to in a manner analogous to A doctor came in. I immediately trusted this lady. Anaphorically linked verbs constrain the same eventuality (modulo bridging anaphora), and therefore must be of the same aspectual class. That is the basis of this aspectual test. Before using it as a test, let me elaborate on my characterization of verb doubling as a verbal analogue of nominal anaphora. In Kalaallisut verb doubling may involve any combination of verbal roots, suffixes, or clitics. Anaphoric chains may be of any length and may span several sentences in discourse. Elaborating verbs (ELA) are anaphorically linked to verbs they elaborate, unless this default is defeated, e.g., by an explicit connective. For instance, in (25#) there is a three-membered chain of eventpredicates introducing and then constraining a single event—a gesture: (25) (25#)
I called to my father [waiting by a seal’s breathing hole] to give up the hunt . . . I called again. Slowly he raised a hand to quiet me. [D: 8]
Ataata-ma¼li arriillillar-lu-ni father-1s.sg.ERG¼but move.slowly-ELA>-3s>
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
5.2 Prospective statives evoke states
Maria Bittner 359
talli-mi aappa-a-nik arm-3s>.sg.ERG pairmate-3s?.sg-MOD nipaar-saa-qqu-llu-nga. yell-stop-order-ELA>-1s (25$)
ussirar-pa-anga signal-IND.TV-3s.1s [D#: 15]
But my father, moving slowly, signalled to me with one arm, ordering me to stop yelling.
(26)
My mother . . . was crying with happiness. Tears ran down her red cheeks. She had been cooking for a long time, getting food ready for me. [D: 75]
(26#)
Qulli-ngiariar-lu-ni uqar-pu-q tear(s)-shed-ELA>-3s> say-IND.IV-3s qangali¼li long.ago¼since taaka¼guuq there¼x.say
(26$)
nirisas-siu-uti-lir-sima-nirar-lu-nga, foodpl-make-for-begin-prf-say-ELA>-1s
piariir-aluar-pu-t be.ready-unrealized-IND.IV-3p
[D#: 89]
Crying, she spoke: she reported having started to cook foodpl for me long ago; here, [she] said, itpl’s ready (if you want it).
Turning now to canonical state-predicates, in (8#) -sima ‘prf ’ evokes the result state of calming down, which is then elaborated by sinig‘sleep’. Likewise in (27#), kisimiit- and ¼it- co-specify an undesirable state concept: (27)
That car’s new, he’s not sitting in it alone.
(27#)
Kisimiil-lu-ni be.alone-ELA>-3s>
(27$)
[I] don’t want him to be alone in the car.
[R: 30]
biili-mi¼i-ssa-nngi-la-q car-sg.LOC¼be-be.desired-not-IRR-3s [R#: 30]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Eventuality anaphora may serve in lieu of nominal anaphora. For instance, the evidential clitic ¼guuq ‘x say’ is an impersonal event-predicate. That is, it only evokes a discourse referent for a speech event, without specifying its agent. But in (26#) this speech event, where the narrator heard what she reports here, is anaphorically linked to the event introduced by the root uqar‘say’. Since uqar- does specify the agent (mother), the event-level anaphoric link in effect identifies the unspecified agent of ¼guuq.
360 Future Discourse in a Tenseless Language Crucially, prospective statives can form anaphoric chains with canonical state-predicates. Thus, in (28#) the same state (wait) is referred to by the canonical state-predicate utaqqi- and the prospective stative -(t)sii: (28) [Two tramps want to rob Pippi.] The tramps waited quite a while until they were sure Mr. Nilsson would have gone to sleep. [L: 107] (28#) Angala-innar-tu-t utaqqi-laa-qqaar-pu-t wander-just-iv\cn-pl wait-a.bit-first-IND.IV-3p hr. Mr.
Nilsson Nilsson
[L#: 95]
(28$) The tramps first waited a bit: they waited for Mr. N. to fall asleep. Similarly, in (29#) the same mental state is referred to by the prospective stative -ssangatip and the canonical stative -suri. These two v\tv suffixes simply form a series, without the aid of the elaborating mood. (29)
[Ron wants to see Hagrid’s dragon egg hatch; Hermione won’t skip Herbology.] ‘Hermione, how many times in our lives are we
going to see a dragon hatching?’ (29#)
inuuni-tsin-ni life-1p.pl-LOC uumasurujum-mik dragon-sg.MOD
qassi-riar-lu-ta what.number(s).n-do.n.times-ELA>-1p tukir-tu-mik hatch-iv\cn-sg.MOD
taku-ssangatis-suri-vi-sigut? see-expect.?.to-believe.?.to-QUE-2s.1p (29$)
[R: 253]
[R#: 289]
Hermione, what do you believe to be our prospects? For what n,
do you expect us to see a dragon hatching n times in our lives? Anaphoric chains may also consist entirely of prospective statives. We have already seen several examples: ilimagi- ‘expect’. . . -ssa ‘expect’ in (14#); -nia-ssa ‘be.intended-be.desired’ in (19#); -niar ‘intend’. . . -ssaq ‘desired’ in (21#); and -niar ‘intend’. . . -niaq ‘intending’ in (22#). In summary, verb doubling confirms that prospective statives are indeed verbal predicates. In addition, since they can form anaphoric chains with canonical state-predicates, they too must belong to this aspectual class.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
sini-lir-sii-llu-gu. sleep-begin-wait.for.?.to-ELA>-3s?
Maria Bittner 361
5.3 Prospective statives evoke current states
(30)
Later when we went back to the hotel, she asked me how I {was going
to spend, had spent} all the money I had won. {[D: 101], MB} (30#) apir-a-anga ask-IND.TV-3s.1s aningaasar-pa-ssui-t money-sum-big-pl su-mut what-sg.DAT {atu-ssamaar-, {use-plan-,
aki-ssar-sia-kka payment-expected-received-1s.pl
atur-sima}-nir-i-kka. use-prf}-wonder-ELA?.TV-1s.3p {[D#: 116], MB}
(30$) she asked me a question: how {did I plan to use, had I used} the large sum of money I had received to spend, she wondered. Relative to the mental state of -nir, the base ‘use-plan-’ or ‘use-prf-’ is temporally de se in the sense of Lewis (1979): the mother wonders about the realization of a concurrent state concept (plan or result). Likewise for other report suffixes, e.g. -suri ‘believe’ in (31#) and -nirar ‘say’ in (32#). Instead of ‘current at the topic time’, we get ‘current at the report time’. But the temporal parallel between prospective statives and canonical state-predicates persists. As illustrated in (30#)–(32#), the mental states of prospective statives (-ssamaar ‘plan’, -ssa ‘expect’, and -gunar ‘be likely’) are located in time exactly like the states of canonical statepredicates (-sima ‘prf ’, -siur ‘experience’, and ilumuur- ‘be right’).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
If prospective statives are state-predicates, then temporal anaphora should treat them like canonical members of this class. And indeed it does—down to the most intricate details—as I now proceed to demonstrate. We have already seen one example of the predicted parallel—to wit, (8#) and (11#). These examples, as well as (30#) below, illustrate the basic temporal pattern: a (mental) state evoked by a (prospective) stative is understood to be current at the topic time—i.e. now, or the contextually relevant then. Thus in (30#), the mother’s (mental) state evoked by -nir ‘wonder’ holds at the topic time which includes the event of her inquiry.
362 Future Discourse in a Tenseless Language (31)
The fire engine came clanging down the street, and the little children . . . cried from fright because they were sure their own houses {would
catch fire, were in danger} too. (31#)
{[L: 135], MB}
. . . aamma namminniq illu-rtik . . . also self.pl.ERG house-3p>.sg {ikualla-ssa-, navianartur-siu-}-sura-lu-gu. {catch.fire-expect-, danger-experience-}-believe-ELA>-3s? {[L#: 115], MB}
(32) [In a Quidditch dispute between Ron and Dean] Hagrid {declared Dean likely to win, was on Dean’s side}. {MB, [R: 204]} (32#) Hagridi-p Hagrid-sg.ERG
Dean Dean
{ajugaa-gunar-, ilumuur-}-nirar-pa-a. {win-be.likely-, be.right-}-say-IND.TV-3s.3s
{MB, [R#: 233]}
(32$) Hagrid declared Dean to {be likely to win, be right}.5 Temporal de se is most clearly shown—by temporal co-variation—in habitual reports such as (33#). The temporal parallel between prospective and canonical statives (here, -rusug ‘feel like’ and qasu‘be tired’) persists. (33)
Everyone wanted to take us into their homes, but my mother {refused,
kept claiming to be tired}.
{[D: 9], MB}
(33#) . . . anaana-ga¼li quja-innar-tar-pu-q . . . mother-1s.sg¼but thank-just-habit-IND.IV-3s {uagutsin-nu¼ka-rusun-niru-, qasu-}-nirar-lu-ni. {our.home-DAT¼go-feel.like-more-, be.tired-}-say-ELA>-3s> {[D#: 17], MB} (33$) . . . but my mother kept just saying (no) thank you, claiming to {feel more like going home, be tired}. 5
To capture temporal de se I use non-finite reports in my back-to-English translations, even if this means misrepresenting -nirar, which is neutral like ‘say’, as emotionally charged ‘declare ? to . . .’, ‘report ? to . . .’, or ‘claim ? to . . .’.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(31$) . . . believing their own house, too, to {be about to catch fire, be in danger}.
Maria Bittner 363
Examples (30#)–(33#) show that the exact sense of current state varies. If the next suffix is a mood inflection, then it means current at the topic time; if it is a report suffix (e.g. -nir ‘wonder’, -suri ‘believe’, or -nirar ‘say’), then it means current at the report time. Further variation is illustrated in (34#)–(37#). But crucially, whatever the exact sense of current, temporal anaphora treats prospective and canonical statives on a par. The parallel continues to hold, e.g. for nominalized dependents of habitual verbs (prospective stative in (34#), canonical stative in (35#)). (34)
John Evans was the only doctor between Womengo and Nome. He traveled around, making regular calls at fishing villages along the Sound. Dr. Evans
(34#)
Nakursa-p doctor-sg.ERG
[D: 14]
taassuma inup-pa-ssui-t that.ERG person(s)-group-big-pl
tuqu-ni-ssa-ralu-an-nit die-v\n-expected-unrealized-3p?.sg-ABL annaat-tar-sima-va-i. rescue-habit-prf-IND.TV-3s.3p
[D#: 22]6
(34$)
That doctor had been saving a lot of people from an otherwise expected death.
(35)
As much money and life as you could want! The two things most human beings would choose above all—the trouble is, humans do have
a knack of choosing precisely those things which are worst for them. [R: 320] (35#)
inui-t ajurnar-tur-siut-aat person-pl.ERG be.difficult-iv\cn-experienced-3p?.sg una-u-vu-q this-be-IND.IV-3s
namminir-min-nut self-3p>.sg-DAT
qinir-niar-tar-a-mikku. choose-try-habit-FCT>-3p>.3s (35$) 6
ajur-nirpaa-q be.bad-what’s.most-sg [R#: 365]
. . . the difficulty of the human experience is that [people] keep trying to choose what is worst for their own selves.
I have corrected [D#: 22], which has an ungrammatical nominative: nakursaq taanna.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
had saved many lives.
364 Future Discourse in a Tenseless Language The temporal parallel also continues to persist for habitual states systematically following in the wake of habitual events (two prospective statives in (36#), two canonical statives in (37#)). (36) The trail wound through steep hills, straight up and straight down. Going up, I had to get off and push hard on the sled to help the dogs. Going down, I had to press hard on the brake and the rubber mat. [D: 64] (36#) . . . am-mul¼li . . . down-DAT¼but
aallar-aanga-tta nakkakattar-uma-na-nga set.off-HAB>-1p roll.down-want-NON>-1s
(36$) . . . but whenever we set off downhill, I was forced to hold the sled back a fair bit, not wanting to roll down the hill. (37) Wood was working the team harder than ever. . .The Weasleys complained. . ., but Harry was on Wood’s side.. . .[he] found that he had fewer nightmares when he was tired out after training. [R: 234] (37#) . . . paasi-sima-va-a . . . find.out-prf-IND.TV-3s.3s qasu-ni-ssa-mi be.tired-v\n-expected-3s>.sg.ERG tamaviaar-lu-ni use.all.strength-ELA>-3s> sinnattu-pilun-na-ni dream-badly-NON>-3s>
tunga-a-nut direction-3s?.sg-DAT
sungiusar-aanga-mi train-HAB>-3s> sini-lluar-tar-lu-ni. sleep-well-habit-ELA>-3s>
[R#: 267]
(37$) . . . he had discovered that whenever he trained with all his might until the state of exhaustion, he slept well, without bad dreams. Being both variable and predictable, temporal anaphora in Kalaallisut (30#)–(37#) is reminiscent of non-finite constructions in English (Stump 1985). For instance, non-finite complements of English report verbs predictably concern either the de se now of the experiencer or speaker (believe her to be fair, be tired of losing, claim to know this) or his de se future (expect her to be fair, dread losing, predict her to know this). Moreover, in Kalaallisut as in English, more complex temporal relations can be conveyed, with equal precision, by
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
qamuti-t nuqi-ngaatsiar-tariaqar-tar-pa-kka sled-pl hold.back-a.fair.bit-be.necessary-habit-IND.TV-1s.3p [D#: 77]
Maria Bittner 365
composing suitable constructions. This is illustrated by the more complex examples in (38#)–(41#). The discourse in (38#) mentions no fewer than three reports, which could be weeks or even months apart. It starts with a report (‘said’) by Erik the Red to another Greenland settler, Herjulf. Erik reports a prior report (‘spoken of ’. . . -nirar ‘say’) received from Sverre’s folk. And Sverre’s folk, in turn, report a yet another prior report (uqar- ‘say’), by Erik’s son Leif about his then current intentions. Temporally, (38#) is interpreted like (38).
(38#)Uqar-sima-nirar-pa-at say-prf-say-IND.TV-3p.3s nuna-nut kian-niru-su-nut land-pl.DAT be.warm-more-iv\cn-pl.DAT niuvvag-iar-qaar-lu-ni trade-go.to-first-ELA>-3s> angun-ni ukii-vvig-iartur-niar-a-a. father-3s>.sg spend.winter-at-go.to-intend-ELA?.TV-3s?.3s [G: 10] Another example of complex temporal relations captured by composition is the demand for clarification in (39#). Whitey has just been talking about his silly dog and his beloved blind colt in one breath, referring to both as he, and Uncle Torwal has trouble with nominal anaphora. The temporal anaphora in Kalaallisut (39#) is similar to English (39$). (39)
‘What in the world is eatin’ on you?’ Torwal wanted to know. ‘You claim his brains is addled, and he sure is acting it, then you say you
allus knowed he was too smart to have anything happen to him. [P: V] (39#)
Uqa-riar-lu-tit say-have.just-ELA>-2s . . . uqar-pu-tit . . . say-IND.IV-2s
paatsivi-irus-sima-nirar-lu-gu sense(s)-lose-prf-say-ELA>-3s?
nalunngin-nirar-lu-gu know-say-ELA>-3s?
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(38) In the fall Erik the Red, while visiting Herjulf ’s folk for several days . . . , said that his son Leif was in Wineland. Sverre’s folk had spoken of him on their last trading visit in Erik’s fiord. They reported him to have said that he intended, after first going trading to warmer lands, to go to his father to spend the winter. [MB]
366 Future Discourse in a Tenseless Language ajuquti-ssar-si-ssa-llu-ni harm-expected-find-be.expected-ELA>-3s> silatu-allaar-tu-q. be.smart-too-ELA?.IV-3s?
[P#: 37]
(39$) Having just said something declaring him to have lost his senses —and he’s sure acting like it—you say something claiming to know that he’s too smart to be expected to come to harm.
(40) After Whitey had thrown the hay out, he hung around [the pasture] as long as he dared, admiring the colt [who had strayed in] and trying to think of a way he could keep him without Uncle Torwal finding out about it. But after
a while he had to leave for fear Torwal would wonder what had happened to him and come out to look. [P: VII] (40#) Sivitsu-nngit-su-r¼li be.long-not-iv\cn-sg¼but Whitey Whitey
angirlar-tariaqa-lir-pu-q go.home-need-begin-IND.IV-3s
annilaangagi-ga-mi-uk ajuqu-sir-sima-sura-lu-ni be.afraid.of-FCT>-3s>-3s? harm-come.to-prf-believe-ELA>-3s> Thorvald’i-mit Thorvald-ABL
taku-niar-niqar-ni-ssa-ni. see-try-pass-v\n-expected-3s>.sg]
[P#: 44]
(40$)But before long Whitey> began to feel the need to go home, being afraid of [the prospect of Torwal trying to find him>, believing him> to have come to harm]. (41) DO NOT OPEN THE PARCEL AT THE TABLE. It contains your new Nimbus Two Thousand, but I don’t want everybody knowing you’ve got a broomstick or they’ll all want one. [R: 179] (41#)
. . . ila-vil¼li . . . mate-2s.pl.ERG¼but
paasi-ssa-nngi-la-at find.out-be.desired-not-IRR-3p.3s
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Finally, (40#) and (41#) illustrate how this tenseless system also has the capacity to express complex temporal relations in a feared, desired, or otherwise anticipated future. Once again, temporal anaphora in Kalaallisut is similar to the corresponding back-to-English translations (40$) and (41$).
Maria Bittner 367
saniguti-taar-tu-tit broomstick-get.new-ELA.IV?-2s
— —
paasi-gu-nikku find.out-HYP>-3p>.3s?
pi-uma-li-qina-mm-ata. do.so-want-begin-be.liable-FCT?-3p?
[R#: 203]
(41$) . . . but [I] don’t want your schoolmates to find out you got a new broomstick—if they do find out about it, they’re liable to start wanting one too.
5.4 Prospective statives evoke attitude states Having shown that prospective statives are state-predicates, I now turn to distinguish them as a natural subclass. As a first step, in this section I show that prospective statives evoke attitude states. More precisely, they express three-place relations between a mental attitude state—belief, expectation, fear, considering something likely, unlikely, etc—the experiencer of this mental state, and a modal concept for the object of the attitude. In Kalaallisut the presence of a modal object can be detected by several diagnostic tests. I focus here on two reliable tests, provided by the counterfactual suffix -galuar and modal discourse anaphora. The counterfactual suffix -galuar indicates a currently unrealized possibility. This possibility may be the modal object of the host predicate, e.g. expectation in (34#) (‘rescue from an otherwise expected death’), belief (42#), or what might be necessary if it did not conflict with a desire (43#). (42)
Pippi as a Thing-Finder : ‘Oh, dear! I was sure I saw a lump of gold.’
[L: 28]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The evidence presented in this section shows that Kalaallisut has a genuinely tenseless system, which, nevertheless, is a true match for the English tense system. Although the means are very different, both systems have the capacity to convey even complex temporal information with great precision by discourses of comparable length. The Kalaallisut system systematically treats prospective statives, including the putative ‘future tenses’ (-ssa, -niar, -jumaar), like canonical state-predicates (qasu- ‘be tired’, -sima ‘prf ’, etc). All varieties of statepredicates in Kalaallisut refer to a state that is understood to be current in a variable, but predictable, sense.
368 Future Discourse in a Tenseless Language (42#)
Kuulti-minir-mik gold-lump-sg.MOD taku-suri-i-vik-kaluar-a-ma see-believe-already-really-unrealized-FCT>-1s
[L#: 26]
(42$) I really already believed myself (alas, wrongly) to have seen a lump of gold. (43)
[Snowed in at a check point, we built an igloo.] Oteg had brought a lamp with him. . .it glowed and made blue shadows dance against the walls. I should have felt snug but I didn’t. I wanted to be on the trail.
[D: 46]
taama¼li misigi-sima-nngi-la-nga thus¼but feel-prf-not-IRR-1s — ingirla-qqik-kusuk-kaluar-nir-mik — travel-again-wish-unrealized-v\n-sg.MOD (43$)
[D#: 58]
I should#ve enjoyed myself, but I didn’t feel that way—(just) a frustrated wish to be on the trail again.
But if the host predicate has no modal object, then the counterfactual -galuar instead targets the predicate itself. For example, in (44#) -galuar is hosted by a predicate, ‘forget’, whose object is a fact—a race currently in progress. The se of (44#) is a participant who three days ago got caught up in the break-up of the ice on the trail. Her father has just come to rescue her. What -galuar implies here is that the race is now no longer fogotten. (44)
‘We are losing time,’ my father said. ‘You’re still in the race.’
I had given up all thoughts of the race. (44#)
[D: 91]
Sukka-niun-niq puigur-aluar-pa-ra. go.fast-compete-v\n forget-unrealized-IND.TV-1s.3s [D#: 106]
(44$) I had forgotten (but was now reminded) about the ongoing race. Similarly, in (45#) -galuar induces an implicature that a hunting habit is incompatible with the speaker’s idea of a ‘real’ woman. (45) She is a woman, but she goes hunting.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(43#) Nuannisaar-tariaqar-aluar-pu-nga, enjoy.self-be.necessary-unrealized-IND.IV-1s
Maria Bittner 369
(45#) Arna-u-galuar-lu-ni piniar-tar-pu-q. woman-be-unrealized-ELA>-3s> go.hunting-habit-IND.IV-3s
(46)
My father and I harnessed up the team while my mother watched. We
would meet again in Ikuma. Then we would pack two big sleds and take the trail back to Norton Sound and Womengo. [D: 102] (46#) Ikuma-mi taku-qqi-ssamaar-pu-gut. Ikuma-sg.LOC see-again-plan-IND.IV-1p Taava¼lu puurtu-i-riar-lu-ta then¼and pack-apass-quickly-ELA>-1p Womengo-mut uti-ssamaar-pu-gut Womengo-sg.DAT return-plan-IND.IV-1p
[D#: 117]
(46$) We planned to meet again in Ikuma. And then, after quickly packing up, we planned to return to Womengo.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In general, the suffix -galuar provides a reliable diagnostic for modal arguments. If the host predicate has a modal argument, then that is what is currently unrealized, as in (42#) and (43#) (see also (34#), (49#), and (65#)). Otherwise, it is the predicate itself, as in (44#), or some implicature, as in (45#) (see also (26#)). This diagnostic test reveals modal arguments for all prospective statives—on a par with canonical attitude states, such as -suri ‘believe’ (see (42#)), uppiri- ‘believe’, or isuma-qar- ‘idea-have-’. Another reliable diagnostic is modal anaphora, exemplified in (46#)– (51#). These examples further show that the modal objects of Kalaallisut attitudes are eventuality concepts, rather than propositions. First of all, unlike propositions, eventuality concepts are directly associated with temporal concepts, without the benefit of tense. So if the modal object of the first -ssamaar ‘plan’ in (46#) is a discourse referent for the planned reunion—a modal concept of an event—then it can directly antecede taava ‘then’ in the second sentence, which elaborates the plan. Also, the planned packing can be readily located just before the planned return to Womengo—the modal object of the second instance of -ssamaar. (See also (48#), where an analogous story can be told about -riaannaa ‘be (considered) easily possible’, taava ‘then’, and -ssa ‘be expected’.)
370 Future Discourse in a Tenseless Language Secondly, the anaphor taama- ‘thus’ shows that modal objects of Kalaallisut attitudes are aspectually typed—as states or habits (¼it- ‘be’ in (47#)), state changes (-ili ‘become’ in (48#)), or actions (-iliur ‘do’ in (49#)). This makes sense for eventuality concepts, but not for propositions. (47)
‘It was papa made me leave. I am a boy and I must obey him.’ ‘I know,’ the old man said. ‘It is quite normal.’ [H: 10]
(47#)
‘Ataata-ma¼mi¼una father-1s.sg.ERG¼for¼this
Nukappiara-u-ga-ma¼lu suli still boy-be-FCT>-1s¼and tassa that’s.it
naala-innar-tariaqar-sima-va-ra.’ obey-just-be.necessary-prf-IND.TV-1s.3s
. . . ‘Taama¼it-tu-ssa-u-vu-r¼mi.’ . . . ‘thus¼be-iv\cn-expected-be-IND.IV-3s¼for
[H#: 6]
(47$) ‘It’s because papa wanted me to get off [your] boat. And being a boy still, that’s it—I have been obliged to just obey him.’ ‘I know,’ the old man said. ‘For it is expected to be so.’ (48)
[Prof. McGonagall to Hagrid sobbing loudly in a Muggle neighborhood.]
. . . get a grip on yourself, Hagrid, or we’ll be found. (48#)
. . . iqqissi-sariaqar-pu-tit, . . . calm.down-need-IND.IV-2s
[R: 23]
Hagrid. Hagrid
Taama-ili-nngik-ku-it thus-become-not-HYP>-2s takkut-tu-qa-riaannaa-vu-q show.up-iv\cn-have-be.easily.possible-IND.IV-3s taava¼lu then¼and
taku-niqa-ssa-u-gut. see-pass-be.expected-IND.IV-1p
[R#: 22]
(48$) . . . you need to calm down, Hagrid. If you don’t calm down, someone can easily show up here and then [I] expect us to be seen. (49)
[Pippi, Tommy and Annika are looking for Mr. Nilsson lost in the woods.]
Tommy suggested they all three go in different directions and
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
niu-qqu-ga-anga; get.off.boat-want.?.to-ELA?.TV-3s?.1s
Maria Bittner 371
hunt. At first Annika didn’t want to because she was a little afraid. [L: 83] (49#)
Tommy Tommy
siunnirsu-i-vu-q avissaar-lu-tik suggest-apass-IND.IV-3s [separate-ELA>-3p>
ujar-li-ssa-llu-tik. Annika seek-apass-be.desired-ELA>-3p>] Annika taama-iliu-rusu-nngik-kaluar-pu-q. thus-do-wish-not-unrealized-IND.IV-3s (49$)
annilaanga-ga-mi be.afraid-FCT>-3s> [L#: 73]
Thirdly, (49#) further reveals that modal objects can have discoursetransparent parts available for eventuality anaphora—here, by the implicitly anaphoric annilaanga- ‘be afraid (of that)’. Intuitively, Annika is afraid only of the first part of Tommy’s plan—the proposed separation—not of the proposed search. Formally, these two parts of Tommy’s plan can be individuated as two discourse referents for eventuality concepts. Tommy wants both concepts to be realized— first the proposed separation and then, during the result state, the proposed search. Annika is afraid of realizing the first part of the plan, and therefore does not wish to do that (until Tommy calls her a ’fraidy cat). Finally, consider modal anaphora by means of agreement. Modal subject agreement is exemplified in (50#), modal object agreement in (51#). Both sentences contain predicates that can be sensibly applied to eventuality concepts—e.g. being really suitable for a nice ending to a pleasant Sunday is applicable to a desired horse ride—but not to propositions. (50)
She remembered that she hadn’t been riding for a couple of days and
decided to go at once. That would be a nice ending to a pleasant Sunday. [L: 133] (50#) . . . massakkur-lluinnar¼lu . . . now-absolute¼and hiistir-uma-llir-pu-q. ride.horse-want-suddenly.begin-IND.IV-3s Sapaati-siur-nir-mut nuannir-su-mut Sunday-experience-v\n-sg.DAT be.enjoyable-iv\cn-sg.DAT naggasiu-ssa-ssaqqi-vip-pu-q. ending-desired-be.suitable.for-really-IND.IV-3s
[L#: 114]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Tommy made a suggestion: [he] wanted them to separate and search. Annika was afraid so she didn’t wish to do that (but . . .).
372 Future Discourse in a Tenseless Language (50$) . . . and she suddenly felt like going riding right away. That was really suitable for a nice ending to an enjoyable Sunday. (51)
Percy—he’s a Prefect, he’d put a stop to this.
(51#)
. . . iliusiqar-niar-ni-rsi¼lu . . . do.something-intend-v\n-2p.sg¼and uni-tsi-riaannaa-va-a. stop-cause-be.easily.possible-IND.TV-3s.3s
[R: 170]
[R#: 193]
(51$) . . . and what you intend to do, he can easily stop that.
5.5 Prospective statives evoke attitudes toward de se prospects In the foregoing sections I have shown that prospective statives evoke current attitudes toward eventuality concepts. So far, prospective statives have patterned like canonical attitude states, e.g. -suri ‘believe’. However, -suri normally evokes belief states about the experiencer’s now. In contrast, prospective statives are always about the experiencer’s future. First of all, to clarify the difference, I present some minimal contrasts in English and Kalaalisut. We begin with a sample of emotional attitudes: future-oriented (52a), in contrast to now-oriented (52b). (52) a. Ron {wanted, hoped for, was eager for} Snape to suffer. b. Ron was {glad, amused, pleased} to watch Snape suffer. In the future-oriented (52a) Ron feels (at a past topic time) a wish that Snape suffer. Snape may in fact suffer while Ron feels this way, but until Ron discovers this fact, he cannot feel glad about it. So for him, the joy of a wish come true is still a future prospect. In contrast, in the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In summary, the counterfactual -galuar reveals that prospective statives have modal objects, like canonical attitude states. This is confirmed by modal anaphora, which further reveals that this object is an eventuality concept. For unlike a proposition, it participates in temporal anaphora in modal contexts directly, without the benefit of tense; it is aspectually typed, as a state/habit, state change, or action; it can have discourse transparent part-whole structure; and it can serve as an argument of a predicate that can be sensibly applied to eventuality concepts, but not to propositions.
Maria Bittner 373
(53) Harry muttered angrily as Snape limped away. ‘Wonder what’s wrong with his leg?’ ‘Dunno, but I hope it’s really hurting him,’ said Ron bitterly. [R: 198] (53#) . . . niriup-pu-nga¼li annirna-ssa-qi-su-q. . . . hope-IND.IV-1s¼but be.painful-be.desired-very-ELA?.IV-3s? [R#: 225] (53$) . . . but I have a hope: [I] want it to be very painful. (54) . . . c-can’t t-tell you how p-pleased I am to meet you. [R: 80] (54#) Nuannaa-qa-u-nga be.pleased-very-IND.IV-1s
naapik-ka-kkit. meet-FCT> -1s.2s
[R#: 87]
(54$) I am very pleased I have met you. Turning now to intellectual attitudes, these too can be either futureoriented (55a) or now-oriented (55b). Again, the key is the perspective of the experiencer. For the se of (55a), the expected verification (by DNA tests?) is still a future prospect, even if the expected state in fact holds now. (55)
a. I {expect, count on} this husky to be a son of that wolf. b. I {believe, know} this husky to be a son of that wolf.
For intellectual attitudes, too, Kalaallisut marks de se prospects by prospective statives (in (56#), -ssa ‘be expected’), whereas the de se now is unmarked (e.g., nalunngi- ‘know’. . . -tu ‘ELA?.IV’ in (39#)). (56)
[Observing the wolf closely] I saw that his body was exactly the same pure white as Black Star’s. Their faces were the same too . . . Suddenly it struck
me that Black Star was the son of this wolf.
[D: 72]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
now-oriented (52b) Ron watches Snape suffer and, concurrently, feels glad about it. In Kalaallisut this contrast involves the presence versus absence of a prospective stative. In addition, modal objects of future-oriented attitudes are explicitly marked as prospects—usually by -ssa ‘(be) expect(ed), (be) desire(d)’, -ssaq ‘expected, desired’, or -niar ‘(be) intend(ed), be about to’. For instance, in English (53) Ron’s wish that Snape’s currently unclear state be painful is in the present tense, is. But in the Kalaallisut translation (53#) this is rendered by a prospective stative, -ssa. In contrast, the complement of pleased to in English (54) is rendered in Kalaallisut (54#) as a presupposed fact (FCT), without any prospective stative.
374 Future Discourse in a Tenseless Language (56#)
Tassa that’s.it
isuma-liu-lir-pu-nga idea-make-begin-IND.IV-1s
Milalinnguaq Black.Star
amaqqu-p [wolf-sg.ERG
taassuma this.ERG]
kinguaari-ssa-ga-a. be.sire.of-be.expected-ELA?.TV-3s?.3s
[D#: 86]
(56$) That’s it! I got an idea: this wolf is Black Star’s sire, [I] reckoned.
(57) a. Snape is not likely to be a fair referee. b. Snape does not seem to be a fair referee. In Kalaallisut we again get the by now familiar contrast: futureoriented attitudes have their modal objects marked as de se prospects, as in (58#). In contrast, now-oriented attitudes do not, as (59#) attests. (58)
‘Snape’s refereeing? . . . He’s not going to be fair if we might
overtake Slytherin.’ (58#) Slytherini-mut Slytherin-DAT
ajugaa-lir-u-tta win-begin-HYP>-1p
[R: 234] naapirtuilluar-tu-mik be.fair-iv\cn-sg.MOD
dummiri-u-ni-ssa-a ilimana-nngi-la-q. referee-be-v\n-expected-3s?.sg be.likely-not-IRR-3s[R#: 267] (58$) The prospect of his being a fair referee is not likely, if we begin to overtake Slytherin.7 (59) ‘Harry—yer a wizard.’. . . ‘I don’t think I can be a wizard.’ (59#) Uppiri-nngi-la-ra angakku-u-ni-ra. believe-not-IRR-1s.3s wizard-be-v\n-1s.sg 7
[R: 67] [R#: 72]
The prospect of Snape being a referee is expected. What is unlikely is his being fair.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Impersonal attitudes can likewise be future- or now-oriented. For English, this contrast is illustrated in (57). The future-oriented be likely in (57a) evokes an attitude state whose experiencer considers the prospect of Snape being a fair referee unlikely. In contrast, the noworiented seem in (57b) evokes a state whose experiencer considers Snape’s concurrent refereeing (state or habit) to have the appearance of unfairness.
Maria Bittner 375
(59$)
I don’t believe it—my being a wizard.
Predicates like forget also exhibit this contrast. One can forget a fact—e.g. a current race in English (60b) and Kalaallisut (44#). But one can also forget to realize an expectation—e.g. the expected sign-in in English (60a) and Kalaallisut (61#). (60)
a. The race participant forgot {to sign-in, the expected sign-in}. b. While fighting for her life, the participant forgot the race. [Arriving at the check point in my home village] I was too tired to enjoy the excitement. I almost forgot to check in. [D: 76]
(61#)
Allaammi even
unammi-qata-u-su-tut compete-fellow-be-iv\cn-sg.EQU
alla-tsin-ni-ssa-ra write-apass-v\n-expected-1s.sg piugu-ngaja-vip-pa-ra. forget-almost-really-IND.TV-1s.3s (61$)
[D#: 90]
Even my expected sign-in as a race participant—I really almost forgot that.
The prospective sapir- ‘be unable, not dare’ in (62#) and (63#) at first blush looks like a counterexample to the generalization that nominalized modal arguments of prospective statives are explicitly marked as de se prospects (e.g. by -ssaq in (61#) and (58#), or by -niar in (51#)). (62)
And the trunk of the tree was smooth and had no branches for climbing on.
Even Pippi wouldn’t be able to climb it. qallur-niq climb.up-v\n
[L: 137]
sapir-pu-q. be.unable-IND.IV-3s [L#: 117]
(62#)
Pippi¼luunniit Pippi¼even
(62$)
Even Pippi was unable to climb up.
(63)
I dare not say anything to him.
(63#)
Uqar-vigi-niq sapir-pa-ra. say-to-v\n not.dare-IND.TV-1s.3s
[SL: 246]
But in fact, sapir- is not a counterexample, because the nominalized verb is not a subject or object, but part of a complex predicate. Unlike a subject
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(61)
376 Future Discourse in a Tenseless Language or object, it determines the transitivity inflection on sapir- (intransitive in (62#), transitive in (63#)). Nothing can intervene between sapir- and the nominalized verb, and the order is fixed. All of this shows that what we have in (62#) and (63#) are complex predicates, not predicate-argument structures. Therefore, there is no syntactic argument to be marked as a de se prospect, and sapir- is enough to mark prospectivity. Thus, in Kalaallisut as well as English future-oriented attitude states concern de se prospects—that is, future prospects of the experiencer viewed from his own perspective. It is therefore not surprising that only future-oriented attitudes give rise to purposive readings, e.g. in (64):
In (64a) the purposive reading arises because the matrix action (flew up) is understood to be concurrent with a mental state of the agent having a suitable attitude—intention, wish, hope, or the like—toward a de se prospect—by implicature, a purpose—of saving Harry. (64b) is similar, except that the attitude is impersonal. Still, the mental state of considering a prospect possible or likely must have an experiencer, so the prospect could be his purpose. In contrast, (64c), with noworiented attitude/speech reports, has no purposive reading for lack of any element to evoke a de se prospect. The same story can be told about the purposive use of prospective statives in Kalaallisut. Parallel to (64a), we have (22#) (-niar-ELA> ‘intending to’), and (36#) (-uma-NON> ‘not wanting to’). And parallel to the impersonal (64b), we have (19#) (-nia-ssa-IND.TV ‘[I] intend and desire that. . .’) and the truly impersonal (65#). (65) The mare . . . knew . . . that those extra long legs of [her newborn colt] were specially made that way so that by the time he was a day or two old he would be able to travel as fast and as far as the grown horses in case of danger. [P: I] (65#) . . . navianartu-mik . . . danger-sg.MOD naammattuu-i-ssa-galuar-u-ni encounter-apass-be.expected-unrealized-HYP>-3s>
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The Weasleys flew up, {intending, wishing, hoping, expecting, eager, determined, . . .} to save Harry. b. The Weasleys flew up, making it {possible, likely for Woods} to save Harry. c. The Weasleys flew up, {seeming, claiming} to save Harry.
(64) a.
Maria Bittner 377
histi-tut horse-sg.EQU
inir-sima-su-tut¼li grow.up-prf-iv\cn-sg.EQU¼but
pangalis-sinnaa-li-riir-sima-qqu-llu-gu run-be.able-begin-already-prf-enable.?.to-ELA>-3s? [P#: 9] (65$) . . . enabling him by the age of one or two days old to have already become able to run like an adult horse if he had what would otherwise be expected to be a dangerous encounter.
(66)
a. The ground was too wet for the colt to want to lie down. b. The ground was too wet for the colt to be pleased to lie down.
(67)
a. He’s too smart to {be expected, be likely} to die. b. #He’s too smart to {be known, seem} to die.
Analogous uses of prospective statives in Kalaallisut are illustrated in (68#) and (69#). Again, one story—whatever the details might be— will hopefully generalize across English (66a)–(67a) and Kalaallisut (68#)–(69#). (68)
. . . for [the ground] was much too wet to lie down.
(68#)
. . . nuna¼mi . . . ground¼for
[P: V]
nallar-vigi-ssa-llu-gu lie.down-on-desire-ELA>-3s?
masap-pallaa-qi-mm-at. be.wet-too-very-FCT?-3s?
[P#: 33]
(68#)
. . . for the ground was much too wet [for him] to want to lie down.
(69)
[Uncle Torwal says the winter is hard on wild horses and he thinks that the blind colt, Whitey’s favorite, is already dead. Whitey replies:] ‘I haven’t seen him for quite a spell, but I don’t think he’s dead; he’s too smart.’ [P: V]
(69#)
. . . tuqu-ssa-llu-ni¼mi . . . die-be.expected-ELA>-3s>¼for
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Another distinctive use of future-oriented attitudes are scalebased predictions. For English, this use is illustrated in (66a) and (67a). The (b) examples, with now-oriented attitudes, are included for contrast.
378 Future Discourse in a Tenseless Language silatu-allaa-qa-u-q. be.smart-too-very-IND.IV-3s
[P#: 36]
(69$) . . . for he’s much too smart to be expected to die. In summary, reports of attitude states or speech acts can concern either the de se future prospects of the experiencer/speaker, or his de se now. Only future-oriented reports support purposive construal and scale-based predictions. In Kalaallisut, diagnostic tests consistently show that prospective statives report current attitude states toward de se prospects.
The evidence presented here conclusively demonstrates that Shaer’s (2003) conjecture—that Kalaallisut is tenseless—is in fact correct. This language does not have tense, in any sense of the term tense found in descriptive work (e.g., Chung and Timberlake 1985, Comrie 1985, Klein 1994). This conclusion holds even under the most liberal criteria for what counts as a tense marker. For example, by the criteria of Chung & Timberlake (1985) and Klein (1994), the English tense markers arguably include not only the past and present tense inflections, but also the future auxiliaries, will/would and is/was going to. That is, tense markers are allowed to be fused, e.g. with mood or aspect, and to be realized by various morphosyntactic categories, including complex auxiliaries. But even with all of these allowances, there is still nothing in Kalaallisut that qualifies as a tense marker—in spite of the grammarians’ reports of ‘future tenses’, dating back to Kleinschmidt (1851) and reinforced by Fortescue (1984). According to these reports, Kalaallisut has three ‘future tense’ suffixes corresponding to Germanic future auxiliaries. Out of context, these three ‘future tenses’ are indeed likely to be used by native consultants to translate the English future auxiliaries will/ would and is/was going to (or their Danish equivalents). But a very different picture emerges from text studies examining how the same auxiliaries are rendered in published translations—that is, under optimal conditions. For professional translators have excellent command of English (or Danish). They also have full knowledge of the context and are doing their best to render the meaning.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
6 CONCLUSIONS
Maria Bittner 379
To translate the future uses of will/would and is/was going to under these conditions, Kalaallisut translators go far beyond the three putative ‘future tenses’—they use nearly thirty morphemes. Morphologically, the attested translations are diverse. But semantically, they form a natural class, with three subclasses conforming to the following Prospectivity Thesis: PROSPECTIVITY THESIS Kalallisut translations of future auxiliaries comprise three related classes:
In Bittner (2005) some of the claims of the Prospectivity Thesis are explicated by combining the logic of change of Muskens (1995) with a suitable ontology and the formal theory of centering of Bittner (2001). But here, my goal was not to argue for any particular theoretical framework, but rather to contribute to the data base for any framework aspiring to universality. Therefore, the goal was strictly descriptive—to provide a comprehensive, in-depth, theory-neutral description8 of a hitherto unknown tenseless system for conveying temporal information. The data reported here show how Kalallisut does that without the benefit of tense, but still with the same precision as the tense system of English. To give some logical structure to this mass of data—and to jumpstart the debate—I have presented the raw data of Kalaallisut together with my current understanding of this linguistic system. But whereas my current understanding may be—and probably is—flawed, the raw data are facts of life. From now on they are part of the data base that any theory of temporal anaphora and quantification must account for, if it is to have any hope of universality. Hopefully, the effort of rethinking our theories to accommodate tenseless mood languages like Kalaallisut will be rewarded, not only by more general and more representative theories, but also by new insights into tensed languages such as English. 8
To be sure, I have used terms that have been theorized about—e.g., event, state, topic time, discourse referent, de se attitude, etc—but I have only used them in their dictionary sense, or in the pretheoretical sense of Klein (1994), Karttunen (1976), or Lewis (1979).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A. prospective statives evoking (current) attitude states to de se prospects, B. prospective inchoatives evoking (realized) starts of expected processes, C.prospective matrix moods marking the speech act as a request or wish.
380 Future Discourse in a Tenseless Language APPENDIX A: TEXTS AND DICTIONARIES Berthelsen, C. et al. 1990. Oqaatsit: Kalaallisuumiit Qallunaatuumut. (Words: From Kalaallisut to Danish.) Atuakkiorfik, Nuuk. [G] Gedionsen, I. 1980. Qallunaatsiaq Herjulf. Kalaallit nunaanni naqiterisitsisarfik, Nuuk. [H] Hemingway, E. 1952. The Old Man and the Sea. Scribner, New York. [H#] Hemingway, E. 1991. Angutitoqaq imarlu. Translated from the Danish translation by Thomas Efraimsen. Atuakkiorfik, Nuuk. [LH] Lennert Olsen, L. and Herling, B. 1988. Grønlandsk tilhængsliste. (List of Greenlandic suffixes.) Pilirsuiffik. [L] Lindgren, A. 1997. Pippi Longstocking. Translated by Florence Lamborn. Puffin, New York. [L#] Lindgren, A. 2000. Pippi Langstrømpe—ikinngutaalu. Translated by R.Grønvold Benjaminsen. Atuakkiorfik, Nuuk. [D] O’Dell, S. 1988. Black Star, Bright Dawn. Fawcett Books, New York. [D#] O’Dell. S. 1994. Milalinnguaq Sikkersorlu. Translated by Svend Møller. Atuakkiorfik, Nuuk. [P] Rounds, G. 1941. The Blind Colt. Holiday House, New York. [P#] Rounds, G. 1988. Hesti piaraq tappiitsoq. Translated by Mariane Petersen. Atuakkiorfik, Nuuk. [R] Rowling, J. K. 1997. Harry Potter and the Philosopher’s Stone. Bloomsbury, London. [R#] Rowling, J. K. 2002. Harry Potter ujaraallu inuunartoq. Translated by Stephen Hammeken. Atuakkiorfik, Nuuk. [SL] Schultz-Lorentzen. 1974. Den grønlandske ordbog: Grønlandsk-Dansk. Sydgrønlands Bogtrykkeri, Nuuk. [B]
Examples (a) justify the glosses; examples (b) illustrate the translations of English future auxiliaries. A1. ilimagi- (tv) ‘expect’ (a) Ilimag-a-a. (expect-IND.TV-3s.3s) Han forventer det. (‘He expects it.’)
[SL: 68]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
APPENDIX B: KALAALLISUT TRANSLATIONS OF FUTURE AUXILIARIES
Maria Bittner 381
(b)
See (14#).
A2. ilimanar- (iv) ‘be likely’ (a) Ilimanar-pu-q. (be.likely-IND.IV-3s) Det er sandsynligt. (‘It’s probable.’) (b) See (58#).
[B: 103]
A3. -gunar (v\v) ‘be likely to’ (a) See (18#). (b) See (13#).
A5. -jumaar (v\v) ‘(be) hope(d), (be) dread(ed)’ (a) See (23#). (b) He’ll turn up. (b#) Takkuti-qqik-kumaar-pu-q. show.up-again-be.hoped/dreaded-IND.IV-3s (b$) There is a hope/risk of his showing up again.
[R: 116] [R#: 130]
A6. -navianngi (v\v) ‘be very unlikely to’ (a) Pi-uma-navianngi-la-q. [B: 217] (do.so-want-be.very.unlikely-IRR-3s) Han vil sikkert ikke. (‘He certainly will not want to do this.’) (b) At the checkpoint, the race judge advised me not to leave. ‘You’ll not get far,’ he said. [D: 78] (b#) Ungasilli-ngaar-navianngi-la-tit. get.far-very-be.very.unlikely-IRR-2s [D#: 93] (b$) You’re very unlikely to get very far. A7. -niar (v\v) ‘(be) intend(ed), be about to, (process use: try)’ (a) See (18#), (19#), (22#), (38#), (51#). Process use: (35#), (40#), A9(b#), C1(b#). (b) See (21#).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A4. -juma (v\v) ‘want to, be willing to’ (a) See (36#), (41#), (50#), A6(a). (b) ‘. . . won’t you be a good girl and put the ladder back so that we can get down?’ ‘Of course I will,’ said Pippi. [L: 44] (b#) Aap, suurunami taama-iliur-uma-vu-nga. [L#: 41] yes of.course thus-do-be.willing-IND.IV-1s (b$) Yes, of course I am willing to do that.
382 Future Discourse in a Tenseless Language A8. -qina (v\v) ‘be liable to’ (a) Nakka-qina-vu-tit. [SL: 335] (fall.down-be.liable-IND.IV-2s) Pas pa˚, ellers falder du ned. (‘Watch out, or you’ll fall down!’) (b) See (41#). [B: 293]
[H: 13]
[H#: 8]
A10. -qqu (v\tv) ‘want __ ? to, enable __ ? to, (event use: order __ ? to)’ (a) Aviisi-p tunniun-nir-a-nut akili-qqu-vu-q. [R#: 77] paper-sg.ERG give-v\n-3s?.sg-DAT pay-want.?.to-IND.IV-3s [R: 72] He [the owl] wants payin’ fer deliverin’ the paper. State or event use: (25#), (47#) (b) See (65#). A11. -riaannaa (v\v) ‘be well able to, be easily possible’ (a) Asiru-riaannaa-vu-q. (break-be.easily.possible-IND.IV-3s) Det kan let ga˚ i stykker. (‘It can easily break.’) (b) See (48#), (51#).
[LH: 65]
A12. -rusuk (v\v) ‘wish to, feel like’ (a) See (33#), (43#), (49#). (b) I still haven’t got yeh a birthday present . . . I’ll get yer animal. [R: 92] (b#) Uumasu-mik tuni-rusup-pa-kkit. [R#: 102] animal-sg.MOD give-feel.like-IND.TV-1s.2s (b$) I feel like giving you an animal. A13. sapir- (v) ‘be unable, not dare’ (a) See (63#). (b#) See (62#).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A9. -qqajaa (v\v) ‘be just about to, be ready to’ (a) Pitaruti-qqajaa-vu-q. (get.through-be.just.about.to/ready.to-IND.IV-3s) Det kan ventes hvert øjeblik. (‘It can be expected any moment.’) (b) ‘You didn’t steal [these baits for me]?’ ‘I would,’ the boy said. ‘But I bought these.’ (b#) Pillir-vigi-sinnaa-nngit-suu-gu-kkit offer-to-be.able-not-nonfact-HYP>-1s.2s tillin-nia-qqaja-qa-u-nga. steal-try-be.ready-very-IND.IV-1s (b$) Were I unable to offer you this, I am quite ready to try to steal.
Maria Bittner 383
A14. sapirnar- (iv) ‘be impossible, be hard to do’ (a) Sapirnar-pu-q. [SL: 246] (be.hard/impossible-IND.IV-3s) Det is vanskeligt/uoverkommeligt. (‘It’s hard/not feasible.#) (b) That I’ll never believe. [L: 54] (b#) Uppiri-niq sapirna-ngajap-pu-q. believe-v\n be.impossible-almost-IND.IV-3s [L#: 51] (b$) That is almost impossible to believe.
A16. -sinnaa (v\v) ‘be able to, be possible for __ > to’ (a) See (65#), A9(b#) (b) ‘Say you’re back [from your injury] in three weeks . . . Your team needs to run fifty miles a day to get in shape . . .’
‘Bright Dawn will train the dogs for me.’ [D: 28] Sikkirsu-p qimuttu-t B.D.-sg.ERG sled.dog-pl sungiusar-tar-sinnaa-va-i. [D#: 39] train-habit-be.able-IND.TV-3s.3p (b$) Bright Dawn is able to keep (on) training the dogs for me. (b#)
A17. -ssa (v\v) ‘(be) expect(ed), (be) desire(d)’ (a) ‘(be) expect(ed)’: (14#), (31#), (39#), (56#), (65#), (69#), A15(b#). ‘(be) desire(d)’: (19#), (27#), (41#), (49#), (53#), (68#). (b) See (48#). A18. -ssaq (n\n) ‘expected, desired’ (a) ‘expected’: (30#), (34#), (37#), (39#), (47#), (58#), (61#), A22(b#). ‘desired’: (21#), (50#), A15(b#). (b) See (23#), (40#).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A15. -sariaqar (v\v) ‘need to, be necessary for __ > to’ (a) See (36#), (40#), (43#), (47#), (48#). (b) Igloos are very good in the wind . . . If we stay here another day and night, we will build a porch for ourselves, a very good place to cook in. [D: 46] (b#) Aqagu-mut aalla-ssa-nngik-ku-tta next.day-sg.DAT leave-be.expected-not-HYP>-1p illuiga-rput paa-liur-tariaqar-pa-rput igloo-1p.sg porch-make-need-IND.TV-1p.3s iga-vvi-ssa-tsin-nik [D#: 58] cook-location-desired-1p.sg-MOD (b$) If we’re not expected [by the marshal] to leave tomorrow, then we need to add to our igloo a porch, a desirable place for us to cook in.
384 Future Discourse in a Tenseless Language A19. -ssamaar (v\v) ‘plan to, be about to’ (a) Avi-ssamaar-pu-t. [LH: 76] (split-plan-IND.IV-3p) De har planer om at blive skilt. (‘They plan to get divorced.’) (b) See (30#), (46#) (two instances).
A21. -ssangatip (v\tv) ‘expect __ ? to’ (a) Aalla-ssangatip-pa-a. (set.out-expect.?.to-IND.TV-3s.3s) Han regner med at hun tager af sted. (‘He expects her to leave.’) (b) See (29#).
[LH: 77]
A22. -ssaqqaar (v\v) ‘be confidently expected’ (a) Sini-ssaqqaar-pu-q. [LH: 77] (sleep-be.confidently.expected-IND.IV-3s) Tag det roligt, han skal nok sove. (‘Relax, he’ll sleep all right.’) ‘If you can’t see it, how will you know where you are?# (b) ‘My lead dog will know,’ I said. [D: 78] (b#) Ittuquti-ma aqquti-ssa-rput lead.dog-1s.sg.ERG route-expected-1p.sg nani-ssaqqaar-pa-a. [D#: 93] find-be.confidently.expected-IND.TV-3s.3s (b$) [I] confidently expect my lead dog to find our expected route. A23. -ssaqqip (n\iv) ‘be suitable for’ (a) Illu-ssaqqip-pu-q. [LH: 78] (house-be.suitable.for-IND.IV-3s) Det er velegnet til beboelse. (‘It’s suitable to live in.’) (b) See (50#).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A20. -ssanga (v\v) ‘expect to’ (a) Aalla-ssanga-vu-q. [LH: 77] (set.out-expect-IND.IV-3s) Han regner med at tage af sted. (‘He expects to leave.’) One racer to another: ‘Maybe I come first.’ His eyes glittered. (b) ‘Where will you be?’ [D: 36] (b#) Illit nurmu qassi-u-ssanga-vi-t? you number what.number.n-be-expect-QUE-2s [D#: 48] (b$) What number do you expect to be?
Maria Bittner 385
A24. -(t)sir, -(t)sii (v\tv) ‘wait for__ ? to’ (a) Unnus-sii-va-a. [LH: 73] (become.evening-wait.for.?.to-IND.TV-3s.3s) Han venter til det bliver aften. (‘He’s waiting for the evening.#) (b) See (28#). B1. -lir (v\v) ‘begin’ (a) See (21#), (26#), (28#), (40#), (41#), (56#), (58#), (65#). (b) See (15#).
C3. -niar- (please-IMP) ‘please, . . . /let’s . . . ok?# (a) See (6). (b) See (16#).
Acknowledgements For thought-provoking comments and questions, I am deeply indebted to Hans Kamp and two reviewers for Journal of Semantics. Thanks are also due to the participants in my 2004 colloquia at CUNY and ZAS (Berlin), and in my 2003 and 2005 Rutgers seminars. I am also grateful to the former editor Peter Bosch for helpful editorial feedback.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
C1. -li/la (-OPT) ‘let____ . . .’ (a) See (6). (b) ‘He’s taken [the bait],’ [the old man] said. ‘Now I’ll let him [H: 44] eat it well.’ (b#) Iqqissi-vil-lu-ni calm.down-really-ELA>-3s> ii-lluar-titaar-nia-rallar-li-uk. swallow-well-bit.by.bit-try-for.now-OPT-3s.3s [H#: 29] (b$) For now let him really calm down and try to swallow it well bit by bit. C2. -niar-li (please-OPT) ‘please let____ . . .’ (a) See (6). (b) Race sponsors ask an injured competitor in training what to do: ‘My daughter will run the race.’ [D: 29] (b#) Pani-ga taarsir-lu-nga daughter-1s.sg replace-ELA>-1s sukka-niuti-qata-u-niar-li [D#: 40] go.fast-compete-fellow-be-please-OPT.3s (b$) Please let my daughter participate as a competitor instead of me.
386 Future Discourse in a Tenseless Language MARIA BITTNER Rutgers University Linguistics Department 18 Seminary Place New Brunswick, NJ 08901–1184 USA. e-mail:
[email protected]
Received: 26.06.04 Final version received: 15.02.05 Advance Access publication: 24.08.05
REFERENCES
Bergsland, K. (1955) A Grammatical Outline of the Eskimo Language of West Greenland. Micro-Editions of Interdocumentation Co. Zug, Switzerland. Bittner, M. (2001) ‘Surface composition as bridging’. Journal of Semantics 18:127–77. Bittner, M. (2003) ‘Word order and incremental update.’ Proceedings of CLS 39. CLS. Chicago. (pdf at http://www.rci.rutgers.edu/ ;mbittner/ou.html) Bittner, M. (2005) ‘Online update: Temporal, modal, and de se anaphora in polysynthetic discourse’. To appear in C. Barker and P. Jacobson, (eds), Direct Compositionality. Oxford University Press, Oxford. (pdf at http://www.rci.rutgers.edu/ ;mbittner/ou.html) Bohnemeyer, J. (2002) The Grammar of Time Reference in Yukatek Maya. Lincom Europa. Munich. Chung, S. and Timberlake, A. (1985) ‘Tense, aspect, and mood’. In T. Shopen (ed), Language Typology and Syntactic Description, Vol. III. Cambridge University Press, Cambridge, 202–58. Comrie, B. (1985) Tense. Cambridge University Press. Cambridge.
Givon, T. (1972) ‘Studies in ChiBemba and Bantu grammar’. Studies in African Linguistics, Supplement 3:1–247. Fortescue, M. (1980). ‘Affix ordering in West Greenlandic derivational processes’. International Journal of American Linguistics 46:259–278. Fortescue, M. (1984) West Greenlandic. Croom Helm. London. Haiman, J. (1980) Hua: A Papuan Language of the Eastern Highlands of New Guinea. John Benjamins. Amsterdam. Karttunen, L. 1976. ‘Discourse referents’. In J. McCawley (ed.), Syntax and Semantics 7. Academic Press, New York, 363–385. Klein, W. (1994) Time in Language. Routledge. London. Kleinschmidt, S. (1851) Grammatik der gro¨nla¨ndischen sprache. Reimer. Berlin. Lewis, D. (1979) ‘Attitudes de dicto and de se’. The Philosophical Review 88:513–543. Muskens, R. (1995) ‘Tense and the logic of change’. In U. Egli, P. Pause, C. Schwarze, A. von Stechow and G. Wienhold (eds.) Lexical Knowledge in the Organization of Language, John Benjamins, Philadelphia. 147–84. Ritter, B. and Wiltschko, M. (2004) ‘The lack of tense as a syntactic category: Evidence from Blackfoot and Halkomelem’. In J. Brown and
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Baker, M. and Travis, L. (1997) ‘Mood as verbal definiteness in a ‘‘tenseless’’ language’. Natural Language Semantics 5:213–69.
Maria Bittner 387 T. Peterson (eds), UBC Working Papers in Linguistics, Vol. 14. Sapir, E. (1922) Language. Harcourt and Brace. New York. Shaer, B. 2003. ‘Toward the tenseless analysis of a tenseless language’.
Proceedings of SULA 2, GLSA, University of Massachusetts at Amherst, 139–56. Stump, G. (1985) The Semantic Variability of Absolute Constructions. D. Reidel. Dordrecht.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Journal of Semantics 22: 389–400 doi:10.1093/jos/ffh031 Advance Access publication May 9, 2005
More Than Bare Existence: An Implicature of Existential Bare Plurals ARIEL COHEN Ben-Gurion University
Abstract
1 EXISTENTIAL BARE PLURALS Carlson (1977) has established that there are semantic differences between the existential reading of bare plurals (henceforth BPs) and other existentially interpreted noun phrases. For example, existential BPs may receive narrow scope only, whereas other indefinites are scopally ambiguous. Thus, (1a) is ambiguous: it may mean either that for everyone there is a book (are some books) on giraffes that he or she read, or that there is one book (are some books) on giraffes that everyone read. Sentence (1b), in contrast, can only receive the first reading. a book (1) a. Everyone read on giraffes. some books b. Everyone read books on giraffes. It is important to note that, while acknowledging that existential BPs and other indefinites behave differently (e.g. in terms of scope), Carlson (and many others who have analysed the meaning of existential BPs) still maintains that they mean the same: both make an existentially quantified statement. However, while they do seem to have the same truth conditions, it is not clear that existential BPs and explicit indefinites are really Ó The Author 2005. Published by Oxford University Press. All rights reserved.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Existential bare plurals (e.g. dogs) have the same semantics as explicit existentials (e.g. a dog or some dogs) but different pragmatics. In addition to entailing the existence of a set of individuals, existential bare plurals implicate that this set is suitable for some purpose. The suitability implicature is a form of what has been variously called informativeness-based or R-based implicature. Condoravdi (1992, 1994) and others have claimed that bare plurals have a third reading (in addition to the generic and the existential), sometimes called quasi-universal. However, the suitability implicature is sufficient to account for the quasi-universal interpretation, without the need to stipulate a distinct reading of bare plurals.
390 More Than Bare Existence interpreted in the same way. This paper argues that their interpretations are, in fact, different: existential BPs carry, in addition to their existential force, an implicature to the effect that the set denoted by the BP is suitable for some purpose. We will investigate the characteristics of this implicature, and the mechanisms by which it is triggered. Finally, we will evaluate proposals for yet a third, quasi-universal interpretation of BPs, and consider reasons to believe that these do not constitute distinct readings, but come about as a consequence of the suitability implicature.
2.1 The implicature of existential BPs Consider the following examples: (2) a. This tractor has wheels. b. This tractor has some wheels. Suppose the tractor in question has only two wheels. Then (2a) would be odd, but (2b) would be fine. Sentence (2a), but not (2b), suggests that the tractor has four wheels, suitably arranged: two large ones in the rear, two smaller ones in front. Let us propose that, in general, existential BPs entail the existence of some set, and implicate that it is suitable. Suitable for what? Suitability must be defined relative to some purpose of the speaker. Presumably, the speaker’s purpose in (2a) is to drive the tractor, so the speaker implicates that the tractor is ready to run (at least as far as the number and arrangement of wheels are concerned). Note that if the speaker’s purpose were different, the criterion for suitability would be different. For example, suppose we hear (2a) in a junkyard, while we are looking for a wheel to replace the broken wheel on our own tractor. In this case, (2a) would be quite acceptable even if the tractor in question did not have a complete set of wheels. For another example, consider the following pair, suggested by Tova Rapoport (pc): (3) a. John has playing cards. b. John has Victorian playing cards. Sentence (3a) implicates that John has a set of cards suitable for playing, e.g. a full deck. Sentence (3b), on the other hand, may be felicitously uttered even if John has only a few Victorian cards; in the context of this sentence, John is most probably a collector of Victorian cards, and any number of cards is suitable for collecting purposes.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
2 SUITABILITY
Ariel Cohen 391
Sometimes any group is suitable, e.g. in the case of (3b), or (4) below. (4) This shirt has stains. Presumably, the speaker wishes to indicates that the shirt is dirty; since any number of stains, and any arrangement of stains on the shirt is suitable for the shirt to be considered dirty, the reading of the BP is purely existential, and (4) means the same as (5). (5) This shirt has some stains.
(6) In this forest, trees are dying. The speaker who utters (6) presumably wishes to indicate some disease, or other problem, with the forest as a whole. Hence, if only a handful of trees, or only trees in a certain grove were dying, (6) would not be felicitous. There must be a non-negligible number of dying trees, and they must be distributed more or less evenly in the forest. There are cases where only very specific groups count as suitable. As an example, take the ‘classic’ example of an existential reading of a BP, namely the subject of available. Suppose we wish to send a spaceship to the moon. We contact NASA, and get the following response: (7) Astronauts are available Sentence (7) says more than simply that there exist some available astronauts. Rather, it implies that there is a set of available astronauts that is suitable, in terms of its size, the training of its members, etc., for our mission: there are, say, three astronauts, who have the respective roles of command module pilot, LEM pilot, and mission commander. Note that the suitability implicature is perceived by the hearer, who may felicitously respond to it as if it were explicitly made: (8) a. A: B: b. A: B: c. A: B: d. A: B:
This tractor has wheels. So where do you want to go with it? John has playing cards. Great, let’s start a game. In this forest, trees are dying. Well, it was going to be cut down anyway. Astronauts are available. But the mission cannot go ahead, because the rocket is still malfunctioning.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In other cases, suitability implies constraints on the size and distribution of a set. For example:
392 More Than Bare Existence
(9) a. b. c. d.
This tractor doesn’t have wheels. John doesn’t have playing cards. In this forest, trees aren’t dying. Astronauts aren’t available.
Sentence (9a) does not mean that the tractor doesn’t have a suitable set of wheels, but that it doesn’t have any wheels at all; in other words, the negation applies to the existential claim made by (2a), not to its suitability inference. Similarly, (9b) means that John has no cards, not that he fails to have a full deck; (9c) means that no trees are dying, not that no suitable set of trees are dying; and (9d) negates the existence of any available astronauts, not only of a suitable set of astronauts. Another indication that suitability is an implicature is that it is cancellable; the following sentences are all quite acceptable, and are certainly not contradictions. only three (10) a. This tractor has wheels, but . they are all the same size not enough for a game b. John has playing cards, but . the Queen of Hearts is missing only a few c. In this forest, trees are dying, but . only near Eeyore’s house not enough d. Astronauts are available, but for this kind they aren’t trained of mission. If suitability is an implicature, it must be derived by some pragmatic process. What is the nature of this process?
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
B’s response in (8a) relates to the implicated claim that the tractor has a set of wheels suitable for the purpose of riding it; in (8b), B responds to A’s implicature that John has a set of playing cards suitable for playing; in (8c), B perceives that A is upset about the forest as a whole, not about the fate of a handful of dying trees; and in (8d), B understands A’s purpose to indicate that the mission is ready (and the set of available astronauts is suitable for it). How do we know that suitability is implicated by, rather than part of the literal meanings of, the sentences under discussion? One piece of evidence comes from their behaviour under negation. When a sentence containing an existential BP is negated, the negation applies to the existential statement, not to the suitability inference:
Ariel Cohen 393
2.2 Two types of implicature Horn (1984) reformulates Grice’s (1975) in terms of an opposition between two principles: the Q-principle, which says that the speaker ought to provide as much information as possible, and the R-principle, which says that the speaker ought to be as economical as possible. Sometimes one of them wins, and sometimes the other. As an example of the Q-principle, note that an utterance of (11) implicates that John employs exactly one secretary. (11) John employs a secretary.
Horn notes that, while (12a) means simply that the speaker’s brother changed his location to the specified location—the church (the jail, the school), (12b) implicates that the brother went there for the purpose of
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
This is an implicature based on lack of information: because the utterance does not say that John has more than one secretary, we conclude that the statement that he does is false (as far as the speaker knows). So, whatever is not said in the sentence is implicated to be false. The R-principle, namely the requirement that communication be as economical as possible, leads to an additional type of implicature. Atlas and Levinson (1981) note that we normally infer from (11) that John employs a female secretary. This type of implicature, which they call informativenessbased implicature, works in the opposite direction: information that is not in this sentence is implicated to be true, rather than false. One of the cases where the R-principle wins is when the speaker intends to convey some stereotypical information. In this case, the effort exerted by the speaker to make this information explicit is not worth the informational gain: since the information is stereotypical, hearers can probably infer it on their own. This, then, is the explanation for (11): since the stereotypical secretary is female, a speaker who knows that John’s secretary is female would not bother to say so, knowing the hearer is quite capable of drawing the appropriate inference. R-based implicature is often triggered when a word, or morpheme, is left out. The following examples are from Horn (1993): 9 8
394 More Than Bare Existence performing the stereotypically associated function (praying, being incarcerated, studying). One test for R-based implicature is that it cannot be denied, not even by metalinguistic negation. Horn (1989) argues that metalinguistic negation can deny almost anything about an utterance, from its phonetics, through its register, to its conversational implicatures. He exemplifies metalinguistic negation with sentences such as the following: (13) a. I didn’t call the po´lice, I called the polı´ce. b. Grandpa isn’t feeling lousy, Johnny, he’s just a tad indisposed. c. You didn’t eat some of the cookies—you ate them all.
(14) a. John doesn’t employ a secretary—he employs a male secretary. 9 8 church = < —he went there but b. My brother didn’t go to jail ; : school 8 9 didn’t pray < = wasn’t incarcerated . : ; didn’t study
2.3 Suitability and R-based implicature The difference between an existential BP and an explicitly existentially quantified DP is that the former lacks a determiner that the latter has. This is exactly the difference seen in (12), whose stereotypical interpretation is a consequence of R-based implicature. It is attractive, therefore, to explain suitability too as a case of R-based implicature. Indeed, in many cases of sentences implicating suitability, the BP is interpreted stereotypically. An utterance of (2a) implicates suitability since the stereotypical tractor has all its wheels; (3a) implicates suitability since the stereotypical set of cards is a full deck; (6) implicates suitability since the stereotypical case of trees dying in a forest is when the whole forest is gravely ill,1 and (7) implicates 1 Of course, some trees are dying in every reasonably sized forest, however healthy. But in this case the forest is explicitly mentioned, and the stereotypical case of a forest with dying trees is a forest in trouble.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Example (13a) rejects an utterance on the basis of its intonation, (13b) rejects an utterance on the basis of its improper register, and (13c) denies the Q-based implicature of an utterance. However, it is impossible to use metalinguistic negation to negate an R-based implicature, as can be seen by the unacceptability of the following examples:
Ariel Cohen 395
suitability since the stereotypical set of astronauts is a team suitable for a mission. It should be noted, however, that there is more to suitability2 than stereotypicality. As we have seen above, in the right context, a nonstereotypical interpretation may still satisfy suitability (recall, for example, the context where we are looking for spare parts in a junkyard, and encounter a tractor with three wheels). If suitability is an R-based implicature, we predict that metalinguistic negation ought not to deny it. This is, indeed, the case: This tractor doesn’t have wheels—it has two. John doesn’t have playing cards—he misses the King and Ace. In this forest, trees aren’t dying—at most ten. Astronauts are not available—we are missing a LEM pilot. 3 QUASI-UNIVERSAL READINGS
3.1 A third reading of bare plurals? In this article, we have considered the existential reading of BPs and its suitability implicature. It is well known that, in addition to the existential interpretation, BPs have a second reading: generic. But does this exhaust the space of readings of BPs? Condoravdi (1992, 1994) claims that BPs have yet a third reading. This reading is claimed to be exemplified by (16b), when uttered following (16a). (16) a. In 1985 there was a ghost haunting the campus. b. Students were aware of the danger Condoravdi claims that the reading of the BP in (16b) is neither generic nor existential. The sentence is clearly not a generic because it is not lawlike; for example, it does not support the counterfactual (17). (17) If the captain of the Titanic had been a student on campus in 1985, he would have been aware of the danger.3 2 And perhaps to R-based implicature in general; but a formal analysis of R-based implicature will have to wait for another occasion. 3 Let us get one red herring out of the way: it would be wrong to argue that the BP must be interpreted generically because be aware of is an individual-level predicate. The fact is that transitive individual-level predicates do not require their BP subjects to be interpreted generically:
(i)
a. b.
Monkeys live in that tree (Glasbey 1998). Criminals own this club (Cohen & Erteschik-Shir 2002).
Sentence (i.a) says that some monkeys live in that tree, not that living in that tree is a general property of monkeys; and (i.b) says that some criminals own this club, not that owning the club is a property characterizing criminals in general.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(15) a. b. c. d.
396 More Than Bare Existence Condoravdi claims that the BP is not interpreted existentially either. The sentence says more than simply that some students were aware of the danger; in her view, (16b) means that all (or almost all) students were. She calls this interpretation the functional reading; let us, instead, follow Dobrovie-Sorin’s (1998) terminology in referring to this as the quasi-universal interpretation. Crucially, Condoravdi argues that the quasi-universal interpretation is a distinct reading. She claims that the following sentences are ambiguous: (18) a. Details will be presented tomorrow. b. Prices went up today.
3.2 Quasi-universality is suitability Are we, then, forced to conclude that quasi-universality constitutes a distinct reading? Not really. There is an alternative: the quasiuniversal force comes from the suitability implicature. Here is how this implicature is plausibly generated. Consider again the discourse in (16). What is the purpose of the speaker in uttering it? Presumably, it is to indicate that the ghost was detected, with the appropriate steps (whatever those might be!) taken. Maybe it is to 4 Interestingly, Greenberg (2002) proposes a theory of genericity that uniformly treats generics in episodic and non-episodic contexts, but she denies that quasi-universal readings are generic.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
According to Condoravdi, these sentences have the usual existential reading, according to which some details will be presented, or some prices went up. But, in addition, they also have a quasi-universal reading: all (or almost all) details will be presented, and all (or almost all) prices went up. Condoravdi has discovered an interesting set of facts. But is her interpretation of them correct? Is this really a third reading of BPs, or can this interpretation be subsumed by the well established generic or existential reading? Dobrovie-Sorin (1998) has questioned Condoravdi’s interpretation, and proposed that, rather than a third reading, this is actually the generic reading. Its lack of lawlikeness is explained by the ‘episodic context’ of sentences such as (16a) and (18). If this account is to succeed, however, Dobrovie-Sorin ought to have provided a unified semantics for generics, which will explain why they are law like in some contexts, but not in others. Unless such a theory is forthcoming, we are still left with a three-way ambiguity: existential, lawlike generic, and episodic generic BPs.4
Ariel Cohen 397
indicate that the danger was averted. In this context, it clearly seems that a suitable set of students who were aware of the danger must include all, or almost all of the students. If a few students had been left unaware of the danger, they might have gotten hurt by the ghost, and the speaker’s purpose would be moot. Note that if the speaker had another purpose, the suitable set of students might be different. Suppose the point were to indicate that the students were more vigilant than the professors. Then, the speaker might say something like (19). (19) Students were aware of the danger, but professors had no idea what was going on.
(20) A: The ghost roamed the campus completely undetected. B: That’s not true! Students WERE aware of the danger. With respect to the examples in (18), note that they can only be perceived to be ambiguous out of context. In a given context, when the purpose of the speaker is clear, they can only have one reading: (21) a. If you want to know the complete story, details will be presented tomorrow. b. We have to cut our expenses: prices went up today. For sentence (21a), the suitable details are those that will enable the hearer to understand the complete story; presumably, these are all of the relevant details. And with respect to (21b), a suitable set of prices will be that of many, but not necessarily all (or almost all), of the commodities typically bought by the speaker and hearer. Even out of context, the sentences in (18) are not really ambiguous, and have only one reading. The different interpretations come about because, out of context, the purpose of the speaker is not clear, hence the criteria for suitability are not clear. This fact can be established by applying tests for ambiguity. The word bank, for example, is lexically ambiguous between the senses river bank and financial bank. Hence, (22) is ambiguous: it can mean either that there is a river bank near John’s house, or that there is a financial bank there. (22) There is a bank near John’s house. Sentence (23) is also ambiguous.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In this case, it does not seem necessary that all, or almost all students were aware of the danger, so long as a substantial number were. An even smaller set is suitable for (20), where B’s aim is presumably to refute A’s claim that the ghost was not detected at all.
398 More Than Bare Existence (23) There is a bank near John’s house, and the same goes for Mary’s. However, the type of bank (river bank or financial bank) near Mary’s house has to be the same as the type of bank near John’s house. This fact can be demonstrated by the unacceptability of (24). (24) There is a bank near John’s house, and the same goes for Mary’s, the only difference being that his house is near a financial bank, and her house is near a river bank.
(25) a. John employs a secretary. b. John employs a female secretary. c. John employs a male secretary. Indeed, (26) is perfectly acceptable: (26) John employs a secretary, and the same goes for Mary, the only difference being that his secretary is female, and her secretary is male. How do quasi-universal BPs behave? Consider: (27) a. Details will be presented tomorrow, and the same goes for next week, the only difference being that tomorrow almost all the details will be presented, and next week it will be just a few insignificant details. b. Prices went up today, and the same will happen tomorrow, the only difference being that today everything became more expensive, while tomorrow just a few luxury items will be affected. The sentences in (27) are quite acceptable, indicating that this is not a case of ambiguity, in contrast with Condoravdi’s theory. Crucially, it follows from the claim made here that the quasiuniversal reading is not part of the literal meaning of the BP (as claimed by both Condoravdi and Dobrovie-Sorin), but is brought about by a suitability implicature. Let us consider this point in more detail. If the quasi-universal interpretation were part of the literal meaning of the sentence, then negating the sentence should negate the quasiuniversal meaning. For example, the sentences in (28) should be true in case not all (or almost all) of the students were aware of the danger.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In contrast, R-based implicature does not behave in this way. Sentence (11), repeated as (25a) below, is not ambiguous between (25b) and (25c).
Ariel Cohen 399
(28) a. Students weren’t aware of the danger. b. It is not the case that students were aware of the danger. In fact, this is not the case. If, say, 10% of the students were aware of the danger, the sentences in (28) would clearly be false. This is exactly what would be expected if the literal interpretation of the BP is existential, but it is problematic for a theory that takes the quasi-universal interpretation to be part of the literal meaning of the sentence. Perhaps the clearest indication that we are, in fact, dealing with an implicature is the fact that quasi-universality is cancellable. (29) Students were aware of the danger, but only a handful.
Acknowledgments I would like to thank Nomi Erteschik-Shir, Bart Geurts, Larry Horn, Anita Mittwoch, Tova Rapoport, and two anonymous reviewers for helpful suggestions and insightful discussions.
Received: 29.09.04 ARIEL COHEN Final version received: 16.02.05 Department of Foreign Literatures and Linguistics Advance Access publication: 09.05.05 Ben-Gurion University of the Negev Beer Sheva 84105 Israel e-mail:
[email protected]
REFERENCES Atlas, J. D. & Levinson, S. C. (1981) ‘Itclefts, informativeness, and logical form: radical pragmatics (revised standard version)’. In P. Cole (ed.), Radical Pragmatics. Academic Press, New York. 1–61. Carlson, G. N. (1977) ‘Reference to kinds in English’, PhD dissertation,
University of Massachusetts, Amherst. (Published 1981, Garland, New York.) Cohen, A. & Erteschik-Shir, N. (2002) ‘Topic, focus, and the interpretation of bare plurals’. Natural Language Semantics 10:125–165. Condoravdi, C. (1992) ‘Strong and weak novelty and familiarity’. In C. Barker
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Sentence (29) is perfectly acceptable; but if quasi-universality were part of the literal meaning of the sentence, it should be contradictory. According to the theory presented here, namely that quasi-universality is an implicature, this is exactly as it should be. We can conclude, then, that quasi-universal interpretations of BPs are not a distinct reading, and their quasi-universal force comes from a suitability, R-based implicature.
400 More Than Bare Existence Grice, H. P. (1975) ‘Logic and conversation’. In P. Cole & J. Morgan (eds), Syntax and Semantics 3: Speech Acts. Academic Press. New York. 41–58. Horn, L. R. (1984) ‘Toward a new taxonomy for pragmatic inference: Q-based and R-based implicature’. In D. Schiffrin (ed.), Meaning, Form, and Use in Context: Linguistic Applications. Georgetown University Press. Washington. 11–42. Horn, L. R. (1989) A Natural History of Negation. University of Chicago Press. Chicago. Horn, L. R. (1993) ‘Economy and redundancy in a dualistic model of natural language’. In S. Shore & M. Vilkuna (eds), SKY 1993: 1993 Yearbook of the Linguistic Association of Finland. 33–72. Lawson, A. (1998) Proceedings of the Seventh Conference on Semantics and Linguistic Theory. Cornell University. Ithaca.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
and D. Dowty (eds), Proceedings of the Second Conference on Semantics and Linguistic Theory. Ohio State University, Columbus. 17–37. Condoravdi, C. (1994) ‘Descriptions in Context’, PhD dissertation, Yale University. (Published 1997, Garland, New York.) Dobrovie-Sorin, C. (1998) ‘Types of predicates and the representation of existential readings’. In A. Lawson, Proceedings of the Seventh Conference on Semantics and Linguistic Theory. Cornell University, Ithaca. 117–134. Glasbey, S. (1998) ‘I-level predicates that allow existential readings for bare plurals’. In A. Lawson, Proceedings of the Seventh Conference on Semantics and Linguistic Theory. Cornell University, Ithaca. 169–179. Greenberg, Y. (2002) ‘Manifestations of Genericity’, PhD dissertation, Bar-Ilan University, Israel. (Published 2003, Routledge, London and New York.)
Journal of Semantics 22: 401–438 doi:10.1093/jos/ffh019 Advance Access publication March 14, 2005
The Semantic and Syntactic Decomposition of get: An Interaction Between Verb Meaning and Particle Placement ANDREW McINTYRE Universita¨t Leipzig
VPs with get and a PP/particle provide an argument for lexical decomposition in syntax. Get (and German kriegen) has a ‘hindrance’ reading, which does not denote causative events and resembles manage in that the result is portrayed as hard to achieve, and in that possibility operators do not affect the meaning under negation: I didn’t (¼couldn’t) get the key in. These effects surprisingly follow from an analysis where hindrance-get VPs are nothing more than inchoatives of have-VPs of the type have the key in. In get out one’s wallet, we see another reading which is genuinely causative and is not found with German kriegen. Hindrance-get VPs (like VPs with have, want and need, which decompose with HAVE, and unlike causative get and other causative-agentive verbs) disallow particle-object order (get/take out your wallet vs. *get/have/want/ need in the key). The effects of semantics on word order are shown to be unmysterious only if the HAVE predicate in the meaning of hindrance-get is a syntactic head.
1 INTRODUCTION Many linguists believe that verb meanings (partly) decompose into primitives like CAUSE and BECOME (e.g. Dowty 1979; Jackendoff 1990; Wunderlich 1997). Some of these linguists assume what I call syntactic decomposition, in that the primitives are treated as morphemes present in the syntax. For instance, Harley (1995) and Richards (2001) express a semantic decomposition of give in a syntax like (1). Causativeagentive morphemes like CAUSE (also answering to names like v, VOICE) are now common in generative studies (e.g. Baker 1997; Hale & Keyser 1993; Harley 1995; Kratzer 1996; Marantz 1997; Pesetsky 1995; Pylkka¨nen 2002; Stechow 1996). (1) Mary gave John a book: [VP Mary[V# a book]]]]
CAUSE[PP
John [P#
HAVE
Ó The Author 2005. Published by Oxford University Press. All rights reserved.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Abstract
402 The Semantic and Syntactic Decomposition of get
(2) a. b. c. d.
I got the screws out; I got him to the station I got foutg my wallet foutg to pay. The lock is rusty, and I can’t get f*ing the key fing. I had/wanted/needed f*ing the key fing.
A foretaste of the odd behaviour of get is the puzzle in (2b–d), which involves restrictions on the position of a complementless preposition (‘particle’). Particles can usually appear before or after a direct object, cf. (2b). (2c) shows that get sometimes disallows particle-object order. Replacing get in (2c) with push makes both sequences possible. Blockages on particle-object order are observable with have (and other verbs argued to have a silent have in their complements, e.g. want and need, cf. (2d) and section 4.4). I argue that the uses of get like that in (2c) which disallow particle-object order are inchoatives of have, while those that allow particle-object order like (2b) have a genuinely causal semantics. Independent differences between the use of get in (2c) and that in (2b) are that (2c), but not (2b), can be translated with German kriegen or bekommen and displays a peculiar form of non-agentivity I call ‘hindrance specialization’: it suggests that the result is achieved despite resistance. The effect of the meaning of get on word order in
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The syntactic decomposition hypothesis differs radically from theories decomposing verbs at a lexical-semantic level or not decomposing them. Methodological arguments avail us nothing in choosing between these approaches. Syntactic decomposition simplifies the syntaxsemantics mapping, but the received conception of the verb as syntactically atomic, standard outside minimalism, simplifies the mapping between syntax and its audible output. Syntactic decomposition requires an abstract, complex syntax, which many find methodologically repugnant, while a syntax where ‘what you see is what you get’ (Jackendoff 1990: 159) entails abstraction and complexity in a lexical/semantic component, which is methodologically unappealing to other linguists. Conceptual arguments are a flimsy basis for deciding the issue, so we need empirical arguments. This study offers a new empirical argument for syntactic decomposition, complementing those in Baker (1997), Hale & Keyser (1993) and Stechow (1996). However the argument is received, I hope it will become clear that there is intrinsic worth in someone’s attempt at understanding the data on which it is based, structures like (2a), where a DP and PP/particle appear in the complement of get. The data have languished in obscurity, despite displaying quirks affording insights into agentivity, causation, event structure, and the syntax–semantics interface.
Andrew McIntyre 403
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(2b, c) is shown to follow naturally only if the elements into which getVPs decompose semantically (BECOME, HAVE) are present in syntax. We proceed as follows. Section 2 introduces basic semantic observations about get + PP structures. Section 2.1 discusses the hindrance reading. Section 2.2 discusses what I call the ‘unintentional’ reading (I got ants in my room). This is irrelevant to particle verbs but must be mentioned because it helps us to understand the hindrance use. Section 2.3 describes what I call the ‘(genuinely) causative’ use in (2b), which (like take and unlike put) entails that the object comes to be possessed or manually controlled by the subject. This requirement suggests that these VPs originated as resultative constructions based on agentive DP complement uses (get/take the book), but some idiosyncrasies speak against the synchronic validity of this analysis for get. A causative semantics is formulated for this reading, supplemented by conditions stipulating the contexts where it may appear. Section 3 gives a more detailed analysis of the hindrance reading, arguing that it is an inchoative of have. Surprisingly, this analysis captures (a) the fact that the subject of hindrance-get is invariably seen as responsible for having brought about the result, and (b) hindrance specialization. Section 4 discusses the particle position facts. After an introduction to the basic facts on particles, differences between speakers in their acceptance of a particle before an object in a given combination are shown to follow from differing degrees of liberality in the use of causative get. Section 4.3 introduces the light verbs used here. In a framework indebted to distributed morphology, I survey most uses of get, concluding that it is a spellout of BECOME in all its uses. Section 4.4 defends an account of the particle shift in which HAVE—seen as an item present in the syntax—disallows the incorporation of particles, thus disqualifying them from movement to a head position to the left of the object. Section 4.5 argues that only syntactic lexical decomposition can give a non-arbitrary account of the data. Section 5 summarizes the main properties of the readings of get discussed here. Space permits only a focus on transitive get + PP VPs, with side glances at other uses of get, e.g. possessive uses (get a book), unaccusative uses (get off the boat), AP uses (get him drunk), ‘passive’ uses (get killed), ‘inchoative’ uses (get working) and ‘causative’ uses like those in get him to go, get him working. Ditransitive structures like (3a) are irrelevant here, pace a reviewer. Ditransitive get VPs are augmentations by a beneficiary of the agentive DP complement use of get in the sense ‘fetch, buy’. I assume with Pylkka¨nen (2002) that these are licensed by
404 The Semantic and Syntactic Decomposition of get a benefactive-applicative head below the agent-licensing head. (3a) is interpreted like (3c), not (3b), for (a) and (c), but not (b), are true if I have obtained the book but not yet given it to Ann. Linguists wanting to derive ditransitive get VPs by dative shift would have to start with structures like (c), not (b). It is reasonable to exclude (c) from the purview of this study. The for-PPs are not directional complements like the PPs in (3b) and (2a), but are adjuncts (cf. do so ellipses: I got a book for Ann, then I did so for John, v. *I got a book to Ann, then I did so to John).
The literature on get does not help with the questions discussed here. There are several studies on ‘passive’ uses (e.g. Haegeman 1985). General treatments of get are scarce, although (or because?) get dwarfs the oft-discussed verb have in the complexity of the issues it raises. Givon & Yang (1994), Kimball (1973) and Tobin (1994) offer worthwhile insights, but do not delve into complexities like those discussed herein. 2 THREE READINGS FOR get + PP STRUCTURES I now introduce the three readings of get + PP structures. Since the empirical terrain covered in this section is substantial, a summary of the main points is provided in section 5.
2.1 The hindrance reading My introduction to the uses of get + PP structures begins with a use called ‘hindrance-get’. It suggests that the result is hard to attain. Put otherwise, this use of get displays ‘hindrance specialization’, being specialized to contexts where the subject overcomes some kind of obstacle. Consider (4), where # indicates acceptability contingent on special contextual assumptions. In (4a) get is good in a context where the arm-lifting is difficult, for instance if the subject is injured or holding a heavy object. In (4b–d), the use of get in describing the physically easy act of carrying a book somewhere is acceptable only if obstacles to the reaching of the goal are either explicit (as in (4c, d)) or contextually inferred (for instance if it is known that the book in (4b) could be confiscated). (4e) is acceptable in any context if the time limit expressions are included, apparently because time is construable as
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(3) a. I fgot/fetched/boughtg Ann the book b. I fgot/*fetched/*boughtg the book to Ann c. I fgot/fetched/boughtg the book for Ann
Andrew McIntyre 405
a factor competed against, as an obstacle. (4f) can be reconciled with the above remarks if we extend the notion of hindrance to cover imagined difficulties. Were there no imagined difficulties in (4f), there would be no point in denying them with without difficulty. (4)
Hindrance-specialized get + PP structures with goal PPs are achievements, denoting only arrival at the goal, while VPs with other verbs are accomplishments, i.e. also express progress toward the goal. Hindrance-get is not usable if the theme’s progress toward the goal is interrupted, as in (5a), while other caused motion verbs pattern with standard accomplishments in permitting the progressive with an uncompleted event (cf. the imperfective paradox, e.g. Dowty 1979: 133). With get in (5b), the assertion is that we will arrive at the station in ten minutes’ time, while with take/drive it is that the journey will start in ten minutes’ time. In (5c), on the way implies that the journey has started. This contradicts the negation of take/drive, but not the negation of get. (5)
a. I was ftaking/driving/*gettingg them to France, but we never reached the border. b. She’ll fget/take/driveg us to the station ten minutes from now. c. Because the car broke down on the way, she didn’t fget/ *take/*driveg us to the station.
Another facet of the achievementhood of hindrance-get VPs is that they do not denote causing events or agentive acts. In (6a) and (6c) almost may take scope over the actions causing the results, while (b) and (d) lack wide scope readings denying that the subject acted. This follows if hindrance-get VPs, unlike the causative/ resultative VPs, do not denote agentive acts. (6)
a. b. c. d.
I almost fput/stamped/blewg the fire out. I almost got the fire out. I almost fput together/assembledg the machine. I almost got the machine together.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
a. She fraised/moved/#gotg her arm up to shoulder level. b. He ftook/carried/#gotg the book through the room. c. He got the book through the room funimpeded by the crowd/unnoticedg. d. He got the book through fcustoms/the narrow openingg. e. She got me to the station # (in ten minutes/before the train left). f. I got the thread through the needle without difficulty
406 The Semantic and Syntactic Decomposition of get (7) makes the same point. In the deviant variants, the subordinate VPs have causing events in their denotation. The matrix verbs express acts construed pragmatically as identical to the causing events in the unacceptable subordinate VPs. By consequence, the interval occupied by the main clause actions is a subpart of the interval occupied by the subordinate clause events with the causative verbs. This yields a non sequitur, since the complementizer before indicates that the subordinate clause events occur after (not during) the main clause events. That no such temporal paradox exists with get in (7) is explicable if the get-VPs do not denote causing events.
Thus, hindrance-get VPs do not include agents’ acts in their denotation (although they presuppose them, see below). They are thus non-agentive. It may be objected that the non-agentivity of hindranceget is at best weakly supported by judgments with normal agentivity tests like (8) (Cruse 1973; Dowty 1979: 112f; the contexts in (8a–c) diagnose volitionality, and the sense of the test in (8d) seems to be that non-auxiliary do is an underspecified agentive verb). (8) a. I told him to fdrive/?getg the car to the top of the hill. b. He fhelped/??gotg the old man across the road to court his favour. c. Please fbring/??getg him to my office at some stage. d. What I did was fput/?getg the key in the lock. However, it has been known (or forgotten) since Cruse (1973) that other non-agentive verbs pass these tests. In (9a) the tests are passed by states (cf. *he’s having it ready, *he’s being absent). It seems to me that these cases and any structures like (8) where get is judged acceptable involve a metonymy in which the VP with get/have/be stands for the actions which the subject must perform to bring about the situation named by the VP. With regard to imperative, a further extraneous factor is a possible optative reading. The German imperative seems to lack this reading, and the German equivalents of hindrance-get, kriegen and bekommen, resist imperative strongly, cf. (9b). In sum, I feel justified in upholding the claim that hindrance-get VPs do not include the actions of an agent in their denotation.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(7) a. They had to fight the fire for hours before they fgot it out/ *put it out /*extinguished it/*caused it to go outg. b. I had to twist the key for ages before I fgot/*put/*stuck/ *forcedg it in the rusty lock. c. I had to pull on the boot for ages before I fgot/*took/ *pulledg it off.
Andrew McIntyre 407
(9)
a. What you must do is be absent that day; Please have your passports ready; Please understand that this will ruin our lives. b. fHa¨mmer’/*krieg’/*bekomm’g bitte den Nagel in die Wand rein! Please fhammer/get/getg the nail into the wall.
(10)
I didn’t get them to the stadium ¼ I was unable to get them to the stadium. a#. I didn’t take them to the stadium 6¼ I was unable to take them to the stadium. b. I didn’t get the key in the lock ¼ I couldn’t get the key in the lock. b#. I didn’t put/insert/stick the key in the lock 6¼ I couldn’t put/insert/stick the key in the lock. a.
There are other hindrance-specialized verbs. German kriegen and bekommen are good translations of hindrance-get in all respects, including hindrance specialization. Unaccusative get + PP structures also have a hindrance-specialized use, cf. I got through fa thick book/#a half-page abstractg, where the latter variant is odd unless illness or 1 Other effects of the presupposition: Answering ‘no’ to (i) admits attempted envenomation. In (ii) get suggests that I could not have attained the result unaided (though I would have tried to), but with take I could have refused to act towards attaining the result unaided, though I perhaps could have attained it.
Did you get poison into anyone’s beer? (ii) I wouldn’t have fgot/takeng Basil home unaided. The variants with cannot/be unable in structures like (10a, b) sound more natural than those without. The German translations of get show the opposite preference, cf. the literal and idiomatic glosses in (i). I cannot explain this, but an explanation seems unnecessary given that the AngloGerman contrast is not specific to get, cf. the glosses in (ii). (i)
(i)
2
Ich kriege den Nagel nicht durchs Brett; I get the nail not through the board; ‘I can’t get the nail through the board’; (ii) Ich finde/verstehe/sehe/ho¨re/fu¨hle/rieche/schmecke I find/understand/see/hear/feel/taste it not ‘I can’t find/understand/see/hear/feel/smell/taste it’
Wir kommen nicht da rein We come not there in ‘We can’t get in there’ es nicht
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(10) showcases another oddity of hindrance-get. Under negation, possibility operators do not affect the interpretation of hindrance-get VPs, while negation of causative-agentive VPs without possibility operators is possible if the subject can attain the result but chooses not to, a reading which is unavailable if possibility expressions are present. Hindrance-get behaves as in (10) because it presupposes an attempt at achieving the result. The fact that presuppositions survive under negation (if not explicitly negated: I didn’t get the key in because I didn’t try) coupled with the non-occurrence of the result trivially suggests that the subject is unable to attain it.1,2
408 The Semantic and Syntactic Decomposition of get
2.2 Unintentional get (11) instantiates unintentional get. Here there is no hindrance specialization. The subject is not responsible for the result. Unintentional get does not combine with particles and is not directly relevant to my main argument, but section 3 requires some knowledge of it. (11)
Ii got ants in myi room; Shei got her i hair in a tangle (because the wind blew); The camerai got dust in iti; Hei got blood all over himi and shrapnel in hisi arm in the explosion; The car i got paint on the top of itsi fender
(11) makes clear that unintentional get requires co-indexation between the subject and an item somewhere in the complement of get, a requirement also known to hold with certain uses of have (see section 3). Unintentional get disallows unambiguously directional PPs like into: I got ants into my room has a hindrance-specialized, not an unintentional reading. It is clear that (11) involves small clause (predicative) complementation. An analysis where the sole complement is a DP (e.g. get [DP ants in my room] or [VP [VP get ants] in my room]) is falsified by cases in (11) where the subject does not get the object: *the camera got dust, *she got her hair, etc.
2.3 Genuinely causative get + PP structures I now introduce and analyse the use of get in (12a), where get is a normal manner-unmarked causative position change verb. It is not
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
disability make the reading of the abstract difficult. Di Meola (1994: 59–72) sees the overcoming of obstacles as one of the non-deictic meanings of German kommen, which normally means ‘come’ but also translates some types of unaccusative get + PP structures. Manage and succeed are also hindrance-specialized: #manage to cross/succeed in crossing the street are odd unless hindrances like disability or traffic are presupposed. It is hard to detect a difference between a get + PP structure in the complement of manage and one which is not (I managed to get the key in the lock v. I got the key in the lock). The 1995 Oxford English Reference Dictionary defines get + PP uses using succeed (‘succeed or cause to succeed in coming or going’). I offer a more explicit analysis of hindrance-get in section 3. Before doing so, it is necessary to acquaint readers with other uses of get + PP structures.
Andrew McIntyre 409
hindrance-specialized. In I almost got out my wallet the scope of almost may include a causing actions, which is impossible with hindrance-get, cf. (6). The use of get in (12a) also fails to display the phenomenon in (10) in that I didn’t get out my wallet suggests not that I was unable to but that I chose not to. I call get in (12a) ‘(genuinely) causative get’. (In this study, ‘causative get’ only refers to the PP complement use relevant to us, not to other causative uses like I got him to sing. Section 3 argues that hindrance-get is not genuinely causative.) (12)
If we replace out (of) in (12a) with in(to), we must replace get with put (unless we want the hindrance reading, as in contexts like (12b)). This is a reflection of a possession constraint which affects causative get (but not put and hindrance and unintentional get), in that the subject must come to possess the object, with ‘possession’ understood as control or access, not just ownership. If I get out my wallet, I do not affect my ownership of it, but I do move it into my hands, where I have more immediate access to and control over it. This is not true of I got the cat outside, where get has the hindrance rather than the causative reading, witness, for example, the almost test. Causative get is not only found with out. Get in the police and get together some money involve bringing the respective objects into a domain where they are accessible to the subject, so they obey the possession constraint, and the constructions do not exhibit hindrance effects. The possession constraint on causative get suggests a connection to the likewise possessive DP complement use: if I get the milk out of the fridge, I in a sense get the milk. That this connection is real is confirmed by other verbs with both a possessive DP complement use and a causative position change use with a possession constraint like that found with causative get. Examples are take (take the key fout of/*intog the lock) and, in some varieties, fetch (witness internet attestations like he fetched it out of his pocket and she fetched it out of her bag), as well as German nehmen ‘take’ and holen ‘fetch’. (Holen, unlike fetch, is in common use, freely allowing constructions like es aus der Tasche holen ‘get it out of one’s pocket’). Verbs like get, take, fetch, nehmen and holen with both a causative PP complement use and a possessive DP complement use all have in common that the latter is agentive, i.e. requires an
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
a. I got out my wallet; I got the milk out of the fridge; I got the washing out of the dryer b. So full of money was my wallet that I couldn’t get it in my pocket; I got the milk in the fridge before it went off
410 The Semantic and Syntactic Decomposition of get act induced by a decision on the part of the subject. Take the book expresses a volitional initiation of manual control of the book, while get the book can express deliberate initiation of possession, such as buying or fetching it. Receive illustrates this phenomenon well. With objects like guests, spiritual influences, receive is agentive, requiring an act of the will, and directional PPs may occur exactly in these cases, cf. (13). German kriegen/ bekommen + PP VPs are never agentive; they have hindrancespecialized and unintentional readings, but not genuinely causative ones. This tallies with the fact that their DP complement uses only express passive receipt.3 a. She decided to receive fguests/Christ/*booksg. b. She received fguests/*booksg into her home; She received Christ into her life.
The above facts initially favour an analysis where the grammar of causative get directly refers to the agentive possessive sense. Earlier versions of this essay therefore treated causative get VPs as resultative constructions based on the agentive monotransitive use. On this view, get/take one’s wallet out is grammatically parallel to pull one’s wallet out; both express causation of motion with the verb root identifying the causing event as getting/taking (in the monotransitive agentive senses of the verbs) or pulling. That only agentive readings of possession verbs are suitable as causing events in these structures should fall out from a general theory of resultatives (cf. I rammed/*fell the door open). Wunderlich (1997) and McIntyre (2004) offer different theories of resultatives from which the matters just discussed could be made to follow. An objection to the resultative analysis is that, context free, (14a) does not seem to entail (14b). (14b), but not (14a), implies that the subject moves some distance toward the object (the same holds of German holen). This does not refute the resultative analysis, for the motion intuition in (14b) seems to be an implicature. In (14b), where ownership is irrelevant to an event denoted by get VPs, the type of possession initiated must be assumed to be temporary access or control. Objects in reach of the subject are liable to be already possessed in this way, so we infer 3 In Max und Moritz, Erster Streich (Wilhelm Busch, 1865) we read . . .Kriegt sie jetzt das Messer her, literally ‘gets she now the knife hither’. The context suggests that (her)kriegen should be glossed with ‘fetch’, suggesting that a (now impossible) causative use of kriegen + PP once existed. This exception proves the rule, for kriegen + DP was formerly used agentively, true to its original meaning ‘obtain by war’.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(13)
Andrew McIntyre 411
that they are out of reach of the subject, who must thus move towards the object to gain control over it. Like other implicatures, this inference can be cancelled. (14b) would be possible if it were known that the knife had been within the subject’s reach. PPs like out of my pocket in (14a) are simply another way of cancelling the motion implicature. (14)
a. I got my knife out of my pocket
b. I got my knife
(15)
a. They didn’t get their clothes off b. The band members get their clothes off on stage to attract publicity
Other problems for the resultative analysis concern cases like (16a). Here the subject gets the object, but by the test in (10), (16a) involves the hindrance, not the causative reading, since we do not change the meaning by replacing didn’t with couldn’t. (16b) shows the opposite problem. Get has the genuinely causative reading, although one cannot get the objects, since they cease to exist as soon as they are out/off (stitches become threads once removed). (16)
a. He didn’t get the children home b. I didn’t get fout the stitches/out the stain/off the rustg because that’s not my job
The resultative analysis is thus empirically suspect, but the parallels between get and take, receive and holen show that it is real in some sense. I respond to this paradox by suggesting that the resultative analysis applied to causative get at an earlier stage (as it does to receive + PP VPs currently, cf. (13)), but that causative get has undergone lexical drift. The possessive constraint, when it holds, is a relic from the period when the resultative analysis was valid. Synchronically, it must be stipulated in some way, since a stipulation-free account of it like the resultative analysis would
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Unfortunately, other data speak against the resultative analysis. In (15), clothing is the theme and the human body is the implicit ground/ reference object of off. In varieties like mine, get in (15a) can only have the hindrance reading: it implies that the subject cannot undress, e.g. due to disability. Other speakers can use (15a) of a situation where the subject decides not to undress, and allow get in clearly agentive contexts like (15b). These speakers have a use of causative get which must be replaced by take in other varieties. The resultative analysis cannot capture varieties which lack a causative use for get off in (15) but use take off in this context.
412 The Semantic and Syntactic Decomposition of get both over- and under-generate with regard to examples like (15) and (16). I approach these facts by assuming that causative get and other causative position change verbs like put are semantically identical, but that their lexical entries stipulate different use conditions, constraints on the contexts where the (relevant senses of) the verbs may appear. The proposal is in (17) and (18). (17)
Semantics for [VP x get y [PPP (z)]] with get in the causative use: kP fkZg kY kX kS CAUSE (ACT(X), BECOME P(Y f,Zg))(S)
For expository clarity (17) and other semantic representations before section 4.3 ignore my belief that the decomposition predicates are syntactic heads. These representations can be thought of as interpretations or LFs of syntactic phrases. I ignore irrelevant matters like tense, focussing on the VP, assuming VP-internal subjects. The syntax in (17) and (18) also ignores finer points about the structure of VP. BECOME is defined more precisely in section 3.4, where it becomes crucial. For now, it suffices to take BECOME as symbolizing that what is in its scope changes from being untrue to true. P is a placeholder for the semantics of the preposition. In particle verbs like get out your wallet the preposition’s internal argument z is absent in syntax (and perhaps in conceptual structure in some cases, hence the braces around z). The causal relationship holds between two events. The causing event is assumed to be some unspecified action(s) of the subject, represented as a one-place function ACT. (18) gives use conditions for causative get. (18a) gives varietyspecific versions of the possessive constraint. ‘Control’ in (18a1) is prototypically manual control, where the subject holds and may manipulate, the object. Speakers accepting (15b) follow (18a1), not (a2), for people removing clothing gain manual control over it, but are not motivated by a desire to use or otherwise interact with it. More
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(18) [VP x get y [PPP (z)]] with get in the causative use is usable if either (a) or (b) holds. a1. (some speakers:) x moves y into a position where x has control over y a2. (other speakers:) x’s primary motivation in moving the object is to bring it into a position where, after the event, x can do something with it (i.e. use it or perform some action involving it). b. y ceases to exist as a result of the event
Andrew McIntyre 413
confirmation for (18) is given in section 4.2, where it is discussed with reference to particles, which would have disrupted the exposition here.4 (18b) is motivated by the causative interpretation of get in (16b). I do not know how this use of get constitutes a natural class with uses of get obeying the possession constraint (including its variants in (18a)), but there is evidence that the two uses are somehow related. The get data in (16b) have parallels in (19) with take ((19b, c) are varietyspecific). Recall that take normally resembles causative get in imposing a possession constraint (take the ball fout of/*intog the box), so the disappearing theme use is somehow connected to the possession constraint. (19)
3 HINDRANCE-get AS AN INCHOATIVE OF HAVE I now analyse hindrance-get in detail. I suspect that many linguists would at first glance surmise that (20a) is a causativization of (b) (cf. Hale & Keyser 1993: 86f ). (20)
a. John got the lion in the cage.
b. The lion got in the cage.
Doubts about an analysis where (20a) differs from (b) in the presence of a causal relation emerge from the discussion of (6)–(10). There it was shown that hindrance-get VPs do not denote causing events and are not agentive, properties which do not hold of genuinely causative get and of other causative verbs. A causal analysis for hindrance-get must give way to 4 (18a2) is attested elsewhere as a direction of semantic specialization. It is found with some uses of before. I sat before the piano entails intent to do something with the piano (e.g. play it, clean it, inspect it), while I sat in front of the piano does not force this entailment (cf. The sofa was taken, so I sat fin front of/*beforeg the piano while watching the film. (18a2) also distinguishes take up from pick up. One can take up a pen or stone if one intends to do something with it (e.g. use it, look at it), but not in the context of streetcleaning work, where only pick up is usable. That (18a2) constrains both take up and causative get in some varieties may suggest principled connections between (18a2) and possession, but (18a2) is an idiomatic property of take up not found with other uses of take, for (ia) entails no intended use of or interaction with the sword. A final note on the bigger context of (18a2) is that it should not be confused with functional specializations like that in (ic), which entails intended canonical use of the sword, while (ib) need not entail this, but (in varieties constrained by (18a2)) does entail intent to do something with it.
(i)
a. He ftook out/pulled outg the sword in order to clean the sheath b. He got out the sword fto look at it/*to clean the sheathg c. He fbrandished/ pulled/ drewg the sword f*to look at it/*to clean the sheathg
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
a. The doctor didn’t take out the stitches b. she couldn’t take out the stain [www.cs.washington.edu/ homes/notkin/jstories/node1.html] c. You can take off the rust [www.automotivehelper.com/ topic24462.htm]
414 The Semantic and Syntactic Decomposition of get a representation which asserts that the subject is the initiator of the result state without referring to causing events. Perhaps unexpectedly, it turns out that the approach in (21) gives us this, since the use of have in question holds the subject responsible for having brought about the state (an example of this use is He had the lion in the cage). I argue that hindrance specialization should not be stipulated but follows from reasoning based on the use of the BECOME operator, correctly defined, in certain contexts. (21)
The BECOME HAVE theory: Hindrance-get VPs are nothing more than inchoatives (embeddings under BECOME) of states expressed overtly by a certain type of have + PP structure.
I firstly indicate why it would even be desirable to argue for (21). German kriegen and bekommen are cognate with neither get nor each other, but display many of the same uses as transitive get, cf. (22). (Bekommen matches kriegen in all uses.) The result states of all these sentences can be expressed by replacing kriegen/get with haben/have. The possessive use in (22a) and the unintentional PP use in (22b) are straightforwardly analysable as inchoatives of the have variants. Beside these uses, we find uses which (i) are hindrance-specialized and (ii) hold the subject responsible for having brought about the result. We find that uses with these properties occur under the same conditions in both languages. AP complement uses like (22d) seem always to be hindrance-specialized in both languages (cf. She didn’t get her hair straight ¼ she couldn’t . . . , and analogously in German). The participial use in (22c) is hindrance-specialized in both languages under the same conditions. If the subject is unconscious while being bandaged (and hence not responsible for the result), there is no hindrance specialization. The expressions are hindrance-specialized, e.g. if they denote bandaging one’s own arm (I didn’t (¼ couldn’t) get my left arm bandaged because my right hand was injured).5 5 Two complications with participial structures like (22c): Firstly, some scholars see the nonresponsibility reading of the German structure in (22c) as a passive based on dative structures like ich verband ihm den Arm (lit. ‘I bandaged himdat the arm’). English lacks a comparable ditransitive structure (*I bandaged him his arm), but get is still possible in (22c), showing that the passive analysis of the German in (22c) is redundant. I show elsewhere that German kriegen-subjects can paraphrase datives because they are both arguments of the same HAVE relation. Secondly, beside the hindrance reading, the English in (22c) has a genuinely causative reading ‘get someone to bandage one’s arm’ which is impossible with kriegen. This difference between get and kriegen may be systematic since there are other causative-agentive uses of get which are impossible with kriegen, namely agentive get + DP structures (I tried to get the book) and causative get + PP structures.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
3.1 Initial plausibility arguments for the BECOME HAVE theory
Andrew McIntyre 415
(22)
a. b. c. d. e.
Briefe kriegen ¼ get letters (in sense ‘receive’, not ‘fetch’) einen Kugel in den Arm kriegen ¼ get a bullet in one’s arm den Arm verbunden kriegen ¼ get one’s arm bandaged die Haare glatt kriegen ¼ get one’s hair straight den Lo¨wen in den Ka¨fig kriegen ¼ get the lion in the cage
3.2 The responsibility requirement I firstly discuss the responsibility requirement, the fact that subjects of hindrance VPs are responsible for having initiated the situation: I got the lion into the cage is untrue if I am an inactive beneficiary of a voluntary return of the lion into its cage or of actions performed by people not acting at my behest. The responsibility requirement also occurs with some have + PP structures. (23a) is untrue if Cuthbert did nothing to induce the attacker to flee (e.g. if he was tied up and the attacker fled because the police arrived). (23b) entails that Clive had (asked someone to) put the goods in the warehouse. However, not all have + PP structures exhibit the responsibility requirement, cf. (23c). (23)
a. Cuthbert has the attacker on the run b. Clive has the stolen goods in a warehouse c. The box has sand in it
[responsibility] [responsibility] [unintentional]
My approach to these matters is encapsulated in (24). (24)
a. Semantics for [VP x get y [PPP (z)]] with get in the hindrance reading: kP fkZg kY kX kS BECOME HAVEresp(X, P(Yf,Zg))(S) resp HAVE is provisionally defined as meaning what responsibility have in (23a, b) means, and defined more exactly in (32). b. Semantics for [VP x get y [PPP (z)]] with get in the unintentional reading: kP fkZg kY kX kS BECOME HAVEunint(X, P(Yf,Zg))(S)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The coexistence in three historically unrelated verbs of (i) uses which are plausibly inchoatives of have, and (ii) hindrance-specialized uses, is surely not arbitrary. The question is not whether the uses are connected, but how. My proposal is that both types of uses of get are inchoatives of have and that differences them are due to different uses of have. To uphold this proposal, I must show how it captures hindrance specialization and the concomitant requirement that the subject be interpreted as responsible for having brought about the result. I begin my attempt at this now.
416 The Semantic and Syntactic Decomposition of get unint
is provisionally defined as meaning what unintentional have in (23c) means.
HAVE
(25)
a. John got/had a nail in my tyre. [responsibility/*unintentional] b. Johni got/had a nail in hisi tyre. [responsibility/unintentional]
Secondly, get behaves analogously to have with regard to Harley’s (1998) observation that a reflexive in the complement of have forces a responsibility reading. (26a) attributes the location of the paint to the subject’s carelessness or deliberate self-decoration. In (26b), the subject could be the passive victim of someone else’s act. In (26c) the sergeant may be an inconvenienced by enemy spies (if the camp is his camp), while (26d) forces the responsibility reading where the sergeant had
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(24), a more precise statement of the BECOME HAVE theory in (21), claims that the hindrance and unintentional uses of get are inchoatives respectively of the responsibility and unintentional uses of have. Although it is not crucial to my purposes, I assume that HAVEresp and HAVEunint are two context-specific manifestations of one highly underspecified HAVE relation. If this were not so, the fact noted in section 3.1 that get, kriegen, have and haben all have these readings would be arbitrary. The ensuing sections elaborate on (24a) by defining HAVEresp explicitly and by showing how (24a) captures hindrance specialization. ((24b) is discussed only insofar as it helps us to understand the hindrance reading. It is not directly relevant to the main argument in this study since unintentional get does not combine with particles.) Extra support for the idea in (24) that the responsibility requirement in hindrance-get is inherited from a predicate overtly expressible by have comes from striking parallels between have and get with respect to the conditions under which the responsibility and unintentional readings are found. Firstly, many studies (e.g. Belvin 1993, 1996; Belvin & den Dikken 1997; De´chaine et al. 1994; den Dikken 1997; Harley 1995, 1998; Ritter & Rosen 1997) note that unintentional have (¼ ‘experiencer have’ in these works) requires coindexation between the subject of have and an item in its complement, unlike responsibility-have (¼ ‘causative have’ in these works). (25) shows that this applies to get as well as have. (25a) holds John responsible for the puncture, while (b), where there is coindexation, need not. Belvin & den Dikken (1997: 167) note that covert coindexation suffices for the unintentional reading. Replacing his with the in (25b) allows the unintentional reading provided the tyre is understood as John’s tyre.
Andrew McIntyre 417
stationed himself in the camp, as opposed to an unintentional reading (e.g. he is emprisoned in the camp). (26)
a. hei had/got paint on himselfi b. hei had/got paint on himi c. The sergeant had/got spies in the camp responsibility] d. The sergeant had/got himself in the camp *unintentional]
[responsibility] [unintentional] [unintentional/ [responsibility/
(27)
a. Smith had/got Jones into his team (*due to circumstances beyond Smith’s control) b. Smith had/got Jones in his team (due to circumstances beyond Smith’s control)
3.3 The semantics of HAVEresp (24a) decomposes hindrance-get VPs using HAVEresp, which was said to mean whatever the responsibility-have with PP complements (henceforth: haveresp) means. I now analyse haveresp and equivalently HAVEresp. The studies on responsibility-have cited above (25) treat infinitival complements and ignore have + PP structures, so I now conduct my own analysis of these. Haveresp + PP VPs are states, witness the use of the simple present in contexts like (28), where habitual, generic or historic narrative readings compatible with non-states are unavailable. (28)
Egbert has the stolen car on the street now
6 The tie-in between directional prepositions and responsibility recalls De´chaine et al. (1994), who derive responsibility have by incorporating a specifically directional preposition into a copula, but the account suffers from the lack of a plausible connection between the directionality of the preposition and responsibility semantics. The incompatibility of unintentional have/get with small clauses headed by inherently directional PPs also recalls the claim (e.g. Belvin 1993, Harley 1998, Ritter & Rosen 1998) that unintentional have requires stative complements, while responsibility have may have eventive complements. I know of no satisfying account for this generalization.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Thirdly, responsibility readings with both have and get are forced in the presence of exclusively directional prepositions like into and onto. With both get and have, (27a) portrays Smith as a selector or captain responsible for Jones’ inclusion in the team, while in (27b) Smith could be a player in the team who is not responsible for Jones’ inclusion. I cannot explain (27),6 but it is a further case where the responsibility readings with get as with have are sensitive to the same principles, as expected under the BECOME HAVE theory.
418 The Semantic and Syntactic Decomposition of get Rough equivalence between (29a) and (29b) should not tempt us to view haveresp + PP VPs as some type of elliptical structure involving the perfect auxiliary have. Have in (a) has the syntax of a lexical verb, not that of the auxiliary in (b), witness, e.g. the relative position of have and now, and the fact that the tag question for (a) would be doesn’t she while that for (b) would be hasn’t she. (29)
a. She now has my car in John’s garage b. She has now put my car in John’s garage
(30)
a. When Egbert heard the police siren, he had the stolen car on the street b. Grandma has her wrestling opponent on the floor c. Egbert didn’t have the car on the street
Although subjects of haveresp must have performed actions aimed at bringing about the small clause state before the interval occupied by the haveresp state, these actions are not presuppositional, i.e. need not have occurred if the VP is negated, as in (30c). The fact that haveresp + PP structures hold an entity responsible for bringing about a situation, yet are states and do not express causing events, may seem strange, but has a precedent in (31). (31)
She is responsible for this situation; This situation is due to him
Putting the above observations together, we arrive at the definition in (32) of haveresp (and thus of the HAVEresp relation with which I decompose hindrance-get VPs). (32)
Interpretation of [VP x have [SC y [PPP (z)]]] with have in the responsibility reading: kP fkZg kY kX HAVEresp(X, [P(Yf,Zg)]) This is true during interval I if (i) the state denoted by the small clause is true at I and (ii) x had performed actions before I which caused this state to come about.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The state expressed by haveresp + PP VPs and the state expressed by the complement small clause hold during exactly the same interval. Actions performed by the subject before the start of the interval in which the embedded state holds are not included in the interval occupied by either state. Hence, (30a) is true only if the car was on the street when the siren sounded. It is false if Egbert was still moving the car towards the street at that point. The subject may do something to sustain the small clause state during the time that it holds, cf. (30b), but need not, cf. (30a).
Andrew McIntyre 419
3.4 BECOME and the Aktionsart of hindrance-get (33) (cf. (24) and (32)) is the semantics for hindrance-get VPs which I am proposing. (34) introduces the definition for BECOME, based on Bierwisch’s (2004) revision of that in Dowty (1979: 141). Bierwisch and Dowty may be right to treat (34c) as an implicature. I state it explicitly since its consequences are important here. (33)
a. Semantics for [VP x get y [PPP (z)]] with get in the hindrance reading: kP fkZg kY kX kS BECOME HAVEresp(X, P(Y f,Zg))(S) b. kP fkZg kY kX HAVEresp(X, [P(Y,Z)]) is true during interval K if (i) the state [P(Y f,Zg)] is true at K and (ii) x had performed actions before K which caused this state to come about.
(34)
[BECOME /] is true at interval I iff a. there is an interval K containing the final bound of I with / implied to be true at K.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(32) is provisional. I will not describe uses of HAVE/have other than those manifesting themselves in responsibility have + PP structures, which I see as part of the meaning of hindrance-get VPs crucial to my main argument. Belvin (1993, 1996), Harley (1998) and Ritter & Rosen (1997) and others may be right in assuming an underspecified semantics covering most uses of have. If this were wrong, the observation in section 3.1 that several unrelated verbs have similar ranges of meanings would be mysterious. However, the literature known to me has not brought forth an underspecified meaning for have which predicts exactly the generalisations captured by (32). Theories claiming that have has next to no semantic content (e.g. Ritter & Rosen 1997) cannot prevent I had the lion in the cage from meaning ‘I saw that the lion was in the cage’. The same overgeneration problem attends the claim that have-sentences decompose semantically and, in some theories, syntactically, into a metaphorically construed location of the entity or situation expressed by have’s complement with respect to have’s subject (Benveniste 1966; Belvin 1993;1996; Belvin & den Dikken 1997; De´chaine et al. 1994; den Dikken 1997; Freeze 1992; Harley 1998). This may be right, but I know no suitably predictive variant of this proposal. I therefore leave (32) as a purely descriptive generalization, trusting that successful attempts at more general meanings covering more uses of have and further decompositions of it will be able to be brought together with the central aims of this study without spoiling its main message.
420 The Semantic and Syntactic Decomposition of get b. there is an interval J containing the initial bound of I with :/ presupposed to hold at J. c. there is no interval I# such that I# is included in I and conditions (a) and (b) hold for I#.
(35)
My mobile rang twice while I was getting the cart to the top of the mountain
7 An open (apparently undiscussed) question is how BECOME interacts with PPs. My earlier dealings with motion sentences eschewed BECOME in favour of Jackendoff ’s (1990) GO function relating entities to paths. I found the claim of the BECOME analysis that motion sentences like (ia) assert the inchoation of states like (ib, c) problematic in view of hostility of the states to directional prepositions found in motion sentences.
(i)
a. Cuthbert got (the books) finto/to/ontog the truck b. Cuthbert had the books fin/at/on/*into/*to/*ontog the truck c. The books were fin/at/on/*into/*to/*ontog the truck
The BECOME analysis can be upheld if we assume that directional PPs differ from locational PPs only in that the former must be in the same clause (or subevent) as BECOME. I cannot elaborate on this here. This is a gap in my analysis, but the alternative, the GO analysis, now seems untenable to me. Its non-recognition of result states precludes it from handling restitutive scope (e.g. Stechow 1996) with re- and back in cases like when I left the country for the first time, I freentered/got back ing after a week. Using GO for PPs and BECOME for APs cannot explain the existence of verbs like get taking both types of complements (I got to the shops/I got sick).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The definitions capture the Aktionsart properties of hindranceget VPs. (33b) reflects the finding of section 3.3 that the HAVEresp relation is temporally co-extensive with the small clause state, i.e. the state of a theme at a location. Thus, the only parts of the meaning of hindrance-get VPs relevant to Aktionsart are BECOME and a target state. (34c) ensures that a BECOME transition does not hold until the negation of the small clause proposition ceases to hold. This matches the results of tests like (5) and (6), which showed that hindrance-get VPs are achievements, i.e. are not yet assertable if the theme is known to be moving towards the goal or the subject is acting towards realizing the result state. The term ‘achievement’ is sometimes associated with punctuality, but (35) shows that hindrance-get VPs need not be punctual. (34) allows BECOME transitions to be non-punctual if the initial and result states are separated by an interval where the speaker cannot decide whether the result state holds. Such truth value gaps give a plausible account for cases like (35), where the goal region (here: at the top of the mountain) has fuzzy boundaries, so that the theme must pass through a state where the theme is neither at the goal but nor not at it.7
Andrew McIntyre 421
3.5 Deriving hindrance specialization and the trying presupposition I now show how the semantics for hindrance-get in (33) captures hindrance specialization and the trying presupposition, the presupposition that the subject of hindrance VPs must have tried to bring about the result. Recall from section 2.1 that this presupposition manifests itself in the intuitive equivalence of (36a) to I couldn’t get the screw in the wall. The trying presupposition coupled with the non-occurrence of the result trivially suggests that the subject cannot attain it. (36)
The trying presupposition seems to be related to presuppositions triggered in (36b, c). (b) presupposes that I had been moving towards the shop, and (c) that I started reading the book. The examples in (36) are all achievements and all express situations which, judged by world knowledge, are ‘(inherently) subeventive’, i.e. are intrinsic final parts of larger events also comprising characteristic leadup actions (cf. Pin˜o´n’s 1997 view that achievements are boundaries of larger situations). (36b, c) are subeventive in this sense because one cannot simply reach a location without prior motion, or finish a book without prior reading. Hindrance-get VPs like (36a) are subeventive in the following sense. Hindrance VPs always show the responsibility requirement (captured in my theory by HAVEresp), but do not actually denote actions of the subject which lead to the goal state. Yet the subject cannot be responsible for this state without having performed such actions, i.e. without having tried to attain it. Given this notion of sub-eventiveness, it is easy to see why, if we reformulate (36) without negation, the subject is inferred to have acted with a view to inserting the screw in (a), to have been moving prior to reaching the goal in (b) and to have been reading in the book in (c). What requires explaining is why these inferences are retained in the negative variants in (36), i.e. why they are presuppositions. I put this down to the following reasoning. Speakers can easily avoid these inferences by replacing the VPs in (36) with accomplishment VPs (e.g. with put/insert in (a), with go/drive in (b) and by omitting the aspectual verb in (c)), so that the leadup actions would have fallen under the scope of negation. The choice to bypass these expressions in favour of the relatively marked subeventive ones is therefore plausibly a sign that the leadup actions are to be excluded from negation.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
a. I didn’t get the screw in the wall b. I didn’t freach/get to/arrive at/make it tog the shop c. I didn’t finish (reading) the book
422 The Semantic and Syntactic Decomposition of get
4 Get WITH VERB PARTICLES
4.1 Basic facts about particles We can now begin explaining the effects of get on particle position. I firstly review some facts about particles. Most English particle verbs (alias ‘phrasal verbs’, ‘verb-particle combinations’) allow either objectparticle order or particle-object order: (37)
I carried/took/pushed fout/in/away/downg the box fout/in/ away/downg.
Particle verbs should not be confounded with so-called ‘prepositional verbs’ like (38), where the post-verbal DP is complement of the preposition (and semantically its ground/reference object). Prepositional verbs trivially disallow V-DP-P order, since English has
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
We turn now to hindrance specialization, the intuition that the result of hindrance-get is hard to attain. My account for this begins with the observation in section 2.1 that hindrance-get VPs are non-agentive. While the BECOME HAVEresp representation presupposes actions on the part of the subject aimed at bringing about the result, as just noted, it has no causing event or agentive component explicitly attributing the attainment of the result to actions performed by the subject. In choosing such a representation, and in bypassing normal causativeagentive verbs, the speaker downplays the importance of these actions as the cause of the result. A natural conclusion which can be drawn from the failure to credit the attainment of the result to the subject’s actions is that results with hindrance-get are perceived as non-automatic, i.e. are not guaranteed to eventuate just because the subject tries to attain them. By its very nature, a non-automatic result occurs partly due to propitious circumstances such as luck, external help or lack of resistance. One can try to bring about a non-automatic result, but one cannot decide to. Non-automatic results are by definition hard to achieve, which is the intuition we set out to explain. I do not claim that all hindrance-specialized verbs are to be dealt with in the manner just sketched (though I will show elsewhere the approach extends to unaccusative get and German non-deictic kommen). It is likely that something else must be said about hindrance-specialized verbs with infinitival complements like manage. However, it seems reasonable to conclude that hindrance-specialization can be inferred from the BECOME HAVE representation. This is fortunate given the observation in section 3.1 that any other approach would be arbitrary.
Andrew McIntyre 423 PREpositions
rather than POSTpositions. Compare this with particle verbs like (37), where the post-verbal DP is a theme or figure, whose final/initial location is indicated by the particle, often with respect to an implicit reference object. Thus, the particle verb take in the box entails motion of the box into something, while the prepositional verb fall in the box expresses motion into a stationary box.
(38)
go fing the house f*ing; fall foffg a cliff f*offg; look fthroughg the glass f*throughg
(39) a. *we carried fthrough/inside/upwards/aroundg the materials b. She got/hurt/hit f*backg her enemy fbackg c. She got fbackg her book fbackg. We should also exclude back from the discussion. (39b) involves monotransitive verbs (e.g. get in the sense ‘inflict harm on’) and reciprocal back. This cannot precede an object because it is an adverbial, not a particle, cf. post-PP uses like punch him in the face back. Radford (1997: 444-8) and Koizumi (1993: 125) noted that restitutive back in (39c) shows greater flexibility than normal particles in double object constructions. It can precede objects in have/want/need back the book, though these verbs otherwise disallow particle-object order (cf. (51)). One can receive back a book, although no normal particle combines with receive (*receive in guests).9 When discussing particle position, we must use examples where objects are not weak pronouns, which insist on object-particle order: I took f*outg it foutg. This is due to general information-structural principles, and/or clitic-like properties of weak pronouns (Dehe´ 2002). 8 One could redefine ‘particle’ to include the few non-prepositions that can separate verbs and objects (let slipV a chance, let goV the rope, set freeA the captives, cut shortA the meeting), but these are irrelevant to get. The only superficially non-prepositional item relevant to us is home in take home the books, but home distributes like a PP in other contexts like coordination (they went [home or to work]), use in the complement of way (the way home/down/into the house) and locative inversion (home/into the house/down ran the children). 9 *Receive in DP obeys a ban on particle verbs with Latinate stems (Fraser 1976) which, though not exceptionless (partition off DP), seems real, witness ‘Germanicizing’ truncations like sum(*marise) up and (*con)fess up ‘own up’.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
‘Particle’ is shorthand for ‘complementless preposition which may occur before or after a direct object’.8 The latter clause is needed because some intransitive prepositions shun the pre-object position. This applies to through and some morphologically complex intransitive prepositions (exceptions: away, aside), cf. (39a). When studying particle verbs with get, we will ignore prepositional elements like those in (39a), since these independently resist particle-object order.
424 The Semantic and Syntactic Decomposition of get One should also be wary of objects consisting of more than a determiner and an article; ‘heavy’ DPs may appear clause finally due to an extraneous phenomenon known as heavy-NP-shift. (40) shows why heaviness should be controlled for. (40)
a. I threw [the article] in the fire v. *I threw in the fire [the article]. b. I threw in the fire [the article about dative-ablative syncretism in phonetically null case infixation in English adjectives].
While certain particle verbs with get allow particle-object order (get out your wallet), (41) gives cases where get (unlike the verbs given in brackets with each example) disallows particle-object order in varieties like mine. Some speakers allow particle-object order in some cases in (41), but all reject particle-object order in at least some of these cases. I offer an account of speaker variation shortly. Prior to that, all judgments are based on varieties like mine. (41)
a. I wish I could get f*offg this straitjacket foffg. [take] b. I can’t get f*ong the lid fong, maybe it doesn’t go with this jar. [screw] c. The lock was rusty, but I eventually got f*ing the key fing. [push] d. How can we get f*downg the cat fdowng? It wants to stay on the roof. [lure] e. You’d better get f*offg that letter foffg soon. [send] f. I’m so sick I can’t get f*downg my food fdowng. [hold] g. We eventually got f*outg the fire foutg [put] h. We got f*homeg the children fhomeg by six. [take] i. I can’t get f*upg the blind fupg; it’s stuck [pull] j. I can’t get f*ong the heater fong; I think it’s blown a fuse. [turn, switch]
We can dismiss an account of (41) which says that, e.g. get on the lid in (41b) is avoided because it is homophonous with the prepositional verb meaning ‘climb onto the lid’. Apart from doubts about whether weird readings involving climbing onto a lid or, for example, climbing down food in (41f) would really block the desired readings, such an explanation fails in cases like (41g, h) where the particle has no DP complement use (out would require of insertion: get out of the fire).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
4.2 Particle-object order with get
Andrew McIntyre 425
The vast literature on particle verbs only discusses data like (41) twice. It does not generalize about get. Pesetsky (1992: 250, 304) stars *Her remarks really got down Bill and admits to being unable to explain why this is worse than other psych particle verb structures (?It pissed off Bill). (The contrast is ignored in the published version, 1995: 284.) Harley & Noyer (1998) note the contrast he got out a gun vs. *he got out the drunk sailor without trying to explain it. I now establish the generalizations in (42). (42)
a. Causative get allows both particle-object and object-particle order. b. Hindrance-get disallows particle-object order.10
(43)
a. I got together fsome money/some friends/*an estranged couple/ *a machineg b. She got out fher wallet/the milk/a shotgung c. I got down the details12
Since speakers differ in when they accept get-particle-object order, I solicited judgments and comments on the get sentences in (44) and the corresponding object-particle variants. Eleven informants (five Australian, four American and two British) judged all sentences, and other informants were consulted on a less complete set of data. (44)
a. [The cat is on the roof:] How can we get down the cat? b. [The lock was rusty, but] I eventually got in the key.
10 The account in section 4.4 would also block particle-object order with unintentional get, but unintentional get does not seem to combine with particles, perhaps because of its aversion to specifically directional prepositions. 11 Had there been an English verb krig with the same set of senses as kriegen, I surmise that this verb would always have disallowed particle-object order, since kriegen + PP has hindrance-specialized and unintentional readings, but no genuinely causative use. If krig had existed, this essay would have been much shorter, as there would have been no need to distinguish causative from hindrance readings. 12 (43c) is not an idiom. Combinations like take down, write down, scribble down, note down show that the particle has a semi-productive sense ‘onto paper’.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In (43) particle-object order is good if the structure fulfils the possessive constraint on causative get noted in section 2.3. The acceptable variants in (43) are not hindrance-specialized, for negation of the sentences does not suggest that the subject was unable to attain the result (recall (10)), and the hindrance-specialized German verbs kriegen and bekommen cannot accurately translate the acceptable sentences.11 If we put the particles after the objects, the unacceptable variants become acceptable, but have the hindrance reading.
426 The Semantic and Syntactic Decomposition of get
In my variety, only (44j) and (l) are acceptable. Some informants remarked that (44a), (44b) and (44k) are legitimate only if they express climbing down a cat, into a key or up a flag. This is irrelevant here because these readings reflect a parsing the V-P-DP strings as prepositional verbs with unaccusative get rather than as particle verbs. Speaker variation may also stem from different versions of the possession constraint on causative get, described in the different use conditions in (18a). (44a), (44d) and (44m) enjoyed acceptance by five informants. These speakers follow the use condition in (18a1), for the subject gains (e.g. manual) control over the object. Two of these speakers also accept (44i), but not if down is replaced with up. This is understandable: one momentarily holds a blind after pulling it down, but not after putting it up, and blinds are fixed to window frames, so speakers may differ on whether the subject’s holding the blind constitutes control over it. The varieties of several informants and myself are less permissive particle-object order with get because they have the strict use condition on causative get given in (45) (from (18a2)). (45)
The subject’s primary motivation in moving the object is to bring it into a position where, after the event, the subject can do something with it (i.e. use it or perform some action involving it).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
c. [Said by someone trying to put a lid on a jar:] I can’t get on the lid d. [The prisoner was handcuffed:] Eventually, he saw a way to get off the handcuffs. e. [The boss said] You’d better get off that letter soon f. [Maria has trouble swallowing:] She can’t get down her food. g. We eventually got out the fire h. We got home the children i. [A blind often gets stuck, requiring one to pull on it, untangle the cord etc.:] After fiddling around with it for five minutes, I finally managed to get down the blind. j. [Person talking to neighbours making loud noises:] We’ll get in the police k. [The natives were protesting about the conquerors flying the flag of their home country, and tried to prevent them from doing this:] The soldiers didn’t get up the flag. l. I got out the washing. m. [People were being held captive in a building:] The police got out the hostages.
Andrew McIntyre 427
(46)
The dentist got out my tooth; I got out the screws
We also find particle-object order in (47), where get fulfils the use condition in (18b) that causative get is possible with disappearing themes (cf. the remarks on (16b), (19)). (47)
a. He spilt wine on the carpet but didn’t get out the stain b. I didn’t get off the rust before painting the metal c. The doctor didn’t get out the stitches [cf. Harley 2004: 266]
I thus conclude that particle-object order with a given sentence is possible when get has the causative reading. Differences between speakers on whether particle-object order is possible in a given case
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
This constraint (call it the purpose constraint) is fulfilled in the acceptable sentences in (43) and (44j, l), which are good in all varieties, but not in (44a, d, i, m), which are good only in varieties with the control constraint just discussed. As the parenthetic disjunction in (45) suggests, purpose constraint varieties allow causative get (and hence particle-object order) in contexts where either the object fulfils some function or the subject performs actions involving it. Neither part of this disjunction subsumes the other. Get in the police involves using the police for some purpose, but the subject need not perform further actions, while get out a knife is suitable if the subject intends to look at or clean the knife without using it. The purpose constraint is relevant to get out the cork or get off the lid, which are unacceptable to me, but are internet-attested. If I remove a lid or cork from a bottle, I end up with manual control over it, but intend to interact with the bottle, not the lid/cork. Compare this with *get on the lid/cork, which is bad to all informants, since the subject neither intends to do something with the lid/cork, nor gains manual control of the lid/cork. A final example: In varieties with the purpose constraint, (46) does not suggest removal of the objects from their respective canonical interiors (i.e. mouths, wood), although we might have expected the objects to prime this interpretation strongly (as indeed they do if the particle is placed after the object, allowing the hindrance reading, which has no possessive constraint or purpose constraint). What (46) does suggest in varieties with the purpose constraint is that the objects are removed from storage containers (drawers, boxes, pockets etc.). The extraction of teeth or screws from their canonical interiors does not fulfil the purpose constraint (since in this case the objects cease to fulfil any kind of function), while removing them from storage containers is likely to be motivated by the intention to do something with the objects.
428 The Semantic and Syntactic Decomposition of get result from differences in the conditions on the use of causative get noted in section 2.3. I now move towards an account for why particle placement should be sensitive to the semantics of get in this fashion.
4.3 Syntactic decomposition and the lexical entry for get
Vbecome is a light verb taking a small clause complement. It is interpreted as in (34), the only additional point to note being that / is the proposition expressed by the small clause. Vcause takes a VP as complement and a DP as specifier, and asserts (i) the existence of an event which is the cause of the event named by the inner VP and (ii) that the entity denoted by the specifier of Vcause is the agent of the causing event. Taking (i) and (ii) together, we can say that Vcause introduces the semantic material symbolized by CAUSE and ACT in (17). Pylkka¨nen (2002) argues that separate heads may perform the tasks named in (i) and (ii), but not in English. Vhave is a polysemous morpheme with the same set of functions as English non-auxiliary have. It takes a DP as specifier and a small clause or DP complement. Since only the responsibility reading found with PP complements defined in (32) is relevant to my main argument, my account could equally well be formulated in terms of a morpheme Vhave.resp with the semantics in (32). Although I cannot yet offer a precise semantics for the broader Vhave, I will use this symbol in the belief that some variant of an underspecified semantics for have (and hence Vhave) must be correct, as noted in section 3.3. I take Vhave to be of the category V for simplicity. It would not harm my proposal if it is a preposition, as in Harley (1995). Vbecome, Vcause and Vhave are the only light verbs used here. These (and others irrelevant here) can by hypothesis combine in any
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
My account of particle placement with get appeals to the idea that semantic decomposition predicates are syntactic heads. For expository convenience, the semantic representations given above ignored this idea. I now integrate it into the picture. The approach below draws partly on distributed morphology (DM, e.g. Marantz 1997; Harley & Noyer 1998). A crucial tenet is that VPs are projections not of open class lexical verbs (get, sing etc) but of closed class items roughly corresponding to semantic primitives in lexicalist theories like Wunderlich 1997 or Jackendoff 1990. I now describe the closed-class items (light verbs) used here.
Andrew McIntyre 429
permutation observing selection restrictions. To ensure the interpretability of the combination of light verbs, a VP must contain an audible item functioning as a spellout of the light verbs. The spellout is by hypothesis inserted post-syntactically. I now describe how get fits into this scheme of things. (48) gives part of get ’s lexical entry (in DM terms, its encyclopaedia entry). I take get to be a spellout of V.become The multiplicity of uses of get and the fact that get cannot be used in all cases where BECOME is present is handled by conditions stipulating its syntactic context. Excerpt from the lexical entry for get: a. /get/ spells out Vbecome. It is usable in the following contexts: b. [VPVbecome [SC DP XP]] (e.g. I got inside/drunk/working/shot) c. [VPVbecome [VP DP Vhave [sc DP PP/AP/PartprogP/ PartperfP]]](e.g. I got him inside/drunk/working/shot) d. [VP Vbecome [VP DP Vhave DP]] (e.g. I got a letter, on nonagentive reading)) e. [VPDPi Vcause[VPVbecome [VP PROi Vhave DP]](e.g. I got a book, on agentive reading) f. [VPDPa Vcause[VPVbecome [SC DPb PP]] under the following conditions, cf. (18): i. (some varieties:) DPa comes to have control of DPb as a result of the event (other varieties:) DPa moves DPb in order to bring it into a position where, after the event, DPa can use or do something with DPb. ii. DPb ceases to exist as a result of the event
(48)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
I see get as a spellout of Vbecome since it has unaccusative uses where it only spells out Vbecome (I got sick) and since Vbecome is present in all its uses. Approaches not seeing get’s uses as elaborations of BECOME would require several homonymous gets. (48) does not mention the ditransitive use noted in connection with (3), since this can reasonably be predicted from general principles. (48) stipulates the existence of other readings of get because this seems necessary in an empirically responsible theory. For instance, opting to capture causative uses of get in (e, f) by productive causativization processes rather than by lexical stipulation would prevent us from capturing its use conditions and would wrongly predict that all spellouts of Vbecome could be productively causativized (*Beer became me a bad linguist; *fear went him pale; *it flew him into a rage; *the plague fell him sick).
430 The Semantic and Syntactic Decomposition of get
4.4 An explanation for particle position with get (49) is the syntax I assume for particle verbs with get. (49a) involves hindrance-get and (b) causative get. Vbecome is spelt out as get. In (a) get incorporates into the silent Vcause head. (49)
a. Ann got f*ong the lid fong. [TP Anni [T PAST] [VP Vbecome [VP ti Vhave [SC [DP the lid] [P(P)on]]]]] b. (I gave her the bill and) Ann got foutg her wallet foutg. [TP Anni [T PAST] [VP ti Vcause[VP Vbecome [SC [DP her wallet] [P(P)out]]]]]
Various linguists (e.g. Haider 1997; Haegeman & Gue´ron 1999: 258; Harley & Noyer 1998; Radford 1997: 374) assume that English particle-object order reduces to the ability of particles to incorporate into a verb (e.g. Vbecome in (49b)). The now complex verb moves to a verbal head position to the left of the direct object (e.g. Vcause in (49b)) by the short verb movement used in all shell theories. Another approach (Dehe´ 2002; Johnson 1991; Koizumi 1993; Olsen 2000) initially generates English particle verbs as morphological objects, forming a complex verb which may optionally be treated as a V° for the purposes of short verb movement. The incorporation approach and the morphological one share the idea that the particle-object sequence involves a V° node dominating the particle and a verbal element. The idea that syntactic incorporation produces a configuration which is subject to peculiarly morphological principles (and idiosyncrasies) was
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Further work on get would doubtless occasion emendations to (48). It ignores non-semantic information and reflects some marginal decisions. (48b) may need to list the possible categories of small clause heads given that get a nun cannot mean ‘become a nun’, unless a theory of blocking can predict this given the existence of become or the readings in (48d) and (48e). (48c) and (48d) could be a single use specifying only that the lower VP is headed by Vhave. (I separated (48c) and (48d) and listed the SC head categories because VP-headed SCs as in I had him sing are impossible with get.) Uncertainties notwithstanding, the framework is flexible enough to express subtle constraints on uses of spellouts. The framework is no less flexible than lexicalist theories, for it has a lexicon (encyclopaedia) constraining the contexts where a ‘verb’ (or, in this theory, a phonological item spelling out a light verb) may be used. We now apply the system to particle verbs with get.
Andrew McIntyre 431
(50)
Vhave cannot incorporate anything, including particles. (Alternative formulation: Vhave cannot head a morphologically complex X°.)
(50) predicts that VPs decomposing with Vhave with audible verbs other than get should disallow particle-object order. We find this with have, want and need, cf. (51a, b). Various writers (Dowty 1979: 244–50, 269– 71; Fodor & Lepore 1998; Harley 2003; Larson et al. 1997) note that transitive want and need are equivalent to want/need to have. Positing a silent Vhave makes sense of the scope of the italicized adverbials in (51c). (51d) applies the scope test to show that a silent have is present in particle verbs formed with these verbs. Generative semanticists worked with ‘have deletion’ based on want/need to have. An update of this is to assume a silent Vhave in the complement of want/need (as in Harley 2003 and Larson et al. 1997). (51)
a. *He had off his jacket; *I soon had out the splinter; *He was having on the people b. *I need off the light; *I need out the fire; *The doctor wants out the stitches. c. I need your apartment until next week; Last week Bill wanted your car yesterday d. I need that light on until tomorrow; At noon you wanted the heater on tonight
(50) may not follow from general principles of grammar, but the formation of complex heads, and thus incorporation/head movement, is known to be fraught with idiosyncrasy. We see this in [vPV] compounds (downsize, downplay, download v. *downbring, *downpull, *downtear; overturn v. *overfall, *overtopple; offload v. *offcast, *offtake) and in morphologically complex prepositions (into, onto but *underto, *byto).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
endorsed by Baker (1988: 68ff). The approaches differ in whether this relationship arises before or during the syntactic derivation (see Koopman 1995: 147, fn. 15 for the suggestion that base generation of particle verbs as morphological objects is not incompatible with assuming that a PP is projected, as in the incorporation approach). My account for the blockage on particle-object order in (49a) starts with (50). A standard constraint on head movement, the head movement constraint (e.g. Baker 1988), says that heads may move only to the next highest head position. Thus, if the particle cannot head-move to Vhave, as (50) claims, then it will be unable to move any higher than Vhave, and thus stays after the object. Upholding this account requires us to confirm the empirical and theoretical status of (50).
432 The Semantic and Syntactic Decomposition of get
13 If (50) is a morphological idiosyncrasy, (50) need not hold of all languages with Vhave. My account merely predicts that items decomposing with HAVE in a language should all behave alike in either allowing or disallowing incorporation. While German shows no word order facts speaking for head movement of particles, Zeller (2001: chap. 6) suggests that affixation of German particle verbs requires incorporation of the particle into the verb. A preliminary check suggests that particle verbs with kriegen, bekommen and haben resist affixation (*heraushabbar ‘have-out-able’, *herausbekommbar ‘get-out-able’, *heraushaber ‘haver-outer’, *herauskrieger ‘getter-outer’, cf. herausnehmbar ‘take-outable’, herausnehmer ‘taker-outer’), but it is hard to control for extraneous factors here.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(50) may fall under a larger generalization, though this is hard to prove. If what I notate as Vhave is (or decomposes with) a preposition (Belvin & den Dikken 1997; De´chaine et al. 1994; den Dikken 1995, 1997; Freeze 1992; Harley 1998, 2003; Richards 2001), then (50) would fall under a larger generalization about the incomplete productivity of complex preposition formation just noted.13 Many analyses of English particle verbs (see Dehe´ et al. 2002 for an overview) eschew incorporation in the form assumed here. Den Dikken (1995) denies overt particle incorporation, Svenonius (1996) moves particles to a position below the verb and Olsen (2000) rejects incorporation altogether. It is left to proponents of particle verb analyses inimical to mine to check whether these analyses can handle get. For reasons given in section 4.5, I maintain that all principled accounts of these facts will converge on syntactic decomposition. Here I merely address the fact that the structure [V°V P] used in incorporation analyses of particle-object order in English violates the right-hand head rule (RHR). Den Dikken (1995: 88f ) assumes that the RHR rules out an incorporation analysis like mine. His citing of Williams (1981) in this connection is unfortunate, because Williams did not intend the RHR to be exceptionless (see p. 249f ). I do not find Williams’ exceptions compelling, but the use of double inflection patterns like sisters in laws, hangers-ons, passers-bys (attestable under www.google.com) confirms the existence of left-headed X° items. If sister were not the head of sister in law, we would not expect it to host inflection. If sister in law were not a morphological object, one wonders why inflection appears on its outer edge. Thus, the RHR can be violated in English, so it does not argue against the incorporation analysis for particle verbs. (49) can be altered in various ways without detriment to my proposal. Firstly, SC is often used as an abbreviation for an endocentric structure with a head mediating agreement or predication (e.g. den Dikken 1995: 25f, Svenonius 1996). If this is right, ‘incorporation of particles’ should be understood as shorthand for ‘incorporation of functional elements heading an SC which have themselves incorporated a particle’. Secondly, if English objects move to some specifier or
Andrew McIntyre 433
adjunction position to receive case, the proposal will not be affected, since all object movement theories move the object to the left to a position just to the right of the spellout position of the verb (with or without an incorporated particle). My own assumption is that Vcause and Vhave, but not Vbecome, can assign case to the specifier of their complement. The assignment of case to specifiers of small clauses is seen transparently in With [SCthe psycho in gaol], we could relax. In (49b), I assume that the object adjoins to the lower VP to receive Case from Vcause, roughly as in Johnson (1991) and Kratzer (1996, section 2).
4.5 Alternative accounts Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
My account of particle verbs with get counts as an argument for syntactic decomposition only if theories doing without it are problematic. I try to show this now. Construction grammar (e.g. Goldberg 1995) would see verbparticle constructions as idioms equipped with semantic representations and open slots for verbs, objects and particles. As far as I can see, this theory could only capture the get-facts by positing a verb + particle + object template which has a causative semantics, with which noncausative verbs like have and hindrance-get would be semantically incompatible, and a semantically broader verb + object + particle template (or several homophonous verb + object + particle templates) with which they are compatible. Nothing predicts that object-particle order should be possible in more cases than the reverse, while theories using head movement predict this automatically, since head movement can be blocked under certain circumstances. While construction grammar is able to capture semantic constraints on constructions not yet captured in other theories, the very flexibility which permits this actually drains the theory of the power needed to block hindrance-get from the particle(-first) construction. The construction’s causative semantics does not actually suffice for this, since the responsibility requirement threatens to make hindrance-get close enough to a causative for it to be a viable candidate for insertion in the construction, and since constructions are assumed to be polysemous, so that the particle(-first) construction could a priori have a hindrancespecialized responsibility reading alongside the causative one. What the theory cannot do is refer to the presence of HAVE in the semantics of hindrance-get in describing the blockage on particle-object order. Apart from forcing learners to rely on negative evidence, admitting this negative generalization as a possible constraint on constructions begs the question as to whether there are any principled limits on the
434 The Semantic and Syntactic Decomposition of get
(52)
a. [VP [V ] fPrtg NP fPrtg] b. [vP NP v [VP get [SC NP Prt]]] c. [vP NP v [VP NP get Prt]]
Another alternative to syntactic decomposition would be one where causative verbs project a complex predicate as in (52c) while hindranceget and have take small clauses as in (52b). The particle-object order could by hypothesis only be derived from (52c) (e.g. because particleobject order requires re-analysis which requires adjacency), but not from (52b). This is the best alternative to syntactic decomposition, since a verb’s meaning can uncontroversially affect complement selection, which can itself affect word order. This approach would divest itself of the arbitrariness tainting the other approaches, provided it could be supplemented with predictions on the differing conditions under which a verb will project a small clause or a complex predicate. The latter proviso seems unfulfillable. Take (53), used of inserting a cork in a bottle. In both cases, the particle is a goal and a predicate on the object, so one cannot attribute the putative difference in argument projection to a difference in the particle’s semantic role. One might assume that the complex predicate analysis is motivated when the verb and particle share an argument, but if there is argument sharing, one wonders why the cork in (53) is less an argument of get than of jam, stick or wedge, which do not select the object: *I wanted to seal the bottle, so I jammed/wedged/stuck the cork. Moreover, I see argument sharing with resultatives and particles as questionable. McIntyre (2004) notes that obligatorily transitive verbs need not retain their argument structure in such constructions, witness examples like I lit *(a cigar) v. I lit up (a cigar). None of the particle constructions supporting the anti-sharing claim shows any tendency towards resisting particle-object order. If the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
number and nature of semantic constraints to which any construction could be sensitive. Language learners would need to assess the compatibility of particle-object order with an indeterminate number of semantic verb classes, e.g. inherently punctual acts (spit out the tablet, shoot down the can or knock over the chair), verbs of sustained motion causation (drag/pull/push in the car), etc. If we assume a non-abstract syntax like (52a) but see the VP as a projection of a lexical verb, the arbitrariness problems just discussed remain. The verb would need to carry a syntactically visible tag indicating some or all of its semantic content, and the sensitivity of the generation of particle-object order to this tag would raise tricky questions about how and why word order should relate to verb semantics.
Andrew McIntyre 435
particle’s role and the verb-selected status of the object are not enough to predict the distinction between small clauses and complex predicates, then this distinction yields no principled explanation of the particle order facts with get. (53)
get f*ing the cork fing v. jam/wedge/put/force/stick/bung/whack/ twist/hammer fing the cork fing
5 SUMMARY OF THE MAIN PROPERTIES OF THE READINGS OF get + PP STRUCTURES A. Hindrance-get: I got the nail in (the wall) It disallows particle-object order. It can be translated by German kriegen and bekommen. It is hindrance-specialized, i.e. suggests that the result is hard to attain (2.1, 3.5). It yields achievement VPs (2.1, 3.4). It does not denote agentive acts (2.1) although it does show the responsibility requirement, i.e. characterizes the subject as responsible for achieving the result (3.2). Possibility operators do not effect the interpretation of negated hindrance-get VPs (I didn’t (¼couldn’t) get the nail in the wall), due to a presupposition that the subject tried to realize the result (2.1, 3.5). It is argued to be nothing more than an inchoative of the responsibility reading of have (e.g. I had the nail in the wall).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
With the failure of this initially appealing alternative, we are left with a choice between syntactic lexical decomposition, which leads one to expect interactions between verb meaning with word order like that seen with get, and other theories, where such interactions require mysterious stipulations. To be sure, many linguists may prefer stipulations, however arbitrary, to a syntax replete with invisible, hard-to-detect heads, however closely they match the invisible primitives located in abstract lexical/semantic/conceptual structures in other theories. Or perhaps there is an account I have overlooked which can explain the data naturally without syntactic decomposition. This study will have been worthwhile if it inspires the formulation of viable alternative accounts, or even if it does no more than to encourage linguists of diverse persuasions to include the data it discusses in their deliberations about how syntax and semantics interact.
436 The Semantic and Syntactic Decomposition of get It is decomposed semantically as in (33) (and, assuming syntactic decomposition, as in (49a)). B. (Genuinely) causative get: I got my wallet out (of my pocket)
C. Unintentional get: The camera got dust in it It does not seem to combine with verb particles. The subject of this use of get is not in any way responsible for the result (2.2). It disallows exclusively directional prepositions like into and mostly requires co-indexation between the subject and something in its complement (2.2, 3.2). It is an inchoative of unintentional have structures like the camera has dust in it (3.2, 3.3).
Acknowledgements This article benefited greatly from extensive and fair comments by the journal’s anonymous readers, and from discussions with Willi Geuder, Heidi Harley, Jaume Mateu, Irene Rapp and the audience at the 2002 Sinn und Bedeutung conference. Credit for my initial reflections on get goes to the Australian schoolteachers who enjoined me to avoid get when I write because it is ‘boring’.
ANDREW McINTYRE Institut fu¨r Anglistik Universita¨t Leipzig Beethovenstr. 15 04107 Leipzig e-mail:
[email protected]
Received: 20.08.03 Final version received: 08.12.04 Advance Access publication: 14.03.05
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The term causative is used here only of uses with PP/particle complements, not of other causative uses of get. It allows particle-object order or object-particle order. It does not translate with German kriegen and bekommen. It is genuinely causative and agentive (2.3). In most uses, it requires that the object come to be possessed or manually controlled by the subject (probably because it was originally used in resultative contructions based on agentive DP complement uses, cf. parallels between get/take out the key and get/take the key). (17) lists the details of the possession constraint for different speakers. Its semantic representation is as in (18) (and, assuming syntactic decomposition, as in (49b)).
Andrew McIntyre 437
REFERENCES Baker, M. (1988) Incorporation. University Press. Chicago. Baker, M. (1997) ‘Thematic roles and syntactic structure’. In L. Haegeman (ed.), Elements of Grammar. Kluwer. Dordrecht. Belvin, R. (1993) ‘The two causative haves are the two possessive haves’. MIT Working Papers in Linguistics 20:19–34.
Belvin, R. & den Dikken, M. (1997) ‘There, happens, to, be, have’. Lingua 101:151–183. Benveniste, E. (1966) Proble`mes de Linguistique Ge´ne´rale. Gallimard. Paris. Bierwisch, M. (2004) BECOME and its Presuppositions. Ms. Humboldt-Universita¨t. Berlin. Cruse, D. A. (1973) ‘Some thoughts on agentivity’. Journal of Linguistics 9:11–23. De´chaine, R.-M., Hoekstra, T., & Rooryck, J. (1994) ‘Augmented and non-augmented have’. In L. Nash and G. Tsoulas (eds.) Proceedings of Langues et Grammaire 1:85–101. Dehe´, N. (2002), Particle Verbs in English. Benjamins. Amsterdam. Dehe´, N., Jackendoff, R. McIntyre, A., & Urban, S. (eds.), (2002) Verb particle explorations. Mouton de Gruyter. Berlin. den Dikken, M., (1995), Particles. Oxford University Press. Oxford. den Dikken, M. (1997) ‘The syntax of possession and the verb ‘‘have’’’. Lingua 101:129–150. Di Meola, C. (1994) Kommen und Gehen. Niemeyer. Tu¨bingen.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Belvin, R. (1996), Inside Events. Unpublished Ph.D. thesis, University of Southern California.
Dowty, D. (1979) Word meaning and Montague Grammar. Reidel. Dordrecht. Fodor, J. A. & Lepore, E. (1998) ‘The emptiness of the lexicon’. Linguistic Inquiry 29:269–288. Freeze, R. (1992) ‘Existentials and other locatives’. Language 68:553–596 Givon, T. & Yang, L. (1994), ‘The rise of the English get-passive’. In B. Fox and P. J. Hopper, Voice. Benjamins. Amsterdam. 119–149. Goldberg, A. (1995) Constructions. University of Chicago Press. Chicago. Haegeman, L. (1985) ‘The get-passive and Burzio’s generalisation’. Lingua 66:53–77. Haegeman, L. & Gue´ron, J. (1999) English Grammar. Blackwell. Oxford Haider, H. (1997) ‘Precedence among predicates’. Journal of Comparative Germanic Linguistics 1:3–41. Hale, K. & Keyser, S. (1993) ‘On argument structure and the lexical expression of syntactic relations’. In K. Hale & S. Keyser (eds), The View from Building 20. MIT Press. Cambridge, MA. 53–110. Harley, H. (1995) Subjects, events and licensing. Unpublished Ph.D. Thesis. MIT. Harley, H. (1998) ‘You’re having me on! Aspects of have ’. In J. Gue´ron & A. Zribi-Hertz (eds), La Grammaire de la Possession. Publedix. Nanetter. 195– 226. Harley, H. (2004) ‘Wanting, having and getting’. Linguistic Inquiry 35:255– 267. Harley, H. & Noyer, R. (1998) ‘Mixed nominalisations, short verb movement and object shift in English’. NELS 28:143–157.
438 The Semantic and Syntactic Decomposition of get Pesetsky, D. (1995) Zero Syntax. MIT Press. Cambridge, MA. Pin˜o´n, C. (1997) ‘Achievements in an event semantics’. A. Lawson and E. Cho (eds), Proceedings of Semantics and Linguistic Theory 7. CLC Publications, Cornell University. Ithaca, NY . Pylkka¨nen, L. (2002) Introducing Arguments. Unpublished Ph.D. thesis. MIT. Radford, A. (1997) Syntactic Theory and the Structure of English. Cambridge University Press. Cambridge. Richards, N. (2001) ‘An idiomatic argument for lexical decomposition’. Linguistic Inquiry 32:183–192. Ritter, E. & Rosen, S. T. (1997) ‘The function of have ’. Lingua 101:295– 321. Stechow, A. von (1996) ‘The different readings of wieder ‘‘again’’’. Journal of Semantics 13:87–138. Svenonius, P. (1996) ‘The verb-particle alternation in the Scandinavian languages’. Ms. Tromsø. www.hum.uit. no/a/svenonius Tobin, Y. (1994) Aspect in the English verb. Longman. London. Williams, E. (1981) ‘On the notions ‘‘Lexically Related’’ and ‘‘Head of a Word’’’. Linguistic Inquiry 12:245– 74. Wunderlich, D. (1997) ‘CAUSE and the structure of verbs’. Linguistic Inquiry 28:27–68. Zeller, J. (2001) Particle Verbs and Local Domains. John Benjamins. Amsterdam.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Jackendoff, R. (1990) Semantic Structures. MIT Press. Cambridge, MA. Johnson, K. (1991) ‘Object Positions’. Natural Language and Linguistic Theory 9:577–636. Kayne, R. (1993) ‘Toward a modular theory of auxiliary selection’. Studia Linguistica 47:3–31. Kimball, J. (1973) ‘Get’. In J. Kimball (ed.), Syntax and Semantics 2. Academic Press. New York. 205–15. Koizumi, M. (1993) ‘Object agreement phrases and the split VP hypothesis’. MIT Working Papers in Linguistics 18:99–148. Koopman, H. (1995) ‘On verbs that fail to undergo V-second’. Linguistic Inquiry 26:139–163. Kratzer, A. (1996) ‘Severing the external argument from the verb’. In J. Rooryck & L. Zaring (eds), Phrase Structure and the Lexicon. Kluwer. Dordrecht. 109–137. Larson, R., den Dikken, M., & Ludlow, P. (1997) ‘Intensional transitive verbs and abstract clausal complementation’. http://semlab2.sbs.sunysb.edu/Users/ rlarson/larson-papers.html Marantz, A. (1997) ‘No escape from syntax’. University of Pennsylvania Working Papers in Linguistics 4:201– 225. McIntyre, A. (2004) ‘Event paths, conflation, argument structure and VP shells’. Linguistics 42:523–571. Olsen, S. (2000) ‘Against Incorporation’. Linguistische Arbeitsberichte 74:149– 172. Pesetsky, D. (1992) Zero Syntax. (Version 2.0) Ms. MIT.
Journal of Semantics 22: 439
Editor’s Note We wish to express our gratitude to the following colleagues who are not members of the Editorial Board of the Journal but have kindly helped us in the past year with the refereeing of papers submitted:
Ó Oxford University Press. All rights reserved
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Maria Aloni Jay David Atlas Giosue` Baggio Josef Bayer Sigrid Beck Walter Daelemans Tim Fernando Martin Hackl Jack Hoeksema Sabine Iatridou Gerhard Ja¨ger Jacques Jayez Stefan Kaufmann Laura Kallmeyer Ji-yung Kim Utpal Lahiri Claudia Maienborn Reinhard Muskens Diarmuid O’Seaghdha Massimo Poesio Susan Rothstein Kjell-Johan Sæbø Roger Schwarzschild Scott Schwenter Ben Shaer Junko Shimoyama Martin van den Berg Jan van Eijck Frank Veltman Yoad Winter
