Modals in Haitian Creole: A Lexical Semantic Investigation Tamina Stephenson Massachusetts Institute of Technology Several modal expressions in Haitian Creole are classified according to their major semantic characteristics, in particular, their force (necessity or possibility), their available interpretations (epistemic, deontic, and so on), and their scope properties. The general system of modals is then connected to cross-linguistic phenomena.
1. Introduction This is a preliminary descriptive overview of modal expressions in Haitian Creole, with particular focus on their semantic properties. I look at several lexical items, mostly verbs, which clearly have modal meanings, and try to classify them according to their force (necessity or possibility), their available interpretations (epistemic, deontic, and so on), and their scopal interactions with negation. I also discuss their basic syntactic properties, but this is mostly restricted to aspects that bear an obvious relation to the semantics. This paper is organized as follows: Section 2 reviews the basic semantics of modals that I will assume, as well as the parameters of variation along which I will classify individual items. Section 3 contains a short subsection on each of the modals, classifying them and discussing their properties individually. The particular items I look at include fo ke, gen pou, ka/kapab, mèt, and dwe, which each correspond roughly to English must, have to, might, or can; and the verbs bezwen (‘need’) and vle (‘want’). Finally, Section 4 concludes with some brief general observations on how Haitian modals fit in with common patterns observed in other languages.
2. Semantic Assumptions I will assume a simplified version of the semantics for modals developed by Kratzer (1977, 1981) and ensuing tradition. The basic idea, which comes out of work in logic, is that words like might, can, must, and have to (along with their equivalents in other languages) are quantifiers over possible worlds. In this way they are parallel to expressions that quantify over individuals, such as every or some. For example, it is standardly claimed that the meaning of (1) can be represented as in Figure I. (1)
Every student came to the party.
1
Tamina Stephenson
Haitian Modals
Figure I The set of students is a subset of the set of individuals that came to the party.
came to the party Students
Similarly, we can express the meaning of sentence (2) as shown in Figure II.1 Note that although words like have to and can seem to be part of the verb phrase, I will follow common practice in assuming that they combine semantically with an entire clause. (2)
John has to wash the dishes. = has to [John washes the dishes]
Figure II John washes the dishes John fulfills his obligations
The set of worlds where John fulfills his obligations is a subset of the set of worlds where he washes the dishes.
The expressions every in (1) and have to in (2) both contribute a requirement for a subset relation, and they are called universal quantifiers by virtue of this property. Modals that have the force of universal quantification, such as have to, are called necessity modals. It is also standardly claimed that (3) can be represented as in Figure III: (3)
Some students came to the party.
Figure III came to the party
Students
The set of students has a non-null intersection with the set of individuals that came to the party.
Similarly, we can express the meaning of (4) as in Figure IV. (4)
John can play outside. = can [John plays outside]
[expressing permission]
1 To simplify matters, I am abstracting away from the distinction between the modal base and the ordering source, which Kratzer argues for based on examples which are not relevant here. Readers familiar with the doubly-relativized system should take this set (e.g., the set of worlds where John fulfills his obligations) to be the set that is obtained after the ordering source has been applied over the modal base, i.e., the set that is ultimately quantified over.
2
Tamina Stephenson
Haitian Modals
Figure IV John plays outside
John fulfills his obligations
The set of worlds where John fulfills his obligations has a non-null intersection with the set of worlds where he plays outside.
The expressions some in (3) and can in (4) both contribute a requirement for an intersection relation, by virtue of which they are called existential quantifiers. Modals that have the force of existential quantification, such as can, are called possibility modals. The most basic characteristic of any given modal is whether it has the force of necessity or possibility. Necessity modals have the force of universal quantification, requiring a subset relation, whereas possibility modals have the force of existential quantification, requiring non-null intersection. A common notation uses the symbol “ ” to indicate a necessity modal and “◊” to indicate a possibility modal. Another crucial ingredient in the interpretation of a modal is the set of worlds which is used as the basis for comparison – in (2) and (4), this was the set of worlds where John fulfills his obligations. For my purposes we can call this set the modal base.2 Kratzer (1977) and others argue that to a great extent this set is provided by the context. For example, an English sentence of the form x has to P could take as its modal base the set of worlds where x follows the dictates passed down by x’s parents, the laws of the United States, etc., depending on the context. However, particular modal expressions typically carry restrictions on what type of modal base they can have. Certain words such as want and need seem to take more specialized modal bases – something like the set of worlds in which all of the subject’s needs (for need) or desires (for want) are met. Other modal expressions tend to have restrictions that cluster around fairly well-defined areas which have been given labels. A modal (or use of a modal) is called deontic when it takes a modal base determined by something like rules, laws, or obligations. Sentences (2) and (4) were examples of deontic modals. A modal is called epistemic when it’s based on something like the set of facts that are known by someone. In (5) and (6), for example, must and might are epistemic. (5)
[Seeing a woman walking in with John] That must be John’s wife. (He said he was going to bring her to meet me.)
(6)
John might not be home now. (He didn’t answer the phone when I called him.)
2 As I alluded to in note 1, what Kratzer ends up calling the “modal base” is something different, and what I’m calling the modal base ultimately needs to be derived in a more complicated way.
3
Tamina Stephenson
Haitian Modals
Another type of modal that is commonly referred to is an ability modal, e.g., can in John can run fast. In this case the modal base is something like the set of worlds in which John only does things that – in the actual world – he is able to do.3 These three labels – deontic, epistemic, and ability – are the main ones I’ll need for my classification. There’s just one more ingredient that will be important for my description of Haitian modals, which is the way they combine with negation. When a sentence with a modal ( p or ◊p) combines with negation, there are in principle two different ways this could happen, resulting in different meanings. I’ll illustrate with a necessity modal. The first possibility in this case is that the whole modal sentence is negated (~ p); then we say that the modal has scope below negation. This is the meaning we get in English when we negate sentence (2). This is shown in (7), with the meaning illustrated in Figure V: (7)
John doesn’t have to wash the dishes. = NOT [have to (John washes the dishes) ]
Figure V John washes John fulfills his the dishes obligations
=
John fulfills his obligations
The set of worlds where John fulfills his obligations is NOT a subset of the set of worlds where he washes the dishes. = ~ (John washes the dishes) = ◊~ (John washes the dishes)
John doesn’t wash the dishes
Now, we’re saying that the set of worlds where John fulfills his obligations is not a subset of the set of worlds where he washes the dishes, which is equivalent to saying that there is a non-null intersection between the set of worlds where he fulfills his obligations and the set of worlds where he doesn’t wash the dishes, as illustrated in the lower part of Figure V. This is the same meaning we would have gotten if we had first negated the lower proposition and then combined it with a possibility modal, which means that ~ p is logically equivalent to ◊~p. The second possibility is that negation just combines with the lower proposition and then the modal combines with the resulting negative proposition in the normal way ( ~p); then we say that the modal has scope over negation. This is the meaning we get when we combine must with not, as in (8). This time we get the meaning shown in Figure VI. (8)
John must not be home.
3 In all of these cases, it’s crucial that the rules, laws, knowledge, abilities, etc. which determine the modal base are the ones that hold in the actual world, not in the other worlds.
4
Tamina Stephenson
Haitian Modals
Figure VI John isn’t Everything that home we know holds
=
John is home
Everything that we know holds
The set of worlds where everything that we know holds is a subset of the set of worlds where John isn’t home. = ~ (John is home) = ~◊ (John is home)
This time, we’re saying that the set of worlds where everything that we know holds is a subset of the set of worlds where John isn’t home, which is equivalent to saying that the set of worlds where John is home is disjoint from (does not overlap with) the set of worlds where everything we know holds. This is the same meaning we would have gotten if we had first combined the lower proposition with a possibility modal and then negated the whole thing, which means that ~p is logically equivalent to ~◊p. Although there are always two ways that a modal could combine with negation in principle, typically both are not freely available. Specific modal expressions carry restrictions on how they can take scope with respect to negation. These restrictions thus constitute one more way in which a modal can be classified. I have discussed three major properties of modal expressions: their force (necessity or possibility), their allowable readings (deontic, epistemic, and so on), and their scope properties with respect to negation. Now I turn to Section 3, where I classify several Haitian modals in terms of these properties.
3. Classification of Modal Expressions in Haitian Creole 3.1 fo ke / fòk / fò The item fo ke, also realized as fòk or fò, has the force of necessity. These three phonological variations are interchangeable, although the shorter fòk and fò are more common. Syntactically fo ke takes a complete sentence as its complement, its closest English equivalents in that respect being pseudo-logical locutions such as it is necessary that…. (The second element, ke, is a complementizer.) The construction is illustrated in (9). (9)
Fo ke Jan lave asyèt yo. FO C J. wash dish the-pl. ‘John has to wash the dishes’ = p
(10)
Fòk Jan lave asyèt yo. FO+C J. wash dish the-pl. = (9) 5
Tamina Stephenson
(11)
Fò Jan FO+C J. = (9)
Haitian Modals
lave asyèt yo. wash dish the-pl.
It is not possible for fo ke (or its variants) to appear between the subject and the VP, as shown in (12): (12) cf. (13)
* Jan/Li/Mwen fo ke / fò /fòk lave asyèt yo. J. /3sg/1sg fo+C wash dish the-pl. Jan lave asyèt yo. J. wash dish the-pl. ‘John washed the dishes’4
Fo ke can have both deontic and epistemic interpretations, as illustrated in (14)-(15) below. (14)
Fòk Jan lave asyèt yo. FO+C J. wash dish the-pl. ‘John has to wash the dishes’ = deontic p
[repeated from (9)]
(15)
[Standing in front of John’s house and seeing that all the lights are on.] Fò
Jan la J. there ‘John must be home.’ = epistemic p
FO+C
It can also take what has sometimes been called a “dispositional reading” (see, e.g., Kratzer 1997), as in (16). (16)
Fòk m etènye FO+C 1sg sneeze ‘I have to sneeze’ = “disp” p
The negative particle pa can appear in its normal position in the lower clause, as illustrated in (17)-(18). As might be expected, the modal takes scope over negation.
4 The reason (13) has a past interpretation relates to independent facts about the Haitian tense/aspect system. The simple present of non-stative verbs cannot be used to describe ongoing actions, for which the progressive marker ap must be used. A past marker, te, also exists, but is optional in many cases. As a result, non-stative sentences with no overt tense or aspect marking get a past interpretation by default.
6
Tamina Stephenson
Haitian Modals
(17)
a.
Fò mwen pa wè Mari. FO+C 1sg NEG see M. ‘I must not (I’m not allowed to) see Mary.’ = deontic ~ p * ‘I don’t have to see Mary.’ * ~ deontic p
cf.
b.
Mwen pa wè Mari. 1sg NEG see M. ‘I don’t see Mary’ / ‘I didn’t see Mary’
(18)
a.
[Standing in front of John’s house and seeing that all the windows are dark.] Fò
Jan pa la. J. NEG there ‘John must not be home.’ = epistemic ~ p * ‘John might not be (is not necessarily) home’ * ~ epistemic p
FO+C
cf.
b.
Jan pa la. J. NEG there ‘John isn’t home’
There is no way that negation can combine with fo ke other than appearing in the lower clause and taking low scope, as in (17)-(18) above. For example, it’s impossible to put pa before fo ke as in (19), between the two elements fo and ke as in (20), or between fo ke and the subject as in (21). (19)
* Pa NEG
fo ke / fòk / fò FO+C
Jan lave asyèt yo. J. wash dish the-pl
(20)
* Fo pa ke Jan lave asyèt yo. FO NEG C J. wash dish the-pl
(21)
* Fo ke / Fòk / Fò pa Jan lave asyèt yo. FO+C NEG J. wash dish the-pl Summary – fo ke: This is a general purpose necessity modal which takes a full sentence as its syntactic and semantic argument. Negation with fo ke is confined to the complement clause, with the modal taking high scope.
7
Tamina Stephenson
Haitian Modals
3.2 gen pou Gen pou is also a necessity modal. Its two components gen and pou have independent uses in the language: gen is the verb of possession (‘have’) and pou is used as a preposition (‘for’) and as a complementizer. In this sense, then, gen pou is analogous to English have to. Unlike fo ke, the expression gen pou looks more like the kind of modal verb familiar from many other languages, since it can appear after the subject and before the VP, as in (22). However, the subject can also be “doubled” – appearing before the second verb as well, as in (23). (22)
Li gen pou wè Mari. 3sg have C see M. ≈ ‘He has to see Mary’ = deontic? p
(23)
Li gen pou li 3sg have C 3sg = (22)
wè Mari. see M.
Gen pou has a more restricted use than fo ke. It cannot take an epistemic reading at all. For example, (24) is not appropriate in the context given. (24)
[Standing in front of John’s house and seeing that all the lights are on.] # Jan gen pou (li) la J. have C 3sg there attempted: ‘John must be home’ * epistemic p
It’s actually not even clear whether gen pou can take a run-of-the-mill deontic reading, which is why I put a question mark on “deontic” to the right of (22). Sabine Iatridou and Kai von Fintel (class notes, spring 2004) have suggested that gen pou is restricted to a reading where the subject is “scheduled” to do something, as is the case with the Greek modal construction using echo, ‘have.’ One speaker did reject (25), which would give support to this claim; however, another speaker consistently accepted examples of the same kind. (25)5
% Li gen pou wè doktè, men li poko gen randevou. 3sg have C see doctor but 3sg not-yet have appointment ≈ ‘He has to see the doctor, but he doesn’t have an appointment yet.’
There may be a dialect difference here, or it could be that these examples don’t distinguish the context well enough to evoke the relevant judgments. For example, it’s possible that the first speaker rejected (25) for some irrelevant reason; or, on the other hand, the second speaker might have been thinking of a 5
I use the “%” sign to indicate that judgments differ among speakers.
8
Tamina Stephenson
Haitian Modals
scenario in which the subject was somehow “scheduled” to see the doctor (e.g., he had a firm plan to do so) despite lacking an appointment. The “scheduledness” of gen pou comes out more clearly with an example like (26): (26)
# M gen pou (m) etènye 1sg have C 1sg sneeze attempted: ‘I have to sneeze’
The speaker who accepted examples like (25) nevertheless rejected (26), saying that it suggested that the subject is supposed to sneeze at a certain time. This suggests that gen pou may indeed be restricted to something like a schedule reading, but it doesn’t prove that a regular deontic reading is not available, since the intended meaning of (26) is not deontic. It would be necessary to work with more speakers and contexts to determine conclusively what readings are available with gen pou, so I’ll leave it open for now. Normally gen pou appears after negation and takes semantic scope below it, as illustrated in (27). With negation appearing after gen pou as in (28), the sentence becomes highly degraded. (27)
M pa gen pou wè Mari. 1sg NEG have C see M. ‘I don’t have to see Mary.’ =~ p * ‘I can’t (must not) see Mary’ * ~p
(28)
M gen pou pa wè Mari. 1sg have C NEG see M. ‘I have to not see Mary.’ ?* ~ p
Negation can, however, appear low at the end of a string of coordinated clauses, as in (29). In this case the semantic scope matches the surface scope.6 (29)
[Context: The doctor has ordered the subject to follow a healthy regimen.] M gen pou manje pi byen epi dòmi 1sg have C eat more good & sleep
pi plis more
epi pa bwè kafe. & NEG drink coffee ‘I have to eat better and sleep more and not drink coffee.’ = deontic?(p & q & ~r)
6
The judgment of (29) came from the speaker who accepted examples like (25).
9
Tamina Stephenson
Haitian Modals
Summary – gen pou: This necessity modal is more restricted than fo ke; it seems to only allow a deontic reading and may be even more restricted than that. Syntactically, it takes a VP as a complement, with possible doubling of the subject. Gen pou normally appears after negation and scopes below it. (When it can be forced to appear before negation, it takes high scope.) 3.3 ka / kapab The verb kapab (with its shortened variant ka) has the force of possibility. The variants ka and kapab are not completely interchangeable, but for present purposes we can treat them as such. The only difference is that the shorter variant ka is impossible in certain phrase-final positions, presumably for prosodic reasons.; apart from that they seem to be realizations of the same item. Ka/kapab takes a VP complement, and can have a wide variety of interpretations, including epistemic, deontic, and ability readings. These are illustrated in (30)-(32). (30)
[Context: A friend didn’t show up for work today, and we’re wondering why.] (M kwe) li ka malad. 1sg believe 3sg KA sick ‘(I think) he might be sick.’ = ◊epistemic p
(31)
[Child has asked for permission to go play outside. Parent responds.] (Wi,) Ou kapab jwe deyò a. yes 2sg KA play outside the-sg. ‘(Yes,) you can play outside.’ = ◊deontic p
(32)
Jan ka kouri vit. J. KA run fast ‘John can run fast’ = ◊ability p
Unlike with gen pou, the subject cannot be “doubled,” i.e., appear overtly as the subject of both the modal and the lower verb. For example, (33) is ungrammatical. (33)
* Ou kapab ou jwe deyò a. 2sg KA 2sg play outside the-sg.
Negation can occur either before or after ka. In both cases the surface scope relation always holds – that is, ka pa corresponds to ◊~ and pa ka corresponds to 10
Tamina Stephenson
Haitian Modals
~◊. When ka has an epistemic reading, both structures are possible, as shown in (34)-(35). (34)
[A co-worker called in sick, but we’re not sure if he’s telling the truth.] Li ka pa malad. 3sg KA NEG sick ‘He might not be sick.’ = ◊epistemic ~p
(35)
[Same context as (34)] Li pa ka malad. (M wèl yè.) 3sg NEG KA sick. 1sg see+3sg yesterday. ‘He can’t be sick. (I saw him yesterday.)’ = ~◊epistemic p
When ka has a deontic or ability reading, however, negation can only appear before ka, with the modal taking low scope. This is illustrated in (36)-(38). (36)
[Child has asked for permission to go play outside. Parent responds.] a.
b.
(37)
(Non,) Ou pa ka jwe deyò a. no 2sg NEG KA play outside the-sg. ‘(No,) you can’t play outside.’ = ~◊deontic p # Ou ka pa jwe deyò a. 2sg KA NEG play outside the-sg. * ◊deontic ~p
a.
Jan pa ka kouri vit. J. NEG KA run fast ‘John can’t run fast’ = ~◊ability p
b.
# Jan ka pa kouri vit. J. KA NEG run fast * ◊ability ~p
In the right context, however, two instances of negation can appear in the same sentence; in this case one goes in each position, as in (38): (38)
M pa ka pa chante. 1sg NEG KA NEG sing ‘I can’t not sing.’ = ~◊ability? ~p
11
Tamina Stephenson
Haitian Modals
Summary – ka/kapab: This is a general purpose possibility modal which takes a VP as complement. Negation can appear either before or after ka/kapab, in each case taking surface scope. Epistemic readings allow both structures, but deontic and possibility readings are normally restricted to the scope ~◊. This results in restrictions on the position of pa with deontic and ability readings. 3.4 mèt Mèt is also a possibility modal, but its use is much more restricted than ka. It can only have a deontic reading, as illustrated in (39)-(41). (39)
[Child has asked for permission to go play outside. Parent responds.] (Wi,) Ou mèt jwe deyò a. yes 2sg MÈT play outside the-sg. ‘(Yes,) you can play outside.’ = ◊deontic p
(40)
[A friend didn’t show up for work today, and we’re wondering why.] # (M kwe) li mèt malad. 1sg believe 3sg MÈT sick * ◊epistemic p
(41)
[We’re talking about our friends’ athletic skills.] # Jan mèt kouri vit. J. MÈT run fast * ◊ability p
Furthermore, the negative particle pa cannot appear with mèt at all: (42)
a.
* Ou pa mèt jwe deyò a. 2sg NEG MÈT play outside the-sg. * ~◊ p, * ◊~ p
b.
* Ou mèt pa jwe deyò a. 2sg MÈT NEG play outside the-sg. * ~◊ p, * ◊~ p
As with ka, subject doubling is impossible with mèt: (43)
* Ou mèt ou jwe deyò a. 2sg MÈT 2sg play outside the-sg.
12
Tamina Stephenson
Haitian Modals
Summary – mèt: Like ka, mèt is a possibility modal which takes a VP as complement. However, it is restricted to a deontic reading and cannot combine with negation at all regardless of position or scope. 3.5 dwe The verb dwe is a necessity modal. In affirmative sentences, its meaning is similar to that of fo ke and it is usually translated as should, must, or have to, or with an adverb such as probably. Examples are given in (44)-(45). (44)
(Si m pa peye enpo mwen), m dwe peye amand. if 1sg NEG pay tax 1sg 1sg DWE pay fine ‘(If I don’t pay my taxes,) I’ll have to pay a fine.’ = deontic p
(45)
[Standing in front of a friend’s house and seeing that all the lights are on.] Li dwe lakay li. 3sg DWE house 3sg ‘He must be home’; ‘he’s probably home’ = epistemic p
Subject-doubling is impossible with dwe: (46)
* Li dwe li lakay li. 3sg DWE 3sg house 3sg
When dwe appears with negation, dwe takes high scope, even though pa appears before it. This is illustrated in (47.a). It is not grammatical for pa to appear after dwe, as illustrated in (47.b). (47)
[Child has asked for permission to go play outside. Parent responds.] a.
b.
(Non,) Ou pa dwe jwe deyò a. no 2sg NEG DWE play outside the-sg. ‘(No,) you can’t play outside.’ = deontic ~p * ‘You don’t have to play outside’ * ~ deontic p * Ou dwe pa jwe deyò a. 2sg DWE NEG play outside the-sg.
The same situation holds for sentences with epistemic readings, such as (48.a-b).
13
Tamina Stephenson
(48)
Haitian Modals
[We’re knocking on a friend’s door, and there’s no answer.] a.
b.
Li pa dwe la. 3sg NEG DWE there ‘He must not be home.’ = epistemic ~p * ‘He might not be home’ * ~ epistemic p * Li dwe pa la. 3sg DWE NEG there
In an example like (48.a) it’s actually a bit tricky to show that dwe pa represents ~ rather than ~ or, equivalently, ◊~. (Recall from Section 2 that ~ and ◊~ are equivalent.) The usual way to make sure that a sentence has the force of necessity rather than possibility is to coordinate it with its counterpart in which the lower clause is negated and show that the resulting sentence is judged as incoherent. For example, the English sentence he might be home, but he might not is perfectly sensible, whereas the sentence he must be home, but he might not is incoherent (if both must and might are epistemic). However, when we try to do this with (48.a), it is acceptable, as indicated below. (Note that I have to use ka in the second conjunct, since dwe by itself has the force of necessity.) (49)
[We’re knocking on a friend’s door, and there’s no answer.] Li dwe pa la, men li ka la. 3sg DWE NEG there but 3sg KA there “He must(?) / might(?) not be home, but he might be.”
The acceptability of (49) might be taken to indicate that epistemic dwe, unlike deontic dwe, takes low scope with respect to negation. This would predict that (49) would mean ‘he might not be home, but he might be,’ which would be perfectly coherent. Another possibility, though, is that dwe is similar to English should, which is considered a necessity modal but (at least on certain readings) leaves a bit of room for doubt. Thus (50) is acceptable. (50)
[We’re at the train station. It’s 5:00. John’s train is scheduled to arrive at 5:30.] John shouldn’t be here yet, but (I guess) he might be.
Therefore, shouldn’t p is acceptable in a situation where it’s very likely but not certain that p is false. However, shouldn’t p becomes unacceptable if p and ~p are known to have equal probability. For example, suppose I am arguing with an officemate about who a certain book belongs to, and we decide to flip a coin for it: if it’s heads, I get to keep the book, and if it’s tails, my officemate gets it. Someone has flipped the coin and is about to uncover it. In this case it would be 14
Tamina Stephenson
Haitian Modals
natural for me to utter (51) but odd to utter (52), unless perhaps I believed that the coin was weighted.7 (51) (52)
I want the book, but it might not be heads. # I want the book, but it shouldn’t be heads / it probably isn’t heads.8
Indeed, in the same situation, (53) is acceptable but (54) is not. (53)
M vle liv la, men li ka pa tèt. 1sg want book the but 3sg KA NEG head ‘I want the book, but it might not be heads.’
(54)
# M vle liv la, men li dwe pa tèt. 1sg want book the but 3sg DWE NEG head ‘I want the book, but it’s probably not heads.’
It appears, then, that dwe is basically a necessity modal but that it’s somewhat weaker than Haitian fo ke or English must or have to. Summary – dwe: Like fo ke, dwe is a necessity modal that allows both deontic and epistemic readings. Its force seems to be somewhat weaker – more like ‘should’ or ‘probably’ than ‘must’ or ‘have to.’ Negation can only appear before dwe, but the modal takes scope over negation – that is, pa dwe represents ~ rather than ~ . It is not possible for negation to appear after dwe (*dwe pa). 3.6 bezwen, ‘need’ Bezwen corresponds closely to English need. It can take either a VP or a CP as complement, as illustrated in (55)-(56). (55)
Ou bezwen ale nan libreri pou sa a 2sg need go to bookstore for demon. ‘You need to go to the bookstore for that.’ = need p
(56)
M bezwen ke li ale nan libreri 1sg need C 3sg go to bookstore ‘I need him to go to the bookstore.’ = need p
The higher subject in (56) is obligatory:
7
Thanks to Martha McGinnis for suggesting this kind of example as a test.
8
(52) is also acceptable with but it probably isn’t on a reading where the speaker is simply being unreasonably pessimistic about the likelihood of winning.
15
Tamina Stephenson
(57)
Haitian Modals
* Bezwen ke li ale nan libreri need C 3sg go to bookstore
Bezwen can also take a DP complement, as in (58). (58)
Plant sa a bezwen anpil dlo. plant demon. the need a lot water ‘This plant needs a lot of water.’
Negation can only appear before bezwen, with bezwen taking low scope. This is illustrated in (59). (59)
M pa bezwen ale nan libreri 1sg NEG need go to bookstore ‘I don’t need to go to the bookstore.’ = ~ need p * ‘I need to NOT go to the bookstore.’ * need ~p Summary – bezwen: This verb means ‘need.’ It can take VPs, CPs, and DPs as arguments. Negation can only appear before bezwen, in which case bezwen takes low scope.
3.7 vle, ‘want’ The verb vle means ‘want,’ which has a meaning of necessity. It can take VP complements, as in (60). (60)
M vle danse. 1sg want dance ‘I want to dance.’ = want p
It can also take what looks like a full clause to its right, as in (61). The complementizer ke can optionally appear after vle in this case, suggesting that the lower clause is finite. (61)
M vle (ke) ou danse. 1sg want C 2sg dance ‘I want you to dance.’ = want p
The structure in (61) not possible, however, if the two subjects are the same, as illustrated in (62). (62)
* M vle (ke) m danse. 1sg want C 1sg dance “I want me to dance.” 16
Tamina Stephenson
Haitian Modals
Vle can also appear with a CP headed by pou, an item roughly equivalent to ‘for.’ In this case, the lower subject can either match or differ from the higher one, but in either case is obligatory. This is illustrated in (63)-( 64). (63)
M vle pou *(m) danse. 1sg want C 1sg dance ‘I want to dance.’
(64)
M vle pou *(li) danse. 1sg want C 3sg dance ‘I want him to dance.’
Like bezwen, vle can also take a DP complement, as in (65). (65)
M vle yon liv. 1sg want a book ‘I want a book.’
If ke is absent, then the only position where negation can appear in a sentence like (60) or (61) is immediately before vle, as illustrated in (66)-(70). (66)
(67) (68)
M pa vle danse. 1sg NEG want dance ‘I don’t want to dance.’ = want ~p * M vle pa danse. 1sg want NEG dance M pa vle ou danse. 1sg NEG want 2sg dance ‘I don’t want you to dance.’ = want ~p
(69)
* M vle pa ou danse. 1sg want NEG 2sg dance
(70)
* M vle ou pa danse. 1sg want 2sg NEG dance
If ke is present, then pa may appear in the lower clause as well as before vle, but not between vle and ke. This is illustrated in (71)-(73). (71)
M pa vle ke ou danse. 1sg NEG want C 2sg dance ‘I don’t want you to dance.’ = want ~p 17
Tamina Stephenson
(72) (73)
Haitian Modals
* M vle pa ke ou danse. 1sg want NEG C 2sg dance M vle ke ou pa danse. 1sg want C 2sg NEG dance ‘I want you to not dance’ = want ~p
Like English want and equivalents in many languages, vle has logical scope over negation even when the surface syntactic scope goes the other way – that is, when we say that x doesn’t want p, we’re not just saying that it is consistent with x’s desires that p not hold, but rather that p is inconsistent with x’s desires. Summary – vle: This verb means ‘want.’ It can take DP, VP, and clausal complements. Negation must normally appear before vle, but the modal takes high scope.
4. General Observations: Let me conclude with two brief observations about the modal expressions I’ve looked at. The first observation relates to a general cross-linguistic tendency for epistemic modals to take higher scope than deontic modals with respect to various operators, such as negation, quantifiers, and other modals. Cinque (1999), for example, proposes that epistemic modals belong to a higher functional projection than deontic modals, thus predicting that an epistemic modal can take scope over a deontic modal within the same clause but not vice versa.9 Von Fintel & Iatridou (2003) claim that quantifiers can take scope over deontic modals but not over epistemic modals. There are also cases where a single modal will scope differently with respect to negation depending on which reading it takes. A well-known example of this is English may, which takes scope over negation in epistemic readings but below negation in deontic readings. This is illustrated in (74)-(75). (74)
There may not be any sign of water on Mars. = ◊epistemic ~p
(75)
Students may not have triple-majors. = ~◊deontic p
In (74), where may has an epistemic reading, may not is understood as ◊~; however, in (75), where may has a deontic reading, may not is understood as ~◊. (Picallo 1990 discusses similar facts about Catalan poder, ‘can/may.’) Haitian ka is similar to English may in only allowing low scope with respect to negation on a deontic reading, although it differs from may in allowing both scopes on an 9
Kai von Fintel (class notes, spring 2004) cites apparent counterexamples to this claim from German, which he attributes to Kratzer (1976, Was “können” und “müssen” bedeuten können müssen, Linguistische Berichte 42).
18
Tamina Stephenson
Haitian Modals
epistemic reading. Also, whereas the surface position of negation with English may is the same either way (may not), the surface position of negation with ka in Haitian changes to match the semantic scope. Thus pa ka always corresponds to ~◊ and ka pa always corresponds to ◊~. Haitian does not always keep such a close match between the interpretation and the surface scope of modals with negation, however, which brings me to my second observation. We have seen that dwe takes scope over negation even though negation normally appears before it – that is, dwe corresponds to but pa dwe corresponds to ~. This is illustrated in (76)-(77), repeated from (47)-(48). (76)
Ou pa dwe jwe deyò a. 2sg NEG DWE play outside the-sg. ‘You can’t play outside.’ = deontic ~p
(77)
Li pa dwe la. 3sg NEG DWE there ‘He must not be home.’ = epistemic ~p
This seems to be an example of the more general phenomenon of “neg-raising” seen in many languages, where in certain contexts the surface position of negation seems to be structurally higher than the position corresponding to its interpretation. Horn (1989) discusses neg-raising in some detail. This effect is common not only with modal verbs, but also with verbs of thinking and believing, as in the English example in (78). (78)
I don’t think John is here. = I think (John is not here)
(78) seems to mean not just that it isn’t the case that I think John is here (which would be consistent with my being completely ignorant of John’s whereabouts), but that in fact I think he isn’t here. As is generally the case with neg-raising, the actual meaning is stronger than the expected meaning would be. The Haitian verb kwe, ‘believe,’ shows a similar effect, as illustrated in (79). (79)
cf. (80)
M pa kwe Jan malad 1sg NEG believe J. sick ‘I don’t think John is sick’ [= it’s my belief that he isn’t sick] = believe (~p) M kwe Jan malad 1sg believe J. sick ‘I believe John is sick’ = believe (p) 19
Tamina Stephenson
Haitian Modals
Thus in examples like (78) and (79), there seems to be negation associated with the lower clause whose source is the upper clause. Even with this kind of negation, dwe still takes high scope. We can see this by putting dwe in the lower clause of a sentence like (79), as in (81). (81)
M pa kwe li dwe malad 1sg NEG believe 3sg DWE sick ≈ ‘I don’t think he could be sick’ / ‘I think it’s unlikely that he’s sick’ = believe ( ~p)
The fact that dwe takes high scope even in cases like (81) suggests that the scope properties of this modal ought to be connected to the more general phenomenon of neg-raising, however this is to be accounted for in the end.
References Cinque, Guglielmo (1999). Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford: Oxford University Press. von Fintel, Kai and Sabine Iatridou (2003). Epistemic Containment. Linguistic Inquiry 34(2): 173–198. Horn, Laurence R. (1989). A Natural History of Negation. Chicago: University of Chicago Press. Kratzer, Angelika (1977). What ‘must’ and ‘can’ must and can mean. Linguistics and Philosophy 1: 337–355. Kratzer, Angelika (1981). The notional category of modality. In H.-J. Eikmeyer and H. Rieser (eds.), Words, Worlds, and Contexts. New Approaches in Word Semantics, Berlin: de Gruyter, pp. 38–74. Picallo, M. Carme (1990). Modal verbs in Catalan. Natural Language and Linguistic Theory 8(2): 285–312.
MIT Dept. of Linguistics and Philosophy 77 Massachusetts Ave., 32-D808 Cambridge, MA 02139 USA
[email protected]
20
