UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

3 Morphological number, semantic4 number and bare nouns

5 Viviane Deprez

6 Institute for Cognitive Sciences, Lyon, Rutgers University, 18 Seminiary Place,

7 New Brunswick, NJ 08903, USA

8 Received 8 January 2004; accepted 8 January 2004

9

10 Abstract

11

12 This paper compares the properties of bare nouns in a non-plural marking French Lexifier Creole

13 to those in plural marking languages in order to better understand the role played by number, both

14 semantic and morphological, in determining the meaning of bare nouns. A parameter is proposed that

15 distinguishes languages according to whether or not the structure of their nominal expressions

16 obligatorily includes a functional projection for Number. The presence of NumP is taken to enforce

17 the realization of a kind into its instances, offering a link between the morpho-structure of bare nouns

18 and their interpretation that obviates the needs for a semantic parameterization.

19 # 2004 Published by Elsevier B.V.20

21 Keywords: Bare nouns; Semantic plural; Morphological plural; Parameterization of plurality22

23 1. Introduction

24 The question of how to properly capture the meanings and distribution of nominal

25 expressions that lack overt determiners has been the topic of intense research for many

26 years, both in the semantic and in the syntactic literature. Yet, it is only recently that cross-

27 linguistic considerations have begun to be taken into account, or more specifically that

28 approaches attempting to account for attested cross-linguistic variations in the meaning of

29 bare nominals have flourished. The broad and influential proposal of Chierchia (1998) has

30 prompted much investigation into the meaning of bare nominals in a variety of languages

31 (Cheng and Sybesma (1999), Dayal (2001), Deprez (1999a, 2001), Longobardi (1999),

32 Munn and Schmitt (2001) among others). The results have challenged some aspects and

33 predictions of his proposals, such as the distribution of bare singular nominals, while

Lingua xxx (2004) xxx–xxx

E-mail address: deprez@rci.rutgers.edu (V. Deprez).

1 0024-3841/$ – see front matter # 2004 Published by Elsevier B.V.

2 doi:10.1016/j.lingua.2004.01.006

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

34 supporting others, such as, the importance of plural morphology in the determination of

35 parametric choices.

36 The present paper is an attempt to bring into this line of research the questions involved

37 in accounting for the distribution and the meanings of bare nominals (BN) in Creole

38 languages. It focuses centrally on a Romance based Creole, Haitian Creole for two reasons:

39 First, it is mostly within Romance at large that comparative concerns have been critically

40 examined, and second, it is mainly with their superstrate that Romance based Creoles

41 seem, at least superficially, to differ most radically in the interpretation and in the

42 distribution of their bare nominals. Most characteristically, while data on Romance

43 languages have brought to the forefront important constraints on the meaning and

44 distribution of BN, questioning in particular the general availability of kind readings

45 and of bare nominals in general, Romance based Creoles illustrate in contrast an intensive

46 use and a rather free distribution of nominal expressions without determiners. For instance,

47 while French is often taken to illustrate one of the most constraining languages with respect

48 to the distribution of bare nominals—with few exceptions it disallows bare nominals in

49 argument positions—French based Creoles on the contrary, allow bare nominals quite

50 freely and extensively with a variety of interpretations (Deprez, 1999a, in press).

51 As is well known, Romance based Creoles are languages with a much ‘poorer’

52 inflectional morphology than their superstrate counterparts. Comparative concerns with

53 Romance based Creoles thus raise interesting questions on the role of morphology in the

54 distribution and interpretation of bare nominals. In Chierchia’s approach, morphological

55 plural is seen as a factor that helps a native speaker determine the basic interpretation of

56 nouns as primary predicates (sets of objects of type <e,t>) or as primary kind terms

57 (individuals of type <e>). On this view, morphology is taken to serve as a trigger for a

58 semantic parameter, the Nominal Mapping Parameter, in the sense that it determines an

59 essentially ‘lexical’ choice N ¼ <e,t> or N ¼ <e>, but morphology does not by itself play a

60 direct role in the interpretation of bare nominals.

61 The present paper explores a different approach to the role of morphology in the

62 distribution and interpretation of bare nominals. In Deprez (1999a, 2001), an alternative

63 parametric approach based on a syntactic parameter is sketched to provide an account of the

64 syntactic and semantic properties of Haitian Creole bare nominals and to better understand

65 the role of morphology in licensing/constraining the meaning of bare nominals. In this

66 approach, morphology plays a direct role in the interpretation of bare nominals at two levels.

67 First, the richness of plural morphology determines whether a syntactic node NumP is

68 obligatorily projected or not in a given language. Second, the presence of NumP in a nominal

69 projection plays a compositional role in determining its interpretation. The present paper

70 confronts this syntactically inspired approach to more extensive cross-linguistic data, it

71 explores its ability to integrate what is known about a variety of non-Creole languages and

72 assesses its consequence for the interpretation of bare nominals in predicative positions.

73 The paper is organized as follows. Section 2 briefly summarizes the findings on bare

74 nominals in HC and introduces new data on the number interpretation of Haitian BN and on

75 BN in predicative positions. Section 3 develops the alternative proposal first sketched in

76 Deprez (1999a, 2001). Section 4 tests the proposal against a wide set of empirical data from

77 Creole as well as non-Creole languages and discusses in detail its consequences. An

78 analysis of bare nominal predicative sentences is proposed that provides evidence for the

2 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

79 syntactic presence of a null functional projection NumP in the predicative use of Haitian

80 bare nominals. The proposal is also shown to provide a straightforward explanation for the

81 restricted use of bare singular nominals in Romance and English predicative and argu-

82 mental constructions.

83 2. Properties of Haitian bare nouns

84 This section summarizes the properties of bare nominals in Haitian Creole1 and points

85 out some problems for Chierchia’s model that have motivated the alternative parametric

86 proposal explored here. It also investigates in greater detail the number interpretation of

87 bare nominals in HC as well as the properties of HC bare nominals in predicative positions.

88 As (1) illustrates, Haitian Creole clearly manifests an unrestricted distribution of bare

89 nouns in argument positions. They can be subjects and objects of verbs, as in (1a), as well

90 as objects of prepositions, as in (1b). Both the existential reading (1a) and the generic

91 reading (1c) are available for HC bare arguments and their distribution characteristically

92 depends on the predicate they associate with. That HC bare nouns also allow for a kind

93 reading is shown in (1d) and (1e), where BN occur with typical kind taking predicates in

94 subject and object positions.

(1) a. Moun koumanse ap pran batonPeople started taking sticks

b. Jan ap danse ak tifiJohn is dancing with (some) girl/s

c. Jouromou pa donnen kalbas

The ‘jouromou’ cannot produce calabashes

d. Elefan ap vin ra

Elephants are/the elephant is becoming rare

e. Edison (te) envante anpoul elektrikEdison invented the light bulb

127 In their existential readings, HC BN have obligatory low scope. They contrast in this

128 respect with singular indefinites with the determiner yon. As illustrated below, indefinite

129 DPs with yon can have scope either above or below negation. Only the second reading is

130 available for bare nouns.

(2) a. Ou pat we yon tach ate a

You did not see one spot on the floor

There is one spot on the floor you did not see

b. Ou pat we tach ate a

You did not see spots on the floor�There are spots on the floor you did not see

1 The data in this section were collected over a number of years both from informant sessions and in the

literature. For their invaluable help with the Haitian data, I would like to thank Wilson Douce, Francois Canal,

Wisly Paul and particularly Emanuel Vedrine.

V. Deprez / Lingua xxx (2004) xxx–xxx 3

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

151 Similarly, HC bare nouns cannot take scope over sentence final time adverbs. As the

152 infelicity of (3a) indicates, indefinites with yon necessarily have wider scope than these

153 adverbs, yielding an absurd interpretation. With bare nouns in contrast, the adverbial has

154 wider scope and the sentence is fine.

(3) a. #Jan te tuiye yon lapen pluzie fwa

John killed a rabbit several times

b. Jan te tuiye lapen pluzie fwa

John killed rabbits several times

169 Finally, as (4) illustrates, HC bare nouns also contrast with yon indefinites in opaque

170 contexts. While yon indefinites have wide and narrow scope readings with respect to the

171 intentional predicate vle (want), HC bare nouns only allow narrow scope.

(4) a. Jan vle rankontre avek yon dokte

There is a doctor John wants to meet

John wants to meet some doctor

b. Jan vle rankontre avek dokte

Jan wants to meet doctors

189 Modified bare nominals that cannot correspond to kinds, however, do not show the same

190 obligatory low scope properties. As illustrated in (5), they can have either wide scope or

191 low scope with respect to negation.

(5) Jan pa t we moso machin yo te poze sou tab la

John not past see parts machine they Past put on the table

John did not see parts of the car that were put on the table

200 Similarly in (6), the Haitian equivalent of ‘John killed rabbits that were running several

201 times’, the modified bare nominal has an obligatory wide scope reading that leads to an

202 incoherent interpretation for the sentence, i.e. repeated killing of the same rabbits.

(6) Jan te tiye lapen ki t ap kouri plizye fwa

J. killed rabbits that were running several times

209 In sum, HC bare nouns clearly display the basic features argued by Carlson (1977) and

210 others to be characteristic of English bare nominals.

211 Haitian Creole clearly differs from English, however, with respect to number morphol-

212 ogy. In contrast to English, Haitian Creole clearly lacks a standard manifestation of plural

213 inflection. There is no plural inflection on nouns, or on predicates, whether adjectival or

214 verbal, or on any tense related particles as (7) and (8) show:

(7) a. Jan achte anpil/plizie/de livJohn bought many/several/two books

b. Jan achte yon livJohn bought one book

4 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

(8) a. Yon tifi ap rele ou a0. Anpil tifi ap rele ou

One girl is calling you Many girls are calling you

241 Plural is marked in the pronominal paradigm, the clearest contrast being between the 3rd

242 person singular pronoun and its corresponding plural, two clearly distinct morphemes.

(9) li He/she/it yo they

250 HC also distinguishes a singular from a plural definite determiner. Characteristically in

251 fact, the plural definite determiner is homophonous (and ambiguous) with the 3rd person

252 plural pronoun, as (10b) shows.

(10) a. Jan achte liv laJohn bought the book

b. Jan achte liv yoJohn bought the books/their books

267 Thus it seems fair to conclude that plural distinctions in HC are essentially limited to the

268 pronominal domain and, consequently, that the language displays no inflectional plural

269 morphology. While the lack of inflectional plural morphology is not particularly surprising

270 in itself, it turns out to raise trouble for Chierchia’s proposed parametric approach to bare

271 nominals. With respect to number marking, Haitian Creole seems indeed to be typolo-

272 gically close to Chierchia’s Class-I-languages, i.e., the Chinese/Japanese class. For

273 Chierchia, Class-I bare nouns have the basic denotation of kind terms with corresponding

274 mass properties and are [þarg][-pred]. These languages manifest: (1) a rather free

275 distribution of bare nominals in argument positions—since kind terms are individuals

276 of type <e> and good argument types—(2) a lack of plural morphology, (3) the obligatory

277 presence of classifiers for count nouns, and (4) no distinction between mass and count

278 nouns—since properties derived from kind terms have a mass denotation.

279 While Haitian Creole clearly displays the first two properties of Class-I-languages, it just

280 as clearly differs from them with respect to the two remaining ones. First in Haitian Creole,

281 (and all other French based Creoles)2 number terms and quantifiers directly associate with

282 count nouns. The presence of a classifier is never required:

(11) a. Jan achte twa liv

John bought three books

b. Twa tifi vini

Three girls came

c. �Mwen bwe twa dlo

I drank three waters

d. Mwen bwe twa gout dlo

I drank three drops of water

2 Although data from other French Based Creoles are not reviewed here, the generalizations mentioned for

Haitian Creole also obtain for other French based Creoles. In Seychelles Creole, however, BN can additionally

have a definite interpretation. See Deprez (in press) for a discussion.

V. Deprez / Lingua xxx (2004) xxx–xxx 5

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

309 This, however, is not true of HC mass terms, which, like in English or in French, require the

310 presence of a measure phrase to be adequately partitioned. In contrast to count nouns, mass

311 nouns also resist pluralization in HC. They do not allow the plural determiner yo, unless

312 they have a sub-kind or taxonomic reading:

(12) a. Mwen manje diri �yo

I ate the(plur) rice(s) (ok I ate several types of rice)

321 As shown in (13) moreover, HC mass nouns must be referred to with a singular pronoun, a

322 plural one being excluded. This, however, is not true of count nouns, for which a plural

323 pronoun can be used (13b):

(13) a. Paske lo (se bagay ki) ra, li/�yo che

Because gold is rare, it is expensive

b. Paske elefan se bet ki ra, yo che

Because elefants/the elefant are/is a/rare animal(s), he/they is/are

expensive

339 HC bare count nouns further differ from HC mass nouns in the accessibility of their ‘minimal

340 parts’. These seem quite directly accessible for bare count nouns, but not for mass nouns. As

341 shown in (14), HC bare count nouns are compatible with distributive reciprocal predicates,

342 i.e., predicates ranging over the atomic parts of a plurality, but mass nouns are not.

(14) a. Neg, se yon rayi lot

As for people, one hates the other

b. �Sab, se yon sanble lot

As for sand, one resembles the other

357 Note, interestingly, that in this respect, HC bare count nouns differ from other known non-

358 plural kind expressions such as singular definite kinds, which generally fail to support

359 reciprocal predication:

(15) �The lion hate each other

364 Given the above-discussed differences, HC bare count nouns cannot simply be assumed to

365 have the same semantic value as mass nouns, a result not predicted on Chierchia’s

366 approach, since HC has no regular plural morphology.

367 We have seen above that HC count BN differ from mass BN in a number of respects.

368 Below, we see that they also differ from the count BN of regular plural marking languages.

369 To begin with, HC count BN can have either a singular or a plural construal, depending on a

370 variety of contextual and pragmatic factors. Consider the examples in (16):

(16) a. Jan achte liv pou Pol

John bought book for Paul

b. Jan achte kay pou Pol

John bought house for Paul

6 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

385 While in (16a), the BN liv is preferably understood as a plural, the opposite is true in (16b)

386 for kay. Yet there are, strictly speaking, no linguistic differences between these two

387 sentences. The plural vs. singular construal of the BN appears here to be due to socio-

388 cultural and pragmatic factors. Joseph (1988: 88) gives the following explanation. As

389 books are rather affordable objects to buy, the speaker that would have wanted to signify

390 singularity in (16a) could have used the indefinite/numeral marker yon ¼ one, a. In absence

391 of yon, the bare noun in this context signals a plural interpretation. In (16b) in contrast,

392 socio-cultural factors make the singular interpretation most salient, houses being objects

393 that are not commonly bought in quantity. Thus some markers such as anpil (many) would

394 have been needed to signify plurality. In the absence of such markers, the singular

395 interpretation is favored. This singular interpretation is clearly not available for an English

396 bare plural, despite comparable pragmatic presuppositions on house buying events. In such

397 contexts at least,3 an overt determiner is required to express singularity in English.

398 Similarly in (17), the interpretation of the BN piki is preferably singular, but nothing

399 properly linguistic forces it to be:

(17) li te bay m pikiS/he gave me a shot

406 In (18), the interpretation of the BN kalbas is dependent on the overtly marked number of

407 the subject. As it is known that children can usually carry only one calabash at a time, (18a)

408 has a singular construal. For the same reasons exactly, (18b) has a plural (distributive) one:

(18) a. Ti moun nan pa pote kalbas pou l pran dlo

The child did not bring a calabash to take water

b. Ti moun yo pa pote kalbas pou yo pran dlo

The children did not bring calabashes to take water

423 Confirming this dual interpretative possibility, either a singular and or a plural pronoun can

424 be used to reference a BN:

(19) a. Zwazo fe nich li/yo nan prentan

Bird(s) build its/their nest in the Spring

b. Chen se bet ki jape, li/yo jape anpil nan nwit

The dog/Dogs is/are (an) animal(s) that barks. It barks a lot at night

c. Bourik fe pitit se pou do l’ ka repoze Haitian Proverb

A donkey has an offspring so that its back can rest

d. Bouch manje tout manje, men li pa pale tout pawol (Savain, 1993: 103)

The mouth can eat all food but not say all words

e. Vwayel kon chante. Yo gen bel vwa (Dejean, 1985: 3)

Vowels can sing. They have a nice voice.

3 Singular interpretations of English BN are possible in so called dependent plural contexts: Unicycles have

wheels.

V. Deprez / Lingua xxx (2004) xxx–xxx 7

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

466 That HC BN are clearly compatible with semantic plural requiring predicates is shown in

467 (20). In this respect, they pattern in similarity with bare plurals.

(20) a. Zotolan gaye nan tout mon (Joseph: 1989: 190)

Ortolans are spreading in the whole mountain

b. �Yon zotolan gaye nan tout mon

An ortolan is spreading in the whole mountain

486 But they also resemble singular definite kind terms that can felicitously associate with

487 semantically plural predicates, in spite of their morphological singularity:

(21) The lion gathers near acacia trees when it is tired (Gen: 90)

493 Further Haitian examples, in fact, clearly show that both the ‘singular’ and the ‘plural’

494 construal of BN is available with semantically pluralizing predicates. In (22), we see that a

495 reflexive form of the same verb allows for either a plural or a singular pronominal reference

496 to the bare noun.

(22) a. Zotolan gaye ko li nan tout mon (Joseph: 192)

Ortolan spread body his in whole mountain

The ortolan is spreading (itself) in the whole mountain

b. Zotolan gaye ko yo nan tout mon

Ortolans spread body their in whole moutain

Ortolans are spreading (themselves) in the whole mountain

523 Data in (1–6) have shown that Haitian BN have a number of properties that are identical to

524 those of English BN. In particular, they allow for existential, generic, and kinds readings

525 and have systematic low scope. However, Haitian and English BN differ with respect to

526 number marking and interpretation. In contrast to English, Haitian BN show no number

527 marking and are compatible with both a singular and a plural interpretation in existential

528 and generic contexts. The facts reviewed above thus clearly suggest that Haitian BN are

529 under-specified for number, both morphologically and semantically. Yet, as there are some

530 clear distinctions between HC count BN and mass nouns, it seems empirically clear that

531 being under-specified for number is not equivalent to having a mass denotation. HC count

532 BN do not need a classifier or a measure phrase to felicitously associate with numerals or

533 quantifiers. The plural definite determiner can determine them and they are compatible

534 with reciprocal predicates. HC mass BN in contrast, manifest none of these properties and

535 seem rather to correspond to a morphological singular, as they can be referenced by

536 singular pronouns only.

537 Put together, these data raise problems for Chierchia’s proposed typology of bare

538 nominals. Like Chinese-Japanese BN, HC BN can be said to be morphologically unmarked

539 for plural. Yet, they do not require the presence of classifiers and manifest a count/mass

540 distinction. HC BN cannot be easily assumed to be English-like either, i.e., [-arg] [-pred],

541 as they have a possible singular construal. Moreover, HC lacks the required trigger for the

542 [-pred] setting, namely morphological plural. HC BN also differ from the Class-III BN, i.e.,

8 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

543 the Romance BN, in distribution and in interpretation. They readily allow a kind

544 interpretation and manifest no distributional subject/object asymmetry.

545 As Munn and Schmitt (this volume) discuss, Chierchia’s proposal also indirectly makes

546 some predictions for bare nominals in non-argument positions. The occurrence of Class-I

547 BN, with features [þarg][-pred], should either be impossible or severely restricted. The

548 reverse is expected for Class II and III BN, which as [þpred] should occur in non-argument

549 positions without restrictions. Munn and Schmidt show that for [þpred] languages like

550 English, Chierchia’s predictions fail. In English, for instance, it is clear that predicative BN

551 manifest as much of a plural restriction as argument BN (23):

(23) �John is doctor�The lion is mammal

558 Since for Chierchia, the plural restriction is a consequence of the type-shift to kind

559 denotation required in argument positions to provide a suitable argument of type <e>, it is

560 not motivated in predicative positions.4

561 Chierchia’s predictions also seem to fail with respect to a potential class-I-language like

562 HC. If BN are [þarg], their distribution should be severely restricted in predicative

563 positions. But as examples (19) repeated here in (24) and example (25) show, HC BN, in

564 fact, occur quite freely as predicates:

(24) a. Chen se bet ki jape. Li jape anpil nan nwit

Dog Pres animal that barks. It barks a lot at night

b. Chen se bet ki jape. Yo jape anpil nan nwit.

Dogs Pres animals that bark. They bark a lot at night

(25) a. Michel se mason

Michel is mason

b. Loulou se tigason

Loulou is a boy

c. Yon tomate se legim

A tomatoe is a vegetable

599 Interestingly, as (24a) with a singular pronoun and (25) show, bare nominal predicates are

600 not limited to plural predication in HC. They can have a singular interpretation. In this

601 respect, their distribution is even less constrained than in English. HC BN, in sum, fail to

602 adequately fit Chierchia’s proposed typology both in argument and in predicative positions.

603 3. An alternative approach: the plural parameter

604 Chierchia (1998) proposes to derive cross-linguistic variations in the meaning of BN

605 from the interaction of a flexible mapping between syntactic categories and semantic

4 As noted by Zamparelli (pc), the restriction to plural in predicative position is a mystery for any approach

that assumes a basic predicative denotation for BN.

V. Deprez / Lingua xxx (2004) xxx–xxx 9

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

606 types, a semantic parameter governing this mapping and some economy constraints on a

607 universal set of type shifting operations (cf. the Blocking Principle). The model explored

608 here offers instead a rigid category to type match, as in the original spirit of Montague

609 grammar, and as advocated in Zamparelli (1995), and uses null structure and syntactic

610 principles and parameters to derive the possible meaning range of bare nominals. In so

611 doing, it attempts to increase recourse to syntactic structures and parameters that have

612 been independently motivated in recent works on DP in an effort to better interface

613 semantic interpretation with the results of syntactic findings. The aim, moreover, is to

614 follow the Borer-Chomsky (1995) line of thought in its attempt to limit the domain of

615 parametric variation to morphology and to deduce, whenever possible, interpretative

616 distinctions from overt morphological differences. The central ingredients of this

617 proposal, first sketched in Deprez (1999a, 2001), are briefly summed up in the table

618 below. They are discussed in the present section.

619

620 The proposal explored here has at its heart a morpho-syntactic parameter called the Plural

621 Parameter. This parameter distinguishes two broad sets of languages, the þPL and the

622 �PL, according to whether or not the structure of their nominal expressions obligatorily

623 includes a functional projection for Number.5 In þ PL languages, NumP must system-

624 atically project, for count nouns at least—that is, even when the noun phrase is singular

625 with no apparent overt ‘‘plural’’ or ‘‘singular’’ morphology—and this NumP must contain

626 a counter, i.e., a measure function, which I take, as a first approximation, to be the

627 translation of countabilty. In �PL languages, on the other hand, the projection of NumP is

628 optional and when it occurs, NumP does not have to contain a counter.

629 Like other parametric distinctions in current day syntactic theory, the proposed Plural

630 Parameter concerns functional structure and capitalizes on an overt morphological difference

631 among languages, namely the presence vs. absence of a ‘rich’ plural morphology. However,

632 the frontier betweenþPL and�PL languages, I suggest, is not so much in the ‘richness’ of the

633 morphological marking, as in its ‘unavoidability’. Understood this way, the parameter

634 separates on the one hand, languages in which, in the most general case, no morphological

635 form of count nouns is understood outside of the number system—that is, nominals are either

636 singular or plural and interpreted as such, so that the absence of plural morphology entails a

637 singular interpretation—and on the other hand, languages in which one form of the noun

638 manifests what Corbett (2000) has called ‘general number’—with the meaning of ‘one or

639 more x’,— what I have here called an under-specified form.

5 It is important to note that despite its name, the proposed parameter is not claiming that some languages,

the –PLL ones, fail to express plurality.

The Plural Parameter (Deprez, 1999a, 2001)NumP must project and contain a semantic counter (�)

NP Ł <e> kind

NumP Ł <e,t> realization of a kind (object level individuals or sub-kinds)

NumP ¼ lKlnlwlx[R/Tw (x,K) & OU/Kuw(K)(x) ¼ n]

NumP introduces a predicate over objects (R) or subkinds (T) of a kind (and a counter

of object units (OU) or kind units KU)

10 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

640 The spirit underlying this proposal is as follows. I follow Bosweld de Smet (1997) in

641 assuming that notions such as mass and count are lexical in nature and distinguished from

642 notions such as uncountability and countability, which are structural and thus composi-

643 tional. On this view, the mass/count distinction is a lexical feature, assigned on the basis of

644 the ease with which a noun appears in unambiguously countable or uncountable contexts

645 (Allan’s 1980 ‘Countability preference’ test) but it can be overruled by contextual or

646 syntactic factors so that mass terms can become count and count terms uncountable. The

647 basic denotation of all nouns, however, is assumed to be the same: They denote kinds and

648 are under-specified for number. All N have thus the same semantic denotation, with

649 countability being a compositional syntactic notion that is superimposed on their basic

650 denotation. In the view explored here, it is NumP that assumes the role of bringing about

651 countability, as it retrieves on the one hand instantiations of a kind (objects or sub-kinds

652 (following Krifka, 1995)) and imposes, on the other hand (but in some cases only), a

653 measure function on these instantiations. Formally, counting can be seen as an application

654 of the cardinality function to sets of individuals. Measure functions can be seen as the

655 application of some measure function that results in a proper partition. It has been

656 suggested by Colban (1991) and Nerbonne (1993) that measures are perhaps more general

657 than cardinalities and that the latter are a special case of the former. Although my concerns

658 in this paper are limited to count nouns, this suggests that even though the ‘measures’ of the

659 measure function may change for mass nouns and count nouns, their role as well as their

660 syntactic realization is essentially the same. Both are located in NumP. In this respect, I

661 follow Cheng and Sybesma (1999) in assuming that classifiers are largely the equivalent of

662 plural morphology in languages that have them. These languages, however, still differ from

663 the þPL languages in allowing for an optional projection of NumP rather than an

664 obligatory one.

665 The proposal further assumes that NumP can have two distinct possible realizations:

(1) NumP can be a pure instantiation of the Carlsonian realization rule R (without stages

668 cf. Krifka, 1995) from kind to objects and generalized to taxonomic sub-kinds, that

669 gives the set of objects (or sum of objects) or sub-kinds (or sums of sub-kinds)

670 instantiating a given kind in a given world:

(26) NumP ¼ lw lx. Rw(x, K)

In this case, the resulting denotation of a NumP projection is that of a property true of

individuals or sum individuals and under-specified for number. Closure of this

property yields the existential reading of bare nouns that are under-specified for

number. As the present paper does not bear directly on how this closure occurs, I will

leave this choice open, various options being compatible with the present proposal.

Given the formulation of the Plural parameter above, this type of NumP is allowed in

�PL languages only.

(2) NumP can be the Carlsonian realization rule associated with a measure function that

returns a countable property. I follow here Krifka (1995) in assuming a measure

function (OU ¼ object Unit), which, when applied to a kind and an object, returns a

number, giving us the number of objects instantiating the kind in a given world.

V. Deprez / Lingua xxx (2004) xxx–xxx 11

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

Krifka also generalizes this function to sub-kinds (KU ¼ Kind Unit) for languages

like English that do not distinguish measure functions for object units and kind units.

(Three bears ¼ three object bears or three subkind bears). The denotation of NumP

for this case is given in (27):690

(27) NumP ¼ lKlnlwlx[R/Tw (x,K) & OU/KUw (K)(x) ¼ n]

695 Note importantly that the counting function OU/KU introduces a number argument. The

696 presence of this number argument in turn requires the presence of some element to saturate

697 it. I take it that various morphological elements can saturate the counting function in

698 various ways, but what is crucial is that there be some element to accomplish this

699 saturation. For instance, if NumP contains a number term, then this number term can

700 saturate the number argument. This, for example, derives the following denotation for

701 NumP applied to the kind BEAR and the number term three:

(29) [NumPThree bears] lwlx [Rw (x,BEAR) & OUw (BEAR) (x) ¼ 3]

707 The same is true I assume, if the number term is one or is an indefinite determiner with

708 essentially the same meaning as a. A second way of saturating the number argument is with

709 plural morphology. If NumP contains a morphological plural marker (s for English), then

710 following Krifka again, I assume that the number argument can be existentially quantified

711 over as follows:

(30) Bear þs 9n lwlx [Rw (x,BEAR) & OUw(BEAR) (x) ¼ n > 1]

717 In the absence of any morpheme, however, and if nothing happens at a higher functional

718 level, such as for instance the introduction of a determiner, the number argument will

719 remain unsaturated and the NumP will be un-interpretable. This, I suggest, is what is at the

720 source of the impossibility of singular bare arguments in þ PL languages, where the

721 projection of NumP is obligatory and must contain a counter.

722 The proposed Plural Parameter locates variation in the lexicon, or more specifically in

723 the inventory of morpho-syntactic features a language may chose to instantiate. It adds,

724 however, a twist to this broadly accepted view of parametric variation, in the sense that

725 morphological ‘richness’ is here tied to a specific interpretation with the result that the

726 parametric choice now has semantic consequences in addition to expected syntactic ones.

727 The key idea is that languages that manifest a ‘rich’ plural morphology project a nominal

728 structure that contains a semantic counter over realizations of a kind. Languages that do not

729 have a rich plural morphology, in contrast, do not enforce the presence of such a projection,

730 nor of such a counter, although, they can allow them optionally.6 It thus follows from this

12 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

731 proposal that languages that have a richer plural morphology are expected to manifest more

732 constraints on the possible interpretation of their bare nominals.

733 The proposed Plural Parameter as formulated above has three direct semantic con-

734 sequences.

(1) Only the �PL languages allow direct access to the basic kind denotation of nouns.

737 For the þPL languages, access to kind denotation requires the presence of a relevant

738 operator.

(2) Only the �PL languages can have bare nominals that are under-specified for number

740 and thus compatible with either a plural or a singular construal depending on

741 contextual factors.

(3) In þPL languages, in the absence of relevant morphology, bare ‘singular’ nominals,

743 i.e., NumP that contain an unsaturated counter, are excluded.744

745 This simple parametric proposal has some far-reaching semantic and syntactic con-

746 sequences. The following section will test these predictions against a number of cross-

747 linguistics empirical facts. The proposal offers a rigid category to type match in the sense

748 that the syntactic category NP and the syntactic category NumP are always assumed to

749 rigidly map into a single semantic denotation. NPs are always mapped to a denotation of

750 type <e>. In this respect, the claim of the present proposal is that the category NP always

751 denotes a kind and never anything else, cross-linguistically. The category NumP, on the

752 other hand, always denotes the realizations of a kind, that is, the set of objects/subkinds that

753 instantiates a kind in a given world, and in this sense, it always maps onto an <e,t>

754 denotation.

755 Syntactically, the proposed Plural parameter entails that bare arguments can be NPs in

756 �PL languages only, and that whenever they are NPs, they always denote a kind. For þPL

757 languages, in contrast, the parameter entails that bare arguments cannot map onto NPs. The

758 syntactic consequence of the proposal is that bare arguments must always be minimally

759 NumPs. If we take the NumP projection to be in a sense structurally equivalent to a null

760 determiner—it is a null syntactic functional head endowed with number features—the

761 proposed parameter can then be said to entail that bare arguments always have null

762 determiners in þPL languages, but not necessarily in �PL ones. The presence of a null

763 syntactic determiner-like-head on top of bare nominals has been argued for extensively in

764 the works of Longobardi (1994 and following) as well as in Chierchia (1998), for the

6 The obligatory versus optional projection of NumP is one dimension of variation. The presence or absence

of a semantic counter is presumably another. Here, as in my 2001 paper, these two dimensions are directly

linked, but further cross-linguistic comparison could justify a move to consider independent variation of the two.

Since in the proposed model, the former dimension governs direct access to kinds while the later concerns

number interpretation, the possible dissociation of the two factors would make room for languages that allow

bare singular (underspecified) arguments but no direct access to kind readings (obligatory NumP, optional

counter). Brazilian Portuguese has been argued to be such a language in Muller (2000). However, there seems to

be disagreement on the basic facts, since Munn and Schmitt (2001) have noted in contrast that bare singulars in

BP can have a kind reading. These may of course represent two distinct dialects, a conjecture that seems quite

likely given the weakening force of plural marking in BP. As described by Muller (2000) ‘The bare plural

belongs very much to the written language register. The most usual oral nominal forms that express genericity in

BP are either the definite singular or the bare singular.’ The dissociation of the two factors is explored in further

work.

V. Deprez / Lingua xxx (2004) xxx–xxx 13

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

765 Romance languages. There are, however, fewer consensuses on the presence of a null

766 determiner for English BN. But interesting synchronic and diachronic syntactic arguments

767 can be found in the works of Longobardi (1994, 1999) and Crisma (1991) among others.

768 The presence of null NumP heads has also been proposed for Chinese (Cheng and

769 Sybesma, 1999) and Brazilian Portuguese (Munn and Schmitt, 2001). The arguments

770 given in these works are not repeated here. Instead, a new syntactic argument is provided

771 based on the distribution of Haitian Creole bare nominals in predicative sentences. Notably,

772 it is a consequence of the proposal made here that BN should always be NumPs whenever

773 they are used predicatively, kinds being a priori not adequate predicate types. This makes

774 the prediction that predicate BN should always have a null determiner like projection even

775 in languages where NumP is optionally projected. A study of Haitian BN in predicate

776 position will provide empirical support for this prediction.

777 4. Testing the empirical ground

778 In this section, the proposed alternative described in Section 3 is tested against a variety

779 of empirical facts to evaluate the validity of its predictions. I first return to Haitian Creole

780 BN, our present example of a �PL language, and show how the above model accounts for

781 its properties. I then turn to a more detailed investigation of Haitian BN in predicative

782 position and consider how the present model fares with respect to predicate BN in a �PL

783 language. I then consider descriptive generalizations about bare nominals in distinct types

784 of þPL languages, in particular, English, Italian, French and Hindi.

785 4.1. Haitian Creole: a �PL language

786 As described above, Haitian Creole is very clearly a �PL language as it only optionally

787 marks number, whether singular or plural. It was shown above that HC BN could have kind

788 readings as well as existential and generic readings. In terms of the Plural Parameter

789 summarized above, direct access to a kind reading is predicted, as NumP does not have to

790 project in this language. Note moreover, that on this view, directly accessed kinds are

791 under-specified for number. It thus follows that BN interpreted as kinds should be able to

792 correspond indifferently to a morphological plural or a singular. This prediction is verified

793 in HC with examples like (22) that clearly have a kind reading and allow either singular or

794 plural pronominal reference. The existential and generic readings of HC BN are also

795 predicted. In these cases, NumP is taken to project, leading to a property that can be closed

796 under existential quantification or binding by a generic operator. Assuming that closure of

797 NumP under existential quantification proceeds essentially in the same way in �PL

798 languages, existential bare nominals are expected to have essentially the same scopal (or

799 more exactly scopeless) properties, as was indeed shown to be the case. The main

800 difference with English was shown to concern number interpretation. In this respect,

801 since Haitian Creole is �PL, the prediction of the Plural Parameter is that NumP does not

802 have to contain a counter with the result that bare NumPs can remain under-specified for

803 number. Lexical, pragmatic and contextual information will prevail to determine the plural

804 or singular interpretation of HC bare nominals under their existential as well as generic

14 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

805 interpretations.7 It is for this reason, in my view, that HC permits what can be considered as

806 semantically singular bare arguments in examples like (16b) and (17). Note, however, that

807 specification for number is optionally possible for HC NumP. In this case, however, some

808 morpheme must be present to saturate the number argument. Such NumPs must thus

809 correspond to nominals with an overt determiner (not to BN), such as for instance a numeral

810 term like the numeral one yon that also serves as a marker for singular indefinite nouns.

811 Note that the proposal put forth here entails that bare nominals in a �PL language do not

812 have the same syntactic structure on their kind denotation and on their property denotation.

813 On the former they are bare NP projections, on the later, they have a null NumP. Empirical

814 evidence for such a claim would of course considerably reinforce the proposal. Yet finding

815 empirical evidence for the presence of an invisible and inaudible element is never easy. One

816 traditional type of evidence adduced for the presence of a null syntactic head in BN has

817 been the distributional subject/object asymmetry observed in the Romance languages.

818 However, as was shown above in Section 2, no such asymmetry occurs in HC. To explain

819 this lack of asymmetry, Deprez (1999a, 2001) argued that the null NumP heads of HC are

820 internally licensed by the movement of NP to Spec NumP and provided arguments in

821 support of internal phrasal movements in Haitian Creole nominal projections (see also

822 Deprez (forthcoming a.) for more detailed evidence for DP internal phrasal movements in

823 all French Based Creoles). Being internally licensed, an HC null NumP needs no external

824 licensing, hence the lack of detectible asymmetry.

825 Distributional asymmetries can thus not provide evidence for the presence of a null

826 NumP head in HC. The presence of a syntactically active null determiner would never-

827 theless gain empirical grounds if it could be shown that, in some contexts, apparently

828 determiner-less nominal phrases have the same syntactic effects as nominal phrases that

829 contain overt determiners. As argued in Section 4.1.2, such conditions are observable in the

830 predicative nominal constructions of HC. As these constructions are surely one of the

831 clearer cases of bare nominals requiring a predicate denotation, evidence of the presence of

832 a null head in these bare nominals can be seen as providing support for the syntactic

833 realization of NumP.

834 4.1.1. Haitian Creole bare nominal predicative constructions

835 It has often been observed that in HC, predicative constructions are most commonly

836 direct, i.e., making no use of a copula. As shown in (31), when the predicate is an adjective,

837 an adverbial or a (locative) preposition, the presence of an element like se, sometimes

838 analyzed as a copula (Pompilus, 1976), is neither required, nor possible and predication

839 must be direct.

(31) a. Jan entelijan a0. �Jan se entelijan

John is intelligent

b. Jan nan lakou a b0. �Jan se nan lakou a

John is in the courtyard

7 Here I am tentatively suggesting that it is the lexical feature of mass nouns that commands the singular

pronominal reference.

V. Deprez / Lingua xxx (2004) xxx–xxx 15

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

862 However, when the predicate is a full noun phrase with an overt determiner, the reverse is

863 observed. The element se is required and direct predication by simple juxtaposition is

864 impossible.

(32) a. �Jan fre m a0. Jan se fre m

John is my brother

b. �Jan yon pwofese b0. Jan se yon pwofese

John is a teacher

c. �Jan pwofese a c0. Jan se pwofese a

John is the teacher

897 Assuming this generalization to be correct (see Pompilus (1976), Damoiseau (1987)

898 DeGraff (1992a, 1992b) for similar conclusions), the interesting question at this point is

899 what happens in predicative sentences with bare nominals. Observe first that, in these

900 cases, the presence of the element se is possible:

(33) a. Moun sa yo se dokte

These people are doctors

b. Chen se bet ki jape

Dog is an animal that barks

915 Bare nominal predicates then clearly pattern like nominal predicates with an overt

916 determiner in allowing the presence of se. This of course is entirely expected if both

917 types of predicates have the same syntactic structure, more specifically, if both are full noun

918 phrases with, respectively, null and overt determiners. Yet the data provided so far does not

919 suffice to enforce this conclusion. It does not eliminate a plausible alternative account of

920 the presence of se based on the nominal nature of the predicate. The presence of se could

921 indeed be enforced by the syntactic category of the predicate rather than by the presence of

922 a null determiner. That is, if se were obligatory with [þN,�V] predicates and impossible

923 with predicates of all other syntactic categories, the above data would also follow. As I

924 argue below, however, there are clear empirical arguments against this categorical

925 alternative and in favor of the null determiner approach.

926 To begin with, examples like (34) show that se can occur with certain types of

927 prepositions, not just with nominal categories:

(34) a. li se tankou se’ m

3rdS se like sister my

She is like my sister

b. Tout sa se pou ou

All this se for you

All this is for you

948 Second and perhaps, more interestingly, se turns out to be optional with some predicates

949 that are quite clearly nominal in nature.950

(35) a. Jan chapantje

John (is a) carpenter

16 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

b. Michel mason

Michel (is a) mason

967 Examples like (35) thus show clearly that the presence of se cannot be enforced solely on

968 the basis of categorical information. There are some clearly [þN] predicates that do not

969 obligatorily require it. Interestingly, the interpretation of bare nominal sentences with se

970 differs from those without se. As noted by one of our informants, an example like (35a)

971 without se is felicitously uttered in a situation where John has just become a carpenter,

972 perhaps because he just finished school or a carpentry degree. The reading in this case is

973 something like: John is now a carpenter, although he may not have been one before (or the

974 speaker may not have known about it). In other words, the property of being a carpenter is

975 located in time and is not a general identifying feature of the subject. The reverse is true of

976 the version with se. A comparable interpretative difference is noted in Pompilus (1976) for

977 examples like (36) below:

(36) a. Nou malad a0. Nous sommes malades

We are sick (presently)

b. Nou se malad b0. Nous sommes des malades

We are sick people

1000 For the variant without se, Pompilus gives an adjectival and temporal interpretation. For the

1001 variant with se in contrast, he gives a nominal and characterizing interpretation that asserts the

1002 membership of the subject in the class of sick people. It is then interesting to note that the

1003 corresponding French translations—given by Pompilus himself and reproduced here in

1004 (36a0) and (36b0)—manifest an alternation between the presence and the absence of an overt

1005 determiner for a comparable interpretative effect. In (36a0) with no determiner, malade is an

1006 adjectival predicate with a temporal interpretation; in (36b0) malade ‘nominalized’ by the

1007 presence of an overt determiner is understood as a characterizing property.

1008 The parallelism between French and HC bare nominal predicative sentences does not

1009 stop here. As observed by Pompilus, bare nominal predication without se is possible in HC

1010 only with a restricted class of nominal predicates, roughly speaking, the professional ones.

1011 With other types of nominal predicates, as in (37), the presence of se is strongly preferred

1012 and direct predication is judged quite poor:8

8 These restrictions have been reapeatedly noted in the literature on HC. See for instance Wingerdt-Philips

(1982) among others. However, since they are not absolute, the exact nature of the predicates involved seems rather

difficult to establish. What is eminently clear is that whenever a bare nominal is used without se it must receive a

temporally bounded interpretation . The class of nominal predicates that licence the bare nominal construction is

equally hard to define in French as various factors enter acceptablity judgements on this type of sentences.To

illustrate, although an example like (ia) is strongly excluded, (ib) with the same nominal predicate is perfectly

interpretable and attested. In (ib), the use of an indefinite marker would in fact simply not make sense.

(i) a. �Un pingouin est oiseau

A pinguin is a bird

b. Mon corps est oiseau (L’amour sorcier: Claude Nougaro)

My body is bird-like

The relevant distinction closely resembles the individual level/stage level difference in its flexibility. Predicates

seem lexically more prone to a given classification but they can be contextually coerced into a different one.

V. Deprez / Lingua xxx (2004) xxx–xxx 17

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

1013 (37) a. �Jan tigason

John is boy

b. �Yon tomat legim

A tomato is vegetable

1028 These restrictions, in fact, closely parallel comparable restrictions observed in French

1029 on the possible occurrence of singular nominals without determiner in predicative

1030 sentences. It is well known that in French, to be acceptable predicates in nominal

1031 predicative sentences, most types of singular nouns require the presence of an

1032 indefinite determiner as in (38a). Yet certain types of nouns, roughly speaking the

1033 so-called professional ones, can be used predicatively without a determiner as in

1034 (38b):

38) a. Jean est un garcon �Jean est garcon

John is a boy John is boy

b. Jean est medecin/dentiste/professeur

John is doctor/dentist/professor

1053 This is rather unexpected given that French quite generally disallows other uses of BN.

1054 The point of interest to us here is the great similarity between the limitations observed

1055 on bare nominal predication in French and those observed for the direct nominal

1056 predication without se in HC. Both manifest comparable interpretation and comparable

1057 restrictions on the possible types of predicates. Moreover, there is a point-by-point

1058 correspondence between the constructions that lack se in HC and those that lack an

1059 overt determiner in French and those that feature se in HC and an overt determiner in

1060 French. In this regard, the interesting meaning distinction that contrast these two types

1061 of constructions in the Haitian examples of (39) is quite telling. In (39a) with se, the

1062 predicate bet only has a characterizing and nominal interpretation, meaning ‘animal’.

1063 Without se in constrast, bet in (39b) only has a purely adjectival interpretation, meaning

1064 ‘stupid’.

(39) a. Elefan se bet

Elefants are animals

b. Elefan bet

Elefants (are) stupid

1079 Strikingly this distinction is again exactly replicated in French by pairs like (40) that differ

1080 in the presence or absence of the determiner.1081

(40) Un elephant est une bete An elefant is a beast

Un elephant est bete An elefant is stupid

1090 In sum, we observe that when se is missing, HC bare nominal predicates receive an

1091 adjectival-like and temporally bounded interpretation that parallels the interpretation of

18 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

1092 French nominal predicative sentences without determiners. In contrast, when se is present,

1093 HC predicates receive a nominal and identifying interpretation that parallels the French

1094 nominal predicative constructions with an overt determiner.9

1095 In Deprez and Vinet (1992, 1997) (hence D&V) the element se in HC is analyzed as a

1096 functional projection that surfaces as a last resort in a predicative projection AspP/PredP

1097 when the head position fails to be occupied by a lexical predicate. With adjectival predicative

1098 sentences such as (31) above, where se cannot be present, the lexical predicate comes to

1099 occupy the head of the Asp/PredP projection after movement, as in the structure (41):

(41) [IP Jan [PredP entelijan ½XP t]]]

1104 When the predicate is a noun phrase with an overt determiner as in (32) above, the presence

1105 of the determiner blocks the movement of the predicate. Se must then surface to lexicalize

1106 Predo.

(43) a. �[IP Jan Io [PredP pwofes�e ½DP yon Do ½XP t ]]]]

b. [IP Jan Io [PredP se [DP yon Do [NP pwofese ]]]]

1115 In other words, se must occur when the movement of a lexical predicate to the functional

1116 projection Pred is blocked by the presence of any intervening (functional) head (cf. the

1117 Head Movement Constraint), and more particularly, in the case at hand, by the presence of

1118 an overt determiner.

1119 In light of this analysis, let us return to the case of bare nominal predicates. If their

1120 structure contains a bare NP as in (44), nothing should prevent the movement of the lexical

1121 nominal predicate into the functional head. The expectation is then that se should not occur

1122 since the head of the functional projection is filled.

(44) a. [IP Jan Io [PredP chapantye ½NP t ]]]]

1128 If on the contrary, bare nominals contain a null determiner (or null NumP) as in (45), it is

1129 plausible to think that, just like with overt determiners in HC, the movement of the

1130 predicate to the functional projection is blocked. In this case, the presence of se is required.

(45) a. �[IP Nou Io [PredP chapantye ½NumPNumo [NP t]]]]

b. [IP Nou Io [PredP se [NumP Numo [NP chapantye]]]]

1139 Only this second structure can explain why se is allowed with bare nominals. This clearly

1140 suggests that, professional predicates aside, HC bare nominals predicates have in the

1141 general case a phonetically null determiner with the syntactic effects of an overt one,

9 For similar observations about the disctinct interpretations see among other Bentolila (1978) Ferere (1974)

and Damoiseau (1996).

V. Deprez / Lingua xxx (2004) xxx–xxx 19

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

1142 namely blocking the movement of the lexical predicate to Pred0.10 Professional predicates,

1143 it would seem, have the option of moving directly to the Asp/Pred position. This may be

1144 attributed to their adjectival-like nature. Perhaps, as suggested by Munn and Schmitt (this

1145 volume), they have an event argument in their thematic grid. Besides accounting for the

1146 distribution of se, the structures (45a) and (45b) elegantly explain Pompilus’ contrast in

1147 (36). When se is missing, malad is a bare predicate that behaves like an adjective and raises

1148 to the functional Pred head. When se is present, malad is embedded in a nominal projection

1149 with a null determiner that blocks the movement of the predicate to Predo, enforcing the last

1150 resort presence of se. Given these two structures, the interpretation differences also follow.

1151 In the first case, malad is truly like a (stage level) adjective and is interpreted accordingly.

1152 In the second case, malad being dominated by a NumP is comparable to a noun, i.e., it is a

1153 ‘‘nominalized’’ adjective.

1154 This account also establishes a simple and clear parallelism between HC and French. In

1155 HC as in French, the distinction between the two cases now merely boils down to the

1156 presence/absence of a determiner, the only difference being that in HC, this ‘determiner’,

1157 understood here as Num0, remains silent.11

1158 Given this analysis, the above observed facts fit a rather simple generalization. In HC, se

1159 is absent whenever a predicate, whatever its categorial nature, can move into the head of the

1160 functional ASP/Pred projection. Se is required whenever an intervening determiner (or

1161 functional (Po)) head blocks this movement. This approach provides, on the one hand, an

1162 elegant account of the apparent optionality of se with bare nominals and on the other hand,

1163 independent evidence for the presence of a null determiner in HC bare nominals with a

1164 predicative interpretation.

1165 A question that remains at this point is why should French nominal predicates,

1166 professional nominals apart, require the presence of an overt determiner, unlike HC?

1167 That is, assuming that NumP in French has a predicative interpretation, why is a null

1168 determiner excluded? As discussed shortly in Section 4.2 below, this question turns out to

1169 have a simple answer in the present approach, for French as well as for other þPL

1170 languages, that is determined by the Plural Parameter.

1171 This section discussed how the proposed model accounts for the properties of HC BN.

1172 The main characteristics of HC BN, i.e., their possible kind, existential and generic

1173 readings, their scopelessness and their under-determination for number have been shown to

1174 receive simple accounts. Furthermore, one interesting syntactic consequence of the Plural

1175 Parameter, namely that BN interpreted predicatively must project a null functional head,

10 A question may arise as to why a nominal head could not move head to head through Num0 and then on to

Pred0. There are several possible answers to this question. Recall first that in HC NP moves to Spec NumP to

license the null Num head. N movement to Numo would then have to come from a left branch. As has been

argued in Deprez (forthcoming a.), moreover, there is no overt N to D movement in French based Creole DPs.

Covert head movement in DP is thus all the more unlikely. Concerning predicate head movement to Predo, V&D

consider this process as similar to head movement within a VP projection (V to v) available in languages like

English (Chomsky, 1999), where no other instance of head movement occurs.11 If, in French, the copula lexicalizes To and not Predo, its constant presence is expected. Movement to Predo

may thus take place in French adjectival predicative constructions just as in HC. The presence of a lexical

element in To moreover, can be assumed to be sufficient to prevent lexicalisation of Pred0. In HC when Tense is

expressed, se is absent with all predicates. See Deprez and Vinet (1992, 1997) Deprez (1999a, forthcoming b.)

and DeGraff (1992a, 1992b) for more detailed analyses of predicative constructions in HC.

20 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

1176 has been confirmed in a detailed study of the properties of HC bare predicate nominal

1177 constructions. The next section considers the consequence of the proposed approach for

1178 þPL languages.

1179 4.2. BN in þPL languages

1180 This section turns to a brief evaluation of the consequences of the proposed Plural

1181 parameter for the þPL languages. English, Italian and French are on our terms all þPL

1182 languages. Although these languages manifest a number of common properties in regard to

1183 bare nominal constructions—for instance, they all disallow singular bare nominals—they

1184 also manifest a number of well-known differences. It is the purpose of this section to

1185 discuss these similarities and differences in regards to the predictions of the present

1186 proposal.

1187 Let us begin with the restriction against singular bare nominals. The generalization that

1188 seems most salient across English, French and Romance is that these languages all disallow

1189 singular bare nominals, arguments or predicates. This general statement surely needs some

1190 qualification, since as discussed above, French, as well as the other Romance languages,

1191 allow a restricted type of bare singular nominals in predicative positions with professional

1192 predicates. Whether this is also true of English, where the range of singular bare nominal

1193 predicates seems even more limited, is unclear to me at this point, so I will leave it aside.

1194 But the phenomenon is clearly manifest in German, as attested by (46):

(46) Johan ist Artz

John is doctor

1201 As observed above, professional bare nominal predicates also seem to have a peculiar

1202 behavior in�PL languages like HC. It thus seems in other words, that these predicates have

1203 an unusual behavior in all types of languages, þPL and �PL included. I do not conjecture

1204 here as to the causes of this property, but simply assume that these constructions

1205 are in a sense orthogonal to the more important common ground discussed here, namely

1206 the absence, in ordinary cases, of singular bare nominals across the different þPL

1207 languages.

1208 Now it is a straightforward prediction of the Plural Parameter that singular bare nominals

1209 should be ruled out quite generally in þPL languages. Let me briefly recall why, before

1210 dealing with some potential counterexamples, of which only one, namely the case of Hindi,

1211 is discussed below. Consider for instance BN in predicative positions. In the proposed

1212 approach, BN with a predicative <e,t> interpretation are NumP. Recall that given the Plural

1213 Parameter, NumP must project in þPL languages and contain a counter introducing a

1214 number argument that must be saturated. Assuming that in þPL languages, singular

1215 morphology, which corresponds in fact to lack of plural morphology, is not semantically

1216 represented, it turns out that in a singular NumP, the number argument will remains

1217 unsaturated. If so bare singular predicates are simply un-interpretable: They contain a

1218 variable that is unbound, namely, the number variable of the counter. It follows then that in

1219 þPL languages, singular bare nominals predicates will always require the presence of an

1220 overt determiner to be licensed, putting aside, the exceptional cases of the so-called

V. Deprez / Lingua xxx (2004) xxx–xxx 21

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

1221 professional predicates. The same of course is equally true of bare singular arguments.

1222 Since for argument as well, the number argument of the counter must be saturated, either

1223 the presence of an overt determiner, or the presence of the plural morpheme is needed to

1224 satisfy this requirement. In short, in the proposed model, the plural restriction on bare

1225 nominals follows directly from the obligatory presence of the semantic counter in the

1226 projection of NumP and bare singular count nominals are ruled out because they feature an

1227 unbound variable.

1228 Now French differs from both English and Italian in disallowing even plural bare

1229 nominals in predicate or argument positions. This more stringent restriction follows quite

1230 naturally on the proposed view if the French plural morpheme (perhaps there is none?) is

1231 assumed, in contrast with English or Italian, to fail to provide an appropriate saturation for

1232 the number argument of the NumP counter.12 As has been often discussed in the literature,

1233 the French morphological plural on nouns is overall quite weak, in comparison with

1234 English or Italian, having in fact almost no phonological effect, apart from a few

1235 exceptional nouns (cheval/chevaux). If plural is disqualified, the consequence will be

1236 that, in French, there is essentially only one way to saturate the number argument

1237 introduced by the NumP counter, namely through the presence of some overt determiner.

1238 It is beyond the scope of the present paper to provide a formal account of how this

1239 saturation is effected by a variety of determiners. Language specific differences may of

1240 course arise, leading to important variations in the overall resulting picture. But the

1241 consequence of the Plural Parameter is that in þPL languages, saturating the number

1242 argument must be the minimal role that any (count) determiner must fulfill in the absence

1243 of the relevant plural morphology.

1244 Does the present approach generally predict that bare singular nominals are excluded in

1245 all þPL languages? The answer is: almost but not quite. Dayal (2001) has shown

1246 convincingly that a rather clear case of a þPL language, namely Hindi, allows bare

1247 singular arguments quite readily. However, simplifying somewhat, Dayal’s central con-

1248 clusion is that Hindi bare singulars correspond to two distinct types of bare nominals,

1249 definite bare nominals and incorporated ones. Although I will not offer an account of BN in

1250 this language here, the results of Dayal’s research allow us to see how the Hindi case may in

1251 fact fit with the proposed Plural Parameter. First, it must be assumed that Hindi has a null

1252 definite determiner that is essentially equivalent to an overt one. This is, in fact, indirectly

1253 what Dayal argues for. In her view, the definite reading of a Hindi BN comes from a type

1254 shifting operation rather than from a distinct null determiner projection. Although it is

1255 clearly beyond the scope of this paper to provide empirical evidence for the existence of a

1256 separate Definite projection, I will refer the readers to works of Zamparelli (1995) and

1257 Heycock and Zamparelli (2002) as well as to the work of Lyons (1999) that provide rather

1258 convincing arguments for its existence in a variety of languages. Within the framework of

1259 the proposed model, it will then suffice to assume that this null definite determiner, like the

1260 overt one, can serve as the binder for the number argument of the counter in Hindi, thus

1261 allowing apparently bare argument nominals in a þ PL language. As for the incorporated

1262 BN, the same assumption can be made. It may be that the number argument of the counter

12 This proposal reinterprets with some differing consequences the idea that the weakness of plural

morphology is responsible for the absence of bare nominals in French (Delfitto and Schroten, 1991).

22 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

1263 comes to be saturated in this case indirectly from quantification over events. I will leave a

1264 formal approach to the present speculation for further work, my current goal being merely

1265 to suggest how an apparent counter example language like Hindi may be accommodated on

1266 the current proposal. Note finally, that in conjunction with the Plural parameter, the

1267 suggested solution for the Hindi case allows an interesting generalization with respect to

1268 þPL languages. As presently formulated, the Plural Parameter makes the strong prediction

1269 that bare singulars nominals will be found in þPL languages only if the relevant languages

1270 can be argued to have a null definite determiner.13

1271 Leaving now our preoccupation with number, it is time to return to more semantic

1272 considerations and to discuss how the various interpretations of bare arguments in þPL

1273 languages can be accounted for in the current approach. Let us begin by recalling the

1274 straightforward predictions of the Plural Parameter. Since NumP must obligatory project,

1275 it follows that the primary interpretation of bare nominals in þPL languages will be

1276 predicative. Here as above, I leave open how the existential and generic interpretations

1277 arise from this basic denotation, whether it is through some DKP like procedure, or

1278 through existential closure, as the current proposal does not at present directly bare on this

1279 question. The Plural Parameter, however, turns out to make a strong prediction with

1280 regards to kind readings. The prediction is as follows: The kind readings that arise in the

1281 þPL languages must be built on the predicative denotation of bare nominals. That is,

1282 although I have proposed that the basic meaning of all nouns is equivalent in all languages

1283 and denotes a kind, this meaning is only directly accessible in �PL languages, not in þPL

1284 ones, where NumP always obligatorily projects. Kind readings in þPL languages must

1285 then involve the intervention of a particular semantic operator. The prediction turns out to

1286 be correct for the Romance languages, since as has been extensively argued by Long-

1287 obardi (1999) for Italian, (see also Zamparelli, 1995) and by Benedicto (1997) among

1288 others for Spanish, bare nominal arguments do not allow for a kind reading in these

1289 languages. For a kind reading to obtain, a definite determiner must be present. If this is

1290 empirically correct, then there are at least some þPL languages for which kind readings

1291 are not available with bare nominals. It is interesting to note that, to my knowledge, such a

1292 restriction has never been mentioned for any �PL language. And this is in fact a

1293 distinction that is straightforwardly predicted on the current approach. However, while

1294 the prediction has a desired result for the Romance languages, it turns out to be

1295 problematic for English. To accommodate the existence of kind readings for English

1296 bare nominals, it must assumed, on the present view, that English licenses some particular

1297 null operator whose role is to retrieve the meaning of a kind from that of a ‘predicative’

1298 NumP. Such an operator has, in fact, been proposed by Krifka (1995) and I reproduce his

1299 formulation below:

1300

1301 (46) [NP bears] s (lwlx9n(R/Tw (xBEAR) & OKU (BEAR)(x) ¼ n])

¼ i y 8 w 8x [9n (R/Tw( xBEAR ) & OKU (BEAR) (x) ¼ n] iff R/Tw (x,y)

13 Striclty speaking, the approach allows for yet another possibility, namely the existence of a singular

morpheme that has the capacity to saturate the number argument. Note that with such a morpheme, however,

bare nominals should manifest the same properties as nominals with an overt singular determiner in languages

like English or French. In particular, they should show the typical wide scope properties of indefinites, and not

the characteristic scopelessness of bare plurals.

V. Deprez / Lingua xxx (2004) xxx–xxx 23

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

The unique entity y such that for all x that are realizations of the kind BEAR

(in n number) in all worlds, they are realizations of the entity y

1313 As noted by Krifka, this denotation should be identical to the kind BEAR, as any

1314 realization of the kind BEAR will be in the extension of the predicate and vice versa.

1315 Assuming that such a null operator exists, however, raises a number of delicate questions.

1316 First, it is unclear what kind of evidence could bear on its existence. If we take into

1317 consideration what this operator translates into in the other þPL languages, the most

1318 obvious answer is that it corresponds to a particular interpretation of a definite

1319 determiner. That English and Romance definite determiners have distinct meanings

1320 has been argued in Dayal (2001). Dayal has also made interesting conjectures as to why in

1321 English the plural definite determiner fails to allow for a kind reading (see also

1322 Zamparelli, 2000). One could perhaps imagine that the null operator of English is the

1323 morphological counterpart of this distinct meaning. Heycock and Zamparelli (2002) have

1324 argued for the existence of a null definite in English that can be licensed under

1325 constrained abstract movement of nominal constituents to its Specifier. It could be,

1326 modifying somewhat Longobardi’s (1994) original parametric proposal, that English has

1327 a syntactic way of licensing this null determiner that is lacking in Italian and Romance.

1328 At this point of our research, however, evidence supporting such conjectures is missing

1329 and the question thus remains open. Yet, although I will not speculate further as to why

1330 English disallows bare definite plural and allows instead bare nominals as kind referring

1331 expressions, I would just like to point out that plural morphology seems again to be an

1332 important factor here. It is well known indeed that English differs from the Romance

1333 languages with respect to its agreeing capacity in the noun phrase. In particular, although

1334 in Romance number is always marked on the determiner, sometimes redundantly, this is

1335 not the case in English. In the present context, this difference suggests an interesting

1336 generalization:

(47) Only definite determiners overtly marked for plural allow for a (plural) kind

reading.

1342 This generalization definitely seems to play a role, but at this point in our research,

1343 it remains to be understood what this role can be. I leave this question for further

1344 research.

1345 5. Concluding remarks

1346 This paper has explored an alternative parametric approach to derive and predict the

1347 possible meanings and distribution of bare nominals in a cross-linguistic perspective. At

1348 the heart of this proposal is the Plural Parameter, a morphological parameter that controls

1349 the syntactic presence of a functional projection for Number. The parameter distinguishes

1350 two broad types of languages, the �PL and the þ PL languages, for which it makes a

1351 number of predictions that have been shown to be largely upheld across a variety of

24 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

1352 languages for BN in argument as well as in predicative positions. The main predictions of

1353 the model are again summarized below.

(1) Only �PL languages allow direct access to the kind denotation of bare nouns. Thus,

1356 the kind reading is predicted to never be missing for BN in �PL languages, as it is the

1357 most basic reading for BN in these languages. In þ PL languages in contrast, the

1358 existential and the generic readings built on a predicative NumP are the most basic

1359 ones. The kind reading, in contrast, requires the presence of a particular operator and

1360 could well be missing for bare nominals in þ PL languages.

(2) Only the �PL languages can have bare nominals that are underspecified for number

1362 and thus compatible with either a plural or a singular reading depending on

1363 contextual factors. In þ PL languages, in the absence of relevant morphology, bare

‘singular’ nominals—i.e., nominals whose NumP contains an unsaturated number

1365 argument—will be excluded. This should be true in both argument and predicative

1366 positions. Bare singulars can nonetheless be allowed in þPL languages, if relevant

1367 morphology is present. For instance, a language can feature a null definite determiner

1368 that like an overt one can saturate the number argument. Singular BN in this case will

1369 manifest distinct properties allowing definite readings and other readings built on

1370 definite denotations (see Dayal, 2001) but not the generalized existential and generic

1371 readings found commonly with plural BN. Conversely, there can also be þPL

1372 languages in which the plural morphology fails to appropriately saturate the number

1373 argument, as in French. In such cases, bare nominals will be largely excluded, and

1374 determiners always required to saturate the number argument contained in the

1375 obligatorily projected NumP.1376

1377 For stimulating comments and discussion that have greatly improved this work, I am

1378 gratefully indebted to Roberto Zamparelli, Veneeta Dayal, Karina Wilkinson and Richard

1379 Kayne. Its remaining shortcomings, of course, are my sole responsibility.

1380 Uncited reference

1381 Carlson and Pelletier (1995), Deprez (1999b), Kupferman (1991), Pollock (1983) and

1382 Zamparelli (1998).

References

1384 Allan, K., 1980. Nouns and countability. Language 56 (3), 541–567.

1385 Benedicto, E., 1997. Verb movement and its Effects on Determinerless Plural Subjects. Paper presented at

1386 LSRL.

1387 Bentolila, A., 1978. ‘Creole d’Haıti: Nature et Fonction- Fonction Naturelle’ Etudes Creoles. AUPELF,

1388 Montreal, pp. 65–76.

1389 Bosweld de Smet, L., 1997. On mass and plural quantification: the case of French des/du NP. Groeningen

1390 dissertation in Linguistics.

1391 Carlson, G., 1977. A unified analysis of the English bare plural. Linguistics and Philosophy 1, 413–456.

1392 Carlson, G., Pelletier, J. (Eds.), 1995. The Generic Book. The University of Chicago Press.

1393 Chierchia, G., 1998. Reference to kinds across languages. Natural Language Semantics 6 (4), 339–400.

V. Deprez / Lingua xxx (2004) xxx–xxx 25

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

1394 Cheng, L., Sybesma, R., 1999. Bare and not so bare nouns and the structure of NP. Linguistic Inquiry 30, 4.

1395 Chomsky, N., 1995. The Minimalist Program. MIT Press.

1396 Chomsky, N., 1999. Derivation by Phases. MIT Occasional Papers in Linguistics, vol. 18. Department of

1397 Linguistics. MIT

1398 Colban, E., 1991. Three Studies in Computational Linguistics. Doctoral Dissertation. Oslo University

1399 Corbett, G., 2000. Number. Cambridge University Press, Cambridge.

1400 Crisma, P., 1991. Functional Categories inside the NP: a Study on the Distribution of Nominal Modifiers. Tesi di

1401 laurea, University diVenezia.

1402 Damoiseau, R., 1987. Situation de communication et fonctionnement de la langue en Creole haitien. Etudes

1403 Creoles 10, 106–290.

1404 Damoiseau, R., 1996. «Les adjectivaux en creole haıtien», dans Materiaux pour l’etude des classes

1405 grammaticales dans les langues creoles. In: Daniel, V. (Ed.), Publications de l’Universite de Provence,

1406 pp. 150–161.

1407 Dayal, V., 2001. Number Marking and (In)definiteness in Kind Terms. Ms Rutgers University.

1408 DeGraff, M., 1992a. Creole Grammars and The Acquisition of Syntax: The Case of Haitian. Ph.D. University of

1409 Pennsylvania.

1410 DeGraff, M., 1992b. The syntax of predication in Haitian. In: Proceedings of NELS 22. GLSA, University of

1411 Massachusetts.

1412 Delfitto, D., Schroten, J., 1991. Bare plurals and the number affix in DP. Probus 3.2, 155–185.

1413 Deprez, V., forthcoming a. Determiner architecture and phrasal movement in French Lexifier Creoles. In:

1414 Proceedings of the Amsterdam Going Romance Workshop on Determiners. John Benjamins, Amsterdam.

1415 Deprez, V. forthcoming b, Haitian Creole se: a copula, a pronoun, both or neither. In: Adone, D. (Ed.), Creoles

1416 and Generative Grammar, John Benjamins, Amsterdam.

1417 Deprez, V., in press. Constraints on the meanings of bare nouns. In: Nunez-Cedeo, R., Lopez, L., Cameron, R.

1418 (Eds.), Current Issues in Romance Linguistics: Selected Papers from the 31st Linguistics Symposium on

1419 Romance Languages. John Benjamins, Amsterdam.

1420 Deprez, V., 2001. On the syntactic and semantic nature of Haitian Bare NPs. In: Cresti, D., Tortora, C.,

1421 Saterfield, T. (Eds.), Current Issues in Romance Linguistics: Selected Papers from the 29th Linguistics

1422 Symposium on Romance Languages. John Benjamins, Amsterdam.

1423 Deprez, V., 1999a. De la nature semantique des nominaux sans determinant en creole haitien. Langues 2.4,

1424 289–300. John Libbey Eurotext, Agence Universitaire de La Francophonie.

1425 Deprez, V., 1999b. Empirical quicksand or empirical smokescreen. Journal of Pidgin and Creole Languages 14

(2), 351–357. John Benjamin, Amsterdam.

1427 Deprez, V., Vinet, M.-T., 1997. Predicative constructions and functional categories in Haitian creole. Journal of

1428 Pidgin and Creole Languages 12.2, 1–32. John Benjamin, Amsterdam.

1429 Deprez, V., Vinet, M.T., 1992. Une structure predicative sans copule. La Revue Quebecoise de Linguistique 22

(1), 11–44.

1431 Dejean, Y., 1985. Ann Aprann otograf kreyol la, Port-au-Prince, Haiti.

1432 Ferere, G.A., 1974. Haitian Creole Sound System, Form classes, Text. Ph.D. Dissertation In Linguistics,

1433 University of Pennsylvania.

1434 Frantz L.J., 1988. La determination nominale en creole haıtien, These de Doctorat de 3emecycle, Universite Paris VII.

1435 Heycock, C., Zamparelli, R., 2002. Coordinated bare definites, Ms University of Edimburg and University of

1436 Bergamo. Linguistic Inquiry, in press.

1437 Krifka, M., 1995. Common nouns: a contrastive analysis of Chinese and English. In: Carlson, G., Pelletier, F.

1438 (Eds.), The Generic Book, Chicago Press.

1439 Kupferman, L., 1991. Structure evenementielle de l’alternance un/0 devant les noms humains attributes.

1440 Language 101, 52–75.

1441 Longobardi, G., 1999. How Comparative is Semantics. Ms. University of Trieste.

1442 Longobardi, G., 1994. Reference and proper names: a theory of N-movement in syntax and logical form.

1443 Linguistic Inquiry 25 (4), 609–665.

1444 Lyons, C., 1999. Definiteness. Cambridge University Press, Cambridge.

1445 Muller, A. 2000. The expression of Genericity in Brazilian Portuguese. In: Kuusmoto, K., Villalta, E. (Eds.),

1446 University of Massachusetts Occasional Papers 23: Issues in Semantics and Its Interface.

1447 Munn A., Schmitt C., this volume. Number and Indefinites.

26 V. Deprez / Lingua xxx (2004) xxx–xxx

UN

CO

RR

EC

TED

PR

OO

F

Lingua 1074 1–27

1448 Munn A., Schmitt. C., 2001. Bare nominals and the morphosyntax of number. In: Cresti, D., Tortora, C.,

1449 Saterfield, T. (Eds.), Current Issues in Romance Linguistics: Selected Papers from the 29th Linguistics

1450 Symposium on Romance Languages, Ann Arbor. John Benjamins, Amsterdam.

1451 Nerbonne, J., 1993. Semantics of nominal comparatives. In: Dekker, P., Stokhof, M. (Eds.), Proceedings of the

1452 Ninth Amsterdam Colloquium. ILLC-Publication, Amsterdam, Part II, pp. 487–506.

1453 Pollock, J.-Y., 1983. De quelques proprietes des phrases copulatives en francais. Langue Francaise 58, 89–125.

1454 Pompilus, P., 1976. Contribution a l’etude comparee du Creole et du Francais: Morphologie et Syntaxe. Edition

1455 Caraibes. Port-au-Prince. Haiti.

1456 Savain, R., 1993. Haitian-Kreol in Ten Steps. Schenkman Books Inc. Rochester. Vermont.

1457 Wingerdt-Philips, J., 1982. A Partial Grammar of the Haitian Creole Verb System: Forms, Function and Syntax.

1458 Ph.D. SUNY, Buffalo.

1459 Zamparelli, R., 1995. Layers in Nominal Structure. Ph.D. dissertation, University of Rochester.

1460 Zamparelli, R., 1998. A theory of kinds, partitives and of/z possessives. In: Wilder, C., Alexiadou, A. (Eds.),

1461 Possessors, Predicates and Movement in the Determiner Phrase. John Benjamins, Amsterdam.

1462 Zamparelli, R., 2000. Definite and Bare Kind-Denoting Noun Phrases. In: Beyssade, C., Bok-Bennema, R.,

1463 Drijkoningen, F., Monachesi, P. (Eds.), Romance Languages and Linguistic Theory. Selected Papers of

1464 Going Romance 2000. John Benjamins, Amsterdam.

V. Deprez / Lingua xxx (2004) xxx–xxx 27