The categorial status of determiners. The New Comparative Syntax ed. by Liliane Haegeman

The New Comparative Syntax Edited by

LILIANE HAEGEMAN

'sa LONGMAN LONDON AND NEW YORK

Addison Wesley Longman Limited Edinburgh Gate Harlow, Essex CM2O 2JE England

and Associated Companies throughout the world

Published in the United States of America by Addison Wesley Longman Inc., New York

Addison Wesley Longman 1997

All rights reserved; no part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the Publishers or a licence permitting restricted copying in the United Kingdom issued by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London WIP 9HE.

First published 1997

ISBN 0 582 27943-7 Paper ISBN 0 582 27942-9 Cased

British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library

Library of Congress Cataloging-in-Publication Data The new comparative syntax / edited by Liliane Haegeman.

p. cm. — (Longman linguistics Iibrary Includes bibliographical references and index. ISBN 0-582-27943-7. — ISBN 0-582-27942-9 (cased) 1. Grammar, Comparative and general—Syntax. I. Haegeman,

Liliane M. V. Il. Series. P291.N48 1997 415—dc21 96-39472

CIP

Set by 8 in 10/12 pt Times Produced by Longman Singapore Publishers (Pte) Ltd. Printed in Singapore

Table of Contents

List of Contributors viii Editor's preface and acknowledgements ix Publisher's acknowledgements

1 Introduction: on the interaction of theory and description in syntax 1

Haegeman

2 Subjects and clause structure 33 Anna Cardinaletti

3 Verb syntax in, and beyond, creolization 64 Michel DeGraff

4 The categoria) status of determiners 95 Giuliana Giusti

5 Romance causatives 124 Maria Teresa Guasti

6 Focus and the CP domain 145 Genoveva Puskas

7 Event nominais and the construct state 165 Tal Siloni

8 V°-to-I° movement and inflection for person in all tenses 189 Sten Vikner

9 Negation and verb movement 214 Raffaella Zanuttini

10 The Germanic SOV languages and the Universal Base Hypothesis 246 Jan-Wouter Zwart

11 A parametric approach to comparative syntax: properties of the pronominal system 268 Luigi Rizzi

lndex 287

vii

List of Contributors

Anna Cardinaletti, Department of Linguistic and Literary Studies, University of Venice, Italy Michel DeGraff, Department of Linguistics and Philosophy, MIT, USA Giuliana Giusti, Department of Linguistic and Literary Studies, University of Venice, Italy Maria Teresa Guasti, Department of Cognitive Science, Foundation San Raffaele, Milan, Italy Genoveva Puskas, Department of Linguistics, University of Geneva, Switzerland Luigi Rizzi, Department of Communication Science, University of Siena, Italy and Department of Linguistics, University of Geneva, Switzerland Tal Siloni, Department of Linguistics, Tel Aviv University, Israel Sten Vikner, Department of General and Germanic Linguistics, University of Stuttgart, Germany Raffaella Zanuttini, Department of Linguistics, Georgetown University, Washington, USA Jan-Wouter Zwart, Institute of General Linguistics, University of Groningen, The Netherlands

vili

Chapter 4

The categorial status of determiners1

Giuliana Giusti

1 Introduction

1.1 Functional categories in the noun phrase

It is currently assumed in the literature that the verb builds a lexical projection VP which reflects its argument structure and further projects a functional structure including AgrOP, TP, AgrSP, CP,2 etc. (cf. the introduction to this volume section 3.1). In a parallel fashion, the noun also builds a lexical projection NP, which reflects its argument structure, and further projects a functional structure which includes a certain number of functional heads. Some of these will be motivated in this section. We will refer to the constituent formed by the NP and its functional projections as `noun phrase'.

Hungarian provides evidence for the strict parallelism with both an inflection-like and a complementizer-like functional projection in the nominai structure. The for-mer will be called Agr(eement) P(hrase), due to the fact that in some language (e.g. Hungarian) it hosts Agreement morphology with a possessor, the latter will be called DP, due to the fact that its head is filled by a determiner (namely an article, as will be claimed in section 2).

Szabolcsi (1994) notes that the possessive argument of a N in Hungarian is marked with nominative Case and triggers person agreement with the noun, exactly as a subject does in relation to a verb:

(1) a. az en kalap-om the I hat-lst.s. `my hat'

b. a te kalap-od the you hat-2nd.s. 'your hat'

c. a Peter kalap-ja the Peter-Nom hat-3rd.s. `Peter's hat'

This argues for a functional head in the noun phrase (parallel to AgrS in the clause), where the agreement morphology on the noun is generated (or checked), and whose specifier is the nominative possessive.

95

Spee.""".

96 The New Comparative Syntax

(2) DP

D2 AgrP

Spec Agr'

/\

(N+)Age Peter kalap-ja

Possessors can also occupy a higher position inside the noun phrase in Hungarian. In addition to (3a), parallel to (1), we find (3b):

(3) a. a Mari kalap-ja the Mari-Nom hat-poss.3rd.s.

b. Mari-nak a kalap-ja Mari-DAT the hat-3rd.s. `Mary's hat'

Szabolcsi (1994) proposes that (3b) derives from (3a) by movement of the posses-sor from the nominative assigning position SpecAgrP to the dative assigning posi-tion SpecDP. The resulting structure is in (4):

(4) DP

Marij-nak a

tj

Agre A

.

Agr2

kal

‘,4‘

- ja

SpecDP is an operator position, as shown by the fact that the dative possessor, con-trary to the nominative possessor, can be moved out of the DP, as a focus (5) or as a wh-element (6):

(5) a. [Marinak]i Peter latta [t'i a ti kalap-ja-t]. Mari-DAT Peter-Nom saw the hat-poss.3rd.s.-ACC `Peter saw Mari's hat'

b. *[Mari]i Peter latta [t'i a ti kalap-ja-t].

(6) a. Kati ki-nek a kalap-ja-t latta? Kati-Nom who-DAT the hat-poss.3rd.s.-ACC saw Whose hat did Kati see?'

b. *a ki kalap-ja the who-NOM hat

Parametrized N-to-Agr movement (in overt syntax vs Logical Form) has been proposed by Cinque (1994) to account for the postnominal position of (some) adjec-tives in Romance and the prenominal position of the same adjectives in Germanic. In (7) we see an object denoting nominai and in (8) an event nominai:

The categorial status of determiners 97

(7) Evaluating Size Color a. Eng: the beautiful big red ball b. It: la bella grande palla rossa

(8) Manner Theme (Complement) a. Eng: the terrible Italian invasion of Albania b. It: la terribile invasione italiana dell'Albania

English (7a) and (8a) have the structure in (9a) while Italian (7b) and (8b) have the structure (9b):

(9) a. DP ...........--------

D2 AgrP

Spec Agr' AP

Agr2 AgrP

-----------\. Spec Agr' AP ......„.....-------

Agr° NP

Spec N' AP ...../----....., A N(' PP

i the beautiful big red ball .L. the terrible Italian invasion of Albania

b. DP

D2

Spec AP

la bella la

AgrP

Agr'

Agr2 AgrP

Spec Agr' AP

Agro NP

Spec N' AP

N2 PP

grande pallai rossa ti dell'Albania terribile invasioni italiana ti

N-movement to the D° position has been proposed by Grosu (1988) to account for the following Rumanian data:

(10) a. un frumos bàiat romàn a nice boy Rumanian

b. bàiatul frumos (cel romàn)3 boy-the nice (the Rumanian)


When an indefinite article is present as in (10a), the word order in the Rumanian DP is basically the same as in Italian, with some prenominal and some postnominal adjectives. According to Cinque (1994), in these cases N moves to an intermediate Agr. When the definite article is present and the noun is the host for the encliticiza-tion of the article, as in (10b), N is the leftmost element in the DP. N+Agr move-ment to D straightforwardly captures the word order in (10b), as shown in (11):

DP

D° AgrP

Spec

Agr

Agr'

AgrP Spec

____,• Agr'

AP Agr -NP N' N°

un frumos baiati roman Ci ti

bàit- ul frumos t"i [cel romani ei ti

The trigger for N-to-D movement in Rumanian is the enclitic article, given that this movement does not take piace when the indefinite article is present.

Many intermediate projections have been proposed in the literature, which will not be reviewed here (cf. Num(ber)P by Ritter (1991), W(ord marker)P by Harris (1991) and Bernstein (1993), K(ase)P by Giusti (1995) severa) AgrPs, one for each class of adjectives by Cinque (1994)). In this introductory section, we just want to emphasize that at least some functional heads are needed to host N-movement across APs, and some functional specifiers are needed to provide landing sites for movement of possessors (e.g. in Hungarian).

According to Abney (1987: 64f.) functional elements have the following properties:

(12) a. They constitute closed lexical classes. b. They are generally phonologically and morphologically dependent.

They are generally stressless, often clitics or affixes, and sometimes even phonologically null.

c. They permit only one complement, which is in generai not an argument. d. They are usually inseparable from their complement. e. They lack `descriptive content'. Their semantic contribution is second-

order, regulating or contributing to the interpretation of their comple-ment. They mark grammatica) or relational features, rather than picking out a class of objects.

Each of these properties constitutes a tendency, not a criterion, to decide whether an, element is lexical or functional. (12a) refers to the fact that, for example, compie-mentizers, determiners, prepositions constitute closed classes. (12b) refers to the fact that articles in many languages are stressless (e.g. in Italian) or clitic (e.g. th

The categoria! status of determiners 99

enclitic articles discussed above). In (12c), it is stated that functional heads do not have a selectional frame such as the one assumed for lexical elements. They embed a unique category as a complement. For example, in the clause C embeds an AgrSP, AgrS embeds a TP, etc.; in the noun phrase D embeds an AgrP, etc. (12d) refers to the fact that, for example, a complementizer cannot be left in piace while the rest of the clause is topicalized, as in the Italian (13) below:

(13) a. Che Gianni è un cretino, ho detto! That Gianni is an idiot [I] said

b. *Gianni è un cretino, ho detto che!

Finally, (12e) captures the traditional characterization of functional elements which are said to be void of meaning. For example, the typical functional preposition is of, It. di etc. This preposition can introduce any kind of argument of the noun: posses-sor (14a), agent (14b), theme (14c), experiencer (14d):

(14) a. il libro di Gianni `Gianni's book'

b. la telefonata di Gianni `Gianni's cali'

c. il furto dell'icona `the theft of the icon'

d. il dispiacere di Gianni `Gianni's regret'

Grimshaw (1991) proposes a more principled conception of functional projections as the 'extended projection' of the lexical category they embed. According to her, functional categories do not select a complement in the sense lexical categories do. On the contrary, they are projected by (inflectional) properties of the lexical head they embed. In this way she provides an explanation for the descriptive statement (12e). Grimshaw also derives the stipulative (12c) from the fact that thematic roles are assigned to the whole extended projection, which is for this reason, impossible to break.

1.2 Multiple prenominal elements and the structure of the DP

In this chapter we consider those elements that have often been treated as repre-sentative of the class of determiners: articles, demonstratives and quantifiers. In the current literature it is usually assumed that all these prenominal elements belong to the category of determiners and hence occupy the D° position, (one of) the top func-tional projection(s) of the nominai projection (see introduction to this volume, sec-tion 4.3). Trivial evidence in favour of such a view is the complementary distribution of demonstratives, articles and quantifiers in the first position in the nominai string (15).

(15) these/the/many students

However, the complementary distribution found in English is by no means universal.


In Rumanian the demonstrative can be in second position following a noun which has incorporated into the definite article:

(16) a. un frumos bàiat a nice boy

b. bàiatul acesta frumos boy-the this nice `this nice boy'

In many languages, including English, quantifiers cooccur with articles and demon-stratives in quite a complex way. Some quantifiers behave like adjectives in being preceded by an article or a demonstrative (17a); others behave like independent heads taking a full extended nominai projection as their complement with a definite article or demonstrative (17b); finally some others cannot cooccur with any deter-miner at all (17c):

(17) a. these/the many (*these/the) boys b. (*these/the) all these/the boys c. (*these/the) several (*these/the) boys

Data such as (16) and (17) above make a unified treatment of the prenominal elements in (15) less appealing than it might have looked like at first sight. A unified analysis of prenominal elements makes it difficult to provide a coherent analysis of the systematic cross-linguistic contrasta between the different prenominal elements such as articles, demonstratives and quantifiers. The cross-linguistic parallelisms between articles, for instance, and the generai cross-linguistic contrast with demonstratives would be hard to express if we were to assume that both articles and demonstratives occupy D.

1.3 Aim of the chapter

The Purpose of this chapter is to re-examine the status of the various prenominal ele-ments in (15), which are standardly referred to as `determiners'. I will consider articles, demonstratives and quantifiers in turn. I will propose that these elements do not constitute a homogeneous category. Rather, I will assign each of the distinct types to a different category. In particular, I will argue that only articles are extended heads of the noun phrase; demonstratives are lexical elements and occupy specifier positions; quantifiers are also lexical elements: they can either be adjectives (when they are preceded by another determiner) or lexical heads that select a full DP, that is a noun phrase with its complete extended projection, as its complement.

1.4 The need for a comparative approach

The comparative approach to the analysis of determiners plays a crucial role in our research. For many of the properties we are concerned with, data drawn from one singie language cannot give us a full insight into ali the relevant properties. Two examples have already been given above. One was the review of Cinque's (1994) proposal which derives the different word order in noun phrases. A second example

The categoria) status of determiners 101

of the need of comparative research in order to better understand generai pheno-mena is the syntax of the enclitic article in Rumanian. In section 1.1. it has been assumed that the need for N-to-D movement in overt syntax is related to the pres-ence of a definite article (which is a clitic) in D°.

At first sight, in fact, N-to-D movement is not found in Italian, where the noun follows the determiner:

(18) a. la ragazza the giri

b. *ragazza-la girl-the

One might be tempted to conclude that N-to-D movement is irrelevant for the Italian DP. However, observe the variable word orders in (19) which correlate with the presence vs absence of the article with proper names:

(19) a. il mio Gianni the my Gianni

b. Gianni mio Gianni my my Gianni'

The incorporation hypothesis established in (11) on the basis of Rumanian data gives us ari insight into the varying word order in (19). Longobardi (1994) proposes that (19a) reflects the underlying word order and that in (19b) the N Gianni moves to D°. The only difference with respect to the Rumanian data is the trigger of N movement to D: in Italian, it is the property of the proper name to be intrinsically referential and to check its [+Ref] feature in D° in overt syntax; in Rumanian, it is the property of the enclitic article to trigger N-to-D movement in order to have a proper host to encliticize on.

These comparative data suggest a unified treatment of the DP not only in terms of the internai structure but also in terms of movement. Just as was the case for the V-movement as discussed in the introduction to this volume, we might say that univer-sally N moves to D. In some languages this happens in overt syntax (cf. Rumanian (11)), while in others the movement is procrastinated and takes piace at Logical Form. Again, we then interpret the variation in terms of strength of the features.

In the discussion below we will give evidence for the importance of a compara-tive approach to the study of the DP. The data to be discussed in this chapter are drawn from the following languages: Albanian, Bulgarian, Danish, English, Greenlandic Eskimo, French, Hebrew, Irish, Modern Greek, Italian, Kiswahili, Niuean, Norwegian, Rumanian, Spanish, Swedish, Southern Tiwa.

1.5 Organization of the chapter

In each section of this chapter one type of determiner will be discussed. Section 2 deals with articles, section 3 deals with demonstratives, section 4 deals with quantifiers and section 5 is the conclusion.


2 Articles

2.1 Articles form a closed class

Some languages have no article (most Slavic languages, Kiswahili, Latin). Others have only one (usually the definite one, cf. Bulgarian, Modern Greek), or two: the definite and the indefinite one (Romance and Germanic languages). Some lan-guages also have a special article for generics and proper names,4 or differentiate the definite article with respect to proximity to the discourse participants.5

2.2 Articles are phonologically and morphologically dependent on the noun or on other nominai elements

In some languages articles are clitics, Rumanian examples seen above. In other lan-guages they are unstressed elements and phonologically cliticize to the following word displaying some allomorphs, the occurrence of which depends on the initial sound of the following word, as in Italian:

(20) a. il ragazzo, lo scolaro, l'amico

b. la ragazza, la scolara, l'amica the boy/girl the student.M/F the friend.M/F

Articles are trivially morphologically dependent on the head noun in all languages with noun—article agreement for gender, number and case. An even stricter depen-dency is found in languages like Bulgarian, Rumanian and Albanian, in which the form of the article changes according to the class of the noun stem:

(21) a. Bul: momce-te, xora-ta `the boys, the people'

b. Rum: balat-ul, frate-le `the boy, the brother'

c. Alb: dial-i, shok-u `the boy, the friend'

In Bulgarian (21a), momce and xora are both masculine plural nouns,6 while Rumanian bàiat and frate (21b) and Albanian diale and shok (21c) are masculine singular nouns. The different morphological form of the article uniquely depends on the morpho-phonological properties of the root. This suggests that the article is in fact part of the inflectional morphology of the head noun, and not an independent lexical element.

2.3 Articles are strictly inseparable from their complement

Articles do not give rise to floating constructions, parallel to demonstratives but contrary to (some) quantifiers (cf. section 3 below), as shown by the contrast in Italian (22):


(22) a. Ragazzi, ne sono arrivati molti Boys, CL have arrived many

b. *Ragazzi, (ne) sono arrivati i/questi Boys, (CL) have arrived the/these

More crucially, contrary to both demonstratives and quantifiers, they cannot pro-nominalize; in other words, their complement cannot be empty, as shown by the Italian contrast (23):

(23) a. Ho visto il/un *(ragazzo). I saw the/a *(boy)

b. Ne ho visto uno. CL [I] saw one

c. Ho visto quello. [I] saw that

2.4 Articles are extended nominai heads (in the sense of Grimshaw 1991)

In Abney's (1987) terms, articles take an NP or an extended projection of it as their unique complement.

Apparent evidence that articles appear on adjectives, prepositions, and other nominai modifiers in some languages actually supports the claim that they are func-tional heads, under the assumption that nominai modifiers project their own extended projection which may happen to be morphologically identical to the nominai projec-tion. As a matter of fact, in many languages the declensions of nominai and adjectival (and prepositional) articles overlap only partially in the paradigm.

As an example, Hebrew adjectives are preceded by a definite article if the noun also is:

(24) a. ha-bayit ha-gadol the house the big `the big house'

b. bayit gadol house big à big house'

In Rumanian, an article (encliticized on the dummy preposition a-) must appear on a genitival phrase if this is not adjacent to a nominai definite article (Grosu 1988):

(25) a. potretul (*al) unei fete portait-the a.Gen giri

b. portetul frumos *(al) unei fete portrait-the beautiful A-the.Masc a.Gen giri

c. un portret *(al) unei fete a portrait A-the a.Gen giri


For Hebrew, the apparent generalization in (24) is that adjectives agree for 'definite-ness' with the noun they modify. This is not true for Rumanian genitival articles, as shown by the contrast between (25a) and (25c).

Furthermore, in Albanian adjectival articles are inserted only depending on the properties of the adjectival stem and independently of the definiteness of the noun phrase they modify:

(26) a. nje djale i mire, a boy the good, à good boy',

b. djali i mire boy-the the good

nje diale besnik a boy faithful à faithful boy' djali besnik boy-the faithful

The data in (26) suggest that, contrary to other determiners, articles are not inserted for semantic, but rather for grammatical (morpho-syntactic) reasons.

Further evidence for the functional (syntactic) nature of the (definite) article is provided by the following Rumanian vs Italian contrast. In Italian the masculine singular form of the indefinite quantifiers uno (òne'), nessuno (`no one, none')

lacks the -o ending in (27) where it appears adjacent to the noun phrase. When it appears far from the noun phrase (in this case the clitic) in (28) or as a pronoun in (29), this ending must be overt:

(27) a. un(*o) ragazzo à boy'

b. nessun(*o) ragazzo `no boy'

(28) a. (ne) ho visto un*(o) `[I] saw one'

b. non (ne) ho visto nessun*(o)7 [I] saw none'

(29) a. un*(o) ha detto che ... someone said that ...'

b. nessun*(o) è venuto `none arrived'

The -o ending is a marker for (D-features (masc. sing.), which are recoverable when the quantifier is adjacent to the noun phrase, but which must be overt when the noun is distant from the quantifier as in (28) or is absent as in (29). (30)—(31) show the same phenomenon in Rumanian. Interestingly, the relevant ending in the nomina-tive/accusative form is homophonous to the definite article:

(30) a. un(*ul) bàiat à boy'

b. nici un(*ul) bàiat `no boy'


(31) a. Un*(ul) a spus c'à one said that ...

b. Nici un*(ul) a spus c'à No one said that ...

The same is the case with other quantifiers such as tot(ul) intreg(ul) whole', vreun(ul) 'some', alt(ul) lan]other', mult(ul) `much', putin(ul) whose Italian counterparts display the word marker even in the non-pronominal form (cf. Italian tutto àli', un altro ànother' , molto `much', poco `little', etc.). The article -ul in Rumanian, at least when it appears on quantifiers, is the lexical realization of I- features. Crucially, these (1)-features do not include definiteness, since -ul appears on existential quantifiers.

We can conclude that the definite article, at least in some languages, behaves like part of the inflectional morphology of the lexical head (in this case the quantifier).

2.5 Articles have no semantic value

A definite article may be present in a noun phrase with non-specific interpretation, as in (32a). This is not the case for a demonstrative, as in (32b):

(32) a. Scommetto che non troverai mai la segretaria di un onorevole che sia disposta a testimoniare contro di lui. I bet you'll never find the secretary of a deputy who is-SUBJ willing to witness against him

b. *Scommetto che non troverai mai questa segretaria di un onorevole che sia disposta a testimoniare contro di lui. I bet you'll never find this secretary of a deputy who is-SUBJ willing to witness against him

The noun phrase la segretaria di un onorevole must be analysed as non-specific when it takes a relative clause in subjunctive mood. Such a relative clause is ungrammatical when a demonstrative is present, as in (32b). The non-specific interpretation of the noun phrase in (32a) is triggered by the presence of the indefinite genitive. In fact, as soon as the embedded genitive is turned into a definite noun phrase, the subjunctive relative clause becomes impossible again, as in (32c):

(32) c. *Scommetto che non troverai mai la segretaria di quell'onorevole che sia disposta a testimoniare contro di lui. I bet you'll never be able to find the secretary of that deputy who is-SUBJ

willing to witness against him

Conversely, a definite description may Jack a definite article, in certain cases, cf. English at home, at school, which certainly refer to a definite piace, namely the one which is salient in the discourse.

In Rumanian, the absence of the article in prepositional phrases is a very produc-tive phenomenon and requires a syntactic analysis. All prepositions except cu


(` with') enforce the absence of the definite article and only of the definite article in unmodified nominals in the singular. An indefinite article, a demonstrative or a quantifier are allowed:

(33) a. Merg la profesor(*ul). I'm going to professor-(*the)

b. Merg la un/acest/vreun profesor. Ì'm going to [the]/a/this/some professor['s piace]'

The absence of the definite article in (33a) would be quite mysterious if the unique trigger of the insertion of this element was its definite interpretation. On the other hand, if we consider the definite article as the realizer of a bundle of (19-features, including abstract Case, we can conjecture that the presence of the preposition is sufficient to make these features recoverable and to dispense with the insertion of the article. These features become necessary again when a modifier is present, like an adjective in (34a) or a possessive in (34b) or a PP in (34c):

(34) a. Merg la profesor*(u1) mare. I'm going to the eminent professor['s piace]

b. Merg la profesor*(u1) Cau/Mariei. I'm going to your/Maria's professor['s piace]

c. Merg la profesor*(u1) de la Paris. I'm going to [the piace of] the professor from Paris

I propose that the relation between the preposition and the modifier of the noun is not a direct one, but is mediated through the relationship between the preposition and the functional head of the projection where the modifier occupies the Spec posi-tion. Because of that, it is not sufficient to licene the (D-features on the adjective. The article, on the contrary, is a functional head of the noun phrase. It is in the proper relation to the functional head in whose specifier the adjective is inserted in order for noun—adjective agreement to take place.8

Whatever the correct analysis of the contrast in (33)—(34) is, this phenomenon provides evidence for a syntactic analysis of articles that is not shared by other determiners.

2.6 Some conclusions

The result that articles are not (only) inserted on semantic grounds is welcome given that, differently from demonstratives and quantifiers, they do not have a straightfor-ward distribution across languages. Furthermore, in languages in which previous stages have been recorded, such as those in the Indo-European family, articles have developed at some intermediate stage, usually to make up for loss of inflectional morphology on the noun and/or on adjectives (cf. Giusti 1995 for further discussion).

Note also that the different interpretation that we derive from the (non-)occur-rence of a definite article is very much like the different interpretation that we get


in some languages with the accusative/partitive distinction in object position (cf. de Hoop 1992; Enc 1991, a.o.). Examples (35)—(36) are taken from En( (1991: 4):

(35) Ali bir piyano-yu kiralamak istiyor. Ali one piano-Acc to-rent wants Àli wants to rent a certain piano.'

(36) Ali bir piyano kiralamak istiyor. Ali one piano to-rent wants Àli wants to rent a (nonspecific) piano.'

In sum, it can be safely assumed that articles are functional heads. Languages vary with respect to whether they are found in the extended projection of nouns, articles or prepositions.9

3 Demonstratives

In this section we will see that the arguments for considering articles as functional heads do not extend to demonstratives.

3.1 Demonstratives cooccur with a definite article

The cooccurrence of demonstrative and definite articles show that they cannot occupy the same position. Rumanian presents two constructions: (37) with a phrase initial demonstrative in complementary distribution with an article, and (38) with phrase initial N, inflected with the enclitic definite article and the demonstrative in second position:1°

(37) a. acest/acel (frumos) bàiat (frumos) this/that (nice) boy (nice)

b. **acestul bàiat this-the boy

c. **acest bàiatul this boy-the

(38) a. bàiatul acesta/acela (frumos) boy-il thisA/thatA (nice)

b. frumosul (*acesta) bàiat c. bàiatul frumos (*acesta) d. frumosul bàiat (*acesta)

Let us first consider the more èxotic' construction (38). The structure of (38a) is given in (39):

108 The New Comparative

(39)

D

Syntax

DP

AgrP

Spec Agr' Dem

Agro AgrP

Spec Agr' AP

Agr2 NP 14' 14°

bàiati- ul acesta t 'i frumos t'i ti

In (39) N-to-D movement crosses the demonstrative as well as the adjective, show-ing that the demonstrative is neither in D (which is occupied by N+Art) nor in an intermediate extended nominai head, otherwise it would be picked up by the strictly local N-movement.11 Furthermore, (38b) shows that the presence of a demonstrative blocks AP-movement to SpecDP, which is otherwise possible in Rumanian.

A unified analysis of the demonstrative in (37)—(38) must assume demonstratives in Rumanian to be in Spec positions in all cases. In (37), the demonstrative is in the highest Spec and in complementary distribution with an article. I propose that it has moved to SpecDP. Once SpecDP is filled with an element that has enough features to license the whole projection, no article needs to be inserted, as will be motivated in 2.1.1 below. The structure of (37a) is given in (40):

(40) DP

Spec D'

D AgrP

Spec Agr' Dem

Agro AgrP

Spec Agr' AP

Agr° AgrP

Spec Agr'

acestj t. (frumos) (frumos) ti

The demonstrative checks its referential features in SpecDP. This movement is obligatory in Logical Form, and it may be procrastinated until this level in some languages, among which is Rumanian. Movement of the AP to SpecDP is ruled out in (38b) because it fills the landing site of Dem movement, giving rise to an unallowed structure at Logical Form.

In Rumanian, the demonstrative behaves like a high modifier of the noun, in that it can be preceded by N+Art, but it cannot be preceded by any AR In other lan-guages, demonstratives are found in lower positions.

The categoria) status of determiners 109

In Modern Greek the demonstrative can be found in severa) different positions from the extreme left of the noun phrase where it precedes the article and occupies SpecDP (41a) to the extreme right of the noun phrase which I propose to be in the Spec of the lowest AgrP (41c). In ali cases the article must appear, showing that the article and the demonstrative never competes for the same position as the article:

(41) a. afto to oreo to vivlio this the good the book

b. to oreo afto to vivlio the good this the book

c. to oreo to vivlio afto the good the book this `this good book'

It is possible to unify the analysis of the distribution of the demonstrative in Rumanian and in Modern Greek by assuming that in both languages the demonstra-tive starts from a very low Spec and then it moves to a high Spec before reaching SpecDP where it must land at the latest at Logical Form. The only difference between the two languages with respect to this phenomenon is that in Rumanian no article appears in case this movement takes piace in overt syntax.12

In the present proposal, two crucial assumptions require a more principled explanation: why is it the case that in some languages an element in SpecDP is in complementary distribution with an article in D°?; why do demonstratives move optionally or obligatorily to SpecDP and from exactly which position? I will now address these two questions in turn.

3.1.1 Doubly-filled DPs

The complementary distribution of an element in Specifier position with a func-tional element in the head, reminds us of the `doubly-filled COMP Filter' of Chomsky and Lasnik (1977), which is notably not universal.13 A strict parallelism between DP and CP does not surprise us in view of the fact that both are topmost extended projections (of the clause and the noun phrase respectively). The same parametrized principle that rules the cooccurrence of lexical elements in the Spec and head of CP also applies to DP.

Let us assume that a functional projection is instantiated in order to realize some feature (1), and that this feature must be `visible' in order to be properly interpreted in Logical Form. Let us also assume that the relevant relation to the satisfaction of this visibility condition imposed on functional features is universally the Spec-head relation: if the Spec is empty or does not have a `strong' specification for the rele-vant feature, the head must be filled. Otherwise the head can (and by minimalist requirements must) be abstract. Under these quite minima) assumptions, variation across languages can be reduced to variation across (inflectional) morphological systems. In other words, if an element in Spec makes the relevant features (morpho-logically) `visible', the corresponding head in agreement with it will be empty. On the other hand, if the relevant features are not morphologically `visible', or if the


Spec is empty tout court, the relevant head must be inserted in order for the projec-tion to be properly interpreted in Logical Form.14

An independent assumption to be made is that languages vary with respect to the level at which the demonstrative moves to SpecDP (its final destination).15

Let us now review some evidence in favour of this proposal. Well-studied lan-guages such as English, which have one or more articles and no cooccurrence of the article with the demonstrative, display the demonstrative in first position. For the present proposal, this means that the demonstrative obligatorily moves to SpecDP in overt syntax in these languages,16 and that it has a strong specification for the rele-vant features.'7 This is not the case in Modern Greek (41) where the article is always necessary. In Rumanian, when the demonstrative is in SpecDP in overt syntax, it makes insertion of the article unnecessary and, by minimalist assumptions, impos-sible. The procrastination of movement of the demonstrative to SpecDP makes the article obligatory to make the features in DP visible in the overt syntax.

3.1.2 The basic position of the demonstrative

Having established that the demonstrative is base-generated in a Specifier which is lower than the article, we must further specify the level of the nominai functional structure to which this Spec position belongs.

Brugè (1994) shows that the demonstrative is the lowest modifier of the noun phrase in that it follows all adjectives (even thematic ones) but precedes all PPs. A partial structure is given in (42d):

(42) a. la reacción alemana esa a las crfticas the reaction German this to the criticisms `This German reaction to the criticisms'

b. *la reacción esa alemana (as las crfticas) c. *la reacción alemana a las crfticas esa d. [DP la [Agrp reaccióni [Agrp alemana t'i [Agrp esa [NP ti [PP a las críticas]]]]]]

Brugè suggests that the position of the demonstrative in Spanish is to be taken as the basic position for this kind of element in UG. This is actually supported by cross-linguistic data in unrelated languages. Ernst (1992), for example, gives the following word order for noun phrases in Irish: Art + NumP + N + AP* + Dem + GenP + Re1C1. Notice that when the demonstrative is present, a definite article is required (43a), whereas when a genitive phrase is present, the definite article is absent (43b):

(43) a. *(an) fear seo the man this

b. (*an) hata an fhir hat the man `the man's hat'

Here I will not attempt an explanation for the contrast in (43). I will just note that, on the one hand, the obligatory cooccurrence of article and demonstrative confirms


the proposal put forth here that they belong to different categories and, on the other hand, the incompatibility of the article and a genitive phrase in a language in which they clearly do not appear in the same structural position shows that the comple-mentarity of two elements cannot be taken as evidence for their insertion in the same structural position or for their unified categorial status.

3.2 Opacity effects

The analysis of demonstratives in SpecDP in English-type languages can capture in a nove) way the opacity effects triggered by these elements. In Italian, for example, it is not enough to say that definiteness triggers opacity effects in the noun phrase, since it is possible to extract and to have wide scope of a quantifier contained in a genitive complement of a noun phrase introduced by a definite article:

(44) a. Di chi hai la foto sulla tua scrivania? of whom do you have the picture on your desk whose picture do you have on your desk?'

b. Ho la foto di ogni mio nipote sulla mia scrivania. I have the picture of every grandchild of mine on my desk Ì have the picture of every grandchild of mine on my desk.'

Under an analysis of definite articles as heads, the grammatica) status of (44a) and the possibility of wide scope of interpretation of ogni (` each') in (44b) is predicted. If demonstratives are analysed as specifiers, contrary to definite articles, we also expect parallel cases to be excluded, as is the case. (45a) is out, while in (45b) the quantifier can only have narrow scope:

(45) a. *Di chi hai questa foto sulla tua scrivania? of whom do you have this picture on your desk?

b. Ho questa foto di ogni mio nipote sulla mia scrivania I have this picture of each grandchild of mine on my desk

Up to this point, I have shown that the syntax of demonstratives in a certain number of unrelated languages argues against a trivial identification of this kind of nominai modifier with articles. Let us now consider their behaviour in the light of the diag-nostics in (12).

3.3 Demonstratives have a semantic value

Although demonstratives are parallel to articles in that they lack `descriptive con-tent', they are crucial for the interpretation of the referential index of the noun phrase. More crucially, even if they build a closed class in the sense that they are a very small number and it is impossible to make nove) ones, they belong to the broad semantic field of deixis which includes adverbials, pronominals, possibly aspect morphemes. This is a highly heterogeneous class that includes lexical (e.g., adver-bials) as well as functional (e.g., aspect morphemes) categories. Belonging to a closed class, therefore, is no evidence for the functional status of demonstratives.


3.4 Demonstratives are neither phonologically nor morphologically dependent

This holds at least in the languages that we have been looking at, and which display such properties for the (definite) article. As is shown by the fact that they can appear independently from the presence of the noun they modify, see I want this and that,

these and those, etc. (See also the discussion of (22)—(23) above.)

3.5 The behaviour of demonstratives in languages lacking articles

Kiswahili is one language which has demonstratives and lacks articles (see Carstens 1991). In this language, the DP constituents have the following distribution. Adjectives are only postnominal (46), demonstratives are either postnominal or prenominal (47):

(46) a. magari marya car new

b. *marya magari

(47) a. kitabu hiki book this

b. hiki kitabu this book

When a noun precedes the adjective, N has moved leftwards, to a higher position:

D. This means that the adjective cannot be a head. A demonstrative in postnominal position always precedes the arguments of the noun:

(48) a. *usomaji wangu huu wa esimu study my this of linguistics

b. usomaji huu wangu wa esimu study this my of linguistics

but it can be preceded or followed by adjectives:

(49) a. shati hili langu zuri shirt this my good

b. shati langu hili zuri shirt my this good

c. shati langu zuri hili shirt my good this

The basic argument brought about by Carstens in analysing the demonstrative in Kiswahili is the same as the one independently presented for Rumanian above. When the demonstrative is postnominal, it can occupy a position similar to that of adjectives; importantly, if the postnominal order is due to N-movement, then the demonstrative cannot be a head: based on the Head Movement Constraint (see note


11), an intervening head should block the N-movement. When the demonstrative precedes the N, Carstens proposes that it has moved to SpecDP.

Notice, finally, that demonstratives can be left in piace by noun incorporation into the verb in incorporating languages. Baker and Hale (1990) note that in Niuean (an Oceanic language) Ps cannot be stranded by noun movement, while in Greenlandic Eskimo the noun first incorporates into the preposition and then the combination of the two incorporates into the verb, supporting the hypothesis of strict locality in X°-movement:

(50) a. Ne tutala a au ke he tau tagata. past-talk abs-I to pl-person. Ì was talking to (the people).'

b. *Ne tutala tagatai a au [pp ke he [NP ti]]. past-talk-person abs-I to Ì was people-talking (to)'

(51) Qaqqa-nu-kar-put. mountain-to-go-3pS `They went to the mountains.'

In Southern Tiwa (data from Allen et al. 1984), demonstratives can be left in piace by noun incorporation, showing that they are not heads:'8

(52) a. [Yede seuan-ide] a-mu-ban that man-suf 2sS/A-see-past `You saw that man'

b. [DP Yede [NP [Ni e]] a-seuani-mu-ban. that 2sS/A-man-see-past

You saw that man'

Our approach predicts that in languages in which adjectives are strandable, demon-stratives also are.19


In this section, it has been argued that demonstratives are lexical elements inserted in a low specifier and further moved to SpecDP (through an intermediate position in a high SpecAgrP immediately lower than D). Being in a specifier position, demon-stratives must be taken to have the status of maximal projections. I will leave open the question of what kind of category they are; in other words, whether we must assume a new category Ind(exical), or we could take demonstratives as A(djectives), since they are modifiers of the noun. Another open question is whether bare demon-stratives are DPs with SpecDP filled by the demonstrative and the rest of the struc-ture empty, or whether they simply are the portion of the structure which appears in SpecDP when they are followed by a noun, namely the extended projection of IndP or AP, according to the choice we operate on the previous question. In any case, it is clear that demonstratives are not in D°, contrary to articles.


4 Quantifiers

Quantifiers, different from demonstratives, are strandable in a larger number of lan-guages. This supports a different treatment for this class of determiners, which will be pursued in the next section.

4.1 A taxonomy

At a descriptive level, quantifiers should be divided into three different classes: those that must precede an article (53), those that may follow an article (54), and those that can neither precede nor follow an article (55):20

(53) a. tutti *(i) ragazzi àli the children'

b. *i tutti ragazzi c. li ho visti tutti

Ì saw them all'

(54) a. molti 0*i ragazzi `many boys'

b. i molti ragazzi `the many boys'

c. ne ho visti molti/*ne ho visti i molti Ì saw many of them'

(55) a. alcuni/*i ragazzi `some boys'

b. *gli alcuni ragazzi c. ne ho visti alcuni

Ì saw some of them'

In Giusti (1991a), it is proposed that quantifiers that are not preceded by a definite article such as tutti in (53a), molti in (54a) and alcuni in (55a) are external to DP. They are heads (Q), selecting a DP and projecting a QP, as in (56a) while quan-tifiers that are preceded by a determiner are adjectives in a high specifier, as in (56b):

(56) a.

Spec

QP

(2.2

tutti all

Q'

DP

i ragazzi the boys


b.

D

DP

AgrP

Spec AP

Agro

Agr'

NP

i due ragazzi; simpatici the two boys nice

Spec N' AP

N° ./. i

ti

This proposal is supported by the possibility of extracting a pronoun (of category DP), leaving the quantifier in piace, as in (53c), (54c), (55c). The impossibility of a following article in (54a) and (55a) is due to the selectional properties of these quantifiers that require the noun phrase to be indefinite.21

Being the quantifier external to the noun phrase, it is expected that an article can-not precede it, as is the case for tutti in (53b) and alcuni in (55b). The unexpected possibility of (54b) suggests that molti has the ambiguous status of a Q and of a `quantitative adjective'. Quantifiers preceded by an article do not display selectional properties over the noun phrase and behave as regular adjectives.

Assuming a different structural position for Qs and quantitative As, it is expected that they give different results with respect to floating (cf. section 4.5 below). Qs are expected to allow their complement to move out leaving them in piace, much in the same way as Ns and Vs do. On the other hand, we do not expect this to be possible for quantitative As, since they are internai to the DP and, we assume, movement cannot affect chunks of an extended projection.

Suggestive evidence in favour of this proposal is found in Rumanian (57) which displays synonymous lexical items specialized for the Q (amindoi) and the adjectival function (ambii), as argued in detail in Giusti (1991b, 1994). In Rumanian, the defin-ite article is enclitic either on the noun or on a prenominal adjective. In section 1.1 above, we have shown that a noun to which an article has encliticized is in D° with SpecDP empty. It is therefore the leftmost element in the noun phrase. There is only one exception: namely when the noun phrase is preceded by a universal Q tofi, parallel to Italian tutti (àli') shown above, and dual amindoi (`both'), as shown in (57). On the other hand, ambii carries the enclitic article on itself, like prenominal adjectives, as shown in (58):

(57) a. toti/amindoi bàielii all/both children-the

b. amindoii bàieli all-the/both-the children

(58) a. ambii/frumo5ii bàieli both-the/nice-the children

b. *ambirfrumo5i bàielii both/nice children-the


Floating is allowed with the Q amindoi, as well as with the Q toli and prohibited

with the adjective ambii, as well as with the adjective frumosi:

(59) a. Bàie1ii am vàzut pe toti/amIndoi children-the [I] have seen all/both

b. *Baieti(i) am vàzut pe ambi(i)/frumo5i(i) children-(the) I saw both/nice

After this brief discussion, there are already many reasons to believe that quanti-fiers are not functional heads with respect to the diagnostics in (12). We will provide some futher data supporting this analysis in the rest of this section.

4.2 Quantifiers are lexical categories

Quantitative adjectives are certainly lexical heads, probably with their own func-tional extended projection, since they usually inflect for number and gender in agreement with the features of the noun they modify, much in the same way as other adjectives. Qs proper have complex selectional properties and select full extended nominai projections which have case (cf. the (trace of the) clitics li and ne in (53)—(55) above). They also have their own functional projection, since they are also inflected for nominai features.22

4.3 Quantifiers do not form a closed class

It is possible to make novel quantifiers (both Qs and quantitative As) from descrip-tive adjectives,23 cf. Italian: numerosi (`numerous'), altri (òther'), diversi (` differ-

elle), vari, etc., all of which are undoubtedly descriptive adjectives since (except for altro) they can appear in postnominal position (60). They can also be quantitative adjectives, and in this case they can only appear in prenominal position (61) pre-ceded by a determiner. Finally, they can be Qs and in this case they are in the first position in the noun phrase and cannot be preceded by a determiner (nor followed, since they are indefinite Qs), as in (62):

(60) a. Ha una famiglia numerosa [(s)he] has a big family

b. Ha un'opinione diversa dalla tua [(s)he] has a point of view different from yours

c. Una dieta varia fa bene alla salute a varied diet is healthy

(61) a. Conosco i tuoi numerosi problemi [I] know (the) your numerous problems

b. Ho questi diversi problemi da risolvere [I] have these different problems to solve Ì have to solve these many problems'

c. Le sue varie pretese mi hanno stancato His various requests have tired me Ì am tired of his many requests'

The categoria! status of determiners 117

(62) a. Ho numerosi/diversi/vari libri da recensire. [I] have numerous/different/various books to review Ì have several books to review'

b. Ne ho numerosi/diversi/vari da recensire. CL [I] have numerous/different/various to review Ì have to review several of them'

c. *Ne ho interessanti da recensire. CL [I] have interesting to review

A diagnostics for their Q status in (62) is that they can appear in a floating construc-tion with ne (62b), which is excluded for descriptive adjectives (62c).

Quantitative adjectives and descriptive adjectives can cooccur without being coordinated, showing that they occupy different positions:

(63) a. Conosco le molte famiglie numerose che vivono in questo quartiere. I know the many families numerous that live in this district

b. Mi interessano i molti tipi diversi di cucina che si possono trovare in India. I'm interested in the many types different of cooking that you can find in India

Notice finally that these lexical items have different interpretation according to whether they are used as descriptive adjectives or as quantity expressions (Q or quantitative adjectives), as appears clear from the glosses of examples (60)—(63) above, and as is shown by (64):

(64) a. le numerose famiglie che abitano in questo quartiere (cf. (60a)) the numerous(=many) families that live in this district

b. le diverse cucine dell'India (cf. (60b)) the different(=many) cookings of India

c. le varie diete che ha fatto (cf. (60c)) the various(=many) diets [(s)he] was on

4.4 Quantifiers are not phonologically or morphologically dependent on the noun

Quantifiers have their own morphological features that agree with the noun in gender and number. For example, in Italian we find: molte scarpe (many-fem+pl shoes-fem+pl), molti stivali (many-masc+pl boots-masc+pl). They can also quantify adjec-tival phrases. For example, in Italian we find: molto bello (`very beautiful'), tutta rotta completely broken'), troppo caro (`too expensive').

4.5 Quantifiers are not inseparable from their complement

We have seen that neither articles nor demonstratives can be separated from their complement (65a—b). This is not true for quantifiers, which can be floated of their complement (65c):


(65) a. *Boys have the left. b. *Boys have these left. c. The boys have ali left.

Sportiche (1988), followed by Shlonsky (1991) for Hebrew and Giusti (1990) for German, argues for a derivational approach of floating in terms of noun phrase movement and Q stranding. Basically, whenever NP-movement is allowed in a language, this can strand a Q in its basic position.

According to Sportiche, subject movement in French is driven by the fact that the external 0-role and nominative assignment take piace in different positions: the for- mer in SpecVP, the latter in a functional Spec:

(66) [The boys]; have [vp [all[ti]] [v, left]]

Sportiche's approach is precedent to the IP split proposed by Pollock (1989) and Belletti (1990). He therefore takes the landing site of the subject to be SpecIP. This Spec position is AgrSP (cf. section 4.1 of the introduction to this volume).

Hebrew shows that the Q is not an adjunct but a head, in that it allows move-ment to and through its Spec. The relevant data are discussed in detail by Shlonsky

(1991):

(67) a. Kol ha-yeladim axlu lexem.

ali the-boys ate bread

b. Ha-yeladim kulam axlu lexem.

the-boys all-them ate bread

c. Ha-yeladim axlu kulam lexem.

the-boys ate all-them bread

In (67a) the quantified noun phrase has moved to the IP-subject position with no internai transformation. The quantifier has the uninflected form kol. In (67b) the

whole quantified noun phrase has moved to the IP-subject position. At the same time, the noun phrase which is internai to the QP and that can be of any complexity (according to Shlonsky p.c.) has moved to a position to the left of the Q triggering obligatory agreement on the Q. This is analysed asm

e ovement of th noun phrase to

SpecQP by Shlonsky. The categoria! status of Q is that of a head selecting a full noun phrase, which can move to its Spec, and must use it as an escape hatch to leave the Q in piace, as shown by the obligatory status of the agreement appearing on the Q in (67b—c). In (68) we have the partial structure:

(68) [Qp [ha-yeladim]i [[Q. kul-am] [DP ti]]]

The possibility for Qs and no other determiner (or modifier of the noun) to appear in discontinuous position with respect to the noun phrase can be straightforwardly derived by restricting movement to full extended projections, and by stating that Qs, contrary to other determiners, are external to the extended nominai projection, as represented in structure (56a) above.



In this section, I have argued that quantifiers should be divided into two different classes, Qs proper and quantitative adjectives. Neither of the two types of items is in D. Neither is a functional category. Neither competes for the position of demon-stratives (SpecDP). Articles, demonstratives and quantifiers should therefore be considered as different kinds of elements and cannot be unified under the generai label of `determiners'.

5 Final remarks

In this chapter, I have tried to argue for the different syntactic status of determiners and for the unified categorial status of each class of determiners across languages. The classes considered here were articles, demonstratives and quantifiers. The first have proven to be the only elements that are functional heads in the extended projec-tion of the noun phrase. The demonstrative is generated in the Specifier of a func-tional head and must be moved to SpecDP at the latest at Logical Form. Qs are of two different categories: Qs proper which are external to the extended projection of the noun phrase and select a full extended noun phrase; As which appear, on a par with other As, in a maximal projection inserted in a functional specifier. Both cate-gories of quantifiers are lexical.

This proposal can predict a wide range of syntactic phenomena involving deter-miners across languages that would otherwise force us to establish a language-specific syntax of determiners.

Notes

1 This chapter is the written version of a seminar given at the University of Venice in the winter semester 1994. I thank the audience, among which were Laura Brugè, Anna Cardinaletti, Guglielmo Cinque, Mario D'Angelo, Ernesto Napoli. I also wish to thank Richard Kayne, Istvan Kenesei, Peter Ludlow, Nuria Martí, Ljiljana Progovac, Melita Stavrou, Cristina Tortora for comments on a draft version of this work.

2 Agr(eement) Object P(hrase), T(ense) P(hrase) and Agr(eement) Subject P(hrase) result from the splitting of the features in I(nflection), the head of IP, see Haegeman (this volume). For the DP/CP parallelism, see also Schonenberger and Penner (1994).

3 For reasons of space I must omit the discussion of a number of implications conceming the position of adjectives.

4 Cf. Frison as discussed by Ebert (1970). 5 Cf. some Bulgarian dialects as discussed by Stolting (1970). 6 For a preliminary approach to Bulgarian DP syntax along the lines proposed here, see

Dimitrova-Vulchanova and Giusti (1996). 7 The absence of ne in Italian forces a [+human] interpretation (cf. Cardinaletti and Giusti

1992). The correct glosses for the examples without ne are therefore 'I saw some-body/nobody'. This is inelevant to the argument here.

8 I am assuming with Cinque (1994) that adjectives are inserted in the specifier of nominai functional projections. In this position, they agree with the (1)-features of the noun through


Spec-Head agreement with the functional head whose Spec they occupy. In order to obtain the correct results in most Romance languages, where adjectives agree in ali features of the noun, each functional head with an adjective in its Spec must share ali the (13-features of the functional extended ptojection of the noun. In a minimalist framework (cf. Chomsky 1994) it is conceivable to assume that the N-to-D-movement that is overt in Rumanian takes piace in all languages at Logical Form. In other words, at least at Logical Form a chain is built in which the head is the highest functional position (D) and the foot is the lexical head (N).

9 I will leave open here the question of whether the apparent generalization ìf prepositions have an article then also nouns and adjectives do and, if adjectives have an article then also nouns do', that appears to hold (looking at Indo-European and Semitic languages) is actually correct, and if it is, how it can be derived.

10 In the postnominal position the demonstrative displays an invariable morpheme -a which

appears as the rightmost morpheme in the word after all specification for gender number and Case. Since this morpheme also appears when the demonstrative is pronominal, in Giusti (1993) I propose that it is a marker of Agreement with an empty head, be it the trace of the head N which has moved through the Agr whose Spec is occupied by the demonstrative, or lie it a pro.

11 The Head Movement Constraint formulated, by Travis (1984) and currently assumed in the literature, states that head movement is strictly local. In other words, it cannot skip an

intervening head. 12 Notice that the demonstrative can appear in all positions inside the noun phrase, as if it

could be scrambled' inside the noun phrase. For the time being I will assume, following Stavrou (p.c.), that the demonstrative in (41b—c) is a clitic-like element, parallel to the possessive in the same language: i. to oreo mou to vivlio to kalo

the good his the book the nice ii. to oreo to vivlio mou to kalo Notice, however, that the clitic possessive cannot be in first position (iii), arguing for the

non-clitic status of afto, at least in (41a). Instead, a non-clitic root must be inserted to diko

(iv): iii. *mou to oreo to vivlio iv. to diko mou to oreo to vivlio It is not unreasonable to analyse afto in all positions except the first as a weak element in

the sense of Cardinaletti and Starke (1994). 13 Cf., Haegeman (1994). 14 The very concept of `visibility' must unfortunately be left quite vague for the moment. In

fact, it is necessary to relativize it in view of the redundancy of natural languages in real- izing functional features morphologically.

15 This parallels similar assumptions made for wh-movement to SpecCP. It is premature to decide whether it is some morphological property of the demonstrative that forces the movement in overt syntax, or rather that the child learns where to piace the demonstrative directly from the empirical data.

16 For the principled explanation of this assumption, see the following section. 17 It is not crucial here to enquire which features they are, whether it is [+def], or

Rarg(umental)], or even [number]. In any case demonstratives are stronger in inflection than the article in English.

18 Baker and Hale (1990: 291) give a different account of the contrasts in (50)—(52), in terms of relativized minimality (see Rizzi 1991). They propose to assume that Ps are lexical heads on a par with Ns. According to them, this is why Ps give a relativized minimality


effect on N-movement, contrary to determiners (which they take to be functional heads, following Abney 1987). According to their proposal, .movement of N (a lexical head) could skip a D (a functional head) but not a P (a lexical head).

A lexical/functional relativization of minimality for head-movement, however, is strongly contradicted by the data on enclitic articles discussed in section 1.1 above. More generally, a treatment of suffixes as functional heads would exclude determiner incorpora-tion in noun incorporating languages as well. These languages, in fact, do not display stranded nominai suffixes.

19 Adjectives can be stranded in Greenlandic Eskimo (cf. Baker 1988: 94, quoting Sadock 1980): i. Kusanartu-mik sapangar-si-voq

beautiful-INsTR bead-get-INDIc/3ss `Fle bought a beautiful bead'

Unfortunately, the strandability of adjectives in Southern Tiwa is discussed neither by Baker (1988) nor by Baker and Hale (1990).

20 For the purpose of this discussion I will give Italian data that permit us to abstract away from the possibility of a null determiner.

21 Following Cardinaletti and Giusti (1992), I reduced the indefinite interpretation of the noun phrase to partitive Case assignment of the quantifier to the noun phrase. The indefinite interpretation of partitive noun phrases was inspired by Belletti (1988), and is confirmed by the more recent work by Eng (1991) on partitive Case in Turkish. Under this proposal, it is expected that in certain languages a partitive article appears after a quantifier. This is the case in Ladino, a Romance language spoken in North-Eastern Italy, where the cardinal una (one-fem) can precede an indefinite article such as na (a-fem) in una na ragaza (one a girl=a girl/one girl), as reported in Haiman and Benincà (1992).

22 The complex pattern of quantity expressions in Slavic languages can successfully be analysed by means of a tripartition: quantity nouns, Qs proper, quantitative adjectives. The latter two categories have been argued for above. The tripartition is proposed by Giusti and Leko (1995), to which I refer the interested reader.

23 Consider also that cardinals are the most open class that one can imagine, since they are of an infinite number by definition, as Guglielmo Cinque has pointed out to me.

References

ABNEY, S.P. (1987) The English noun phrase in its sentential aspect. PhD dissertation, MIT, Cambridge, Mass.

ALLEN, B., GARDINER, D. & FRANTZ, D. (1984) Noun incorporation in Southern Tiwa, IJAL 52, 292-311.

BAKER, M.C. (1988) Incorporation. A Theory of Grammatical Function Changing. University of Chicago Press, Chicago.

BAKER, M. & HALE, K. (1990) Relativized minimaly and pronoun incorporation, Linguistic Inquiry, 21, 289-97.

BELLETTI, A. (1988) The case of unaccusatives, Linguistic Inquiry 19, 1-34. BELLETTI, A. (1990) Generalized Verb Movement. Rosenberg and Sellier, Turin. BERNSTEIN, J.B. (1993) Topics in the syntax of nominai structure across Romance, PhD dis-

sertation, CUNY. BRUGÈ, L. (1994) Alcune Considerazioni sulla sintassi del dimostrativo in spagnolo, ms,

Universities of Padua and Venice.


CARDINALETTI, A. & GIUSTI, G. (1992) Partitive ne and the QP-hypothesis, Proceedings of

the XVIII Meeting of Grammatica Generativa Trieste. -Feb. 22-24 1991, E. Fava (ed.),

Rosenberg and Sellier, Turin, 121-41. CARDINALETTI, A. & STARKE, M. (1994) The typology of structural deficiency. On the three

grammatica) classes, (to appear in) Clictics in the Languages of Europe, H.V. Riemsdijk

(ed.) Vol. 8 of Language Typology, Mouton, Berlin.

CARSTENS, V. (1991) The morphology and syntax of determiner phrases in Kiswahili, PhD

dissertation, UCLA, California. CHOMSKY, N. (1994) The minimalist program, A view from Building 20, K. Hale and S.J.

Keyser (eds), MIT Press, Cambridge, Mass, 1-52. CHOMSKY, N. & LASNIK, H. (1977) Filters and control, Linguistic Inquiry 8,425-504.

CINQUE, G. (1994) On the evidence for partial N-movement in the Romance DP, Paths

Towards Universal Grammar, G. Cinque, J. Koster, J.-Y. Pollock, L. Rizzi, R. Zanuttini

(eds), 85-110. DIMTTROVA-VULCHANOVA,

M. & GIUSTI, G. (1996) Quantified noun phrase structure in

Bulgarian, Formai Approaches to Slavic Linguistics 3, J. Toman (ed.), Slavica, Ann Arbor,

Mich. DOBROVIE-SORIN (1987) A propos de la structure nominai en Roumain, Rivista di

Grammatica Gemerativa 12,126-51. EBERT,

K. (1970) Referenz, Sprechsituation und die bestimmten Artikel in einem Nordfriesischen Dialekt, Dissertation, University of Kiel.

ENc, M. (1991) The semantics of specificity, Linguistic Inquiry 22,1-26

ERNST, T. (1992) Phrase structure in Irish, Journal of Linguistics 28,415-43. GIUSTI, G. (1990) Floating quantifiers, scrambling and configurationality, Linguistic Inquiry

21,633-41. GIUSTI, G. (1991a) The categoria) status of quantified nominals, Linguistische Berichte 136,

438-52. GIUSTI, G. (1991b) La sintassi dei nominali quantificati in romeno, Rivista di Grammatica •

Generativa 16,29-57. GIUSTI, G. (1993) La sintassi dei determinanti. Unipress, Padova. GIUSTI, G. (1994) Heads and Modifiers across Determiners: Evidence from Rumanian, G.

Cinque & G. Giusti Advances in Rumanian, Linguisti/t Aktuell 10, Benjamins,

Amsterdam, 103-25. GIUSTI, G. (1995) A unified structural representation of (abstract) case and article. Evidence

from Germanic, Studies in Comparative Germanic Syntax, H. Haider, S. Olsen, S. Vikner,

Kluwen, Dordrecht, 77-93. GIUSTI, G. & LEKO, N. (1995) On the syntax of quantity expressions in Bosnian, R.

Benacchio, F. Fici & L. Gebert Deterrninatezza e Indeterminatezza nelle lingue Slave, Unipress, Padva, 127-45.

GRIMSHAW, J. (1991) Extended Projections, ms, Brandeis University, Mass.

GROSU, A. (1988) On the distribution of genitive phrases in Roumanian, Linguistics 26,

931-49. HAEGEMAN, L. (1994) Introduction to Government and Binding Theory, 2nd Edition.

Blackwell: Oxford. HAIMAN, L. & BENINCA, P. (1992) The Raetho-Romance Languages. Routledge, London.

HARRIS, J. (1991) The exponence of gender in Spanish, Linguistic Inquiry 22,27-62.

HOOP, H. DE (1992) Case configuration and noun phrase interpretation, PhD dissertation,

University of Groningen. LONGOBARDI,

G. (1994) Reference and proper names: a theory of N-movement in syntax and

Logical Form, Linguistic Inquiry 25,609-65.


OLSEN, S. (1989) AGReement in German noun phrase, C.H. Bhatt, E. Lobel, Ch. Schmidt (eds), Phrase Structure Phenomena, Linguistik Aktuell 6,39-50.

POLLOCK, J.-Y. (1989) Verb movement, Universal Grammar and the strutture of IP, Linguistic Inquiry 20,365-424.

Rrrit,R, E. (1991) Two functional categories in noun phrases: Evidence from Hebrew. S.D. Rothstein (ed.) Perspectives on Phrase Structure: Heads and Licencing, Syntax and Semantics 25,37-62.

Rizzi, L. (1991) Relativized Minimality, MIT Press, Cambridge, Mass. SADOCK, J. (1980) Noun incorporation in Greenlandic, Language 56,300-19. SCHONENBERGER, M. & PENNER, Z. (1994) The distribution of nominai agreement features in

German dialects and the strong DP/CP parallelism hypothesis, ms, Universities of Geneva and Bern.

SHLONSKY, U. (1991) Quantifiers as functional heads: a study of quantifier float in Hebrew, Lingua 84,159-80.

SPORTICHE, D. (1988) A theory of floating quantifiers and its corollaries for constituent struc-ture, Linguistic Inquiry 19,425-49.

STAVROU, M. (1992) Adjectives in Modern Greek, ms, Aristotle University of Thessaloniki. STOLTING, W. (1970) Beitreige zur Geschichte des Artikels im Bulgarischen. Slavistische

Beitrage 44, Verlag Otto Segner, Munich. SZABOLCSI, A. (1994) The noun phrase, Syntax and Semantics 27. The Syntactic Structure of

Hungarian, E. Kiss (ed.), Academic Press, New York, 179-274. TAPPE, H.TH. (1990) Determiner phrases and agreement in German, ms, University of TRAVIS, L. (1984) Parameters and effects of word order variation. PhD, MIT, Cambridge,

Mass.

The categorial status of determiners. The New Comparative Syntax ed. by Liliane Haegeman

Documents