73
Tel-Aviv University Faculty of Humanities Department of Linguistics The Syntax of Determiners in Hebrew Thesis submitted in partial fulfillment of the requirements for the M.A. degree in Linguistics at Tel-Aviv University by Gabi Danon Thesis Supervisors: Dr. Tal Siloni Prof. Tanya Reinhart May, 1996

The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

Tel-Aviv University

Faculty of Humanities

Department of Linguistics

The Syntax of Determinersin Hebrew

Thesis submitted in partial fulfillment of the

requirements for the M.A. degree in

Linguistics at Tel-Aviv University

by

Gabi Danon

Thesis Supervisors:

Dr. Tal Siloni

Prof. Tanya Reinhart

May, 1996

Page 2: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

Table of ContentsAcknowledgments......................................................................................................... iAbstract ........................................................................................................................ ii1. Introduction .............................................................................................................. 12. Structure of the Hebrew DP ..................................................................................... 3

2.1. The marking of nominal features............................................................... 32.2. Syntactic definiteness ................................................................................ 42.3. Construct state nominals............................................................................ 72.4. The position of determiners..................................................................... 12

2.4.1. Previous proposals .................................................................... 122.4.2. An alternative solution.............................................................. 15

2.5. Additional nominal properties of determiners......................................... 212.5.1. Plural, definiteness and gender marking................................... 212.5.2. Marginal and ambiguous lexical items ..................................... 222.5.3. Definiteness in CS..................................................................... 232.5.4. Modification and word-order.................................................... 24

3. Distributional restrictions on determiners .............................................................. 253.1. The central problem................................................................................. 253.2. Determiners vs. nouns: the distinctive feature......................................... 263.3. Definiteness and raising........................................................................... 303.4. Plurality of the complement .................................................................... 323.5. Genitive case............................................................................................ 343.6. Number and determiners: a comparison with Ritter................................ 393.7. Formalization & summary of CS licensing ............................................. 41

3.7.1. Review of formal properties of DPs ......................................... 413.7.2. Summary of cases ..................................................................... 43

4. Partial nominal projections..................................................................................... 454.1. Determiners in specifier of DP ................................................................ 454.2. Bare determiners...................................................................................... 48

4.2.1. Ellipsis ...................................................................................... 484.2.2. “Intransitive” determiners......................................................... 51

4.3. Floating quantifiers.................................................................................. 545. Determiners and the semantic interface ................................................................. 63References .................................................................................................................. 67

Page 3: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

i

Acknowledgments

I would like to thank all those who have helped me do this work. In

particular, many thanks to Tali Siloni, for introducing me to Hebrew syntax

and for all her help from the very beginning of this research; and to Tanya

Reinhart, for initiating my interest in the topic of determiners and for her

helpful remarks at all stages of my work. I would also like to thank everybody

else at the Linguistics department here at TAU, especially those who attended

my talk at the department’s colloquium. And finally, thanks to my Arab

informants for giving me the nice piece of data in §3.3 that I was so afraid to ask

about.

Page 4: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

ii

Abstract

Despite many advances in the understanding of the syntax of noun phrases,

the position of determiners and the distributional restrictions they exhibit raise

many unsolved problems. Abney's (1987) DP Hypothesis has led to new

approaches to these problems, but it accounts mainly for articles in languages

such as English, not giving a systematic account of other kinds of determiners.

This paper centers on determiners in Hebrew, and tries to give a unified

account which links the following properties of Hebrew:

• Most determiners have two (synonymous) forms, the one derived from the

other by the same morphology that derives nominal heads of construct-state

nominals; these two forms appear in complementary distribution.

• Definiteness, which is marked by a bound morpheme rather than by a

lexical article, acts like a syntactic feature on par with φ-features;

indefiniteness isn’t marked at all.

• Certain determiners, those that have overt plural morphology, have a

different syntactic distribution from those that lack such morphology.

The main hypothesis of this work is that Hebrew has no lexical category

‘Determiner’; instead, I claim that determiners are actually nouns, which differ

syntactically from “ordinary” nouns only in terms of their formal number

features. Under the framework of Chomsky's (1993, 1994, 1995) Minimalist

Program, and assuming basically a variant of Ritter’s (1991) analysis of

construct-state nominals suggested in Siloni (1994), an account is given of the

syntactic difference between determiners and other nouns, and of the

alternations and restrictions among the class of determiners.

Finally, two phenomena related to determiners are discussed and analyzed

under the hypothesis sketched above: “bare” determiners, and floating

Page 5: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

iii

quantifiers. The proposed structure for Hebrew DPs involving determiners

leads to a natural analysis of these phenomena.

Page 6: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

1

1. Introduction

The position of the determiner1 within the noun phrase has been a topic of

interest for years. A variety of structures has been suggested– most notably, the

traditional analysis of the determiner heading a “determiner phrase” in the

specifier position of NP, which has been mostly abandoned in favor of

alternative analyses: in those following Abney’s (1987) DP hypothesis, the

determiner is assumed to be the head of a functional category (D) selecting the

NP as complement. An additional extension to this approach is the “QP

hypothesis”, claiming that quantifiers head a QP projection which selects a DP

complement.

All these suggestions should be applicable to noun phrases universally. In

this paper I will investigate data from Semitic languages, modern Hebrew in

particular, which can suggest a new perspective to these questions. The

phenomenon under discussion is this: in Hebrew, many determiners have two,

apparently synonymous, forms, one derived from the other by a certain

morphophonological rule; these two forms appear in complementary

distribution, in a manner which seems at first somewhat unsystematic. For

instance, the derived form of the numeral 3 (s&los&et) selects a definite noun, while

the derived form of the numeral 2 (s&ney) can precede any plural noun (the basic

form s&nayim appears only when no noun follows). If this were the only

irregularity, it could have been excused as idiosyncratic properties of a single

lexical item; but further alternations are displayed by other determiners, and

1Throughout this paper, I will use the term “determiner” in its broadest sense, which

includes words that are often refered to as articles, quantifiers, numerals etc. Since this paper

aims at giving a unified account of the distribution of these words, any a-priori classification

should be avoided.

Page 7: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

2

the phenomenon is not limited to numerals. The central issue of this paper is to

give a systematic account of these alternations.

Given the dual nature of determiners in Hebrew, a natural idea is that more

than one structure might exist for phrases including a determiner. I will

investigate the hypothesis that there are two positions which are available to

determiners within the Hebrew DP:

a. D°

b. Specifier of DP

My claim will be that the two forms of determiners correspond to these two

positions.

Another central issue in recent studies of the noun phrase in Semitic

languages is the so-called construct-state (henceforth CS) construction, which

has been analyzed by Ritter (1991), Fassi-Fehri (1989, 1993), Siloni (1991, 1994),

Hazout (1990), Borer (1994) and others. The idea that these structures are the

result of head-movement, which raises N to D, has been a leading idea in these

studies. I will argue for a strong relationship between the distribution of

determiners and the structure of nominal CS. My analysis of determiners,

which relies heavily on the process of CS formation, will in turn offer

additional insight into the mechanism of this process.

The organization of this paper is as follows: In section 2, I discuss the basic

properties of determiners in Hebrew, and suggest which are the positions that

can be occupied by determiners in Hebrew. Section 3 deals with the

alternations displayed by different determiners, and gives a formal account of

N-to-D raising in Hebrew in terms of feature checking. Section 4 discusses

additional constructions involving determiners, extending the analysis of the

previous sections to cover these additional cases as well. And in section 5 I

discuss the effect of the syntactic processes developed in sections 2-4 at the

semantic interface.

Page 8: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

3

2. Structure of the Hebrew DP

2.1. The marking of nominal features

In Chomsky’s minimalist program (Chomsky 1993, 1994, 1995), particularly

in its more recent versions, much attention is turned to the role of feature-

checking as the motivation for movement. Since much of this paper deals with

word order inside the Hebrew DP, it is important to first review the way DP-

internal features are marked. The following properties, related to features of

definiteness, plurality and genitive case, should be noted:2

1. Definiteness is marked as a bound morpheme (the prefix ha-) on the head

noun.

2. Indefiniteness is not marked; there is no indefinite morpheme or article.3

2Another feature that might play a role in DP formation is gender. However, since I will not

refer to this feature throughout this paper I will not discuss it here either. See Ritter (1991) and

references cited there for a discussion of gender and its relevance to DP structure.

3In colloquial Hebrew, however, both the word /exad (‘one’) and the word /eyze

(‘some’/’which’) seem to be devolping into optional indefinite articles, which can co-occur.

Therefore we have the following four sentences, which are all grammatical and mean virtually

the same (‘a/some dog bit me’):

a. kelev nas&ax /oti

b. kelev /exad nas&ax /oti

c. /eyze kelev nas&ax /oti

d. /eyze kelev /exad nas&ax /oti

For most speakers, the first sentence is slightly less acceptable than the others. See

Fruchtman (1979) for a detailed descriptive discussion of these phenomena.

Page 9: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

4

3. Similarly, plurality is marked by a morpheme (the suffix -im/-ot,

corresponding to masculine and feminine nouns, respectively), while

singular nouns are not overtly marked for singularity.

4. There are two ways of marking genitive case: free genitives (FG), which

use the genitive marker s&el, and construct-state nominals (CS), in which

there is no overt case marker.

The first two of these properties will be discussed in some detail in the next

subsection. The issue of genitive case marking is discussed in §2.3. Discussion

of number marking will be delayed until §3.2.

2.2. Syntactic definiteness

In most languages that mark definiteness, this marking is achieved by a

definite article. In the terminology of the minimalist program, such an article is

usually an independent item chosen from the numeration. Definiteness, in this

case, is a semantic property of noun phrases.

Hebrew, as well as other Semitic languages, has a definite article which is a

bound morpheme. A suggestion made in the literature (Borer 1989 & 1994,

Siloni 1994, and others) is that definiteness in these languages is a syntactic

feature of the noun, similar to phi-features such as number and gender; the

definite article, then, is not an independent lexical item. This idea is supported

by the following facts:

a. As opposed to languages such as English, nothing can intervene between

the Hebrew definite article, ha-, and the noun:

(1) a. s&los&et ha-sfarim

three DEF-books

‘the three books’

b. *ha-s&los&et sfarim

b. When an AP modifies the noun, there is definiteness agreement, as well

as number and gender agreement, marked on the AP:

Page 10: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

5

(2) ha-sfar-im ha-/av-im

DEF-book-PL DEF-thick-PL

‘the thick books’

As Galit Adam has pointed out to me (PC), acquisition studies show that

definiteness marking on the AP and number agreement appear at the

same stage, just as we would expect of two occurrences of the same

syntactic phenomenon.

c. The object marker, /et, is sensitive to the definiteness of the direct object:

when the object is definite, /et is obligatory, and otherwise it is

forbidden.4 This sensitivity, of a formal syntactic item to definiteness, is

totally unexpected if definiteness is just a semantic property of DPs; if, on

the other hand, this property is also marked as a syntactic feature, then

there is nothing special about the sensitivity of /et to it.

(3) a. ra/iti (*/et) yeled

(I-)saw (*ACC) boy

‘I saw a boy’

b. ra/iti *(/et) ha-yeled

Note also that although the notions of semantic and syntactic definiteness

usually coincide, this is not always the case. For example, semantic definiteness

can also be achieved by using a demonstrative, which may or may not co-occur

4Some exceptions include proper names, pronouns (which cliticize to /et), and the word mi

(‘who’), which also must be preceded by /et when in object position. These may be argued to be

inherently marked for definiteness in the lexicon.

Page 11: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

6

with the definite article5 6; whether the article appears or not does not affect the

semantics of the noun phrase containing a demonstrative:

(4) a. ha-bayit ha-ze

DEF-house DEF-this

‘this house’

b. bayit ze

I will therefore conclude that definiteness is a formal feature of noun phrases

in Hebrew, and will henceforth refer only to syntactic definiteness throughout

this paper, unless mentioned otherwise.

A direct consequence of the analysis of the definite article as a realization of

the [def] feature on the noun, is that I am not assuming it to be generated in the

D position. This should be contrasted with definite articles in other languages,

which are generally analyzed as being full heads inserted from the lexicon

under D; some discussions of Hebrew DPs (such as Ritter 1991) assume the

same to be true for Hebrew– an assumption that I will not adopt in this work.

Some of the empirical differences between DPs in Hebrew and those in other

languages should therefore follow from this distinction, as I will later show.

In contrast to definiteness, there is no sign that indefiniteness is a

syntactically relevant feature in Hebrew. There is no overt marking of

indefiniteness, and no syntactic phenomena I am aware of can support the

existence of such a feature. I will therefore assume that there is no [-def]

feature, and indefiniteness is a purely semantic notion. This differs from

Siloni’s (1994) assumption that [def] is a binary feature with both values [±def]

possible; it also differs from Borer (1994), which assumes [±def] as well as the

possibility of a noun being unspecified for this feature. The significance of such

5Thanks to Alex Grosu for pointing this out to me .

6Since demonstratives in Hebrew are APs, the demonstrative agrees with the noun in

definiteness; thus, either both bear an article, or neither does.

Page 12: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

7

a decision is not just a matter of notation; in the framework of the minimalist

program, which I am assuming here, the checking of formal features is

assumed to be the only reason for movement. Therefore, if definiteness– as

opposed to indefiniteness– is such a feature, it is expected to play a role in

driving syntactic movement. This asymmetry between definiteness (which is

marked by the syntactic [def] feature), and indefiniteness (which is not a

syntactically relevant feature), will play a central role in my analysis of DP

structure in Hebrew.

2.3. Construct state nominals

A well known fact about Semitic languages is that when a noun has a

genitive argument, there are two ways in which the phrase can be expressed.

These are illustrated below (the two phrases are semantically equivalent):

(5) a. ha-bayit s&el ha-mora

the-house of the-teacher

‘the teacher’s house’

b. beyt ha-mora

The first form is called free genitive (FG), and is characterized by a dummy

case marker s&el. The second is called construct state (CS), and involves no such

case marker. Apart from the presence of the dummy case marker, there are

several special properties that mark CS nominals distinct from free genitives;

these properties have been extensively discussed in the literature (see for

instance Borer 1994 for a detailed discussion), and are briefly reviewed here:

• The head noun in CS differs morphophonologically from the free form of

the noun; most notably, it loses its main stress, and certain other

phonological changes may follow as a result.

Page 13: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

8

• CS nominals are head-initial and right-branching.7

• When the head of a CS is modified by an AP, the modifier isn’t adjacent to

the modified noun, but instead it follows the genitive DP.

• The definite article cannot appear on the head noun in CS.

• The definiteness value of the most deeply embedded branch of the CS is

“inherited” by the entire construct.

These properties, and the contrast between FG and CS, are illustrated in the

following example:

(6) a. ha-bayit ha-yafe s&el ha-mora

the-house(-masc) the-pretty(-masc) of the-teacher(-fem)

‘the pretty house of the teacher

b. beyt ha-mora ha-yafe

house(-masc) the-teacher(-fem) the-pretty(-masc)

‘the pretty teacher’s-house’

(6a) is an example of FG, where the genitive argument is marked by the

dummy case marker s&el. In the CS (6b), which has the same interpretation, the

head noun loses its main stress (this results in the further changes bàyit/bèyt);

the position of the AP, which modifies the head (as seen from its gender

agreement), is different from its position in the FG; and although only the head

of the genitive argument mora bears the definite article, the entire CS must be

interpreted as definite.

Most recent analyses of CS are based on the general idea developed by Ritter

(1991), who suggested that the head noun is base-generated under N and then

undergoes movement to the D position. The genitive argument (which I will

informally refer to as “the complement”) is assumed, under this general

7This property is central to the discussion of CS nominals with a complex argument

structure, most notably derived nominals. See Siloni (1994) for a discussion.

Page 14: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

9

approach, to occupy [spec, NP]. The simplest structure of a CS under this

approach is illustrated in the following example:

(7) a. beyt ha-mora

house the-teacher

‘the teacher’s house’

b.

DP

D’

D

beyti

NP

N’DP

ha-mora N

ti

This raising thus derives the head-initial word order typical of CS.

Where the different analyses vary is with respect to the following questions:

1. What additional functional categories exist between NP and DP (if any)?

2. What is the reason for N-to-D raising?

3. How is genitive case assigned?

4. What is the (DS and SS) position of arguments?

5. What is the position of APs?

The first of these questions is tied closely to the last two, since deriving the

different word orders which were illustrated in (6) is achieved in most analyses

through movement of the noun and arguments to different positions within

functional categories. Assuming that APs are left adjoined to NP, then the word

order in FG (as in (6a)) is straightforward and needs no further explanation,

except for the simple assumption that N-to-D raising occurs in FG as well as in

CS. The problematic case is that of (6b), with an AP modifying a CS; left

adjunction of the AP, and head movement from N to D, would wrongly predict

Page 15: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

10

the same word order as with FG. To solve this problem, it was first suggested

by Ritter that an additional functional category (which she calls NumP), exists,

and it is the specifier of this intermediate projection which is the S-structure

position of the genitive argument in CS (and not in FG). Furthermore, the head

of this category is an intermediate landing site for the raised N, where it gets its

number specification (the idea that Num is the host of the number feature of

the DP will be discussed further in §3.6). Thus, the following structure derives

the observed word order N DP AP in (6b):

(8)

DP

D’

D

beyti

NumP

Num’

NP

AP

ha-yafe

NP

N’

N

ti

Num

ti

DPj

ha-mora

tj

A similar proposal was made by Siloni (1994), who suggested that in CS

nominals, there is an agreement projection that intervenes between NP and DP;

again, movement of the genitive argument to the specifier of AgrP derives the

observed word order, the structure being basically the same as in (8) apart from

the labeling of the nodes. In FG nominals, on the other hand, Siloni assumes

that NP is dominated directly by DP, resulting in the following structure

(which is simply the unmarked structure where no movement other than head

movement has occurred):

Page 16: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

11

(9)

DP

D’

D

ha-bayit

NP

NP

DP

s&el ha-mora

N’

N

ti

AP

ha-yafe

The distinctions between Siloni’s proposal and Ritter’s are not limited to tree

structure, however. First, there are theory-internal assumptions on the nature of

functional categories: Ritter assumes D to be both a determiner position and the

host of Agr features, an assumption rejected by Siloni. A closely related

problem is the second question presented earlier, that of the motivation for N-

to-D movement– a question which has been a central one in many of the

analyses of CS. Both Ritter and Fassi-Fehri (1989, 1993) assume that N-raising is

needed to support the empty D (Dgen) that assigns case, under government, to

the specifier of NP (or NumP). Siloni, on the other hand, argues that the D

position is not related to case assignment at all; rather, she gives evidence that

shows that the case in CS, as opposed to case in FG, is structural case, and

therefore it must be checked in a spec-head configuration in an agreement

projection; further head-raising is driven by strong N features (in particular,

[±def]) on D. I will return to the issue of case assignment in §3.5.

A different approach to accounting for head-movement is taken by Borer

(1994). Arguing for lack of definiteness specification on the nominal head, Borer

claims this to be the defining property of heads of CS, out of which all other

properties follow. According to her proposal, a filter on the syntactic licensing

of Nmax requires it to fulfill the following 2 conditions:

Page 17: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

12

a. having a definiteness value

b. standing in a particular configuration with a Dmax having the same

definiteness value.

The exact details of Borer’s analysis are quite complex and I will not go over

them here, but the central idea is that in the absence of definiteness on the N,

the filter is met by (eventually) raising of the N to incorporate with D.

As a starting point, I will not commit myself to any specific version of the

standard analysis of CS as head-movement within the DP. Initially I will not

assume any additional functional category other than D, but later it will be seen

that an agreement projection does seem to be a natural addition. I will postpone

the question of genitive case marking until §3.5. As to arguments and APs, they

will play no role in this paper, since the problems they present seem to be

independent of the ones I am investigating. Therefore, we can abstract away

from many of the complications that are involved in particular theories.

2.4. The position of determiners

2.4.1. Previous proposals

One of the oldest problems in the syntax of noun phrases is posed by the

position occupied by determiners. As Jackendoff (1977) has shown, the specifier

position of NP is not sufficient for accounting for all possible combinations of

determiners. He suggested to solve this problem with 3 bar levels inside the

NP, with different determiners occupying the two specifier positions.

This idea is obviously problematic to more recent versions of X-bar theory.

The DP analysis of Abney (1987) offered a way out of this problem, that seems

more consistent with the standard theoretical assumptions on phrase structure.

Abney proposed the existence of a functional projection which he labeled DP,

headed by the determiner, that selects an NP complement. This gives a natural

position for articles and perhaps some other “atomic” determiners; however, as

Abney himself has noted, this can’t account for all possible combinations of

Page 18: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

13

determiners and nouns. For instance, there are syntactically complex

determiners that clearly can’t occupy a head position. Abney mentions some

examples from English; some examples from Hebrew are:

(10) a. ben s&nayim le-s&los&a (sifrey s&ira)

between two to-three (books poetry)

‘between two and three (poetry books)’

b. lefaxot s&nayim ve-/ulay /afilu s&los&a (sifrey s&ira)

at least two and-maybe even three (books poetry)

c. harbe paxot (sifrey s&ira)

much less (books poetry)

Compound determiners such as these are what motivated the older

“determiner phrase” analyses. Abney’s solution to the problem of locating

these determiners is basically a variation of Jackendoff’s; adopting the idea that

complex determiners are maximal projections, and in view of the English data,

Abney locates them in [spec, NP]. To account for the fact that there are obvious

distributional restrictions on these complex determiners, both Abney and

Jackendoff classify what I refer to as determiners into subclasses: numerals,

measure phrases (‘one half’, ‘two parts’ etc.), articles and quantifiers, and treats

each as belonging to a different syntactic (sub-)category: numerals head NPs,

measure phrases are DP[+measure], articles are D, and quantifiers are heads of

QPs (or DegP[+Q]). This classification, as well as the distribution of each

category, is stipulated, without any clear criterion for deciding what is a

determiner, what is an article and so forth. Furthermore, there is no way of

predicting in Abney’s system what phrase (NP, measure phrase etc.) will

occupy which position. For instance, why should the word order ‘the three

books’ be grammatical while ‘*three the books’ isn’t? Clearly, an analysis that

derives word order from independently motivated factors and that does not

classify determiners into stipulated subcategories would have obvious

advantages.

Page 19: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

14

Once one compares different languages, the problems with Abney’s analysis

are intensified. Even a superficial examination of Hebrew data shows that this

analysis can’t be accepted without some substantial modification. First, Hebrew

word-order, which is not identical to that found in English, poses empirical

problems to the stipulated ordering suggested for English. Furthermore,

theory-internal assumptions regarding the structure of DPs in Semitic

languages pose serious problems; for example, the specifier of NP is widely

assumed to be occupied in Hebrew by the possessor/agent argument, and

therefore a (complex) determiner can’t be located there as Abney has proposed

to be the case for English. It has also been proposed that possessors occupy this

same position in English prior to movement to [spec, DP]– a proposal which

sheds some more doubt on Abney’s analysis of (complex) determiners as

specifiers of NP.

An alternative approach to the position of determiners is what has come to

be known as the “QP hypothesis” (Valois 1991, Cardinaletti & Giusti 1991,

Shlonsky 1991, Szabolcsi 1994, among others). The general idea is that some

determiners are heads of a projection QP, which selects a DP as its complement.

The presence of an extra projection is then supposed to account for the possible

co-occurrence of more than one determiner, as well as other syntactic

phenomena.

There are several problems with the QP analysis that I want to point out. The

most important to my analysis is the similarity in syntactic behavior of the two

categories DP and QP, which makes the existence of an independent category

Q doubtful. First, note that some Qs may again be followed by a Q-headed

phrase:

(11) a. kol s&los&a /anas&im

every three men

Page 20: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

15

b. kol s&los&et ha-/anas&im

every three the-men

‘all three men’

This means that Q may select either a DP or a QP as its complement. If the

addition of the category Q was supposed to provide a unique position for every

lexical item, the need for recursive QPs makes this solution vacuous; recursive

DPs can solve this problem just as well, as I will soon show.

Second, with respect to case assignment, it is evident that QPs, as well as

DPs, must receive case. Again, this questions the distinction between the

categories– case usually assumed to be a unique characteristics of DPs.

Finally, the QP is assigned a θ-role– again, a property of DPs. For instance, in

case of collective predicates with quantified noun phrases as arguments (such

as 12), it is the entire phrase which assumes the θ-role, and not just the

quantified DP:

(12) kol ha-yeladim herimu psanter

all the-children lifted piano

‘all the children lifted a piano’

To sum up, quantified noun phrases display some characteristic properties

of DPs, a fact that makes the distinction between QP and DP questionable.

Morphological evidence from Hebrew, which will be presented in the next

subsection, will show some additional similarities between determiners

(“quantifiers”) and nouns. This will lead me to propose an alternative account

of quantified phrases, which does not need to distinguish QPs from DPs.

2.4.2. An alternative solution

In the previous subsection I have discussed some analyses of determiners

that make use of projections other than NP and DP. I want now to try and see

whether a structure which does not enlarge the syntactic inventory already

assumed is possible for the Hebrew data. If such an analysis is possible, it will

have obvious conceptual advantages over more complex analyses.

Page 21: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

16

As the discussion of QPs has shown, a recursive step in the formation of

quantified DPs is probably unavoidable. If we limit ourselves to DPs with an

NP complement, a recursive step is still independently assumed in the theory;

this is the step which allows the DP to have a genitive argument, which is itself

a DP. In the case of CS, the argument in [spec, NP] can itself be a full CS DP:

(13) sifrey morey ha-kfar

books teachers the-village

‘the village teachers’ books’

Turning now to determiners, there is a large class of determiners which are

derived from the basic forms by the same phonological rules that derive heads

of CS. For example:

Determiners Nouns

basic derived basic derived

s&nàyim s&nèy raglàyim raglèy

s&los&à s&lòs&et macevà macèvet

I will refer to the items in the second column as “derived determiners”.8

Apart from their phonological properties, these derived forms display the

following syntactic properties of heads:

1. Like other heads in Hebrew (N, P, V), they can host pronominal clitics

(see discussion in Siloni 1994).

2. They are necessarily one-word elements, and cannot be syntactically

complex.

8Not all determiners display this alternation; there are some determiners, such as harbe

(‘many’) and kama (‘some’), which have only a free form:

(a) harbe/kama sfarim

(b) *harbey/kamaey (ha-)sfarim

Page 22: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

17

Furthermore, they share the following properties with nominal heads of CS:

3. They assign (abstract) genitive case.

4. A sequence of determiners can appear in a recursive configuration.

Properties 1-2 are illustrated by the following examples:

(14) a. ra/iti /et s&ney ha-yeladim

(I) saw ACC two the-children

‘I saw the two children’

b. ra/iti /et s&ney-hem

(I) saw ACC two-them

‘I saw the two of them’

c. *ra/iti /et s&ney /o s&los&et ha-yeladim

(I) saw ACC two or three the-children

The third property, which is somewhat difficult to prove, due to the abstract

nature of genitive case in Hebrew, has overt phonological realization in other

Semitic languages such as Arabic. Compare the case marking in (15) to (16):

(15) /arba÷at-u al-/awlad-i

four-nom def-children-gen

‘the four children’

(16) bayt-u il-walad-i

house-nom def-children-gen

‘the children’s house’

Furthermore, when the noun is realized as a pronominal clitic on the

determiner, as in (17a), it is the same clitic as the one used to mark genitive

arguments on nouns (17b), as opposed to accusative clitics (17c):9

(17) a. kul-i/*kul-eni

all-me

‘all of me’

9This was pointed out to me by Tali Siloni.

Page 23: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

18

b. beyt-i/*beyt-eni

house-my

‘my house’

c. hu nisa le-hars&im-eni/*le-hars&im-i

he tried to-impress-me

The fourth property is limited and can’t be displayed by all combinations of

determiners10; but it can be illustrated with the following examples headed by

the determiner kol11:

(18) a. kol s&ney yeladim12

every two children

b. kol /esrot /alfey ha-/anas&im

every tens thousands the-men

‘all tens of thousands of men’

All the properties of Hebrew determiners that were discussed above make it

reasonable to try and draw a parallel between derived determiners and heads

of construct state nominals. In fact, the idea that determiners such as s&ney are

10It seems reasonable to assume that the limitation on “determiner-chaining” is at least

partially semantic; for example, it is hard to make sense of DPs such as the following:

*s&los&et kol ha-yeladim

three all the-children

11 It can be argued that since kol is possibly ambiguous between the free and derived form

(as are all monosyllabic nouns), this example doesn’t necessarily illustrate a sequence of

derived determiners. I will ignore this possibility here, as the next example does not rely solely

on kol.

12 This is a counterexample to Shlonsky's 1991 (and others') claim, that kol selects either a

definite plural or an indefinite singular complement. An alternative generalization suggested to

me by Tali Siloni (PC) is that indefinite complements result in a distributive reading (‘each’),

while the collective reading (‘all’) is associated with definite complements.

Page 24: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

19

heads of CS is so natural, that many speakers take it to be obvious, and Ritter in

her discussion has also assumed it to be taken for granted. Note that in Hebrew

there are also adjectival CS (19a), CS headed by present-tense verbs (19b), and

(arguably) also CS headed by gerunds (19c) (see also Siloni 1994):

(19) a. (yalda) kxul-at /enayim

(girl) blue-FEM eyes

‘blue eyed girl’

b. /oxley s&fanim

eaters rabbits

‘rabbit eaters’

c. cet ha-nes&ama

exiting the-soul

Therefore, the existence of CS headed by determiners is a very attractive

idea. The minimal hypothesis is that there is no syntactic difference between

these and nominal CS, the distinction being purely semantic. I will start from

this zero hypothesis, which will later be slightly modified. In other words, I will

start from the hypothesis that in Hebrew, determiners are actually nouns.

Hopefully, this approach will allow us to give a simple explanation that

requires no additional machinery other than what is already assumed to exist.

Since I am assuming, following the recent research on CS that was

mentioned earlier, that nominal CS structures are generated by N-to-D raising,

I will assume the same for those headed by derived determiners: the

determiners are generated under N, and then raise to D, for reasons that will be

discussed later.

So far I have only discussed derived determiners, and suggested that they

are heads of CS. But there are still two other groups of determiners to consider:

the free forms of those determiners; and complex determiners.

There is good reason to believe that none of these are heads of the DP: the

complex ones obviously can’t occupy a head position; and the free determiners

Page 25: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

20

seem to have the same distribution as the complex ones. The impossibility of

realizing the complement as a clitic on free determiners (21) as opposed to

derived determiners (20), which is illustrated below, gives further evidence for

the non-head status of free determiners:

(20) a. s&los&et ha-sfarim

three the-books

‘the three books’

b. s&los&t-am

three-them

(21) a. *s&los&a sfarim

three books

b. *s&los&-am

Note that in the examples in (10), complex determiners precede anything else

within the noun phrase, including the raised head of CS. A suggestion which

has been made by Hazout (1990), reminiscent of Jackendoff’s analysis of

determiners, is that such determiners occupy a specifier position. The idea that

noun phrases are simply DPs is still sustainable, if we assume that complex and

free determiners in Hebrew occupy the [spec, DP] position. Note that it has

usually been assumed, either implicitly or explicitly (Borer 1994) that this

position is not used in Hebrew. This contrasts with languages such as English

or German, where the standard assumptions are that [spec, DP] is occupied by

possessors (where they are either base generated or raised to). Since there is no

clear reason why Hebrew DPs shouldn’t have a specifier position, my

suggestion seems rather natural. As I hope to show later, this structure makes

correct predictions with respect to phenomena such as floating quantifiers.

As to the question of what are the label and internal structure of these

complex determiners, at the moment I will just label these as XP, and return to

these questions in §4.1.

Page 26: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

21

To sum up, the following two structures illustrate the two positions I am

assuming to be possible for determiners:

(22)

DP

D’

D NP

N’

N

ti

sfarimi

XP

s&los&a

DP

D’

D NP

N’

N

DP

ti

s&los&eti

ha-sfarim

It should be noted that both of these trees show head-movement from N to

D, even though in the non-CS case this is not at all obvious. This point will be

justified in section 3, where I discuss the reason for N-to-D raising.

Finally, regarding the idea that more than one structure is possible when a

determiner is present, it should be mentioned that similar proposals have been

made for other languages as well. See for instance Giusti (1995) for a recent

discussion.

2.5. Additional nominal properties of determiners

The idea that (some) determiners are generated in the same position as

nouns is not such a radical one. Similar suggestions have been made for a wide

range of languages. I will now review some additional data from Hebrew that

shows additional similarities between these two classes of lexical items.

2.5.1. Plural, definiteness and gender marking

If determiners are actually nouns, we would expect them to display some

nominal morphology, such as phi-features. Indeed, like nouns, some

determiners can be marked by the plural suffix– either in their free form (23b)

or in the derived form (23c):

Page 27: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

22

(23) a. /elef/hamon (/anas&im)

thousand/multitude (men)

b. /alaf-im/hamon-im

thousand-PL/multitude-PL

c. /alf-ey/hamon-ey /anas&-im

Similarly, the definite article can sometimes be added to some determiners:

(24) ha-kol / ha-rov / ha-s&nayim

the-every /the-most / the-two

‘everything’ / ‘the majority’ / ‘the two’

Finally, numerals have both a masculine and a feminine form, whose

marking is similar to the marking on nouns (although for some reason the

marking is “reverse”, with the suffix -a marking feminine nouns but masculine

numerals):

(25) a. s&los&-a xatul-im

three-MASC cat-MASC-PL

‘three cats’ (MASC)

b. s&alos& xatul-ot

three cat-FEM-PL

2.5.2. Marginal and ambiguous lexical items

Some words are very hard to classify into one of the categories under

discussion, and might in fact be ambiguous between the two. For instance, the

words hamon and rov are ambiguous between referring and quantifying uses, as

the following examples demonstrate:

(26) a. hayu s&am hamon /anas&im

were there multitude men

‘there were lots of people there’

b. ha-hamon daras& pe/ula

the-crowd required action

(27) a. rov ha-sfarim mes&a/amemim

most the-books boring

Page 28: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

23

b. ha-rov kove/a

the-majority decides

Some additional examples of words that are marginal are mispar (‘number’),

xelek (‘part’) and xeci (‘half’).

All these examples show that at least in Hebrew, the lexical distinction

between nouns and determiners is often unclear, what might again suggest that

the two categories are not distinct from each other. Similar data is available

from other languages– for instance, Babby (1987) has argued for a similar

diachronic change in Russian.

2.5.3. Definiteness in CS

Strong support for the analysis of certain determiners as heads of CS comes

from the fact that they share with nominal CS the property of being

“transparent” for definiteness: the definiteness of the entire DP is determined

by the definiteness of the genitive DP, regardless of the semantic value of the

determiner. Two simple tests for the definiteness value of a DP which were

mentioned earlier are:

1. The object marker, /et, appears only in front of (syntactically) definite

objects.

2. In the presence of an AP, it must agree in definiteness with the whole

DP.

Given these tests, the following examples illustrate the fact that definiteness

is indeed independent of the determiner, and show the parallelism between

nominal heads and determiners with respect to it:

(28) a. ra/iti tmunat perax /adom

(I) saw picture flower red

‘I saw a picture of a red flower’

b. ra/iti *(/et) tmunat ha-perax *(ha-)/adom

Page 29: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

24

(29) a. ra/iti s&ney /anas&im yerukim

(I) saw two men green

‘I saw two green men’

b. ra/iti *(/et) s&ney ha-/anas&im *(ha-)yerukim

The role of definiteness in det-CS structures will be discussed in more detail

in §3.3.

2.5.4. Modification and word-order

Although it is hard to find examples of prototypical determiners that can be

modified, some of the marginal cases discussed in §2.5.2 can be modified.

When this happens, the position of the modifier with respect to the determiner

is the same as that of an AP with respect to the noun it modifies, as discussed in

§2.3: if the determiner is in its derived form, the word order is like in CS

nominals– following the complement ((30a) and (31a); compare with (6b));

when it is in its free form, just as in FG nominals it follows the head ((30b) and

(31b), and compare with (6a)):

(30) a. mispar ha-mis&tatfim ha-mu/at

number the-participants the-little

‘the small number of participants’

b. ha-mispar ha-mu/at s&el ha-mis&tatfim

(31) a. rov ha-mis&tatfim ha-muxlat

majority the-participants the-absolute

‘the absolute majority of the participants’

b. ha-rov ha-muxlat s&el ha-mis&tatfim

Page 30: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

25

3. Distributional restrictions on determiners

In this section I will discuss the distribution of different forms of

determiners, and try to give an account for:

a. the non-uniform distribution of determiners, and

b. the difference in distribution between determiners and nouns.

3.1. The central problem

If, as I have claimed in the previous section, the structure of DPs headed by a

determiner parallels that of CS headed by a noun, then we would expect a

similar distribution of both kinds of heads. However, this is not the case.

Ordinary CS can, in general, appear with all combinations of

definite/indefinite and singular/plural complements. Determiner heads, on the

other hand, have a much more limited distribution: most determiners are

sensitive to the definiteness and plurality of their complement. Furthermore,

even within the class of determiners there is a great deal of variation. The

following examples illustrate these problems:

(32) a. kri/at ha-sfar-im/sfar-im/ha-sefer/sefer

reading the-book-s/book-s/the-book/book

‘the reading of (the) book(s)’

b. s&los&et ha-sfar-im/*sfar-im/*ha-sefer/*sefer

three (the-)book(-s)

c. /esrot ha-sfar-im/sfar-im/*ha-sefer/*sefer

tens (the-)book(-s)

d. kol ha-sfar-im/*sfar-im/ha-sefer/sefer

every/all (the-)book(-s)

e. rov ha-sfar-im/*sfar-im/ha-sefer/*sefer

most (the-)book(-s)

Page 31: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

26

Therefore, in order for my analysis to be plausible, it must account for these

differences– that between noun-headed CS and determiner-headed CS, and

that among determiner-headed CS.

Another question that arises at once from my analysis is: if both determiners

and nouns are generated in N and behave syntactically in the same way, what

distinguishes the ones we call “determiners” from the rest of the nouns? There

is a clear semantic distinction between nouns and determiners, but in light of

the fact that the two classes differ somewhat in their syntactic distribution,

some syntactic classification seems necessary. The question is whether there

exists some syntactic feature which distinguishes nouns from determiners,

apart from the [±det] or [±N] features which might be stipulated– a stipulation

without any explanatory content.

3.2. Determiners vs. nouns: the distinctive feature

A key observation towards solving the problems just presented in the

previous subsection is the following:

• Determiners which have overt plural morphology have a different

distribution from those which lack it.

This phenomenon can best be demonstrated by minimal pairs such as /asara

(‘ten’, free) vs. /asarot (‘tens’, free), and /aseret (‘ten’, derived) vs. /esrot (‘tens’,

derived):

(33) a. /aseret ha-sfarim/*sfarim

ten the-books/books

‘(the) ten books’

b. /asara *ha-sfarim/sfarim

(34) a. /esrot ha-sfarim/sfarim

tens the-books/books

b. /asarot *ha-sfarim/*sfarim

Page 32: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

27

Comparing (33a) with (34a) we see that /aseret can take only a definite noun,

while /esrot can also take an indefinite; the free forms (33b) and (34b) are

complementary to the corresponding derived forms.

The same alternations are observed with pairs such as me/a/me/o t

(‘hundred’/’hundreds’), /elef//alfey (‘thousand’/’thousands’), etc. It is

interesting to note that this pattern carries over even to plural-marked

determiners which don’t have a corresponding “singular” form, and whose

plural marking is therefore less obvious; in particular, s&nayim/s&ney (‘two’)

which is marked by the dual suffix -ayim, patterns with other plural-marked

determiners (as in (32c) and (34)), and contrasts minimally with non-marked

determiners such as s&los&a/s&los&et (‘three’) (see (32b) and (33)).

This data suggests that number marking plays an important role in CS

formation– a fact realized already by Ritter (1991). The existence in the syntax

of a formal number feature is something which is widely accepted. Since

number morphology is overt in Hebrew, the zero hypothesis is that plural

morphology surfaces if and only if a [+plural] feature is present. An idea that I

want to investigate is that determiners, as opposed to nouns, do not have their

own syntactic number specification when they don’t bear the plural suffix, even

though semantically they obviously do. If we follow Chomsky’s (1993, 1994,

1995) approach of the minimalist program, that only the checking of formal

features can drive movement, then the lack of a number feature on determiners

is a fact that should indeed affect the possibility of head movement from N to

D. So the hypothesis I will test is that only determiners that have a plural

morpheme have a number feature, and that this is the minimal syntactic

difference between determiners and nouns.

If determiners lack a number feature, we would expect this to be reflected in

number agreement with the VP, when the DP headed by the determiner is in

subject position. This indeed seems to be the case. Consider, for instance, the

Page 33: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

28

determiner /elef (‘thousand’) that has no number morpheme and is therefore

assumed not to have a [+plural] feature; as the following example shows, this

determiner can agree either with a singular or with a plural VP, despite the

semantic plurality. But once the plural suffix appears on this determiner, the

plural-marked form must agree with a plural VP:

(35) /alaf-im higi-u hayom, ve-maxar kanir/e yagi-a/yagi-u /od /elef

thousand-PL arrived-PL today, and-tomorrow probably will arrive-SG/arrive-PL another thousand

Since we have no reason to assume that /elef is ambiguous, I will conclude

that the number specification of the DP in the second conjunct is not

determined by the numeral, which simply lacks a number feature.

It should also be noted that when a numeral such as /elef is followed by a

plural noun, the entire DP is plural (regardless of the “singularity” of the

determiner):

(36) /elef klav-im navx-u

thousand dog-PL barked-PL

This phenomena will be analyzed in §3.4.

Note that the underspecification of the syntactic number feature is not

limited to numerals. The following examples, including the determiners rov

(‘most’), zug (‘pair’) and harbe (‘many’/’much’), illustrate similar

singular/plural alternations:

(37) a. ha-rov hicli/ax

the-most succeded(SG)

b. ha-rov /ahav-u /et ha-seret

the-most liked-PL ACC the-film

(38) zug tayarim nixnas/nixnes-u la-xeder

pair tourists entered(SG/PL) to-the-room

Page 34: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

29

(39) a. harbe his&tana me-/az

much changed(SG) from-then

‘much has changed since then’

b. harbe /azv-u ba-/emca13

much left-PL in-the-middle

‘many have left in the middle’

To sum up, my claim so far is that unless optionally marked for plurality,

determiners lack a syntactic number specification; such optional marking is

reflected by the appearance of plural morphology. The lack of an inherent

number feature is the substantial point where determiners differ from nouns,

which are always selected from the lexicon as either singular or plural (either

one of [±plural]). The number feature on the N head, I propose, is a strong

feature that must be checked in D. Therefore, its presence is sufficient to drive

overt N-to-D raising, while in its absence an independent reason for such

raising is needed.

An important outcome of the last paragraph is that in “regular” DPs headed

by a noun, the head noun always raises to D– even in free nominals. This has

also been claimed by others for different reasons, mostly related to word order

(particularly to account for NSO order in derived nominals): Ritter claimed that

in free nominals N raises to NumP (the functional projection she assumes

between NP and DP), and Siloni assumes raising to D in all cases to check

definiteness. These claims have been based on reasons which are independent

of my feature-checking analysis; thus, my theory-internal conclusion is

probably independently justified. Also, in view of the growing evidence from

recent research (Longobardi 1994, Cardinaletti 1994, Giusti 1994, Vergnaud

13As to the possibility that this example might involve ellipsis at some level inside the DP, I

will delay discussion of this idea until §4.2.

Page 35: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

30

1992) in support of N-to-D raising, the conclusion that the noun always raises

in Hebrew seems to be unproblematic, and even welcome.

3.3. Definiteness and raising

My proposal that the [±plural] feature drives movement allows us to account

for some of the central problems which we aim to explain; the differences

which are observed between nouns and determiners, and between plural-

marked determiners and those that are not, in the formation of CS, is now

explained by the presence or absence of the number feature. But there still

remains the problem of accounting for the fact that some (but not all)

determiners in CS are sensitive to the definiteness of their complement. This is

illustrated by the following examples:

(40) s&los&et/rov ha-sfarim/*sfarim

three/most (the-)books

(41) /alfey/s&ney ha-sfarim/sfarim

thousands/two (the-)books

‘(the) thousands of/two books’

A simple descriptive generalization that emerges from these examples and

the discussion in §3.2 is that det-CS formation is only possible when:

• The determiner is plural-marked (41), or

• The complement is definite (40).

To account for this additional fact, I will return to the discussion in §2.2

regarding syntactic definiteness. There, it was suggested that there is no [-def]

feature which can participate in feature-checking; this was opposed to the [def]

feature, which was argued to exist as a formal feature on nouns (as well as

other categories). Just like the number feature, which may or may not exist on

the head N, let’s assume that [def] is an optional feature on the head that is

selected from the numeration. Assume at the moment that the existence of the

definite article on the complement is a sign for the existence of the [def] feature

Page 36: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

31

on the head; a formal account of this will be given in §3.7. As a consequence, if

the complement isn’t definite, the head has no definiteness feature to check in

D. Put together with my claim that determiners lack their own number

specification, we get the result that a CS-determiner may have neither a number

nor a definiteness feature to check. If number and definiteness are the only

strong features that N must check in D, then we have an explanation for the

ungrammaticality of cases such as (40) with an indefinite.

This idea is somewhat reminiscent of that developed by Borer (1994). There,

it was claimed that the defining property of nouns-in-construct is their lack of

definiteness specification. Leaving aside the details of Borer’s analysis, the

central point shared by my analysis and hers is that the absence or presence of a

[def] feature can account for differences in N-to-D raising. However, Borer’s

theory assumes both [+def] and [-def] in addition to the possibility of no

specification; therefore, it does not predict any difference between definite and

indefinite complements– a difference which is observed only with determiner

heads, not with nominal heads, and this is why it might have been overlooked.

My analysis, which assumes only a [+def] feature, succeeds in explaining the

definite/indefinite contrast.

There is one strong prediction that the last discussion makes: given a

language which differs minimally from Hebrew in having a [-def] feature in

addition to [+def], we would expect it to allow det-CS formation with indefinite

complements; the presence of the [-def] feature would allow head movement,

just as [+def] allows it in the definite case. Standard Arabic seems to be such a

language: since indefinite nouns are morphologically marked by a suffix, it is

reasonable to assume that this is the realization of a [-def] feature. And indeed,

numerals can head det-CS constructions in Arabic with indefinite nouns (as in

(42a), in exactly the same way as with definites (42b); the numeral has precisely

the same form in both cases, and the difference is only in the definiteness

marking on the noun:

Page 37: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

32

(42) (a) /arba÷at-u /awlad-i-n

four-nom children-gen-indef

‘four children’

(b) /arba÷at-u al-/awlad-i

four-nom def-children-gen

‘the four children’

One minor problem still remains– that of ruling out constructions that

involve no head movement at all. This is needed in order to account for the

ungrammaticality of phrases such as:

(43) *sfarim rov/s&los&et

books most/three

One simple solution is to stipulate that D cannot remain empty; perhaps this

can be explained by some general prohibition on empty functional heads (see

Longobardi 1994 and references cited there for similar suggestions); or perhaps

the need to check the case feature (which will be discussed in §3.5) is what

makes these examples ungrammatical. I will leave this question open.

3.4 Plurality of the complement

So far, I have discussed only the features which are marked on the head.

Since both number and definiteness can be absent from the head, my analysis

leaves the possibility that a head raise to check only one of these features, with

the other feature left unspecified. With respect to definiteness, this is

unproblematic, since there is no clear need for a DP to be specified for this

optional feature. But number is a feature that seems to be necessarily specified

for DP, as obligatory number agreement with the VP shows.

In nominal CS, the head is always specified for number; therefore, the DP

always has the number value of the head:

(44) a. tmunat xatul

picture cat

‘a picture of a cat’ (singular)

Page 38: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

33

b. tmunat xatul -im

picture cat-s

‘a picture of cats’ (singular)

In det-CS, on the other hand, I have already shown in §3.2 that the number

of the entire DP is sometimes determined by the complement, as can be shown

by number agreement with the VP:

(45) a. rov ha-seret hirgiz /oti

most the-film annoyed(SG) me

‘most of the film annoyed me’ (singular)

b. rov ha-srat-im hirgiz-u /oti

most the-film-s annoyed-PL me

‘most of the films annoyed me’ (plural)

We see that even though there is no overt number marking on the head in D,

it gets the feature from the genitive argument. One possible explanation for this

fact is that LF movement of the argument is involved: if the head noun isn’t

specified for a number feature, the complement raises covertly from [spec, NP]

to [spec, DP] to check it. In terms of feature strength, we can say that the

number feature on D is weak, and thus it can be checked after spellout by the

complement. Alternatively, in more recent terms (Chomsky 1995) we can argue

for feature movement after spell-out.14

14An alternative solution to the problem of the number specification of the DP in such cases,

that was suggested to me by Tali Siloni, is that the same mechanism that gives the CS the

definiteness value of the embedded DP also applies here. Although this idea seems very

natural and attractive at first, it has the problem of predicting that determiners unspecified for

syntactic number get a number feature somehow, which doesn’t seem compatible with my

analysis. This contrast between definiteness and plurality might support the idea that [def], as

opposed to [±plural], is generated on the Agr head, and therefore it can be inherited and force

movement to D.

Page 39: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

34

This hypothesis makes a simple prediction: The specifier of DP must be

empty at spell-out in order for such covert movement to be possible. Therefore,

we do not expect to find a specifier when the head isn’t plural-marked. On the

other hand, when the head is marked for a number feature, a specifier might be

present (as far as other reasons don’t block it). If we accept the assumption that

the free determiner kama (‘some, several’) is located in [spec, DP] (see §4.1 for

discussion), then the following contrasts could support our prediction:

(46) a. kama sifrey yeladim

some books children

‘some children’s books’

b. kama /esrot ha-mis&tatfim

some tens the-participants

‘the several tens of participants’

c. * kama /aseret ha-mis&tatfim

some ten the-participants

In (a) and (b), the number feature ([±plural]) on D is checked by the raised N;

therefore, there is no need for the complement to raise after spellout to [spec,

DP], and this position can be occupied by kama. In (c), on the other hand, the

complement must raise to the position which is already occupied in order to

check the number feature of D– therefore, the derivation crashes.

3.5. Genitive case

Up till now, I have only treated the [±plural] and [def] features in the DP.

Another feature that plays a role in CS formation is genitive case, which is

assigned to the argument.

Another approach to accounting for the number “inheritance” phenomenon is that perhaps

some mechanism of feature percolation, as suggested to me by Outi Bat-El, is involved here; but

it is unclear to me how such a mechanism should be formalized in minimalistic terms.

Page 40: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

35

Before turning to case assignment in det-CS, we must address the question of

how genitive case is assigned to the complement DP in nominal CS. I have

mentioned in §2.3 the two suggestions which have been made in the literature

on this subject: Either case is assigned by the raised N under government to the

specifier of NP/NumP (Fassi-Fehri 1993, Ritter 1991); or, as Siloni (1994) has

suggested, being structural case, genitive case in CS is assigned in spec-head

configuration under an agreement projection, which immediately dominates

NP. The arguments for claiming that this is structural case have to do with the

possibility of assigning genitive case to subjects of small clauses, in ECM

constructions involving derived nominals. As this is possible only in CS and

not in free genitives, the conclusion is that genitive case in CS (as opposed to

that in FG) is not θ-related, and thus structural.

Consider now how Fassi-Fehri’s and Ritter’s account of genitive case

assignment relates to my proposals. Central to this account is the claim that

case is the motivation for N-to-D raising: the head noun must raise to a position

where it will c-command the specifier of NP (or NumP in Ritter’s version)– the

D position. The discussion of Hebrew det-CS so far makes this idea

incompatible with my analysis, which relies on the presence of [def] and

[plural] features for N raising to D. If genitive case were a possible cause for

head movement to D, then we would expect all derived determiners (which

presumably are specified for the [assign gen] feature in the lexicon) to be able to

raise, and the impossibility of N-to-D movement of non-plural determiners

would remain unexplained.

Turning back to Siloni’s account of genitive case assignment, it seems to

show no incompatibility with my analysis of raising. If the case feature drives

raising of the head to an intermediate Agr, where case is checked, then further

raising to D still remains to be accounted for. In Siloni’s analysis, the need to

check [±def] is what drives this further raising to D; here, I suggest that it is

Page 41: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

36

[def] and [±plural] features that still need to be checked in D, and my analysis

can be maintained without additional modification.

If we accept the idea that there is an AgrP between NP and DP, we must

answer the question which features (other than case features) are checked in

this projection. In §3.3, I have assumed that definiteness of the complement of

CS is somehow shared (or “inherited”) by the head. It is a decisive fact that the

definiteness value of the complement is also the value of the entire DP. Now we

can give a formal account for this fact, namely– that [def] is checked under the

agreement projection dominating NP.15 Or, alternatively, one can adopt Siloni’s

(1994) idea that the definiteness value is generated on the Agr°, and checked

with the raised DP.

This, of course, means that heads of CS are ambiguous with respect to the

[def] feature, which they don’t display any overt marking of. So, a head such as

batey (‘houses’, derived), can be either [def] or unspecified for this feature

(which is, as we have assumed, the syntactic realization of “indefiniteness” in

Hebrew). The lack of phonological realization of [def] in derived forms might

be due to the morpho-phonological rules deriving these forms, which might be

incompatible with the presence of the definite article.

To sum up, I have claimed that the definiteness “inheritance” in CS is simply

a matter of agreement between the head, which shows no overt realization of

this feature, and its genitive argument. In view of the fact that definiteness

agreement is such a widespread phenomenon in Hebrew (with adjectives and

demonstratives, for instance), then such agreement within the DP seems to fit

in naturally.

15Since definiteness is an interpretable feature, we can accept Chomsky’s (1995) idea that

this feature is not deleted after being checked in Agr, and it can be checked again in D. Thus the

presence of [def] can still drive movement to D, even if it is checked in Agr.

Page 42: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

37

Going back to Siloni’s claim, that case in free genitives, as opposed to that in

CS, is inherent case which is assigned in conjunction to a theta-role, we should

note that it has another important implication to an analysis of determiners. As

has been noted before (by Ritter, for instance), determiners, as opposed to

nouns, can’t head FG or clitic-doubled constructions. Compare (47), where a DP

headed by a lexical noun can be either a CS (a), a FG (b), or involve clitic-

doubling (c), with (48), where the head is a determiner:

(47) a. beyt ha-/ikar

house the-farmer

‘the farmer’s house’

b. ha-bayit s&el ha-/ikar

c. beyt-o s&el ha-/ikar

(48) a. s&los&et ha-sfarim

three the-books

‘the three books’

b. *ha-s&los&a s&el ha-sfarim

c. *s&los&t-am s&el ha-sfarim

Accounting for this paradigm is straightforward if we accept the above

inherent vs. structural case distinction. Since determiners such as s&los&a/s&los&et

don’t assign a possessor/theme/agent theta role (which is available when the

head is a noun), they can’t license inherent case. The same behavior is observed

in adjectival CS, in which the head also lacks the ability to assign these theta

roles:16

16An apparent problem for this explanation that was pointed out to me by Tali Siloni is that

present tense verbs, even though they do have a theta role to assign, can nevertheless assign

genitive case only in CS, and can’t form FG or clitic-doubled construction:

Page 43: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

38

(49) a. ha-yalda kxul-at ha-/enayim

the-girl blue-FEM the-eyes

‘the blue eyed girl’

b. *ha-yalda ha-kxula s&el ha-/enayim

c. *ha-yalda kxulat-an s&el ha-/enayim

There are, however, some apparent counterexamples to the generalization

that determiners can’t license s&el-phrases. The following examples illustrate free

genitives (50b) and clitic doubling (51b) with a determiner head:

(50) a. hamon-ey nemalim

multitudes ants

‘multitudes of ants’

b. hamon-im s&el nemalim

(51) a. rov ha-sfarim

most the-books

‘most of the books’

b. rub-am s&el ha-sfarim

It can be argued that these determiners that allow s&el-phrases are those

marginal lexical items discussed in §2.5.2, which are ambiguous between the

determiner and the noun interpretation. An alternative line of explanation,

which has to do with the semantic value of these determiners, will be suggested

in §4.3.

I will therefore conclude that the distinction between nouns and determiners

with respect to the ability to license s&el-phrases is due to the inability of

(a) /oxley s&fanim

eaters rabbits

‘rabbit eaters’

(b) * /oxlim s&el s&fanim

(c) * /oxleyhem s&el s&fanim

Page 44: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

39

determiners to assign a theta role to the argument, and to the fact that genitive

case in s&el-phrases is an inherent case.

3.6. Number and determiners: a comparison with Ritter

It is evident that the number feature plays a central role both in Ritter’s

analysis and in mine. Ritter assumes this feature to be located in Num;

furthermore, she claims that derived determiners are base generated in this

position. There are some clear similarities, as well as crucial differences,

between this analysis and mine. I will now discuss some of these differences.

The most obvious difference that stands out between the two analyses is the

position of determiners. While Ritter assumes determiners to be lexical Nums,

bearing no direct relationship to nouns, I propose the opposite– that

determiners are Ns, which simply lack a number feature. This alone might seem

to be a theory-internal description which makes no direct predictions; but

Ritter’s proposal makes it difficult to explain facts such as the role of the plural

suffix on determiners such as /alaf-im, and the contrast between plural-marked

and unmarked determiners, which is a central issue in my approach. In fact,

under Ritter’s approach, all variation among determiners must be accounted

for by means of idiosyncratic selectional restrictions.

To support her proposal of a NumP, Ritter also claims that determiners, as

opposed to nouns, have no gender specification. But at least with respect to

numerals, this is not true: numerals have both a masculine and a feminine

form, as the following examples show17:

(52) a. s&los&-a yelad-im

three-MASC boys

17It was already mentioned in §2.5.1 that gender marking on numerals is ‘reverse’: the suffix

-a, which marks feminine nouns, marks masculine numerals. Masculine nouns and feminine

numerals are not marked by any suffix.

Page 45: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

40

b. s&alos& yelad-ot

three girls

This of course poses no problems to my analysis– being nouns, it is only natural

to find gender specification on numerals.

As was noted in §2.3, apart from the labeling of the nodes, Ritter’s structure

for noun-headed CS (53) and Siloni’s (1994) structure ((54), which is essentially

the structure I am also assuming) are basically the same:

(53)

DP

D’

D

beyti

NumP

Num’

NP

N’

N

ti

DPj

ha-mora

tj

Num

ti

(54)

DP

D’

D

beyti

AgrP

Agr’

NP

N’

N

ti

DPj

ha-mora

tj

Agr

ti

Page 46: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

41

However, when it comes to determiner-headed CS, Ritter claims that the

structure differs in having no NP level:

(55)

DP

D’

D

s&neyi

NumP

Num’DP

ha-yeladim Num

ti

As I try to show throughout this paper, there is great similarity between the

two types of CS; therefore, a theory that proposes the same structure for both

kinds of DPs seems to me preferable than one that claims such a syntactic

difference as the one between (53) and (55).

Finally, addressing the fact that free genitives headed by determiners are

impossible, Ritter proposes that this is due to the fact that case marking via s&el

is only available to arguments inside the NP. I believe this to be a mere

description of the facts, with no explanatory content. As the discussion in §3.5

shows, this fact might be reduced to the lexical content of determiners and their

lack of ability to assign a theta-role.

3.7. Formalization & summary of CS licensing

In §3.7.1 I will review and summarize everything that I have argued for so

far; in §3.7.2 I will show examples of how my theory accounts for the major

types of DPs.

3.7.1. Review of formal properties of DPs

Three features are checked inside the DP, and can therefore drive movement

of its constituents:

Page 47: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

42

1. Number. Lexical nouns are always specified for this feature, namely– they

are always [±plural]. Determiners, on the other hand, are not inherently

specified for this feature, but some may optionally be [+plural]– in this

case, they show overt plural morphology. The syntactic number feature

should not be confused with the semantic plurality of most determiners.

2. Definiteness. Both nouns and determiners can be either [+def] or

unspecified for definiteness, in which case they are usually interpreted as

indefinites. Heads of CS (nouns or determiners), although they show no

morphological indication of the presence of the [def] feature, are

ambiguous between being [def] and being unspecified.

3. Genitive case. Derived forms, of both nouns and determiners, are

specified for the feature [gen]; this feature is checked with the genitive

arguments in a spec-head configuration, in an AgrP which dominates

NP.

Of these three features, the first two must be checked in D whenever they are

marked on the head which is generated in N. The agreement projection

between NP and DP (which exists only in CS and not in free genitives) involves

checking of case and (optionally) definiteness; after checking these features

there, further raising to D must again be motivated by the need to check some

feature. When the definiteness feature is available, after checking agreement

between the head and the genitive phrase in AgrP it can be checked again in

head-head configuration in D, since this is an interpretable feature. Otherwise,

[±plural] is the only possible cause for raising to D.

Formalizing this in terms of strong and weak features, the features [±plural]

and [def] are strong features of the head noun/determiner, and therefore

whenever they are present they must be checked with D prior to spell-out. The

case feature is also strong and must be checked in AgrP. Agr also has an

Page 48: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

43

optional [def] feature, and this accounts for the phenomenon of “definiteness

inheritance” from the argument to the head.

Finally, D can never be empty at spell-out. This means that DPs where the

head generated under N has no strong feature can’t be grammatical, either with

or without head movement: if the head moves, it violates the principle of

Greed, which allows movement only when this is necessary for checking

features; and if the head doesn’t move, then D remains empty.

3.7.2. Summary of cases

1. Nominal heads (in particular, CS)

When the head of a DP is a noun, it is always specified as [±plural]. Since

this is a strong feature which must be checked in D, raising is always possible

(and actually required). This accounts for the insensitivity of nominal heads to

the definiteness of their complement with respect to raising to D, as opposed to

the sensitivity of determiners.

2. Det-CS: syntactically plural determiners

When the determiner, which is generated under N, has the feature [+plural],

it can raise to D and check this feature, so movement is allowed. This explains

why determiners with plural morphology (which is the realization of this

feature) can head a det-CS, regardless of the definiteness of their complement–

just like nouns.

3. Det-CS: syntactically definite determiners

Heads of CS can optionally have a [def] feature. When this feature is present,

the head must first raise to Agr, where it checks definiteness agreement with

the complement (which also raises to [spec, AgrP]). From there, the head raises

again to D. The lack of phonetic realization of [def] on heads of CS is probably

due to an incompatibility with the phonology deriving these heads. Thus, we

have an account of the fact that det-CS is possible when the complement is

(overtly marked as) definite, even if the head has no number feature.

Page 49: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

44

4. Det-CS: syntactically indefinite, non-plural, determiners

When the determiner which is generated under N has no [±plural] feature

and no [def] feature, movement is impossible. This is due to the principle of

Greed, which states that feature-checking is the only reason for movement. The

case feature can only drive movement to the intermediate Agr, where it is

checked and deleted (being an uninterpretable feature), but not further to D.

Thus, D is left empty at spell-out, and this is presumably not allowed for some

independent reason. Therefore, indefinite non-plural-marked determiners can’t

be heads of CS.

Page 50: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

45

4. Partial nominal projections

My goal in this section is to discuss various constructions involving

“incomplete” nominal projections. These include “determiner phrases”, ellipsis

within the DP and movement from within the DP. I will try to characterize

some of the restrictions that limit these constructions, and to show that the

theory that was developed so far can be naturally extended to include these as

well.

4.1. Determiners in specifier of DP

Under ordinary assumptions of X-bar theory, a specifier position can be

filled only by a maximal projection. I now turn to discuss those determiners

that I am claiming to occupy the [spec, DP] position, and that therefore must be

analyzed as forming some sort of maximal projection.

In pre-DP literature, the term “determiner phrase” is frequent. Considering

particular English data, Jackendoff (1977) labeled some of these as QP, some as

DegP and some as NP. Since this classification is based on a specific analysis of

the English data that does not seem to be easily applicable to the Hebrew data, I

will not adopt it here. Rather, I will try to extend my analysis of the previous

section in order to cover these “determiner phrases” as well.

There are several questions that should be answered:

1. What is the internal structure of the maximal projections headed by free

determiners?

2. What is their categorial status?

3. Are there any limitations on the determiners that can head such XPs?

To answer these questions, we return to the minimal difference between

determiners and nouns, which has been central to my analysis so far.

Continuing the reasoning of previous sections, the simplest and most obvious

option is that these XPs form some sort of a nominal projection. The two

Page 51: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

46

possibilities to consider are thus NP and DP. One important observation, that

might help us decide between these possibilities, is that both plural-marked

and definite-marked determiners can’t occur in their free state in front of a

noun:18

(56) a. /elef/s&los&a /anas&im

thousand/three men

b. */alaf-im/s&nayim19 /anas&im

thousands/two men

c. *ha-s&los&a (ha-)/anas&im

the-three (the-)men

Since the only grammatical cases are those in which the determiner lacks

both definiteness and number specification, we are led to the natural idea that

these determiners are NPs. In the absence of any reason to assume otherwise,

the internal structure of phrases headed by numerals and other simple

determiners can then be assumed to be trivial– these are just simple NPs,

headed by the determiner and having no complements or specifiers.20

Why should plural or definite determiners be barred from [spec, DP] in the

presence of a noun? Recall that in §2.4.2 it was mentioned that other languages

18In contrast, when there is no noun, the free determiner may be specified for definiteness

and plurality:

ra/iti /et ha-/alaf-im

(I) saw ACC the-thousand-s

Such structures involving no noun will be discussed in §4.2.

19The dual suffix -ayim can be considered plural marking.

20 This, of course, does not extend to more complex “determiner phrases”, whose internal

structure I will not discuss here; the crucial property, that I believe should hold of these as well,

is their lack of DP features; the reasoning behind this is given in the following paragraph.

Page 52: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

47

(such as English) do have a fully specified possessor DP in this position. One

line of explanation for this difference is to derive it from genitive case

assignment: in a language such as English, [spec, DP] is a case position, and

therefore a DP can be licensed in it; in Hebrew, on the other hand, genitive case

is assigned in a different manner, as discussed in §3.5, and thus a DP in [spec,

DP] would not be able to check case; NPs, on the other hand, can be argued not

to be subject to the case filter.

An alternative explanation for this cross-linguistic variation has to do with

the fact that in languages of the English-type, there is no raising of N to D. This

seems to suggest some sort of “doubly-filled D” effect: either there is N-to-D

raising, or the specifier of DP can be filled by a full DP, but not both. Under the

feature-checking approach of the minimalist program, we might pursue the

following idea: since the definiteness/number features on the D are checked by

the N which is raised, these features can’t be marked at the same time on the

specifier. If the specifier possessed these features, then it would try to check the

same feature twice (the second time in spec-head configuration), a possibility

which is ruled out, and the derivation would crash.

A similar prediction is made by Borer (1994) under a different theoretical

framework; discussing only the possibility of another DP in the specifier

position of DP, she predicts that in Hebrew [spec, DP] must be empty for

precisely the reasons I have mentioned– to avoid having conflicting features.

Although this possibility isn’t discussed in her paper (Borer ignores all

determiners other than ha-), her analysis actually predicts that featureless NPs

should be possible in this position. These are the only nominal projections that

won’t conflict with features which are checked otherwise on the head.

To sum up, free determiners project NPs, and since they have no number or

definiteness features, they can fill the specifier of DP position without

conflicting with the features of the raised noun. This lack of features in the

specifier is what allows us to explain contrasts such as those given in (56).

Page 53: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

48

4.2. Bare determiners

So far, I have discussed DPs which contain an overt noun, either as the head

of the NP, or else inside the genitive argument in [spec, NP]. Now I will turn to

determiners which appear in isolation, without a noun, and yet behave like full

noun phrases. Some examples are:

(57) a. /axalti s&nayim

(I) ate two

b. harbe /azvu

many left

c. ha-kol nigmar

the-all finished

‘everything is over’

The questions one should answer are how these phrases are licensed, and

what is their internal structure. There are at least three possibilities that come to

mind:

a. These cases involve ellipsis of the noun or of one of the projections below

the DP.

b. These are not full DPs, but just NPs (parallel to those which I claimed in

the previous section to occupy [spec, DP] in full DPs).

c. These are full DPs headed by the determiner, which raises from N to D

and takes no complement.

In §4.2.1 I will examine the first possibility; evidence for the existence of

structures of the third kind will be presented in §4.2.2. As to the second option,

I will leave the question of whether such structures exist as well open.

4.2.1. Ellipsis

A well-known property of elliptic structures is the strict/sloppy ambiguities

they give rise to. Considering this, the ellipsis approach to explaining

Page 54: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

49

“nounless” noun-phrases seems quite straightforward, in view of the fact that

the same ambiguity is witnessed in sentences such as the following:

(58) dani /ohev s&los&a xaverim s&elo, ve-yosi /ohev /arba/a

Dani loves three friends of-his, and-Yosi loves four

This sentence can either mean that Yosi loves four of his own friends, or four

of Dani’s. This is exactly the ambiguity pattern which is typical of other ellipsis

structures (see for instance Fox 1993 for a discussion of VP ellipsis).

If this is in fact a case of ellipsis, then the obvious question is what level is

deleted21. Since the determiner in the full DP of the last example occupies [spec,

DP], the ellipsis must be of some projection below the DP level. The following

example shows that the only possibility in this case is D’ deletion:

(59) dani /axal s&alos& /ugot gvina, ve-yosi /axal /arba

Dani ate three cakes cheese, and-Yosi ate four

‘Dani ate three cheese-cakes, and-Yosi ate four’

Since the head noun, /ugot, which is deleted in the second conjunct, is

assumed to raise to D°, the deleted projection must be at least D’; the presence

of the specifier (the determiner) means that the missing projection is indeed D’.

Under this analysis, there is an obvious prediction: ellipsis should not be

possible in det-CS DPs, because the determiner, which must not be deleted,

occupies in these case the D° position. This prediction is supported by data

such as the following:

(60) *dani /axal s&alos& /ugot gvina, ve-yosi /axal s&tey

Dani ate three cakes cheese, and-Yosi ate two

21I will not commit myself to any particular theory of ellipsis, and speak informally as the

missing element as being “deleted”.

Page 55: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

50

In fact, in this case the only way to have an elliptic construction is with the

free form s&tayim, which can’t co-occur with the noun for reasons that were

discussed in §4.1.

Elliptic conjunctions show a similar pattern:

(61) a. hayu s&am s&ney /anas&im /o le-xol hayoter s&los&a

were there two men or at most three

‘there were two, or at most three, men there’

b. *hayu s&am s&los&a /anas&im /o le-xol hapaxot s&ney

were there three men or at least two

‘there were three, or at least two, men there’

c. hayu s&am s&los&a /anas&im /o le-xol hapaxot s&nayim

Here, too, it can be argued that s&ney, being the head of a CS, occupies a

position within D’, and therefore (61b) is a case of NP-ellipsis– an option which

apparently is ungrammatical.22

One crucial problem to the ellipsis analysis of these structures is posed by

the following sentences, in which an AP (62b) or an argument of the missing

noun (62a) is still present:

(62) a. ra/iti s&los&a sratim s&el felini ve [DP /arba/a [ ] s&el kurosawa]

(I) saw three films of Felini and four of Kurosawa

22An alternative, and perhaps simpler, line of explanation for the ungrammaticality of the

last examples, is related to genitive case checking. Assuming that derived forms of determiners

are selected from the lexicon with the case-assigning feature, this feature must be checked by

assigning case to some DP. In the absence of a complement, the derivation crashes, just as it

does when a noun with this feature has no genitive complement:

a. batey ha-kfar

houses the-village

‘the village houses’

b. *batey

Page 56: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

51

b. ra/iti s&los&a praxim /adumim ve [DP /arba/a [ ] sgulim]

(I) saw three flowers red and four purple

Under standard analyses of Hebrew DP structure (such as Siloni 1994), both

APs and genitive arguments are assumed to be located inside the NP or

adjoined to it; this means that the sentences in (62) can’t be generated by D’

deletion. However, the ellipsis analysis can still be saved in this case, by

claiming that the deleted level here is N’, rather than D’.23 This seems to

suggest the following descriptive generalization:

• Only X’ can be deleted.

As pointed out to me by Tanya Reinhart, similar observations have been

made with respect to ellipsis of other categories as well; therefore, the

structures I have proposed for DPs seem to fit into a general pattern, a fact

which gives my analysis further support.

4.2.2. “Intransitive” determiners

In the previous subsection I have considered a class of “bare” determiners

that appears uniform in its behavior. Now I want to consider some cases that

seem to involve a different structure. Specifically, I’m referring to the bare

determiner ha-kol (‘everything’), and possibly also to ha-rov (‘the most’).

23 A stronger problem is posed by examples such as the following, in which both the head

noun and the genitive argument are missing, while the AP and s&el-phrase are present:

ra/iti s&los&a sirtey kolno/a tovim s&el felini ve [DP /arba/a [ ] mecuyanim s&el kurosawa]

(I) saw three films cinema good of Felini and [four [ ] excellent of Kurosawa]

Depending on the exact theory of ellipsis that one is assuming, it might perhaps be argued

that the N’ is deleted before head-movement; this is even supported by the possibility of

s&nayim, which can’t coexist with a raised head. But then there is the problem of the exact

position of the genitive argument and the AP– a problem which is beyond the scope of the

present work.

Page 57: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

52

There is much evidence in support of the claim that ha-kol differs structurally

from the elliptic constructions of §4.2.1. First, ha-kol is much more restricted in

its interpretation: as opposed to the elliptic determiners discussed earlier, ha-kol

cannot denote an animate object:

(63) a. /axalti ha-kol

(I) ate the-all

‘I ate everything’

b. * pagas&ti ha-kol

(I) met the-all

‘I met everyone’

Second, the strict/sloppy ambiguity displayed in the examples of §4.2.1 can’t

be reproduced by ha-kol:

(64) dani ge/e be s&los&a sfarim s&elo ve yosi ge/e be ha-kol

Dani proud of three books of-his and Yosi proud of the-all

Here the bare determiner in the second conjunct can’t refer to all of Dani’s

books (the “sloppy” reading), and not to all of Yosi’s (the “strict” reading)

either; the only reading is that in which ha-kol means ‘everything’, thus refering

to all contextually relevant entities.

The difference between the two structures is not limited to interpretation,

however. ha-kol can’t take s&el-phrases or APs, as opposed to examples such as

(62):

(65) a. * /axalti ha-kol s&el ha-mis/ada

(I) ate the-all of the-restaurant

b. * /axalti ha-kol ta/im

(I) ate the-all tasty

Finally, as opposed to elliptic determiners, which require an antecedent and

are therfore only marginally acceptable if no antecedent is available in the

Page 58: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

53

sentence (unless there is a discourse antecedent), ha-kol is perfectly acceptable

in such positions:24

(66) a. ? ra/iti s&los&a

(I) saw three

b. ? s&los&a naflu /al-ay

three fell on-me

(67) a. ra/iti ha-kol

b. ha-kol nafal /al-ay

To sum up, what all this shows is that ha-kol deserves a different analysis

than the one suggested for the bare determiners of §4.2.1.

As was already mentioned in §4.1, definiteness can’t be marked on the

determiner in the presence of a head noun. However, note that kol can appear

in isolation only if it has the definite article attached to it:

(68) a. ha-kol/*kol ne/elam

the-all/all disappeared

‘everything disappeared’

b. ra/iti ha-kol/*kol

(I) saw the-all/all

‘I saw everything’

This fact suggests a simple explanation, which is a direct continuation of the

analysis of section 3. A standard assumption is that argument noun phrases

must be DPs; thus, if ha-kol projects to a DP without involving any ellipsis, and

if we maintain the previous idea that determiners are generated under N and

must raise to D, the role of the definite article in ha-kol becomes clear: Just like

in det-CS, the definiteness feature serves here to license raising of the

determiner to D. Since there is no number feature on kol, the only other feature

24It should be noted, however, that indefinite noun phrases in subject position are often

awkward, even when no ellipsis is involved.

Page 59: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

54

that might license this movement is [def]. Therefore, in the presence of the [def]

feature, ha-kol is simply a full DP, headed by the determiner, and taking no

genitive argument. Direct morphological evidence for the last claim is hard to

find, due to the fact that kol has just one syllable and therefore phonetically

ambiguous between the free and the derived form; but the existence of the

article makes it clear that it is the free form here, and the DP is parallel to a

simple nominal such as ha-kise (‘the-chair’).

I have discussed so far only the form ha-kol, but a similar account can be

given to other bare determiners which have the article ha- attached to them,

such as ha-s&los&a (‘the-three’) and ha-rov (‘the-most’), which display similar

behavior with respect to the diagnostics given above. Thus, the presence of the

[def] feature, as witnessed by the presence of the definite article, is what

licenses a full DP in all of these cases.

Note, finally, that rov in isloation can also be licensed by an AP, an option

which isn’t available with kol:

(69) ha-haca/a hitkabla be-rov *(gadol)

the-proposal was-accepted in-majority (large)

‘the proposal was accepted by a large majority’

Two explanations can be given to this fact. One is that this AP, presumably,

provides a [-plural] value through some sort of agreement; or, alternatively, it

can be argued that the presence of an AP forces the nominal (referential)

interpretation of rov- that is, ‘majority’ rather than ‘most’, and this lexical item

has an inherent number feature, just like any other noun.

4.3. Floating quantifiers

One area, where my proposal for two distinct positions occupied by

determiners might gain additional support and make certain predictions, is in

the phenomenon referred to as quantifier float (Q-float). This typically involves

Page 60: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

55

a determiner (quantifier) in an argument position, with the quantified noun

located in a pre-quantifier position:

(70) The children are all gone

If such structures exists in Hebrew, we would expect the structural

differences between the two types of DP discussed in this paper to be reflected

by differences in the possibilities of Q-float.

It turns out that the phenomenon of floating quantifiers occurs in Hebrew

with a much wider range of determiners than it does in English. Following

Sportiche (1988), it was argued by Shlonsky (1991) that Q-float is the result of

leftward movement of the nominal projection which is quantified over, to a

position that precedes the quantifier. Thus, sentences like (71b) and (71c) are

derived from (71a) by leftward movement of the quantified phrase ha-yeladim:

(71) a. ne/elmu kol ha-yeladim

disappeared all the-children

b. ne/elmu ha-yeladim kul-am

c. ha-yeladim ne/elmu kul-am

According to Shlonsky’s analysis, ha-yeladim first raises to the pre-quantifier

specifier position, which is its position in (71b); and from that position, the

noun phrase can move again to its position in (71c).25

25One point that should be noted is that sentences similar to those argued to involve Q-float

may involve a left-dislocated noun phrase (and a resumptive pronoun on the determiner). This

kind of construction is insensitive to any of the standard constraints of movement (although

judgements on this are not absolutely clear):

ha-yeladim s&ama/ti /et ha-s&mu/a s&e kul-am hiclixu ba-bxina

the-children (I) heard ACC the-rumor that all-3PL succeeded in-the-exam

The clitic in this example can be easily grouped together with the clitics that appear in other

left-dislocation sentences in Hebrew, such as:

Page 61: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

56

Turning now to the details of this analysis, we need to decide on the label of

the phrase that moves. As was mentioned earlier, Shlonsky assumes quantified

noun phrases to be QPs, taking a DP complement. Trying to adapt the basic

idea of his analysis to the structures I am arguing for, the simplest idea is to

simply replace the labels of the phrases. So, while Shlonsky argues that the DP

moves through [spec, QP], we might consider the idea that NP moves through

[spec, DP]. But this, however, does not work. Remember that I have adopted

Siloni’s claim that in CS nominals there is an intermediate AgrP; since the

embedded DP raises to the specifier of this projection, we would have to say

that it is AgrP that raises– not a welcome conclusion. A more natural

adaptation of Shlonsky’s analysis to my structure is to assume that it is only the

embedded DP, after moving from [spec, NP] to [spec, AgrP], which moves to

the left of the determiner. The fact that the moved projection can end up in a

case-marked position supports the idea that what we have is really DP

movement.

Combining this analysis of Q-float with my analysis of determiners makes

the following simple prediction: Q-float should only be possible with derived

determiners, since only then is what follows the determiner a full DP. In fact,

what follows the determiners which occupy [spec, DP] is not even a maximal

projection, so movement is obviously ruled out. Furthermore, only in the case

of CS is the specifier position available as a landing site (but still, adjunction to

DP could be a possibility). As the following data shows, the prediction appears

to be correct:

ha-yeladim s&ama/ti /et ha-s&mu/a s&e mis&ehu s&afax /ale-hem mayim

the-children (I) heard ACC the-rumor that somebody poured on-3PL water

One must be careful not to confuse these structures with those involving movement of the

NP, which are the “real” cases of Q-float.

Page 62: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

57

(72) a. ha-yeladim ne/elmu s&los&t-am/kul-am

the-children disappeared three-3PL/all-3PL

‘the three children/all the children disappeared’

b. *ha-yeladim ne/elmu s&los&a/harbe/kama

the-children disappeared three/many/some

We see that in contrast to the derived determiners in (a), the free determiners

in (b) don’t allow Q-float. This contrast strongly supports my claim that there

are two distinct structures corresponding to the two forms (free/derived) of

determiners. For a large class of determiners, we now have a simple

explanation of why these can’t be floated.

A central question which comes up now is whether Q-float is possible with

all derived determiners. Judgements vary with respect to the following cases:

(73) ?/* ha-yeladim ne/elmu rub-am/xelk-am

the-children disappeared most-3PL/part-3PL

‘most/part of the children disappeared’

What could explain the difference between these and the perfectly

grammatical sentences given in (72a)? The first thing that comes to mind is a

semantic distinction; given a DP of the form:

(74)

DP1

D’

D NP

N’

N

DP2

t1

N+D1

The two DPs, DP1 and DP2, may either refer to the same entity, or to two

distinct entities. In the case of coreference, Q-float is acceptable to all speakers,

as in the examples of (72a). For instance, s&los&et ha-yeladim and ha-yeladim both

Page 63: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

58

refer to the group of (contextually relevant) children (which is presupposed to

include 3 members). On the other hand, when the referents are distinct, Q-float

is marginal or ungrammatical, as in (73): rov ha-yeladim does not refer to the

same group of children as the embedded DP, but to one of its subsets. Note that

this generalization carries over to other languages: in English, for instance, Q-

float is available only with all and both– neither of which modify the referent of

the phrase they quantify over.

This generalization, that Q-float is allowed only when the two DPs corefer, is

supported by the following observation: Since I am suggesting that the same

structure is involved in CS headed by nouns and in CS headed by determiners,

we would expect a phenomenon similar to Q-float to exists with nominal CS.

But except for limited examples that can be argued to involve a left-dislocated

noun phrase, this possibility seems not to exist.26 The following data illustrates

the impossibility of leftward movement of the genitive argument in nominal

CS:

(75) a. ne/elmu sifrey ha-s&ira

disappeared books the-poetry

‘the poetry books disappeared’

26 While the (b) sentence is stylistically marked, it is nonetheless grammatical:

(a) kavu /orot ha-/ir

went-out lights the-city

‘the city’s lights went out’

(b) ha-/ir kavu /orot-eha

However, in contrast to unquestionable cases of Q-float, the complement in this case (ha-/ir)

can’t come between the verb and the head noun:

(c) * kavu ha-/ir /orot-eha

This suggests that perhaps (b) is just a case of left-disloaction of ha-/ir.

Page 64: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

59

b. * ne/elmu ha-s&ira sfare-ha

disappeared the-poetry books-3-SING-FEM

c. * ha-s&ira ne/elmu sfare-ha

The generalization outlined above, that leftward movement of the embedded

DP is possible only if it corefers with the entire DP, can capture the

ungrammaticality of the (75b) and (75c) sentences: the DPs ha-s&ira and sifrey ha-

s&ira clearly refer to two distinct entities.

In fact, there is one case in which a nominal CS does bear the same referential

index as its argument: the case of reflexives. Hebrew marks reflexivity by using

the word /ecem, which is in all respects a noun that can be translated as ‘self’.

When used as a reflexive marker, the person and number specification on this

noun come from a pronominal clitic, as the following examples illustrate:

(76) a. ha-yeled /ohev /et /acm-o

the-boy loves ACC self-3-SING-MASC

‘the boy loves himself’

b. ha-yelad-ot /ohav-ot /et /acm-an

the-girls love ACC self-3-PL-FEM

‘the girls love themselves

In addition to its appearance as a reflexive marker, the word /ecem can also

head a CS, in which case it has no clitic adjoined to it; in this case, it acts as a

logophor, i.e., an anaphor which does not reflexive-mark any predicate (see

Reinhart & Reuland 1993):

(77) /ecem ha-ra/ayon margiz /oti

self the-idea irritates me

‘the idea itself irritates me’

Now, since the same referential index is shared by the internal DP and the

CS containing it, there is no apparent reason, as far as the discussion so far is

considered, why the internal DP can’t move leftward in this case, just like in Q-

Page 65: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

60

float. The resulting sentence (in which the clitic shows up on the head, just as it

does in Q-float), is indeed grammatical:27

(78) ha-ra/ayon /acm-o margiz /oti

the-idea self-3-SING-MASC irritates me

We can therefore conclude that the descriptive generalization arrived at earlier

is correct: Coreference of the genitive DP and the CS containing it is a necessary

condition for leftward movement of the internal DP.

Since we are discussing a syntactic phenomenon, it is clearly desirable to

formulate this last generalization in syntactic, rather than semantic terms. The

first step is to replace coreference with coindexation:

• Leftward movement of the embedded DP from a CS is possible only if it

has the same index as the CS itself.

What could be the reason for this restriction? The data involving reflexives

suggests one line of explanation: in Reinhart & Reuland’s recent theory of

reflexivity, it is claimed that “self”-type anaphors must be part of a chain,

which is subject to a certain well-formedness condition. I want to suggest that

the two DPs involved in Q-float, too, must form a chain– which is roughly

equivalent to saying that the two DPs must have the same referential index.28

27For some reason, the embedded DP must remain adjacent to /ecem, as the following

example shows:

*ha-ra/ayon margiz /oti /acm-o

One speculative idea is that the reason for this has to do with the role of /ecem as a reflexive

marker, and from the fact that Hebrew doesn’t seem to allow logophors as other languages do.

28Note that this requires us to assume that the moved DP adjoins to the CS, rather than

occupying its specifier position; this way, the two-segment category DP1 is c-commanded by

the adjoined DP2. As far as I can see, there is no problem with this assumption.

Page 66: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

61

The restriction that Q-float must form a chain might be the result of factors

related to case assignment. After movement of the quantified DP, there is only

one case available for the two DPs (nominative, in the following example):

(79) ha-yeladim ne/elmu kul-am

the-boys disappeared all-3-PL-MASC

The first DP, ha-yeladim, receives nominative case from the tensed I; the

second DP (kul-am), however, receives no case. An analysis of these

constructions as involving a chain solves this problem: Most formulations of

the chain condition (see for instance Fox 1993 for a discussion of chains in the

context of Reinhart & Reuland’s theory) require a chain to have one, and only

one, link which is case-marked.29 Similar considerations related to theta-role

assignment can also be given.

We are now in a position to consider one last question related to Q-float in

Hebrew: the role of the clitic on the determiner. According to Shlonsky, this

clitic is an agreement marker necessary for meeting the ECP. However, another

idea that Shlonsky mentions in a footnote is that the role of the clitic is to

“absorb” the case feature of the head. Since, under my proposal, derived

determiners have a case feature which must be checked, this last idea seems

very attractive. Shlonsky rejects this idea, because it does not allow him to

explain the ungrammaticality of leftward movement from nominal CS (as in

75); but since I have already given an independent account of this fact, I can see

29As Ziva Wijler and Naama Friedman pointed out to me, when the floated quantifier is

preceded by the preposition be- (‘in’), it is no longer subject to the coindexation requirement:

(a) * ha-yeladim ne/elmu rubam

(b) ha-yaladim ne/elmu be-rubam

This can easily be accounted for by the preposition’s ability to case-mark its complement; I

am not completely certain, however, whether the underlying structure is the same in the

presence of the preposition.

Page 67: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

62

no reason not to accept the idea that the clitic indeed absorbs the “extra” case

feature of the head.30 As Shlonsky mentions, this is similar to the role of clitics

in clitic-doubling constructions, which again absorbs the structural case

assigned by the head:

(80) a. beyt-am s&el ha-/anas&im

house-3-PL-MASC of the-people

b. * beyt s&el ha-/anas&im

So I will conclude that if the clitic didn’t absorb the genitive case in the Q-

float examples such as (79), then there would be an unchecked case feature on

the determiner, and the derivation would crash. In the presence of the clitic, the

result of movement is a chain with exactly one case-marked node.

30Or, alternatively, that the clitic is attached to a head which is drawn from the lexicon with

no case feature to check; this formulation might be more consistent with the discussion above

regarding case in chains.

Page 68: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

63

5. Determiners and the semantic interface

Throughout this paper, I have concentrated on purely syntactic properties of

determiners. Much of the research on determiners, however, has focused on

their semantics. I want now to discuss briefly a few questions that arise which

are related to the syntax-semantics interface.

At first sight, it might seem that the two syntactic positions I have argued for

correlate with the well-known weak/strong distinction (Barwise & Cooper

1981, Keenan 1987) in the following way: weak determiners seem to occupy the

[spec, DP] position, while strong ones appear to form det-CS. However, this is

not a correct generalization, since some weak determiners can also head CS :

(81) s&ney//esrot sfarim

two/tens books

‘two/tens of books’

As to cases of strong determiners occupying [spec, DP], it is a bit harder to

find decisive evidence, since the two “candidate” determiners, kol and rov, are

possibly ambiguous between the free and derived forms, and there is no

phonological evidence to help us decide. But consider example (82), which can

be argued to involve kol in its free form occupying [spec, DP]: given its lack of

number/definiteness features, we wouldn’t be able to account for any raising

of the determiner from N to D in this case; the only analysis compatible with

the theory developed in this paper is that this is not a det-CS structure.

(82) kol yeled

every boy

Therefore, it is clear from (81) and probably also from (82) that the syntactic

duality displayed by Hebrew determiners does not correlate with the

weak/strong distinction, and whatever correlation we find is accidental. But

why is it that the correlation seems to be “almost true”?

Page 69: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

64

If my explanation of what licenses raising to D is correct, then the apparent

correlation is easily accounted for by the following reasoning: “strong”

determiners are generally those which take definite complements; therefore,

they have the [def] feature that allows them to raise and form CS. “Weak”

determiners, on the other hand, usually take indefinite complements.

Therefore, they can form CS only in the cases (which are rather limited) when

they are marked for [+plural]. Therefore, with the exception of syntactically

plural determiners such as /alaf-im and /asar-ot, the syntactic feature [def] is

what separates the DPs into two classes. Thus the generalization in these cases

is that det-CS is formed with definite complements, which “accidentally”

coincides with the appearance of a strong determiner.

A related question has to do with the semantic implications of the fact that

definiteness is a syntactic feature in Hebrew. Is syntactic definiteness always

mapped into semantic definiteness? This seems to be the case, but the converse

does not hold; as I have mentioned earlier, semantic definiteness can be also

achieved by means of demonstratives, even without having syntactic

definiteness. This is perfectly natural, since we’re not assuming syntactic

indefiniteness. So, we can simply maintain the idea that syntactic definiteness is

always mapped into the corresponding semantic property.

Finally, a less obvious question is whether free determiners differ in their

lexical semantic value from the corresponding derived forms, or is the

difference purely syntactic. For example, does the free numeral s&los&a, which

appears only with an indefinite NP, have a different semantic value from that

of its derived form s&los&et, which can only head a CS with a definite

complement? I believe that the answer to this question is negative, and that all

the semantic differences between such pairs are obtained compositionally– but

I leave this for further research.

Page 70: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

65

Summary

Two related problems were at the center of this paper: the problem of

determining the position of the determiner within the noun phrase; and the

problem of accounting for the various distributional restrictions displayed by

Hebrew determiners.

The main hypothesis was that determiners in Hebrew are actually nouns.

This allowed us to explain determiner positioning using the same syntactic

processes that are used to explain nominal CS– in particular, head movement

from N to D. This gave us a simple solution to the first problem, without

having to make use of any mechanism which isn’t independently assumed in

the theory of Hebrew DPs.

In order to account for the range of differing distributions among

determiners, as well as for the difference in distribution between nouns and

determiners, two formal features were used: number and definiteness. The

hypothesis that determiners lack specification for the first of these features was

argued to be the defining property of determiners, which are otherwise

syntactically equivalent to nouns. This minimal distinction was the central

point used to explain the variations in determiner distribution. Using the

minimalist program’s principle of Greed, the lack of formal features allows us

to explain the impossibility of certain operations of head movement, resulting

in the observed distributional restriction.

Finally, the theory that was developed was used to explain some additional

data: restrictions on the distribution of “bare” determiners (those not followed

by any overt noun), and the phenomenon of floating quantifiers. The observed

data was shown to fit in naturally with the theory, giving some additional

insight into the processes already explained.

An important goal of this paper was to clearly distinguish the syntactic

behavior of determiners from their semantics. The notion of syntactic

Page 71: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

66

definiteness, as opposed to semantic definiteness, was shown to play a central

role in the formation of noun phrases, regardless of their semantics. Similarly,

plurality as a formal syntactic feature, as opposed to the semantic notion of

plurality, was used throughout the analysis to derive a purely syntactic account

of the distribution of determiners.

Page 72: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

67

References

Abney, Steven Paul: 1987, “The English Noun Phrase in Its Sentential Aspect,” Ph.D.thesis, MIT.

Babby, Leonard H.: 1987, “Case, Prequantifiers, and Discontinuous Agreement inRussian,” Natural Language and Linguistic Theory 5, 91-138.

Barwise, Jon and Cooper, Robin: 1981, “Generalized Quantifiers and NaturalLanguage,” Linguistics and Philosophy 4.

Borer, Hagit: 1994, “Deconstructing the Construct,” Ms., Umass Amherst.

Cardinaletti, Anna and Giusti, Giulina: 1991, “Partitive ne and the QP-Hypothesis: aCase Study,” Ms., University of Venice.

Cardinaletti, Anna: 1994, “On the Internal Structure of Pronominal DPs,” TheLinguistic Review 11, 195-219.

Chomsky, Noam: 1986, Barriers. Cambridge, Mass.: MIT Press.

Chomsky, Noam: 1993, “A Minimalist Program for LinguisticTheory,” in The Viewfrom Building 20, Hale, K. and Keyser, Samuel Jay, Eds. Cambridge, Mass.:MIT Press, pp. 1-52.

Chomsky, Noam: 1994, “Bare Phrase Structure,” Ms..

Chomsky, Noam: 1995, The Minimalist Program. Cambridge, Massachusetts: TheMIT Press.

Cornilescu, Alexandra: 1995, “Rumanian Genitive Constructions,” in Advances inRoumanian Linguistics, Cinque, Guglielmo and Giusti, Giuliana, Eds. JohnBenjamins Publishing Company.

Fassi-Fehri, Abdelkader: 1989, “Generalized IP Structure, Case and VS Word Order,”in MIT Working Papers in Linguistics, Laka, I. and Mahajan, A., Eds. MIT, pp.75-113.

Fassi-Fehri, Abdelkader: 1993, Issues in the Structure of Arabic Clauses and Words.Dordrecht: Kluwer.

Fox, Daniel: 1993, “Chain and Binding– A Modification of Reinhart and Reuland's'Reflexivity',” Ms., MIT, Cambridge, Mass..

Fruchtman, Maya: 1979, “Determinative & Deliminative Categories in Israeli Hebrew,”Ph.D. thesis, Tel-Aviv University, In Hebrew.

Giusti, Giuliana: 1994, “Enclitic Articles and Double Definiteness: A ComparativeAnalysis of Nominal Structure in Romance and Germanic,” The LinguisticReview 11, 241-255.

Giusti, Giuliana: 1995, “Heads and Modifiers among Determiners: Evidence fromRumanian,” in Advances in Roumanian Linguistics, Cinque, Guglielmo andGiusti, Giuliana, Eds. John Benjamins Publishing Company.

Grosu, Alexander: 1994, Three Studies in Locality and Case. Routledge.

Page 73: The Syntax of Determiners in Hebrew - BIUdanong1/papers/Danon1996-MA_thesis.pdf · The Syntax of Determiners in Hebrew 3 2. Structure of the Hebrew DP 2.1. The marking of nominal

The Syntax of Determiners in Hebrew

68

Hazout, Ilan: 1990, “Verbal Nouns: Theta Theoretic Studies in Hebrew and Arabic,”Ph.D. thesis, University of Massachusetts at Amherst.

Jackendoff, Ray: 1977, X'-Syntax: A Study of Phrase Structure. MIT Press.

Keenan, Edward L.: 1987, “A Semantic Definition of Indefinite NP,” in TheRepresentation of (In)definiteness, Reuland, Eric J and ter Meulen, Alice G. B.,Eds. MIT Press.

Longobardi, Giuseppe: 1994, “Reference and Proper Names: A Theory of N-movementin Syntax and Logical Form,” Linguistic Inquiry 25.4.

Reinhart, Tanya and Reuland, Eric: 1993, “Reflexivity,” Linguistic Inquiry 24.4.

Ritter, Elizabeth: 1991, “Two Functional Categories in Noun Phrases: Evidence fromModern Hebrew,” Syntax and Semantics 25.

Shlonsky, Ur: 1991, “Quantifiers as Functional Heads: A Study of Quantifier Float inHebrew,” Lingua 84, 159-180.

Siloni, Tal: 1991, “Noun Raising and the Structure of Noun Phrases,” in MIT workingpapers in Linguistics 14. MIT Press.

Siloni, Tal: 1994, “Noun Phrases and Nominalizations,” Ph.D. thesis, Université deGenève.

Sportiche, Dominique: 1988, “A Theory of Floating Quantifiers and Its Corollaries forConstituent Structure,” Linguistic Inquiry 19.3, 425-449.

Szabolcsi, Anna: 1994, “The Noun Phrase,” Syntax and Semantics 27.

Valois, Daniel: 1991, “The Internal Syntax of DP,” Ph.D. thesis, UCLA.

Vergnaud, Jean-Roger and Zubizarreta, Maria Luisa: 1992, “The Definite Determinerand the Inalienable Constructions in French and in English,” Linguistic Inquiry23.4, 595-652.