Subject positions and copular constructions

Texte intégral


Book Chapter


Subject positions and copular constructions


SHLONSKY, Ur. Subject positions and copular constructions. In: Bennis, H., Everaert, M. &

Reuland, E. Interface strategies . Amsterdam : Royal Netherlands Academy of Arts and Sciences, 2000. p. 325-347

Available at:

Disclaimer: layout of this document may differ from the published version.

1 / 1


Ur Shlonsky

Subject Positions and Copular Constructions·

1. Introduction

This paper studies the positional options open Lo preverbaJ subjects in Hebrew. I am led to cast doubt on the idea that there is a unique, canonical subjecl position. Rather, what emerges is a cartography of ubje l p sition , based on a number of interconnected fa tor : The referential talu f the ubject (i.e., whether iL i referential, argumental or expletive), lhe weak/strong di tinclion and the person fea1ures of pronouns.

Evidence from a brand of negative sentences supports 1he view lhat lhere arc at least) two VP-external positi ns for subject in Hebrew. Summarizing the result. exposed in Shlonsky 1997), the annly is is developed and lhen extended L the realm of copular con truclion . ll i argued thal the tw terms of identity sentences exploit two subje 1

positions. For reasons of brevity, I single out and discus ne long-slttnding sel of problem in the analysis of copular constructions, namely,, that of pronominal ubjects. Some additional problems and a somewhat speculative set of answers are explored in the last section.

2. Subjects above and below negation

When appearing in a sentence with a verbal predicate, the Hebrew negative head 'eyn can either precede the subject or follow it.1 When 'eyn follows the subject, and only

"' Many rhanks to Anna Cardinalelti who fir t uggested this line of research and spent many hours dis us ing it with me. Her own work on subjec1 positions in particular Cardinalctli ( 1997 ) has been for me a source of inspiration. I nm al o gra.~eful to Edit Doron for written comm nts, to Oulgielmo Cinque for lengthy discussions, ro Cecilia Polclto, Luigi Rizzi. audiences al the universities of Venice, Geneva, lomouc, Sienna, Southern California as well as to the participants and organizer or the Dutch Royal Academy colloquium on Interface Strategies in Amsterdam.

1 'eyn is restricted to appear in clauses with 11 present ten e verb or a nonverbal predicate, for

reasons discussed in Shlonsky (1997). rn its nonexlstential, nonpossessive use, 'eyn is a literary fonn. Predicates of all tenses can be negated by the particle lo, which, unlike 'eyn, is insensitive to the tense of the predicate or to its lexical category.

Ur Shlonsky 325


. . . . anifests an agreement suffix, the fonn of which when it follows it, it obhgatonly m b and person features of the clausal

. d with the gender, num er Th

alternates m accor ance ·11 t the two formats of 'eyn sentences. e

t es in (1) below i ustra e 1 d · (2)

subject. The sen enc . h , follows the subject is tabu ate m · paradigm of suffixes which appear w en eyn

(l) a. 'eyn ~ina mgdaberet rusit'.

neg Rina speak:fs Russi'.111 , 'Rina does not speak Russ1~n.

b. Rina 'eyn-a mgdaberet ruR sit..

Rina neg:3fs speak:fs ~ssi~n 'Rina does not speak Russian.

(2) Agreement Paradigm of 'eyn Singular Plural

1 -enu

2m -xa -xem

2f -ex -xen

3m -(gn)o -am

3f -(gn)a -an

alternation illustrated by the sentences in (1) In Shlonsky (1997), I argue


the. . . Hebrew one to the right of and below provides evidence for two subject ?ositions m ,

Nego and one to its left and above it. . (la) is the specifier of a functional It can be shown that the subject below 'eyn •. I~ dir~ct but straightforward. The verb category and is not VP-internal. The. argumehnt ishi~t 'well' as shown by the contrast

d r adverbials sue as e1 ev ' f VP

in (la) must prece e manne 1 sion that the verb has raised out o in (3). The

acceptabilit~ ~f

(3).leads to the.cone u ears to the left of the verb, it must be over that adverbial posiuon. Smee th.e


app me position in the middlefield, below

the case that it is not in VP. Rather, it occupies so . negation and above VP.2

(3) a. *'eyn Rina heitev mgdaberet rusit._

neg Rina well speak:fs Russian 'Rina doesn't speak Russian well.' b. 'eyn Rina mgdaberet heit

11ev rRusit..

neg Rina speak:fs we ussian 'Rina doesn't speak Russian well.'

. . . osition under negation. The agreement In (la) and (3b)'. the subject is licensed

i~ s~:~h~se

examples is participial


morphology which appears on the dver f t res) and is associated with a participial (manifesting only number and gen er ea u

b' · lso raised

2 heitev can also follow the direct object. I assume that in such cases the o Ject ts a above the adverb, cf. Kayne (1994).

. . I C tructions

Subject Positions and Copu ar ons 326

agreement head. The present tense verbal fonns are, morphologically, participles and occur as such in periphrastic constructions such as the past habitual (4).3

(4) Rina hayta mgdaberet rusit.

Rina be:PAST speak:fs Russian 'Rina used to speak Russian.'

Th representation f (I b) is m re familiar. (I b) can be ass ciate I with a conventional phrase marker in which AgrSP dominates NegP aud the subject appears in Spec/AgrS.

The range of po sible subject-types is reslri ted both under nonagreeing 'eyn and ab vc the agreeing ne. onsider frrsl the 'ey11"subjecl format.

A referential null subject is ungrammatical under 'eyn, as shown in (5).

(5) *'eyn mgdaberet rusit.

neg speak:fs Russian

'I.f/You.f/She don't/doesn't speak Russian.'

This is due to the fact that the verbal form appearing under 'eyn (a participial form; see above), lacks person features. Person features are necessary for the identification or assignment of the content of the referential null subject. As shown in (6), a null subject cannot appear as the subject of such verb forms, even when 'eyn is absent.

(6) *m~daberet rusit.

speak:fs Russian

'I.f/You.f/She speak(s) Russian.'

The paradigm in (7)-( I I) is more surprising in that it illustrates, for a variety of non- refercntiat null subj cts, u nlra 't between 'ey11 sen ten es and both negative and affinmuive semences cmpl ying lhe rune verbal form. N n-ref' ren1ial null ubj els (in lhe sen e of Chom ky 19 I), i.e., null expletives, quasi-referential or argumenlal pro, pseudo-exi tentinl arbitrary pr Cinque 1988) etc.) ar perfectly Ii it a subjects of present tense verbs in Hebrew whether r n l !he verb is preceded by the negative particle lo (see note 1.) Under 'eyn, however, all tokens of such null subjects are robustly ungrammatical.4

3 For recent discussi n of Lhe syntax of 1he H 'brl.!W present Lense, see Shlonsky (1997) and Siloni (1997).

4 Concerning the data in (7a) and (9a), it should b n Led that the ungrammaticality of subject inversi n us in the form r example) and ot' arbitrary pro in the latter is suspended when the sentences are interpreted generically. Thus, (7a) contra. ts with the acceptable (i) and (9a) with the grammatical (ii).

(i) •eyn yordim kan gfamim ba-derex klal.

neg fall:pl here rains usually 'Rain doesn't usually fall here.' Lit: 'It doesn't usually rain here.' (ii) 'eyn moxrim kan sigariot

neg sell:pl here cigarettes 'They don't sell cigarettes here.'

Ur Shlonsky 327


(7) a. *'eyn yored gesem b~-Telavi~.

neg falls rain in Tel-Av~v , 'Rain doesn't fall in Tel-Aviv.

Lit. 'It isn't raining in Tel P:viv.' b. (lo) yored gesem b~-TelaAvi~.

(neg) falls rain in Tel- vi~ , 'Rain does(n't) fall in Tel-Aviv'. , Lit. 'It is(n't) raining in Tel Aviv.

(S) a. *'eyn kase li-lmod rusit'.

neg difficult to-learn Russian . , 'It isn't difficult to learn Russian.

b. (lo) kase li-lmod rusit..

(neg) difficult to-learn Russian . , 'It is(n't) difficult to learn Russian.

(9) a. *'eyn dofkim ba-delet.

neg knock:mpl on-the door , 'There isn't anyone knocking on the door.

b. (lo) dofkim ba-delet.

(neg) knock:mpl on-the door . ,

'There is(n't) someone/anyone knockmg on the door.

( 1 O) a. *'eyn carix l~-daber 'it-o.

neg must:3ms to.speak 'One shouldn't speak to him.

b. (lo) carix l~-daber 'it~oh. h.

neg must:3ms to.speak wit : i~

'One should(n't) speak to him.

( 11) a. *'eyn kar.

neg cold 'It isn't cold.' b. (lo) kar.

(neg) cold

'It is(n't) cold.' .

. he (a) and (b) examples of (7)-(11) i~ ~he I suggest interpretmg the contrast

(b) examples occupies a different pos1tton following way. Suppose that the SU


of lar suppose that it is the specifier of some from the one in the (a) examples_. n par tc~l 'biects and is able to assign to them the functional head which formally hcenses nu su J. 5 .

f on referential interpretation. . f

features necessary or a n - . d , h uld be taken to be the specifier o a By a similar logic, the subJeC~ un er


sb? ts Hence all types of null subjects, functional head which does not hcense nu su ~ec . '

. . d b mber specification.

5 Argumental pro is generally. iden~ifiable wh~n .the hcensmg hea ears a nu An expletive requires no ident1ftcat1on. See R1zz1 (1986).

· ns

Subject Positions and Copular Construct10

whether referential or not, are formally barred from that position. Following R" · (1986), take the head formally licensing null subjects to be the head responsible 1


nominative Case and consider (12). or

(12) The head of which a subject under 'eyn is the specifier does not have a nominative Case feature.6

A different interpretation of the data in (7)-(11) might sugge. L ilself, to the effect that the constraint observed by subjects under 'eyn is not that they must be oven, hut r•1th~r

that they must be fully referential. This constraint is both too strong and t o weak. The constraint is too weak since it fails to rule out (5) and it is too strong since iL incon-cctly rules out (13) below, where the subject is the impersonal quasi-referential pr noun

z »

(which in many respects resembles Dutch het, as per Bennis ( 1986), Vikner 19 5), among others; viz. Hazout (1994).) (13) should be compared wilh Sa).

(13) 'eyn ze kase l~-daber rusit.

neg it difficult to-speak Russian 'It isn't difficult to speak Russian.'

I therefore conclude that there is a subject position under 'eyn in which nominative Case is not available and since pro must be licensed by nominative Case it is ruled out from the subject position under 'eyn, which can only be filled by a lexical subject. Lexical subjects, I shall argue below, are licensed under 'eyn by a structural default Case.

Let us turn now to the restrictions which hold of subjects to the left of agreeing 'eyn.

Contrary to the situation obtaining under 'eyn, referential null subjects are licensed above this negative head. Sentence (14) contains a first person singular null subject.

(14) 'eyn-(~n)i m~daberet rusit.

neg: ls speak:f Russian 'I don't speak Russian.'

All combinations of number and person agreement on 'eyn yield grammatical null subject sentences, with the exception of third person singular and plural. The ungrammaticality of (15) is indeed, surprising, since the third person agreement features are discretely represented on 'eyn (see (2).) Third person referential pro has a peculiar distribution in Hebrew, which has been treated in different ways by a number of researchers. I set aside the problem posed by the ungrammaticality of (15).7

6 I as ume here that in the absence of nominative Case, a subject i assigned a default Case, on which. see below. See H nry (l 995) for discussion of nominafrvc subjects which are Case marked by AgrS and non-nominative ones whose Case i a signed by Tense.

7 Por furLher discu ion. see Borer (1986. 1989), Dor n 1983), Eli ha (1997), Ritter (1995) and Shlonsky (1990, 1997).

H. Borer (p.c.) points out that a third person referential pro is admissible as an embedded controlled subject, se below. and patterns, in this respect, with subjects of future tense verbs (see Borer 1989 for further discussion).

Ur Shlonsky 329


(15) *'eyn-(:m)o mgdaber rusit._

neg.3ms speak:ms Ru~s1a?

'He doesn't speak Russian.

. . ver is sufficient to demonstrate that pro is formally The acceptab1hty of (14), howe . ' ' I line with our previous discussion, we can licensed in its position above agreemg eyn. n

state this in the following terms. .

. b' b (agreeing) 'eyn has a nomma- ( 16) The head of the projection hostmg a su ~ect a ove

tive Case feature.

. ' b r a (strong) nominative feature, movement of a If the agreement



~ e~n e~

s th d to check this feature in the overt syntax.

nominative DP to its


is dnven hy 'e


agreement, the subject must precede It is thus correctly predicted that w en ey 7 ) d (l 7b)

'eyn and cannot follow it, as in the contrast between (I a an . (17) a. Rina 'eyn-a mgdaberet rusit..

Rina neg:3fs speak:fs Russian 'Rina does not speak Russian.' b. *'eyn-a Rina mgdaberet rusit..

neg:3fs Rina speak:fs R~ss1~n

'Rina does not speak Russian.

. . nstrates that all instances of non-referential pro are The paradigm unacceptable above m (18) 'eyn. As we ave below demh o JUS . t see ' n this generalization cannot be ascribed to the formal illegitimacy of pro.

( 18) a. *'eyn-o yored ge8em bg-Telavi~

neg:3ms falls rain in T~l-~v1v . , 'Rain doesn't fall/isn't falling m Tel Aviv.

b. *'eyn-o kase li-lmod rusit.

neg:3ms difficult to-learn Rus~ian,

'It isn't difficult to learn Russian.

c. *'eyn-am dofkim ba-delet.

neg:3mpl knock:mpl on-the door 'No one is knocking on the door.' d. *'eyn-o carix lg-daber 'it~o.

3 neg:3ms must:3ms to-spe~ ';1th: ms 'One shouldn't speak to him.

e. *'eyn-o kar.

neg:3ms cold 'It isn't cold.'

Dani amar se eyn-o m;idaber rusit..

Dani said that neg.3ms speak:ms Ru~sta~

'Dani said that he doesn't speak Russian.

330 Subject Positions and Copular Construe io f ns

Tbe generalization unuerlying tl1is paradigm is that. only referential subjects ar admitted in the subject p sition above eyn. This is further onfim1ed by the ungrammaticality f (19) where lhe subjec1 position to the left F agreeing 'eyn is ccupied by the qua ·i- referential pronoun ze (compare witJ1 ( 13) above.)

( 19) *ze 'eyn-o kas l1:i-daber rusi1.

it neg:3ms difficult to-speak RussiUJ1 'lL isn't difficult to speak Russian.'

lf the subject in agreeing 'eyn sentence is in Spec/AgrS and must b referential, then why is it the case that non-referential ubjects in ·enlences with uL 'eyn are pe1t'ecUy acceptable? In olher words, how hould one explain the comrast between ( 18a-e) and lhe (b) examples of (7)- 11 )?

The imp ssibiliLy I' n n-referentiaJ subje I to the left of 'e n can be expr s et.I by saying that in contrast to subje t agreement morphology on finite verbs, th agreement rn rphology on 'ey11 (ll:1bulaled in (2)) calls for tile prnjecti n f an (a)rgumental topic p sition (I tha11k Luigi Rizzi for discu ion of this poinl). Characterizing the pre-agreeing


eyn subject posiLi.011 in Lhis way immediately explains why only fully referential subjects may app ar there whereas quasi-arguments and expletives may not.

Orne insighL into the 't picalily' of this positi n can be gained by considering the formal identity of the agreement suffixes on 'eyn see (2)) and th e borne by nouns, prepositions and, with minor phonetic diff ren ·es, transitive verb in examples such a (20a ) below.

(20) a. l-o to:3ms 'to him' b. beit-o

house:3ms 'his house' c. fa-hazmin-o

to-invite:3ms 'to invite him'

Robert and Shlonsky ( 1996) and Shlonsky ( 1997) argue that the suffixes in (20a-c) are Agr head , to which a lexical head oujoin (P, N and Y_

111N, respectively). The specifier of these Agr head is pro, coindexed wirh Agr.

One of the properties f these clitic-like suffixes is that they must be associated with fully rcl'erential arguments. Verbs like 'Jead' in 'this leads to th conclusion that. .. ' have either fully referential or arbitrary (generic) direct objects, see (2Ia,b). The arbitrarily- refen"ing theme may be realized by e.g. the profonn 'one' or it may be lexically saturated, (see Rizzi (1986).)

(21) a. This leads John to the conclusion that...

b. this leads (one) to the conclusion that ...

Ur Shlonsky



. . ff the theme must be interpreted as fully 'movil' bears a direct-object su ix,



reTwh' i's shown by the interpretations under (22b).

referent.a . is

(22) a. ze movil la-maskana se. · · this leads to-the-conclusion that 'This leads to the conclusion that. .. ' b. ze movil-o la-maskana



this leads:3ms to-the-cone us1~n ' 'This leads him to the conclus.10n that ...

*'This leads one to the conclus10n that. ..

. e's castle' where the nominal possessors are Similarly, expressions




is on l be rendered periphrastically, as in (23a) arbitrary and hence quas1-refere~t1al can on y

(with clitic-doubling) but not as m (23b).

beit-o 8el 'adam hu 'armon-o.

(23) a. house:3ms of person it castle:3ms 'A person's home is his castle.' b. beit-o hu 'armon-o.

house:3ms it castle:3ms 'His home is his castle.'

*'One's home is one's castle'

. . . . osed b Roberts and Shlonsky( 1996) and The formal treatment of Sem1t1c

c.ht1c~ pro~6 J

sis of Romance clitics as heads of Shlonsky (1997) is similar to Sport1cheh.s h(l.9 ')thaenr aypro or a (clitic-doubled) argument.

. . th specifier of w ic is e1 ' . f .



.. e l like Semitic ones (and like subjects o agreemg Romance object cht1cs behave exact y f t' I . terpretation Consider the contrast

1 h e a fully re eren ia m · f · 1

'eyn) in that they can on Y av .


of (25) where a non-re erentia between (24a) and (24b), and the ungrammatica I y ,

interpretation of the clitic is forced.

(24) a. Cela rend heureux.

this makes happy 'This makes one happy.' b. Cela le rend heureux.

this 3ms makes happy 'This makes him/*one happy.

(25) * Je le consioere probable que tu viennes.

I 3ms consider probable that you c~me

'I consider it probable that you come. . . ._

. . iated with the clitic projection with spec1fic1 While Sportiche identified the feature assoc 'fic1'ty feature Cecchetto (1995)

h l.t. roiection houses a spec1 ' 1

ty and argued that t e c J ic p J h' t be entirely correct. For examp e,

, . ) les which show that t is canno ) t c

provides (Italian exa?1p l' . . th Clitic Left Dislocation (CLLD cons ru - generic DPs can be picked up b~ a c ittc m e

tion, as shown by the example m (26).

332 Subject Positions and Copu ar on I C structions

(26) Un Italiano lo riconosci sempre al primo colpo.

an Italian 3ms-(you) recognize always at first sight 'An Italian, you always recognize him at first sight.'

A propos of the example in (27), Cecchetto notes that " ... What seems to be necessary is a linking between the [CLLD'd] DP and the previously established context of discourse."

(27) un pasta Gianni !'ha fatto. a meal Gianni it-has taken 'A meal, Gianni has eaten.'

It seems justified, therefore, to think of tJ1e prop rty associated with pronominal Jjlics as topicality. This characterization is sufficient to exclude all the cases of non-referential clitics, of the sort illustrated in (24b) and (25). The ubject position of agreeing )ey11 i an argumemal topic position and i thus partially assimilated to the specifier p itions of'

litic projection in Semitic and Romance.

A opp sed lo the agreement uffixes on nouns, verbs, prepositions and particles like 'eyn, subject agreement on finite verb in Hebrew imposes no restrictions on the referent- iaJity of the subject. We must lhcrefore distinguish a subject position which is a topic and a subject po ition which is not.R

If Hebrew has two subject positions, one b low negation and one above it, one wonders whether there are cases when b th are used and whether there are additional subject positions in lhe clau e. Both of these questions are taken up in the rest of this article.

3. Subjects in copular constructions

Present tense copular sentences in Hebrew are characterized by the absence of a verbal copula, as shown in (28).


Note, in passing, that quantified DPs cannot be topicalized (i.e., clitic left-dislocated), but can appear as subjects of 'eyn.

(i) *'g, hu 'eyn-o mu~lam.

no one he neg:3ms perfect 'No one, he's perfect.' (ii) •g 'eyn-o mu~lam.

no one neg:3ms perfect 'No one is perfect.'

If a quantified DP is quontifier-raised in LF from an A' topic position, as in (i), it would have no variable to bind, since rhe trace of topicalization in Spec/IP would already be bound by the topic.

No such problem aris in (ii), since the specifier of agreeing 'eyn is at once a topic position and lhe ubjcct Case position.

Ur Shlonsky



(28) Rina zameret rok.

Rina singer rock

'Rina is a rock singer.' . " b l

. h ver where the string subject nonver a There are a variety of circumstances,. olw~ p~rticularly clear example is that of an

. ptable or very margma .

predicate is unacce. h' h the second term is a proper name.

equative sentence, m w ic (29) a. *Rina gveret Levi.

Rina Mrs. Levi 'Rina is Mrs. Levi' b. *xaver-i ha-tov Dan~.

friend: ls the-good Dam

'My best friend is Dani.' .

. . (29) grammatical is to mtroduce an

d · th equative sentences m . f th

One way of ren enng e . between the two constituents o e

. al . f to a third person pronoun h' 1 ent

element ident1c m orm . ( 1987) 1 shall henceforth refer to t is e em copular construction. Following Rapoport '

as H. H appears in (30).

(30) a. Rina hi gveret Levi.

Rina H:fs Mrs. Levi

'Rina is Mrs. Levi' . b. xaver-i ha-tov hu DDan~. friend: 1 s the-good H:ms am

'M best friend is Dani.~ . .

y d G (1976) established that H is neither

1986) f llowing Berman an rosu al h d 9 Th Doron (1983), ( • 0 h the lexicalization of a function ea · e

the clausal subject nor a verb, but rat er d. d Hebrew copular constructions is that the

archers who have stu ie Th " ris

consensus among rese . . t d with different structures. e iorme sentence in (28) and those m (30) ar_e assocta e 10

a small clause while the latter constttute full IPs.

also serve as a lexicalization of a functional head although not

9 The impersonal pronoun ze can

the same head as H. See Sichel (1_997). 95 observe that the distribution of copular constructio~s

iu Rappoport (1987) and Rothstein (_19 ) f I l complements to consider-type verbs.His with H and those without it patterns hke that o . c ~us~ssible. Thus, compare (28)-(30) with the obligatory where a small clause complement is imp


(i) I consider [Rina a roe~ singer]-, (ii) *I consider [my best friend Dam]. .

'd [ best friend to be Dam]. d

(iii) 1 cons1 er my . ('") i's obligatory when the secon . . (') b f II IP structure, as in lll • 83) A reduced clause is possible in I ut a u l' zat1'on due originally to Doron ( 19 '

· · name The genera 1 •

term of the copular constructton is a . h the second term of a copular construc- seems to be that a full IP struct_ure is necessary; et~ev:;oro (1997) (see also den Dikken 1997) tion cannot be construed and hcensed as a pre ica .

C tructions Subject Positions and Copular ons

Although this characterization is too crude, (the possibility f in erting adverbiRI material between 'Rina' and 'rock singer' in (28) suggests Lhat the ubject appears ut. id Lhe projection of its predicate ) it is correct in its essence: The subje l of (28) i I wer than the subject f (30). Let us see if we can be more precise ab ut the e positions.

The pair of ent.ences in (31) show that a predicative sentence which does not require H ca11 be embedded under 'eyn while an equative sentence in which H is obligatory cannot be embedded under 'eyn.

(31) a. 'eyn Rina zameret rok neg Rina singer rock 'Rina is not a rock singer.' b. *'eyn Rina (hi) gveret Levi

neg Rina (H:fs) Mrs. Levi 'Rina is not Mrs. Levi.'

Indeed, the ungrammaticality of (31 b) provide addiLional evidence for Lhe claim that equative entences require m re structure than predicative sentence . Contra ted with the grammatical (32) below, in which there is a subjecl position LO U1e left of 'eyn, (31 b) demonstrate that U1e first DP of an equative c nstruction must access a ubject positi n higher than negation.

(32) Rina 'eyn-(n)a gveret Levi.

Rina neg:3fs Mrs. Levi 'Rina is not Mrs. Levi.'

A number of different proposals have been advanced to deal with the distribution of H.11 A critical assessment of these propo al being beyond the scope of this article, I proceed directly with a presentation of my own views on the matter.

I would like to advance the hypothesis that since equative c nslructions involve two referential expressions both must be licensed in pecifier positi ns. Copular construc- tions, I hall argue. do not make an object position available and the two DPs must therefore make use of 1h lwo positions I have earlier identified as subject positions, the nominative (topic position) and the nonnominative one.

Since the subject position in (3lb) (under 'eyn) is filled by one of the two referential

argues that (iii) is an inverted copular construction, with the predicate raised above the subject and adjoined to IP, whence the need for a full IP structure. However, the fact that the same structural requirement holds of copular constructions involving two referential expressions (as in the (a) examples in (29) and (30)) suggests that his analysis is not general enough, as noted by Krach and Hey cock ( 1996).

With respect to the Hebrew data, Greenberg (1994) shows that Daron's generalization fails to cover predicational generic sentences which also require H. Her own generalization, stated in semantic terms, is in tum vitiated by a numerous exceptions, as she herself notes.

11 See Berman and Grosu (1976), Dechaine (1993), Daron (1983, 1986), Greenberg (1996), Penner (1988), Rapoport (1987), Rothstein (1995) and Sichel (1997).

Ur Shlonsky 335


. . . sub"ect osition is available above non-agreeing expressions and smce no


the s!ond DP. In (32), however, there are two 'eyn, there simply is


pos1t1on to


and one above it. Hence, both arguments of subject positions available, one

b~~o: e~~er

positions. . . the equative sentence can appear I p ak have a syntactic charactenzat10n of

. e I take it that spe ers . "th Let us be more prec1s . ear in an specifier-head


WI an

referential DPs and that


DPs must


t tically licensed. In predicative or better appropriate functional head m o_rder _tot e


ac in Spec/ AgrS where nominative Case still, verbal sentences,. the.

subject~: l~~el

olrepresentation) in Spec/AgrO or in some licenses it while the object is_ (at so · ive Case is available. . . . such position, where accusative or object . Case is unavailable. This is v1s1ble m

Equatives contain two DPs but accusative Russian Latin and Classical Arabic

· · ch diverse languages as ' · f th

copular constructions m su . The modern Romance languages





which display Case


ive Case in equative constructions. Dtrect objects argument for the unavatlab1hty


?bJect h b as in (33). These clitics bear the in French or Italian can be cht1c1zed onto t e ver '

accusative form.

(33) a. Claire verra Madame Lev!.

Claire will see Mrs. Levi 'Claire will see Mrs. Levi.' b. Claire la verra.

Claire her will see 'Claire will see her.'

'b , indicating that it is not l DP however cannot be cliticized onto12 e ,

A postcopu ar ' ' ) d (34)

associated with accusative Case, compare (33 an . (34) a. Claire et Gaston sont les Lev!s.

Claire and Gaston are the , 'Claire and Gaston are the Levis.

b. *Claire et Gaston les sont.

Claire and Gaston them are 'Claire and Gaston are them.,

l DP there must be some other,

. 1 d ·gned to the precopu ar , f

If nominative ts area y ass1 . . C ailable Let us call this Case de au t 1

t" d nonnommat1ve ase av . . h h

(structural) nonaccusa ive an . "d ith the Case associated wit t e Case. In many languages, default Case comc1 es w

12 Non-accusative chttcs can certarn y . . · 1 appear in such sentences.

(i) Claire y est.

Claire there is 'Claire is there.'

(ii) Claire n'est pas peintre, Marie !'est.

Claire is not painter, Marie it is 'Claire is not a painter, Marie is.'

336 Subject Pos1t1ons and Copu ar on . . I



citali n form, i.e., accu alive in English n minaliv in Latin and Arabi , etc.

l w uJd like to argue lhal postcopular DP are Ii ensed in U1e n n-nominative p s11.Jon which J have earlier idemified under >e n in the Hebrew middleficJd. This p ·iti n is where structural default Case is a signed. To recall two subject position were distin- guished. n high, n minative posiri n which rnight be labeled Subl and a lower non- nominalivc positfon, Sub2.

In Hebrew, the nominative po ition is the specifier of l11e agreement h ad lcxkaljzcd by H or by lhe agreement suffixes on >eyn. Th l wer head is phonetically null but we can expe I it lo be overt in ome other language. In (28), the subject 'Rina' occupies Sub2, Lhe lower position, which is why (28) can be mbedded under 'eyn as in (31 a).

Just as speaker have yntactic intuitions a· to wba1 a refer ntiaJ expressi n is, they can identify predicates. The crucial p int ab UL syntactic pr dicates is !hat !hey are licensed differently from referential expressions. ln parti ular, they do not c upy th subject po ition(s) reserved for argumenrs.13 Thus, the predicative expre ion 'rock singer in (28) is .nol in Sub2 and con equently 'Rina· i not forced up to Sub I.


I\ disUnetion, l lhink, can be usefully drawn between syntactic predicates and other constituents (including arguments) which are not syntactic predicu1 s but can be interpr ted predicaUvely, by e.g .. undergoing a semantic type-shi fling operation. or example. French (and German) distinguish determiner-less NPs from indefinit• DPs in po tcopular position; c mpare i) and (ii).

(i) Cet homme e t enseignant.

this mnn is teacher.

'This man is a teacher.' (ii) Cet homme est un enseignant.

this man is a teacher 'This man is a teacher.'

Pollock ( 1983) judges (i) as predicative and (ii) as ambiguou between an predicative and an identity reading (tJ1ough see Reboul and Moeschler ( 1994) for a more detailed las ificalion.) Syntactically, only lhe det rminorless NP in (i) i. a predicate, and cannot be an argumen1 (since a bare NP will not rai e Into a specifier position; ee nbove.) (ii) i not a yntactic predi ate although Pollock's judgement suggc ts that ii can b imerpreted as such. Only the bar NP 'teacher' in (i can be clilicized by the predicnt cliti le (on the rather complex di tribution


which, se SporLiche (1995).) ompare (iii) and iv), with the predicate cliti underlined ..

(iii) Ces homm ci sont en. eignants mais


homm s IA n le ont pns. enseignants.

these men here arc teachers but th . e men IJ1ere neg-Prcd I-arc neg, lcaehers.

'These men arc teachers, but those men are not (teach ·rs).'

(iv) *Ces hommes ci sont des en eignants mais ces hommes la ne le sont pas, theso men here are indef.pJ teachers but these men there neg-PredCI-are neg, des enseignants. teachers.

'These men are teachers, but U10 e men arc 1101 (teachers).'

Definite DPs can be syntactic predicates. a Williams ( 1994) argued (see also Fiengo and May 1994, among others). Definite DPs and names (particularly role an also function syntactically as non-predicates interpreted prcclicari vely, under appropriate pragm:uic conditions, as noted already by Higgins (1979).

Ur Shlonsky



4. Pronominal subjects of copular constructions

ron 1983) that pronominal tokens of DPl in equat!ve It has been often noted (e:g., Do f H nlike nonpronominal tokens which constructions do not reqmre the presence o ' u

require H (recall the contrast between (29) and (30).) (35) a. 'ani (hi) gveret Levi.

I (H:fs) Mrs. Levi 'I am Mrs.Levi.' b. hi (hi) gveret Levi.

She (H:fs) Mrs. Levi

'She is Mrs. Levi.' th "th

This section addresses this fact dealing first with the case where H appears and en w1 the case where it does not.

4_ 1 Pronominal subjects with H

. ossible in (35) if the pronominal subject is It should first be noted t?at H is only


u a context in which the pronoun



focalized and bears a

spe~1al str~ss.

If we d rfp 'Mrs Levi' is (contrastively) focal1zed, construed as old information, or if ~h~ secon . (36) j4

H becomes plainly impossible. This is shown m . . 'ata lo yod'ea mi Rina? hi (*hi) gveret Lev~

(36) a. you:ms not know who Rina she <t,I:fs) Mrs.


'You don't know who Rina is? She ts Mrs. Levi.

hi (*hi) gveret LEVI (lo gveret Cohen.) b. she (H:fs) Mrs. Levi (not Mrs. Cohen.)

'She is Mrs. LEVI (not Mrs. Cohen.) .

h the sentences in (35) with H are mverse Adapting Daron's insi_ght to the


t at ructures in the sense of Moro 1997), let us structures (though not mverse


it is the specifier of H) and the pronoun say that 'Mrs. Levi' first. appears m




above Subl to an A' focus position .. H, in Sub2. Then, the _focahzed


ts ra;se s] and raises into Foco in order to satisfy much like I

~n ~

finite clause, ts

:~~~~J~) %~a)

with H is thus derived from (37) by a the

Focusk~ntt~n~~g\~:~ :~tj!:~a~xiliary

inversion (see Heggie (1988).)

process a m

(37) gveret Levi hi 'ani.

Mrs. Levi H:fs I 'Mrs. Levi is me.'

14 Care must be taken not to introduce a pause between the two hu's in (36), to avoid a left dislocated reading of the subject.

338 Subject Positions and Copular Construct10 · ns

Two piece f evidence can be culled to upport thi prop sal for th derivati n of ( 5).

First, note that in (35a), the ubje l is a lfrst per on pronoun while H i specified for third person. Thus they do n t agree in <j>-fealu.rcs. If H, like Lhe agreement suffix on

'eyn discus ed above, mu t agree in person features with its sp cifier, (35a) ought l · be

ungrammatical. Under the proposed derivation of (35), however, H a tually agrees in cp- fcaturcs with 'Mr . L vi' while t·he pron minal subject need only 'agree' in fo us features with H.

The second piece f evidence i du 10 l. Sichel (p.c.) who note that while the nega1ive particle lo invariably foll ws 1-J (Doran 1983}, it carmot foll w it in (35). (38a) below is grarnma1ical, sine 'Mrs. Levi' is in Sub I, H is unmoved and lo appears below it. ln (38b). on the other hand, Subl is filled by 'Mrs. Levi' H ha been shifted to oc above SubJ and !here i n po ition for negati n between Foc0 and Sub I. 0

(38) a. gveret Levi hi Io 'ani.

Mrs. Levi H:fs not I 'Mrs. Levi is not me.' b. *'ani hi Io gveret Levi

I H:fs not Mrs. Levi 'I am not Mrs. Levi.'

We might ask wha1 prevent (35b) from being a signed a tructure imilar to Lhat f, say, (30a), with the pronoun in Sub! and 'Mrs. Levi' in Sub2. Afler all, that w uld be a implcr and more economical dcrivali n. ((35a) is no I nger relevant, given the lack of cp-fea1ure matching with H). This question actuaJJy harb r two distinct queries: First, what prevem a non-focalized pr noun from appearing in Sub I and . ec nd, what preve111s a fo aJized one from appearing there?

The first issue concern the imp ssibility of non-focalized pronouns in Sub I. It is clarihed by U1c observation that Sub I may only host strong nominals, that is, non- pronominal DPs and lrong pronouns.' .

Unlike say, lhe Romance languages, Hebrew d es n L m rphologically distinguish weak and strong pr nouns. Since U1e weak form is the defaul! form Cardinaletti and Starke 1996), strong pronouns how up only when they are required. Thu , in lhe ab nee of any 'slrengtheniDg' mechanism, it is the weak pr noun which i generated and it cannot appear in Sub I. Weak pronouns are rendered trong by oordinati n and by focalization, discus ed b low). Let us therefore c nsider co rdinated pr nominal subjc t . The sentences in (39) show tbat a oordination of pronouns behaves just like a full DP. H is obligatory in (39a), as it i in e.g., 'c30). In a context where the coordinated pronominal canno1 be taken as I' cus a in (39b), it mu l agree with H. The ungrammat- icalily of this example is due lO lhe lack of pers n agreement: The pronominal subject is second person whjle H is third pers n.


This is precisely the conclusion reached in Cardinaletti (1997).

Ur Shlonsky



(39) a. hi ve hu *(hem) ha-Levi?1 she and he (H:mpl) the-~e;1s

'She and he are the Levis. . )



(*hem) ha-LEVIM. (lo ha-Cohemm b. I and-you (H:mpl) the-Levis, not the-Cohens

'Me and you are the LEVIS, (not the Cohens.) . . . . . dependently motivated m pred1cauonal The restriction of Subl to strong


as (40b)) in that His unf!am-

. (40a) contrasts w1 · t tion

copular construcuons. . . . al . the latter an equauve cons rue . matical in the former whereas it is option m ,

(40) a. hu (*hu) zamar rok. he H:ms singer rock

'He is a rock singer.' b. hi (hi) gveret Levi.

She (H:fs) Mrs. Levi 'She is Mrs. Levi.'

bl that subject position being res~rved fo~ strong Weak pronouns are excluded fr_om Su 40 ) t be derived via inversion, that is, from elements. We must now ascertain that ( _a ca~no cup1"es Sub l (such a derivation would

· h" h 'rock smger oc · ·

an underlying structure m w 1c . (3S)) (4 l) the putative pre-mvers1on parallel the one proposed for the m trasting with the fully acceptable (37).

structure for (40a) is plainly ungrammauca' con (41) *zamar rok (hu) hu.

singer rock (H:ms) he . . h" •

'He is a rock singer.' Lit: 'A rock smger is im. .

. rt

of (41) means that 1here is no way of putting Doron points out that the ungra?1~at1c~ i y t l since Sub I is an A-position and 'rock the predicative DP in Subl. This is quite .nha urath, weak pron un nor the predicate can

. d" t 11 To conclude, ne1t er e '

singer' 1s a pre ica e. . atical derivation for (40).

appear in Sub 1 and there is h_ence


pronouns may not appear in Sub 1. The Let us now tackle the question o w

. . h. h the pronoun hu is not H but

\6 (40a) is fine under the irrelevant left-dislocation reading, m w ic

a nominative (resumptive) pronoun: . ite DPs cannot appear in Sub l - even when the_y

17 Matters are more complicated smce i~defin ' rential DP This is independent of whether t~1s

are referential - when the second




(ii) are


(both are possible with second term is a pronoun or not.Thus, . o i

heavy, contrastive stress on the mdefimte.) (i) *yeled hu Dan~.

child H:ms Dam

'Dani is a child.' . (ii) *yladim hem Rina ve Dan~

children H:mpl Rina and Dam 'Rina and Dani are children.


I C structions Subject Positions and Copu ar on

discussion in Lhe previous paragraphs actually suggests lhc c nlrary. Focalization being one of Lhe ways of rendering a weak pron< un strong we should expect focalizcd pron uns lo appear in Lhe subject p ition re erved for strong n minals. Yet we have cen that

~ calized pronoun to Lhe lefl f H do not have to agree with H and H cannol be immediately followed by negation, both facts. upp 1ting th view thnt when H is pr ceded by a C calized pronoun, the pronoun is not in Sub 1 but in a (higher) ~ cus po ·ition.

Why cru1 Sub I n th st


ci? The reason i lhat like the subject posili n lo the left of 'eyn, ub I i a topic position. I argued that the exclusion of n n-ref erenlial subject· from the ubject p siti n to the left f 'ey11 foJJ ws if that subje t position is a topic positi n.

I would now like to extend t11i characterizalion to the subject p ilion to the left



Dor n ( 1983, pp. 97- 8) notes t11a1 H is obligatory when the subject of a copular construction is relativized and ( ptional) when it i que ti ned by mean of


which NP exprcssion.18

(42) a. ha-'g se- *(hu) more ... the-man that- H:3ms teacher:m

'The man who is a teacher ... ' b. 'eize y~dida selxa (hi) mora?

which friend yours (H:3fs) teacher:f 'Which friend of yours is a teacher.'

The common feature of relative clauses and D-linked questions is that b 111 are types of topicalization. Suppose, now, that movement of a D-linked ivh-expres i. n and of a relative operator (or head, depending on ne' analysis of relative clauses) must implicate the topic projection. The obligatorines of H in (42) should th n be related to the triggering of the topic projection. Con retely, assume that subject relatives and subject wh questions involve movement through the specifier of H, a topic position.

There is thus reason to believe that Sub l, like the subject position we find in agreeing

'eyn sentences is a topic position. Being an A topic position, it cannot host foci,

pronominal or not.

4.2 Pronominal subjects without H

I have discussed equative constructions with pronominal subjects, yet the original motivation for that discussion was the fact that H can be missing in examples such as (35), whereas it i mandatory in equatives ·With nonpronominal subjects. Consider the nature of the problem: The second DP of (35) can be no lower than Sub2. The pronoun, being weak, cannot appear in Sub I and yet the sentence is perfectly grammatical.

Let us suppo e that in addition to Subl and Sub2, UG makes available a subject position which is reserved for weak pronouns (Cardinaletti's (1997) lower subject position).

We have seen that equative constructions cannot be embedded under 1eyn. The relevant

18 Doron shows that the distribution of H is different in long interrogation, but this matter is tangential to the present discussion.

Ur Shlonsky 341


. d b l as (43) I argued that since Sub2 is filled by 'Mrs.

example, (31 b ), is repeate e ow ' . L ev1, ., 'Rina' must go to Subl, above eyn.

(43) *'eyn Rina gveret Lev~

neg Rina Mrs. Levi. , 'Rina is not Mrs. LeVI.

(44) with a pronominal subject strikingly contrasts with (43).

(44) 'eyn hi gveret Lev~

neg she Mrs. Levi . '

'She is not Mrs. Levi. .

. . rd ex lanation under the hypothesis that there is an This contrast receives a stra1ghtforwa b p S b2 which is reserved for pronouns.

additional subject



:;~· ~~~e fir~t o~

second person singular


The reduced acceptab1hty of a ' d l l ones suggests that the pronominal subjects and (4Sb) with first




while first and second person position is actually reserv~d for t .1~ pe~sone 'eyn (45~ b) contrast with (46a,b).

pronouns are associated with a pos1t10n a ov . ' (4S) a. ??>eyn 'ani/'at gveret Lev~.

neg l/you:fs Mrs. Levi . , 'I/You am/are not Mrs. Levi.

b. ??ieyn 'anaxnu/'atem ha-levi~

neg we/you:mpl the-L~v1.s.

'We/You are not the Levis.

(46) a. 'ani/'at 'eyn(::i)-ni/-ex gveret Lev~

l/you:fs neg-1 s/2fs Mrs. Levi 'I/You am/are not Mrs. Levi.' . b. 'anaxnu/'atem 'eyn-::inu/xem hha-leLv1~

we/you:mpl neg-lpl/2ms t e- ev1s.

'We/You are not the Levis.'

. . . d in man Italian dialects studied in Poletto This kind of pronominal spht is_




in the dialect of Polesano ( 1997) and Manzini and Sav01a

~in p~~~s



negation, while third person clitics studied by Poletto, first person subject c it1cs prec

follow it.

(47) a. A no vegno scl: ls neg come.

'I am not coming.' b. No la vien

neg scl:3fs comes.

'She is not coming.' .

s cited above is to split the subject clitic field into The approach taken by ~he author . h t" lar set of morphosyntactic features. The discrete projections associated each wit a par icu

Hebrew data strongly support this research strategy.


· ns Subject Positions and Copular Construct10

5. Further Issues

Copular and 'eyn-negated sentences in Hebrew provide evidence for a clausal cartogra- phy in which several subject positions are distinguished. The following maximal structure emerges from our discussion. Subjects occupy specifier positions associated with heads that I label Agr, neg heads a NegP.19

(48) Su bl Agr SubWeakl Agr Neg SubWeak2 Agr Sub2 Agr














DP H D~011ull1l11nl-l/2 0 'eyn DP +jl!Onominal-3 0 DP 0 Two as umptions have tacilly underpinned the discu sion so far. First, that nonagreeing and agreeing 'ey11. occupy the same base position, namely, Neg0 and second, that the subje t posi1ion above agreeing 'eyn and Spec/H are the same po ition. By way of a conclu ion, I would like to call into question b th of these as umption .

The careful reader will have surely noticed the following problem. If the weak pronoun position housing third per on pronouns is below 'ey11, as (44) uggests, then Lhe possibility of a pronominal subject to the left of agreeing 'yn, as in (48) is iocorreclly excluded, since the pron un,


have argued, cannot ccupy Sub I.

(49) hi 'eyn-(n)a gveret Levi.

she neg:3fs Mrs. Levi 'She isn't Mrs. Levi.'

The grammaticality of (48), in tandem with the fact that unlike H, the agreement suffixes

on 'eyn can expre all c mbinations of number, gender and person, suggests the

folJowing conclu ion: Agreeing 'eyn's initial position is lower than that of nonagreeing 'eyn. Consider therefore the revised hierarchy of positions in (SO) in which two Neg p silions are distinguished.20

19 The features appearing in the Agr n des are merely de criptive. What these features actually mean remains to be determined.

20 The availability of several positions for negation in a single grammar is independently motivated in much recent work, see Cinqu (1998), Manzini and Savoia ( 1998) and Zanuttini ( 1997).

Ur Shlonsky 343


(50) Su bl Agr SubWeakl Agr Negl jsubWeaklj Agr Neg2 Sub2 Agr














DP H DP +pron-1/2 0 'eyn DP-l;jll'OD·3 0 'eyn+Agr DP 0

Agreeing 'eyn thus has the option of raising to any one of the Agr heads to its left, allowing the full range of referential subjects.21. There is therefore no single subject position to the left of agreeing 'eyn, but rather three positions, only the highest of which is the topic position identified earlier.

The occurrence of adverbs between the subject and agreeing 'eyn, as in ( 49), suggests - and under Cinque's (1988) terms - forces the conclusion that agreeing 'eyn does not have to occur in the head position of which the subject to its left is the specifier.22 ( 49) Rina betax/ kanir'e 'eyn-a mgdaberet rusit/ gveret Levi

Rina certainly/apparently neg:3fs speak:fs Russian Mrs. Levi 'Rina certainly/apparently doesn't speak Russian/isn't Mrs Levi.'

Rather than thinking of agreeing 'eyn as raising to an Agr position in the overt syntax, let us say that it (or its relevant Agr features) raise to one of the Agr heads in LF. Such raising creates a representational chain connecting the clausal subject with its lower positions below Neg.

Unlike agreeing 'eyn, H cannot be separated from its subject by adverbial material since H is the head of the highest projection in the subject field. Contrast (53a) with ( 49) above.23

21 The impossibility of the quasi argumental ze to the left of an agreeing 'eyn (viz. (19)) should be correlated with the fact that unlike personal pronouns, quasi argumental ze lacks number and gender features and is hence presumably barred from the pronominal subject position and must be taken to occupy Sub2. Demonstrative ze, which has not been discussed in this paper, inflects for gender and number, and being strong and not weak, can occur to the left of 'eyn.

zot 'eyn-(n)a cipor.

this:f neg:3fs bird 'This isn't a bird.'

22 Such adverbs can also intervene between the subject and a finite verb, a fact with important consequences for Hebrew clause structure.

23 It is not the case that H must be adjacent to its subject (pace Doron 1983, Rapoport 1987),

since parenthetical material may intervene between the two, as in the following example.

Rina, toda la'el, hi zameret rock.

Rina, thank God, H:fs singer rock 'Rina, thank God, is a rock singer.'

344 Subject Positions and Copular Constructions

- - -~---=- - - - - - - ---=---

• . - • . -

(50) a. *Rina betax/ kanir'e hi

R · . zameret rok

' i?a c.ertamly/apparenlly H:fs singer rock

~ma i.s certainly/apparenUy a rock singer.' b. R' Rma h1 betax/ kanir'c zameret rok

' i?a ~:fs certainly/apparently singer ro


. Rma is certainly/<1pparcntly a ro k singer.' This analysis can naturally accommodate the f

grammaticality of (51) establishes. act that H and 'eyn can co-occur, as the (5 l) R~na hi 'eyn-(gn)a gveret Levi.

~1?a ~:fs neg:3fs Mrs. Levi Rma 1s not Mrs. Levi.'

! he final dffference between H sentences and ,

~s t~at.

1-l .i . only possible in sentences with




1 wish t . dwell upon

•estnct1on is unpo'ed Oil 'eyn sentences I (lb) verbal p1ed1cates, whiJe no such verb appears under 'eyn while its




below a (52a), a present ten e

(52) a R' , n er m (52b) yields ungn:unmaticality.

· ~Jila eyn-a mgdaberet rusit.

~1?a neg:3fs speak:fs Russian Rma does not speak Russian ' b. *Rina hi mgdaberet rusit ·

Rina H:fs speak:fs Russ.ian 'Rina speaks Russian.'

Daron, (1983) argued that since H is in J, j[ reclud .

Daron s explanation cannot be incorpor t d

p es

ve1 b mo.vement to I. As stated ample evidence that present tense ve1·b


eH ib.nto the pre ent discussion since there


· · m e rnw do not I ·

project10n (see Section 1 and Sblonsk 1997) rnve to raise to the highest I between H and the verb Indirectly

h /

and o no competition should arise argument can and should .be mainlai~ed wever, and on a more abstract level, Daron's

A subject


Spec/1-J mu I head a (rep;esentation . . .

VP. In order for such a chain co be D d H al) cham rooted m its base position in only way to ensure this coindexing




be coindexed with the verb and the If Chomsky ( 1995) is right selection . y thse. ect1on of V by H or movement of V to H

s b , ' · m is case reduces t D t . ·

u case ol movement. The ungrammaticalit f ( o ea ure attraction, i.e., a

(53) H d y o 52b) can thus be taken care of by (53)

oes not attract (+VJ. ·

Although (53) does not capture the entire a . .

aspectual) type of predicates which can foll g ~ut of restrictions on the (semantic and robust categorial restriction illustrated by




(1994), it expresses the fonnaI anchor for what traditional Hebr g ammaticahty of (52b) and provides a

ew grammar labeled the 'nominal sentence'.

Ur Shlonsky




Sujets connexes :