Abstract

Pseudocleft sentences are copular sentences with a free relative subject, as in (1):

(1) what John saw was a cat.

Higgins (1973) distinguished predicational and specificational pseudoclefts, and showed that, unlike predicational pseudoclefts, specificational pseudoclefts exhibit a variety of syntactic and semantic connectivity effects, i.e. the post-copular phrase behaves in some ways as if it ‘sits’ inside the free relative subject.

In this thesis I give a syntactic-semantic analysis of specificational pseudoclefts in Hebrew. I show that Hebrew makes a lexical distinction between predicational and specificational pseudoclefts in the choice of the copula, and that Hebrew specificational pseudoclefts have an identifiable subset that do not exhibit connectivity. I use these data to argue (a) that specificational pseudoclefts are equative sentences (also known as identity statements) and not, as suggested in Williams (1983), an instance of inverse predication; (b) that connectivity is not an effect of syntactic reconstruction or copying as assumed in a lot of syntactic literature, but rather a semantic effect as suggested in Jacobson (1994) and Sharvit (1997, to appear); and (c) that connectivity is a by-product of equation at high semantic types.

As is well known since the work of Berman & Grosu (1976) and Doron (1983), Hebrew uses personal and impersonal pronominal copulas instead of a present-tense verbal copula that does not exist in Hebrew. The impersonal pronominal copula (pronZ) was not given much attention in the generative literature, as opposed to the personal pronominal copula (pronH). I show, using tests developed for pronH, that pronZ is indeed a copula, rejecting its analysis as a subject pronoun. In pseudoclefts, both pronominal copulas are possible, as in (2):

(2)

a. ma Se-ani macbi'a alav hu xatul
what that-I point on-it H cat

b. ma Se-ani macbi'a alav ze xatul
what that-I point on-it Z cat
“What I’m pointing at is a cat.” (adapted from Higgins, p.212)

I show that a pronH pseudocleft is predicational and a pronZ pseudocleft is specificational, following Higgins’ (1973) battery tests for distinguishing between predicational and specificational pseudoclefts. A closer look at pronZ pseudoclefts reveals that there are two kinds of pronZ - agreeing and non-agreeing (neutral) - and they differ with respect to connectivity effect. Although all these pseudoclefts behave like specificational pseudoclefts on Higgins’ tests, neutral pronZ pseudoclefts exhibit all the connectivity effects (opacity, binding, agreement of predicates and case marking) but agreeing pronZ pseudoclefts do not.

My analysis of specificational pseudoclefts involves two assumptions. First, I use the overt distinction between predicational and specificational pseudoclefts to argue that specificational pseudoclefts are best analyzed not as inverse predication, as suggested in Williams (1983) for English and in Moro (1997) for Italian, but rather as equatives (identity statements), as suggested in Heycock & Kroch (1996, 1997). This means that while pronH is a ‘BE of predication’, pronZ is a ‘BE of identity’ (in the sense of Partee 1987).

Second, I discuss connectivity effects in specificational pseudoclefts. The fact that some specificational pseudoclefts do not exhibit connectivity effects serves as strong evidence against a syntactic analysis of connectivity by means of reconstruction, since the reconstruction account wrongly predicts that both types of specificational pseudoclefts would exhibit connectivity. I assume that the difference between agreeing and neutral pronZ is in the type at which the equation takes place: agreeing pronZ equates individuals (type e) only, while neutral pronZ equates non-individuals only. I show how this predicts that agreeing pronZ pseudoclefts would not exhibit connectivity, and I use the analysis of Sharvit (1997) to account for the connectivity effects in neutral pronZ specificational pseudoclefts, pointing out that the high type equation is what allows for connectivity.