A Left Branch Extraction Perspective on Bound Variables and Pronoun Insertion Strategy

Pronouns like his in (1) are well-known to be ambiguous between the bound and the referential readings. Not all languages pattern with English. As seen in its Serbo-Croat (SC) counterpart (2a), the 3PsSgMascPoss njegov (his) cannot get the bound-pronoun reading. This reading requires the possessive ‘self ’ svoj (2b). Trivially, whereas SC has two possessive pronouns, English has only one. This, however, does not seem to explain the pattern in (1)- (2). Romanian has the same inventory of formatives as SC (3PsSgMasc. lui in (3a) and the counterpart of the possessive ‘self ’ isi in (3b)), but it patterns with SC only partially. Though Romanian isi gets the bound variable interpretation (3b)- which puts it on a par with SC (2b)- the 3PsSgMasc pronoun in (3a) is interpretatively ambiguous between the bound and referential readings, just like its English counterparts and quite unlike SC. The richer inventory of formatives, then, cannot explain why bound reading is illicit for SC njegov. The conclusion is that there is more to be said about the pattern in (1)- (2). Taking Hornstein’s (2001) account of (1a) as a starting point, I argue that the SC data in (2) can be explained and that the pattern can be accounted for. Under Hornstein’s account, English must opt for the less economical derivation with bound pronoun because the movement derivation (4) i