寺田 寛

Successive cyclic wh-movement is an operation contingent on the information available at a later stage of the derivation. For this reason, the conventional bottom-up view of structure-building would inevitably induce look-ahead properties or backtracking of computation. This paper explores successive cyclic wh-movement within Phillips' (1996, 1998) top-down approach to structure-building, assuming contra Chomsky (1995, 1998b) that lexical arrays are eliminated, that computation multiply accesses the lexicon, and that Merge (combined with Agree) does not preempt Move. I speculate on technical implementation of the top-down approach and discuss its extensions to observed properties of A/A'-movement.

The purpose of this article is to elucidate the nature of scope-marking questions in Hindi-Urdu. Chomsky (1993) proposes within a minimalist program that a wh-operator cross-linguistically moves in overt syntax. Based on this proposal, we claim that the overt movement of a null operator for a wh-phrase in Hindi is both partial and multiple: it raises more than one null operator only to the Spec of the closest tensed CP that could be different from its actual scope position. There are some pieces of evidence in favor of the existence of such movement. Thus, a null operator moved constitutes an island for scrambling (Wh-Island effects) and rightward scrambling blocks null operator movement (Freezing effects). Although wh-in-situ in a tensed CP cannot be assigned scope over higher clauses, an operation of inserting a scope marker KYA in such clauses allows the wh-in-situ to take wider scope. The partial movement of a null operator is driven by the morphological necessity that is associated with KYA-insertion: a verb in a clause where KYA is inserted selects a [ + wh] CP, whose head contains a strong operator-feature that must be checked off via the overt partial movement of a null operator. The partially moved null operator is unselectively bound by the scope-marker at LF and then can extend its scope over higher clauses. We thus characterize Hindi-Urdu as a partial wh-movement language on a par with German, Romani and Iraqi Arabic.

The aim of this article is to shed some light on the tense effect on LF movement of wh-in-situ in Hindi-Urdu. Although these languages lack syntactic wh-movement, wh-interrogatives are not allowed to appear in tensed islands, while they are allowed in untensed ones. This is contrary to what is generally known about wh-in-situ in English, which does not obey Subjacency. This tense effect is provided with a desirable account on the basis of Generalized Binding theory and by assuming a parameter of Hindi-Urdu that tensed C qualifies as a SUBJECT. A wh-trace (an A'-anaphor) must be A'-bound by its antecedent in its binding domain, which is tensed CP, not tenseless CP. When the specifier position of tensed CP is occupied by some element such as a relative pronoun and the antecedent of a trace is not allowed to occur in the clause, the trace is not A'-bound in its domain, violating Condition (A) of generalized binding. Hence interrogatives are excluded from tensed relative and adjunct clauses, resulting in the tense effect. Our analysis also accounts for the fact that wh-in-situ is insensitive to other island conditions and lacks subject-object asymmetries. We also show that the Empty Category Principle and Subjacency at LF do not provide a satisfactory account for the tense effect.