Embedded Topics
According to Lasnik and Saito (1992), local topicalization of subjects is not possible. This asymmetry likely falls under the VMH, assuming that topics move to Spec,CP. In light of the two-chain approach to overt movement, consider the following contrast, also noted by Lasnik and Saito (1992):
(1)
The contrast is very clear for those who accept embedded topicaliza tion. In (29a), the reflexive, topicalized from the object position in the embedded clause, takes the matrix subject as its antecedent (cf. the non topicalized counterpart *John thinks that Mary likes himself ). However, if the reflexive is a topicalized subject as in (1b), it cannot take the matrix subject as its antecedent.
I assume that a CP-recursion structure is present in the embedded clauses in (1), as suggested for embedded topicalization by Authier (1992). (1a) shows that pied-piping the reflexive to Spec,CP of the embedded clause allows the reflexive to ‘‘find’’ its antecedent in the matrix clause. If local subject topicalization is blocked by the VMH, then (1b) is on a par with (2), where the reflexive is in the subject position of the embedded clause.
(2) *John thinks that himself likes Mary
Under the classical formulation of the VMH in (3), it is not clear why the subject should remain in situ. Under the two-chain analysis, economy forces the subject category in (1b) to stay in the subject position, even though the feature to be checked has moved to C in overt syntax. Features relevant for binding are pied-piped in (1a) and are therefore high enough to find a local binder in the matrix clause. If the features that mediate binding are not part of the feature(s) checked, then the contrast in (1) again suggests that the location of the category, and not of the formal features, affects binding (contra Chomsky 1995, 272). Note, finally, that the semantic effects derived by the (non)application of Pied-Pipe con firm that Pied-Pipe applies in narrow syntax and not in the post-Spell-Out derivation to PF.
