C, T, and the subject: That-t phenomena revisited

The purpose of this paper is to provide a novel account to ‘that-t’ phenomena, following minimalist assumptions. Taking subject extraction out of complement clauses (the ‘that-t’ effect) as the key case, it is argued that the licensing and identification parts of the ECP correspond to PF and LF requirements respectively. In particular, subject extraction over that yields ungrammaticality because in this case the Agr-features of T fail to be lexicalized; thus the CP phase contains features that have not been satisfied and the derivation crashes. C0 is taken to correspond to the realization of the feature content of T (plus Agr) in C, thus allowing for the lexicalization of Agr. The interpretation of Agr as a variable is determined in connection with the wh-phrase in the matrix Spec,CP via the operation Agree. To this end, I consider the relation between C and T, the one between C and Agr in non-extraction contexts as well (e.g. ‘pro-drop’ under V2), and also the parameterization of subject extraction as a by-product of the different ways of lexicalizing Agr and/or C. The proposed analysis succeeds in deriving the empirical effects of the ECP derivationally, without furthermore postulating extra technical devices such as proper head government and traces.