LINGUISTIC REVIEW, cilt.30, sa.1, ss.25-49, 2013 (SSCI)
In this paper, I propose that Croatian relative clauses (RCs) introduced by the complementizer to 'what/that' do not form a homogeneous class with respect to their derivation: some are derived by movement, and some are derived by a non-movement strategy. Unless the relativized element is the subject, sto-RCs normally require a resumptive pronoun to appear in the site of relativization. However, this requirement is removed under morphological case matching between the head of the RC and the resumptive pronoun: the resumptive pronoun, may be omitted if the pronounced head of the RC, case-marked in the matrix clause, appears in the morpho-phonological form that it would have if it were case-marked by the embedded predicate. I argue that the absence of a resumptive pronoun in a sto-RC indicates that the RC is derived by movement. I propose that the matching analysis of RCs (Bhatt 2002; Citko 2001; Hulsey and Sauerland 2006; Sauerland 1998, 2002), coupled with a version of the Inverse Case Filter (Boskovic 1997, 2002; Martin 1999) and a particular view of inverse (case) attraction offers a natural explanation for why such sto-RCs may receive a movement analysis only if the matching requirement is satisfied.