On Reducing Control to Movement

Hornstein (1999, 2001) proposes a version of the movement theory of control (MTC) in which movement of a DP is driven by the need to “receive” a θ-feature of the verb/predicate it merges with. I argue against the use of θ-features in syntax on the grounds that they are fundamentally semantic, hence properly belong to the bare output conditions of the C-I interface. I propose instead a version of the MTC closer in spirit to Bowers (1973/1986, 1981) in which Merge is driven by interpretable or uninterpretable c-selection features of the standard sort. I then show that, given standard minimalist assumptions, the MTC is not only possible but necessary. I conclude by arguing that the MTC thus formulated supports a purely derivational theory of syntax and interpretation that contains no level of LF.