Compositionality , copy theory , and control

Our ultimate concern is the mapping between representations in the Language of Thought (in the sense of Fodor 1975) and representations at LF in some version of the Minimalist Program (Chomsky 1995). We argue on a variety of grounds that such a mapping is direct and that both these levels of representation must meet the (same) condition of compositionality. We argue that Copy theory, in particular in the version exploited for 'control' phenomena by Hornstein (1999), is inimical to any such condition and must therefore be rejected. The evidence comes from the properties of quantified NPs in English and the Caucasian language Tsez. Specifically we suggest a reanalysis of 'backward control' in Tsez, as described by Polinsky & Potsdam (2001, 2002), thereby removing a major argument for copy theory. Our analysis does not however preclude an assimilation of canonical control to raising, and we provide an alternative to a Hornstein style analysis, giving some justification for the exploitation of the combinator R in Combinatorial Categorial Grammar.