Empty Categories, Chain Binding, and Parsing

Abstract

In Government-Binding (GB) theory, the chief empirical effect of the principle Move-α is the definition of coindexing relations between nodes in phrase structure trees. This situation is explained by the extreme generality of the base component and several substantive constraints on transformations. The constraints on transformations, in particular Emonds’ (1976) Structure Preserving Hypothesis, imply that the base may generate all S-structures of the core constructions of the language in question. The two conditions combine with a shift away from a grammatical model with both conditions on derivations, such as transformational rule ordering, and conditions on representations, such as the Subjacency condition on adjacent chain elements, to a grammatical model that favors representations.1 It follows that the principal function of Move-α in the grammar is not reordering or otherwise rewriting of an input phrase marker for the derivation of S-structure, but rather the above noted establishment of coindexing relations between the nodes in the output representation. The relations in question are trace-antecedent relations. This trend in the role of Move-α in the grammar has been noted in the literature, especially by Koster (1978), Chomsky (1982), and Barss (1983).