ABSTRACT
A central assumption in generative grammar research on the relationship between syntax and the lexicon is that syntax is a projection of the lexicon. The structure of sentences is a refl ection of the lexical properties of the individual lexical items they contain. In the standard view, each lexical item is associated with a lexical entry that contains three kinds of information, as indicated in (1):
(1) The contents of a lexical entry: a. Morpho-phonological information b. Semantic information c. Morpho-syntactic information: –syntactic category –subcategorization –theta grid (argument array, theta-roles, linkings)
In classical theories of lexical structure, the morpho-syntactic component of a lexical entry consisted of a specifi cation of c-selection (or subcategorization) and a theta grid, specifying a given lexical item’s argument array, the thematic roles of its arguments, and the way these theta-roles are linked to the arguments. The lexical entries of fang ‘put’ and zhidao ‘know’ looked like (2) and (3):
(2) fang ‘put’: a. [+V, -N] b. [+ NP PP] c. (x, y, z) d. (Agent, Theme, Location)
(3) zhidao ‘know’: a. [+V, -N] b. [+ NP/CP] c. (x, y) d. (Perceiver, Theme)
In recent years, it has been shown that a lexical entry of this kind suffers from a problem of overspecifi cation, as there is considerable redundancy among the kinds of information provided. It has been observed by Pesetsky (1982), for example, that information about subcategorization (c-selection) is largely predictable from its meaning, which includes s-selection, or the specifi cation of participants of a predicate, in terms of their ontological types. If a verb s-selects a thing, then it c-selects an NP, as in (2). If a verb s-selects a proposition, then it c-selects an NP or CP. Information about c-selection can be predicted by a principle of Canonical Structural Relation (CSR) from information about s-selection.
