This paper claims that the manifest possibility of acquisition and evolution of language requires that the interfaces between modules of grammar be transparent. That is, logical form (or the corresponding model theory), and phonological form (or the corresponding articulatory representation) must all be simply related. In particular, syntax must be both phonology-free and semantics-free, with no potential to alter structure nonmonotonically at either "interface level." The paper reports some recent developments in Combinatory Categorial Grammar (CCG), drawing on work by Jason Baldridge, whose PhD thesis showed how categorial functional types can be elegantly restricted via a type-hierarchy whose values correspond to the sets of applicable combinatory rules. In this form, CCG is a theory of grammar in which the sole language-specific interface component is the lexicon, which pairs (ordered) phonological forms with (unordered) logical forms, and associates each such pair with a syntactic type. A universal set of purely type-driven syntactic combinatory rules then combines the lexical elements onto sentences by concatenating phonological forms and "projecting" logical relations (to both of which the combinatory apparatus is entirely blind) into pairs of sentential phonological and logical forms. (The derivations themselves are not distinguished in the theory as a level of representation.) CCG therefore provides a formal basis for eliminating rules of movement in favor of compositional merger. This paper shows how this version of the theory strengthens earlier results in CCG concerning coordinate and prosodic structure and their interaction with quantifier scope, and addresses some recent criticisms hinging on the reality of the Across the Board Condition and the Strict Competence Hypothesis concerning incremental interpretation by the parser.
Proceedings of the 24th West Coast Conference on Formal Linguistics
edited by John Alderete, Chung-hye Han, and Alexei Kochetov
Table of contents