Computational efficiency in the syntax–phonology interface

Abstract

This paper discusses conceptual aspects of a Multiple Spell-Out theory of the syntax-phonology interface where a domain of spell-out is a prosodic domain. I claim that economy considerations, which have contributed to the development of syntactic and phonological studies of generative grammar, also contribute to the study of syntax-phonology mapping. I argue that computational efficiency should be formulated in an explicit manner so that it has empirical effects on the mapping. Specifically, I propose an efficiency condition, which makes the mapping computationally efficient by prohibiting a prosodic domain from being modified in the course of mapping. I show that the so-called restructuring of phonological phrases allegedly induced by syntactic branching not only violates the proposed condition but also has empirical problems, and that it should be analyzed purely phonologically, without considering syntactic factors.