Two competing views exist on the origin of interface categories (prosodic words, phonological phrases, intonational phrases): rhythmic determinism, which holds that they originate in phonology, and syntactic determinism, which derives them from syntactic constituents. This paper presents empirical evidence from Japanese and Wenzhounese, showing that at least a subset of interface categories cannot be derived from syntax and therefore must be rhythmically determined. These rhythmically determined interface categories call for a generative phonological module, which is incompatible with the classic T-model’s assumption that only syntax can generate hierarchical structure. Consequently, evidence for rhythmic determinism supports a parallel architecture, where phonology can be generative and is not secondary to syntax. In LFG’s parallel architecture, adopting rhythmic determinism entails that a direct correspondence from c-structure to p-structure is empirically insufficient. I discuss two LFG approaches to the syntax-phonology interface, showing that they can capture the Japanese and Wenzhounese data, though some theoretical issues remain to be addressed.
Chen Xie (Thu,) studied this question.