UCLA

This thesis uses empirical data to make the case that the architecture of Grammar is feed-�forward. There are three thematic parts: first, I discuss the separation of different phonetically important variables into modules and how a few prominent planning and production models have regarded the order of those modules. Then, I give a novel, statistical operationalization of the concept of feed-�forwardness. Finally, I present the results of a production experiment — targeting the center of gravity (COG) of American English [ʃ] — wherein multiple phonetically important variables are simultaneously manipulated. The aforementioned models are translated into linear mixed effects models and then compared based on their ability to predict the experimental data; I find that the model which corresponds to Keating & Shattuck�-Hufnagel’s 1989 model and Chomsky’s 1965 T�-model most parsimoniously fits the data, thereby providing phonetic evidence for a feed-�forward model of the Grammar. Discussion of some implications for phonetic and phonological research ensues.