| Linguistic research has documented a wide range of regularities in phonological structure. Within languages, certain sound combinations are ill-formed, never appearing in words of the language; across languages, certain phonological structures are preferred to others. Less well understood are the types of regularities that are encoded by the spoken production system. To explore this question, section 1 describes three theories regarding the types of regularities that are encoded. These theories are: one, the Instance-Based theory—gradient regularities based on within-language token frequency of segmental and supra-segmental structures are encoded; two, the Lexical Distribution theory—gradient regularities based on within-language type frequency of segmental and supra-segmental structures are encoded; and three, the Markedness theory—categorical regularities based on cross-linguistic and within-language markedness of sub-segmental, segmental, and supra-segmental structures are encoded.; Building on previous research, a framework for spoken production processing is described in section 2. The three theories are situated within this general framework. Section 3 then reviews previous research regarding the types of regularities that are encoded. These studies suggest that categorical within-language phonological regularities are encoded by the spoken production system, but fail to distinguish between the three theories.; Section 4 reports the results of two experimental studies designed to contrast the predictions of the three theories. These two experiments are the first to demonstrate that sub-segmental regularities must be encoded by the spoken production system. Experiment 1 uses an implicit learning paradigm. As predicted by the Markedness theory, participants in this experiment are sensitive to sub-segmental regularities. Furthermore, gradient regularities are encoded, supporting the predictions of the Instance-Based and Lexical Distribution theories. Experiment 2 examines biases in speech errors. The biases conform to the regularities of the Markedness theory, but exhibit gradient effects. These results support a theory incorporating elements of all three theories (i.e., gradient as well sub-segmental regularities are encoded).; Section 5 discusses the implications of the results presented in section 2, 3, and 4 for the computational mechanisms implementing phonological processing. Future work to extend this research is outlined, including an extension to existing computational theories that may account for the full range of results. |
