Phonological transfer in second language perception and production

This study provides an account of how a speaker's first language (L1) influences his/her second language (L2) speech, an effect also known as cross-language transfer. The account of transfer proposed here relies on theoretical constructs and techniques from linguistics, psycholinguistics and, to a limited extent, computational modelling. The study contributes to an understanding of the perception-production relationship in L2 speech, how transfer factors into that relationship, and general (universal) principles which constrain transfer.To explain interdental substitutions by speakers of English as a second language (ESL), an analysis based on feature prominence (Halle 1959) is adopted. Phonological segments are assumed to consist of a set of features (a la Chomsky and Halle 1968), each of which has a relative prominence or weight of importance based on a language's segment inventory. By underspecifying (Archangeli 1984) the segment inventory of a language, a feature's prominence relative to other features arises as a result of the minimal phonemic distinctions it makes in the overall inventory. Features with higher prominence are considered to be more salient to speakers of that particular language. Upon hearing unfamiliar aural input, features which are more salient are perceived more easily. This in part dictates which sounds will be substituted for interdentals.The phonemic inventories of four languages are analyzed and the relative prominence of (continuant) in each is derived. Those languages are: Hindi, Turkish, German, and Japanese. ESL speakers of these languages are then tested on their perception of English interdentals across several contexts. Results indicate that the above analysis predicts how three of the four sets of language speakers perceive interdentals. The same ESL speakers were tested on production of interdentals across different contexts, and the results of the perception and production tasks are compared. By looking at four different first languages and the speakers' perception and production abilities, universal and language-specific effects as well as task-specific effects are observed.A computational model based on connectionist processing was designed to simulate the hypotheses and results from the experiments. A connectionist model was used because of its compatibility with functional approaches to (second) language acquisition (Gasser 1991, MacWhinney 1989). Though the results of the modelling were not ideal, they suggest that the modelling is a plausible and useful method for understanding sound substitutions.U of I OnlyETDs are only available to UIUC Users without author permissio