It is widely believed that a critical period exists for second-language (L2) speech learning. However, the actual cause(s) of foreign accent remain uncertain. The research proposed here will evaluate a model that attempts to account for age-related changes in bilinguals' production and perception of vowels and consonants (""""""""sounds"""""""") in their L2 and in their native language (L1). Segmental production errors in the L2 are important because they result in misunderstandings and/or slowed processing, and contribute to negative personal evaluations. A less obvious, but nevertheless serious consequence of L1 vs L2 phonological differences are segmental perception errors that contribute to difficulties in comprehending the L2. Such errors might arise from neurological changes that diminish the ability to add or modify sensorimotor programs, or prevent the formation of perceptual representations for sounds not found in the L1. Deter-mining the cause(s) of segmental production and perception errors is important clinically. For example, if they arise because of maturationally determined limits on neural plasticity or sensorimotor flexibility, then the benefit of foreign accent reduction therapy, or speech therapy for those who have suffered brain trauma, may be severely limited for individuals who have passed the critical period. On the other hand, if aging merely results in a non-optimal use of intact learning mechanisms, then renewed effort must be directed at determining how therapy can best be structured to optimize adults' capabilities. According to the model under evaluation, foreign accents arise because bilinguals are unable to separate effectively the sound systems of their two languages. The model predicts three sources of error in producing an L2 sound: (l) Partial merger with the properties of a neighboring L1 sound in instances where the L2 and L1 sounds persist in being linked perceptually; (2) Incorrect featural specification, in instances where the L2 uses different features or feature weights than the L1; and (3) Deflection away from a neighboring L1 sound when a new category has been formed. Nine experiments will examine the performance of Italian-English or Japanese-English bilinguals who differ primarily according to their age of learning (AOL) English. By hypothesis, the frequency of error types #l and #2 will increase as AOL increases whereas error type #3 will decrease in frequency as the bilinguals' AOL increases. Three proposed experiments will define the acoustic and perceptual relation between the vowels in L1 and L2. Another will assess the accuracy with which bilinguals produce L1 and L2 vowels, and test for the formation of categories for vowels in English using a categorical discrimination task. Other experiments will use similar techniques to examine the production and perception of consonants. One question of interest is whether bilinguals who correctly produced English consonants not found in the L1 inventory formed categories for the English consonants whereas bilinguals who produced the English consonants incorrectly did not form categories for them. The two remaining experiments will assess the relation between AOL and the bilinguals' use of non-L1 features in perceiving English vowels and consonants.
