and SIBYLLA DITTMANN University ofMinnesota, Minneapolis, MinnesotaNative Japanese speakers learning English have difficulty perceptually differentiating the liquid consonants Irl and Ill, even after extensive conversational instruction. Using a samedifferent discrimination task with immediate feedback, eight adult female Japanese were given extensive training on a synthetic "rock"-"lock" stimulus series. Performance improved gradually for all subjects over the 14 to 18 training sessions. Comparisons of pretraining and posttraining categorical perception tests with the training stimuli indicated transfer of training to the more demanding identification and oddity discrimination tasks for seven of the eight subjects. Five of seven subjects also improved in identification and oddity discrimination of an acoustically dissimilar "rake"-"lake" synthetic series. However, transfer did not extend to natural speech words contrasting initiallrl and Ill. It was concluded that modification of perception of some phonetic contrasts in adulthood is slow and effortful, but that improved laboratory training tasks may be useful in establishing categorical perception of these contrasts. 131In the phonological system of a particular language, only a subset of all possible phonetic contrasts are utilized to differentiate lexical items. Different languages utilize different subsets of phonetic contrasts; thus, the patterns of speech production and perception for a speaker-hearer are constrained by the phonological system of his or her native language. It has long been known that, for adults, learning a new pattern of speech production in a foreign language is problematic. The persistence of a foreign accent includes, among other things, a failure to produce phonetic contrasts appropriately in the new language, contrasts that are not

To assess the extent to which adults can discriminate intraphonemic differences in place-ofarticulation, we gave 4 subjects extensive training on a synthetic rock-lock series, using a same-different discrimination task. Transfer of training was evaluated using oddity discrimination tests of the training series and an acoustically dissimilar rake-lake series. Performance during training showed gradual improvement in intraphonemic perception, with asymptotic performance well below 100% accuracy. Oddity discrimination improved for the rock-lock stimuli, but not for the rake-lake series, suggesting that learning was specific to the acoustic parameters of the training stimuli. Distinctions between auditory, phonetic, and phonemic modes of speech perception are discussed.