7th International Conference on Spoken Language Processing
September 16-20, 2002
This paper reports some of the findings of a three-week perceptual training experiment involving Japanese and Korean learners of English as a second language. Other studies have shown that under appropriate conditions, adults can learn to identify nonnative speech sounds such as American English /r/ and /l/ with generalization to novel stimuli, transfer to new tasks and retention. Such conditions of "high variability" have included multiple talkersí voices and a large stimulus training set involving these sounds across word positions. The current study investigated additional sources of variability by exploring the effects of the adjacent vowel and talkersí facial cues on perceptual training. Thus, word position, adjacent vowel, and training type [auditory-visual (AV) vs. auditory-only (A-only); multiple talker vs. single talker] were independent variables. Results indicated significantly greater improvement in identification accuracy for AV vs. A-only training. Visual input contributed the most to the bimodal percept for the more difficult pre-training phonetic environments. Significant effects were found for vocalic environment as well as word position and training talker. Findings also revealed successful transfer to novel stimuli, a new talker, production improvement, and earlier word identification in connected speech.
Bibliographic reference. Hardison, Debra M. (2002): "Sources of variability in the perceptual training of /r/ and /l/: interaction of adjacent vowel, word position, talkers≤ visual and acoustic cues", In ICSLP-2002, 1465-1468.