Identification of American English vowels by native Japanese speakers: Talker-and-token-based analysis

Abstract

Native speakers of Japanese identified American English vowels /i, I, ε, æ, ■, ■/ produced by four female native speakers of American English in /CVC/ contexts. Native speakers of American English served as the control group, and they outperformed the Japanese subjects in identifying all the English vowels in every /CVC/ context. In another experiment the Japanese subjects equated these English vowels with Japanese vowels. In general, English vowels were equated with phonetically close Japanese vowels, but significant talker effect was observed. The /i/ tokens equated with the Japanese long high front vowel /ii/ were much more correctly identified as /i/ than those equated with the Japanese short high front vowel /i/. These tokens were more often misidentified as /I/. The /■/ and /■/ tokens were predominantly equated with the Japanese low vowel /a/. The percent-correct identification of /■/ and /■/ was low in most of the /CVC/ contexts, and these two vowels were often misidentified as each other, and the Japanese subjects’ latency before they decided what vowel they had heard was longer when /■/ or /■/ tokens were presented. The Japanese subjects do not seem to have salient cues to differentiate /■/ and /■/.