Native Japanese speakers often perceive English vowels based on their duration, whereas native speakers use spectral cues (formant frequencies). The current study examined whether 23 Japanese adult learners of English could create a new vowel category along the spectral dimension after phonetic training with the English vowels /i/-/ɪ/ as in beat and bit. As in previous training studies, the Japanese trainees improved their ability to categorize the vowels in tokens included in the training, and were able to generalize to novel tokens and talkers. A cue-weighting task confirmed that the Japanese trainees categorized the vowels based on temporal cues before training, presumably because vowel duration is used phonologically in Japanese. While 12 of the Japanese trainees still relied on vowel duration to distinguish the English vowels after training, 11 of them successfully created a new vowel category along the spectral dimension. However, their category boundary between /i/-/ɪ/ was set at a different location from that of native speakers. We interpret these results as providing evidence that it is possible for late inexperienced learners to create new phonetic categories with phonetic training, but that the new categories may be subject to phonetic category dissimilation, an effect previously documented with early bilinguals.
ASJC Scopus subject areas