Acoustic comparisons of Japanese and English vowels produced by native speakers of Japanese

Abstract

This study explored acoustic similarities/differences between Japanese (J) and American English (AE) vowels produced by native J speakers and compared production patterns to their perceptual assimilation of AE vowels [Strange et al., J. Phonetics 26, 311–344 (1998)]. Eight male native J speakers who had served as listeners in Strange et al. produced 18 Japanese (J) vowels (5 long-short pairs, 2 double vowels, and 3 long-short palatalized pairs) and 11 American English (AE) vowels in /hVbopena/ disyllables embedded in a carrier sentence. Acoustical parameters included formant frequencies at syllable midpoint (F1/F2/F3), formant change from 25% to 75% points in syllable (formant change), and vocalic duration. Results of linear discriminant analyses showed rather poor acoustic differentiation of J vowel categories when F1/F2/F3 served as input variables (60% correct classification), which greatly improved when duration and formant change were added. In contrast, correct classification of J speakers’ AE vowels using F1/F2/F3 was very poor (66%) and did not improve much when duration and dynamic information were added. J speakers used duration to differentiate long/short AE vowel contrasts except for mid-to-low back vowels; these vowels were perceptually assimilated to a single Japanese vowel, and are very difficult for Japanese listeners to identify.