Acoustic patterns of Japanese voiced velar stops

Abstract

This presentation focuses on Japanese voiced velar /g/. The phoneme /g/ in VCV contexts is said to be characterized by a distinctive wedge-shaped formant pattern in which F2 and F3 converge toward one frequency in the transition from vowel to stop closure, and then diverge as the vocal tract opens from the stop release to the following vowel. This pattern was examined using acoustic and kinematic data from an x-ray microbeam database of Japanese speakers, which is comparable to the English language x-ray microbeam speech production database [Hashi et al., J. Acoust. Soc. Am. 104, 2426–2437 (1998)]. Japanese speakers produced the expected wedge-shaped formant pattern in isolated VCV nonsense syllables, but rarely, if ever, produced this pattern in connected speech. In contrast, English speakers more frequently produced the expected formant pattern in connected speech, though the pattern was less reliably present than in isolated VCVs and varied considerably within and across speakers. These observations highlight substantial differences between controlled laboratory speech and meaningful connected speech, as well as differences in the ways that phonemes are manifested by different linguistic communities. These data also potentially illuminate the relationship among phonetic, acoustic, and kinematic levels of speech production.