VOT variability: Within‐subject and between‐subject measurements of stop‐consonant production by female talkers of English, Japanese, Navajo, and Spanish

Abstract

Voice onset time (VOT) presents the language learner with a problem in temporal perception and generalization. However, the variability of this cue in real‐speech tokens has not been reported. Two female native talkers for each of four languages—English, Japanese, Navajo, and Spanish—produced six repetitions of their native‐language stop consonants in a nonsense syllable form: ha‐ba, ha‐da, etc. Talkers sat in an anechoic chamber 0.5 m from an AKGC451E microphone and self‐monitored syllable peak levels using a sound level meter. Productions were digitized (Nakamichi DMP100 pulse‐code modulator) and stored on videotape (Fisher 805A video cassette recorder). The VOTs were measured from waveforms produced on a Kay 7800 digital sonograph. Results illustrate: (1) individual differences in VOT patterns; (2) patterns of absolute‐value range that are language specific; (3) less consistency in actual VOTs than suggested in the synthetic‐speech literature; and (4) the need for a multi‐cue study of phoneme perception in order to examine individually characteristic trade‐offs among cues that may be the rule rather than the exception in real speech. [Work supported by AFOSR.]