Cues to vocal-fold and vocal-tract lengths interact in talker-sex perception from vowel segments

Abstract

Even brief sounds can be richly informative about talker characteristics. Earlier acoustic analysis confirmed predictions that the best statistical classification of vowels by talker sex occurred using acoustic correlates of dimorphism in vocal-fold and vocal-tract lengths [J.-A. Bachorowski and M. J. Owren, J. Acoust. Soc. Am. (in press)]. In follow-up perceptual testing, 24 participants in experiment 1 classified talker sex from 180 short segments that balanced fundamental frequency (F0) and estimated vocal-tract length (VTL) cues. Responses were 98% correct, with a mean latency of 478 ms. However, latencies were longer for male talkers that combined high F0 and short VTL values, as well as for female talkers with low F0 and long VTL. In experiment 2, 24 participants heard stimuli composed of small numbers of waveform cycles. Classification ‘‘threshold’’ was 1.8 cycles, with male talkers being more accurately classified than females. These results show that indexical cueing is an inherent component of even the shortest possible vowel segments, and that the acoustic features most closely related to sexual dimorphism interact in predictable fashion in influencing perceptual processing. F0 and VTL, the particular acoustic cues in question, are fundamentally related to features also known to influence speech processing.