Temporal factors in vowel perception

Abstract

The perceptual role of the temporal fine-structure of vowel waveforms was investigated in five experiments. The interaction of the fundamental frequency and the first formant (F1) was shown to result in temporal patterns consisting of a number of cycles of F1 per fundamental frequency period. Changes in these patterns were shown to correlate with shifts in the perceptual boundary between /i/ and/I/. The results indicate that the perceptual system was responding to either a change in number of cycles of F1 per fundamental period or a change in the harmonic structure of the sounds. The hypothesized temporal cue was then used in synthesizing vowel- like sounds which, while not differing in formant center-frequency or harmonic structure, did differ in temporal structure. Subjects were able to match different sequences if the vowel-like sounds with sequences of natural vowels as predicted by their temporal properties. In a subsequent experi- ment, two pure tones were used as building blocks for synthesizing the vowels a, e, i, o, and u. With careful temporal modeling, the two tones proved sufficient fpr synthesizing intelligible tokens of the five vowels. Although the possibility exists that all the results may be explained in terms of spectrum, the indications are that temporal properties play a considerable role in vowel perception. Subject Classification: [43]70.30; [43]65.75.