Perceptual learning of acoustic correlates of nonspeech categories

Abstract

Although recent evidence suggests that certain phonetic trading relations have an auditory basis, others appear to be the result of listeners' knowledge of articulatory constraints on production, or of their acoustic consequences. The latter group includes the effect of first‐formant (F1) frequency on the voicing boundary in initial stop consonants; as F1 is lowered, longer voice‐onset times (VOT) are required to perceive stops voiceless. Summerfield [J. Acoust. Soc. Am. 72, 51–61 (1982)] and Hillen‐brand [J. Acoust. Soc. Am. 75, 231–240 (1984)] showed for nonspeech stimuli varying in tone‐onset time (TOT), simultaneity‐successivity judgments did not vary systematically with frequency of the lower tone. This study demonstrates that when listeners learn to associate different frequency components with two nonspeech categories (simultaneous and nonsimultaneous onsets), they will use frequency information in making simultaneity judgments. Two TOT series were generated, consisting of two‐component, steady‐state, sine‐wave stimuli; in each series, the onset of the lower tone (f1) lagged that of the higher tone (f2) by 0–120 ms. For all stimuli, f2 was 1240 Hz; f1 was 750 HZ for one series and 250 Hz for the other. When subjects were trained on the simultaneous endpoint from the high f1 series and the nonsimultaneous endpoint from the low f1 series, subsequent identification performance revealed significantly more stimuli were labeled “simultaneous” for the high‐f1 series than for the low‐f1 series. These results suggest that general principles of perceptual learning can account for the interaction between VOT and f1 onset in specifying voicing categories in initial stops. [Work supported by NICHHD.]