The perceptual locus of spectral tilt

Abstract

In a previous study [Richardson and Sawusch, J. Acoust. Soc. Am. Suppl. 1 85, S136 (1989)] using natural, synthetic, and modified voiced stops it was shown that spectral tilt does not play a role in human stop identification. In the present study, subjects were presented with three sets of tone analogs that mimicked the voiced stops [b], [d], and [g] in conjunction with the vowels [ɛ], [al, and [u]. All stimuli were based on the speech tokens described by Richardson and Sawusch (1989). The first set (“natural” tones) was based upon high‐quality, synthetic CV tokens. The second set (“stylized” tones) was also based upon the synthetic CV tokens. In this set, each token with the same vowel (e.g., [ba], [da], and [ga]) was given the same set of parameter values as for the vowel portion of the syllable. The third set (“modified” tones) was based upon the stylized tones and contained identical changes in spectral tilt for each stop place of articulation. Both speech and nonspeech subjects were run in a perceptual grouping task in order to investigate the perceptual locus of any effects of the change in spectral tilt. Results will also be discussed in terms of general human auditory‐phonetic coding ability and the nature of the phonetic representation used in speech perception. [Work supported by NIDCD Grant No. DC00219 to SUNY at Buffalo.]