Rate normalization is not necessary for the identification of coarticulated vowels.

Abstract

Consonant–vowel–consonant syllables, excised from a carrier sentence produced at normal and rapid rates, were presented to subjects in a mixed rate condition (normal and rapid syllables presented in random order) and a blocked rate condition (normal and rapid syllables presented in separate subtests). In the mixed rate condition, subjects made errors in judgments of speakingrate on 25% of the trials (50%=chance). Errors were predictable from absolute syllable durations. Nevertheless, identification of the intended vowels in the mixed rate condition (by an independent group of subjects) was only slightly less accurate than in the blocked rate condition (93% vs 98% correct) and in a control condition in which the syllables were presented in the original sentences (98% correct vowel identification and 98% correct rate judgments). These results argue against the hypothesis that perceivers ‘‘normalize’’ for perceived differences in speaking rate prior to identifying coarticulated vowels. Acoustical analysis of this corpus revealed that ‘‘vowel-intrinsic spectral information’’ remained relatively unchanged across speaking rate and consonantal context, while vowel durations varied systematically with speaking rate. [Work supported by NIDCD.]