Perception of coarticulatory information in normal speech and dysarthria

This study addressed three research questions: (a) Can listeners use anticipatory vowel information in prevocalic consonants produced by talkers with dysarthria to identify the upcoming vowel? (b) Are listeners sensitive to interspeaker variation in anticipatory coarticulation during prevocalic consonants produced by healthy talkers and/or talkers with dysarthria, as measured by vowel identification accuracy? (c) Is interspeaker variation in anticipatory coarticulation reflected in measures of intelligibility?

Stimuli included 106 CVC words produced by 20 speakers with either Parkinson's disease or multiple sclerosis or by 16 healthy controls characterized by an operationally defined normal, under, or over level of anticipatory vowel coarticulation. Ten listeners were presented with prevocalic consonants for identification of the vowel. Ten additional listeners judged single-word intelligibility. An analysis of variance was used to determine differences in vowel identification accuracy and intelligibility as a function of speaker group, coarticulation level, and vowel type.

Listeners accurately identified vowels produced by all speaker groups from the aperiodic portion of prevocalic consonants, but interspeaker variations in strength of coarticulation did not strongly affect vowel identification accuracy or intelligibility.

Listeners appear to be tuned to similar types of information in the acoustic speech stream irrespective of the source or speaker, and any perceptual effects of interspeaker variation in coarticulation are subtle.