Do listeners perceive coarticulatory differences for normal talkers and neurologically impaired talkers?

Abstract

Speech production studies suggest that the extent of coarticulation varies across talkers. The extent of coarticulation may even vary for repetitions of the same utterance produced by a single talker. Thus, an utterance or talker could be characterized by an average degree of coarticulation, a high degree of coarticulation, or a low degree of coarticulation. The perceptual consequences, if any, of this coarticulatory variability are not well understood, particularly for individuals with speech motor control disorders. The current study examined whether the listeners’ speed and accuracy of vowel identification for naturally-produced CV sequences varied depending on the extent of anticipatory vowel coarticulation. Speakers and speech tokens included those characterized by an average degree of anticipatory coarticulation, a high degree of coarticulation, and a low degree of coarticulation. Healthy talkers, speakers with multiple sclerosis, and speakers with Parkinson’s disease produced the stimuli. Consonants in CV syllables beginning with /s/, /k/, or /t/ followed by the vowels /i/ or /u/ were excised and presented to listeners for identification of the following vowel. While identification accuracy may be high across speakers and stimuli, the speed of response should vary with extent of coarticulation, if listeners are sensitive to coarticulatory cues in the acoustic speech stream.