Intonation structure and disfluency detection in stuttering. Paper presented at LabPhon10

Despite traditional evidence that lexically-stressed content words are the most common loci of disfluencies in stuttered speech (Natke, et al., 2002; Prins, et al., 1991; Brown 1938), the fact that certain function words—crucially, those directly preceding stressed content words—are also disfluently produced (Au-Yeung, et al., 1998; Howell & Sackin, 2000), suggests that the actual trigger of a disfluency occurs later than the speech perturbation itself. It was hypothesized in this paper that stutterers ’ disfluencies would be accompanied by more prosodic irregularities prior to the actual disfluency than would non-stutterers ’ disfluencies, and that the underlying disfluencies would be triggered by metrically prominent material in the phrase. Three stutterers and their age-matched controls were recorded performing a spontaneous narration of a picture book. Results supported the hypothesis that metrically prominent, or pitch-accented, words would attract a higher rate of stuttered disfluencies than would unaccented words. Stutterers also produced a higher percentage of their disfluencies on the nuclei of words than did controls; moreover, these disfluencies attracted pitch accents more often than did those of controls, providing further evidence of anomalous derivative disfluencies surfacing in advance of underlying ones