Identifying information-bearing prosodic parameters in severely dysarthric vocalizations

grantor: University of TorontoThis thesis consists of two experiments designed to further our understanding of prosodic control in dysarthria. Speakers with severe dysarthria are known to have reduced range in prosody. Consistent control within that range, however, has largely been ignored. We investigated whether eight speakers with severe dysarthria due to cerebral palsy could control prosodic parameters for sustained vowel and phrase productions. In Experiment I, speakers were able to consistently control sustained production of the vowel /a/ at three distinct levels of duration (short, medium and long). Speakers were more variable in their ability to control distinct levels of pitch (low, medium and high). Nonetheless, seven of eight speakers were able to produce two distinct pitch levels. In Experiment II, we studied whether speakers could vary prosodic cues in phrase level productions. The task was to produce ten unique three-syllable phrases as questions and statements. We inferred the speakers' ability to control prosodic cues by asking a group of 48 naïve listeners to classify dysarthric productions as either questions or statements. Our results indicated that listeners were able to distinguish questions from statements with approximately 87% accuracy. Removal of pitch contour cues from the original recordings dramatically reduced listener accuracy scores to almost chance performance. We can infer that listeners found pitch contour to be an information-bearing prosodic cue in severely dysarthric vocalizations. Some speakers may have also been using syllable duration and loudness cues to mark the question-statement contrast. These findings provide a novel and significant step toward a more complete understanding of the prosodic control abilities of individuals with severe dysarthria. The potential implications for influencing diagnostic and intervention protocols are far reaching. Systematic exploration of the speaker's prosodic control abilities would enable clinicians to develop appropriate treatment regimes to optimally shape residual abilities for the purpose of enhancing communication. This line of research may also guide the development of voice-driven assistive aids that harness information within the prosodic channel.Ph.D