Characterizing listener perception of dysarthric speech beyond speech intelligibility

Speech intelligibility, naturalness, and listening effort are interrelated perceptual dimensions that appear to be similarly affected in dysarthric speech when examined across speakers. However, it is unclear whether within-speaker changes in one dimension relate to changes in the others. This has implications for understanding whether these perceptions are formed from stable speaker-level characteristics or more fine-grained aspects of speech. The current study examined how listener perceptions vary in response to within-speaker variability across different speech tasks-habitual speech, clear speech, and structured interactive speech. Naïve listeners heard /h/-vowel-/d/ words produced by individuals with dysarthria. In two separate tasks, they rated either speech naturalness or listening effort and identified the stimuli they heard. Naturalness ratings, effort ratings, response accuracy, and response times were analyzed. Intelligibility and reported listening effort were improved in clear and interactive speech compared to habitual speech. Furthermore, intelligibility and response time were improved for interactive speech over clear speech. Speech naturalness did not differ across tasks. Finally, greater speech intelligibility was associated with faster response times, but not with naturalness or reported listening effort. These findings suggest that intelligibility, naturalness, and listening effort represent distinct constructs that are affected differently by speech modifications.