Impaired Categorical Perception of Speech Sounds Under the Backward Masking Condition in Adults Who Stutter

Evidence increasingly indicates that people with developmental stuttering have auditory perception deficits. Our previous research has indicated similar but slower performance in categorical perception of the speech sounds under the quiet condition in children who stutter and adults who stutter (AWS) compared with their typically fluent counterparts. We hypothesized that the quiet condition may not be sufficiently sensitive to reveal subtle perceptual deficiencies in people who stutter. This study examined this hypothesis by testing the categorical perception of speech and nonspeech sounds under backward masking condition (i.e., a noise was presented immediately after the target stimuli).Fifteen Cantonese-speaking AWS and 15 adults who do not stutter (AWNS) were tested on the categorical perception of four stimulus continua, namely, consonant varying in voice onset time (VOT), vowel, lexical tone, and nonspeech, under the backward masking condition using identification and discrimination tasks.AWS demonstrated a broader boundary width than AWNS in the identification task. AWS also exhibited a worse performance than AWNS in the discrimination of between-category stimuli but a comparable performance in the discrimination of within-category stimuli, indicating reduced sensitivity to sounds that belonged to different phonemic categories among AWS. Moreover, AWS showed similar patterns of impaired categorical perception across the four stimulus types, although the boundary location on the VOT continuum occurred at an earlier point in AWS than in AWNS.The findings provide robust evidence that AWS exhibit impaired categorical perception of speech and nonspeech sounds under the backward masking condition. Temporal processing (i.e., VOT manipulation), frequency/spectral/formant processing (i.e., lexical tone or vowel manipulations), and nonlinguistic pitch processing were all found to be impaired in AWS. Altogether, the findings support the hypothesis that AWS might be less efficient in accessing the phonemic representations when exposed to a demanding listening condition.https://doi.org/10.23641/asha.20249718.