The Effect of Audio Sampling Rate on the Validity and Reliability of Clinical Measurements of /s/ and /z/ in Children and Adolescents

PURPOSE: Sibilant fricatives have acoustic energy weighted toward the high-frequency range, which researchers and clinicians use to assess articulatory and aerodynamic factors of a speech sound. This study investigates whether the clinical perception of child and adolescent fricatives is impacted by sampling rate, which affects the amount of high-frequency energy represented in a digitized recording. The research is motivated by the rise of speech analysis pipelines that use a 16-kHz sampling rate. METHOD: A total of 150 utterances each of /s/ and /z/, from children and adolescents 7.7-16.2 years of age, were rated by expert listeners trained using the perceptual error rating for the clinical evaluation of phonetic targets (PERCEPT) Rating Scale to evaluate sibilant accuracy at three different playback sampling rates: 44.1, 22.05, and 16 kHz. Linear mixed modeling analyzed whether 16 or 22.05 kHz sampling rates influenced accuracy ratings or the labeling of categorical error subtypes relative to 44.1 kHz audio. RESULTS: There was no evidence that sampling rate systematically influenced PERCEPT Rating Scale scores for /s/ or /z/, with post hoc power simulations indicating that the study achieved adequate power. Sampling rate also did not significantly impact the proportion of error subtype labels assigned by raters, nor the interrater nor intrarater reliability of the ratings. CONCLUSIONS: Clinical perception of these child and adolescent fricatives was not significantly impacted by sampling rates as low as 16 kHz. These findings do not preclude the validity and reliability of speech technologies for child and adolescent fricatives on the basis of 16-kHz digitization for use by well-trained raters evaluating children and adolescents in the 7-16 years age range. SUPPLEMENTAL MATERIAL: https://doi.org/10.23641/asha.34750809.