Challenges in Accommodating Talker Variability in Lexical Tone Perception Among Autistic Adults: Implications for Talker Adaptation

Abstract

Abstract Purpose Successful speech perception requires listeners to accommodate acoustic variability arising both across talkers and within a single talker. This challenge is especially pronounced in tone languages, where fundamental frequency (F0) simultaneously encodes lexical meaning and talker identity. The present study investigated how between-talker and within-talker variability affects lexical tone perception in Cantonese-speaking adults with and without autism spectrum disorder (ASD). Methods Twenty-five adults with ASD and 25 neurotypical (NT) controls completed a speeded word identification task involving minimal word pairs differentiated by level tones. Between-talker variability was manipulated by presenting words produced by either a single talker or multiple talkers, and within-talker variability was manipulated by using either a single exemplar or multiple exemplars from each talker. Performance was assessed using identification accuracy, response time (RT), and a composite efficiency measure. Results Both groups demonstrated reliable processing costs associated with between-talker variability, reflected in lower accuracy, longer RTs, and reduced efficiency, with no group difference in the magnitude of this effect. In contrast, only the ASD group exhibited reduced efficiency under within-talker variability. Furthermore, within-talker and between-talker processing costs were not significantly correlated in either group. Conclusion These findings suggest that autistic adults may experience greater difficulty accommodating subtle within-talker acoustic variability during lexical tone perception, consistent with a more detail-focused processing style. Such difficulty may have implications for everyday speech understanding in socially diverse listening environments while highlighting the importance of distinguishing between-talker normalization from within-talker stabilization in models of speech perception.