Purpose: Adapting to talker variability typically incurs a cost to speech perception. However, it remains unclear whether the underlying mechanism of multitalker processing cost (MTPC) reflects the recomputation of acoustic–phonetic interpretations or the reorientation of auditory attention. This study tests these accounts in Cantonese lexical tone perception, where lexical tone categories and talker identity are both represented in fundamental frequency ( F 0). Method: Two tone pairs, one differing primarily in F 0 height and one differing in F 0 contour, were used in a within-subject tone identification task. The number of talkers was manipulated (one, two, four, and eight), varying the numbers of acoustic–phonetic interpretations required. Measures included accuracy, response time, and an efficiency index. Listeners' expectations about talker variability were also assessed for the modulation of the cost. Results: As the number of talkers increased from one to two to four, both tone pairs displayed a gradual increase in MTPC (reduced accuracy, longer response times, or lower efficiency). No additional cost was observed from four to eight talkers. The magnitude of the cost was modulated by expectations. Conclusions: The results suggest that talker adaptation in lexical tone perception is an active process that initially requires recomputation but plateaus with increasing talker exposure. Importantly, the underlying mechanisms differ from that observed in English vowel perception, indicating that talker variability interacts with language-specific phonetic demands. This highlights the need for cross-linguistic approaches to models of multitalker processing.
Wu et al. (Tue,) studied this question.