Talker variability can facilitate cross-talker generalization (e.g., comprehension of novel L2-accented talkers can increase after multi-talker exposure compared to single-talker exposure). However, current work contains a confound—do multi-talker exposure benefits stem from numerosity (greater number of talkers) or heterogeneity (greater phonological variability)? The current study assigned L1-English participants (n = 815) to one of five exposure conditions: (i) no exposure control; (ii) single-talker, casual speech; (iii) single-talker, hard-of-hearing-directed speech; (iv) multiple apparent talkers, casual speech; (v) multiple apparent talkers, hard-of-hearing-directed speech. There were two test conditions: (i) single-talker, casual speech; (ii) single-talker, hard-of-hearing-directed speech. Listeners transcribed Mandarin-accented English sentences in noise across both phases (a novel talker was always presented at test). Critically, participants only heard a single exposure talker in reality. Apparent talker variability was created through the “Change Gender” Praat function, which we claim to boost the number of perceived talkers while mitigating changes in phonological variation (i.e., an increase in numerosity with minimal adjustment in heterogeneity). We find that multiple apparent talker exposure can enhance generalization (greater test phase accuracy versus control), but only given acoustic similarity in speaking style across exposure and test.
Aoki et al. (Wed,) studied this question.