Natural phonetic variability such as talker differences facilitates second language (L2) word learning. Whether accent variability enhances the talker variability benefit, particularly in the learning of words that differ only in lexical tones, has not yet been examined despite its potential applied value for L2 tone language learning. Two groups of monolingual English speakers completed six training sessions on four minimal-tone quadrads of Mandarin monosyllabic pseudowords, produced either by 12 talkers from Beijing (single accent) or by four talkers each from Beijing, Yantai, and Guangzhou (multiple accents). Bayesian mixed-effects modeling revealed strong learning improvement across sessions in both conditions, but the multiple-accent group improved faster than the single-accent group. In addition, the multiple-accent group demonstrated superior generalization to new talkers with a familiar and an unfamiliar accent, suggesting that natural L2-Mandarin accent variability facilitates English learners’ access to abstract tone-word lexical representations, beyond the benefits of talker variability alone.
