Listeners rapidly adapt to current conditions: “Good enough” adaptation in multi-talker speech perception

Listeners must cope with highly variable input to successfully recognize speech. One way theydo this is by adapting to the systematicities of individual talkers. Research on talker-specificadaptation has found that listeners either generalize talker-specific phoneme categories to newtalkers or individuate them, creating talker-specific categories. Five experiments investigated theconditions under which listeners generalize or individuate talker-specific phoneme categoriesusing a distributional learning paradigm. In Experiment 1, participants (N=413) acquired a novelvoicing boundary for a new talker and generalized this boundary to another novel talker. In alater session (1-3 days later), they were trained on a second novel talker but showed noevidence of learning it and no evidence of retaining the first talker. Experiment 2 (N=355)demonstrated that the lack of retention was a not a product of interference from learning in thesecond session. We also asked if listeners individuate talkers when exposed to multiple talkerssimultaneously in a distributional learning paradigm (Experiment 3, N=113) and in a supervisedlearning paradigm (Experiment 4, N=125). Neither showed evidence for talker-specific learning.Finally, Experiment 5 (N=97) demonstrated rapid learning of new categories in as few as 48trials, which can rapidly be unlearned at test or in a new training block. This argues participantsrapidly adapt a set of categories to the current listening environment, but imperfectly, adapting asingle boundary to each new talker rather than demonstrating talker-specifically. This goodenough adaptation may be sufficient for everyday needs.