When multiple talker exposure is necessary for cross-talker generalization: Insights into the emergence of sociolinguistic perception

Sociolinguistic research finds that: (i) the speech signal contains talker-specific and socio-indexical structure, with talkers varying idiosyncratically within the same social category and systematically across categories; (ii) both talker-specific and socio-indexical variation influence speech perception. What is unclear is how sociolinguistic perception arises—following exposure to an unfamiliar, socially-mediated variant, how do listeners learn that this feature is characteristic of a broader social group and can generalize to other group members? The current study exposed listeners to an unattested variant in L1-English (a /p/ to [b] phonetic shift), investigating how the number of exposure talkers mediates cross-talker generalization. All participants completed an exposure phase (phrase-final keyword identification) followed by a test phase (categorization along a buy–pie continuum for a novel female and male talker in separate blocks). Experiment 1 exposed listeners to a single shifted female talker (“The novel is now in brint”) and a single unshifted male talker. Experiment 2 presented two shifted female and two shifted male talkers. We find: (i) no generalization in Experiment 1 (no difference in buy–pie response between the novel talkers); (ii) robust generalization in Experiment 2 (greater pie response for the novel female than the novel male talker), but only when the novel female block is presented first (i.e., generalization is short-lived). Taken together, the results support a numerosity account: when a previously unheard social variant is presented, multiple talkers per social group seem to be necessary for socially-mediated, cross-talker generalization. This study highlights a critical role of the listener’s social experiences on generalization—multi-talker exposure might be unnecessary when exposed to more familiar types of speech (e.g., L2-accented English) and necessary when exposed to completely unfamiliar variants. Overall, the present experiments enhance our theoretical understanding of cross-talker generalization and offer insights into the emergence of sociolinguistic perception.