Generalization of perceptual adaptation to second-language speech in a single-talker training paradigm: Predictors remain elusive.

First-language (L1) listeners typically show talker general adaptation to second-language (L2) speech after exposure to multiple L2 talkers. However, cross-talker generalization following single L2 talker exposure is inconsistent. We tested two hypotheses regarding generalization following single-talker training. If intelligibility drives generalization, then a high-intelligibility training talker should facilitate cross-talker generalization. However, if similarity drives generalization, then test performance should vary depending on acoustic similarity between training and test talkers. Twenty training-test talker pairs were selected from a dataset consisting of empirical intelligibility measures and pairwise acoustic similarity measures derived from self-supervised model representations for over 100 L2 English talkers. Each talker pair was assessed by 10 L1 listeners in a sentence-in-noise transcription task with an exposure-test format. Most talker pairs showed cross-talker generalization; however, neither intelligibility nor similarity reliably drove generalization. These findings show that single-talker training can be effective, but mechanisms supporting cross-talker generalization following limited exposure remain unexplained. OPEN PRACTICES STATEMENT: The data and materials for the experiments and analyses are available at https://osf.io/b32ka and https://speechbox.linguistics.northwestern.edu/allsstar . The experiment was not preregistered.