What do listeners learn from exposure to a vowel distribution? An analysis of listening strategies in distributional learning

This study first confirms the previous finding that Spanish learners improve their perception of a difficult Dutch vowel contrast through listening to a frequency distribution of the vowels involved in the contrast, a technique also known as distributional training. Secondly, it is demonstrated that learners' initial use of acoustic cues influences their performance after distributional training. To that end, types of unique listening strategies, i.e., specific ways of using acoustic cues in vowel perception, are identified using latent class regression models. The results before training show a split between “low performers”, who did not use the two most important cues to the Dutch vowel contrast, namely the first and second vowel formants, and “high performers”, who did. Distributional training diversified the strategies and influenced the two types of listeners differently. Crucially, not only did it bootstrap the use of cues present in the training stimuli but also the use of an untrained cue, namely vowel duration. We discuss the implications of our findings for the general field of distributional learning, and compare our listening strategies to the developmental stages that have been proposed for the acquisition of second-language vowels in Spanish learners. •A statistical model identified various classes of adult L2 listeners.•Each class represented a specific way of acoustic cue use in vowel perception.•The classes were influenced differently by a 2-min distributional training.•Training increased the use of trained (formants), and untrained (duration) cues.•Cue use confirmed Escudero's L2LP model for the development of L2 perception.