Phonetic variability leads to gradient speech perception

Listeners show sensitivity to within-category variation in speech sounds. This gradient categorization system helps promote flexibility, deal with ambiguity, and maintain plasticity. Prior work has raised the possibility that statistical learning mechanisms might be involved showing that listeners adopt a more gradient categorization when exposed to greater phonetic variability (Clayards et al., 2008; Theodore and Monto, 2019). However, this work relied on a 2AFC task in which a shallower slope could derive from either difference in gradiency or differences in noise in the system. We thus reinvestigated this issue across three experiments. Participants were trained on distributions of voice onset time with either high or low variance. Later, gradiency was assessed with the VAS task, which overcomes these limits. Experiment 1 (n = 84) did not support the hypothesis that gradiency is an adaptation to increased variance. Instead, higher variance increased trial-by-trial inconsistency. Experiment 2 (n = 168) utilized 28 continua to better generalize across stimulus characteristics and found robust evidence for increased gradiency as a result of variability. Experiment 3 (n = 85) introduced a baseline condition to rule out the alternative explanation that listeners were adapting to low variance by becoming less gradient, instead of becoming more gradient in the face of high variance.