Reevaluating perceptual stability in non-native speech categorization: A multi-task perspective

Listeners vary in their use of categorical versus gradient perceptual strategies,which are often less stable and more individually variable in non-native contexts. While two-alternative forced choice (2AFC) and visual analog scaling (VAS) tasks are widely used to assess these strategies, it remains unclear whether individual perceptual profiles remain consistent across paradigms. This study examines whether categorization styles—classified as categorical or gradient—are stable across both tasks. Native Mandarin-speaking undergraduates (N = 137) completed both tasks using an 11-step /da/–/ta/ voice onset time continuum. K-means clustering based on logistic slopes (2AFC) and response distributions (VAS) was used for classification. Only 47.4% of listeners retained the same categorization style across tasks. Agreement was minimal (Cohen’s κ = –0.087), with a significant asymmetry in switching direction (McNemar’s χ² = 4.20, p = .041). Contrary to expectations, some listeners showed more polarized responses in the VAS task. Further analysis revealed that consistent categorizers had steeper 2AFC slopes and lower response variability in VAS. Cross-task correlations were weak, underscoring the task-specific nature of these metrics. These findings highlight the limitations of single-paradigm assessments and support a multi-method approach to better capture individual differences in non-native speech perception.