Within-trial variability improves rule-based, not information-integration, category learning
Categorization is a critical component of cognition and contributes to many complex processes, including speech perception. High variability within the environment is thought to initially slow learning while increasing the ability to generalize to novel exemplars. However, little is understood about the mechanisms driving this benefit of variability. The current study investigates the effect of pairing within-category variability with response and feedback within single category-learning trials. Participants who learned categories defined by boundaries orthogonal to the category dimensions—rule-based categories—had superior learning and were better able to generalize to novel exemplars when they were trained with within-trial variability compared to when only a single exemplar was presented on each trial. In contrast, participants who learned categories defined by boundaries involving reliance on both category input dimensions—information-integration categories—showed no enhancement of learning from within-category variability. This draws a distinction between overall variability in the acoustic environment and variability more tightly coupled with response and feedback. The influence of variability as experienced within a single trial differs substantially depending on the nature of the category learning challenge. The results have implications for learning speech categories and for further understanding the mechanisms that contribute to auditory category learning.