Dissociating Categorization Learning System: Overconfidence and Variations Learning Rates in Accuracy and Confidence Reports
In examination of dual-process models of categorizations, research has typically focussed on the manner in which categorization responses change over time. However, one of the basic assumptions of a prominent dual-process account (COVIS) is that an explicit learning system dominates initial stages of training whereas an implicit learning system dominates later stages of training. In three experiments, we consider the utility of using subjective measures of performance (i.e., confidence reports) to continuously sample from a participantâ€™s explicit representation of the category structure while also examining changes in these reports over the course of training. The results of an examination of learning rates and the block at which participants reached a performance asymptote support multiple processes and representations accounts of categorization.