Abstract
We explored the possibility that implicit learning plays a role in base-rate sensitivity during perceptual category learning. Participants learned to categorize simple stimuli (bar graphs varying in height from trail to trial) with unequal category base-rates (relative exemplar frequencies). Implicit learning was explored via manipulations previously used to test the COVIS (Competition between Verbal and Implicit Systems) theory of categorization. These included learning with different response types (making a categorization response on each trial vs. observational learning) and feedback delays (immediate corrective feedback after each response or after a 5 second delay on each trial). We also manipulated the salience of base-rate information by manipulating base-rate ratio (2:1 or 3:1), discriminability level (d'=1 or d'=2), and category structure training (pretraining or not on category structures prior to base-rate manipulation). We found that performance in response conditions was closer to optimal than performance in observational-learning conditions, and that performance was closer to optimal when feedback immediately followed a response. In other words, under manipulations known to disrupt implicit learning (observational learning, delayed feedback), performance clearly suffered. There was a strong interaction between these trends and category discriminability. When d' was lower and when category structure was not pre-trained (i.e., when the available perceptual information was less informative), the influence of base-rate information on decision-criterion placement was clearly stronger. These results suggest that implicit (or procedural) learning does indeed play a role in the learning of base-rate information.
Meeting abstract presented at VSS 2014