Abstract
Few theories of categorization make predictions about the effect of mere exposure on categorization behavior. Statistical learning studies suggest that mere exposure to stimuli can result in significant learning about the correlational structure of stimulus features. While correlational structure may be learned, there is little or no evidence that this learning might facilitate subsequent visual category learning. We tested whether pre-exposure to a stimulus set enhanced or impaired learning to categorize the stimulus set according to a simple conjunctive rule. During a pre-exposure phase, participants performed 480 trials of a 1-back task on a series of numbers appearing at the center of the screen. In one condition (pre-exposure), each number in the 1-back task was accompanied by one of 24 novel six-featured “alien” stimuli that appeared in the background behind the numbers and were irrelevant to the 1-back task. In the control condition the numbers appeared without the stimuli in the background. In the subsequent category learning phase, both groups learned to categorize the aliens according to a 2-dimensional conjunctive rule. The participants that had seen the alien stimuli during the 1-back task (pre-exposure group) were able to learn the rule in fewer blocks than the control group. This finding suggests that rule-based category learning can benefit from unsupervised learning - possibly visual statistical learning. The finding also suggests that category learning of this kind may be resistant to latent inhibition, a phenomenon observed in classical conditioning studies where mere exposure impairs learning. A second experiment examined the effect of pre-exposure to stimuli with various patterns of correlation between the stimulus dimensions. Results suggest that an effect of pre-exposure on category learning is dependent on the pattern of correlation between pre-exposed stimulus features.
This work was supported by a training grant awarded by the National Institute of Health through the Vanderbilt Vision Research Center (PHS 5-T32-EY07135-14) and by a grant from the James S. McDonnell Foundation to the Perceptual Expertise Network and also by the Temporal Dynamics of Learning Center (NSF Science of Learning Center SBE-0542013).