Abstract
The COVIS theory of categorization (Ashby et al 1998, Psychological Review) posits that verbalizable (explicit) rule learning is mediated in part by prefrontal cortex (PFC), while nonverbalizable (implicit) rule learning is mediated chiefly by basal ganglia structures. COVIS also predicts that both learning systems attempt to determine each category response on each trial (i.e., the systems compete). On this basis, we predicted unique patterns of PFC blood flow depending on how participants performed on perceptual category learning tasks that required either selective attention to a single stimulus dimension ("rule-based" learning) or attention to multiple stimulus dimensions at once ("information-integration" learning). Participants completed rule-based and information-integration category learning tasks with two-dimensional stimuli (gabor patches varying in orientation and spatial frequency across trials). Hemodynamic response was measured in dorsolateral PFC using functional Near Infrared Spectroscopy (fNIRS). As expected, we found similar levels of oxygenated hemoglobin (Hbo2) in DLPFC early in learning for both conditions, and a divergence of activity level across tasks over the series of training blocks. This divergence was mediated by the type of rule used in each task. Participants using the incorrect rule type (a 1-dimensional rule) during information integration learning showed higher PFC activity than those using the appropriate rule type. These results support COVIS and suggest that these perceptual category learning tasks provide a dissociation that may be useful for examining changes in PFC integrity due to injury or aging.
Meeting abstract presented at VSS 2014