Abstract
A prominent neuropsychological theory of category learning called COVIS (for Competition between Verbal & Implicit Systems) posits that separate brain systems compete to learn a classification rule. The prefrontal cortex (PFC) tends to mediate learning of verbalizable category rules while the basal ganglia underlies learning of non-verbalizable category rules. Because the systems compete, the explicit system may persist in seeking a classification rule even when the implicit system is better suited to the task. As a result of this prediction, we expect higher PFC activity when learners attempt to use a verbalizable rule when a nonverbalizable rule is appropriate. When a verbalizable rule is appropriate, we expect PFC activity to decline once the rule has been discovered. The current experiment replicates earlier studies that observed explicit (verbalizable) or implicit (nonverbalizable) classification rule learning. Functional near-infrared spectroscopy (fNIRS) was used to measure hemodynamic changes in the PFC as participants learned over three blocks of training trials. Participants completed either a rule-based (RB, explicit) or information-integration (II, implicit) learning task. Participants classified simple 2D stimuli: lines that varied in orientation and length from trial to trial. Across blocks response accuracy increased in both conditions. Participants in the II condition showed greater PFC activation than those in the RB condition, particularly near the end of training. Dividing participants into early and late learners (based on block 1 training performance), late learners in the II condition again displayed greater PFC activation than in the RB task (amplifying the trend found when all participants were included in the analysis). Focusing on the II condition, late learners showed significantly greater activation than early learners in blocks 2 and 3 (who had already learned the rule) implying that they were still attempting to discover a verbalizable classification rule later in training. These results support our COVIS-based predictions.
Meeting abstract presented at VSS 2016