Common and distinct neural substrates of rule- and similarity-based category learning

Categorization is a fundamental ability in human cognition that enables generalization and promotes decision-making. A categorization problem can be solved by employing a rule-based or a similarity-based strategy. The current study aims to elucidate the brain mechanism for category learning by investigating whether the use of the two strategies is supported by common or distinct neural substrates. We conducted three experiments using stimuli with a rule-plus-similarity category structure and applying an EEG-fNIRS fusion methodology. In Experiment 1, participants were explicitly instructed to use either a rule-based (single feature) or a similarity-based strategy, while in Experiment 3, they were instructed to use a rule-based (multi-feature) or a similarity-based strategy. In contrast, in Experiment 2, participants were required to self-discover categorization strategies. After learning, categorization was tested. The results of the three experiments were largely consistent, revealing distinct decision-making processes associated with each strategy. The results revealed that hypothesis testing and semantic processing, as reflected by the larger P300 and N400 components and increased activation in Wernicke's area, were critical for rule-based category learning, suggesting the role of an explicit system. In contrast, complex visual processing and the integration of multiple features, as indicated by a larger P1 component and the heightened activation in the frontopolar cortex, were critical for similarity-based category learning, suggesting the role of an implicit system. These distinct cognitive processes challenge single-system accounts suggesting a unified neural mechanism for both forms of category learning. Instead, our findings are consistent with the COVIS theory, which implies an explicit system for rule-based category learning and an implicit system for similarity-based category learning.