Behavioral and Neural Predictors of Individual Differences in Concept Generalization

Abstract

Concept learning involves linking related pieces of information to a shared label, such as learning that furry creatures that bark are called “dogs.” People vary in how well they learn concepts and apply them to new situations (generalization). What factors drive these individual differences? In the present study, we tested whether stable aspects of intelligence or transient activations in the brain best predicted concept generalization abilities. To measure aspects of intelligence, subjects underwent an assessment that included measures of working memory, processing speed, perceptual reasoning, and verbal comprehension, which could be combined into an overall IQ. Subjects also completed a concept generalization task while undergoing functional MRI, allowing us to measure activations in brain regions that are part of the explicit rule-learning system (hippocampus, prefrontal cortex) or part of an implicit system that learns without awareness (caudate, posterior visual cortex). To elucidate the shared or dissociable roles of behavioral and neural predictors in concept generalization, we tested the relationship between accuracy in concept generalization and individual differences in measures of intelligence and activation in each brain region of interest. Behaviorally, we found that overall IQ, but not its subcomponents, predicted concept generalization abilities. Neurally, we found that only the activation in the hippocampus predicted concept generalization abilities. Finally, we found that IQ and hippocampal activation each predicted concept generalization independent of each other, indicating that they represent two separate processes that both contribute to generalization success. These results show dissociable contributions of behavioral and neural predictors of concept generalization, suggesting that both stable cognitive abilities and transient brain states influence the ability to learn new concepts.