Integration of contextual reward motivation with category representations

This study examined how representations of categories and contextually signaled reward availability interact to support categorization that maximizes potential reward. Participants performed a categorization task during fMRI scanning in which they viewed a cue indicating the probability of receiving reward, then categorized a stimulus, and then (if correct) received a reward according to the cued probability. We combined representational similarity analysis (RSA) and univariate measures to identify areas important for representing categories and integrating the categorization decision with reward probability. Both boundary distance error (BDE; difference between feedback following category response and actual category membership scaled by the relative distance in stimulus space) and RSA for the interaction of within category variability and discrete category provided convergent evidence for category representations in a right frontoparietal-striatal network including the anterior IPS, inferior frontal gyrus, and posterior caudate. Category and reward probability converged in two locations. The posterior caudate was sensitive to both reward probability and category structure. There was both discrete category coding and an interaction between category and reward coding in areas of motor cortex underlying the hand responses used to indicate category membership. These results indicate two ways that reward availability is integrated with category representation. First, reward availability is represented within the posterior caudate which interacts within a corticostriatal loop with frontoparietal cortex regions representing categories. Second, reward availability is combined with category related response selection in motor cortical regions. Both mechanisms may support efficient selection and execution of categorization responses expected to result in reward.