Relationship of Midbrain and Dorsal Striatal Reward Responses to Noncontingent Learning in Experiment 1
(A) Midbrain responses to contingent rewards in a region consistent with the location of substantia nigra and ventral tegmental area correlate negatively with the degree to which subjects’ behavior was guided by contingent learning as opposed to either PROX or SoECh—the exact opposite pattern of what was observed in lOFC (see Figures 3C and 3D). Contingent reward responses at the peak location were strongly related to both PROX and SoECh; however, the correlation with PROX tended to be stronger than that with SoECh.
(B) A parametric contrast revealed that rewards elicited a stronger response the sooner they occurred following a choice (upper row). In areas associated with model-free learning such as putamen and associated motor cortical areas, this effect was strongly related to the extent subjects’ behavior was guided by PROX (bottom row).