The EEG feature known as the Reward Positivity (RewP) is elicited by reward receipt and appears to reflect sensitively and specifically positive prediction errors during reinforcement learning. Yet, the RewP also is modulated by state and trait affect, suggesting that it has a more complex computational role than simple reinforcement surprise. We conducted a series of experiments aimed to investigate underlying affect processing reflected in the RewP during a reinforcement learning task. In the first experiment (N = 25), we manipulated the type of rewards a person could win (simple points or hedonically-appraised pictures). Although there were no differences in the amplitudes of the RewP for different types of rewards, there was a significant correlation between the individual rating of liking for the images and RewP amplitude. In a second experiment (N = 25), we manipulated reinforcement rates (easy vs. hard) and affective picture content (liked vs. ambivalent) to examine the potential interaction of prediction error and liking on RewP amplitude. We again found a significant relationship between liking and RewP amplitude, however, only in the hard condition. These findings suggest that the RewP reflects cortical computations of reward surprise as well as hedonic liking, identifying it as a possible nexus where multidimensional value is computed.
展开▼
