The mammalian medial frontal cortex (MFC) is involved in reward-based decision making. In particular, in nonhuman primates this area constructs expectations about upcoming rewards, given an environmental state or a choice planned by the animal. At the same time, in both humans and nonhuman primates, the MFC computes the difference between such predictions and actual environmental outcomes (reward prediction errors). However, there is a paucity of evidence about the time course of MFC-related activity during reward prediction and prediction error in humans. Here we experimentally investigated this by recording the EEG during a reinforcement learning task. Our results support the hypothesis that human MFC codes for reward prediction during the cue period and for prediction error during the outcome period. Further, reward expectation (cue period) was positively correlated with prediction error (outcome period) in error trials but negatively in correct trials, consistent with updating of reward expectation by prediction error. This demonstrates in humans, like in nonhuman primates, a role of the MFC in the rapid updating of reward expectations through prediction errors.
Keywords: ACC; Medial frontal cortex; Prediction error; Reinforcement learning; Reward.