Figure 1. Scheme of learning by prediction error. Red: a prediction error exists when the reward differs from its prediction. Blue: no error exists when the outcome matches the prediction, and the behavior remains unchanged.