Beta-gamma oscillation reveals learning from unexpected reward in learners versus non-learners.

Beta-gamma oscillation has been demonstrated to be sensitive to unexpected reward feedback. However, it remains unclear whether beta-gamma activity manifests individual differences in reinforcement learning processing. Given that individuals differ largely in reinforcement learning tasks, we adapted the Friedland task and split subjects into two groups: learners who learned to choose the optimal card after training and non-learners who did not learn well. We used recorded electroencephalography signals to test the difference in brain activity between the learners and non-learners groups when participants conducted a time estimation task and received win and loss feedback in expected and unexpected conditions. The results revealed that the unexpected condition elicited a larger reward positivity amplitude than did the expected condition, but only in the learners group. No significant difference was found between the two groups for the expectancy effect on frontal-midline theta. The current results thus demonstrate that for learners, unexpected win feedback elicits a larger beta-gamma oscillation than expected win feedback, while this was not the case for non-learners. These results indicate that beta-gamma oscillation may reflect effective learning from positive reward prediction error, a finding that adds to the existing theories on learning processes.