Konovalov, A.; Hu, J.; Ruff, C. C.

Successful strategic behavior must be grounded in beliefs about the opponent and her intentions. While many potential models have been proposed to explain choices in such situations, the neural mechanisms that govern learning and choice in complex strategic contexts remain poorly understood. Here, we use a computational model that combines dynamic learning and choice mechanisms to explain both choices and response times of human participants engaged in a competitive strategic task. Using electroencephalography (EEG), we identify temporally structured stages of neural processing that support an evolving value-based decision process, corresponding to first- and second-order belief updates and evidence accumulation related to the comparison of action values. Gamma-band phase coupling between central and parietal EEG signals varied with individual winning rate, suggesting that strategic behavior involves coordinated information transfer across spatially remote areas. Together, our data characterize the temporally evolving neural dynamics of belief and valuation processes that underlie strategic choice and provide neural validation for assumptions embedded in computational models of this behavior.