From Sampling to Stopping: The P300 ERP component and beta power contribute to reward-related decision commitments

Optimal stopping problems provide a framework for studying decision-making under uncertainty, balancing the trade-off between information sampling and decision commitment. We investigated deviations from normative strategies in human decision-making and examined the neural mechanisms underlying decision commitment versus sampling using electroencephalography (EEG). Forty participants viewed sequences of beads drawn from fictitious urns and attempted to infer the majority bead colour in each urn, while their EEG activity was recorded. After viewing each bead, participants could choose to sample more by drawing another bead (draw choice) or to stop sampling and infer the contents of the urn (urn choice). A Bayesian ideal observer model and parametrised models were used to predict participant behaviour. Participants undersampled relative to the ideal observer, particularly in the more uncertain condition where the proportion of bead colours was close to chance (60/40 or 0.6), with the model better capturing behaviour in the easier 80/20 (0.8) condition. P300 amplitudes showed larger responses for urn choices and a gradient of increasing amplitude as draws approached commitment. Larger frontal ERP responses were also observed under higher uncertainty (0.6 condition). Beta oscillatory activity was stronger for urn choices in the high uncertainty condition, with fast beta (20–30 Hz) activity driving final commitment to a decision. Beta power was further predicted by model-derived action values. By integrating behavioural, computational, and neurophysiological data, this study advances our understanding of active information sampling versus decision commitment in reward-related probabilistic environments.