Resting-state alpha scale-free dynamics predict individual differences in the balance between reward-learned and preference-based values during free choice

Values learned through reward can continue to influence choices guided by one's own preferences, such as which painting to hang, to an extent that varies across individuals. Because such preferences build up gradually across choices, their expression may depend on how well internally generated states are sustained over time, a property indexed by long-range temporal correlations (LRTC) in intrinsic brain activity. We tested whether resting-state alpha-band LRTC predicts individual differences in the balance between reward-learned and preference-based values. Eighty-one participants learned through probabilistic feedback which of four abstract shapes tended to be rewarded, and then judged their preferences among these shapes and 11 new shapes without feedback. Alpha LRTC was estimated by detrended fluctuation analysis of resting-state electroencephalography recorded before, between, and after the tasks. A trial-level mixed-effects model showed that reward history influenced preference judgments and that this influence decreased with resting-state alpha LRTC: participants with weak LRTC chose high-reward shapes over new shapes on 75% of trials, whereas those with strong LRTC were close to indifference (54%). The relationship was distributed across the scalp, stable across recordings, and not explained by spectral power or learning performance. Resting-state alpha LRTC is thus related to how strongly reward-learned values are carried into preference-based choice. Given the view that stronger LRTC reflects a longer window over which internally generated states are integrated, the observed relationship could arise because the window is biased toward values built up through one's own choices, or because it attenuates values learned under a criterion that no longer applies; which account holds remains to be determined.