Pre-report SEEG high-gamma activity weakly predicted confidence and added little to behavior in five patients

We tested whether 70–150 Hz SEEG high-gamma activity (HGA), measured before a confidence scale appeared, predicted later confidence ratings and added information beyond behavior. The analysis used archived epochs and behavior tables from five patients (180–300 trials each) performing a random-dot motion decision task. A prewritten plan specified patient-wise ridge regression on log HGA from four 150 ms windows around the response, five-fold cross-validation repeated five times, and Spearman correlation of out-of-fold predictions with confidence. HGA predictions were positive in four of five patients (median rho = 0.067); three patients had permutation p < 0.05, but the exact group sign-flip test was not significant (p = 0.125). Behavior alone had median rho = 0.374. After control for behavior, median partial rho for HGA was −0.002 (p = 0.375), and adding HGA increased rho in one of five patients. In correct trials, median rho was 0.039 (p = 0.19). HGA predictions correlated negatively with response time in three patients, consistent with much of the HGA read-out reflecting timing or motor-vigor information. The result is limited to five patients, whose clinical electrode coverage was not targeted to metacognition areas; the null group result does not establish absence of an effect. The archived work supports at most a weak confidence read-out from pre-report HGA, with little added information beyond behavior in this sample.

confidence · high-gamma activity · SEEG · response time · ridge regression · brain-computer interface · decision-making