Fig. 3: Testing the robustness of subjective-value representations in the OFC.

We performed seven different General-Linear-Modeling (GLM) analyses to examine the robustness of subjective-value representations in the OFC. The GLMs were implemented for the high-gamma activity (80–150 Hz). a GLM-1. This is the original model where brain activity was regressed against the current and the previous subjective value. The high-gamma graph was identical to that shown in Fig. 2b. b GLM-2. This analysis was performed in two steps. In the first step, we regressed brain activity against only the current subjective value. In the second step, we used the residuals from the first step and regressed them against only the previous subjective value. c GLM-3. The model was identical to GLM-1 except that we only included trials where the subjects’ willingness-to-pay were not zero in the analysis. d GLM-4. The model was identical to GLM-1 except that the subjects’ response time (RT) in the current trial was added as a regressor to the model. e GLM-5. The model was identical to GLM-1 except that we added the subjective value of the option encountered two-trials back as a regressor. f GLM-6. The model was identical to GLM-1 except that we included the spectral power from the previous trial as a regressor. g Evaluating GLM-1 at different inter-trial intervals (ITIs). We sorted trials according to the preceding ITI (1 s, 1.5 s, or 2 s) and estimated GLM-1 separately for each possible ITI. Error bands represent ±1 standard error of the mean. Colored horizontal lines with the * symbol on top or beneath indicate the time points with p < 0.05 (familywise error corrected) using permutation test (one-tailed) with the threshold-free-cluster-enhancement (TFCE) statistic as the test statistic. Source data are provided as a Source Data file.