Fig. 4: Dissociation of reward PE (R-Qr) and punishment PE (P-Qp) signals.

a Time course of regression estimates obtained from linear fit of BGA with PE modeled separately for the reward (blue) and punishment (red) conditions (PPE punishment prediction error, RPE reward prediction error). Horizontal bold lines indicate significant difference between conditions (blue: RPE > PPE; red: PPE > RPE; pc < 0.05). Shaded areas represent inter-sites SEM. b Time course of regression estimates obtained from a linear model including both outcome (solid lines) and expected value (dotted lines) components for both reward (R and Qr) and punishment (P and Qp) PE. c Regression estimates averaged over the 0.25–1 s time window (represented as shaded gray areas in panels b). Stars indicate significance (*p < 0.05, one-sample, two-tailed Student’s t test). Error-bars correspond to inter-sites SEM and dots correspond to individual recording sites. The sample size (n) used to derive statistics in all panels was: aINS: n = 83 sites; dlPFC: n = 74; vmPFC: n = 54; lOFC: n = 70.