Plots of the data were created using the rfxplot toolbox for SPM5 (Glascher, 2009), which is capable of dividing a parametric modulator into different bins and estimating the average BOLD response for each bin. We extracted the data for the plots of percent signal change in Figures 3 and 4 using a cross-validation leave-one-out procedure: we re-estimated our second level analysis (repeated measures ANOVA, see above) 18 times always leaving out one subject. Starting at the peak voxels for the SPE signal in IPS and PFC and for the RPE in ventral striatum, we selected the nearest maximum in these cross-validation second level analyses. From that new voxel we extracted the data from the left out subject and sorted all trial into a 3 bins according the size of the SPE and defined by the 33rd, 66th, and 100th percentile of the SPE range. Then three new onset regressors containing all trials of each bin were created and estimated for each left-out subject. The parameter estimates of these onset regressors represent the average height of the BOLD response for all
