Striatal Tracking of Momentum-Biased RPEs

Analyses utilized a region of interest (ROI) approach. The procedure by which we defined our ventral striatum and left and right anterior insula ROIs is described in the Supplement. Parameter estimates representing the mean activity across all voxels in these ROIs for each participant were extracted for each contrast of interest using MarsBaR (http://marsbar.sourceforge.net/) using a standard threshold (p < .05). For the ventral striatum ROI, the contrast of interest was the parametric modulation of outcome-locked activation by trial-by-trial momentum-biased RPE values (i.e., the difference between unbiased and momentum-biased estimates). For the left and right anterior insula ROIs, the contrast of interest was the parametric modulation by trial-by-trial momentum, and Bonferroni correction was used with these 2 tests.

To test for associations with residual mood symptoms, manic and depressive symptoms were entered as predictors into a regression analysis in participants with bipolar disorder. Trait impulsivity was accounted for in this analysis because it had previously been shown to impact reward activations in this dataset (15). Log transformation was used for non-normal variables.