Figure 3. Searchlight decoding analysis for representations of associative value.

(a) Value-based cross-decoding schematic. Patterns of fMRI activity within each searchlight sphere evoked by CS⁺ and CS of one category were used to train the classifier, which was then tested on CS⁺ and CS⁻ patterns of the other category. Post-conditioning accuracies were compared with pre-conditioning accuracies at the group level with paired t-tests. (b) Value representations emerged after conditioning in left vmPFC, which were significantly above chance (c) in the post-conditioning session. (d) Expected value was also represented in right OT, which also demonstrated above-chance classification in the post-conditioning data alone (e). Error bars represent between-subject s.e.m. *P<0.05, paired t-test against chance. Brain activations in b and d displayed at P<0.001 uncorrected (red) and P<0.05 small-volume corrected for family-wise error (yellow). Coordinates in b and d refer to voxel of peak decoding accuracy.