Figure 7.

Repetition suppression for value in the anterior cingulate cortex. A, Repetition suppression analysis logic. We hypothesize a population of neurons tuned to value where the different neurons have overlapping tuning curves spanning the range of values presented. Black arrows denote stimulus value for that trial. If consecutive trials activate nonoverlapping populations of neurons, evoked responses for each stimulus are similarly high on each trial (top panel in orange). However, repeated presentation of the same stimulus produces repeated activation of the same neurons on consecutive trials, leading to a reduction in the neural response (bottom panel in blue). Summation over all neurons in the population (as in the BOLD signal measured in fMRI), leads to higher activity when consecutive stimuli activate unique subsets of neurons (top panel) than when consecutively activated populations overlap (bottom panel). Predicted BOLD activity is thus proportional to the absolute difference in value between consecutive trials. B, Evidence for multiple forms of value coding in the cingulate. We examined cingulate representations of repetition suppression to integrated value (change in value from trial n − 1 to trial n, ΔIntegratedValue; green), and monotonic encoding of integrated value (a standard parametric modulator approach; red). Voxels sensitive to repetition suppression were more posterior, with monotonic encoding stronger in anterior voxels. C, The ACC region identified in the conjunction analysis (Fig. 5A) also shows repetition suppression in the integrated value. We extracted mean parameter estimates for ΔIntegratedValue and for integrated value from the voxels identified in the conjunction analysis. Both were positive on average (ΔIntegratedValue: t₍₂₄₎ = 2.48, p = 0.020; Integrated Value: t = 3.26, p = 0.0034). Error bars represent the SEM across participants.