Figure 2. Logistic choice model, pupillometry, brain areas and reward rate decoding.

A. AIC model weights, showing probability in favor of the logistic choice model with a reward rate x ΔEV interaction term. Weights are the mean and standard error across for each subject, taken over each subject’s sessions. The weights were all greater than 0.5, indicating this model was favored for all subjects. B. The cumulative density choice function of the reward rate x value interaction regression coefficients for all session logistic models. C. Logistic choice curve from average model coefficients fitted across six subjects. Curves are shown for the low reward rate context (purple), high reward rate context (green) and the median reward rate (black line). Subjects show more optimal choices (steeper slope) in the high reward rate condition. D. Average reward gain defined as the normalized ratio of high reward rate to low reward rate contexts. Reward gain increases for the n+1 trial defined after the 3 trial reward rate window. Subjects gained more reward on these trials following high reward rates. E. Baseline corrected mean pupil across subjects, for both the low (purple) and high (green) reward rate conditions. Note that values are greater even before an offer appears. Shading: standard error. Black dots: time points with significant differences. F. MRI coronal slices showing the 4 different core reward regions that were analyzed. G. The decoding of reward rate across all brain areas (lines) and their significant points (dots) for each brain area. Each brain area is colored according to E. OFC: red; VS: blue; PCC: brown; pgACC: pink.