Fig. 1: Task design and behavior.

a Subjects chose from menus of 4 or 7 masked offers (4 shown here). Fixating a mask (white diamond) caused it to disappear, revealing a reward cue (red bar). The filled area of the reward cue indicated juice magnitude. The subjects could freely search through the offers in any order. Subjects could accept offers by continuing to fixate them or else could reject an offer and continue to search by saccading away. b Offer values were uniformly distributed between 0 and a maximum juice value. c Recording sites in vmPFC (area 14). d Probability of choosing an offer as a function of offer value. Grayline = logistic fit. Each dot represents an offer value. e Distribution of the number of offers viewed per trial. Cyan line = prediction from an optimal compare-to-threshold model. Black line = best fit compare-to-threshold model (maximum likelihood). f Probability of returning to a previously seen offer. Each dot represents one session. g Expected value of trials which did (“ret.” = return) or did not (“seq.” = sequential) contain a return to a previous offer, subject J: mean difference = 0.12, 95% CI = [0.11, 0.15], t(88) = 12.01, p < 0.001, two-sample t-test; subject T: mean difference = 0.22, 95% CI = [0.20, 0.24], t(80) = 20.38, p < 0.001, two-sample t-test. Each dot represents one session. h Average value of offers which were returned to (“ret.” = return) versus those which were seen only once (“seq.” = sequential), subject J: mean difference = −0.19, 95% CI = [−0.21, −0.17], t(88) = −15.92, p < 0.001, two-sample t-test; subject T: mean difference = −0.05, 95% CI = [−0.07, −0.02], t(80) = −4.06, p < 0.001, two-sample t-test. Each dot represents one session. i Probability of a given number of returns in a trial, in trials where multiple unique offers were revealed (≥3). Each dot represents one session. j Probability of the return in multiple offer trials as a function of the viewing time at which an offer was finally accepted (at position 0). Each dot represents one session. k Probability of choosing the best option as a function of value of the best offer revealed within a trial (left) or as a function of the difference in value between the best and the 2nd best offer (right), the two collinear factors influencing overall difficulty of the trial, multiple offer trials, see also Supplementary Fig. S2A. i The same as (k) but for returning to a previously seen option, see also Supplementary Fig. S2B. Lines = least squares fit. f–l Error bars indicate ± standard error of the mean across sessions (SEM), subject J: n sessions = 45, subject T: n sessions = 41, *p < 0.001.