Fig. 1: Experiment 1 task design.

A Each block of the reinforcement-learning task included 15 unique stimuli (shown in the gray box), which comprised five exemplars each drawn from three broader categories. For each stimulus set, three additional novel exemplars per sampled category and an additional category with eight novel stimuli were used in a test of subsequent memory (see panel D). B In the category-predictive condition, rewards on every trial were sampled from normal distributions centered on means determined by the stimulus categories. In the exemplar-predictive condition, rewards on every trial were determined by the individual exemplars. C On every trial of the reinforcement-learning task, participants chose whether to approach or avoid a stimulus. Participants won or lost points if they chose to approach the stimulus. While they did not win or lose any points if they chose to avoid, they saw counterfactual feedback showing how many points they would have won or lost had they approached. D Approximately 1 week after completing the reinforcement-learning task, participants completed a test of recognition memory in which they had to decide whether stimuli were old or new on a four-point confidence scale. The images in the figure are illustrative; actual task stimuli differed slightly. Image credit: iStock/GlobalP (https://www.istockphoto.com/portfolio/GlobalP), Life on White (https://www.lifeonwhite.com/).