Figure 3.

Observed (left plot) and simulated group mean conditional error probabilities (all other plots). Error bars indicate ± 1 standard error of the mean. Note that set-loss and perseveration errors follow positive and negative feedback, respectively. AU attentional-updating model, MB-RL only model-based reinforcement-learning model, P-RL parallel reinforcement-learning model, wP-RL weighted parallel reinforcement-learning model, SLE set-loss error, PE perseveration error, Repetition demanded response repetition, Alternation demanded response alternation, Conditional error probability probability of an error given error type (perseveration vs. set-loss) and response demand (repetition vs. alternation).