choice, second-step and transition on the current trial. As the subjects knowledge of the transition/reward probabilities is ambiguous in the period following block transitions where they change, these predictors were coded 0 in the 20 trials following such changes, and ± 0.5 at other times. These analyses included only sessions where we had at least 40 trials in at least two different states of the transition (Figure 5A) or reward (Figure 5B) probabilities.
