The experiment proceeded in two separate scanning sessions of 80 trials each. During the first session, all decisions were predetermined and the subjects simply had to register them. Subjects also received no rewards at the outcome states during this part of the experiment (see Figure 1b). Taken together all trials in this first session reflected the underlying transition probabilities exactly, but they were presented in a randomized order. Subsequently during a break, subjects were exposed to the reward contingencies (see Supplemental Experimental Procedures). Finally, in the second scanning session, subjects made their own choices and were rewarded at the outcome states.
