Fig. 7.

Left: The serial belief reaction time task. In this design, on each trial, participants had to report how many green flowers were received among green clovers according to one of two smurfs (i.e., Papa Smurf or Smurfette). On true trials, the smurf was turned to the screen and participants should report what the smurf could observe (the number of flowers); on false trials, the smurf was turned away from the screen and participants should report what they believed that the smurf saw last. Participants implicitly learned the fixed (but unknown) sequences embedded in the task, in particular the sequence of true and false beliefs. Implicit learning was attested by interspersing blocks with random instead of fixed standard sequences (e.g., random true–false beliefs), and observing significantly increasing response times as a consequence. Right: In a follow-up fMRI study [135], a parallel increasing pattern of posterior cerebellar activation during true–false belief randomization was observed (MNI coordinates − 36, − 64, − 42; n = 18)