Fig. 1: Experimental task in fMRI and task space.

A fMRI task procedure. On each trial, the participants were presented with a compound stimulus. They had to judge the age of either a face or a house based on the information provided by the immediately previous trial. B Task design. The task started with a cue indicating a category (face or house) of stimuli to be judged (The pictures used as stimuli were courtesy of Dr. Dorothea Haemmerer and FACES datasets)⁵⁴. The participants continuously judged the age of the same category until an age switch occurred. In the example, the participants were instructed to judge the age of a face. After three trials the age of the face switched from young to old. At this point, the category to be judged in the following trial switched to house. According to these task rules, we classified all trials into three types, Enter (trials where a category switch occurred), Internal (trials where no switch occurred), and Exit (trials where an age switch occurred within same category) trials. C Task space. Each node represents a particular task state that reflects the current location in the task, composed of observable and unobservable task-relevant information. Each black circle in the node indicates the current trial information, and white circle in the node indicates the previous trial information. Each arrow represents a possible transition between states. Transition probabilities were equal, P = 0.5. Node colors indicated the three trial types.