We designed a Markov decision task in which the subjects had to make 2 sequential choices (“LEFT” or “RIGHT”), one in each of two successive decision states in order to obtain a monetary outcome at the end state. Each state was signaled to the subject by a different fractal image (see Figure 1a for an example), which indicated to them that during the first 2 states they had the choice between left or right button press. The states were intersected by a variable temporal interval drawn from a randomly uniform distribution between 3 and 5 sec. The inter-trial interval was also sampled randomly from a uniform distribution between 5 and 7 sec. Upon each state the subjects had 1 sec to make the button press. If they failed to submit their choice in that time window the trials restarted from the beginning.
