Figure 1.

Two-step decision-making task. (a) Tree-shaped state mapping structure and an example state transition of a trial resulting from first choice (R: right) and second choice (R), and resulting in obtaining the 0 red coin. If the trial was in the flexible-goal condition, the subject would have received 0 shocks. If the trial was a specific-goal condition with a red collection box, the subject would also receive 0 shocks, but if the box was one of the other colours (blue or yellow), the subject would have received a default of 4 shocks. (b) Time sequence of a trial in the specific-goal condition. The initial, intermediate and outcome states are represented as a fractal image. With the decision states, the current goal is shown as a coloured box (e.g. yellow). After two consecutive choices, an outcome state is displayed with electrical shocks delivered on the back of the subject’s non-dominant hand. (c) Diagram of the computational model with arbitration between model-based and model-free learning systems.