Figure 7.

(A) This boxplot shows the average probabilities of revisiting the second-stage state that was visited in the previous trial given low (1~3 points), medium (4~6 point), and high (7~9 points) previous rewards. This probability obviously changed depending on the previous outcome. (B) This boxplot shows the average sensitivity to the previous outcomes (SPO) for choices in model-based (MB) trials and model-free (MF) trials. This sensitivity is calculated as the average ratio of the probability of revisiting the same second-stage state after a high reward to the same probability after a low reward. This score was significantly higher in MF trials than that in MB trials. (C) This panel shows the relationship between the parameter w estimated by the LA-FD model and the difference in SPO between the MF and MB trials (ΔSPO). A negative relationship was found between these variables, showing that participants with a low w were more sensitive to the previous reward in the MF trials vs. the MB trials.