Fig. 3: Computational accounts of event-type-dependent action-selection strategies.

a The mechanisms of the support-stay, conflict-shift (SSCS) model. After each trial, the model performs (1) conflict bias update, (2) event-type-dependent action selection, (3) support bias update, and (4) setting a flag for the conflict bias update in the next trial. Each variable’s superscript indicates the trial index, e.g. e^t represents the event type (support or conflict event) the model experienced at the t^th trial. f^t represents the flag at the t^th trial. b The rat’s choice behavior after the reversal and representations of the SSCS model; Left for the choice behavior after reversal of rat (Black) and the SSCS model (Gray). The x-intercept and time constant of the exponential curve fitted to rat behavior and SSCS model behavior were compared using paired two-sample permutation tests and Spearman correlation analysis. Right for the modulation of decision variables, the support bias difference \({b}_{{{\rm{S}}}}^{+}-{b}_{{{\rm{S}}}}^{-}\) (Top) and the conflict bias difference \({b}_{{{\rm{C}}}}^{+}-{b}_{{{\rm{C}}}}^{-}\) (Bottom). The positive and negative actions are determined by the context after the reversal. The dotted lines indicate the trial when the context reversal occurs, the trial when the quantity of interest exceeds 0, and the time constant of the exponential curve fitted to the quantity of interest after the context reversal from the left. b shows data from n = 21 rats. All statistical tests were two-sided. Error bars indicate mean ± s.e.m. See Supplementary Table 2 for full statistical information. Source data are provided as a Source Data file. SSCS, support-stay, conflict-shift model.