Note: (A) * p < .05, after Benjamini-Hochberg correction. (B) The blue dot indicates a significant positive regression coefficient (p < .05) with the vertical line in upper panels depicting the 95% confidence interval. The lower panels showed each specification model with selection of predictors and control variables. Forced_pv = predicted value in forced switching context, Forced_rv = the actual values in forced context; Natural_pv = predicted value in natural switching context, Natural_rv = the actual values in natural context; Voluntary_pv = predicted value in voluntary switching context, Voluntary_rv = the actual values in voluntary context.

To further explore whether this predictability generalized to the behavioral performance, we conducted SCA to reveal the effect of predicted value calculated by RR models for each context on RTs. The specification curve was shown in Figure 3B, and the bootstrapped results were shown in Table 1. The results showed significant positive relationship between the predicted values in the forced switching context and RTs of congruent trials. However, no significant relationships were detected for other specifications.

Table 1. Bootstrapped results of the specification curve analysis for congruent trials.