(A) Histogram of the difference in AIC between the full model, in which all DDM parameters were allowed to vary by reward context and microstimulation status, and a reduced model, in which all DDM parameters were allowed to vary by reward context but not microstimulation status. Negative ΔAIC implies that the full model is better. The red arrow indicates the criterion we used and corresponds to the gap in the histogram. (B) Map showing the best DDM variant (lowest AIC, black bar) for sessions with significant microstimulation effects (n = 39 sessions to the left of the red arrow in A). In the reduced models, parameters associated with collapsing bounds (β_alpha and β_d, “NoCollapse”), total bound height (a, “NoA”), drift rate scalor (k, “NoK”), bias in drift rate (me, “NoME”), relative bound height (z, “NoZ”), and non-decision times (t_contra and t_ipsi, “NoT0”) were allowed to vary based on reward context, but not microstimulation status. (C) Histograms of differences in AIC between the full model and reduced models for the 39 sessions to the left of the red arrow in A. Mean ΔAIC values were negative for reduced models (t-test, p<0.001 for all). (D) Scatter plots of changes in DDM parameters induced by electrical microstimulation in the Ipsi-LR blocks (abscissa and top histograms) or in the Contra-LR blocks (ordinate and right histograms). Solid lines in histograms: mean values across sessions. Red lines, t-test p<0.05. Labels (a, b) correspond to the example sessions in Figure 3A and B, respectively.