Fig. 4.

Estimated reinforcement learning diffusion decision model (RLDDM) parameters. Left column: development of the latent learning variables (i.e., difference in learned Q values, trial-conflict, contextual value) throughout learning, as predicted by the RLDDM (by context, and across experiments). Right column: 95% Bayesian credible intervals of the estimated posterior distributions of the effects of the latent learning variables on the DDM parameters (respectively, drift rate, threshold, and non-decision time) at the experiment level