Behavioral and model performance.

Average accuracy and RT across subjects (N = 34) as a function of option pairs in the learning phase (A) and option value differences (derived from the experimental reward probabilities) in the transfer phase (B) that indicated small (s), medium (m) or large (l) value differences between presented options. (C): Real and simulated choice accuracy as a function of run number in the learning phase, split by option pair. For all option pairs, simulated and real accuracy was very similar, with real EF accuracy being slightly underestimated by the model. (D): Group-level posterior distributions of the obtained parameter estimates for β, αGain and αLoss. (E): Model estimates of value beliefs for each option at the end of the learning phase. β/100 for visualization; error bars represent mean ± s.e.m.