highly significantp = 8 × 10 −40
The difference in fit of the 13 models was highly significant ( p = 8 × 10 −40 , ANOVA), but there was no significant difference in the fits of the five top models, all of which incorporated a perseverative learning mechanism: Perseveration/Reward-learning model (PR model, ref. 36 ); differential forgetting Q -learning 33 ; ideal observer with perseveration (see Methods); ideal-observer Bayesian filters with perseveration for either a generic probabilistic reversal-learning task (see Methods), or a generic drifting reward probability task 37 .