Bayesian methods for addressing long-standing problems in associative learning: The case of PREE