Working memory prioritization: Goal-driven attention, physical salience, and implicit learning

Abstract Items in working memory (WM) are prioritized if they are relevant to task goals, are physically salient, or have acquired importance from implicit learning. We propose that all forms of prioritization increase the likelihood of recall, but only goal-driven attention will affect the quality of those representations. In a delayed-estimation task with four colors, prioritization was manipulated via a predictive spatial cue (goal-driven attention), a non-predictive peripheral cue (physical salience), or implicit learning of a previously relevant target location. Probabilities of recalling the target (Ptarget) and memory precision were estimated using a Bayesian implementation of the mixture model. Strong evidence was observed that all forms of prioritization increased Ptarget, whereas physical salience and implicit learning had only weak or negligible effects on precision compared to goal-driven attention. We propose that generating and maintaining high-resolution memories is an effortful process that will primarily be invoked when participants voluntarily prioritize memory items.