Statistical learning of target location guides attention proactively

Our perceptual system is highly sensitive to statistical regularities in our environment. In particular, we respond faster to targets that appear in frequently attended locations—a phenomenon known as target-location probability learning (LPL). Is attention proactively aligned with the high-probability target location, or reactively allocated to that location when the learning context is detected? The studies that addressed this question tested whether a spatial attentional bias learned in one task transfers to another. However, they yielded conflicting findings and were open to alternative accounts. We reexamined whether LPL-guided attention is proactively allocated to the high-probability target location while addressing these previous studies’ potential caveats, in two experiments. During learning, the search target appeared more often at one location than elsewhere, and during extinction, all search target locations were equiprobable. In both learning and extinction, letter-probe trials were interspersed among the search trials. We found that LPL acquired during search transferred to the letter-probe task during both learning and extinction. Importantly, during extinction, participants continued to prioritize the previously high-probability location on both search and letter-probe trials, even when they were informed after the learning phase that the bias would be discontinued and were asked to start their search at the location indicated by an arrow precue. We conclude that LPL guides attention proactively and inflexibly.