Statistical learning and the efficiency of visual search

Target selection and distractor suppression can be facilitated through learning-dependent processes. Targets are found faster when appearing in a high-probability color or location and distractors produce less interference when appearing in a high-probability color or location. Such effects of statistical learning on attention have been demonstrated almost exclusively in the context of paradigms that do not systematically vary the number of items in the display, precluding assessment of learning-dependent changes in the efficiency of search through successive items, or search guidance. In the present study, I directly measured the influence of statistical learning on search guidance. In Experiment 1, target color and location were manipulated orthogonally, such that the target could appear in a high-probability color, a high-probability location, both, or neither. Location-based and feature-based statistical learning were found to produce additive benefits for search guidance, each reducing search slope, while only location-based learning reduced search intercept. In Experiment 2, the target was never rendered in a particular color, which resulted in participants to some degree filtering items of this color during search, producing a benefit that scaled with the number of items of the critical color presented in the display (i.e., reducing effective set size). Together, the results of the present study highlight an important role for selection history in facilitating search guidance.