The role of statistical learning in attentional guidance during search through naturalistic scenes

Regularities within the environment have been shown to guide attention and facilitate efficient search; however, previous work utilized real-world objects that participants were already familiar with, making it difficult to discern the relative contribution of statistical learning and semantic knowledge to this effect. In Experiment 1, we associated novel objects (with no apparent semantic meaning) with regions in kitchen scenes (floor, countertop, wall) but placed them in the associated region of the scene on 75% of trials; we found that participants were faster on trials when the target appeared in its high-probability location, consistent with learning-dependent guidance. In Experiment 2, we looked to see if participants could learn multiple context-dependent object-location associations. We found that when the associated location of an object varied by the context provided by the category of scene, participants were unable to learn such associations. In Experiment 3, further showed that the learning in one context containing object-location associations similarly guided attention in a second context that did not contain the same contingency. Overall, our results suggest that learned object-location associations can guide attention in the absence of semantic knowledge concerning the objects, suggesting a role for statistical learning in scene-based guidance, but such learning does not present with the context specificity associated with semantic guidance. (PsycInfo Database Record (c) 2026 APA, all rights reserved).