Visual statistical learning can drive object-based attentional selection

Recent work on statistical learning has demonstrated that environmental regularities can influence aspects of perception, such as familiarity judgments. Here, we ask if statistical co-occurrences accumulated from visual statistical learning could form objects that serve as the units of attention (i.e., object-based attention). Experiment 1 demonstrated that, after observers first viewed pairs of shapes that co-occurred in particular spatial relationships, they were able to recognize the co-occurring pairs, and were faster to discriminate two targets when they appeared within a learned pair ("object") than when the targets appeared between learned pairs, demonstrating an equivalent of an object-based attention effect. Experiment 2 replicated the results of Experiment 1 using a different set of shape pairs, and revealed a negative association between the attention effect and familiarity judgments of the co-occurred pairs. Experiment 3 reports three control experiments that validated the task procedure and ruled out alternative accounts.