Abstract
Human beings are extremely proficient at extracting statistical regularities from sensory inputs to enhance a variety of basic and higher-order cognitive functions, such as attention and working memory. It remains unknown, however, whether statistical learning can facilitate the selection of stimuli into consciousness (i.e., conscious access). To answer this question, we manipulated the probability of target location (left or right side of fixation, Experiment 1) and feature (upright or inverted triangle, Experiment 2), and used the breaking continuous flash (b-CFS) paradigm to measure the speed of conscious access. We find that targets that were initially rendered invisible by CFS masks were reported faster when they appeared at the probable (versus improbable) location or with the probable (versus improbable) feature. While these results suggest that statistical learning may influence conscious access, the differential response times in b-CFS may be attributed to unconscious processing (i.e., differences in conscious access) or conscious processing (i.e., differences emerging after conscious access). We therefore adapted a recently developed detection-discrimination dissociation (DDD) paradigm that allows for excluding effects (of statistical learning) emerging after conscious access. Using this paradigm, we show that [1] probable targets (e.g., upright triangles) are localized better than improbable targets (e.g., downwards triangles), while [2] observers are unable to discriminate between the probable and improbable targets (i.e., downwards vs. upright triangles) in a forced-choice discrimination task. In this way, we established that statistical learning affects the processing of stimuli prior to conscious access. We conclude that the visual system can utilize statistical regularities of sensory inputs to facilitate conscious access of probable events.