Abstract
It is well-established that the brain can exploit statistical regularities to selectively guide attention towards task-relevant information in the external world, even without explicit awareness of these regularities. This study examined whether such ‘statistical learning’ can also guide the selection of previously encoded information held in working memory, using an online experiment. To this end, 69 participants performed a visual working-memory task. In each trial, two tilted colored bars were shown, followed by a sound (low or high in pitch). After a short delay, participants had to either reproduce the orientation of one of the bars (5/6 of trials) or report the pitch of the sound (1/6 of trials). Critically, unbeknownst to the participant, the sound’s pitch functioned as an implicit retrocue, predicting which bar was likely to be the target in that trial. A Bayesian hierarchical model showed tentative evidence that participants exploited the predictive cue, as they were slightly faster in reporting validly cued targets over invalidly cued targets, while showing no explicit awareness of the sound-to-memory-item regularity. However, this effect failed to reach significance in an equivalent frequentist model. In a control experiment (n=31) where participants were informed of the contingency beforehand, validly cued targets were reported significantly faster and more precise than invalidly cued targets, and to a much greater extent. These findings suggest that guidance of internal attention by contextual cues may be possible to some extent in the absence of explicit awareness and may be sensitive to implicit statistical learning. We are currently conducting a follow-up experiment (n=65) using implicit temporal expectations instead of an implicit contextual cue to guide internal attention, to further characterize the boundary conditions of implicitly learning to prioritize information in working memory.