Abstract
Statistical learning of spatial regularities during visual search leads to prioritization of target-rich locations. The resulting attentional bias may subsequently affect orienting and search behavior in similar tasks but its transfer to free viewing has not been demonstrated. We exploited a novel closed-loop paradigm where human observers searched for invisible target locations on a screen only guided by real-time auditory feedback conveying gaze-target distance. Unbeknownst to participants, location probability was biased towards one hemifield. Free viewing during rest, free image viewing, and spatial judgments were assessed before and after the search task. Search performance systematically improved and peaked in the biased hemifield, showing the unfolding of statistical learning. Importantly, the spatial bias transferred to both free-viewing conditions in terms of mean horizontal fixation position, while it did not transfer to spatial judgments. Exploratory results suggest that search performance was influenced by participants' viewing pattern, whereas transfer was modulated by pre-existing (natural) spatial biases. Our results demonstrate that task-based statistical learning transfers to ecological scenarios, paving the way for future research and clinical applications aimed at ameliorating pathological spatial biases.