Abstract
The way that we select and process visual information is shaped by our behavioral goals: attentional filters ensure goal-relevant information is selected, and relevant information is encoded into visual working memory (VWM). According to the Filter Disruption Theory (Dube & Golomb, 2021), salient visual distraction disrupts these attentional filters. What else might disrupt attentional filtering? Saccades have been shown in some cases to have similar consequences for perception as distraction (i.e., inducing feature-binding errors). Here we evaluate whether an attentional filter for VWM persists across saccades, or if saccades disrupt attentional filtering, requiring rapid filter re-instatement post-saccade. On each trial, participants saw a target-shape cue (either a circle or a square). Shortly after, they viewed an array of four colored shapes—either two circles and two squares (2 targets/2 non-targets), or all four of the target shape—and were told to remember the target-shape colors. One target-shape item was then probed, and participants indicated its color via a continuous color wheel. On some trials (saccade trials; E1: intermixed, E2: blocked), participants were prompted to make a saccade before the onset of the memory array, and the memory array appeared after either a short (50 ms) or long (400 ms) post-saccade delay. On no-saccade trials we observed effective attentional filtering: performance was better (i.e., less color report error) when only two items were relevant relative to when all four were relevant. Critically, on saccade trials, filtering was disrupted in the short delay condition but re-emerged following a long post-saccade delay. These results suggest that attentional filtering for VWM is disrupted by saccades but can be quickly re-instated. We propose that the Filter Disruption Theory may apply more broadly, as a general principle where attentional interruptions due to salient distractors, saccadic eye movements, and perhaps more, temporarily disrupt filters regulating VWM encoding.