Abstract
Whether looking for your keys, or your friend’s face at a party – you’ll be actively sensing your surroundings while concurrently holding relevant visual information in mind. Temporary maintenance of visual information, often referred to as Visual Working Memory (VWM), is primarily studied using steadily fixating participants and gray screens, ignoring the everyday reality of ongoing visual exploration via systematic eye movements. Eye movements present uniquely patterned dynamic input to the visual system, against which VWM contents should be robust if they are to be held stably in mind. How is VWM behavior impacted by distractors that closely match visual inputs occurring during natural viewing behavior? We had participants freely view 334 images of natural scenes for 8s each, and created a moment-by-moment “replay” of participants’ central-most visual input by circularly cropping (radius=7º) around the recorded eye position over time. During a subsequent VWM task, participants remembered randomly oriented gratings for 5s before recall via method-of-adjustment, maintaining fixation throughout. The delay period consisted of either a blank screen, or a 4s “replay” of the cropped images presented at fixation in either intact or shuffled order. Contrary to the intuition that VWM should be particularly robust to distraction from inputs matching naturalistic viewing behavior, recall error was larger on trials with intact replay during the delay, compared to trials with shuffled replay or without distractors. This effect was not explained by differences in ocular behavior. In fact, both pupil size and the number of (micro-) saccades decreased with distractors during the delay, and more strongly so for shuffled compared to intact replay. What’s more, trials with larger recall errors were associated with a larger number of saccades. In sum, ocular behavior during VWM can be informative of the type of visual input presented during the delay, as well as recall error.