Abstract
We navigate our visual environments via interactions between attention and visual working memory (VWM): When searching for a target, an early filter constrains attention to target-matching features and, in turn, attention filters VWM encoding to ensure relevant information is represented and used to guide behavior. When distracting (i.e., unexpected/salient) information appears, however, attention is captured, slowing visual search. Does distraction also disrupt ongoing attention/VWM interactions? Across two experiments we test the predictions of our Filter Disruption Theory: distraction disrupts the filters that typically prevent irrelevant information from being attended and stored in VWM. In E1 we measured filters tuned to object categories using fMRI: Participants viewed a 2x2 array of hybrid face/house images and performed a 1-back visual search task on one stimulus category (i.e., attend-faces or attend-houses blocks). The target image was defined by a solid white border, but on some trials, a salient distractor (a white dotted line) also appeared briefly around a nontarget image. On distractor-absent trials, we found the standard pattern of greater BOLD activation in fusiform face area (FFA) during attend-faces blocks and parahippocampal place area (PPA) during attend-houses blocks. However, on distractor-present trials, we observed a boost in activation for the nontarget category (e.g., increased FFA activity during attend-houses), suggesting a loss of control over both spatial attention and the attentional control filter tuned to object category, leading to incidental processing of the task-irrelevant category. In E2, we demonstrated behaviorally that the features of a salient distractor are also incidentally encoded into memory and can drive subsequent attention, even when those features are irrelevant to both the original and subsequent tasks. These data provide direct support for the Filter Disruption Theory, and more broadly suggest that distraction disrupts the interactions between attention and VWM.