Abstract
Items around us constantly change, and visual working memory (VWM) must change its representations accordingly. One way of achieving this is updating, but sometimes changes are too large to be incorporated into existing representations. We suggest that processing such changes relies on a resetting of VWM, a process in which the old representations are removed from VWM, and the items are re-encoded as novel. We presented moving shapes that separated into parts in a shape change-detection task. While the joint movement was task-irrelevant, it supported the interpretation of the parts as a single object. Hence, individuating the parts (which was needed following their separation) was very challenging prior to separation. To determine the online status of VWM, we monitored the contralateral delay activity (CDA), an electrophysiological marker whose amplitude rises as more items are held in VWM. Following the separation, the CDA amplitude sharply dropped, indicating that VWM contents were removed, followed by the re-encoding of the parts as novel objects, which was reflected in a rise in the CDA amplitude. This effect was replicated with different stimuli, to illustrate its generality. Conversely, when individuation was possible already during the joint movement, the separation was followed by an updating process without resetting (i.e., CDA amplitude steadily rose, since more VWM-units were present, without a drop). We demonstrate that resetting does not depend on separation. Rather, we show that under certain conditions, resetting, as indicated by the CDA drop, is evident when an object is replaced by a different object. Finally, we provide behavioral evidence that during the resetting process, VWM is blind to salient changes in the items' shape. Overall, this work introduces the novel process of online resetting in VWM, its neural signature and its behavioral consequences.
Meeting abstract presented at VSS 2016