Updating objects in visual short-term memory

Abstract

Information in visual short-term memory is encoded as unified objects (Luck & Vogel, 1997). Does the updating of stored information also proceed in an object-based manner? If so, any effect of updating one feature of an object would spread to its non-updated feature. In the first experiment, participants were briefly presented with three objects with unique colors and orientations. After a delay, a color patch or white oriented bar appeared in the same location as one of the objects. Participants were instructed to update, that is selectively modify, their memory of either the color or orientation of the corresponding object. Finally, either the updated object or one of the two remaining non-updated objects appeared as a memory probe. The probed objects appeared either accurately or with a change applied to either the updated feature or the non-updated feature. When updated objects were probed, change detection accuracy for updated features (81%) was superior compared to non-updated features (57%), t(10)=3.2, pt(10)=0.6, ns. Two control experiments suggested that this did not result from effects of visual priming or a lack of object-based storage. These results show that features in memory can be selectively processed without affecting memory for other features of the same object. A future control experiment will examine whether this means the effect of updating one feature of an object did not spread to its non-updated features, or that memory for the updated feature was separate from the object. As viewed from object file theory (Kahneman, Treisman & Gibbs, 1992), these results could indicate that updating an object file may be done in the absence of a reviewing process that would integrate the old and new states of the object.