Abstract
According to the framework of Kahneman et al. (1992), the representation of an object's perceptual features (e.g., color, shape) is indexed to a spatial location, forming an object file. When the object moves, the spatial index is updated to the new location, and the object property information comes to be associated with the new location. In previous work, we demonstrated that object file updating depends on the visual short-term memory (VSTM) system studied extensively over the last decade. However, we found that object feature information in VSTM was bound not only to the updated location but also to the original location, and the latter binding was more robust than the former. In the present study, we hypothesized that binding to the original location depends on configural coding in VSTM. Participants saw an array of boxes that were filled briefly with a set of colors. The empty boxes moved to new locations and were simultaneously translated so that the boxes formed the same spatial configuration but in new absolute locations. The boxes were again filled with colors. All were the same, or one was changed; the task was change detection. The test colors appeared either in the appropriate updated positions, in the original positions within the configuration, or in positions corresponding to neither of these (no correspondence). The advantage for the original position condition over the updated condition was maintained, as was the advantage for the updated condition over the no correspondence condition. The configuration benefit remained even in the presence of large-scale, static scene structure. Thus, there appear to be two forms of object-position binding in VSTM, a mechanism that updates object property information with motion (the object file system) and a more robust binding mechanism that is not sensitive to motion and codes object position relative to array configuration.
This research was supported by NIH grants R01 EY017356 and R03 MH65456.