Abstract
Successful object recognition requires the binding of different visual properties (e.g., color, shape, location) into an integrated object-level representation. Theories of feature-integration propose that spatial attention is crucial for binding—but attention is rarely static, instead dynamically shifting and splitting across multiple goals and locations. What happens to object-feature binding when attention must shift or split across multiple objects with multiple features? While maintaining central fixation, participants were briefly presented with an array of four colored, oriented bars. The target bar was defined by a spatial pre-cue that remained stable (Hold) or dynamically shifted from one location to another (Shift) before array presentation. In another condition, attention was split across two simultaneously pre-cued locations (Split). Participants were instructed to reproduce both the color and orientation (i.e., joint continuous-report) of the target item; they also performed a location report to confirm the target location. Object-feature binding was measured by applying probabilistic models to the joint distribution of feature errors: Errors in recalling both features of the same object could be correlated (and thus bound together) or independent (and unbound). Across multiple experiments, splitting attention across multiple objects degraded object integrity, resulting in unbound feature errors. In contrast, rapid shifts of spatial attention maintained object integrity—even when those shifts were inadvertent. For example, on Hold trials, participants sometimes misreported the target location, indicating a lapse of spatial attention. Yet on these trials, participants reported both the color and orientation bound to that incorrectly-attended location. Moreover, converging data from a parallel line of fMRI experiments reveal that neural reconstructions reflect the features of the attended object on a trial-by-trial basis, even when attention selects the incorrect object. Together, these results emphasize the importance of a single focus of spatial attention in object-feature binding, whether that focus is stable or dynamically shifting across multiple locations.
Meeting abstract presented at VSS 2018