Many studies of object and scene recognition focus on the ability to recognize different views of the same display. The problem of recognition across views is fundamental to understanding how we represent our visual world. Do we recognize objects by comparing our current retinal image to a stored set of previously seen images? Or, do we store a generalized description of an object that is independent of the particular views we've already seen? The first, a view-based approach, argues that object recognition should depend on the particular views we've seen before and their similarity to our current view. The second, a view-independent approach, suggests that recognition should be independent of the particular views we've seen before because the representation is equally appropriate for all views. Over the past 30 years, many studies have examined this distinction by showing views of objects on a computer display at study and testing memory for rotated versions of those objects at test. Often these experiments on view changes attempt to generalize their findings to real-world object recognition and real-world view changes. Yet, in the real world, most view changes are not caused by objects rotating in front of a stationary observer. More often, we cause view changes by actively moving our head or our whole body relative to stationary objects. Previous studies assume that the view changes created by rotating displays on a computer monitor are comparable to those produced in the real world, because the change to the retinal projection caused an observer movement can be mimicked by a display rotation. Although the retinal image changes may be equivalent, extra-retinal information for the view change is not.

      In a series of recent studies, we showed that the detection of changes to a spatial array of real objects is differentially affected by display rotations and observer movements (Simons & Wang, 1997; Wang & Simons, 1998). When the array rotates relative to a stationary observer, change detection accuracy is reduced. However, when observers move relative to a stationary array (with the same view change), performance is less affected by the view change. This pattern of results is unaffected by such factors as the visibility of background information, the amount of information specifying the view change, and the degree of active control over the view change. However, performance is disrupted when observers are disoriented as they move relative to the stationary array. These findings suggest an important role for extra-retinal information in updating spatial representations across observer movements. Furthermore, they suggest that representations of spatial arrays are largely view-dependent in this task, although recognition is largely view-independent when observers have sufficient extra-retinal information to update their representation as they move relative to the array.

      We have also explored the role of extra-retinal information for spatial updating using a standard new/old recognition task with single novel objects (rather than spatial layouts). Again, we find that object recognition is better following an obsever movement than following a display rotation (Simons, Wang, & Roddenberry, 2002). This finding provides a direct challenge to existing models of object recogntion, none of which explicitly account for observer movement.