Viewers remember having seen beyond the edges of a picture (boundary extension). Recent research indicates that this unconscious extrapolation is limited to pictures of scenes — suggesting a possible role in the integration of successive views. What implications might this have for the representation of occluded objects in scenes? Occlusion by a view-boundary may elicit extrapolation, whereas occlusion by another object in the scene may not. Alternatively, Kanizsa's “occlusion illusion” raises the possibility that extrapolation of an object may occur whenever occlusion is present.
Two versions of 12 multi-object scenes (digital photographs) were created in which a target object was occluded either by a view-boundary or by another object in the scene. One version of each photo was shown for 10 s, followed by a visual noise mask ( sec, 1sec, or 10 sec), and then it reappeared. Using a mouse, the viewer positioned the target object within the digital photograph (multiple graphics layers allowed viewers to move the objects behind other objects or behind the view-boundaries). Viewers remembered seeing more of the target object than before when it had been occluded by the view-boundary, but not when occluded by another object in the scene. Might this be attributed to the lack of visual information outside the view? In another experiment the view-boundary was composed of occluding objects. Thus the local information at the point of occlusion was controlled. Again, extrapolation occurred only at the view-boundary. Not all occluding borders are equal — extrapolation beyond a view-boundary may serve an adaptive purpose in understanding and integrating successive views.