Scene-based and viewer-centered representations for comparing shapes

Abstract

Five studies examined the use of scene-based and viewer-centered (i.e., head- or retina-based) representations for comparing shapes at different orientations. Two objects that resembled either the line-drawings used by Metzler and Shepard (1974), or helices, were placed far apart on a table top, so that the lines of sight from the viewer to each object were 90 or 150 degrees apart. Observers had to decide whether the objects' shapes were identical or mirror images, and were instructed to physically rotate one object to an orientation that allowed them to make a decision. They often rotated the object until it had the same relationship to the table top (and room) as the other object (thereby achieving scene-based alignment), even though this produced quite different retinal images of the two objects. Responses regressed up to a third of the way toward viewer-centered alignment as principal surfaces of an object were less aligned with significant directions in the scene. Similar patterns of alignments were observed for pairs of objects with very different surface topology. When subjects were instructed to turn one object so that it was perfectly aligned with the other with respect to the scene, responses also regressed toward the viewer-centered alignment, in this case by as much as 15 degrees (as the unmoved standard object's principal surfaces were less aligned with the scene's significant directions). Overall, these results suggest that when comparing shapes in this kind of task people rely more on scene-based representations than on viewer-centered representations.