Surface construal and the mental representation of scenes

What distinguishes scenes from nonscenes? Photographs of objects on both naturalistic and blank backgrounds yielded boundary extension (BE: memory for unseen spatial expanse outside the picture’s boundaries). However, line-drawn objects on blank backgrounds did not (Experiment 1). Perhaps the blank background was construed as depicting a real-world surface in the photograph condition but was construed as depicting nothing in the line-drawn condition. To change background construal, the authors used objects cut out of photographs; these were placed on blank backgrounds while viewers watched (Experiments 2 and 3). BE was eliminated. The authors propose that amodal continuation is a funda-mental aspect of scene perception. However, not all pictures are scenes—only pictures construed as depicting a truncated view of a continuous world. In our interaction with the visual world, most of the information we process comes from complex visual scenes. These scenes typically include multiple objects and complex backgrounds. Our environment extends all around us, but physiological constraints prevent us from ever seeing an entire scene at once. Head move-ments and eye movements provide us with successive views of a continuous world. Yet, in spite of the complexity inherent in scenes, and the spatial limitations on how much of the world we can see at one time, our understanding of the visual environment seems effortless. In this study we focused on the viewer’s ability to remember how much of a scene was contained in a single view. We tested the hypothesis that when faced with a partial view of a continuous world, amodal continuation of scene layout plays a critical role in scene perception. Memory for spatial expanse was examined under conditions in which stimuli either were or were not understood to be scenes. Intraub and her colleagues showed that when remembering a photograph of a scene, viewers tended to remember having seen a greater spatial expanse than was actually shown—they referred to this error as boundary extension (Intraub & Richardson, 1989; se