Abstract
The leaning tower illusion and similar illusions illustrate that the orientation of objects in three-dimensional space is affected when viewing multiple pictures. We directly measured the accuracy of perception when viewing multiple pictures in order to better understand this effect. Across several experiments, 10-20 participants viewed two pictures for 3 seconds. One picture was a sidewalk receding into the distance to the left, right, or straight ahead. The second picture was either another sidewalk, a non-sidewalk nature scene, or nothing. After the stimuli disappeared, participants adjusted the orientation of an on-screen line to match the orientation of a sidewalk. After completing this task, participants viewed all possible combinations of sidewalk pictures twice, judging how parallel the sidewalks appeared to be on a 1 to 5 scale. When judging sidewalk orientation, perception was most affected when the two sidewalks presented were perceived as being parallel to each other. In these cases, the sidewalks were judged to have orientations more similar to each other than when each was paired with a nature scene or nothing. The maximum degree of misperception was approximately twice as large as the misperception in the leaning tower illusion. Additional experiments show that this effect fully survives inversion and partially survives alternating presentation of the two pictures. However, cueing the participants during the stimulus presentation about which of the two sidewalks will be judged eliminated the effect. We conclude that when viewing multiple pictures containing objects with perspective convergence, those objects’ orientations are misperceived, particularly objects that are perceived as parallel or nearly parallel. This misperception occurs over a wide range of conditions, but is eliminated when attention can be focused on only one of the pictures. These data suggest that the perception of elements in a picture are adjusted based on visual integration across pictures.