Abstract
A number of studies reported that the visual space is systematically distorted, although the nature and magnitude of the distortions varied across studies. One of the well known tests was performed by Foley (1972) in his triangle experiment. He asked a binocular subject to adjust the positions of two lights, A and B, placed at the eye level, so that the triangle formed by these two points and the subject herself (represented as point O) was an isosceles right triangle (OB=AB, angle OBA=90 deg). The actual ratio AB/OB of a perceptually isosceles triangle was 0.53. At the same time, the angle AOB was perceived accurately as being 30 deg (instead of 45 deg), which implied that the visual space is a non-Euclidean space with negative curvature. It is important to point out that the viewing conditions in Foley’s experiment were very impoverished. Since 3D vision is a difficult inverse problem whose solution critically depends on the operation of a priori constraints, perception cannot be veridical unless natural viewing conditions are used so that all natural constraints are effective. We identified several constraints which allowed our model to recover 3D space veridically (gravity, horizontal floor, subject’s height). The present study replicated Foley’s experiment in a variety of conditions from very impoverished to natural. Three independent variables were manipulated: lighting of the room (dark vs. bright), the level at which the targets were placed (eye level vs. floor), and the number of targets (2, like in Foley’s experiment vs. 3). The subject’s head was free to move. For 3 targets in a bright room, the perception was close to veridical (the visual space is Euclidean). With a dark room, or with 2 targets, perception was not veridical and the distortion of the visual space was similar to that reported by Foley.
Meeting abstract presented at VSS 2012