Abstract
Psychophysical evidence indicates that, at viewing distances >60 cm, the perceived depth from binocular disparity significantly underestimates the true depth. One geometric consequence of this is that rigid objects should appear to change shape as they rotate in depth. To illustrate, consider a monitor such that 400 pixels subtend 10 cm along the vertical axis but only 8 cm along the horizontal axis. Suppose we draw a square (in pixels) and its two diagonals, and paint two of the resulting four triangles in white (like an upright hourglass) and the other two in black. Because of monitor distortion, this stimulus appears as a rectangle whose white sides are longer than the black. If this stimulus rotates 45 deg in the monitor plane, it appears as a rhombus rather than rectangle. After additional 45-deg rotation, it appears rectangular again, but now the black sides seem longer than the white. Method: 11 observers adjusted the width-to-height (W2H) ratio of a flat-panel monitor to make the stimulus described above appear as an exact square. Importantly, the observers could rotate the stimulus at will by turning a knob. The experimental manipulation was to rotate the monitor itself around its vertical axis so that the angle between the normal vector of the monitor and the line of sight was 0, 38, or 68 degrees, respectively. These viewing angles were counterbalanced across blocks within subjects. Binocular viewing from a chin rest, viewing distance 100 cm from stimulus center, 72 trials (=3 x 24) total. Results: The adjusted W2H ratio increased systematically as the viewing angle increased: mean W2H ratios = 1.04, 1.06, and 1.38. The precision across trials decreased: average standard deviations of the adjustments = 0.06, 0.09, and 0.17, respectively. This indicates a failure of shape constancy for a symmetrical object viewed under natural conditions.
Meeting abstract presented at VSS 2016