In
Experiment 3, the background motion was broken up into short segments, as illustrated in
Figure 4. In
Figure 4a, a circle is broken up into a polygon of 36 sides (arrowed), with each side subtending 10 deg (360/36) at the center of the polygon. This polygon could define the circular orbit of a background, but instead, a duplicate set of these 36 arrows has been assembled in the center of the polygon as the radii of a tiny porcupine or asterisk. Now, each radius in the asterisk, taken in clockwise sequence, represents the motion of a background texture, and the lengths of the lines are proportional to the distances traveled. Instead of a single background that moved around along complete circles subtending 2 deg of visual angle, a new background was randomly selected from a set of eight candidate textures, 36 times per revolution, that is, after the equivalent of 10 deg of angular rotation. Thus, Texture 0 moved vertically upward through a distance of 21 arcmin of visual angle (2 × π × 2 deg/36), taking a time of 167 ms. It was then replaced by a new unrelated texture (Texture 1) that moved upward through 21 arcmin at an angle 10 deg clockwise from vertical. Then, a new texture (Texture 2) moved up at an angle of 20 deg, Texture 3 at 30 deg …, Texture 35 at an angle of 350 deg—in other words, 10 deg counterclockwise from vertical. The sequence then repeated indefinitely. This meant that instead of a continuously circling background, there were only brief snatches of approximately linear motions, whose directions stepped progressively around a circle, and each texture moved through only a small distance, 21 arcmin or about 3% of a circumference, before it was replaced by a fresh texture. This sequence of movements would have been parallel to a single background following a rotary orbit, but the sequence deliberately eliminated any continuity of motion, position, or pattern.
Movie 4 offers a reduced version of the stimulus but with only 8 instead of 36 texture changes per revolution.
The textures were brightly colored lattices of stars, flowers, bull's-eyes, and the like, taken from the texture palette of the Deneba Canvas drawing program. We examined whether this background sequence of interrupted motions would suffice to induce a shape illusion.