The yellow dots were found to disappear alternately for 40% of the viewing time, with durations as long as 10 seconds. Disappearance of the 3 dots was largely independent, and we found that a dot in the upper-left quadrant tends to disappear more for many people. Note that large eye movements tend to bring the dots back, but disappearance can resist small eye movements. Note also that it is possible to pay attention to a specific dot, which will nevertheless disappear.
The yellow dots were found to disappear more when their luminance contrast was increased. Disappearance of any of the dots occurred between 20% and 35% of the viewing time for the lowest and highest luminance respectively. All targets were clearly visible in all luminance levels.
Slow movement of the targets did not eliminate disappearance. Occasionally, targets were reported to disappear in one quadrant and reappear in another. At a speed of 0.2 deg/sec (visual angel), targets still disappeared for 10% of the viewing time.
Targets, which were dynamically changed during presentation, could still disappear. Targets that were flickered at a 3 Hz. disappeared for 25% of the time. Similar results were obtained with targets made of small bars that rotated locally, and with targets that were jittered in position. Position jitter uniformly distributed in the range of 15 arcmin every cycle (33 ms), were reported to disappear over 10% of the time. Overall, large and abrupt transients tended to bring the disappeared patterns back to awareness, but small and regular changes, did not disrupt disappearance.
Large targets disappeared less. Target dots of 0.5 deg of visual angel still disappeared for over 15% of the time. Line patterns of more than 1 deg. long could disappear in parts (see Gestalt effects).
Target dots surrounded by blank “protection zones” in which the blue dots were not visible, disappeared for 20% of the time for zones of 1.5 deg. in diameter. Disappearance decreased with the size of the protection zones.
Different mask types were compared: 3D “sphere”, 2D rotation, 1D translation and Brownian motion. Disappearance increased with motion speed for all types of mask. The 3D “sphere” was found to be the most efficient in causing disappearance. Target dots could still disappear for as few as 10 mask dots.
Two types of Gestalt grouping were tested: Contour smoothness and Proximity. In general, good Gestalts tend to disappear as whole (all parts together) or resist disappearance. For example, the 3 dots in the 2 groups (see demo) tended to disappear together when positioned in proximity, but disappeared in isolation when largely spaced. Other effects are discussed below under “object rivalry”
Two distinct objects tend to disappear in alternation (rivalry). The 2 overlapping ellipses or triangles tend to disappear one at a time. A stronger effect is perceived with 2 adjacent orthogonal Gabor patches, which show rivalry without overlap! In comparison, 2 collinear patches tend to disappear and reappear simultaneously.