Fixate on the white cross at the center. Three yellow dots are placed at equal distances from the fixation point. With the rotating mask running, the dots will begin to vanish — one at a time, or in pairs, sometimes all three. They are still on screen. The canvas code never removes them.
Press and hold SPACE (or the button below) whenever any dot is gone. Release when it returns. After ~30 seconds you'll have a disappearance record.
The effect was reported by Yoram Bonneh, Alexander Cooperman, and Dov Sagi in 2001. A stationary target surrounded by a global coherent motion signal periodically disappears from awareness — for 0.5–5 seconds at a time, with the rate and duration varying between observers.
Three hypotheses compete. The filling-in account: the visual system extrapolates the background pattern across the target location, as it does with the blind spot. The attention-based suppression account: the motion stimulus claims attentional resources, leaving the stationary target suppressed from competition. The neural adaptation account: sustained stationary input reduces the firing rate of neurons encoding the target, and they fall below threshold. All three generate predictions that partially match the data. The choice between them is not settled.
What MIB shares with blindsight and inattentional blindness: a gap between the physical state of the stimulus and the observer's report. But the gap is differently caused. In blindsight, the processing exists but doesn't reach awareness. In inattentional blindness, the target is never selected for representation. In MIB, the target was represented — and then actively suppressed. The transition is visible to the observer (something changed) but the mechanism is not.
hide mask removes the rotating pattern to confirm the dots are present. hide targets shows what the mask looks like without them. Both confirm the ground truth the canvas is maintaining while you perceive something else.