In 1834 the Scottish scientist Robert Addams stood by the Falls of Foyers and stared at the water for a while. When he looked away at the rocks beside it, they seemed to creep slowly upward. They did not go anywhere. They stayed exactly where they were, and still they moved.
People had noticed this for a very long time. Aristotle wrote that things at rest look as if they are moving after you have watched movement, though he did not say in which direction. In 1820 the Czech physiologist Jan Purkyně gave the first clear account, after watching a cavalry parade. Addams's waterfall gave the effect its popular name: the waterfall illusion, or the motion aftereffect.
Press Adapt. The left half of the stripes will scroll down and the right half will scroll up. Keep your eyes on the cross in the middle, not on the stripes. When the time is up the stripes freeze. Keep looking at the cross.
If it worked, the frozen stripes on the left now seem to drift upward and those on the right seem to drift downward, even though nothing on the screen is moving. The effect fades over several seconds. Try a longer adaptation time and compare. Try the other mode, where both halves scroll down, and both will seem to drift up.
Your visual system contains cells that respond to motion in a particular direction. Some prefer downward movement, some upward, others left or right. When the scene is still, the cells for opposite directions fire at their ordinary background rates, and the signals balance. Your brain reads that balance as "nothing is moving".
Watch a downward flow for long enough and the cells that favour downward motion adapt: they respond less, and their background firing drops too. The upward cells have been idle, so they are untouched. Now show a still picture. The upward cells are stronger than the tired downward cells, the balance tips, and the brain reads the tilt as upward motion. The two meters above show this idea. The numbers in them are made up to show the shape of the effect, not measured values.
You might guess that the retina gets tired. Experiments suggest otherwise. If you adapt one eye and then test the other, the aftereffect still shows up. This is called interocular transfer. Since the two eyes only share their signals once they reach the brain, the adapted cells must sit there. Brain imaging in 1995 pointed at a region called MT, also known as V5, which is rich in direction-sensitive cells. A person looking at a still pattern after watching motion showed activity in that area.
Any long, steady flow can do it: a river, a treadmill belt, a window of a moving train, or a long scroll down a page. Stare at it, look away, and the world drifts the other way for a moment. It is a small glitch that shows how your brain works. It never receives "this is moving" as a ready-made fact. It works the fact out by comparing channels, and a tilted comparison gives a wrong answer without any warning.
Sources: Addams (1834); Purkyně (1820); Michael Bach's Visual Phenomena & Optical Illusions page on the motion aftereffect; Tootell et al. (1995), fMRI of the motion aftereffect in human area MT.