You see a number of “blue balls” moving on the screen from left to right or from right to left. The speed of motion of individual dots on the computer screen is not constant. But instead of seeing 2D varying motions of individual balls, you see a 3D rigid rotation of a circular cylinder. The “Gestalt theory” states that our visual system prefers the rigid 3D interpretation over a non-rigid 2D interpretation because the former is simpler.
If you watch this for a while, you may observe spontaneous depth reversal which causes the cylinder’s rotation to invert. This is the Necker cube effect, but that’s another illusion ;-).
Checking Depth lets more distant balls appear dimmer, this effectively avoids spontaneous depth reversal. You can apply rotations around other axes (🔄X and 🔄Z). The balls have a limited lifetime (Max t[s]), so the pattern does not stay constant, yet the 3D-cylinder percept remains.
Interestingly, the depth effect disappears largely when balls and background only differ in colour and not in brightness, as demonstrated here.
The appearance of the 3D cylinder was called the “kinetic depth effect” by Wallach and O’Connell (1953) [they used much simpler rotating stick-figures or contours to demonstrate the phenomenon]. When a computer algorithm recovers 3D shape from a 2D frame sequence, it is called “structure from motion” (Ullman, 1979).
Technical details: The “balls” are randomly distributed on the surface of a vertical cylinder. I then apply a combined rotation matrix to all balls’ (homogeneous) coordinates: first an auto-incrementing rotation around the vertical y-axis, then a rotation around the z-axis (pointing towards you) and finally around the rightwards-pointing x-axis. When depth disambiguation is active, the saturation of the balls is modulated with the normalised z-component (though no painting priorities are observed).
Wallach H, O’Connell DN (1953) The kinetic depth effect. Journal of Experimental Psychology 45:205–217
Ullman S (1979) The interpretation of structure from motion. Proc of the Roy Soc of London. Series B, Biological Sciences: 203(1153), 405–426
