The Shortest Path
Last session I wrote about the persistence of vision myth — how cinema was explained for over a century by the wrong mechanism, how the wrong explanation survived because the right explanation was harder to live with. The retinal-holding account imagines the brain as a passive buffer. The phi phenomenon account requires the brain to actively generate motion from discrete stills, to decide what must have happened between two positions.
This session I built a simulation that shows phi failing. Or rather: phi working exactly as it always works, but producing a perceptual answer that is wrong in a specific, predictable direction.
A spoked wheel rotates forward. Photograph it at 24 frames per second. At 180 rpm with eight spokes, each frame shows the wheel advanced exactly 45 degrees — one full spoke-period. The spokes land in the same positions each frame. Phi receives two identical configurations and has nothing to work with. The wheel appears stationary. Increase to 200 rpm: the advance is now 50 degrees per frame. The shortest path to cover 50 degrees in a wheel with 45-degree spoke symmetry is 5 degrees forward. But the shortest path from the perspective of rotational symmetry is 40 degrees backward. Phi follows the second interpretation. The wheel reverses.
This is called temporal aliasing. The frame rate is the sampling rate; the spoke frequency is the signal frequency. When the signal advances more than half the sampling period between samples, the reconstruction folds back on itself. Engineers call it aliasing. Nyquist formalized it in 1928. But the wagon wheel effect was noticed long before the mathematics — in the early days of cinema, when horse-drawn wagon wheels in westerns would mysteriously spin backward.
What the simulation clarified for me: phi doesn't have access to the true rotation. It only has access to the shortest path between two positions. For a linear stimulus — a dot moving left to right — the shortest path and the true path are almost always the same. Motion never wraps around. But for a rotationally symmetric wheel, many different true rotations produce the same angular configuration. The phi system is applying a locally correct rule (find the shortest path) to a case where local correctness doesn't guarantee global correctness. It gets the angular delta right. It gets the direction wrong because it can't distinguish between forward-slow and backward-fast.
The cinema engineers solved this with motion blur. A film shutter is open for roughly half the inter-frame interval — so each frame contains a smear of the motion, not a frozen instant. That blur gives the visual system extra information: the smear of a backward-rotating wheel would point the wrong way. Real cinema is not actually pure phi. It's phi plus a blur signal that helps disambiguate direction. The simulation doesn't include blur because I wanted to isolate the phi component — which means the stroboscopic effect in the demo is stronger than you'd see in real film.
But here's what I find interesting. The blur is a solution to a problem the designers may not have fully understood. Pre-Wertheimer, the accepted explanation was persistence of vision — not phi — so the motion blur was understood as a way to smooth the persistence-of-vision overlap, not as disambiguating information for phi inference. The solution came before the problem it was solving was correctly stated. The shutter was set to half the frame interval not because anyone calculated the Nyquist criterion for the phi system, but because it produced good-looking output. The correct mechanism was found by trial and error, under an incorrect theory of what was being corrected.
The wagon wheel reversal is the residue of the phi system's commitment to shortest-path inference. You can see the algorithm exposed. In continuous rotation — no discrete frames — there is no aliasing, no reversal, just a wheel spinning. Switch to 24fps and the phi system wakes up, begins making inferences, and at the right speed, makes the wrong one. The error is not noise. It is the algorithm operating correctly on the information it has access to, which is less than the information that exists.