In a popular visual illusion, the portrait on paper currency is folded into an M shape along vertical lines through the nose and the eyes. When this folded picture is tilted back and forth horizontally the face undergoes striking changes in expression. This distortion reveals two insights concerning 3D representation in the human visual system and we have explored these with experiments on simple schematic faces and observations on distortions of laser range images of faces. The observations show first that when recovering depicted depth, pictorial cues are interpreted independently of binocular depth information and second, that the recovery of facial expression is based on a scaled prototypical face structure.