The Copycat Solution to the Shadow Correspondence Problem

Abstract

Cast shadows (henceforth simply ‘shadows’) depicted in artworks can fulfil some of their perceptual roles, such as helping the retrieval of 3-D structure and of relative positions of objects, without having to be geometrically or physically accurate. The visual system displays a wide tolerance for shadows that are mostly inaccurate, at times making use of the scarce accurate but relevant information that can still be retrieved from them. However, the extent of the tolerable inaccuracy is at present still unexplored. I propose that inaccuracy can be not only totally acceptable but in some cases is even likely to be preferable to accuracy if the main perceptual role of shadows is to help locate in a scene the objects that cast them. I examined a small but effective historical corpus of pictorial endeavours, from which it appears that in some cases painters have used a copycat strategy for drawing the terminator of a shadow, ie they have produced a replica of the visible profile of the caster, which in the norm yields an impossible shadow. The copycat strategy is perceptually effective for solving the correspondence problem of associating a shadow with its caster; copycatting can be more effective than other solutions, including the depiction of the geometrically correct shadow, and is complementary to coarser solutions to the correspondence problem. These phenomena provide insight into the computations effectively used by the brain to deal with space perception. In particular, the brain is not relying on some sort of simplified physical model of the world, for the shadows produced by the copycat effect would correspond to a more complicated physical situation than the one encountered in real life.