The existence of colour-generating interactions across the corpus callosum has recently been suggested from observations with a 'split-brain' patient, thus indicating long-range colour computations at the cortical level. Observations on induced colours described here suggest long-range colour computations at the retinal level. If a white surface surrounded by a particular colour is fixated for some time, the resulting after-image has two colours: the surround appears in complementary colour, whereas the white centre takes on the colour of the surround. The question of whether such colour induction is located in the retina or more centrally was tested in a brain-injured patient with hemianopia. It could be demonstrated that areas of the visual field that are no longer represented in the geniculo-striatal pathway still contribute to colour induction, suggesting that colour induction is a retinal phenomenon.