It cannot be required that the philosopher should be a naturalist, and yet his co-operation in physical researches is as necessary as it is desirable. He needs not an acquaintance with details for this, but only a clear view of those conclusions where insulated facts meet.
(Johann Wolfgang von Goethe 1840/1970:283, §717)
Philosophical discussions about the ontology of colour typically proceed by attempting to determine whether colours can be identified with some set of perceiver-independent physical properties. Thus philosophers have argued that the colours of objects can be identified with the wavelengths of light that objects reflect (Armstrong 1968a), or with the surface spectral reflectances of objects (Hilbert 1987). This kind of view about colour is known as objectivism. The position I am calling computational objectivism results when objectivism is defended by appealing to the computational level of explanation for vision. Most philosophers since Newton's time have found objectivism to be unsatisfactory, and so have argued for some version of subjectivism. The traditional version was explored in Chapter 1; it is the view that things are coloured only in so far as they have the disposition to cause sensations of colour in a perceiver. 1 There is another version of subjectivism, however. According to this view, nothing is strictly speaking coloured at all, not even dispositionally. Rather, colours are entirely 'in the head'; they are nothing but sensations of a certain type. Colours are 'projected' on to the world, but there is no further sense in which the world is coloured (see the quotations from Gouras