For many years, the inversion of the retinal image seemed to be a serious problem: do babies have to learn to correct the inversion, to see things the right way up? This question is based on the false notion that the brain is a kind of eye looking at the retinal image, which presumably has its image-with another eye, another image, and so on forever. This notion of an inner eye should have been a non-starter as it can never get anywhere. The point is, one's retinal image is not seen, as an object is seen. The retina is the interface between the optical projection from objects to the neural-coded signals to the brain-arriving down the million fibres of the optic nerve-which are related to touch experience of objects. The inversion in the image does not matter: what matters is the relation of the brain's visual signals to those from touch. Indeed without touch retinal images would have little or no meaning.(Gregory, 1997, pp. 52-53).Here Gregory presents the inferentialist view of visual images quite succinctly: "one's retinal image is not seen, as an object is seen". As we shall see, this is the way around the problems Gibson sometimes describes has with retinal images.
I do wonder if Gregory had all this stuff down in earlier editions of Eye and Brain. I would guess that he did, but our local copy of the first edition from 1966 is checked out.