Another point: If we accept the definition of knowledge given in Chapter VIII, which was framed so far as to be as favourable as possible to behaviourism, a given reaction may be regarded as knowledge of various different occurrences. When we see Jupiter, we have, according to the definition, knowledge of Jupiter, but we also have knowledge of the stimulus at the surface of the eye, and even of the process in the optic nerve. For it is arbitrary at what point we start in the process leading to a certain event in the brain: this event, and the consequent bodily action, may be regarded as a reaction to a process starting at any earlier point. And the nearer our starting-point is to the brain, the more accurate becomes the knowledge displayed in our reaction. A lamp at the top of a tall building might produce the same visual stimulus as Jupiter, or at any rate one practically indistinguishable from that produced by Jupiter. A blow on the nose might make us “see stars”. Theoretically, it should be possible to apply a stimulus direct to the optic nerve, which should give us a visual sensation. Thus when we think we see Jupiter, we may be mistaken. We are less likely to be mistaken if we say that the surface of the eye is being stimulated in a certain way, and still less likely to be mistaken if we say that the optic nerve is being stimulated in a certain way. We do not eliminate the risk of error completely unless we confine ourselves to saying that an event of a certain sort is happening in the brain; this statement may still be true if we see Jupiter in a dream. But, I shall be asked, what do you know about what is happening in the brain? Surely nothing. Not so, I reply. I know about what is happening in the brain exactly what naive realism thinks it knows about what is happening in the outside world. But this needs explaining, and there are other matters that must be explained first.