We arrived a moment ago at what we called “physiological inference” as an essential ingredient in the common-sense notion of a physical object. Physiological inference, in its simplest form, means this: given a stimulus S, to which, by a reflex, we react by a bodily movement R, and a stimulus S′ with a reaction R′, if the two stimuli are frequently experienced together, S will in time produce R′.[2] That is to say, the body will act as if S′ were present. Physiological inference is important in theory of knowledge, and I shall have much to say about it at a later stage. For the present, I have mentioned it partly to prevent it from being confused with logical inference, and partly in order to introduce the problem of _induction_, about which we must say a few preliminary words at this stage. [2] _E.g._ if you hear a sharp noise and see a bright light simultaneously often, in time the noise without the light will cause your pupils to contract. Induction raises perhaps the most difficult problem in the whole theory of knowledge. Every scientific law is established by its means, and yet it is difficult to see why we should believe it to be a valid logical process. Induction, in its bare essence, consists of the argument that, because A and B have been often found together and never found apart, therefore, when A is found again, B will probably also be found. This exists first as a “physiological inference”, and as such is practised by animals. When we first begin to reflect, we find ourselves making inductions in the physiological sense, for instance, expecting the food we see to have a certain kind of taste. Often we only become aware of this expectation through having it disappointed, for instance if we take salt thinking it is sugar. When mankind took to science, they tried to formulate logical principles justifying this kind of inference. I shall discuss these attempts in later chapters; for the present, I will only say that they seem to me very unsuccessful. I am convinced that induction must have validity of some kind in some degree, but the problem of showing how or why it can be valid remains unsolved. Until it is solved, the rational man will doubt whether his food will nourish him, and whether the sun will rise tomorrow. I am not a rational man in this sense, but for the moment I shall pretend to be. And even if we cannot be completely rational, we should probably all be the better for becoming somewhat more rational than we are. At the lowest estimate, it will be an interesting adventure to see whither reason will lead us.