Ancient logic was also familiar with this method of reasoning; but it was branded with the name of incomplete induction as the worst of all because it lacks absolute and unconditional certainty. One must admit, however, that all contemporary science makes use of no form of reasoning other than incomplete induction. It alone permits of prediction, that is to say, the determination of relations which have not yet been directly observed.
But how does science get along with this lack of certainty in its method of drawing conclusions? The reply to this question is that the probability of the conclusions may run the gamut of all possible gradations from mere supposition to the maximum of probability which is no longer to be distinguished, practically, from certainty. The probability is greater the more frequently any given incomplete induction of this kind has been found consistent with subsequent experience. Thus we have at our disposal a number of propositions which in their simplest and most general shape take the form: If the component A is to be found in a given thing, the component B is also to be found in it (in relation to either time or space).
If the relation is one of time, we term this general proposition the law of causality. If it is one of space we speak of the idea (in the Platonic sense) or type of the thing, of substance, etc.
These considerations yield answers to many questions which have been repeatedly propounded in various forms. We have first the question of the universal validity of the law of causality. All attempts to establish this kind of validity have failed, and only the fact remains that without this law we should feel an unendurable uncertainty as to the world. Hence it follows that we have to deal in this matter with a question of fitness. From the constant stream of our experience we select relations which we encounter again and again in order that whenever the component A is given us we may conclude that the component B is also to be found. Hence we do not find these correlations occurring as 'given,' but we ourselves bring them into our experiences, by ourselves regarding the components which show such a connection as belonging together.
We may make quite the same statements in regard to space relations. The components which are always, or at any rate frequently, encountered together we interpret as forming a unit; and we shape from them a concept which includes these components. As in the case of time relation, there is no sense in propounding the question why. There are thousands of correlations to which we pay no attention because they are unique or rare. Knowledge of such a unique correlation leads to nothing, because it does not enable us to infer the presence of one component from the presence of another and therefore does not render prediction possible. Of all possible and actual combina-
