Kant gave a new turn to an old philosophical controversy, as to how far our knowledge is _a priori_ and how far it is based on experience. Kant admitted that without experience we could know nothing, and that what we know is only valid within the realm of experience. But he held that the general framework of our knowledge is _a priori_ in the sense that it is not proved by means of particular facts of experience, but represents the conditions to which phenomena have to conform in order to be capable of being experienced. Before his day, the tendency had been for continental philosophers to regard almost everything as _a priori_ while British philosophers regarded almost everything as empirical. But both sides thought that what is _a priori_ can be proved by logic, at least in theory, whereas Kant held that mathematics is _a priori_ and yet “synthetic”, _i.e._ not capable of being proved by logic. In this he was misled by geometry. Euclidean geometry, considered as true, is “synthetic” but not _a priori_; considered merely as deducing consequences from premisses, it is _a priori_ but not “synthetic”. The geometry of the actual world, as required by engineers, is empirical; the geometry of pure mathematics, which does not inquire into the truth of the axioms but merely shows their implications, is an exercise in pure logic. It should be said, however, that, if the correct analysis of knowledge bears any resemblance at all to that which has been suggested in this book, the whole controversy between empiricists and apriorists becomes more or less unreal. All beliefs are _caused_ by external stimuli; when they are as particular as the stimuli they are of the sort which an empiricist might regard as _proved_ by experience, but when they are more general difficulties arise. A foreigner arrives in America and sees the immigration officials, who lead him to the generalisation that all Americans are rude; but a few minutes later the porter upsets this induction in the hope of a tip. Thus sometimes a given belief will be caused by one event and destroyed by another. If all the events in a man’s life, so far as they affect the belief in question, are such as to cause it, he counts the belief true. The more general a belief is, the more events are relevant to it, and therefore the more difficult it is for it to be such as a man will long consider true. Roughly speaking, the beliefs which count as _a priori_ will be those which well might have been upset by subsequent events, but in fact were confirmed. Here as elsewhere we are driven to the view that theory of knowledge is not so fundamental as it has been considered since Kant.