The distinction between "normal" and "transition" _periods_ suggests a dangerous fallacy: namely, that during one period one sort of causation is working, with the other in abeyance. In fact, whatever causes there are are working all the time. The only legitimate thing is to abstract from one set of causes, and see what the other set, if left to themselves, will bring about. But this sort of abstraction has many dangers, one of which is that the causes abstracted from are frequently thought of as non-existent. The chemist, in his laboratory, can in actual physical fact abstract impurities from his chemicals, and see what they will do. He can even perform experiments in what is practically a vacuum. But the economist has no right to _think in vacuo_! All that he has a right to do is to assume the factors which he does not wish to study _constant_. And even that he must not do if (1) changes in the factors which he wishes to study do in fact lead to changes in the factors abstracted from, or (2) if the factors which he wishes to study can only change _because_ of prior or concomitant changes in the factors from which he is abstracting. Is it, for example, legitimate to assume an increase in M' apart from its usual accompaniment, an increase in PT? The notion, too, that causation can be seen in a state of stable equilibrium should be critically analyzed. Causation is only _revealed_ by a _course of events_, when mechanical causation is involved. The relation of cause and effect may be a contemporaneous relation in fact, and it is possible, where conscious, psychological phenomena are involved, to discern causal relations among the elements in a mental state by direct introspection. It is the not uncommon practice, also, in the theory of mechanics, or in theoretical economics, where the method of investigation is deductive rather than inductive, to abstract from the temporal sequence, and to construe causal relations as timeless, logical relations. But even here, the cause of a _change_ in the general situation precedes the change in time, and it is only by abstraction that the time element is left out. If there is no question as to the causal relations, this abstraction is legitimate, but if all that one knows about the situation be that in a stable equilibrium certain constant ratios obtain, then the question as to which term in the ratio is cause and which is effect remains unanswered. In Fisher's situation, then, assuming that it be true--which I shall deny--that the only stable equilibrium is that which the normal theory requires, it still remains true that the causal relations among the factors can only be revealed by a study of the transitions, by seeing the temporal sequence of changes in the factors of the equation. Even if it be granted that M, M' and P tend to keep a constant relation to one another, the quantity theory falls if, for instance, it can be shown that a change may first occur in