PLACE OF HYPOTHESIS IN EXPEEIMENTAL SCIENCE. 551

that the existing distribution of matter and motion is under- going a gradual transformation which, when complete, will present to future observers a set of relations of coexistence, antecedence and sequence totally different from any which we observe to-day. If causation were really unconditionally invariable antecedence and sequence between phenomena, none but periodical events could with any plausibility be said 'to be subject to causation, and of them only such as could be proved to be not subject to modification by counteracting causes ; nor even so would any date be assignable when post hoc might ripen into propter hoc. In fact, however, it cannot be maintained that any observable events there are of which the recurrence is unconditionally invariable. The ' compo- sition ' of causes stands in the way. It is indeed impossible, consistently with the empirical theory of causation, to give a meaning to this expression 'composition of causes'. Two causes may neutralise one another, or be so compounded as to produce a different result from that which either of them operating singly would have produced. Were causation unconditional invariability of antecedence and sequence, this would be impossible. If un- conditional invariability has any meaning, an unconditionally invariable relation of antecedence and sequence is one which is in no way modifiable. The composition of two uncondi- tionally invariable events is a contradiction in terms. Nay, in strictness of speech, events are not capable of composition at all ; they can only be related in the way either of coexis- tence, or of antecedence and sequence. It may perhaps be said that the empirical theory of the composition of causes merely means that the simultaneous occurrence of several events is invariably followed by an event different from that which would have invariably supervened upon any one of them occurring singly. This, however, is to abandon the idea of unconditionally. An event which is only followed by another event provided some other event also occurs, is not an unconditionally invariable antecedent, and therefore, on the empirical theory of causation, no cause. In short, if we define causation as unconditional invari- ability of antecedence and sequence between phenomena, we have no choice but to hold that only those antecedents upon which the same event always ensues and will ensue "so long as the present constitution of things endures," no matter what other antecedents are combined with them and " under all changes of circumstances," are causes ; but, as we do not know that any such antecedents there are, the consistency of