Research by way of experiment, which is impossible in Astronomy, appears in Physics. It is therefore here where it originates that we must study it. It is also here that it is most successful, and gives the greatest number of results. Indeed, to experiment successfully we must be able to compare two cases “which present no other difference direct or indirect, than that which relates to the course of the phenomenon under analysis.”[135] By experimenting, Comte here clearly designates what John Stuart Mill will call the method of difference, that is to say the most powerful of his methods for the investigation of phenomena. Now, experimenting, so understood, is extremely difficult when very complicated phenomena are concerned. In physiology, for instance, the experiments must be combined in such a way as to maintain the subjects in the living state, and even, as far as possible, in the normal state. But any modification of one part of the organism immediately affects the other parts. The living being reacts instantly, and adapts itself as best it can to the new conditions in which it has been placed by the experimentalist. We can therefore hardly ever establish in physiology what is so easily obtained in physics: two cases exactly similar in all respects, except in the one which we want to analyse. In chemistry, it is true, experimenting would seem to be even easier than in physics, since in it, as a rule, we merely consider facts resulting from circumstances which are produced by man’s intervention. But this is to mistake the nature of the experimental method. The essence of this process does not consist in man’s institution of the circumstances surrounding the phenomena; it lies in the “freest possible choice of the case best suited to show the law of the phenomenon,” whether this case be, moreover, natural or artificial. Now, this choice is nearly always easier in physics than in chemistry. For the chemical phenomena more complex in themselves, in general can only be brought about by the co-operation of a great number of different influences; for this reason in chemistry, it is more difficult to modify the circumstances under which phenomena are produced, and still more difficult to isolate as completely as in physics the various conditions by which phenomena are determined.