ing to each figure of the syllogism. In the case of the general proposition the method of exhaustion is one of addition, such and such a thing is true of this and that and the other, therefore of all. In the case of identity and causation the method of exhaustion is one of subtraction or elimination, the thing or the cause in question cannot be this or that, therefore it must be the other. This method of elimination would not be possible unless we were sure enough of the uniformity of nature to assume that a thing identical with the one in question or that a cause for the given event must surely exist somewhere.

It is noteworthy that in practice questions concerning general truths, concerning identity and concerning causation are interwoven in many ways. When we have proved either by a perfect induction' or by the Inductio per Enumerationem Simplicem that everything with the characteristic A has also the characteristic B we can hardly avoid suspecting that between A and B there is some rather direct relation of causation. Conversely we can prove that everything with the characteristic A has also the characteristic B without the Perfect Induction or Inductio per Enumerationem Simplicem if we have any means of proving that A inevitably causes B. So with Identity. A person can be sure that the book he finds in a certain library to-day is the one he left there yesterday if he can prove that it is the only book like his that is now in the library, and that no books have gone out and come in since his was left. But how can one be sure that a book has not gone out of the library and another come in to take its place? Is it not only because he knows that books cannot pass through solid walls or throw themselves spontaneously through windows and doors, and because he knows also that if anybody had carried one book in and another out the librarian would have seen him or the lock on the door would have been broken, or some other perceptible change would have been caused, when it was not?

In this way knowledge of causal relations helps us to prove identity just as it may help us to prove the truth of a universal proposition. Such a case as this is not exceptional. We prove that the comet now visible is the one that appeared ten years ago because calculations made at that time showed that at precisely this time it must appear at the very spot where a comet is now seen; and since there cannot be two comets in the same place the one we see now must be the one we saw then. So too when a 'medium' tries to prove to you that the spirit with which she now claims to be in communication really is the spirit of your departed friend she tries to show that the spirit does and knows what nobody but that friend would or could do or know; and thus she leaves this problem on your hands: If my friend did not speak through the medium, how in the world am I to account for all the things the medium or the spirit that spoke through her did? In the same way once more, when a prosecuting attorney tries to prove that the prisoner before the court was the person who broke into the bank he attempts to show that there is some effect which could not possibly have been produced if the burglary had been committed by anybody else. Thus a question of identity, like the question of the truth of some general statement, can be resolved into a question of causation.

So, vice versa, when we are investigating a question of causation we always take for granted various relations of identity and various relations of kind such as are expressed in general propositions. When we conclude that a certain man must have been poisoned by arsenic because nothing else would have caused the same symptoms we assume the existence of the man and of the arsenic as real things retaining their identity from the time of the supposed poisoning to the time that the symptoms appeared. We also assume that every bit of arsenic has the same nature as every other bit and acts in the same way under the same conditions, and that men also are similar to each other, so that at least some

<

of the symptoms produced by the arsenic are substantially alike in them all.

In courts of law individuals as such are everything, and general laws or rules of action are only means for adjusting their conflicting claims. Consequently problems of identity in the courts are all-important, and all sorts of devices are resorted to for answering such questions as whether or not this particular man is the one that owned this particular horse or signed this particular bit of paper. But with science it is different; for scientists have very much more to say about the conditions under which certain effects are produced than about the identity of the agents. Since the chemist assumes all atoms of hydrogen to be alike he neither knows nor cares whether the one involved in a given reaction is Atom No. 49 or Atom No. 63. His identification stops when he finds out that the atom in question is some one or other of the many billions of similar atoms that we call hydrogen, and that it has recently been put through such and such a process. For that particular atom as such he cares not a snap of his finger. Its only value for him is to show what any atom of the sort will do under a given set of conditions. This distinction, however, is not absolute. The jury in a trial for murder has to determine whether the injury inflicted by the prisoner upon the deceased really was the cause of his death, and such sciences as geology and history are concerned very largely with the question of what individual person or thing it was that produced this or that given effect. After all, the question of identity and the question of circumstance are inextricably interwoven; for thing and circumstance are. only different aspects of the same concrete fact and neither could act or even exist without the other.

CHAPTER XXV.

INDUCTION BY SIMPLE ENUMERATION AND THE SEARCH FOR CAUSES.

inference, and explanation.

THE Inductio per Enumerationem Simplicem is regarded. by scientists nowadays as a primitive and unsatisfactory kind of inference that should be replaced or supplemented wherever possible by an inference Analysis, based upon the knowledge of causes. The essential difference between the two methods is that the one based upon a knowledge of causes involves analysis and the other does not. Inductio per Enumerationem Simplicem takes cach of the observed relations in the mass just as it stands, notices how often they coincide, and then makes a guess about the future. Causal analysis is not content to take the observed uniformities so roughly; but splits them up, in order to find the simpler and more general uniformities which are involved. And then when this is done it may try to determine the conditions under which they will work together in the same way again. An inference based upon the knowledge of causes is therefore based upon a definite knowledge of all the details rather than upon a confused impression of a whole.

A farmer may happen to notice that red clover grows better near the homes of old maids than elsewhere in the neighborhood. Reasoning per enumerationem simplicem he might conclude that there is always something about an old maid that helps the growth of clover. But he would have not

means of knowing how much faith should be attached to this conclusion. He could not say how likely it was that the next case he noticed would conform to the rule. But suppose he should happen to think that old maids often keep cats, that cats kill mice, that mice destroy bumblebees, and that without bumblebees to carry the pollen from one plant to another red clover cannot develop its seed. Then he would have broken up the first uniformity-between the old maids and the clover-into a series of others, most of which, at least, are much more familiar and therefore much more thoroughly tested; he would know almost exactly how much reliance could be placed upon each of them; and putting them all together he could say without any hesitation whatever whether it is always true that old maids help the clover-meadows. In this way, a knowledge of causes gives much greater certainty than the bare, unanalytic Inductio per Enumerationem Simplicem.

In this example, as in almost any other that might be given, the analysis into causes or simpler uniformities is very incomplete. Each of the causal relations given might be itself reduced to others still simpler. That old maids keep cats; that cats kill and eat mice; that mice destroy the nests and young of bumblebees; and that red clover needs these bees to fertilize it all these may be themselves mere Inductiones per Enumerationem Simplicem, about which we may feel sceptical; and we may wish to test any or all of them by finding the still simpler uniformities that would account for them if they really existed. If cats do always eat mice, the mice are probably good for them; if mice are good for cats to eat, it must be because they can be digested and pass into the system; if digested food passes into the system of a cat or any other animal, it must be because it can get through the membranes lining the alimentary canal; if food can get through these membranes, it must be because certain fluids will pass through moist membranes anywhere. In this way one uniformity after another can be reduced to others still more general, until we can carry the process no further and we