impossibility of distinguishing with absolute accuracy or certainty between relevant and irrelevant circumstances, and partly to our careless way of ignoring the distinction between different conditions so long as they are called by the same name. So long as we know merely that we have indigestion after breakfast' in all the cases mentioned, we cannot tell whether it was caused by the same thing every time or by different things; but if we took the trouble to find out each day how soon after the meal the attack arose, precisely how severe it was, how long it lasted, and every other observable detail, we should soon be able to say whether the trouble really was the same in all the cases, as our first, rough statement implied, or in fact quite different. If it was pre

O cisely the same in every relevant respect, we could be quite sure that in each case the attack was caused by the same thing-namely, the eggs; but if we found constant variations in the symptoms, and if we could be sure (as we cannot) that variations in the things eaten for breakfast, as distinguished from previous conditions or the manner of eating, were the only possible cause of these variations, and that the eggs eaten were all precisely alike in every essential respect, then we could be quite certain that the eggs were not the cause, or at least not the only cause, of the symptoms.

Theoretically I think we must admit that no two causes in the world could be substituted for each other and leave precisely the same results everywhere; and therefore to a perfect intelligence dealing with perfect data there would be no such thing as a plurality of possible causes. I think we must admit, too, that in many cases where we seem to be confronted by such a plurality of possible causes the difficulty is due to careless and avoidable inaccuracy; but, on the other hand, I think it cannot be doubted that in most cases the difficulty is due to the fact that in this practical world we learn a great many things vaguely before we learn anything accurately; so that when any particular question, such as the cause of indigestion, arises we cannot expect anybody to know everything

else so well that he can distinguish the relevant antecedents and consequents from the irrelevant. In a purely hypothetical case a person might be supposed to know that his indigestion after breakfast had absolutely nothing to do with the supper he ate the evening before, or with the soundness of his sleep during the night, or with what he read in the morning newspaper, and so on; and yet not to know what really did cause it. But in most actual cases the facts are reversed and we have to know what caused the trouble before we can find out what circumstances were irrelevant-we cannot state our problem with refinement until we have solved it. And so the characteristic imperfection' of the method of agreement remains. We can only say that the more we know to begin with and the more carefully we distinguish slight differences, the less trouble it will give.

Let us now compare inferences-whether drawn by the method of difference or by that of agreement—which are based upon the comparison of two changeless sitof data com- uations with those which are based upon the observation of changes taking place in one.

Two kinds

pared.

One advantage in having a changing situation when we are seeking for causal relations is that it sometimes enables us, though not always, to distinguish between causes and effects. If we find one place where the soil is moist and vegetation luxurious, and another where moisture and vegetation are both absent, we cannot say whether the moisture causes the vegetation or the vegetation the moisture; but when we notice things begin to grow after a rain we can be quite sure that in this case at least the growth of the vegetation is not the cause and the moisture the effect. If either of them is the cause of the other, it is the moisture. In comparing two changeless situations we talk of antecedents and consequents just as we do in speaking of changes; but here the words are used rather metaphorically, 'antecedent' meaning one of a group of causes, and consequent' one of a group of effects; and we cannot tell which of two circumstances

should be called the antecedent' and which the 6 consequent' until we have found out in some indirect way which of the two is cause and which effect. Where there is a change we can often see directly which event is antecedent, and therefore cause, and which is consequent, and therefore effect.

Yet this advantage in reasoning from a changing situation is rather dubious, so long at least as we are mere observers. Often we notice a situation changing without being able to tell which element in it changes first. We feel the breeze

arising and see the aspen leaves begin to quiver or the trees begin to sway. But do we really observe which came first? Can we prove by direct observation alone that the breeze really came first and was therefore the cause-and not the effect, as children sometimes suppose-of the fan-like swaying of the trees? It is only when there are a number of intermediate links in the chain that the interval in time between causes and effects is really perceptible. Then, again, the observation of the apparent order in time often positively leads us astray. Is the sound of thunder caused by the flash of light because we hear it afterwards? Does the falling of the barometer cause rain because it precedes it? The fact is that observation of the order in time does not help us to tell which of two events is the cause and which is the effect nearly so much as does a knowledge of other causal relations. We do not believe that the wind moves the trees, and not the trees the wind, because we see that the wind comes first, but rather because we do not know anything but wind that is likely to move the trees, while we do know something besides the trees that can stir up the wind. One theory leaves a broken causal chain and the other does not. In other words, a state of affairs in which the wind moves the trees fits into the rest of the world as we know it better than one in which the movement of the trees causes the wind; and so we assume that the former state of affairs and not the latter is the real one. But such reasoning as this is quite as applicable to cases in which there is not any change as to those in which

there is. Thus this advantage which reasoning about change seems to possess over reasoning about situations which do not change is often only apparent.

When we assume that one event must be the effect of another merely because it follows it-post hoc ergo propter hoc' -we commit a very common blunder, which ergo propter is sometimes called the fallacy of 'False Cause', but is more commonly designated by the Latin

Post hoc

hoc.

phrase which describes the reasoning.

Advantages of experiment.

Whether the cases compared compel us to draw our inferences by the method of agreement or by the method of difference, and whether we are comparing two different situations or the changes in one, are not nearly so important questions as one not yet mentioned. This is the question whether we merely observe and compare situations as we happen to find them or whether we deliberately create them for the purpose of answering specific questions the question whether our inference is based upon mere observation' or upon experiment'.

6

Mere observation often raises problems; but when they are once raised, experiment, where it can be tried properly, gives the more satisfactory answers.

In the first place, the very object of experiment is to produce conditions which are thoroughly understood and from which all disturbing factors are, so far as possible, removed. Hence when we experiment there is much less chance that some real cause has been overlooked than when we merely observe some of the occurrences that take place amidst the great confusion of natural conditions.

A second point in favor of experiment is that it enables us to tell, as observation does not, not merely that a pair of circumstances have some direct or indirect causal relation, but that one is the actual cause of the other, and which one that is.

We have just seen (p. 271) how these questions can often be settled indirectly, through our knowledge of the rest of the world. Experiment often enables us to settle them directly.

If I blow air against the branches of a tree, they will begin to move; and from this I conclude that in this case at least the movement of the air is the cause and the movement of the tree the effect. On the other hand, if I sway a tree, a little breeze will arise; and from this I conclude that in this case the movement of the tree is the cause, and the movement of the air the effect. So I say that either may cause the other. Again, if I heat the air around a thermometer, I will find the mercury rise; but if I raise the mercury in some other way, I will not find any noticeable change in the temperature of the surrounding air; so I conclude that the temperature affects the height of the mercury, but not vice versa. Still again, if I forcibly lower the mercury in a barometer, a storm will not follow; and if I make a little rainstorm around it with a watering-pot without lowering its temperature, the mercury will not fall. So I conclude that the fall of the mercury does not cause the rain, or the rain the fall of the mercury, but that when the two are found together they must be joint effects of the same cause; and if I guess what that cause is, I may be able to verify my guess by a new set of experiments designed to test the relative weights of air and watery vapor, and the effect produced upon the mercury in a barometer by the pressure of such different weights.

Assumption in experi

ment.

We can reach all these conclusions by experiment because we take for granted the relative spontaneity of our own voluntary acts. It may be that we ourselves are as much a part of nature as anything else and that every one of our acts is the necessary effect of some preceding cause. But we assume that we know enough about ourselves to be sure that the immediate causes of our acts are very different from the immediate causes of such things as wind, temperature, and rain. If we did not make this assumption, we should have to assume that our own acts in various experiments and the results which seem to follow from them might be merely joint effects of the same causes. Hence we could not be sure that our blowing of the air made