like B which I will call β: I must have observed both α and β, and also I must have observed β preceding α. This, at least, I must have observed. But I do not pretend to say how like α and β must be to A and B; nor do I pretend to say how often I must have observed β preceding α, although it is generally held that I must have observed this more than once. These are questions, which would have to be discussed if we were trying to discover what observations were sufficient to justify the generalisation that the existence of A is generally preceded by that of B. But I am only trying to lay down the minimum which is necessary to justify this generalisation; and therefore I am content to say that we must have observed something more or less like B preceding something more or less like A, at least once.
But there is yet another minimum condition. If my observation of β preceding α is to justify the generalisation that the existence of A is generally preceded by the existence of B, it is plain, I think, that both the β and the α, which I observed, must have existed or been real; and that also the existence of β must really have preceded that of α. It is plain that if, when I observed α and β, α existed but β did not, this observation could give me no reason to suppose that on another occasion when A existed, β would exist. Or again, if, when I observed β preceding α, both β and α existed, but the existence of β did not really precede that of α, but, on the contrary, followed it, this observation could certainly give me no reason to suppose that, in general, the existence of A was preceded by the existence of B. Indeed this condition that what is observed must have been real might be said to be included in the very meaning of the word "observation." We should, in this connection, say that we had not observed β preceding α, unless β and α were both real, and β had really preceded α. If I say "I have observed that, on one occasion, my hearing of the word 'moon' was followed by my imagining a luminous silvery disc," I commonly mean to include in my statement the assertion that I did, on that occasion, really hear the word "moon," and really did have a visual image of a luminous disc, and that my perception was really followed by my imagination. If it were proved to me that this had not really happened, I should admit that I had not really observed it. But though this condition that, if observation is to give reason for a generalisation, what is observed must be real, may thus be said to be implied in the very word "observation," it was necessary for me to mention the condition explicitly. It was necessary, because, as I shall presently show, we do and must also use the word "observation" in a sense in which the assertion "I observe A" by no means includes the assertion "A exists"—in a sense in which it may be true that though I did observe A, yet A did not exist.
But there is also, I think, a third necessary condition which is very apt to be overlooked. It may, perhaps, be allowed that observation gives some reason for the proposition that hens' eggs are generally laid by hens. I do not mean to say that any one man's observation can give a reason for this proposition: I do not assume either that it can or that it cannot. Nor do I mean to make any assumption as to what must be meant by the words "hens" and "eggs," if this proposition is to be true. I am quite willing to allow for the moment that if it is true at all, we must understand by "hens" and "eggs," objects very unlike that which we directly observe, when we see a hen in a yard, or an egg on the breakfast-table. I am willing to allow the possibility that, as some Idealists would say, the proposition "Hens lay eggs" is false, unless we mean by it: A certain kind of collection of spirits or monads sometimes has a certain intelligible relation to another kind of collection of spirits or monads. I am willing to allow the possibility that, as Reid and some scientists would say, the proposition "Hens lay eggs" is false, if we mean by it anything more than that: Certain configurations of invisible material particles sometimes have a certain spatio-temporal relation to another kind of configuration of invisible material particles. Or again I am willing to allow, with certain other philosophers, that we must, if it is to be true, interpret this proposition as meaning that certain kinds of sensations have to certain other kinds a relation which may be expressed by saying that the one kind of sensations "lay" the other kind. Or again, as other philosophers say, the proposition "Hens lay eggs" may possibly mean: Certain sensations of mine would, under certain conditions, have to certain other sensations of mine a relation which may be expressed by saying that the one set would "lay" the other set. But whatever the proposition "Hens' eggs are generally laid by hens" may mean, most philosophers would, I think, allow that, in some sense or other, this proposition was true. And they would also I think allow that we have some reason for it; and that part of this reason at all events lies in observation: they would allow that we should have no reason for it unless certain things had been observed, which have been observed. Few, I think, would say that the existence of an egg "intrinsically points" to that of a hen, in such a sense that, even if we had had no experience of any kind concerning the manner in which objects like eggs are connected with animals like hens, the mere inspection of an egg would justify the assertion: A hen has probably existed.
I assume, then, that objects having all the characteristics which hens' eggs have (whatever these may be) are generally laid by hens (whatever hens may be); and I assume that, if we have any reason for this generalisation at all, observation gives us some reason for it. But now, let us suppose that the only observations we had made were those which we should commonly describe by saying that we had seen a hen laying an egg. I do not say that any number of such observations, by themselves, would be sufficient to justify our generalisation: I think it is plain that they would not. But let us suppose, for the moment, that we had observed nothing else which bore upon the connection between hens and eggs; and that, if therefore our generalisation was justified by any observations at all, it was justified by these. We are supposing, then, that the observations which we describe as "seeing hens lay eggs" give some reason for the generalisation that eggs of that kind are generally laid by hens. And if these observations give reason for this, obviously in a sense they give reason for the generalisation that the existence of such an egg is generally preceded by that of a hen; and hence also, they give us reason to suppose that if such an egg exists, a hen has probably existed also—that unless a hen had existed, the egg would not have existed. But the point to which I wish to call attention is that it is only in a limited sense that they do give reason for this. They only give us reason to suppose that, for each egg, there has existed a hen, which was at some time near the place where the egg in question then was, and which existed at a time near to that at which the egg began to exist. The only kind of hens, whose existence they do give us reason to suppose, are hens, of which each was at some time in spatial and temporal proximity (or, if Idealists prefer, in the relations which are the "intelligible counterparts" of these) to an egg. They give us no information at all about the existence of hens (if there are any) which never came within a thousand miles of an egg, or which were dead a thousand years before any egg existed. That is to say, they do give us reason to suppose that, if a particular egg exists, there has probably existed a hen which was at some time near that egg; but they give us no reason to suppose that, if a particular egg exists, there must have existed a hen which never came near that egg. They do give us reason to suppose that, for each egg, there has probably existed a hen which at some time stood to the egg in question in that relation which we have observed to hold between an egg and a hen, when we observed the hen laying an egg. But they give us no reason to infer from the existence of an egg any other kind of hen: any hen which never stood to the egg in the relation in which we have observed that some hens do stand to eggs.
What I wish to suggest is that this condition is a universal condition for sound inductions. If the observation of β preceding α can ever give us any reason at all for supposing that the existence of A is generally preceded by that of B, it can at most only give us reason to suppose that the existence of an A is generally preceded by that of a B which stands to our A in the same relation in whichβ has been observed to stand to α. It cannot give the least reason for supposing that the existence of an A must have been preceded by that of a B, which did not stand to A in the observed relation, but in some quite different one. If we are to have any reason to infer from the existence of an A the existence of such a B, the reason must lie in some different observations. That this is so, in