(1) It is sometimes contended, and with some plausibility, that what we mean by saying that it is possible for a thing which possesses one predicate F to possess another G, is, sometimes at least, merely that some things which possess F do in fact also possess G. And if we give this meaning to "possible," the corresponding meaning of the statement it is impossible for a thing which possesses F to possess G will be merely: Things which possess F never do in fact possess G. If, then, we understood "impossible" in this sense, the condition for the "internality" of a kind of value, which I have stated by saying that if a kind of value is to be "intrinsic" it must be impossible for two things to possess it in different degrees, if they are exactly like one another, will amount merely to saying that no two things which are exactly like one another ever do, in fact, possess it in different degrees. It follows, that, if this were all that were meant, this condition would be satisfied, if only it were true (as for all I know it may be) that, in the case of all things which possess any particular kind of intrinsic value, there happens to be nothing else in the Universe exactly like any one of them; for if this were so, it would, of course, follow that no two things which are exactly alike did in fact possess the kind of value in question in different degrees, for the simple reason that everything which possessed it at all would be unique in the sense that there was nothing else exactly like it. If this were all that were meant, therefore, we could prove any particular kind of value to satisfy this condition, by merely proving that there never has in fact and never will be anything exactly like any one of the things which possess it: and our assertion that it satisfied this condition would merely be an empirical generalisation. Moreover if this were all that was meant it would obviously be by no means certain that purely subjective predicates could not satisfy the condition in question; since it would be satisfied by any subjective predicate of which it happened to be true that everything which possessed it was, in fact, unique—that there was nothing exactly like it; and for all I know there may be many subjective predicates of which this is true. It is, therefore, scarcely necessary to say that I am not using "impossible" in this sense. When I say that a kind of value, to be intrinsic, must satisfy the condition that it must be impossible for two things exactly alike to possess it in different degrees, I do not mean by this condition anything which a kind of value could be proved to satisfy, by the mere empirical fact that there was nothing else exactly like any of the things which possessed it. It is, of course, an essential part of my meaning that we must be able to say not merely that no two exactly similar things do in fact possess it in different degrees, but that, if there had been or were going to be anything exactly similar to a thing which does possess it, even though, in fact, there has not and won't be any such thing, that thing would have possessed or would possess the kind of value in question in exactly the same degree. It is essential to this meaning of "impossibility" that it should entitle us to assert what would have been the case, under conditions which never have been and never will be realised; and it seems obvious that no mere empirical generalisation can entitle us to do this.
But (2) to say that I am not using 'necessity' in this first sense, is by no means sufficient to explain what I do mean. For it certainly seems as if causal laws (though this is disputed) do entitle us to make assertions of the very kind that mere empirical generalisations do not entitle us to make. In virtue of a causal law we do seem to be entitled to assert such things as that, if a given thing had had a property or were to have a property F which it didn't have or won't have, it would have had or would have some other property G. And it might, therefore, be thought that the kind of 'necessity' and 'impossibility' I am talking of is this kind of causal 'necessity' and 'impossibility.' It is, therefore, important to insist that I do not mean this kind either. If this were all I meant, it would again be by no means obvious, that purely subjective predicates might not satisfy our second condition. It may, for instance, for all I know, be true that there are causal laws which insure that in the case of everything that is 'beautiful,' anything exactly like any of these things would, in this Universe, excite a particular kind of feeling in everybody to whom it were presented in a particular way: and if that were so, we should have a subjective predicate which satisfied the condition that, when a given thing possesses that predicate, it is impossible (in the causal sense) that any exactly similar thing should not also possess it. The kind of necessity I am talking of is not, therefore, mere causal necessity either. When I say that if a given thing possesses a certain degree of intrinsic value, anything precisely similar to it would necessarily have possessed that value in exactly the same degree, I mean that it would have done so, even if it had existed in a Universe in which the causal laws were quite different from what they are in this one. I mean, in short, that it is impossible for any precisely similar thing to possess a different value, in precisely such a sense as that, in which it is, I think, generally admitted that it is not impossible that causal laws should have been different from what they are—a sense of impossibility, therefore, which certainly does not depend merely on causal laws.
That there is such a sense of necessity—a sense which entitles us to say that what has F would have G, even if causal laws were quite different from what they are—is, I think, quite clear from such instances as the following. Suppose you take a particular patch of colour, which is yellow. We can, I think, say with certainty that any patch exactly like that one, would be yellow, even if it existed in a Universe in which causal laws were quite different from what they are in this one. We can say that any such patch must be yellow, quite unconditionally, whatever the circumstances, and whatever the causal laws. And it is in a sense similar to this, in respect of the fact that it is neither empirical nor causal, that I mean the 'must' to be understood, when I say that if a kind of value is to be 'intrinsic,' then, supposing a given-thing possesses it in a certain degree, anything exactly like that thing must possess it in exactly the same degree. To say, of 'beauty' or 'goodness' that they are 'intrinsic' is only, therefore, to say that this thing which is obviously true of 'yellowness' and 'blueness' and 'redness' is true of them. And if we give this sense to 'must' in our definition, then I think it is obvious that to say of a given kind of value that it is intrinsic is inconsistent with its being 'subjective.' For there is, I think, pretty clearly no subjective predicate of which we can say thus unconditionally, that, if a given thing possesses it, then anything exactly like that thing, would, under any circumstances, and under any causal laws, also possess it. For instance, whatever kind of feeling you take, it is plainly not true that supposing I have that feeling towards a given thing A, then I should necessarily under any circumstances have that feeling towards anything precisely similar to A: for the simple reason that a thing precisely similar to A might exist in a Universe in which I did not exist at all. And similarly it is not true of any feeling whatever, that if somebody has that feeling towards a given thing A, then, in arty Universe, in which a thing precisely similar to A existed, somebody would have that feeling towards it. Nor finally is it even true, that if it is true of a given thing A, that, under actual causal laws, any one to whom A were presented in a certain way would have a certain feeling towards it, then the same hypothetical predicate would, in any Universe, belong to anything precisely similar to A: in every case it seems to be possible that there might be a Universe, in which the causal laws were such that the proposition would not be true.
It is, then, because in my definition of 'intrinsic' value the 'must' is to be understood in this unconditional sense, that I think that the proposition that a kind of value is 'intrinsic' is inconsistent with its being subjective. But it should be observed that in holding that there is this inconsistency, I am contradicting a doctrine which seems to be held by many philosophers. There are, as you probably know, some philosophers who insist strongly on a doctrine which they express by saying that no relations are purely external. And so far as I can make out one thing which they mean by this is just that, whenever r has any relation whatever which y has not got, x and y cannot be exactly alike: That any difference in relation necessarily entails a difference in intrinsic nature. There is, I think, no doubt that when these philosophers say this, they mean by their 'cannot' and 'necessarily' an unconditional 'cannot' and 'must.' And hence it follows they are holding that, if, for instance, a thing A pleases me now, then any other thing, B, precisely similar to A, must, under any circumstances, and in any Universe, please me also: since, if B did not please me, it would not possess a relation which A does possess, and therefore, by their principle, could not be precisely similar to A—must differ from it in intrinsic nature. But it seems to me to be obvious that this principle is false. If it were true, it would follow that I can know a priori such things as that no patch of colour which is seen by you and is not seen by me is ever exactly like any patch which is seen by me and is not seen by you; or that no patch of colour which is surrounded by a red ring is ever exactly like one which is not so surrounded. But it is surely obvious, that, whether these things are true or not they are things which I cannot know a priori. It is simply not evident a priori that no patch of colour which is seen by A and not by B is ever exactly like one which is seen by B and not by A, and that no patch of colour which is surrounded by a red ring is ever exactly like one which is not. And this illustration serves to bring out very well both what is meant by saying of such a predicate as 'beautiful 'that it is intrinsic,' and why, if it is, it cannot be subjective. What is meant is just that if A is beautiful and B is not, you could know a priori that A and B are not exactly alike; whereas, with any such subjective predicate, as that of exciting a particular feeling in me, or that of being a thing which would excite such a feeling in any spectator, you cannot tell a priori that a thing A which did possess such a predicate and a thing B which did not, could not be exactly alike.
It seems to me, therefore, quite certain, in spite of the dogma that no relations are purely external, that there are many predicates, such for instance as most (if not all) subjective predicates or the objective one of being surrounded by a red ring, which do not depend solely on the intrinsic nature of what possesses them: or, in other words, of which it is not true that if x possesses them and y does not, x and y must differ in intrinsic nature. But what precisely is meant by this unconditional 'must,' I must confess I don't know. The obvious thing to suggest is that it is the logical 'must,' which certainly is unconditional in just this sense: the kind of necessity, which we assert to hold, for instance, when we say that whatever is a right-angled triangle must be a triangle, or that whatever is yellow must be either yellow or blue. But I must say I cannot see that all unconditional necessity is of this nature. I do not see how it can be deduced from any logical law that, if a given patch of colour be yellow, then any patch which were exactly like the first would be yellow too. And similarly in our case of 'intrinsic' value, though I think it is true that beauty, for instance, is 'intrinsic,' I do not see how it can be deduced from any logical law, that if A is beautiful, anything that were exactly like A would be beautiful too, in exactly the same degree.
Moreover, though I do believe that both "yellow" (in the sense in which it applies to sense-data) and "beautiful" are predicates which, in this unconditional sense, depend only on the intrinsic nature of what possesses them, there seems to me to be an extremely important difference between them which constitutes a further difficulty in the way of getting quite clear as to what this unconditional sense of "must" is. The difference I mean is one which I am inclined to express by saying that though both yellowness and beauty are predicates which depend only on the intrinsic nature of what possesses them, yet while yellowness is itself an intrinsic predicate, beauty is not. Indeed it seems to me to be one of the most important truths about predicates of value, that though many of them are intrinsic kinds of value, in the sense I have defined, yet none of them are intrinsic properties, in the sense in which such properties as "yellow" or the property of "being a state of pleasure" or "being a state of things which contains a balance of pleasure" are intrinsic properties. It is obvious, for instance, that, if we are to reject all naturalistic theories of value, we must not only reject those theories, according to which no kind of value would be intrinsic, but must also reject such theories as those which assert, for instance, that to say that a state of mind is good is to say that it is a state of being pleased; or that to say that a state of things is good is to say that it contains a balance of pleasure over pain. There are, in short, two entirely different types of naturalistic theory, the difference between which may be illustrated by the difference between the assertion, "A is good" means "A is pleasant" and the assertion "A is good" means "A is a state of pleasure." Theories of the former type imply that goodness is not an intrinsic kind of value, whereas theories of the latter type imply equally emphatically that it is: since obviously such predicates as that "of being a state of pleasure," or "containing a balance of pleasure," are predicates like "yellow" in respect of the fact that if a given thing possesses them, anything exactly like the thing in question must possess them. It seems to me equally obvious that both types of theory are false: but I do not know how to exclude them both except by saying that two different propositions are both true of goodness, namely: (1) that it does depend only on the intrinsic nature of what possesses it—which excludes theories of the first type and (2) that, though this is so, it is yet not itself an intrinsic property—which excludes those of the second. It was for this reason that I said above that, if there are any intrinsic kinds of value, they would constitute a class of predicates which is, perhaps, unique; for I cannot think of any other predicate which resembles them in respect of the fact, that though not itself intrinsic, it yet shares with intrinsic properties the characteristics of depending solely on the intrinsic nature of what possesses it. So far as I know, certain predicates of value are the only non-intrinsic properties which share with intrinsic properties this characteristic of depending only on the intrinsic nature of what possesses them.
If, however, we are thus to say that predicates of value, though dependent solely on intrinsic properties, are not themselves intrinsic properties, there must be some characteristic belonging to intrinsic properties which predicates of value never possess. And it seems to me quite obvious that there is; only I can't see what it is. It seems to me quite obvious that if you assert of a given state of things that it contains a balance of pleasure over pain, you are asserting of it not only a different predicate, from what you would be asserting of it if you said it was "good"—but a predicate which is of quite a different kind; and in the same way that when you assert of a patch of colour that it is "yellow," the predicate you assert is not only different from "beautiful," but of quite a different kind, in the same way as before. And of course the mere fact that many people have thought that goodness and beauty were subjective is evidence that there is some great difference of kind between them and such predicates as being yellow or containing a balance of pleasure. But what the difference is, if we suppose, as I suppose, that goodness and beauty are not subjective, and that they do share with "yellowness" and "containing pleasure," the property of depending solely on the intrinsic nature of what possesses them, I confess I cannot say. I can only vaguely express the kind of difference I feel there to be by saying that intrinsic properties seem to describe the intrinsic nature of what possesses them in a sense in which predicates of value never do. If you could enumerate all the intrinsic properties a given thing possessed, you would have given a complete description of it, and would not need to mention any predicates of value it possessed; whereas no description of a given thing could be complete which omitted any intrinsic property. But, in any case, owing to the fact that predicates of intrinsic value are not themselves intrinsic properties, you cannot define "intrinsic property," in the way which at first sight seems obviously the right one. You cannot say that an intrinsic property is a property such that, if one thing possesses it and another does not, the intrinsic nature of the two things must be different. For this is the very thing which we are maintaining to be true of predicates of intrinsic value, while at the same time we say that they are not intrinsic properties. Such a definition of "intrinsic property" would therefore only be possible if, we could say that the necessity there is that, if x and y possess different intrinsic properties, their nature must be different, is a necessity of a different kind from the necessity there is that, if x and y are of different intrinsic values, their nature must be different, although both necessities are unconditional. And it seems to me possible that this is the true explanation. But, if so, it obviously adds to the difficulty of explaining the meaning of the unconditional "must," since, in this case, there would be two different meanings of "must," both unconditional, and yet neither, apparently, identical with the logical "must."