R. M. Hare, The Language of Morals (1952)

PART II

'GOOD'

'Good. . . . The most general adjective of commendation, implying the existence in a high, or at least satisfactory, degree of characteristic qualities which are either admirable in themselves, or useful for some purpose . . .'

Oxford English Dictionary

5
'NATURALISM'

5.1. The first part of this book has served two purposes. First, by examining in some detail the language used for expressing commands -- the simplest form of prescription -- we are now in a better position to understand the more complex logical behaviour of value-words, the other main instrument for prescribing with which our language provides us. Secondly, we have, in the course of this examination, had occasion to look at some of the kinds of situation in which we are accustomed to use prescriptive language, and seen how we do learn to answer questions of the form 'What shall I do?', to which the answer is a prescription.

In the remainder of the book I shall be dealing with some typical value-words, and especially with 'good', 'right', and 'ought'. Although my selection is conventional, three explanations require to be made here. First, I do not wish to imply that the characteristics of value-words to which I shall draw attention are confined to the few typical words that are here examined; it is in fact the case -- and this has been productive of logical confusion -- that almost every word in our language is capable of being used on occasion as a value-word (that is, for commending or its opposite); and usually it is only by cross-examining a speaker that we can tell whether he is so using a word. The word 'brilliant' is a good example. In confining attention to the simplest, most typical and most general value-words, my only object is simplicity of exposition. Secondly, the terms 'value-words' and 'evaluative' are exceedingly hard to define. I shall for the time being content myself with giving examples and illustrations; it is not until later (11.2) that I shall be able to hazard a definition, and even then without much confidence. Thirdly, I shall follow a procedure similar to that used earlier in connexion with the learning of principles; I shall illustrate the peculiarities of value-words by examples drawn from their non-moral uses, and only later ask whether these same peculiarities are to be found in moral contexts. This procedure, though it may seem perverse, has one great advantage; it will enable me, I hope, to show that the peculiarities of these words have nothing to do with morals as such, and that therefore theories which purport to explain them have to be applicable, not only to expressions like 'good man', but also to expressions like 'good chronometer'; and to realize this is to be preserved from a number of errors.

5.2. Let me illustrate one of the most characteristic features of value-words in terms of a particular example. It is a feature sometimes described by saying that 'good' and other such words are the names of 'supervenient' or 'consequential' properties. Suppose that a picture is hanging upon the wall and we are discussing whether it is a good picture; that is to say, we are debating whether to assent to, or dissent from, the judgement 'P is a good picture'. It must be understood that the context makes it clear that we mean by 'good picture' not 'good likeness' but 'good work of art' -- though both these uses would be value-expressions.

First let us notice a very important peculiarity of the word 'good' as used in this sentence. Suppose that there is another picture next to P in the gallery (I will call it Q). Suppose that either P is a replica of Q or Q of P, and we do not know which, but do know that both were painted by the same artist at about the same time. Now there is one thing that we cannot say; we cannot say 'P is exactly like Q in all respects save this one, that P is a gooff picture and Q not'. If we were to say this, we should invite the comment,

'But how can one be good and the other not, if they are exactly alike? There must be some further difference between them to make one good and the other not.'
Unless we at least admit the relevance of the question 'What makes one good and the other not?' we are bound to puzzle our hearers; they will think that something has gone wrong with our use of the word 'good'. Sometimes we cannot specify just what it is that makes one good and the other not; but there always must be something. Suppose that in the attempt to explain our meaning we said: 'I didn't say that there was any other difference between them; there is just this one difference, that one is good and the other not. Surely you would understand me if I said that one was signed and the other not, but that there was otherwise no difference? So why shouldn't I say that one was good and the other not, but that there was otherwise no difference?' The answer to this protest is that the word 'good' is not like the word 'signed'; there is a difference in their logic.

5.3. The following reason might be suggested for this logical peculiarity: there is some one characteristic or group of characteristics of the two pictures on which the characteristic 'good' is logically dependent, so that, of course, one cannot be good and the other not, unless these characteristics vary too. To quote a parallel case, one picture could not be rectangular and the other not, unless certain other characteristics also varied, for example the size of at least one of the angles. And so a natural response to the discovery that 'good' behaves as it does, is to suspect that there is a set of characteristics which together entail a thing being good, and to set out to discover what these characteristics are. This is the genesis of that group of ethical theories which Professor Moore called ^naturalist' -- an unfortunate term, for as Moore says himself, substantially the same fallacy may be committed by choosing metaphysical or suprasensible characteristics for this purpose.1 Talking about the supernatural is no prophylactic against 'naturalism'. The term has, unfortunately, since Moore's introduction of it, been used very loosely. It is best to confine it to those theories against which Moore's refutation (or a recognizable version of it) is valid. In this sense most 'emotive' theories are not naturalist, though they are often called so. Their error is a quite different one. I shall argue below (11.3) that what is wrong with naturalist theories is that they leave out the prescriptive or commendatory element in value-judgements, by seeking to make them derivable from statements of fact. If I am right in this opinion, my own theory, which preserves this element, is not naturalist.

We have to inquire, then, whether there is any characteristic or group of characteristics which is related to the characteristic of being good in the same way as the angle-measurements of figures are related to their rectangularity. In what way are the latter related? This involves answering the question: Why cannot it be the case that one picture is rectangular and the other not unless the angle-measurements of the two pictures also differ? The answer is clearly that 'rectangular' means 'rectilinear and having all its angles of a certain size, namely, 90 degrees'; and, therefore, that when we have said that one picture is rectangular and the other not, we have said that the measurements of their angles differ; and if we then go on to say that they do not differ, we contradict ourselves. Therefore, to say 'P is exactly like Q in all respects save this one, that P is a rectangular picture and Q not', may be self-contradictory; whether it is self-contradictory depends on what we intend to include in 'all respects'. If we intend to include the measurements of the angles, then the sentence is self-contradictory; for it is self-contradictory to say 'P is exactly like Q in all respects, including the measurements of its angles, save this one, that P is a rectangular picture and Q not'; this contains the assertion that the angles of P both differ and do not differ from those of Q.

Thus the impossibility that we are speaking of is a logical one, which depends upon the meaning of the word 'rectangular'. This is a very elementary example of a logical impossibility; there are other more complex examples. Those who in recent times have denied that there can be synthetic a priori truth, have been asserting that all a priori impossibility can be shown to be of this character, i.e. dependent on the meanings assigned to the words used. Whether they are right is still a matter under dispute; but for the purpose of my argument I shall assume that they are. The dispute has reached the stage when it cannot be argued on abstract grounds alone, but only by the painstaking analysis of particular sentences which are claimed to be true a priori and yet synthetic.2

5.4. Let us then ask whether 'good' behaves in the way that we have noticed for the same reason that 'rectangular' does; in other words, whether there are certain characteristics of pictures which are defining characteristics of a good picture, in the same way as 'having all its angles 90 degrees and being a rectilinear plane figure' are defining characteristics of a rectangle. Moore thought that he could prove that there were no such defining characteristics for the word 'good' as used in morals. His argument has been assailed since he propounded it; and it is certainly true that the formulation of it was at fault. But it seems to me that Moore's argument was not merely plausible; it rests, albeit insecurely, upon a secure foundation; there is indeed something about the way in which, and the purposes for which, we use the word 'good' which makes it impossible to hold the sort of position which Moore was attacking, although Moore did not see clearly what this something was. Let us, therefore, try to restate Moore's argument in a way which makes it clear why 'naturalism' is untenable, not onty for the moral use of 'good' as he thought, but also for many other uses.

Let us suppose for the sake of argument that there are some 'defining characteristics' of a good picture. It does not matter of what sort they are; they can be a single characteristic, or a conjunction of characteristics, or a disjunction of alternative characteristics. Let us call the group of these characteristics C. 'P is a good picture' will then mean the same as 'P is a picture and P is C'. For example, let C mean 'Having a tendency to arouse in people who are at that time members of the Royal Academy (or any other definitely specified group of people), a definitely recognizable feeling called "admiration" '. The words 'definitely specified' and 'definitely recognizable' have to be inserted, for otherwise we might find that words in the definiens were being used evaluatively, and this would make the definition no longer 'naturalistic'. Now suppose that we wish to say that the members of the Royal Academy have good taste in pictures. To have good taste in pictures means to have this definitely recognizable feeling of admiration for those pictures, and only those pictures, which are good pictures. If therefore we wish to say that the members of the Royal Academy have good taste in pictures, we have, according to the definition, to say something which means the same as saying that they have this feeling of admiration for pictures which have a tendency to arouse in them this feeling.

Now this is not what we wanted to say. We wanted to say that they admired good pictures; we have succeeded only in saying that they admired pictures which they admired. Thus if we accept the definition we debar ourselves from saying something that we do sometimes want to say. What this something is will become apparent later; for the moment let us say that what we wanted to do was to commend the pictures which the members of the Royal Academy admired. Something about our definition prevented our doing this. We could no longer commend the pictures which they admired, we could only say that they admired those pictures which they admired. Thus our definition has prevented us, in one crucial case, from commending something which we want to commend. That is what is wrong with it.

Let us generalize. If 'P is a good picture' is held to mean the same as 'P is a picture and P is C, then it will become impossible to commend pictures for being C; it will be possible only to say that they are C. It is important to realize that this difficulty has nothing to do with the particular example that I have chosen. It is not because we have chosen the wrong defining characteristics; it is because, whatever defining characteristics we choose, this objection arises, that we can no longer commend an object for possessing those characteristics.

Let us illustrate this by another example. I am deliberately excluding for the moment moral examples because I want it to be clear that the logical difficulties which we are encountering have nothing to do with morals in particular, but are due to the general characteristics of value-words. Let us consider the sentence 'S is a good strawberry'. We might naturally suppose that this means nothing more than 'S is a strawberry and S is sweet, juicy, firm, red, and large'. But it then becomes impossible for us to say certain things which in our ordinary talk we do say. We sometimes want to say that a strawberry is a good strawberry because it is sweet, &c. This -- as we can at once see if we think of ourselves saying it -- does not mean the same as saying that a strawberry is a sweet, &c, strawberry because it is sweet, &c. But according to the proposed definition this is what it would mean. Thus here again the proposed definition would prevent our saying something that we do succeed in saying meaningfully in our ordinary talk.

5.5. It has sometimes been alleged against Moore's refutation of naturalism that it proves too much -- that if it were valid for 'good' it would be valid for any word whatever that is claimed to be definable in terms of other words. Certain phrases of Moore's lay him open to this objection, especially his quotation of Butler's slogan 'Everything is what it is, and not another thing'.3 Of course, what the naturalists are claiming is that 'goodness' is not 'another thing' than the characteristics which they claim to be its defining characteristics. If naturalism were true and were consistently held, the naturalist could argue as follows: 'When I say that x is a good A and when I say that it is an A which is C, I am saying one and the same thing, just as when I say that y is a puppy and when I say that y is a young dog I am saying one and the same thing. A refutation on your lines could be produced of the theory that "puppy" means "young dog". It would proceed as follows: If you accept this definition, then the sentence "A puppy is a young dog" becomes equivalent to "A young dog is a young dog", and this is something that we would never want to say; but we do sometimes say "A puppy is a young dog"; therefore the proposed definition prevents our saying something that in our ordinary talk we do meaningfully say, &c.'

In order to answer this objection, let us inquire on what occasions and for what purpose we use the sentence 'A puppy is a young dog'. It is, I think, clear that we should normally use this sentence as a definition; we should use it when we were explaining what a puppy was or what the word 'puppy' meant. It is not a sentence that would normally be used to say anything of substance about puppies, though I shall consider in a moment one such possible use. Thus this sentence has little difference in meaning, if any, from the original definition ' "Puppy" means "young dog" '. This does not imply that either form of the definition has anything wrong with it as a definition. A definition, if it is a correct one, is always analytic in one sense and synthetic in another. Taken as a sentence about puppies it is analytic; taken as a sentence about the word 'puppy' it is synthetic. It is never a synthetic sentence about puppies; if it were it would not be a definition but something else.

This may be made clear by a consideration of our example. The sentence 'A puppy is a young dog', although it is normally used as a definition of the word 'puppy', is none the less misleading in its form; for it has the same form as some sentences which are not definitions -- e.g. 'A puppy is a queer thing to find inside a beer-barrel'. It is misleading because it is elliptical, and this obscures the fact that it is a definition. We could correct both these faults at the cost of a certain artificiality by saying instead 'The English sentence "If anything is a puppy it is a young dog (and vice versa)" is analytic'. This has the merit of disentangling the synthetic from the analytic elements of the original definition. The part within inverted commas is analytic if the definition is correct; for the function of the definition is to say that it is analytic. On the other hand, the whole sentence is not analytic; it is a synthetic assertion about the part within inverted commas; we find out whether the assertion is correct or not by studying English usage. Thus the whole sentence is a synthetic assertion about words; the part within inverted commas has the form of an assertion about puppies, but asserts nothing about them because it is analytic. Nowhere in the whole sentence is there a synthetic assertion about puppies.

5.6. There is one conceivable case in which 'A puppy is a young dog' might be used to make a synthetic assertion about puppies. It might be parallel to 'A tadpole is a young frog (or other batrachian)', which might be used to inform a person that the class of animals which he had learnt to distinguish by the name 'tadpole' did in fact turn into frogs when they grew up. But this case cannot be used in support of the objection. Suppose that it were objected 'You cannot refute naturalism in the way that you seek to; for in this case you would have to abandon also the definition of "tadpole" as "young frog"; it could be argued that the sentence "A tadpole is a young frog", which we all agree can be used to make a synthetic assertion about tadpoles (namely that they grow into frogs), is, according to this definition, a mere tautology'. It is not difficult to see that this objection rests upon an equivocation. We cannot at the same time maintain that 'tadpole' means the same as 'young frog' and that 'A tadpole is a young frog' is a synthetic assertion. Either we have to define 'tadpole' independently of 'young frog' (for example by ostensive definition, by pointing at a lot of tadpoles swimming in the pond), in which case 'A tadpole is a young frog' will indeed be a synthetic assertion, but 'tadpole' will mean not 'young frog' but 'the sort of animal you can see swimming in the pond there'; or else we have to define 'tadpole' as 'young frog', in which case 'A tadpole is a young frog' becomes analytic and 'Those are tadpoles swimming in the pond there' becomes, not an ostensive definition, but a statement of fact to the effect that those animals swimming in the water will turn into frogs when they grow up. In fact, of course, we learn the meaning of 'tadpole' in both these ways, and it is to that extent equivocal. This does not worry us, because cases do not arise in which animals just like these turn, not into frogs, but into, say, snakes; but if we did find a species of snake that had young just like a tadpole, we should have to make the distinction, by saying 'Before you can tell whether an animal like this is really a tadpole, you have to wait and see whether it turns into a frog or a snake'; or we might adopt other expedients. This is a familiar puzzle in logic; we shall have occasion to recur to it later (7.5; 11.2). It is possible to argue that there is a similar 'equivocation' about the word 'good'; for, as we shall see, it has both descriptive and evaluative force, and these have to be learnt by different means and independently of one another. But we are not yet in a position to explain this.

Here it will suffice to point out that if it is interpreted in this way the objection misses the point. For my argument is that we cannot say that 'x is a good A' means the same as 'x is an A which is C, because then it becomes impossible to commend A's which are C by saying 'A's which are C are good A's'. In the 'tadpole' case the parallel argument would be 'You cannot say that "x is a tadpole" means the same as "x is a young frog", because then it becomes impossible to say that tadpoles turn into frogs by saying "A tadpole is a young frog".' But, of course, if we do stick to the definition of 'tadpole' as equivalent to 'young frog', then it is indeed impossible to say this; it is only because 'tadpole' is sometimes used otherwise than according to this definition, that we are able sometimes to use 'A tadpole is a young frog' as a synthetic assertion. And similarly, it is because 'good' is sometimes (indeed in almost all cases) used otherwise than according to 'naturalistic' definitions, that we can use it in order to commend.

5.7. But let us return to the sentence 'A puppy is a young dog', and, neglecting the possible synthetic use which we have been considering, confine our attention to its use as a definition of 'puppy'. The objection which we are considering maintains that we do sometimes meaningfully say 'A puppy is a young dog', and that by this we do not mean the same as we would if we said 'A young dog is a young dog'. Let us therefore expand both these sentences in the way previously suggested. They become, respectively, 'The English sentence "If anything is a puppy it is a young dog" is analytic' and 'The English sentence "If anything is a young dog it is a young dog" is analytic'. Both these sentences are true, but they do not mean the same; and it is interesting to notice that here is one case in which, although 'puppy' means the same as 'young dog', they cannot be substituted for one another without change of meaning. But this is not in the least paradoxical. It is well known that, if a sentence contains another sentence within it in inverted commas, it is not always possible without changing the meaning of the whole sentence to substitute synonymous expressions for expressions inside the inverted commas. Thus the sentence 'He said "It is a puppy"' does not mean the same as the sentence 'He said "It is a young dog"'; for his actual words are being reported, and it makes a difference what they were. Similarly, the sentence 'It says in the dictionary "Puppy: young dog"' is not the same in meaning as the sentence 'It says in the dictionary "Young dog: young dog"'. Similarly, again, the sentence 'When Englishmen say "puppy" they mean the same as "young dog"' does not have the same meaning as the sentence 'When Englishmen say "young dog" they mean the same as "young dog"'. And so, also, 'The English sentence "If anything is a puppy it is a young dog" is analytic' does not mean the same as 'The English sentence "If anything is a young dog it is a young dog" is analytic'. And therefore the abbreviations of these sentences, 'A puppy is a young dog' and 'A young dog is a young dog' do not mean the same.

But all this is entirely irrelevant to the case of the word 'good'. The force of the objection was, that our attack upon naturalistic definitions of the word 'good' could be pressed equally against definitions of the word 'puppy', and that, since these latter are obviously in order, there must be something wrong with the attack. Now our attack upon naturalistic definitions of 'good' was based upon the fact that if it were true that 'a good A' meant the same as 'an A which is C, then it would be impossible to use the sentence 'An A which is C is good' in order to commend A's which are C; for this sentence would be analytic and equivalent to 'An A which is C is C Now it seems clear that we do use sentences of the form 'An A which is C is good' in order to commend A's which are C; and that when we do so, we are not doing the same sort of thing as when we say 'A puppy is a young dog'; that is to say, commending is not the same sort of linguistic activity as defining. The meaning of expressions like 'A puppy is a young dog' is preserved by expanding them into overt definitions like 'The English sentence "If anything is a puppy it is a young dog" is analytic'. This latter sentence is true, and is verifiable by consulting the usage of educated Englishmen. Which Englishmen are to count as educated is, of course, a value-question about proper word-usage, but that is not here relevant. On the other hand, a sentence of the form 'An A which is C is good' cannot without change of meaning be rewritten 'The English sentence "An A which is C is good" is analytic'. For a sentence of the latter type certainly could not be used for commending, whereas sentences of the former type can be and are; we commend strawberries which are sweet, &c, by saying 'A strawberry which is sweet, &c, is good', but we never do this by saying 'The English sentence "A strawberry which is sweet, &c, is good" is analytic'. This latter sentence, if it were used, would not be a commendation of sweet strawberries; it would be a remark -- and a false one -- about the English language.

5.8. Thus it is not true to say that the means used to upset naturalistic definitions of value-terms could be used equally to upset any definition. Value-terms have a special function in language, that of commending; and so they plainly cannot be defined in terms of other words which themselves do not perform this function; for if this is done, we are deprived of a means of performing the function. But with words like 'puppy' this does not apply; one may define 'puppy' in terms of any other words which will do the same job. Whether two expressions will do the same job is decided by reference to usage. And since what we are trying to do is to give an account of the word 'good' as it is used -- not as it might be used if its meaning and usage were changed -- this reference is final. It is therefore no answer to the above argument to claim that a 'naturalist' might if he pleased define 'good' in terms of some characteristics of his choice. Such an arbitrary definition is quite out of place here; the logician is, it is true, at liberty to define his own technical terms as he pleases, provided that he makes it clear how he is going to use them. But 'good' in this context is not a technical term used for talking about what the logician is talking about; it itself is what he is talking about; it is the object of his study, not the instrument. He is studying the function of the word 'good' in language; and so long as he wishes to study this, he must continue to allow the word the function which it has in language, that of commending. If by an arbitrary definition he gives the word a diffefent function from that which it now has, then he is not studying the same thing any longer; he is studying a figment of his own devising.

Naturalism in ethics, like attempts to square the circle and to 'justify induction', will constantly recur so long as there are people who have not understood the fallacy involved. It may therefore be useful to give a simple procedure for exposing any new variety of it that may be offered. Let us suppose that someone claims that he can deduce a moral or other evaluative judgement from a set of purely factual or descriptive premisses, relying on some definition to the effect that V (a value-word) means the same as C (a conjunction of descriptive predicates). We first have to ask him to be sure that C contains no expression that is covertly evaluative (for example 'natural' or 'normal' or 'satisfying' or 'fundamental human needs'). Nearly all so-called 'naturalistic definitions' will break down under this test -- for to be genuinely naturalistic a definition must contain no expression for whose applicability there is not a definite criterion which does not involve the making of a value-judgement. If the definition satisfies this test, we have next to ask whether its advocate ever wishes to commend anything for being C. If he says that he does, we have only to point out to him that his definition makes this impossible, for the reasons given. And clearly he cannot say that he never wishes to commend anything for being C; for to commend things for being C is the whole object of his theory.


Notes

1 Principia Ethica, p. 39.

2 An excellent example of such analysis is to be found in an article on 'The Incongruity of Counterparts', by D. F. Pears, Mind, Ixi (1952), 78.

3 For an excellent criticism of this side of Moore's refutation, see A, N. Prior, Logic and the Basis of Ethics, ch. i.