Is 'Goodness' a Name of a Simple Non-Natural Quality?

C. D. Broad


Published in the Proceedings of the Aristotelian Society 34 (1933-1934).

    As is well-known, Professor Moore in his Principia Ethica claimed to show that an affirmative answer must be given to the question which forms the title of this paper. There has been a great deal of discussion on the subject during the thirty-one years which have elapsed since the publication of Principia Ethica, and it seemed to me that it might be worth while to review the question on the light of our present knowledge and beliefs.

(1) Statement of Moore's Theory

    I shall begin by stating in my own way what I understand to have been Moore's theory at the time when he wrote Principia Ethica. The theory may be summed up in the following six propositions:

  1. When we use a sentence like: 'That experience is good', we are often, if not always, expressing a judgement in which we ascribe a certain characteristic to the experience. So the word 'good' is often, if not always, used as a name of a characteristic.
  2. The word 'good' is, of course, highly ambiguous. In some of its senses it undoubtedly stands for complex characteristics which can be analysed. When used in any of these senses the word can be defined. In these senses some other word or phrase, such as 'benefic' or 'contributively good' or 'instrumentally efficient', can be substituted for 'good' without loss or gain of meaning.
  3. There is, however, another sense of the word, which is presupposed in some or all of these definable senses of it. This we will call the 'primary sense'. In this sense 'good' stands for a characteristic which is simple and therefore [un]analysable. Consequently the word 'good', in this primary sense, cannot be defined.
  4. It follows at once that the characteristic for which it stands cannot be a relational property, i.e. a characteristic of the form: 'Having the relation R to so-and-so'. For, obviously, all relational properties are complex, and are analysable into a relation and term.
  5. The characteristic is a pure quality and not a pure relation.
  6. The characteristic is of a peculiar kind, which Moore calls non-natural.

    I think that these are the essential points of the theory. They are not all separately stated by Moore, but those which are not stated are clearly implied.

(2) Discussion of the Theory

    I will now take the six points in my statement of the theory in order, and consider the arguments for and against them.

(2.1) Is 'Goodness' a Name of a Characteristic?

    Moore always assumes that 'good' is used as a name of some characteristic or other in sentences like: 'This experience is good'. He evidently thought that this would be admitted by everyone, and that the only question is about the nature of this characteristic or these characteristics.

    Now it has been pointed out by Mr Duncan Jones that it is not safe to let this assumption pass without question. Certainly the sentence: 'This is good' is of the same grammatical form as many sentences which undoubtedly do state that a certain thing has a certain characteristic. It is, e.g., of the same form as: 'This is square', and there is no doubt that anyone who utters the latter sentence is intending to convey the belief that a certain particular has a certain characteristic of which 'square' is a name.

    But we must remember that a sentence, which is grammatically in the indicative mood, may really be in part interjectional or rhetorical or imperative. It may be in part the expression of an emotion which the speaker is feeling. In that case to utter the sentence 'That is good' on a certain occasion might be equivalent to uttering a purely non-ethical sentence in the indicative, followed by a certain interjection. It might, e.g., be equivalent to saying: 'That's an act of self-sacrifice. Hurrah!'. Similarly, to utter the sentence 'That is bad' on a certain occasion might be equivalent to saying: 'That's a deliberately misleading statement. Blast!'. Again, a sentence may be used to evoke a certain kind of emotion in the hearer. In that case to utter the sentence 'That is good' might be equivalent to uttering a purely non-ethical sentence in the indicative in a pleasant tone and with a smile. To utter the sentence 'That is bad' might be equivalent to shouting a purely non-ethical indicative sentence at the hearer with a frown. Here the use of ethical words 'good' and 'bad' is merely a stimulus to produce certain emotions in the hearer, as smiling at him or shouting at him might do. In this case the sentence might be called 'rhetorical'. Lastly, such sentences may be used to command or to forbid certain actions in the hearer. To utter the sentence 'That is good' might be equivalent to uttering a purely non-ethical indicative sentence followed by a sentence in the imperative. It might e.g., be equivalent on a certain occasion to: 'That's an act of self-sacrifice. Imitate it!'. To utter the sentence: 'That is bad' on a certain occasion might be equivalent to saying: 'That's a deliberately misleading statement. Don't do that again!'.

    On this view, words like 'good' and 'bad' do not mean anything in the sense in which words like 'white' and 'square' do. There are no characteristics of which they are names. A person who utters sentences in which they occur as grammatical predicates is not using them to convey the belief that a certain subject has a certain peculiar characteristic of which the grammatical predicate is a name. And a person who hears such sentences and understands them is being exhorted or commanded or emotionally stimulated, but is not receiving any special kind of information about the subject of the sentence. If this be so, Professor Moore's theory breaks down at the first move, and so too do the theories of most of his opponents.

    Mr Duncan Jones points out that his theory fits with two very important facts:
(i) It explains why all attempts to define ethical words in purely non-ethical terms seem unsatisfactory. Suppose you substitute a sentence containing only non-ethical words for one that contains an ethical word. Then the interjectional, rhetorical, or imperative force which the original sentence derived from the ethical word in it, has vanished. You feel that something is missing, and you are quite right. Suppose you have never doubted that ethical words are names of characteristics. Then you will explain this feeling of 'something missing' by saying that the proposed analysis of an ethical characteristic into purely non-ethical characteristics has missed out some essential logical constituent of the ethical characteristic.
(ii) Attempts to define one ethical word, e.g. 'good' partly in terms of another ethical word, e.g. 'right', do not always seem unsatisfactory. It is not, e.g., obviously inadequate to define 'a good experience' as 'an experience which can rightly be desired'. Nor, on the other hand, is it obviously inadequate to define 'right conduct' as 'conduct which is conducive to good consequences.' Now the theory can explain this fact too. Both the original sentence and the proposed equivalent now contain ethical words. Both have therefore interjectional, rhetorical and imperative force. Now it is possible that two different sentences, both of which have this kind of force, may produce precisely similar effects, as evokers of emotion or as commands, in all people of a certain community who may hear them. Suppose you have never doubted that ethical words are names of characteristics. Then you will think that the more complex of two such sentences states the analysis of the ethical characteristic which is named by the ethical word in the simplest of the two sentences. And so you will think that some ethical characteristics can be analysed in terms of other ethical characteristics and of non-ethical characteristics.

    I think that this theory may be further supported by reflecting on how we learn ethical words as children. I suspect that, for a small child, 'good' and 'right' acts are practically co-extensive with those which its mother or nurse names in a certain tone and with a smile or which she exhorts it to do. And 'bad' or 'wrong' acts are practically co-extensive with those which its mother or nurse names in a certain other tone and with a frown or which she forbids it to do. Very soon the ethical words aquire the same rhetorical or imperative force as the tone of voice or the facial expression or the explicit command or forbidding. It may be noted that many words are 'amphibious' in character, i.e. partly non-ethical and partly ethical. Compare, e.g., the two sentences: 'That is a statement made with the intention of producing a false belief', and 'That is a lie'. The first is in purely non-ethical terms. The second has for its grammatical predicate the amphibious word 'lie' which is partly non-ethical and partly ethical. Now it is quite certain that the second sentence does commonly express or stimulate an emotion which the first does not. And it is plausible to hold that this is the only difference between the first, which is purely non-ethical, and the second, which is partly ethical.

    It seems to me then that Mr Duncan Jones's theory is quite plausible enough to deserve very serious consideration. It would have to be refuted before we could be sure that the question: 'Are the characteristics denoted by ethical names analysable or unanalysable?' is a sensible question. If this theory were correct the question would be like asking whether unicorns are or are not cloven-hoofed.

    Henceforth, we will suppose, for the sake of argument, that words like 'good' and 'bad' are names of characteristics. We may agree that when 'good' is used in the sense of 'benefic' or of 'contributively good', it stands for a characteristic which is complex. And we will assume that, when 'good' in these senses is defined, the definition always involves the word 'good' in another sense, which may be called the primary one.

(2.2) Is the Characteristic which 'Good' Denotes Analysable?

    The next question is this: 'Assuming that the word 'good', in the primary sense, is a name of a characteristic, is there any reason to believe that this characteristic is unanalysable?'. It seems to me quite clear that there is no means of proving, with regard to any characterisuc, that it is unanalysable. At most we might be able to show that no analysis so far proposed is satisfactory; and even this is not always so easy as one might think, for the question involves some very fundamental and difficult logical points which I will now try to state.

    Suppose a person raises the question whether the characteristic of which a certain word n is a name is simple or complex, and whether, if it is complex, a certain proposed analysis of it is correct or not. Plainly, in some sense of the phrase, he must 'know what the word n means'. For, otherwise, he does not know what he is asking his question about. Equally plainly this cannot be the same as 'knowing the analysis, if any, of the characteristic which n stands for'. If he knew this in knowing what the word n means, the question whether the characteristic is simple or complex, and what is its correct analysis, if it is complex, could never arise for him. So the question presupposes at least the following three propositions:

  1. That there is a certain one characteristic which the person who asks the question is thinking of whenever he uses the word n in certain kinds of context.
  2. That, whether this characteristic be in fact simple or in fact complex, he can think of it without ipso facto knowing that it is simple or knowing that it is complex.
  3. If it be in fact complex, he can think of it without ipso facto knowing its correct analysis.
In practice a further assumption is, I think, always made. It is assumed
  1. that all or most other people who speak the language of the questioner correctly are thinking of the same characteristic as he is thinking of whenever they use the word n in the same kinds of context.

    Now it might be extremly difficult to justify assumptions (i) and (iv) in many cases. Can I be sure that there is any one characteristic of which I am thinking whenever I use the word 'good' in the primary sense? May there not be, as Mr Braithwaite has suggested, a whole lot of charactcristics, such that I am sometimes thinking of one and sometimes of another of them when I use the word 'good' in the primary sense? Again, can I be sure that, when other people use the word 'good' in certain contexts they are always or generally thinking of the characteristic which I am thinking of when I use the word in such contexts? The only evidence that can be produced is consistency or inconsistency of usage. Do I sometimes call certain things 'good' and at other times call precisely similar things 'bad' or 'indifferent'? Do other people agree among themselves and with me in the things that they call 'good' and in the things that they call 'bad', and in the things that they call 'indifferent'? If there is great inconsistency in applying the words 'good' and 'bad', there is at least a presumption that conditions (i) and (iv) are not fulfilled. Now there certainly is a considerable amount of inconsistency.

    We will suppose, however, that this difficulty can be overcome, and that we can satisfy ourselves that conditions (i) and (iv) are fulfilled. We will now concentrate our attention on conditions (ii) and (iii). There are several grave logical difficulties which I could raise at this point; but I propose to waive them and to pass straight to the following question: 'Supposing that you can think of a certain characteristic c without ipso facto knowing whether it is simple or complex, and without ipso facto knowing its correct analysis if it be complex, how are you to set about answering the question whether it is simple or complex? And, if it is complex, how are you to decide whether a certain proposed analysis of it is right or wrong?'.

    Suppose it is suggested that the characteristic is analysable into tle characteristics cl, c2 and c3. Then
(a) we can reject this suggestion at once if we can think of anything which has c and lacks either cl or c2 or c3. And we can reject it at once if we can think of anything which has cl and c2 and c3 and lacks c.
(b) Suppose that, after applying this test, we are left with one or more suggested analyses of c. We can next proceed as follows. Granted that I know of nothing which has c and lacks any of the characteristics cl, c2 and c3, and that I know of nothing which has cl, c2 and c3 and lacks c, can I conceive that there might be such a thing? If I can, I can reject the proposed analysis of c into cl, c2 and c3. For a characteristic and its analysis would be necessarily co-extensive. The equivalence of their extensions would not be just a contingent fact, like the fact that chewing the cud and having cloven hoofs are co-extensive.
(c) Suppose that we are left with one or more suggested analyses of c which pass this test and can be seen to be necessarily co-extensive with c. There might be several such. The property of being circular, e.g., is necessarily co-extensive with an enormous number of other complicated sets of characteristics. For there are enormous numbers of complicated properties which we can prove must belong to all circles and cannot belong to anything but circles. So we are finally faced with the following question: 'If we know of only one set of characteristics which is necessarily co-extensive with c, how can we tell whether this set is or is not an analysis of c? And, if we know of several such sets of characteristics, how can we tell which, if any, of them is the analysis of c, and which of them are necessarily and reciprocally but synthetically connected with c?'. Suppose, e.g., that it seemed evident that anything that was good would necessarily be a fitting object of desire, and that anything which was a fitting object of desire would necessarily be good. How could we tell whether being a fitting object of desire is the analysis of being good, or whether it is just a complex characteristic which is necessarily and reciprocally but synthetically complicated with goodness?

    It seems to me that, at this stage, further argument would be impossible. All that an objector can say is: 'I feel that your proposed analysis of goodness misses out something which I have in mind when I use the word good'. Or, 'I can't believe that when I use the word good I am thinking of anything so complicated as I should be thinking of if your proposed analysis of goodness were correct'. Now suppose that another person does not feel that the suggested analysis misses out anything that he has in mind when he uses the word 'good'. And suppose that he thinks that what he has in mind when he uses the word 'good' may easily be as complicated as it would be if the suggested analysis were correct. We are assuming that the parties have somehow persuaded themselves that they are both thinking of the same characteristic whenever either of them uses the word 'good' in similar contexts. What further argument is possible between them?

    The actual situation is not, however, quite like this. I think it is true to say that all reasonably simple analyses of goodness in purely non-ethical terms seem to most people to miss out something. (Cf., e.g., 'to be good' means 'to be generally desired as an end'). And all analyses of goodness in purely non-ethical terms, which avoid this defect, seem to most people to be too complex to be correct analyses of what they have in mind. (Cf., e.g., 'to be good' means 'to be something which a man would approve of himself or another for desiring'.) It is only certain definitions which are partly in ethical and partly in non-ethical terms that might seem to many people to avoid both defects. (Cf., e.g., 'to be good' means 'to be a fitting object of desire'.) Now how much weight ought to be attached to a fairly general feeling that suggested analyses of goodness in purely non-ethical terms either miss out something which we have in mind or are too complex to be correct analyses of what we have in mind?

    I think that we commonly make the following assumptions without ever stating them clearly. It is assumed that, if I have thought of a certain characteristic c often enough to have associated a name with it, then any proposed analysis of it which is felt by me to be either inadequate or unduly complex is very probably incorrect. It would be admitted that a proposed analysis might in fact be incorrect even though I did not feel it to be inadequate or unduly complex. But it would be held that, if I do feel it to have either of these defects, then it probably is defective. And, if most people who have frequently thought of a certain characteristic agree in feeling that a proposed analysis of it is inadequate or unduly complex, it would be held to be practically certain that the proposed analysis is defective.

    Now, as regards this general primiple, there are two things to be said:
(i) I am not much impressed with the importance of a widespread feeling that a proposed analysis is unduly complex. We are assuming, it must be remembered, that a person can think of a characteristic without ipso facto knowing its analysis if it has one. Now it seems difficult to suppose that one can estimate the degree of internal complexity of a characteristic when one does not know whether it is simple or complex, and does not know its analysis if it has one.
(ii) More weight should, I think, be attached to a widespread feeling that a proposed analysis is inadequate. This fact has to be accounted for somehow. The most obvious explanation is that the analysis really does omit some logical constituent of the characteristic, or that it analyses, not this characteristic, but some other which is allied to it. Unfortunately this is just the place where Mr Duncan Jones's suggestion becomes highly relevant. It may be that the explanation is simply that the name of the original characteristic has acquired a certain interjectional, rhetorical or emotional force which is lacking in the phrase that expresses the analysis. We feel the lack of this, and we conclude that the analysis is inadequate.

(2.3) Can Goodness be a Relational Property?

    The fourth point in my statement of Professor Moore's theory was that, if the characteristic denoted by 'good' be simple, it cannot be a relational property. It must be either a pure quality or a pure relation This is quite obvious; but, in order to show that goodness is either a quality or a relation, it would be necessary to add the premise that the characteristic denoted by 'good' is simple. I have tried to show that this has not been proved, and that there is no conceivable way of proving it. The utmost that has been shown is that all analyses in purely non-ethical terms, which have so far been suggested, seem to most people to be either inadequate or unduly complex. For reasons which I have given, I do not think that this proves conclusively that none of these proposed analyses is correct; and, even if they all were incorrect, it would would still be possible that there might be a correct analysis in purely non-ethical terms which no one happens to have suggested. Again, it would still be possible that there might be a correct analysis, partly in ethical and partly in non-ethical terms. There is not even a presumption against this, since certain proposed analyses of this kind do not seem to most people to be obviously inadequate or obviously too complex. It seems to me then that no good reason has been produced for holding that the characteristic denoted by 'good', in the primary sense, cannot be a relational property.

(2.4) Can Goodness be a Pure Relation?

    The fifth point in the statement was that 'good', in the primary sense, is not the name of a relation, and must, therefore, be the name of a quality. I think it is obvious that 'good' is not the name of a relation. If it denotes a characteristic at all, the characteristic which it denotes is either a quality or a relational property. So, if one could show that it denotes a simple characteristic, we could admit at once that it denotes a simple quality. The only remark that I wish to make at this point is the following. It does seem to me conceivable that the relation denoted by 'better than' might be more fundamental than the characteristic denoted by 'good'. It might be that the former is simple and unanalysable, and that the latter is complex and definable in terms of the former. The suggestion would be that 'good' is always an abbreviation for 'good of its kind', and that 'good of its kind' means 'better than the average member of its proximate species'. This would make 'good' a name of a relational property of a peculiar kind, in which the relation is that denoted by 'better than'. If it could be shown that 'good', in the primary sense, does not denote a relational property at all, this suggestion could be dismissed at once. But I suspect that some people, who think they have proved this, have not considered the possiblity that 'good' might denote a relational property in which the relation is that denoted by 'better than'. Perhaps they would not be so sure that 'good' might not denote a relational property of this peculiar kind, even though they were convinced that it could not denote a relational property in which any other was involved.

(2.5) Is Goodness a Non-natural Characteristic?

    The last point in Professor Moore's theory is that 'good', in the primary sense, is a name of a 'non-natural' characteristic. Two questions at once arise:

  1. What exactly is meant by the distinction between a 'natural' and a 'non-natural' characteristic?
  2. What connection, if any, is there between the doctrine that 'good', in the primary sense, denotes a characteristic which is simple and unanalysable, and the doctrine that it denotes a characteristic which is non-natural?
We will take these two questions in turn.

    (i) We will begin with complex characteristics. A complex characteristic is natural if it can be analysed into a set of simple characteristics every one of which is natural. A complex characteristic is non-natural if its analysis involves at least one simple characteristic which is non-natural. Thus the question at once arises: 'What is meant by calling a simple characteristic natural or non-natural?'.

    Unfortunately we shall get very little light on this question from Professor Moore's published works. The only place, so far as I know, in which it is explicitly discussed is Principia Ethica, pp. 40 to 41. We are there told that a 'natural object' is any object that is capable of existing in time, e.g. a stone, a mind, an explosion, an experience, etc. All natural objects have natural characteristics, and some natural objects also have non-natural characteristics. We are told that each natural characteristic of a natural object could be conceived as existing in time all by itself, and that every natural object is a whole whose parts are its natural characteristics. We are told that a non-natural characteristic of a natural object is one which cannot be conceived as existing in time all by itself. It can be conceived as existing only as the property of some natural object. Now it seems to me that every characteristic of a natural object answers Professor Moore's criterion of non-naturalness, and that no characteristic could possibly be natural in his sense. I do not believe for a moment that a penny is a whole of which brownness and roundness are parts, nor do I believe that the brownness or the roundness of a penny could each exist in time all by itself. Hence I should have to count brownness, roundness, pleasantness, etc., as non-natural characteristics if I accepted Professor Moore's account of the distinction. Yet he certainly counts them as natural characteristics.

    I think that Professor Moore is intending to explain the distinction between natural and non-natural characteristics in the very difficult essay entitled 'The Conception of Intrinsic Value,' in his Philosophical Studies. So far as I can understand his doctrine in that essay, it may be summarized as follows:

  1. The characteristics of any thing T may be first divided into two great classes, viz., those which do, and those which do not, 'depend solely on the intrinsic nature of T'.
  2. Characteristics of a thing T which depend solely on its intrinsic nature may be sub-divided into those which are, and those which are not, 'intrinsic characteristics' of it. Consider, e.g., an experience which has a certain perfectly determinate kind and degree of pleasantness. Suppose that it also has a certain perfectly determinate kind and degree of goodness. Then, if I understand him aright, Moore would say that both its pleasantness and its goodness are characteristics which depend solely on its intrinsic nature. He would say that its pleasantness is an intrinsic characteristic of it. And he would say that its goodness is not an intrinsic characteristic of it.
  3. Although he does not explicitly say so, I think that he would identify the non-natural characteristics of a thing with those which are determined solely by its intrinsic nature and yet are not intrinsic. The natural characteristics of a thing would be those which are either intrinsic or are not determined solely by its intrinsic nature.

    Unhappily Moore gives no clear account of this distinction between the intrinsic and the non-intrinsic characteristics which depend on the intrinsic nature of a thing. All that he says is this. A complete enumeration of the intrinsic characteristics of a thing would constitute a complete description of it. A description of a thing can be complete even if it does not include characteristics of it which, though determined solely by its intrinsic nature, are not intrinsic characteristics. E.g. a pleasant experience, which is also good, could not be completely described if its pleasantness were not mentioned. But it could be completely described without its goodness being mentioned.

    I find it most difficult to follow or to accept this. I am inclined to think that the fact which Moore has in mind is that goodness, in the primary sense, is always dependent on the presence of certain non-ethical characteristics, which I should call 'good-making'. If an experience is good, this is never an ultimate fact. It is always reasonable to ask: 'What makes it good?'. And the sort of answer that we should expect to get would be: 'Its pleasantness', or: 'The fact that it is a sorrowfully toned awareness of another's distress', or something of that kind. We might, therefore, distinguish the characteristics of a thing into two classes, viz. ultimate and derivative; and goodness will certainly fall into the class of derivative characteristics. Now there is a sense in which one might say that a thing could not be completely described if any of its ultimate characteristics were omitted, but that it could be completely described without mentioning any of its derivative characteristics. In describing a circle, e.g., it is not necessary to mention any of the innumerable properties which follow of necessity from its definition together with the axioms of Euclidean geometry.

    But, although this analogy may throw some light on what Professor Moore had in mind, it certainly does not help us to understand what is meant by saying that goodness is a non-natural characteristic and that pleasantness, e.g., is a natural characteristic. For it is surely quite as evident that pleasantness and unpleasantness are derivative characteristics as that goodness and badness, in the primary sense, are so. If an experience is pleasant, it is always reasonable to ask: 'What makes it pleasant?'. And the sort of answer that we should expect is: 'Its sweetness' or: 'The way in which various sounds are combined in it', or something of that kind. So, if pleasantness is to be counted as a natural characteristic, it is impossible to identify the non-natural characteristics of a thing with the derivative subclass of those of its characteristics which depend solely on its intrinsic nature.

    It seems impossible, then, to extract from Professor Moore's writings any satisfactory account of his distinction between 'natural' and 'non-natural' characteristics. And yet we all recognize fairly well what he is talking about when he makes this distinction. I suggest that the best plan is to start with an epistemological description of the term 'natural characteristic'. I propose to describe a 'natural' characteristic as any characteristic which either

  1. we become aware of by sensing sensa which manifest it or by introspecting experiences which manifest it; or
  2. is definable wholly in terms of such characteristics and the notions of cause and substance.
I think that this covers every characteristic which would be universally admitted to be natural. It would cover, e.g., yellowness, both in the sense in which it is ascribed to sensa and in the sense in which it is ascribed to physical things. It would also cover pleasantness, fearfuhless, intelligence, etc. And it would leave the question whether goodness is a natural or a non-natural characteristic open to discussion. We will therefore take this as our description of a 'natural characteristic' for the rest of the argument.

    (ii) We are now in a position to deal with our second question. What connection, if any, is there between the doctrine that 'good', in the primary sense, denotes a characteristic which is simple, and the doctrine that it denotes a charactcristic which is non-natural?

    It is plain that our epistemological description at once plunges us into questions about how we become aware of the characteristic called 'goodness', assuming that there is such a characteristic.

    (a) It seems to me quite obvious that it is not manifested to us by any of our senses, as, e.g., yellowness, sweetness, squeakiness, etc., are. It is evident that, when 'good' is used in its primary sense, it does not denote a characteristic of which we could become aware by sight or touch or taste or smell or hearing, or any other sense which we have or conceivably might have. It is doubtful whether goodness, in this sense, can belong to the sort of objects that can be sensed or perceived. And, even if it can and does, it is certain that we do not perceive with our senses the goodness of such objects. At most we perceive with our senses certain natural characteristics which are good-making, e.g, certain combinations of colour, of sound, of taste, etc.

    (b) It seems equally clear that no simple characteristic which we can discover by introspecting our experiences can be identified with goodness. We become aware through introspection of experiences which are pleasant or unpleasant, toned with desire or aversion, and so on. We thus become aware of the psychological characteristics of pleasantness, longingness, etc., and their opposites. Now it is true that goodness, in the primary sense, can belong to experiences. Indeed, some people would hold that, in this sense, it can belong to nothing else. Yet I think that a moment's reflection will convince one that by calling an experience 'good' we do not mean that it is pleasant or approving, or that it has any of the other simple psychological qualities of which we become aware through introspecting our experiences. If anyone is tempted to identify goodness with one of these simple psychological qualities, I think that he does so through a confusion. What he really believes is that there is one and only one good-making quality of an experience, e.g. pleasantness. He then fails to notice the distinction between goodness itself and the one and only good-making quality and so he thinks he believes that 'good' and 'pleasant', e.g., are just two names for a single characteristic. And, since pleasantness certainly is a natural characteristic, he will think he believes that 'good' is the name of a natural characteristic. I do not think that the belief that one means the same by 'good' and 'pleasant', e.g., would survive for a moment after the distinction between goodness itself and a good-making characteristic had been pointed out to one. And similar remarks would apply to any other simple psychological characteristic denoted by 'good'.

    We come, therefore, to the following hypothetical conclusion. If the word 'good', when used in its primary sense, denotes a simple quality, then that quality is almost certainly not one which we become aware of either by sensing sensa which manifest it or by introspecting experiences which manifest it. It is, therefore, not a natural characteristic, as described by us. So, with our description of 'natural characteristics', there is an important logical connection between proving that 'good' is the name of a simple quality and proving that it is the name of a non-natural characteristic.

    'I'his, however, does not settle the question whether 'good' is the name of a non-natural characteristic. For I do not think that it has been proved or could be proved, that 'good' is the name of a simplc quality. Indeed, I am now going to argue that there are considerable epistemological difficulties in holding that 'good' is the name of a simple quality. Is there any way of becoming aware of a simple quality belonging to particulars, except by sensing or introspecting particulars which manifest this characteristic to one? Many people would say that there plainly is no other way. If they are right, it follows that we could not possibly have an intuitive idea of goodness if goodness were a non-natural characteristic. For, if goodness were a simple non-natural characteristic, the consequence would follow at once; and, if it were a complex characteristic which contains one or more non-natural characteristics in its analysis, the consequence would follow at the second move. We could not have an intuitive idea of such a complex characteristic unless we had such ideas of its simple non-natural components; and, if the epistemological principle be accepted, we could not have intuitive ideas of these components.

    Now, although this epistemological principle does seem to me highly plausible, I am not prepared to accept it (or any other epistemological principle) as self-evident. I am therefore not prepared to conclude that no characteristic of which I can have an intuitive idea could be non-natural. But I do think it important to point out the following hypothetical fact. If goodness is a non-natural characteristic, then anyone's intuitive idea of this characteristic must be an a priori notion, or must contain a priori notions as elements. For an a priori notion just is an intuitive idea of a characteristic which is not manifested in sensation or introspection and is not definable wholly in terms of such characteristics. Anyone who holds that goodness is a non-natural characteristic and that he has an intuitive idea of it is therefore committed to the view that there are a priori notions and that his notion of goodness is one of them. Now anyone who holds that goodness is a simple characteristic will be almost compelled to hold that it is non-natural. Therefore anyone who holds that goodness is a simple characteristic and that he has an intuitive idea of it will be almost compelled to hold that there are a priori notions and that his notion of goodness is one them.

    There is one other epistemological point to be noticed. Suppose that a person regards goodness as a non-natural characteristic, and admits that it is always dependent on the presence of certain natural characteristics which are good-making. Then, if he holds that the connection between a good-making characteristic and the goodness which it confers is necessary, he will be obliged to hold that there are synthetically necessary facts and that he knows some of them. He will therefore be obliged to admit that he can make synthetically a priori judgments. The necessary connection between those natural characteristics of a thing which are good-making and the goodness which their presence necessarily confers on the thing cannot be analytic. For this would involve that the non-natural characteristic of goodness is contained as a factor in the analysis of a purely natural good-making characteristic, and this would be self-contradictory.

    Now people think it self-evident that all necessary connection must be analytic and that there can be no synthetic a priori judgement. I do not find this principle in the least self-evident myself; but it is worth while to point out that anyone who does so will be compelled to hold either

  1. that goodness is a natural characteristic, or
  2. that the connection between the goodness of a thing and its good-making characteristics is purely contingent and is known only empirically.
He might, of course, combine both views, as Hume did.


Transcribed into hypertext by Andrew Chrucky, Oct. 23, 1998.
Go to: Philosophical Alternatives from C. D. Broad