R. M. Hare, The Language of Morals (1952)
PART I
THE IMPERATIVE MOOD
'Virtue, then, is a disposition governing our
choices'.ARISTOTLE, Eth. Nic. 1106b 36
1
PRESCRIPTIVE LANGUAGE1.1. If we were to ask of a person 'What are his moral principles?' the way in which we could be most sure of a true answer would be by studying what he did. He might, to be sure, profess in his conversation all sorts of principles, which in his actions he completely disregarded; but it would be when, knowing all the relevant facts of a situation, he was faced with choices or decisions between alternative courses of action, between alternative answers to the question 'What shall I do?', that he would reveal in what principles of conduct he really believed. The reason why actions are in a peculiar way revelatory of moral principles is that the function of moral principles is to guide conduct. The language of morals is one sort of prescriptive language. And this is what makes ethics worth studying: for the question 'What shall I do?' is one that we cannot for long evade; the problems of conduct, though sometimes less diverting than crossword puzzles, have to be solved in a way that crossword puzzles do not. We cannot wait to see the solution in the next issue, because on the solution of the problems depends what happens in the next issue. Thus, in a world in which the problems of conduct become every day more complex and tormenting, there is a great need for an understanding of the language in which these problems are posed and answered. For confusion about our moral language leads, not merely to theoretical muddles, but to needless practical perplexities.
An old-fashioned, but still useful, way of studying anything is per genus et diffierentiam; if moral language belongs to the genus 'prescriptive language', we shall most easily understand its nature if we compare and contrast first of all prescriptive language with other sorts of language, and then moral language with other sorts of prescriptive language. That, in brief, is the plan of this book. I shall proceed from the simple to the more complex. I shall deal first with the simplest form of prescriptive language, the ordinary imperative sentence. The logical behaviour of this type of sentence is of great interest to the student of moral language because, in spite of its comparative simplicity, it raises in an easily discernible form many of the problems which have beset ethical theory. Therefore, although it is no part of my purpose to reduce moral language to imperatives, the study of imperatives is by far the best introduction to the study of ethics; and if the reader does not at once see the relevance to ethics of the earlier part of the discussion, I must ask him to be patient. Neglect of the principles enunciated in the first part of this book is the source of many of the most insidious confusions in ethics.
From singular imperatives I shall proceed to universal imperatives or principles. The discussion of these, and of how we come to adopt or reject them, will give me an opportunity of describing the processes of teaching and learning, and the logic of the language that we use for these purposes. Since one of the most important uses of moral language is in moral teaching, the relevance of this discussion to ethics will be obvious.
I shall then go on to discuss a kind of prescriptive language which is more nearly related to the language of morals than is the simple imperative. This is the language of non-moral value-judgements -- all those sentences containing words like 'ought', 'right', and 'good' which are not moral judgements. I shall seek to establish that many of the features which have caused trouble to students of ethics are also displayed by these sorts of sentence -- so much so that a proper understanding of them does much to elucidate the problems of ethics itself. I shall take the two most typical moral words 'good' and 'ought' in turn, and shall discuss first their non-moral uses, and then their moral ones; in each case I hope to show that these uses have many features in common. In conclusion I shall relate the logic of 'ought' and 'good', in both moral and non-moral contexts, to the logic of imperatives by constructing a logical model in which artificial concepts, which could to some extent do duty for the value-words of ordinary language, are defined in terms of a modified imperative mood. This model is not to be taken too seriously; it is intendedonly as a very rough schematization of the preceding discussion, which itself contains the substance of what I have to say.
Thus the classification of prescriptive language which I propose may be represented as follows:
Prescriptive Language Imperatives Value-Judgements Singular Universal Non-Moral Moral This classification is rough only; it will be made more precise in the course of the book; for example, it will be seen that the so-called 'universal imperatives' of ordinary language are not proper universals. Nor do I wish to suggest that the classification is exhaustive; there are, for example, many different kinds of singular imperatives, and of non-moral value-judgements; and there are other kinds of imperatives besides singular and universal. But the classification is good enough to begin with, and explains the plan of this book.
1.2. The writers of elementary grammar books sometimes classify sentences according as they express statements, commands, or questions. This classification is not exhaustive or rigorous enough for the logician. For example, logicians have devoted much labour to showing that sentences in the indicative mood may be of very various logical characters, and that the classification of them all under the one name 'statements' may lead to serious error if it makes us ignore the important differences between them. We shall in the later part of this book see how one kind of indicative sentence, that which expresses value-judgements, behaves logically in a quite different way from the ordinary indicative sentence.
Imperatives, likewise, are a mixed bunch. Even if we exclude sentences like 'Would I were in Granchester!' which are dealt with by some grammarians in the same division of their books as imperatives, we still have, among sentences that are in the imperative mood proper, many different kinds of utterance. We have military orders (parade-ground and otherwise), architects' specifications, instructions for cooking omelets or operating vacuum cleaners, pieces of advice, requests, entreaties, and countless other sorts of sentence, many of whose functions shade into one another. The distinction between these various kinds of sentence would provide a nice logician with material for many articles in the philosophical periodicals; but in a work of this character it is necessary to be bold. I shall therefore follow the grammarians and use the single term 'command' to cover all these sorts of thing that sentences in the imperative mood express, and within the class of commands make only some very broad distinctions. The justification for this procedure is that I hope to interest the reader in features that are common to all, or nearly all, these types of sentence; with their differences he is no doubt familiar enough. For the same reason I shall use the word 'statement' to cover whatever is expressed by typical indicative sentences, if there be such. I shall be drawing a contrast. that is to say, between sentences like 'Shut the door' and sentences like 'You are going to shut the door'.
It is difficult to deny that there is a difference between statements and commands; but it is far harder to say just what the difference is. It is not merely one of grammatical form; for if we had to study a newly discovered language we should be able to identify those grammatical forms which were used for expressing statements and commands respectively, and should call these forms 'indicative' and 'imperative' (if the language were constructed in such a way as to make this distinction useful). The distinction lies between the meanings which the different grammatical forms convey. Both are used for talking about a subject-matter, but they are used for talking about it in different ways. The two sentences 'You are going to shut the door' and 'Shut the door' are both about your shutting the door in the immediate future; but what they say about it is quite different. An indicative sentence is used for telling someone that something is the case; an imperative is not -- it is used for telling someone to make sornething the case.
1.3. It is well worth the moral philosopher's while examumng some of the theories which have been, or which might be, held about the way in which imperatives have meaning. They offer a most arresting parallel to similar theories about moral judgements, and this parallel indicates that there may be some important logical similarity between the two. Let us first consider two theories, similar to the type of ethical theory to which I shall later give the name 'naturalist' (5.3). Both are atempts to 'reduce' imperatives to indicatives. The first does this by representing them as expressing statements about the mind of the speaker. Just as it has been held that 'A is right' means 'I approve of A', so it might be held that 'Shut the door' means 'I want you to shut the door'. There is on the colloquial plane no harm in saying this; but it may be very misleading philosophically. It has the consequence that if I say 'Shut the door' and you say (to the same person) 'Do not shut the door', we are not contradicting one another; and this is odd. The upholder of the theory may reply that although there is no contradiction, there is a disagreement in wishes, and that this is sufficient to account for the feeling we have that the two sentences are somehow incompatible with one another (that 'not' has tne same function as in the sentence 'You are not going to shut the door'). But there remains the difficulty that the sentence 'Shut the door' seems to be about shutting the door, and not about the speaker's frame of mind, just as instructions for cooking omelets ('Take four eggs, &c.') are instructions about eggs, not introspective analyses of the psyche of Mrs. Beeton. To say that 'Shut the door' means the same as 'I want you to shut the door' is like saying that 'You are going to shut the door' means the same as 'I believe that you are going to shut the door'. In both cases it seems strange to represent a remark about shutting the door as a remark about what is going on in my mind. But in fact neither the word 'believe' nor the word 'want' will bear this interpretation. 'I believe that you are going to shut the door' is not (except in a highly figurative way) a statement about my mind; it is a tentative statement about your shutting the door, a more hesitant version of 'You are going to shut the door'; and similarly, 'I want you to shut the door' is not a statement about my mind but a polite way of saying the imperative 'Shut the door'. Unless we understand the logic of 'You are going to shut the door', we cannot understand the logic of 'I believe that you are going to shut the door'; and similarly unless we understand 'Shut the door' we are unlikely to understand 'I want you to shut the door'. The theory, therefore, explains nothing; and the parallel ethical theory is in the same case; for 'I aprove of A' is merely a more complicated and circumlocutory way of saying 'A is right'. It is not a statement, verifiable by observation, that I have a recognizable feeling or recurrent frame of mind; it is a value-judgement; if I ask 'Do I approve of A?' my answer is a moral decision, not an observation of introspectible fact. 'I approve of A' would be unintelligible to someone who did not understand 'A is right', and the explanation is a case of obscurum per obscurius.
1.4. The second attempt to reduce imperatives to indicatives which I wish to consider is that of Dr. H. G. Bohnert.1 This interesting suggestion may be summarized (I hope without injustice) by the statement that 'Shut the door' means the same as 'Either you are going to shut the door, or X will happen', where X is understood to be something bad for the person addressed. A similar theory would be that it meant the same as 'If you do not shut the door, X will happen'. This theory is parallel to ethical theories of the sort which equate 'A is right' with 'A is conducive to Y' where Y is something regarded by the generality as good, for example pleasure or the avoidance of pain. We shall see later that value-expressions sometimes acquire -- by reason of the constancy of the standards by which they are applied -- a certain descriptive force; thus if, in a society whose standards are markedly utilitarian, we say 'The Health Service has done a lot of good', everyone knows that we are implying that the Health Service has averted a lot of pain, anxiety, &c. Similarly, in the case of imperatives which are to a high degree 'hypothetical' (3.2) because we quickly realize, to the attainment of what end, or the prevention of what untoward result, they are directed, Bohnert's analysis is plausible. To take his own example, 'Run', said in a burning house, is somewhat similar in intention to 'Either you run or you burn'. But in cases where the end aimed at is not so easily recognized (the imperative being only to a small degree, or not at all, 'hypothetical') the hearer may be quite at a loss to understand, on this analysis, what he is to supply after the word 'or'. It is very difficult to see how a sentence like 'Please tell your father that I called' would be analysed on Bohnert's theory. It is, of course, always possible to terminate the analysis 'or something bad will happen'; but this expedient succeeds only by reintroducing into the analysis a prescriptive word; for 'bad' is a value-word, and therefore prescriptive. And similarly, teleological theones of ethics which interpret 'right' as 'conducive to Z', where 'Z' is a value-word such as 'satisfaction' or 'happiness' only store up for themselves the difficulty of analysing such words.
The temptation to reduce imperatives to indicatives is very strong, and has the same source as the temptation to analyse value-words in the way called 'naturalistic'. This is the feeling that the proper indicatlve sentence, of which there is thought to be only one kind, is somehow above suspicion in a way that other sorts of sentence are not; and that therefore, in order to put these other sorts of sentence above suspicion, it is necessary to show that they are really indicatives. This feeling was intensified when the so-called 'verificationist' theory of meanmg became popular. This theory, which is in many ways a very fruitful one in its proper sphere, holds, to put it roughly, that a sentence does not have meaning unless there is something that would be the case if it were true. Now this is a very promising account of one of the ways in which a certain class of sentences (the typical indicatives) have meaning. Obviously, if a sentence is claimed to express a statement of fact, and yet we have no idea what would be the case if it were true, then that sentence is (to us) meaningless. But if this criterion of meaningfulness, which is useful in the case of statements of fact, is applied indiscriminately to types of utterance which are not intended to express statements of fact, trouble will result. Imperative sentences do not satisfy this criterion, and it may be that sentences expressing moral judgements do not either; but this only shows that they do not express statements in the sense defined by the criterion; and this sense may be a narrower one than that of normal usage. It does not mean that they are meaningless, or even that their meaning is of such a character that no logical rules can be given for their employment.2
1.5. The feeling, that only 'proper indicatives' are above suspicion, can survive (surprisingly) the discovery that there are perfectly good significant sentences of our ordinary speech which are not reducible to indicatives. It survives in the assumption that any meaning which is discovered for these sentences must necessarily be of some logically inferior status to that of indicatives. This assumption has led philosophers such as Professor A. J. Ayer, in the course of expounding their most valuable researches into the logical nature of moral judgements, to make incidental remarks which have raised needless storms of protest.3 The substance of Ayer's theory is that moral judgements do not ordinarily function in the same way as the class of indicative sentences marked out by his verification-criterion. But by his way of stating his view, and his assimilation of moral judgements to other (quite distinct) types of sentence which are also marked off from typical indicatives by this criterion, he stirred up dust which has not yet subsided. All this might be closely paralleled by a similar treatment of imperatives -- and it seems that writers of the same general line of thought as Ayer would have said the same sort of thing about imperatives as they did about moral judgements. Suppose we recognize the obvious fact that imperatives are not like typical indicatives. Suppose, further, that we regard only typical indicatives as above suspicion. It will be natural then to say 'Imperatives do not state anything, they only express wishes'. Now to say that imperatives express wishes is, like the first theory which we considered, unexceptionable on the colloquial plane; we would indeed say, if someone said 'Keep my name out of this', that he had expressed a wish to have his name kept out of it. But nevertheless the extreme ambiguity of the word 'express' may generate philosophical confusion. We speak of expressing statements, opinions, beliefs, mathematical relations, and so on; and if it is in one of these senses that the word is used, the theory though it tells us little, is harmless. But unfortunately it is also used in ways which are unlike these; and Ayer's use (in speaking of moral judgements) of the word 'evince' as its rough synonym was dangerous. Artists and composer and poets are said to express their own and our feelings; oaths are said to express anger; and dancing upon the table may express joy. Thus to say that imperatives express wishes may lead the unwary to suppose that what happens when we use one, is this: we have welling up inside us a kind of longing, to which, when the pressure gets too great for us to bear, we give vent by saying an imperative sentence. Such an interpretation, when applied to such sentences as 'Supply and fit to door mortise dead latch and plastic knob furniture', is unplausible. And it would seem that value-judgements also may fail to satisfy the verification-criterion, and indeed be in some sense, like imperatives, prescriptive, without having this sort of thing said about them. It is perfectly unexceptionable, on the colloquial plane, to say that the sentence 'A is good' is used to express approval of A (The Shorter Oxford English Dictionary says: 'Approve: . . . to pronounce to be good') but it is philosophically misleading if we think that the approval which is expressed is a peculiar warm feeling inside us. If the Minister of Local Government expresses approval of my town plan by getting his underlings to write to me saying 'The Minister approves of your plan' or 'The Minister thinks your plan is the best one', I shall in no circumstances confirm the letter by getting a private detective to observe the Minister for signs of emotion. In this case, to have such a letter sent is to approve.
1.6. There could be no analogue, in the case of singular imperatives, of the 'attitude' variety of the approval theory of value-judgements;4 but it is possible to construct such a theory about universal imperative sentences. If someone said 'Never hit a man when he is down', it would be natural to say that he had expressed a certain attitude towards such conduct. It is extremely hard to define exactly this attitude or give criteria for recognizing it, just as it is difficult to say exactly what moral approval is as opposed to other sorts of approval. The only safe way of characterizing the attitude which is expressed by a universal imperative is to say 'The attitude that one should not (or should) do so and so'; and the only safe way of characterizing the attitude which is expressed by a moral judgement is to say 'The attitude that it is wrong (or right) to do so and so'. To maintain an attitude of 'moral approval' towards a certain practice is to have a disposition to think, on the appropriate occasions, that it is right; or, if 'think' itself is a dispositional word, it is simply to think that it is right; and our thinking that it is right may be betrayed or exhibited --behaviourists would say constituted -- by our acting in certain ways (above all, doing acts of the sort in question when the occasion arises; next, saying that they are right; applauding them in other ways, and so on). But there is in all this nothing to explain just what one thinks when one thinks that a certain sort of act is right. And similarly, if we said that 'Never hit a man when he is down' expressed an attitude that one should not hit, &c. (or an attitude of aversion from hitting, or a 'contra-attitude' towards hitting), we should not have said anything that would be intelligible to someone who did not understand the sentence which we were trying to explain.
I wish to emphasize that I am not seeking to refute any of these theories. They have all of them the characteristic that, if put in everyday terms, they say nothing exceptionable so far as their main contentions go; but when we seek to understand how they explain the philosophical perplexities which generated them, we are either forced to interpret them in such a way as to render them unplausible, or else find that they merely set the same problems in a more complicated way. Sentences containing the word 'approve' are so difficult of analysis that it seems perverse to use this notion to explain the meaning of moral judgements which we learn to make years before we learn the word 'approve'; and similarly, it would be perverse to explain the meaning of the imperative mood in terms of wishing or any other feeling or attitude; for we learn how to respond to and use commands long before we learn the comparatively complex notions of 'wish', 'desire', 'aversion', &c.
1.7. We must now consider another group of theories which have often been held concurrently with the group just considered. These hold that the function in language of either moral judgements or imperatives (which the theories often equate) is to affect causally the behaviour or emotions of the hearer. Professor R. Carnap writes:
But actually a value-statement is nothing else than a command in a misleading grammatical form. It may have effects upon the actions of men, and these effects may either be in accordance with our wishes or not; but it is neither true nor false.5and Professor Ayer writes:Ethical terms do not serve only to express feeling. They are calculated also to arouse feeling, and so to stimulate action. Indeed some of them are used in such a way as to give the sentences in which they occur the effect of commands.6More recently this sort of view has been elaborated by Professor Stevenson.7 Here again we have a type of theory which may be on the colloquial plane harmless, but uhich suggests philosophical errors by seeming to assimilate the processes of using a command or a moral judgement to other processes which are in fact markedly dissimilar.It is indeed true of imperative sentences that if anyone, in using them, is being sincere or honest, he intends that the person referred to should do something (namely, what is commanded). This is indeed a test of sincerity in the case of commands, just as a statement is held to be sincere only if the speaker believes it. And there are similar criteria, as we shall later see, for sincerely assenting to commands and statements that have been given or made by someone else. But this is not quite what the theories suggest. They suggest, rather, that the function of a command is to affect the hearer causally, or get him to do something; and to say this may be misleading. In ordinary parlance there is no harm in saying that in using a command our intention is to get someone to do something; but for philosophical purposes an important distinction has to be made. The processes of telling someone to do something, and getting him to do it, are quite dictinct, logically from each other.8 The distinction may be elucidate by considering a parallel one in the case of statements. To tell someone that something is the case is logically distinct from getting (or trying to get) him to believe it. Having told someone that something is the case we may, if he is not disposed to believe what we say, start on a quite different process of trying to get him to believe it (trying to persuade or convince him that what we have said is true). No one, in seeking to explain the function of indicative sentences, would say that they were attempts to persuade someone that something is the case. And there is no more reason for saying that commands are attempts to persuade or get someone to do something; here, too, we first tell someone what he is to do, and then, if he is not disposed to do what we say, we may start on the wholly different process of trying to get him to do it Thus the instruction already quoted 'Supply and fit to door mortise dead latch and plastic knob furniture' is not intended to galvanize joiners into activity; for such a purpose other means are employed.
This distinction is important for moral philosophy; for in fact the suggestion, that the function of moral judgements was to persuade, led to a difliculty in distinguishing their function from that of propaganda.9 Since I am going to draw attention to some similarities between commands and moral judgements, and to classify them both as prescriptions, I require most emphatically to dissociate myself from the confusion of either of these things with propaganda. We have here, as often in philosophy, a mixture of two distinctions. The first is that between the language of statements and prescriptive language. The second is that between telling someone something and getting him to believe or do what one has told him. That these two distinctions are quite different, and overlap each other, should be clear after a moment's consideration. For we may tell someone, either that something is the case, or to do something; here there is no attempt at persuasion (or influencing or inducing or getting to). If the person is not disposed to assent to what we tell him, we may then resort to rhetoric, propaganda, marshalling of additional facts, psychological tricks, threats, bribes, torture, mockery, promlses of protection, and a variety of other expedients. All of these are ways of inducing him or getting him to do something; the first four are also ways of getting him to believe something; none of them are ways of telling him something, though those of them which involve the employment of language may include telling him all sorts of things. Regarded as inducements or expedients for persuasion, their success is judged solely by their effects -- by whether the person believes or does what we are trying to get him to believe or do. It does not matter whether the means used to persuade him are fair or foul, so long as they do persuade him. And therefore the natural reaction to the realization that someone is trying to persuade us is 'He's trying to get at me; I must be on my guard; I mustn't let him bias my decision unfairly; I must be careful to make up my own mind in the matter and remain a free responsible agent'. Such a reaction to moral judgements should not be encouraged by philosophers. On the other hand, these are not natural reactions either to someone's telling us that something is the case, or to his telling us to do something (for example, to fit a latch to the door). Telling someone to do something, or that something is the case, is answering the question 'What shall I do?' or 'What are the facts?' When we have answered these questions the hearer knows what to do or what the facts are -- if what we have told him is right. He is not necessarily thereby influenced one way or the other, nor have we failed if he is not; for he may decide to disbelieve or disobey us, and the mere telling him does nothing -- and seeks to do nothing -- to prevent hum doing this. But persuasion is not directed to a person as a rational agent, who is asking himself (or us) 'What shall I do?'; it is not an answer to this or to any other question; it is an attempt to make him answer it in a particular way.
It is easy to see, therefore, why the so-called 'imperative theory' of moral judgements raised the protests that it did. Because based on a misconception of the function, not only of moral judgements but also of the commands to which they were being assimilated, it seemed to impugn the rationality of moral discourse. But if we realize that commands, however much they may differ from statements, are like them in this, that they consist in telling someone something, not in seeking to influence him, it does no harm to draw attention to the similarities between commands and moral judgements. For, as I shall show, commands, because they, like statements, are essentially intended for answering questions asked by rational agents, are governed by logical rules just as statements are. And this means that moral judgements may also be so governed. We remember that the greatest of all rationalis, Kant, referred to moral judgements as imperatives; though we must remember also that he was using the latter term in an extended sense, and that moral judgements, though they are like imperatives in some respects, are unlike them in others (11.5).
Notes 1 'The Semiotic Status of Commands', Philosophy of Science, xii (1945), 302.
2 See my article 'Imperative Sentences', Mind, lviii (1949), 21, from which some material is here used.
3 See especially Language, Truth and Logic, 2nd ed., pp. 108-9. For a later and more balanced statement, see 'On the Analysis of Moral Judgments' Philosophical Essays, pp. 231 ff.
4 See, for example, C. L. Stevenson, Ethics and Language.
5 Philosophy and Logical Syntax, p. 24.
6 Language, Truth and Logic, 2nd ed., p. 108.
7 Ethics and Language, especially p. 21.
8 For a fuller treatment of this question see my article, 'Freedom of the Will', Aristotelian Society, Supplementary Vol. xxv (1951), 201, from which I have used some material here and in 10.3.