VOL. XLIV No. 169. JULY 1969



First published in Philosophy, XLIV, July 1969, pp.175-186
page numbers at bottom of each page

"Ich liebe dich 3" the swains in mountain valleys of Austria inscribe on their presents to those to whom they plight their troth. The pun is a rare one in German. Only in remote valleys does the word for `three' rhyme with joy; and the word for `true' is usually `wahr' not `treu'. `Wahr' is more prepositional, less evaluative than our `true'. So too in Latin and the romance languages `verum', `vrai' `vera' `verdadero' carry a lighter load of feeling. Only the Hebrews resembled us in giving their word for `true' strong moral overtones, and using the same or some cognate word to praise him who had used no deceitfulness to his neighbour as to recommend some statement as being meet to be believed. 2

Our habit of using the same word of friends, dice, lines, aims, jurors and verdicts as well as of statements and propositions is illuminating. Words are winged, and statements and propositions have a tendency to fly upwards into those levels of philosophic abstraction where cloudiness is the normal climatic condition. It may be necessary on occasion to be abstract. But it is well at least to start on the ground, and a true friend is likely to be a better guide than any considerations of Wahrscheinlichkeitrechnung.

A true friend is one I can trust. And trust, too, is what I can repose in true propositions. In telling you that a friend, a die, a line, or a proposition, is true, I am telling you that you can trust him or it, that he or it is trustworthy, that he or it is worthy of your trust and will not let you down. In ascribing truth to propositions, I am plighting myself to you, not for my own future conduct but for their reliability; sometimes, as Strawson points out, 3 conceding, but much more often insisting, that they are to be believed, and making myself answerable to you, should they turn out to be incorrect. To say of a proposition that it is true is like saying, on Austin's analysis, that I know it. If I say that it is true that I was in College last night; I am promising you that I was in College last night; I am vouching for it, guaranteeing it, staking my credit on it; and if you discover that I was not, you will be entitled never to trust me, never to believe me, again. 4



On this account ` . . . . is true' plays a similar, though contrasted, role to ` . . . . is probable'. I use `...... is probable' to hedge, ` . . . . is true' to stick my neck out `........ is probable' weakens the force of the unqualified assertion and could be rendered by `I think but am not sure', whereas ` . . . . is true' strengthens its force and could be rendered `I do not merely think but am absolutely sure'. In spite of the New Testament injunction, our unqualified assertions, our simple `Yea' and `Nay', are not enough: there are occasions when we need to give less than the ordinary warranty, and occasions when we may be asked to, and perhaps can, give more. There are various ways in which we can do this, but the words `probable' and `true' have become peculiarly apt for this purpose, and appropriated to it.

From this function of the word `true' as applied to propositions and statements stem its entailment and other inference patterns. To say of a proposition "It is true, but I do not believe it" is barely intelligible. If we take `I believe' in its parenthetical, not its autobiographical, sense, it is a straight contradiction. We can, just, attach meaning to it as a statement about myself, that I cannot bring myself to believe it, because it is so shocking, so terrible, so surprising, although, really, I know it to be true: but we cannot accept it in its standard autobiographical sense without imputing to the speaker an insincerity that would vitiate everything he said. 5 `It is true' therefore implies `I believe it'. Equally we can ground an imperative doctrine of truth on the unintelligibility of the conjunction `It is true' and `Do not believe it', and hence the entailment from `It is true' to `Believe it'. A latter-day Stevenson 6. might analyse `this is true' by conjoining both these entailments and saying that it is equivalent to `I believe this: do so as well'. But we shall object to this analysis of truth, as we do to Stevenson's analysis of `this is good', that it appears to be subjective and arbitrary, and not to do justice to the rationality implicit in the words defined. In saying that some proposition is true, I am not merely telling you to believe it, imperatively, as in an order, but saying, gerundively, that you should believe it, for it is worthy of belief. In saying that a proposition is true, I am not merely staking my credit on it irresponsibly, or arbitrarily ordering you to believe it, but am implying that there are reasons why I do, and why you should, believe it. I am not merely endorsing it, `a la Strawson, but saying that it is fit to be endorsed, not only by you, but by any rational being. So Walsh, 7 so also St. Paul `This is a true saying, and worthy of all men to be received'. 8

To choose is to reject. To commend a proposition as worthy of belief is significant only if there are other propositions which are not being commended and are not worthy of belief. The point of having the word `true' is that it enables us to pick out some propositions, statements, theories, accounts, well-formed formulae or descriptions, as being worthy to be believed, by contrast with others which are not to be believed. If there were not a contrast, there would be no point in the recommendation. True



propositions have a scarcity value. Not every meaningful proposition or well-formed formula (save in an absolutely inconsistent system) is true, and therefore to say of some that they are true is to give, them a further value they did not have merely by being meaningful or well-formed. It is only because there is a further distinction within the realm of meaningful propositions, well-formed formulae, etc., that they are any use to us. It is never enough merely to have criteria of meaningfulness or formation rules. These by themselves would make language merely a game. If it is to be more, if our propositions, formulae etc,, are to pack any punch, there must be further division between, so to speak, propositional sheep and propositional goats.

This logical doctrine of election explains the black-and-white distinction between truth and falsehood. The distinction is black-and-white because the concepts of truth and falsehood, and of proposition, well-formed formula, etc., depend on there being such a distinction. Information is defined technically 9 in terms of choice, actual choice within a range of possible choice. Linguistic utterances and symbolic formulae will convey "Information" (in its most extended sense) only if they admit of a sharp distinction between those of them that are true and are to be selected and those that are false and are to be rejected. Like Tweedledum and Tweedledee they each need the other, and agree in everything except being different. And in their quarrel no mediation is possible, for the middle is by law excluded.

Although `false' is the prime contrast to `true', it is not the only one. Sometimes it is being contrasted not so much with `false' as with `inaccurate', `inexact', `imprecise', `inadequate'; and whereas in the former case there is no comparative---either a proposition is true or else it is false, tertium non datur---in the latter case degrees of truth are intelligible. One historical account is truer than another, although the other is not false, only less penetrating. If it is not true that France is hexagonal, it is not because it is false, but because it is too inaccurate to be helpful. Although it is not in doubt that of two options the correct rather than the incorrect option has been taken, it may be felt that some other range of options would have enabled a more felicitous choice to have been made. The answer was the right one rather than the wrong one, but the question was badly put. A better question would have divided the bones of the matter neatly, at the joints, instead of cutting across the main issues. It is in these cases that we want to use the comparative, and say that one account is truer than another, without saying that that other one is false. It is characteristically to very complicated exercises---descriptions, accounts, narratives and novels---that we ascribe degrees of truth. One historian may give a less true account of a historical period than another does, not because what he says is false---his account would not become more true by being negated---but because the events he selects, the features he describes, the explanations he offers, although fairly important, reasonably



typical, and partially valid, are less worthy of emphasis than those selected by the other man. We distinguish, among portraits, bad likenesses of a certain man from likenesses, possibly good ones, of somebody else. The former are less true than a good likeness of the man in question would be, but are not in any way false: whereas the latter, if represented as portraits of the man in question, would indeed be false ones.

Philosophers of history sometimes confuse the two ways in which a historical account can fail to be true, and are driven to the conclusion that only a total statement of everything can be allowed as fully true. We are helped to resist this conclusion, if we distinguish the central contrast of true with false from the peripheral one of more true with less true. If the question arises, it can never be right to give the false rather than the true answer to it. Truth in this context is mandatory, falsehood fatal. But it is a big `if'. Often the question does not arise, often it would be wrong, because irrelevant, to raise it. The witness swears to tell the whole truth. We must not suppress anything relevant. But telling the whole truth does not require a step by step account of his walk to the scene of the accident. Although, if he is asked, he should say whether he had been walking steadily or sauntering, whether he had stopped to do some shopping or had been admiring the view, it would be wrong to volunteer these details, and the judge would soon lose patience if he did. Every account, however detailed, can be replaced by another, more detailed. But we are not obliged to give that more detailed one, if we are trying to tell the whole truth. Indeed, if we were, the whole truth could never be told.

When asked to assess the truth of an account or description, we should consider what alternative account could have been given. If the alternative to describing France as hexagonal is calling it a quadrilateral, then it is pointless to protest that really France does not have rectilinear boundaries at all. A brief answer to a short question may be true, although inaccurate if contrasted with a comprehensive account as an acceptable alternative. Our purposes vary, and therefore also the range of possible answers we might give, the range of possible options we might take. To this extent truth varies too. But always it continues to indicate that answer which we should give, and which should be believed, and which is most worthy of trust.

The logical doctrine of election explains also the systematic ambiguity in the application of the words `true' and `false' to propositions, statements, sentences, descriptions, accounts, theories, well-formed formulae, and the rest. Philosophers have often worried whether we should properly describe only propositions as true or false, or perhaps statements, or perhaps theories. They need not have worried. It does not matter. So far as the words `true' and `false' are concerned, they can be applied in almost any case where we are using a symbolism and have a choice between some to be selected and others to be rejected. The proposition that all men are mortal is true, although it is conceivable that with the advance of



medical science it might turn out to be false. The statement I make to the police is true, although some of the statements made by the criminal fraternity are not. The sentence `Snow is white' is true, as is also the Gödelian sentence; and similarly the sentences `All swans are white' and `Two and two make five' are false, although many other sentences, like `It is raining here today', are context-dependent, and cannot be said to be true or false apart from some occasion of utterance. So too descriptions, accounts and theories, as well as well-formed formulae, can be said to be true, just because there are other, equally eligible as far as their formal structure goes, descriptions, accounts, theories and well-formed formulae, which, although eligible are not elected. The failure of the phlogiston theory gives point to our putting forward the atomic theory as true.

That `true' and `false' are available to a multitude of entities makes them popular with thinkers, but makes them also fickle and unreliable when thought about. We hypostatize. In the higher reaches of the Fregelian Absolute, truth becomes an abstract quality, existing by itself apart from all contextual contamination. Not that it is wrong to have a concept of Truth. We may quite properly use the abstract noun, especially when we reflect, metalogically, metalinguistically or metaphysically, on our first-order uses of the adjective: but it is something of a term of art; we confuse ourselves if we divorce it from its context; and whatever else it is, it is not a property.

It is not wrong to talk about truth. But metalogic, like moral philosophy, can suffer from the distortion of detachment. When we talk about evaluative words or use them only hypothetically, instead of actually using them in a full-blooded categorical way, the illocutionary and locutionary meanings begin to diverge. `If Grimbly Hughes is the best shop in Oxford, go to Grimbly Hughes' does not commend Grimbly Hughes, and neither does `If p is true, then so is q' commend either p or q. It is both a great facility and a very dangerous one that we can talk about truth, and propositions being true, without there being any proposition of which we can actually say that it is true. For what we mean in entertaining the hypothesis that a proposition be true must be different from what we mean in asserting that it is true. We are not vouching for it, warrantedly asserting it, saying that it is worthy of our rational acceptance, but only applying, or considering applying, a metalogical label, in order to further our philosophical purpose of logical investigation. We are entitled to do this. But we need to be clear what our purposes are, because although the sophisticated uses of the word `true' develop out of the simple one, the exact way in which it is used will depend on the exact purpose for which it is being used, and the difference between the various senses given to the word has often given rise to controversy and confusion.

The primitive meaning of the word `true' is constituted both by its function of commanding for rational acceptance and by its implication patterns. If we mute the evaluative force of the word, we shall forget that



it has a function and shall concentrate only on its implication patterns, and indeed on only some of those. We shall no longer construe the word as a sort of assertion operator with a part to play in the give-and-take conversation between rational autonomous agents, and regard it as merely a sort of abstract metalogical property---a "truth-value"---whose meaning is given by static patterns of implication. Indeed, even the autobiographical and imperative implications will seem inappropriate, and the only safe account of the meaning of the words `true' and `false' will be that `p' is true if and only if p, and `p' is false if and only if not p, and we are well on the way to both the correspondence theory of truth and the "no-truth" theory of Ramsey and Ayer. 10 The necessary conditions are indeed necessary. It is part of the meaning of the word `true', in contrast to that of `probable', that we commit ourselves unreservedly, and are at fault if we prove mistaken in the event, however well-grounded our assertion was when we made it. But if we make it a sufficient condition too, we are in danger of confusing ourselves by using the word `true' sometimes as a new tailor-made metalogical property and sometimes in its old, partly operational sense. The controversy between the Intuitionists and classical mathematicians is engendered partly by this ambiguity, as also the smoky fog of confusion that has enveloped tomorrow's sea-battle for the past two thousand years.

The Intuitionists have a more primitive notion of truth than classical mathematicians. To say a proposition is true in intuitionist logic is to be able to vouch for it, to be warranted in asserting it, to have, at least in the mind's eye, a construction which shows that it is worthy to be received, The classical mathematicians use the word not in order to make claims, but to consider possibilities and what follows from them. The words `true' and `false' ascribe truth-values, alike in each being symmetrical to the other in respect of negation, different in that the rule of Modus Ponens applies in the one class of propositions, but not in the other. It is a legitimate usage, but one that can mislead, because the original sense of the word `true' is liable to reassert itself, and make us uneasy in considering that a proposition may be true unless we have good reason for believing it. It is the same with the sea-battle. We may follow the traditional logic and apply the metalogical truth-values, True and False timelessly to timeless entities (which I shall call propositions---usage varies), or we may follow Prior and ordinary language, and apply them temporally to tensed entities, statements which, whether actually uttered by any particular person or not, must have been, either actually or potentially, uttered at a particular time. We can take either option, but not both at once. If we take the former, we need have no doubts about the Law of the Excluded Middle and the Principle of Bivalance, but no problem about Future Contingents, since we can never formulate in that logic the question of whether it was true that the sea-battle would be fought. Propositions are senseless, and somewhat tenuous, entities. Although we always can



affirm that it is (tenselessly) true either that a sea-battle is (tenselessly) fought on a certain date or that it is not, we need not be, and before the event are not, able to say which of these two propositions is (tenselessly) true, nor can we tie down either to having been true, even though we did not know it, all along, before the event as well as after it. If, on the other hand, we allow the locutions `was true' `is (tensedly) true' and `will be true', then we may not be able to stipulate that the Law of the Excluded Middle or the Principle of Bivalence shall hold universally. Different tense-logics are possible, and in some the Future Operator does not commute with negation, and there is a difference between saying that it is not the case that there is going to be a sea-battle, and that it is going to be the case that there is not a sea-battle. This represents a facet of our ordinary usage, where we often find ourselves unable at a particular time to commit ourselves either to its being true there will be a sea-battle or to its being true that there will not. We can say that either there will or there will not, but we cannot say whether there will or will not, and so cannot use the word `true' to pick out one statement as worthy of belief rather than another.

A useful analogy may be drawn between the different concepts of truth and the different theories of denotation espoused by Russell and Strawson. We can have a Platonistically perfect, no-truth gap theory, in which the Law of the Excluded Middle obtains rigorously, but in which many ordinary locutions have to be radically reformulated or extruded altogether: or we can have a messier system which is closer to the actualities of discourse, but depends more on common sense, less on the mechanical application of formal rules, to make it work and yield sensible results. If we play the traditional Russellian game, we can lay it down that "it is necessary in every case to affirm or to deny",11 and can ensure that every sentence containing a definite description is to be so understood that it must be true or false. But if we aim to keep closer to our ordinary usage, then we cannot claim that it is necessary in every case to affirm or deny, for often we say `I don't know', and as a matter of conversational fact tertium datur; and, in much the same way, definite descriptions sometimes fail to refer. We have different ideals. If we want to bring out what follows from saying something is true, we idealise in company with Russell and the traditional logicians. If we want to stress the contextual conditions under which it is proper to use the word `true', we make True less of a truth-value and more of an operator, whose application is not timeless, and which, together with False, can only sometimes and not always be applied each to one of a pair of contradictory statements.

Even this account is not negative enough. Although logicians can operate with a metalogical property something like our concept of truth, they cannot construct it, as they would like, formally, and it would be impossible to undertake a rational reconstruction of our primitive, partly operational concept, as a metalogical property formally defined and



formally construed. This is a consequence of Tarski's theorem, which shows that we cannot, on pain of inconsistency, have in a formal system a formal metalogical property of being true, which satisfies the minimum implicational requirements adumbrated above, that `p' is true if and only if p. Truth is not only not the same in all disciplines and in all contexts, but is inherently inexpressible, and perhaps unknowable too. It is a surprising result. Even when we have made all allowances for differences of discipline and differences of context, we tend to assume that at least then there is a class of true propositions in each discipline and granted a specific context, and that if we were gods and omniscient, we could survey the whole class of propositions in each discipline and granted a specific context, and see exactly which ones are true. But Tarski's theorem shows that any such assumption leads to inconsistency. Tarski's theorem is like Gödel's, only it is concerned with a predicate tantamount to `. . . . is true' instead of ` . . . . is provable'.12 If a formal system contains the natural numbers together with the operations of addition and multiplication, and has a predicate with the quite essential formal properties of ` . . . . is true', then we can use Gödel's diagonalization method to produce a contradiction that is, in effect, a fool-proof form of the Epimenides paradox. For, if a formal system contains the natural numbers together with the operations of addition and multiplication, we can assign Gödel numbers to each formula of the system; and if there is a definite predicate which holds of each of those well-formed formulae that are true, there is likewise a predicate of their Gödel numbers, say Tr, such that if k is the Gödel number of a,

Tr(k) if and only if a is true.

The property of being Tr can he represented formally in the system, since being true can, and so also the property of not being Tr. We can therefore give a Gödel number to the formula ~Tr(x) where x is a variable, and similarly a Gödel-number to each of the substitution instances of it, of the general form ~Tr(m) where m is a particular number. We then construct, by Gödel's diagonal method, the formula


whose Gödel number is itself h; h then is the Gödel number of Epimenides' paradoxical assertion `This statement is false'. For if Tr(h) holds, then h is the Gödel number of a true well-formed formula, namely ~Tr(h), and so this well-formed formula ~Tr(h), is true, just because h is its Gödel number. If, on the other hand, ~Tr(h) holds, it is true that ~Tr(h), and therefore the Gödel number of ~Tr(h)---which is in fact h---is one to which the predicate Tr applies. Thus, either way, we have an inconsistency.

The only assumptions we have made are that if a proposition or well-formed formula is true then we can assert the proposition or well-formed formula, and conversely that we can go from `p' to " `p' is true"; and these seem to be essential to the metalogical concept of truth as I have expounded it. And if we are not willing to abandon either of these, then, since almost



every discipline must contain the natural numbers together with addition and multiplication, it follows that we cannot represent within it a predicate `. . . . is true' which is always definite in its application, or a class of all and only those propositions which are true. If I retain the predicate `. . . . is true' at all, I can no longer regard it as an ordinary sort of predicate having an extension that can be manipulated like an ordinary class.

Some philosophers see Tarski's theorem as showing the concept of truth to be internally inconsistent, and therefore needing to be extruded from the wise man's vocabulary. But it is formalisation, rather than truth, that is at fault. We cannot (except in degenerate systems in which Gödelian self-reference cannot be contrived) have a class of all and only all the true propositions of the system: that is to say that `. . . . is true' is not a simple predicate ascribing a definite property to propositions, such that we can always in principle say of every proposition either that it is true or that it is false. But there is no need for us to suppose that any worthwhile intellectual discipline is formalised, or that it is sufficiently closed for there to be a class of all and only all its true propositions. Intellectual disciplines are activities rather than formal systems, and to say of any proposition that it is true is to commend, rather than to ascribe any simple metalogical property. The word `true' is meaningful not because there is some absolute property of truth which it denotes, but because it has a function and implication patterns, and can be rightly or wrongly used.

Many philosophers have not been content to know the meaning of the word `true', but have sought a compendious criterion for its applicability, which would enable them to tell us exactly what truth is. They feel that Our Lord should not have let the opportunity, occasioned by Pilate's question, slip: and take it on themselves to remedy the deficiency and describe briefly---say in the compass of a one- or two-volumed book---the whole of truth. Such endeavours do discredit to philosophy. It is not to be supposed that a philosopher can tell better than the workers in other disciplines what the fruits of their labours are. And even though he may be more articulate than they in formulating the criteria they use, it is, at the very least, a matter for give and take.

The philosopher can only learn by constant discussion and by continually referring back to those who work in some other discipline whether the criterion he proposes is an adequate formulation of the rules they work by. In particular, philosophers who seek to state criteria of truth, have often erred in considering only one discipline, or only one range of disciplines, and then generalising from the special case to others, quite inappropriately. The correspondence theory is passable, if we consider it as a theory of geographical truth: but if we then say that propositions---all propositions---are true in virtue of corresponding to facts, or to the world, or something, then either we shall have to be Platonists and invent realms of mathematical and moral entities for true propositions of mathematics or morality to



correspond to, or we shall have to be sceptics, and deny that the word `true' can ever be applied to the pseudo-propositions of mathematics or morality. The coherence theory is a possible one if we have only mathematics, or, with an important qualification, mathematical physics, in mind; and some metaphysical schemes are so highly integrated, and have their propositions depending so much on one another for their truth, that they too may be assessed on the score of internal coherence more than anything else. But what may be allowed in assessing the Theory of Groups, The General Theory of Relativity, or the philosophy of Spinoza, becomes ludicrous when applied to questions of history or matters of everyday observation. Only an Orwellian could maintain that however much we lost the last war, it will be true that we won it, if our having won fits in better with our other accepted views. Coherence, as one criterion among several, has its part to play in assessing truth, just as correspondence does. But neither can constitute a whole theory of truth, for neither constitutes a sole criterion applicable in all cases, without regard to subject matter.

There is no single criterion of truth. Different disciplines have different criteria, often unspecified, sometimes, where specified, liable to conflict. Any account of criteria is bound to be crude and inexact. All we can say, briefly, is that in mathematics those propositions are true whose negations would be inconsistent with the axioms, and in physics a proposition is often to be taken as true if its negation would be inconsistent with a theory which yields predictions that turn out, on the whole, to be true. With predictions, as with retrodictions, it is much more difficult to say in general what criteria are needed to establish their truth, although usually we have no difficulty in particular cases. Often coherence remains a criterion---we are readier to allow that the expected has happened than the unexpected, and a witness's testimony in history or the law courts is often more likely to be true if it is borne out by that of other witnesses: but not always.

We should note also as a special case of the coherence criterion, of particular importance in assessing metaphysical and other general philosophical theses, that of "skew-inconsistency" where we reject a philosophical thesis as itself showing its own falsehood. Ayer's youthful belief that all meaningful propositions were either analytic or empirical, is, if meaningful, a counter-example to what is being maintained. So too the view, which Hume held, that all reasons must be either deductive or inductive, cuts off the branch on which it metaphorically sits, in that if it is true, then there can be no reason to believe it is. In a similar fashion we find our Marxist and Freudian friends hoist by their own petard, when they make out that all the views a man espouses are determined completely by the economic interests of his class or his own experiences in early childhood. These are the large-scale analogues to the logicians' puzzles of the man who says `I am asleep' or `I do not exist. In each case, there is an inconsistency, although not a straight inconsistency. The utterance belies the truth of what is said. Such a thesis, which is thus "skew-



inconsistent", is to be rejected as false, although many that are not "skew-inconsistent" are nevertheless not true either. The criterion is a negative one, although Descartes' Cogito argument can be viewed as an attempt to use it for positive effect.

If we cannot formalise the concept of truth, or give any rigid application rules, we can never make it proof against error. I regard this as a virtue, but many will view it as a vice, a fatal fault, and again will seek to avoid using so fallible a concept, and to confine themselves to the safer substitute of provability. Proof-sequences are by definition fool-proof. Why have, or at least why use, the less reliable concept of truth, whose use cannot be completely governed by formal rules, and always may need nous which one may have not got? It is a fair question. The answer lies in the nature of language-users, and hence the nature of the language they use. Language, contrary to the assumption of much mathematical logic, is not only used by men but addressed by them to other men. It is, as I have argued elsewhere, 13 a dialogue not a monologue; and, moreover, it is a dialogue between communicators who can be both worse and better than was to be expected. Men, that is, are often wrong and occasionally original. Because we are fallible, it is essential that communications are not automatically accepted and acted upon, but are assessed by an autonomous intellect and may be rejected. But because the autonomous intellect too may be wrong, we need both to offer guidance at the outset and to be able to insist on our own view if it is being wrongly rejected by him. We need both to be able to commend a proposition by saying that it is true, and to reject a rejection by still saying that it is true despite our hearer's denials. We need truth for the same reason as we need negation: fallible men are often wrong, and we need to be able to deny what they say, and draw afresh the distinction between these propositions that are true and those that are false. We could not reduce language to an absolutely consistent but negationless calculus---like the primitive positive calculus of Hilbert 14---unless men were automata who never went wrong. Since men are not automata, they cannot react to language automatically, but have to be able to stand back and consider, metalinguistically, whether some linguistic communication they have received is to be accepted or not. And for this they need not only `not' but the metalinguistic and inherently defensible 15 concept of truth. If we were all infallible, intellectual life would be simpler and we could manage without the concept of truth since it would have no scarcity value: but seeing the error of our ways, we have to think a lot about truth, since we often do not have it.

Mistakes are not the only argument for truth. Men are sometimes original. They discover new truths, occasionally even whole new systems of truth. Intellectual reformers need to preserve autonomy as much as moral ones do; they need to be able to say "Ye have heard of old time that the earth is flat: but I am telling you, it is not true". The word `true' cannot be defined, as some of the logical positivists maintained, as what



the scientific establishment believes. Not only does the word not mean that, but it is necessary to have a word that gives intellectual commendation without being tied to antecedent intellectual acceptance, if knowledge is to grow and falsehood dispelled. We need to be able to vouch for our own view, to stake our reputation on the reliability of the propositions we propound, independently of whether ours is an orthodox opinion or not. If a man cannot ask his intellectual friends to trust him, then he is no longer being regarded as an autonomous intellectual agent who makes an independent assessment of reasons, but merely as an automaton whose sole function is to toe the accepted line, and say the same thing as anybody else would say. Granted, if one is free to chance one's arm one may be non-conformist only to be wrong: if one sticks out one's neck, one may have it chopped off. The quest for truth is hazardous. Truth is a perpetual possibility of being wrong. But I should be happy to settle for that; even, almost, as a definition.

Merton College, Oxford.



1. A paper read to the Jowett Society on 3rd December, 1968. I owe a deep debt to Dr. F. Waismann, whose answers to a question by Mr. Michael Ryle in the course of a lecture first set me thinking on the line developed here.

2. See C. H. Dodd, The Interpretation of the Fourth Gospel, Cambridge, 1953, ch. 4, pp. 170-178, esp. p. 173; and pp. 139-140.

3. P.F.Strawson, "Truth", Analysis, 1949, pp. 93, 95.

4. J.L.Austin, "Other Minds", Proceedings of the Aristotelian Society, Supplementary Volume XX, 1946, pp. 169-175; reprinted in J. L. Austin, Philosophical Papers, Oxford,

1961, pp. 66-68.

5. For a fuller discussion, see A. N. Prior, "On Spurious Egocentricity", Philosophy, XLII, 1967, S 2, pp. 326-328; reprinted in A. N. Prior, Papers on Time and Tense, Oxford, 1968, pp. 15-17.

6. C. L. Stevenson, Ethics and Language, New Haven, 1944, p. 21.

7. W.H.Walsh, "A note on truth", Mind, 1952, LXI, pp. 72-74.

8. I Timothy 1, 15 (Prayer Book translation).

9. Norbert Wiener, Cybernetics, 2nd ed., Cambridge, Mass., 1961, ch. 111, p. 61. See also K. M. Sayre, "Choice, Decision and Information", in Frederick J. Crosson and Kenneth M. Sayre, eds., Philosophy and Cybernetics, Notre Dame, 1967, pp. 73-74. Leon Brillouin, Science and Information Theory, 2nd ed., New York, 1962, ch. 1, section 1.

10. F.P.Ramsey, The Foundations of Mathematics, London, 1931, pp. 142-3.

11. Aristotle, Metaphysics, B, 2, 996b29

12. Alfred Tarski, Logic, Semantics, Metamathematics, tr. J. A. Woodger, Oxford, 1956, pp. 187-188, 247.

13. "Not `Therefore' but `But'", Philosophical Quarterly, 16, 1966, pp. 289-307, and "The Philosophy of the Reasonable Man", Philosophical Quarterly, 13, 1963, pp. 97-106.

14. See Alonzo Church, Introduction to Mathematical Logic, Princeton, 1956, ch. 11, section 26, pp. 140-141.

15. See H.L.A.Hart, "The Ascription of Responsibility and Rights", Proceedings of the Aristotelian Society, 1948-9, pp. 171-194; reprinted in A.G.N.Flew, Logic and Language, Series 1, Oxford, 1951, pp. 145-165.




Return to bibliography

Return to home page