Philosophy

The question of truth and falsehood has been wrapped in unnecessary mystery owing to a number of causes. In the first place, people wish to think that their beliefs are more apt to be true than false, so that they seek a theory that will show that truth is normal and falsehood more or less accidental. In the second place, people are very vague as to what they mean by “belief” or “judgment”, though persuaded that they know beliefs or judgments to be the objects to which the predicates “true” or “false” apply. In the third place, there is a tendency to use “truth” with a big T in the grand sense, as something noble and splendid and worthy of adoration. This gets people into a frame of mind in which they become unable to think. But just as the grave-diggers in Hamlet became familiar with skulls, so logicians become familiar with truth. “The hand of little employment hath the daintier sense,” says Hamlet. Therefore it is not from the logician that awe before truth is to be expected.

There are two questions in our present subject: (1) What are the objects to which the predicates “true” and “false” apply? (2) What is the difference between such as are true and such as are false? We will begin with the first of these questions.

Prima facie, “true” and “false” apply to statements, whether in speech or in writing. By extension, they are supposed to apply to the beliefs expressed in those statements, and also to hypotheses which are entertained without being believed or disbelieved. But let us first consider the truth and falsehood of statements, following our practice of going as far as we can with the behaviourists before falling back on introspection. We considered the meaning of words earlier; now we have to consider sentences. Of course a sentence may consist of a single word, or of a wink; but generally it consists of several words. In that case, it has a meaning which is a function of the meanings of the separate words and their order. A sentence which has no meaning is not true or false; thus it is only sentences as vehicles of a certain sort of meaning that have truth or falsehood. We have therefore to examine the meaning of a sentence.

Let us take some very humble example. Suppose you look in a time-table and find it there stated that a passenger train leaves King’s Cross for Edinburgh at 10 A.M. What is the meaning of this assertion? I shudder when I think of its complexity. If I were to try to develop the theme adequately, I should be occupied with nothing else till the end of the present volume, and then I should have only touched the fringe of the subject. Take first the social aspect: it is not essential that anybody but the engineer and fireman should travel by the train, though it is essential that others should be able to travel by it if they fulfil certain conditions. It is not essential that the train should reach Edinburgh: the statement remains true if there is an accident or breakdown on the way. But it is essential that the management of the railway should intend it to reach Edinburgh. Take next the physical aspect: it is not essential, or even possible, that the train should start exactly at ten; one might perhaps say that it must not start more than ten seconds before its time or more than fifty seconds after, but these limits cannot be laid down rigidly. In countries where unpunctuality is common they would be much wider. Then we must consider what we mean by “starting”, which no one can define unless he has learnt the infinitesimal calculus. Then we consider the definitions of King’s Cross and Edinburgh, both of which are more or less vague terms. Then we must consider what is meant by a “train”. Here there will be first of all complicated legal questions; what constitutes fulfilment of a railway company’s obligations in the way of running “trains”? Then there are questions as to the constitution of matter, since evidently a train is a piece of matter; also of course there are questions as to methods of estimating Greenwich time at King’s Cross. Most of the above points have to do with the meaning of single words, not with the meaning of the whole sentence. It is obvious that the ordinary mortal does not trouble about such complications when he uses the words: to him a word has a meaning very far from precise, and he does not try to exclude marginal cases. It is the search for precision that introduces complications. We think we attach a meaning to the word “man”, but we don’t know whether to include Pithecanthropus Erectus. To this extent, the meaning of the word is vague.

As knowledge increases, words acquire meanings which are more precise and more complex; new words have to be introduced to express the less complex constituents which have been discovered. A word is intended to describe something in the world; at first it does so very badly, but afterwards it gradually improves. Thus single words embody knowledge, although they do not make assertions.

In an ideal logical language, there will be words of different kinds. First, proper names. Of these, however, there are no examples in actual language. The words which are called proper names describe collections, which are always defined by some characteristic; thus assertions about “Peter” are really about everything that is “Peterish”. To get a true proper name, we should have to get to a single particular or a set of particulars defined by enumeration, not by a common quality. Since we cannot acquire knowledge of actual particulars, the words we use denote, in the best language we can make, either adjectives or relations between two or more terms. In addition to these, there are words indicative of structure: e.g. in “A is greater than B”, the words “is” and “than” have no separate meaning, but merely serve to show the “sense” of the relation “greater”, i.e. that it goes from A to B, not from B to A.

Strictly speaking, we are still simplifying. True adjectives and relations will require particulars for their terms; the sort of adjectives we can know, such as “blue” and “round”, will not be applicable to particulars. They are therefore analogous to the adjective “populous” applied to a town. To say “this town is populous” means “many people live in this town”. A similar transformation would be demanded by logic in all the adjectives and relations we can know empirically. That is to say, no word that we can understand would occur in a grammatically correct account of the universe.

Leaving on one side the vagueness and inaccuracy of words, let us ask ourselves; in what circumstances do we feel convinced that we know a statement to be true or false as the case may be? A present statement will be regarded as true if, e.g. it agrees with recollection or perception; a past statement, if it raised expectations now confirmed. I do not mean to say that these are the only grounds upon which we regard statements as true; I mean that they are simple and typical, and worth examining. If you say “it was raining this morning”, I may recollect that it was or that it wasn’t. One may perhaps say that the words “this morning” are associated for me with the word “raining” or with the words “not raining”. According to which occurs, I judge your statement true or false. If I have neither association, I do not judge your statement either true or false unless I have material for an inference; and I do not wish to consider inference yet. If you say “the lights have gone out”, when I can see the lights shining, I judge that you speak falsely, because my perception is associated with the words “lights shining”. If you say “the lights will go out in a minute”, you produce a certain familiar kind of tension called “expectation”, and after a time you produce a judgment that you spoke falsely (if the lights do not go out). These are the ordinary direct ways of deciding on the truth or falsehood of statements about past, present, or future.

It is necessary to distinguish between direct and indirect grounds for accepting or rejecting statements. Pragmatism considers only indirect grounds. Broadly speaking, it considers a statement false when the consequences of accepting it are unfortunate. But this belongs to the region of inference. I ask you the way to the station, you tell me wrong, and I miss my train; I then infer that you told me wrong. But if you say “the lights are out” when I see them shining, I reject your statement without inference. In this case, something in my present circumstances is associated with words different from yours, and different in ways which I have learnt to regard as involving incompatibility. The ultimate test of falsehood is never, so I think, the nature of the consequences of a belief, but the association between words and sensible or remembered facts. A belief is “verified” when a situation arises which gives a feeling of expectedness in connection with it; it is falsified when the feeling is one of surprise. But this only applies to beliefs which await some future contingency for verification or refutation. A belief which is an immediate reaction to a situation—e.g. when you are waiting for a race to begin and presently you say “they’re off”—has no need of verification, but verifies other beliefs. And even where the confirmation of a belief is in the future, it is the expectedness, not the pleasantness, of the consequences that confirms the truth of the belief.

I think it is a mistake to treat “belief” as one kind of occurrence, as is done in traditional psychology. The sort of belief which is based upon memory or perception alone differs from the sort which involves expectation. When you find in the time-table that a train leaves King’s Cross at ten, your belief that this statement occurs in the time-table does not await future confirmation, but your belief about the train does: you may go to King’s Cross and see the train start. A belief which concerns an event may be a recollection, a perception, or an expectation. It may be none of these, in the case of an event which you have not seen and do not expect to see—e.g. Cæsar crossing the Rubicon, or the abolition of the House of Lords. But such beliefs always involve inference. I do not at this stage consider logical and mathematical beliefs, some of which must be, in a sense, non-inferential. But I think we shall find that this sense is different from that in which memories and perceptions are non-inferential.

A belief, I should say, interpreted narrowly, is a form of words related to an emotion of one of several kinds. (I shall give a broader meaning later.) The emotion is different according as the belief embodies a reminiscence, a perception, an expectation, or something outside the experience of the believer. Moreover, a form of words is not essential. Where the emotion is present, and leads to action relevant to some feature of the environment, there may be said to be a belief. The fundamental test of a belief, of no matter what sort, is that it causes some event which actually takes place to arouse the emotion of expectedness or its opposite. I do not now attempt to decide what an emotion is. Dr. Watson gives a behaviouristic account of emotions, which would, if adopted, make my definition of “belief” purely behaviouristic. I have framed the definition so as not to involve a decision on the question of introspection.

The subject of truth and falsehood may be subdivided as follows:

It is very hard to disentangle these questions. The first question, as to formal theories, leads to the fourth, as to the relations of propositions to facts. E.g. “Brutus killed Cæsar” is true because of a certain fact; what fact? The fact that Brutus killed Cæsar. This keeps us in the verbal realm, and does not get us outside it to some realm of non-verbal fact by which verbal statements can be verified. Hence our fourth problem arises. But this leads us to our second problem, as to causes and effects of what is true or false, for it is here that we shall naturally look for the vital relation between propositions and facts. And here again we must distinguish between thinking truly and speaking truly. The former is an individual affair, the latter a social affair. Thus all our problems hang together.

I will begin with C, the difference between a belief and a statement. By a “statement” I mean a form of words, uttered or written, with a view to being heard or read by some other person or persons, and not a question, interjection, or command, but such as we should call an assertion. As to the question what forms of words are assertions, that is one for the grammarian and differs from language to language. But perhaps we can say rather more than that. The distinction, however, between an assertion and an imperative is not sharp. In England, notices say “Visitors are requested not to walk on the grass”. In America, they say “Keep off! This means you.” Effectively, the two have the same meaning: yet the English notice consists only of a statement, while the American notice consists of an imperative followed by a statement which must be false if read by more than one person. In so far as statements are intended to influence the conduct of others, they partake of the nature of imperatives or requests. Their characteristic, however, is that they endeavour to effect their aim by producing a belief which may or may not exist in the mind of the speaker. Often, however, they express a belief, without stopping to consider the effect upon others. Thus a statement may be defined as a form of words which either expresses a belief or is intended to create one. Our next step, therefore, must be the definition of “belief”.

“Belief” is a word which will be quite differently defined if we take an analytic point of view from the way in which we shall define it if we regard the matter causally. From the point of view of science, the causal point of view is the more important. Beliefs influence action in certain ways; what influences action in these ways may be called a belief, even if, analytically, it does not much resemble what would ordinarily be so called. We may therefore widen our previous definition of belief. Consider a man who goes to the house where his friend used to live, and, finding he has moved, says, “I thought he still lived here”, whereas he acted merely from habit without thought. If we are going to use words causally, we ought to say that this man had a “belief” and therefore a “belief” will be merely a characteristic of a string of actions. We shall have to say: A man “believes” a certain proposition p if, whenever he is aiming at any result to which p is relevant, he acts in a manner calculated to achieve the result if p is true, but not otherwise. Sometimes this gives definite results, sometimes not. When you call a telephone number, it is clear that you believe that to be the number of the subscriber you want. But whether you believe in the conservation of energy or a future life may be harder to decide. You may hold a belief in some contexts and not in others; for we do not think in accordance with the so-called “Laws of Thought”. “Belief” like all the other categories of traditional psychology, is a notion incapable of precision.

This brings me to the question whether the truth or falsehood of a belief can be determined either by its causes or by its effects. There is, however, a preliminary difficulty. I said just now that A believes p if he acts in a way which will achieve his ends if p is true. I therefore assumed that we know what is meant by “truth”. I assumed, to be definite, that we know what is meant by “truth” as applied to a form of words. The argument was as follows: From observation of a person’s acts, you infer his beliefs, by a process which may be elaborate as the discovery of Kepler’s Laws from the observed motions of the planets. His “beliefs” are not assumed to be “states of mind”, but merely characteristics of series of actions. These beliefs, when ascertained by observation, can be expressed in words; you can say, e.g. “This person believes that there is a train from King’s Cross at 10 A.M.” Having once expressed the belief in words of which the meaning is known, you have arrived at the stage where formal theories are applicable. Words of known meaning, put together according to a known syntax, are true or false in virtue of some fact, and their relation to this fact results logically from the meanings of the separate words and the laws of syntax. This is where logic is strong.

It will be seen that, according to what we have said, truth is applicable primarily to a form of words, and only derivatively to a belief. A form of words is a social phenomenon, therefore the fundamental form of truth must be social. A form of words is true when it has a certain relation to a certain fact. What relation to what fact? I think the fundamental relation is this: a form of words is true if a person who knows the language is led to that form of words when he finds himself in an environment which contains features that are the meanings of those words, and these features produce reactions in him sufficiently strong for him to use words which mean them. Thus “a train leaves King’s Cross at 10 A.M.” is true if a person can be led to say, “It is now 10 A.M., this is King’s Cross, and I see a train starting”. The environment causes words, and words directly caused by the environment (if they are statements) are “true”. What is called “verification” in science consists in putting oneself in a situation where words previously used for other reasons result directly from the environment. Of course, given this basis, there are innumerable indirect ways of verifying statements, but all, I think, depend upon this direct way.

The above theory may be thought very odd, but it is partly designed to meet the fourth of our previous questions, namely, “How can we get outside words to the facts which make them true or false?” Obviously we cannot do this within logic, which is imprisoned in the realm of words; we can only do it by considering the relations of words to our other experiences, and these relations, in so far as they are relevant, can hardly be other than causal. I think the above theory, as it stands, is too crude to be quite true. We must also bring in such things as expectedness, which we discussed earlier. But I believe that the definition of truth or falsehood will have to be sought along some such lines as I have indicated.

I want in conclusion to indulge in two speculations. The first concerns a possible reconciliation of behaviourism and logic. It is clear that, when we have a problem to solve, we do not always solve it as the rat does, by means of random movements; we often solve it by “thinking”, i.e. by a process in which we are not making any overt movements. The same thing was sometimes true of Köhler’s chimpanzees. Now what is involved in the possibility of solving a problem by verbal thinking? We put words together in various ways which are not wholly random, but limited by previous knowledge of the sort of phrase that is likely to contain a solution of our problem. At last we hit upon a phrase which seems to give what we want. We then proceed to an overt action of the kind indicated by the phrase; if it succeeds, our problem is solved. Now this process is only intelligible if there is some connection between the laws of syntax and the laws of physics—using “syntax” in a psychological rather than a grammatical sense. I think this connection is assumed in logic and ordinary philosophy, but it ought to be treated as a problem requiring investigation by behaviourist methods. I lay no stress on this suggestion, except as giving a hint for future investigations. But I cannot think that the behaviourist has gone far towards the solution of his problem until he has succeeded in establishing a connection between syntax and physics. Without this, the efficacy of “thought” cannot be explained on his principles.

My second speculation is as to the limitations which the structure of language imposes upon the extent of our possible knowledge of the world. I am inclined to think that quite important metaphysical conclusions, of a more or less sceptical kind, can be drawn from simple considerations as to the relation between language and things. A spoken sentence consists of a temporal series of events; a written sentence is a spatial series of bits of matter. Thus it is not surprising that language can represent the course of events in the physical world; it can, in fact, make a map of the physical world, preserving its structure in a more manageable form, and it can do this because it consists of physical events. But if there were such a world as the mystic postulates, it would have a structure different from that of language, and would therefore be incapable of being verbally described. It is fairly clear that nothing verbal can conform or confute this hypothesis.

A great deal of the confusion about relations which has prevailed in practically all philosophies comes from the fact that relations are indicated, not by relations, but by words which are as substantial as other words. Consequently, in thinking about relations, we constantly hover between the unsubstantiality of the relation itself and the substantiality of the word. Take, say, the fact that lightning precedes thunder. We saw earlier that to express this by a language closely reproducing the structure of the fact, we should have to say simply: “lightning, thunder”, where the fact that the first word precedes the second means that what the first word means precedes what the second word means. But even if we adopted this method for temporal order, we should still need words for all other relations, because we could not without intolerable ambiguity symbolise them also by the order of our words. When we say “lightning precedes thunder”, the word “precedes” has a quite different relation to what it means from that which the words “lightning” and “thunder” have to what they respectively mean. Wittgenstein12 says that what really happens is that we establish a relation between the word “lightning” and the word “thunder”, namely the relation of having the word “precedes” between them. In this way he causes relations to be symbolised by relations. But although this may be quite correct, it is sufficiently odd to make it not surprising that people have thought the word “precedes” means a relation in the same sense in which “lightning” means a kind of event. This view, however, must be incorrect. I think it has usually been held unconsciously, and has produced many confusions about relations which cease when it is exposed to the light of day—for example, those which lead Bradley to condemn relations.

In all this I have been considering the question of the relation between the structure of language and the structure of the world. It is clear that anything that can be said in an inflected language can be said in an uninflected language; therefore, everything that can be said in language can be said by means of a temporal series of uninflected words. This places a limitation upon what can be expressed in words. It may well be that there are facts which do not lend themselves to this very simple schema; if so, they cannot be expressed in language. Our confidence in language is due to the fact that it consists of events in the physical world, and, therefore, shares the structure of the physical world, and therefore can express that structure. But if there be a world which is not physical, or not in space-time, it may have a structure which we can never hope to express or to know. These considerations might lead us to something like the Kantian a priori, not as derived from the structure of the mind, but as derived from the structure of language, which is the structure of the physical world. Perhaps that is why we know so much physics and so little of anything else. However, I have lapsed into mystical speculation, and will leave these possibilities, since, by the nature of the case, I cannot say anything true about them.