In this classic essay, written in the heat of logical positivism (1936-37), Ayer covers a lot of ground, including: (i) the correspondence theory of truth, (ii) the coherence theory of truth, and (iii) "protocol-statements".

For (ii) and (iii) Ayer has in mind those defenses and analyses given by leading philosophers of his era: Carnap, Hempel, and Neurath. Ayer finds in all of them a failure to appreciate the notion that one requires verification for each of those alleged (what the German philosophers call) Protokollsätze that make reference to immediate experience, an obvious feature of the correspondence theory of truth. To make sense of this, in the end, Ayer takes refuge in the suggestion from G. E. Moore that "we cannot be mistaken about them ['protocol-statements'] in the way that we can be mistaken about other empirical propositions."

Famous Works Mentioned/Quoted/Referenced: (1) Pragmatism, William James; (2) Logik der Forschung, Karl Popper; (3) The Logical Syntax of Language, Rudolf Carnap; (4) The Unity of Science, Rudolf Carnap; (5) Foundations of Mathematics, F. P. Ramsey.

Allen, J. T. © 2007

Verification and Experience by A. J. Ayer

WHAT IS IT that determines the truth or falsehood of empirical propositions? The customary answer is, in effect, that it is their agreement or disagreement with reality. I say "in effect" because I wish to allow for alternative formulations. There are some who would speak of correspondence or accordance rather than agreement; some who for the word "reality" would substitute "facts" or "experience." But I do not think that the choice of different words here reflects any important difference of meaning. This answer, though I believe it to be correct, requires some elucidation. To quote William James; "Pragmatists and Intellectualists both accept (it) as a matter of course. They begin to quarrel only after the question is raised as to what precisely may be meant by the term 'agreement' and what by the term 'reality' when reality is taken as something for our ideas to agree with."[1] I hope at least to throw some light upon this question in the course of this paper. It will simplify our undertaking if we can draw a distinction between those empirical propositions whose truth or falsehood can be determined only by ascertaining the truth or falsehood of other propositions and those whose truth or falsehood can be determined directly by observation. To the former class belong all universal propositions.

     We cannot, for example, directly establish the truth or falsehood of the proposition that gold is dissoluble in aqua regia, unless of course we regard this as a defining attribute of gold and so make the proposition a tautology. We test it by establishing the truth or falsehood of singular propositions relating, among other things, to particular pieces of gold. We may indeed deduce one universal proposition from another, or even infer it by analogy, but in all such cases we must finally arrive at a proposition for which the evidence consists solely in the truth or falsehood of certain singular propositions.

    It is here to be remarked that no matter how many such singular propositions we succeed in establishing we are never entitled to regard the universal proposition as conclusively verified. However often we may have observed the dissolution of pieces of gold in aqua regia, we must still allow it to be possible that the next piece with which we experiment will not so dissolve. On the other hand the falsity of any one of the relevant singular propositions does entail the falsity of the universal proposition. It is this logical assymetry in the relationship of universal and singular propositions that has led some philosophers[2] to adopt the possibility of falsification rather than that of verification as their criterion of empirical significance. We said that the way to test the validity of a universal proposition about the dissolubility of gold was to ascertain the truth or falsehood of singular propositions referring to particular pieces of gold. But these propositions in their turn depend for their verification upon the verification of other propositions. For a piece of gold is a material thing; and to test the validity of propositions referring to material things we must ascertain the truth or falsehood of propositions referring to sense-data. Here we have another instance of logical assymetry.

    A proposition referring to a material thing may entail propositions referring to sense-data but cannot itself be entailed by any finite number of them. Now at last we seem to have reached propositions which need not wait upon other propositions for the determination of their truth or falsehood, but are such that they can be directly confronted with the given facts. These propositions I propose to call basic propositions. If the distinction which we have drawn between them and other propositions is legitimate, we may confine ourselves, for our present purpose, to questions concerning the nature of basic propositions and the manner in which our determination of their validity depends upon our experience. It is noteworthy that the legitimacy of the distinction which we have drawn is implicitly acknowledged even by philosophers who reject the notion of agreement with reality as a criterion of truth. Neurath and Hempel, for example, have recently been maintaining that it is nonsensical to speak of comparing propositions with facts or reality or experience.[3] A proposition, they say, can be compared only with another proposition. At the same time they assign a status corresponding to that of our basic propositions to a class of propositions which they call protocol propositions.

   According to Neurath, for a sentence to express a protocol proposition it is necessary that it should contain the name or description of an observer and some words referring to an act of observation. He gives the following as an example. "Otto's protocol at 3.17/Otto's speech-thought at 3.16 was (there was in the room at 3.15 a table observed by Otto)/." This is not regarded by Neurath as the only legitimate way of formulating a protocol proposition. If others care to adopt a different convention, they are, as far as he is concerned, at liberty to do so. But he claims for the particular form that he has chosen that it has the advantage of giving protocol propositions greater stability than they might otherwise have. It is easy enough to see why he says this. He is thinking of the case in which it turns out that Otto has been having a hallucination or that in which he is found to be lying. In the former case the proposition in the interior bracket must be held to be false; in the latter, the proposition in the main bracket. But the whole proposition is not a truth-function of the propositions within the brackets, any more than they are truth-functions of one another. We may therefore continue to accept it even when we have rejected them. In itself, this is a valid point. But it is surely inconsistent with Neurath's main position. For how, if we are debarred from appealing to the facts, can we ever discover that Otto has lied or had a hallucination? Neurath makes the truth and falsehood of any proposition whatsoever depend upon its compatibility or incompatibility with other propositions. He recognizes no other criterion. In this respect, his protocol propositions are not allowed any advantage. If we are presented with a protocol proposition and also with a non-protocol proposition which is incompatible with it we are not obliged to accept the protocol proposition and reject the other. We have an equal right to reject either.

   But if this is so we need not bother to devise a special form for protocol propositions in order to ensure their stability. All we have to do if we wish a proposition to be stable is to decide or accept it and to reject any proposition that is incompatible with it. The question whether such a decision is empirically justified or not is one which, according to the implications of Neurath's doctrine, no meaning can be attached. One wonders indeed why he and Hempel pay so much attention to protocol propositions, inasmuch as the only distinction which they are able draw between them and other propositions is a distinction of form. They do not mean by a protocol proposition one which can be directly verified by observation, for they deny that this is possible. They use the term "protocol" purely for a syntactical designation for a certain assemblage of words. But why should one attach special significance to the word "observation"? It may be that there is no error involved in constructing sentences of a peculiar type and dignifying them with the title Protokollsätze, but it is arbitrary and misleading. There is no more justification for it than there would be for making a collection of all the propositions that could be correctly expressed in English by sentences beginning with the letter B, and choosing to call them Basic propositions. If Neurath and Hempel do not recognize this it is probably because, in writing about Protokollsätze, they unconsciously employ the forbidden criterion of agreement with experience.

    Though they say that the term "protocol" is nothing more than a syntactical designation, they do not use it merely as such. We shall see later that Carnap equivocates with this term in a similar way. It is not, however, a sufficient reason for rejecting a theory that some of its advocates have failed consistently to adhere to it. And it is necessary for us to investigate more closely the view that in order determine the validity of a system of empirical propositions one cannot and need not go beyond the system itself. For if this view were satisfactory we should be absolved from troubling any further about the use of the phrase "agreement with experience." The theory which we now have to examine is that which is commonly known as the coherence theory of truth. It should be noted that the theory is not, as we interpret it, concerned with the definition of truth or falsehood but only with the means by which they are determined. According to it a proposition is to be accepted if it is found to be compatible with other accepted propositions, rejected if it is not. If, however, we are anxious to accept a proposition which conflicts with our current system we may abandon one or more of the propositions which we had previously accepted. In such a case we should, it is sometimes said, be guided by a principle of economy. We should make the smallest transformation of the system which ensured self-consistency. I think it is usually assumed also that we have, or ought to have, a preference for large and highly integrated systems; systems containing a great number of propositions which support one another to a high degree.

    One strong objection to this theory is well put by Professor Price in his lecture Truth and Corrigibility. "Suppose," he says, "we have a group of mutually supporting judgments. The extraordinary thing is that however large the group may be, and however great the support which the members give to each other, the entire group hangs, so to speak, in the air. If we accept one member, no doubt it will be reasonable to accept the rest. But why must we accept any of them? Why should we not reject the whole lot? Might they not all be false, although they all support each other?"[4] He goes on to argue that we cannot consider such a system of judgments to have even any probability unless we can attribute to at least one of its constituents a probability which is derived from some other ground than its membership of the system. He suggests therefore that the only way to save the theory would be to maintain that some propositions were intrinsically probable. But this, though he does not say so, is to reduce it to absurdity. There is no case at all to be made out for the view that a proposition can be probable independently of all evidence. The most that could be said in favor of anyone who accepted Price's suggestion would be that he had chosen to give the word "probability" an unfamiliar sense. A point which Price appears to have overlooked is that according to a one well-known version of the coherence theory there can be only one completely coherent system of propositions. If this were so the theory would give us at least an unequivocal criterion for determining the truth of any proposition; namely, the possibility of incorporating it in this single system. It would not, however, afford us any ground for supposing that the enlargement of an apparently coherent system of propositions increased its probability. On the contrary, we ought rather to hold that it decreased it.

  For ex hypothesi any set of propositions which is internally coherent is the only one that is so. If, therefore, we have a set of propositions which appears to be self-consistent, either it is the unique coherent system or it contains a contradiction which we have failed to discover; and the larger the set the greater the probability that it contains a contradiction which we have failed to discover. But in saying this we are assuming the truth of a proposition about the limited powers of the human understanding, which may or may not find a place in the one coherent system. Perhaps, therefore, it would be better to say that the advocates of this form of the coherence theory dispense with the notion of probability altogether. But now we must ask, Why should it be assumed that only one completely coherent system of propositions is conceivable? However many empirical propositions we succeed in combining into an apparently self-consistent system we seem always able to construct a rival system which is equally extensive, appears equally free from contradiction, and yet is incompatible with the first.

    Why should it be held that at least one of these systems must contain a contradiction, even though we are unable to detect it? I can see no reason at all for this assumption. We may not be able to demonstrate that a given system is free from contradiction; but this does not mean that it is probable that it contains one. This indeed is recognized by the more recent advocates of what we are calling the coherence theory. They admit the possibility of inventing fictitious sciences and histories which would be just as comprehensive, elegant and free from contradiction as those in which we actually believe. But how then do they propose to distinguish the true systems from the false? The answer given[5] is that the selection of the true system does not depend on internal features of the system itself. It cannot be effected by purely logical means. But it can be carried out inside the realm of descriptive syntax. We are to say that the true system is that which is based upon true protocol propositions; and that true protocol propositions are those which are produced by accredited observers, including notably the scientists of our era. Logically, it might be the case that the protocol propositions which each of us expressed were so divergent that no common system of science or only a very meagre system could be based upon them. But fortunately this is not so. People do occasionally produce inconvenient protocol propositions. But being in a small minority they are over-ridden. They are said to be bad observers or liars or, in extreme cases, mad. It is a contingent, historical fact that the rest of us agree in accepting an "increasingly comprehensive, common, scientific system." And it is to this, so the theory runs, that we refer when out of the many coherent systems of science that are conceivable we speak of only one as being true. This in an ingenious answer; but it will not do.

One reason why we trust "the scientists of our era" is that we believe that they give an accurate account of their observations. But this means that we shall be involved in a circle if we say that the reason we accept certain evidence is merely that it comes from the scientists of our era. And furthermore, How are we to determine that particular system is accepted by contemporary scientists expect by appealing to the facts of experience? But once it is conceded that such an appeal is possible there is no longer any need to bring in the contemporary scientists. However great our admiration for the achievements of the scientists of our era we can hardly maintain that it is only with reference to their behavior that the notion of agreement with reality has any meaning. Hempel[6] has indeed attempted to meet this objection by telling us that instead of saying that "the system of protocol-statements which we call true may only be characterized by the historical fact that it is actually adopted by the scientists of our culture circle" we ought to express ourselves "formally" and say: "The following statement is sufficiently confirmed by the protocol-statements adopted in our science; 'Amongst the numerous imaginable consistent sets of protocol-statements, there is in practice exactly one which is adopted by the vast majority of instructed scientific observers; at the same time, it is just this set which we generally call true.'" But this does not remove the difficulty.

   For now we must ask, How is it determined that the protocol-statements which support the statement quoted really are adopted in our science? If Hempel is really speaking formally, as he says he is, then the phrase "adopted in our science" must be regarded merely as an arbitrary syntactical designation of a certain set of sentences. But it is clear that he does not intend it to be nothing more than this. He intends it to convey the information that the propositions expressed by these sentences actually are adopted. But this is to re-introduce the reference to historical fact which he is trying to eliminate. We have here a fallacy which is akin to the fallacy of the ontological argument. It is not legitimate to use the phrase "adopted in our science" simply as a means of naming certain statements and then proceed to infer from this that these statements really are adopted in it. But Hempel cannot dispense with this fallacious inference. For each of many incompatible systems might contain the statement that it alone was accepted by contemporary scientists, together with the protocol propositions that were needed to support it. We may conclude then that the attempt to lay down a criterion for determining the truth of empirical propositions which does not contain any reference to "facts" or "reality" or "experience," has not proved successful. It seems plausible only when it involves a tacit introduction of that very principle of agreement with reality which it is designed to obviate. Accordingly, we may return to our original question concerning the nature of basic propositions and the manner in which their validity depends upon fact. And first of all I wish to consider how far this question admits of a purely conventional answer.

     According to Professor Carnap it is wholly a matter of convention what propositions we take as basic. "Every concrete proposition," he tells us,[7] "belonging to the physicalistic system-language can in suitable circumstances serve as a protocol proposition. Let G be a law (that is a general proposition belonging to the system language). For the purpose of verification one must in the first instance derive from G concrete propositions referring to particular space-time points (through substitution of concrete values for the space-time co-ordinates x, y, z, t which occur in G as free variables). From these concrete propositions one may with the help of additional laws and logico-mathematical rules of inference derive further concrete propositions, until one comes to propositions which in the particular case in question one is willing to accept. It is here a matter of choice which propositions are employed at any given time as the terminating point of this reduction, which serves the purpose of verification, must be brought to an end somewhere. But one is never obliged to call a halt at any one point rather than another." In reasoning thus, Carnap says that he is following the example of Karl Popper. Actually Popper adopts a rather narrower convention. He proposes, and takes the view that there can in this matter be no warrant for anything more than a proposal, that basic proposition should have the form of singular existentials. They must, according to his convention, refer to particular spatio-temporal points and the events which are said to be occurring at these points must be observable events.

     But in case anyone should think that the use of the word "observable" brings an element of psychology he hastens to add that instead of an "observable" even he might equally well have spoken of an event of motion located in (macroscopic) physical bodies.[8] His views concerning the verification of these propositions are summed up as follows: "The basic propositions are accepted by an act of will, by convention. Sie sind Festsetzungen."[9] The verification of all other empirical propositions is held to depend upon that of the basic propositions. So that if we take the remark I have quoted literally, we are presented with the view that our acceptance or rejection of any empirical proposition must be wholly arbitrary. And this is surely wrong. Actually, I do not think that Popper himself wishes to maintain this. His stipulation that basic propositions should refer to observable events suggests that he recognizes that our acceptance of them somehow depends upon our observations. But he does not tell us how. There is indeed this much truth in what Popper says. The propositions which he calls basic refer to material things. As such, they can be tested by observation, but never conclusively established. For, as we have already remarked, although they may entail propositions referring to sense-data they cannot be entailed by them. It follows that there is in our acceptance of them an element of convention.

     I cannot carry out all the tests which would bear upon the truth of even so simple a proposition as that my pen is lying on my table. In practice, therefore, I accept such a proposition after making only a limited number of tests, perhaps only a single test, which leaves it still possible that it is false. But this is not to say that my acceptance of it is the result of an arbitrary decision. I have collected some evidence in favor of the proposition, even though it may not be conclusive evidence. I might have accepted it without having any evidence at all; and then my decision would, in fact, have been arbitrary. There is no harm in Popper's insisting that our acceptance of such propositions as he calls basic is not wholly dictated by logic; but he ought still to distinguish the cases in which our acceptance of a "basic" proposition is reasonable from those in which it is not. We may say that it is reasonable when the proposition is supported by our observations. But what is meant by saying that a proposition is supported by our observations? This is a question which in his discussion of the "Basic-problem" Popper does not answer.

    We find, therefore, that this "discovery" or Popper's which has been fastened on to by Carnap amounts to no more than this; that the process of testing propositions referring to physical objects can be extended as far as we choose. What is conventional is our decision to carry it in any given case just so far and no farther. To express this, as Carnap does, by saying that it is a matter of convention what propositions we take as protocols is simply to give the term "protocol proposition" an unfamiliar meaning. We understand that he now proposes to use it to designate any singular proposition, belonging to "the physicalistic system-language," which we are prepared to accept without further tests.

    This is a perfectly legitimate usage. What is not legitimate is to ignore the discrepancy between it and his former usage according to which protocol propositions were said to "describe directly given experience." And in abandoning the original usage he has incidentally shelved the problem which it was designed to meet. Elsewhere,[10] Carnap has suggested that problems concerning the nature of basic propositions, in our sense of the term, depend for their solution only on conventions about forms of words. I think that this, too, can be shown to be a mistake. Most people are by now familiar with his division of propositions into factual propositions such as "the roses in my garden are red," pseudo-factual propositions such as "a rose is a thing," which are also said to be syntactical propositions, expressed in the material mode of speech, and propositions such as "'rose' is a thing-word," which are syntactical and expressed in the formal mode of speech.

     Now when he raises the question "What objects are the elements of given, direct experience?" he treats it as if it were a syntactical question, expressed in the material mode of speech. That is, he considers it to be a loose way of raising the question "What kinds of word occur in protocol-statements?"[11] And he sets out various possible answers both in what he calls the material and in what he calls the formal mode. Thus, he says that it may be the case that "the elements that are directly given are the simplest sensations and feelings" or "more complex objects such as partial gestalts of single sensory fields" or that "material things are elements of the given"; and he takes these to be misleading ways of saying that "protocol-statements are of the same kind as 'joy now,' 'here, now, blue'" or that "protocol-statements are of the forms similar to 'red circle, now'" or that they have "approximately the same kind of form as 'a red cube is on the table.'"[12] In this way he assumes that questions about the nature of immediate experience are linguistic in character. And this leads him to dismiss all the "problems of the so-called given or primitive data" as depending only upon our choice of a form of language.[13] But this is to repeat the error of Neurath and Hempel, which we have already exposed. If the term "protocol-statement" was being used merely as a syntactical designation for certain combinations of symbols then our choice of the sentences to which we applied it would indeed be a matter of convention. It would involve no more reference to truth than a decision to apply the designation "basic" to all English sentences beginning with B. But this is not the sense in which Carnap is supposed to be using the term.

     He is using it not to mark out the form of certain statements, but rather to express the fact that they refer to what is immediately given. Accordingly, our answer to his question "What kinds of word occur in protocol-statements?" cannot depend simply upon a conventional choice of linguistic forms. It must depend upon the way in which we answer the question "What objects are the elements of the given, direct experience?" And this is not a matter of language, but a matter of fact. It is a plain question of fact whether the atomistic or the gestalt theory of sensation is correct. Thus we see that the proposition that "the elements that are directly given are the simplest sensations and feelings" which Carnap takes to be a syntactical proposition expressed in the material mode of speech, is not syntactical at all. And the proposition which he gives as its formal equivalent, namely, that "protocol-statements are of the same kind as: 'joy now,' 'here, now, blue; there, red'" is not syntactical either. If we want to give it a label we may call it a pseudo-syntactical proposition.

     And by this we shall mean that it seems to be about words but is really about objects. It is important that the existence of such propositions should not be overlooked; for they are quite as dangerous in their way as the pseudo-factual propositions of which Carnap has made so much. In this instance the source of confusion is the use of the term "protocol." It cannot without contradiction be interpreted both as a purely formal designation and as involving a covert reference to a matter of fact. But this is precisely how Carnap does interpret it; and it is thus that he is led to make the mistake of supposing that questions about the nature of basic propositions can be decided merely by convention. It is indeed a matter of convention that we should use a word consisting of the letter "j o y" to denote joy. But the proposition that joy is immediately experienced, which is implied in saying that "joy" is a protocol word, is one whose truth or falsehood is not to be decided by convention but only by referring to the facts.

     The psychology of sensation is not an a priori branch of science. We conclude therefore that the forms of basic propositions depend partly indeed upon linguistic conventions but partly also upon the nature of the given; and this is something that we cannot determine a priori. We may hold indeed that a person's sensations are always private to himself; but this is only because we happen so to use the words that it does not make sense to say "I am acquainted with your sense-data" or "You and I are experiencing the same sense-datum."[14] This is a point about which we are apt to be confused. One says mournfully "I cannot experience your toothache" as though it revealed a lack of mental power. That is, we are inclined to think of the contents of another person's mind, or the immediate objects of his experience, as being concealed from us by some sort of natural obstacle, and we say to ourselves: "If only we had a ray which would penetrate this obstacle!" (Intuition!) or "Perhaps we can construct a reflector which will show us what is going on behind." But in fact there is no obstacle but our usage of words.

   To say that whatever is directly "given" to me is mine and mine only is to express a tautology. A mistake which I, for one, have made in the past is to confuse this with the proposition "Whatever is directly 'given' is mine." This is not a tautology. It is an empirical proposition, and it is false. A further point which it is advisable to make clear is that we are not setting any arbitrary boundaries to the field of possible experience. As an illustration of this let us consider the case of the man who claims to have an immediate, non-sensory experience of God. So long as he uses the word "God" simply as a name for the content of his experience, I have no right to disbelieve him. Not having such experiences myself I cannot understand him fully. I do not myself know what it is like to be acquainted with God. But I can at least understand that he is having some experience of a kind that I do not have. And this I may readily believe. I should certainly not be justified in assuming that the sort of experiences that I myself had were the only sort that could be had at all.

   At the same time it must be remarked that "God," in this usage, cannot be the name of a transcendent being. For to say that one was immediately acquainted with a transcendent being would be self-contradictory. And though it might be the name of a person who in fact endured for ever one could not say that one was immediately acquainted with Him as enduring for ever. For this, too, would be self-contradictory. Neither would the fact that people were acquainted with God, in this sense, afford a valid ground for inferring that the world had a first cause, or that human beings survived death, or in short that anything existed which had the attributes that are popularly ascribed to God. And the same thing applies to the case of moral experience. We should certainly not be justified in denying a priori the possibility of moral experience. But this does not mean that we recognize that there is ground for inferring the existence of an ideal, objective world of values. It is necessary to say this because the use of "God" or "value" as a designation of the content of a certain kind of experience often misleads people into thinking that the are entitled to draw such inferences; and we must make it clear that in admitting the possibility of such experiences we are not also upholding the conclusions which are illegitimately drawn from them. We have tried to show that neither the form nor the validity of basic propositions is dependent merely on convention.

    Since it is their function to describe what can be immediately experienced, their form will depend upon the general nature of the "given," their validity upon the agreement with it in the relevant particular case. But what is this relation of agreement? What kind of correspondence do we suppose to exist between basic propositions and the experiences that verify them? It is sometimes suggested that this relation of agreement is of the same kind as that which holds between a picture and that of which it is a picture. I do not think that this is true. It is possible indeed to construct picture-languages; no doubt they have their advantages; but it surely cannot be maintained that they alone are legitimate; or that a language such as English is really a picture-language although we do not know it. But if English is not a picture-language and propositions expressed in English are sometimes verified, as they surely are, then it cannot be the case that this relation of agreement with which we are concerned is one of picturing. Besides, there is this further difficulty. If any propositions are pictures, presumably false propositions are so as well as true ones. In other words, we cannot tell from the form of the proposition, that is, merely by looking at the picture, whether it depicts a real situation or not. But how then are we to distinguish the true picture from the false?

     Must we not say that the true picture agrees with reality whereas the false one does not? But in that case the introduction of the notion of picturing does not serve our purpose. It does not enable us to dispense with the notion of agreement. The same objections hold against those who say that this relation of agreement is one of identify of structure. This is to treat propositions as if they were maps. But then it is to be supposed that a false proposition is also a map. The mere form of the proposition will not tell us whether the country which it purports to map is imaginary or real. Can we then avoid saying that we test the truth of such a map by seeing whether it agrees with reality? But then the notion of agreement is still left unclarified. And, in any case, why should it be assumed that if a proposition is to describe what is directly given it must have the same structure as the given? One might, perhaps, allow the possibility of creating a language in which all basic propositions were expressed by sentences functioning as maps, though I am by no means sure that it would be possible to draw a map of our internal sensations; but I can see no ground at all for assuming that only a language of this kind is legitimate, or that any of the European languages with which I am acquainted is a language of this kind. Yet propositions, expressed in these languages, are frequently verified.

     There is, perhaps, a historical connection between the view that basic propositions must be identical in structure with the facts that verify them and the view that only structure can be known or expressed.[15] But this too is arbitrary, and indeed self-defeating. To maintain that content is inexpressible is to behave like Ramsey's child. "'Say breakfast.' 'Can't.' 'What can't you say?' 'Can't say breakfast.'"[16] What is being assumed in the theories which we have just been discussing is not so much that proposition cannot be verified as that it, or, to speak more accurately, the sentence expressing it, cannot have a sense at all unless it is a picture or a map. The difficulty with regard to sentences that express false proposition is got round by saying that they depict or map possible facts. But surely this assumption is quite gratuitous. If I am speaking English I may use the words "I am angry" to say that I am angry. You may say, if you like, that in doing so I am obeying a meaning-rule[17] of the English language. For this to be possible it is not in the least necessary that my words should in any way resemble the state of anger which they describe. That "this is red" is used to say that this is red does not imply that it bears any relation of resemblance, whether of structure or content, to an actual or hypothetical red patch. But if the words "I am angry" are used to say that I am angry, then it does not seem in any way mysterious that my being angry should verify the proposition that they express. But how do I know that I am angry? I feel it. How do I know that there is now a loud sound? I hear it. How do I know that this is a red-patch? I see it. If this answer is not regarded as satisfactory, I do not know what other can be given.

     It may be suggested that we ought in this connection to introduce the notion of causation. The relation, it may be said, between the proposition "I am in pain" and the fact that verifies it is that the fact causes me to assert the proposition, or at any rate to believe it. That such a relation often exists is not to be denied. But we cannot analyze verification in terms of it. For if I am a habitual liar my being in pain my cause me to deny that I am in pain; and if I am a sufficiently hidebound Christian Scientist it may not cause me to believe it. But in either case my being in pain will verify the proposition that I am in pain. Why? Because when I say "I am in pain" I mean that I am in pain, and if p then p. But how do I establish p? How do I know that I really am in pain? Again the answer can only be "I feel it." Does this mean that basic propositions must be regarded as incorrigible? I find this question difficult to answer because I do not know what precise meaning those who have discussed it have been giving to the term "incorrigible." Probably, different philosophers have given it different meanings.

     Professor Price, for example, when he argues that basic propositions are incorrigible appears to mean no more than that our reasons for accepting them are found in our experience; that if one is justified in saying of a visual sense-datum "this is red," it is because one sees it so. For the only arguments which he gives in favor of the view that some first-order propositions are incorrigible are arguments against the coherence theory of truth.[18] I should of course agree that basic propositions were incorrigible, in this rather unnatural sense. Dr. von Juhos makes the same statement.[19] But what he appears to mean by it is that there can never be any ground for abandoning a basic proposition; that once it is accepted it cannot subsequently be doubted or denied. In a sense, we may agree that this is so. For we may say that what is subsequently doubted or denied is always a different proposition. What I accept now is the proposition "this is red"; what I may doubt or deny in thirty seconds' time is the proposition "I was seeing something red thirty seconds ago." But in this sense every proposition which contains a demonstrative is incorrigible, and not only basic propositions. And if von Juhos wishes to maintain that some special sacrosanctity attaches to propositions which purport to be records of our immediate experiences, I think that he is wrong. If I find the sentence "I feel happy" written in my diary under the heading February 3rd I am not obliged to believe that I really did feel happy on February 3rd, merely because the sentence has the same form as that which I should utter if I felt happy now. I may indeed believe it on the ground that I am not in the habit of writing down false statements in my diary. But that is a different matter.

      Professor Moore has suggested to me that what some of those who say that basic propositions are incorrigible may have in mind is that we cannot be mistaken about them in the way that we can be mistaken about other empirical propositions. If I say "I am in pain" or "this is red" I may be lying, or I may be using the words wrongly; that is, I may be classifying as "pain" or as "red" something that would not normally be so classified. But I cannot be mistaken in any other way. I cannot be mistaken in the way that I can be mistaken if I take this red patch to the cover of a book. If this is a fact, it is not a fact about human psychology. It is not just a merciful dispensation of Providence that we are secured from errors of a certain kind. It is, if anything, a fact about language.[20] If Moore is right, it does not make sense to say "I doubt whether this is red" or "I think that I am in pain but I may be mistaken," unless it is merely meant that I am doubting whether "pain" or "red" is the correct word to use. I believe now that Moore is right on this point. But whether it is a fact from which any important conclusions follow I do not profess to know.

Notes:

[1] Pragmatism, p. 198

[2] Notably Karl Popper. See his Logik der Forschung.

[3] Otto Neurath, "Protokollsätze." Erkenntnis, Volume III, p. 223, and "Radikaler Physikalismus und 'Wirkliche Welt,'" Erkenntnis, Vol. IV; Carl Hempel, "On the Logical Positivists' Theory of Truth," Analysis, Vol. II, "Some Remarks on Empiricism," Analysis, Vol. III, and "Some Remarks on 'Facts' and Propositions," Analysis, Vol. II.

[4] Truth and Corrigibility (Inaugural lecture, Oxford University Press, 1936), p. 19.

[5] E. g., by Rudolf Carnap, "Erwiderung auf die Aufsätze von E. Zilsel und K. Duncker," Erkenntnis, Vol. III, pp. 179-180.

[6] Analysis, Vol. III, pp. 39-40.

[7] "Über Protokollsätze." Erkenntnis, Vol. III, p. 224.

[8] Logik der Forschung, p. 59.

[9] Op. cit., p. 62.

[10] Logical Syntax of Language, pp. 305-6.

[11] The Unity of Science, p. 45.

[12] The Unity of Science, pp. 46-7.

[13] The Logical Syntax of Language, pp. 305-6.

[14] This point has been forcibly made by G. A. Paul, vide "Is there a Problem About Sense-Data?" Supp. Arist. Soc., 1936, also reprinted in A. Flew (ed.), Essays in Logic and Language, First Series.

[15] Cf. E. Zilsel, "Bemerkungen zur Wissenschaftslogik," Erkenntnis, Vol. III, p. 143.

[16] Foundations of Mathematics, p. 268.

[17] Cf. K. Ajdukiewicz, "Sprache und Sinn," Erkenntnis, Vol. IV, pp. 114-116.

[18] Vide Truth and Corrigibility.

[19] See his articles in Analysis, Vol. II, and Erkenntnis, Vol. IV.

[20] Cf. John Wisdom, "Philosophical Perplexity," Proc. Arist. Soc., 1936-7, p. 81.