'Basic Propositions', as I wish to use the term, are a subclass of epistemological premisses, namely those which are caused, as immediately as possible, hy perceptive experiences. This excludes the premisses required for inference, whether demonstrative or probable. It excludes also any extra-logical premisses used tor inference, if there be such -- e.g., 'what is red is not blue', 'if A is earlier than B, B is not earlier than A'. Such propositions demand careful discussion, but whether premisses or not, they are in any case not 'basic' in the above sense.

I have borrowed the term 'basic proposition' from Mr A. J. Ayer, who uses it as the equivalent of the German Protokollsatz employed by the logical positivists. I shall use it, perhaps, not in exactly the same sense in which it is used by Mr Ayer, but I shall use it in connexion with the same problems as those which have led him and the logical positivists to require such a term.

Many writers on theory of knowledge hold that from a single occurrence nothing is to be learnt. They think of all empirical knowledge as consisting of inductions from a number of more or less similar experiences. For my part, I think that such a view makes history impossible and memory unintelligible. I hold that, from any occurrence that a man notices, he can obtain knowledge, which, if his linguistic habits are adequate, he can express in sentences. His linguistic habits, of course, have been generated by past experiences, but these only determine the words he uses. The truth of what he says, given the meanings of his words, can, given adequate care, be wholly dependent upon the character of one occurrence that he is noticing. When this is the case, what he is asserting is what I call a 'basic proposition'.

The discussion of basic propositions has two parts. First, it is necessary to argue, as against opposing opinions, that there [131] are basic propositions. Secondly, it is necessary to determine just what sort of thing they can affirm, and to show that this is usually much less than common sense asserts on the occasions on which the basic propositions in question are epistemologically justifiable.

A basic proposition is intended to have several characteristics. It must be known independently of inference from other propositions, but not independently of evidence, since there must be a perceptive occurrence which gives the cause and is considered to give the reason for believing the basic proposition. Then again, from a logical point of view, it should be possible so to analyse our empirical knowledge that its primitive propositions (apart from logic and generalities) should all have been, at the moment when they were first believed, basic propositions. This requires that basic propositions should not contradict each other, and makes it desirable, if possible, to give them a logical form which makes mutual contradiction impossible. These conditions demand, therefore, that a basic proposition should have two properties:

1. It must be caused by some sensible occurrence;
2. It must be of such a form that no other basic proposition can contradict it.

As to (1): I do not wish to insist upon the word 'caused', but the belief must arise on the occasion of some sensible occurrence, and must be such that, if questioned, it will be defended by the argument 'why, I see it' or something similar. The belief refers to a certain time, and the reasons for believing it did not exist before that time. If the event in question had been previously inferred or expected, the evidence beforehand was different from that afforded by perception, and would generally be considered less decisive. Perception affords for the belief evidence which is considered the strongest possible, but which is not verbal.

As to (2): the judgements that common sense bases upon perception, such as 'there is a doe', usually go beyond the present datum, and may therefore be refuted by subsequent evidence. We cannot know, from perception alone, anything [132] about other times or about the perceptions of others or about bodies understood in an impersonal sense. That is why, in the search for data, we are driven to analysis: we are seeking a core which is logically independent of other occurrences. When you think you see a dog, what is really given in perception may be expressed in the words 'there is a canoid patch of colour'. No previous or subsequent occurrence, and no experience of others, can prove the falsehood of this proposition. It is true that, in the sense in which we infer eclipses, there can be evidence against a present judgement of perception, but this evidence is inductive and merely probable, and cannot stand against 'the evidence of the senses'. When we have analysed a judgement of perception in this way, we are left with something which cannot be proved to be false.

We may then define a 'basic proposition' as follows: it is a proposition which arises on occasion of a perception, which is the evidence for its truth, and it has a form such that no two propositions having this form can be mutually inconsistent if derived from different percepts.

Examples would be: 'I am hot', 'that is red', 'what a foul smell'. All basic propositions in the above sense are personal, since no one else can share my percepts, and transitory, for after a moment they are replaced by memories.

In place of the above definition, we can adopt a logical definition. We can consider the whole body of empirical knowledge, and define 'basic propositions' as those of its logically indemonstrable propositions which are themselves empirical, i.e., assert some temporal occurrence. This definition, I think, is extensionally equivalent to the above epistemological definition.

Some among logical positivists, notably Neurath and Hempel, deny that any set of propositions can be singled out as 'basic', or as in any important epistemological sense premisses for the remainder. Their view is that 'truth' is a syntactical, not a semantic concept: a proposition is 'true' within a given system if it is consistent with the rest of the system, but there may be other systems, inconsistent with the first, in which the proposition in question will be 'false'. There is no such process, according to them, as deriving the truth of a proposition from [133] some non-verbal occurrence: the world of words is a closed self-contained world, and the philosopher need not concern himself with anything outside it.

In logic and mathematics, the view that 'truth' is a syntactical concept is correct, since it is syntax that guarantees the truth of tautologies. Truth, in this sphere, is discoverable by studying the form of the proposition concerned; there is no need to go outside to something that the proposition 'means' or 'asserts'. The authors in question assimilate empirical to logical truth, thus reverting unconsciously to the tradition of Spinoza, Leibniz, and Hegel. In rejecting their view, as I shall contend that we must, we are committing ourselves to the opinion that 'truth' in empirical material has a meaning different from that which it bears in logic and mathematics.

The coherence theory of truth, as I have just said, is that of Hegel. It is worked out, from a Hegelian point of view, in Joachim's book The Nature of Truth, which I criticized, from the standpoint of the correspondence theory, in Philosophical Essays (1910). The Hegelian theory, however, differs from that of Neurath, since it holds that only one body of mutually coherent propositions is possible, so that every proposition remains definitely true or false. Neurath, on the contrary, takes the view of Pirandello: 'so it is, if you think so'.

The theory of Neurath and Hempel is set forth in articles in Erkenntnis and Analysis. The following are quotations or paraphrases of their words.

An assertion is called righ when we can fit it in (eingliedern).

Assertions are compared with assertions, not with 'experiences' (Erlebnissen).

There are no primary Protokollsätze or propositions needing no confirmation.

All Protokollsätze should be put into the following form: 'Otto's protocol at 3.17: {Otto's word-thought at 3.16 (In the room at 3.15 was a table perceived by Otto)}.'

Here the repeated use of the word 'Otto' instead of 'I' is essential.

Although, according to the above, it would seem as if we were debarred from knowing anything about the physical world [134] except that physicists make certain assertions about it, Neurath nevertheless commits himself to the statement that sentences are mounds of ink or systems of air-waves (Erkenntnis Iv, 209). He does not tell us how he discovered this fact; presumably he only means that physicists assert it.

Neurath in 'Radikaler Physikalismus und Wirkliche Welt' (Erkenntnis IV, 5, 1934), maintains the following theses:

1. All Realsätze of science including Protokollsätze are chosen as the result of Entschlüsse (decisions), and can be altered.
2. We call a Realsatz false when it cannot fit into the edifice of science.
3. The control of certain Realsätze is compatibility with certain Protokollsätze: instead of die Wirklichkeit we have a number of mutually incompatible but internally coherent bodies of propositions, choice between which is 'nicht logisch ausgezeichnet'.

The practice of life, Neurath says, quickly reduces ambiguity; moreover the opinions of neighbours influence us.

Carl G. Hempel 'On the logical positivist's theory of truth' (Analysis II, 4 January 1935) sets forth the history of the views of logical positivists as to Protokollsätze. He says the theory developed step by step from a correspondence theory into a restrained coherence theory. He says that Neurath denies that we can ever compare reality with propositions, and that Carnap agrees.

We started, he says, from Wittgenstein's atomic propositions; these were replaced by Protokollsätze, at first thought to express the results of observation. But then Protokollsätze were no longer the result of observation, and then no class of statements was admitted as basic.

Carnap (Hempel continues) says there are no absolutely first statements for science; even, for Protokollsätze further justification may be demanded. Nevertheless:

'Carnap and Neurath do by no means intend to say: "There are no facts, there are only propositions"; on the contrary, the occurrence of certain statements in the protocol of an observer or in a scientific book is regarded as an empirical fact, and the [135] propositions occurring as empirical objects. What the authors do intend to say, may be expressed more precisely thanks to Carnap's distinction between the material and the formal mode of speech. . . .

'The concept of truth may be characterized in this formal mode of speech, namely, in a crude formulation, as a sufficient agreement between the system of acknowledged Protokollsätze and the logical consequences which may be deduced from the statement and other statements which are already adopted. . . .

'Saying that empirical statements "express facts" and consequently that truth consists in a certain correspondence between statements and the "facts" expressed by them, is a typical form of the material mode of speech.' (p. 54) [i.e., 'truth' is syntactic, not semantic]

'In order to have a relatively high degree of certainty, one will go back to the Protokollsätze of reliable observers.' [Two questions arise: A. How do we know who are reliable? B. How do we know what they say?]

'The system of Protokollsätze we call true . . . may only be characterized by the historical fact, that it is the system which is actually adopted by mankind, and especially by the, scientists of our culturecircle.

'A Protokollsätze, like every other statement, is at the end adopted or rejected by a decision.'

Protokollsätze are now superfluous. It is implied that there is no definite world with definite properties.

I think Neurath and Hempel may be more or less right as regards their problem, which is the construction of an encyclopaedia. They want public impersonal propositions, incorporated in public science. But public knowledge is a construction, containing less than the sum of private knowledges.

The man who is constructing an encyclopaedia is not expected himself to conduct experiments; he is expected to compare the opinions of the best authorities, and arrive, so far as he can, at the standard scientific opinion of his time. Thus in dealing with a scientific question his data are opinions, not direct observations of the subject-matter. The individual men of science, however, whose opinions are the encyclopaedist's premisses, have not [136] themselves merely compared other investigators' opinions; they have made observations and conducted experiments, on the basis of which they have been prepared, if necessary, to reject previously unanimous opinions. The purpose of an observation or experiment is to give rise to a perceptive experience, as a result of which the percipient has new knowledge, at first purely personal and private. Others may repeat the experiment, and in the end the result becomes part of public knowledge; but this public knowledge is merely an abstract or epitome of private knowledges.

All theory of knowledge must start from 'what do I know?' not from 'what does mankind know?' For how can I tell what mankind knows? Only by (a) personal observation of what it says in the books it has written, and (b) weighing the evidence in favour of the view that what is said in the books is true. If I am Copernicus, I shall decide against the books; if I am a student of cuneiform, I may decide that Darius did not say what he is supposed to have said about his campaigns.

There is a tendency -- not confined to Neurath and Hempel, but prevalent in much modern philosophy -- to forget the arguments of Descartes and Berkeley. It may be that these arguments can be refuted, though, as regards our present question, I do not believe that they can be. But in any case they are too weighty to be merely ignored. In the present connexion, the point is that my knowledge as to matters of fact must be based upon my perceptive experiences, through which alone I can ascertain what is received as public knowledge.

This applies, in particular, to what is to be found in books. That Carnap's books say whatever they do say is the sort of thing that would be generally accepted as public knowledge.

But what do I know?

1. What I see when I look at them
2. What I hear when others read them aloud
3. What I see when others quote them in print
4. What I see when I compare two copies of the same book.

Hence, I pass, by elaborate and doubtful inferences, to public knowledge. [137]

On Neurath's view, language has no relation to non-linguistic occurrences, but this makes many every-day experiences inexplicable. For instance: I arrived in Messina from a sea voyage in 1901 and found flags at half-mast; on inquiry I learnt that McKinley had been murdered. If language has no relation to the non-linguistic, this whole procedure was frivolous.

As we saw, Neurath says the proper form of a protocol sentence is: 'Otto's protocol at 3.17: {Otto's word-thought at 3.16 was: (In the room at 3.15 there was a table perceived by Otto)}.'

It seems to me that, in giving this form to protocol sentences, Neurath shows himself far more credulous than the man who says 'there's a dog'. In the inside bracket he perceived a table, which is just as bad as perceiving a dog. In the outside bracket he finds words for what he has perceived, viz.: 'in the room at 3.15 there was a table perceived by Otto'. And a minute later he writes down the words at which he has arrived. This last stage involves memory and the continuity of the ego. The second stage involves memory also, and in addition involves introspection.

Let us take the matter in detail.

To begin with the inner bracket: 'in the room at 3.15 there was a table perceived by Otto'. We may take the words 'in the room' as merely meaning that the table had a perceptual background, and in that sense they may be allowed to pass. The words 'at 3.15' imply that Otto was looking at his watch as well as at the table, and that his watch was right. These are grave matters, if taken seriously. Let us suppose that, instead of 'at 3.15' we say 'once upon a time', and instead of '3.16' we say 'a little later', and instead of '3.17' we say 'a little later still'. This eliminates the difficulties of time-measurement, which surely Neurath cannot have intended to introduce. We come now to the words 'there was a table'. These are objectionable on the same grounds as 'there's a dog'. It may not have been a table, but a reflection in a mirror. Or perhaps it was like Macbeth's dagger, a phantasm called up by the intention of committing a murder on a table. Or perhaps a very unusual collocation of quantum phenomena caused a momentary [138] appearance of a table, which was going to disappear in another moment. It may be conceded that this last hypothesis is improbable, that Dr Neurath is not the sort of person who would think of murdering anybody, and that his room probably contains no mirror large enough for the reflection of a table that is elsewhere. But such considerations ought not to be necessary where protocol-sentences are concerned.

I come now to a still more serious matter. We are told, not only that there was a table, but that there was a table 'perceived by Otto'. This last is a social statement, derived from experience of social life, and by no means primitive; in so far as there is reason to believe it, it is based upon argument. Otto perceives the table, or rather a tabular appearance -- well and good -- but he does not perceive that Otto perceives it. What is 'Otto'? So far as he can be known, either to himself or others, he is a series of occurrences. One of them is the visual appearance which he rashly calls a table. By the help of conversation, he is led to the conclusion that the occurrences people mention form bundles, each of which is one person, and that the appearance of the table belongs to the same bundle as the subsequent word-thought and the still more subsequent act of writing. But all this elaboration is no part of the visual datum. If he always lived alone, he would never be led to distinguish between 'there's a table' and 'I see a table'; in fact, he would always use the former phrase, if one could suppose him using phrases at all. The word 'I' is a word of limitation, meaning 'I, not you'; it is by no means part of any primitivejdatum. And this is still more evident when, instead of 'I', Neurath says 'Otto'.

So far we have only been concerned with what happened at 3.15. It is now time to consider what happened at 3.16.

At 3.16, Otto put into words what had happened at 3.15. Now I am willing to admit that the words he used are such as well might be employed by a man who was not on the lookout for pitfalls. There is, therefore, less to criticize at this stage. What he thought may well not have been true, but I am quite willing to concede that he thought it, if he says so.

At 3.17, Otto carried out an act of introspection, and decided that, a minute ago, a certain phrase had been in his thoughts, [139] not just as a phrase, but as an assertion concerning an earlier perception which, at 3.16, he still remembered. It is only what happens at 3.17 that is actually asserted. Thus according to Neurath the data of empirical science are all of the following form:

'A certain person (who happens to be myself, but this, we are told, is irrelevant) is aware at a certain time that a little while 'ago he believed a phrase which asserted that a little while before that he had seen a table.'

That is to say, all empirical knowledge is based upon recollections of words used on former occasions. Why recollections should be preferred to perceptions, and why no recollections should be admitted except of thought-words, is not explained. Neurath is making an attempt to secure publicity in data, but by mistake has arrived at one of the most subjective forms of knowledge, namely recollection of past thoughts. This result is not encouraging to those who believe that data can be public.

The particular form given to protocol-sentences by Neurath is, perhaps, not an essential part of his doctrine. Let us therefore examine it more generally.

Let us repeat some quotations. [In what follows, 'N' stands for Neurath' and 'H' for 'Hempel'.]

• 'Statements are compared with statements, not with experiences' (N.)
• 'A protocol-statement, like every other statement, is at the end adopted or rejected by a decision' (N).
• 'The system of Protokollsätze we call true . . . may only be characterized by the historical fact, that it is the system which is actually adopted by mankind, and especially by the scientists of our culture circle' (H).
• 'Instead of reality we have a number of mutually incompatible but internally coherent bodies of propositions, choice between which is not logically determined (logisch ausgeseichnet)' (N).

This attempt to make the linguistic world self-sufficient is open to many objections. Take first the necessity of empirical statements about words, e.g., 'Neurath says so-and-so'. How do I know this ? By seeing certain black marks on a white ground. But this experience must not, according to Neurath and Hempel, be made a ground for my assertion that Neurath says so-and-so. Before I can assert this, I must ascertain the opinion of mankind, [140] and especially of my culture circle, as to what Neurath says. But how am I to ascertain it? I go round to all the scientists of my culture circle, and say: 'what does Neurath say on p. 364?' In reply I hear certain sounds, but this is an experience, and therefore does not give any ground for an opinion as to what they said. When A answers, I must go round to B, C, D, and the rest of my culture circle, to ascertain what they think A said. And so on throughout an endless regress. If eyes and ears do not enable me to know what Neurath said, no assemblage of scientists, however distinguished, can enable me to know. If Neurath is right, his opinions are not known to me through his writings, but through my decisions and those of my culture circle. If we choose to attribute to him opinions completely different from those which he in fact holds, it will be useless for him to contradict, or to point to pages in his writings; for by such behaviour he will only cause us to have experiences, which are never a ground for statements.

Hempel, it is true, denies such consequences of his doctrine. He says: 'Carnap and Neurath do by no means intend to say: "there are no facts, there are only propositions"; on the contrary, the occurrence of certain statements in the protocol of an observer or in a scientific book is regarded as an empirical fact, and the propositions occurring as empirical objects.' But this makes nonsense of the whole theory. For what is an 'empirical fact'? To say: 'A is an empirical fact' is, according to Neurath. and Hempel, to say: 'the proposition "A occurs" is consistent with a certain body of already accepted propositions'. In a different culture circle another body of propositions may be accepted; owing to this fact, Neurath is an exile. He remarks himself that practical life soon reduces the ambiguity, and that we are influenced by the opinions of neighbours. In other words, empirical truth can be determined by the police. This doctrine, it is evident, is a complete abandonment of empiricism, of which the very essence is that only experiences can determine the truth or falsehood of non-tautologous propositions.

Neurath's doctrine, if taken seriously, deprives empirical propositions of all meaning. When I say 'the sun is shining', I do not mean that this is one of a number of sentences among [141] which there is no contradiction; I mean something which is not verbal, and for the sake of which such words as 'sun' and 'shining' were invented. The purpose of words, though philosophers seem to forget this simple fact, is to deal with matters other than words. If I go into a restaurant and order my dinner, I do not want my words to fit into a system with other words, but to bring about the presence of food. I could have managed without words, by taking what I want, but this would have been less convenient. The verbalist theories of some modern philosophers forget the homely practical purposes of every-day words, and lose themselves in a neo-neo-Platonic mysticism. I seem to hear them saying 'in the beginning was-the Word', not 'in the beginning was what the word means'. It is remarkable that this reversion to ancient metaphysics should have occurred in the attempt to be ultra-empirical.