SENSATIONALISM

Sensationalism is the traditionally important doctrine according to which all our knowledge of the world comes to us through the senses. The chief aim of these pages is to systematize the traditional arguments against sensationalism, to show their incompleteness, and to supplement them with some modern arguments.

There are two traditional divisions of sensationalism, yielding four possible sensationalist schools of thought. The first division is that be­tween the sensationalists who think that informative theoretical knowledge is possible - the inductivists - and the sensationalists who think that in­formative theoretical knowledge is not possible - the conventionalists. The second division is that between naive and sophisticated sensationalists: the naive sensationalists assert that all (well attested) reports of observa­tion are entirely reliable; the sophisticated sensationalists assert that only some sort of observation-reports are entirely reliable.

Sensationalism

inductivism conventionalism

Both of these divisions of sensationalism, it should be noticed, are exclusive and exhaustive. A complete criticism of sensationalism must be the criticism of all the four versions, or classes of versions, mentioned above.

For the sake of simplicity the discussion of the present essay will be confined to our knowledge of the external world. The problem of our knowledge of ourselves will be avoided in order to avoid reference to any psychological theory except perception theory, and in order to avoid exegesies of the classical sensationalist texts. The very famous dictum ‘nothing is in the mind which has not been previously in the senses’ may be understood - prior to exegesies - to refer solely to our knowledge of the external world. It is therefore understood here in this restricted sense, and criticized together with the four philosophical schools of thought which endorse it.

1. Sensationalism vs. Theoretical Knowledge

According to sensationalism all knowledge of the world comes through the senses. This obviously entails that knowledge consists exclusively of observation-reports and statements derivable from them. It is therefore inconsistent with the view that there exists theoretical knowledge about the world, since (by the very meaning of the word “theoretical”) theoreti­cal knowledge of the world is that knowledge which is not a derivable from observation-reports alone.

Sensationalists are well aware of this criticism; they view the problem of how to answer it as an integral part, if not the core, of the problem of induction. They vacillate between two alternative answers to this criticism, inductivism and conventionalism.

The inductivist answer to the above criticism is that theoretical knowl­edge is “indirectly” derived from the senses, being based on observation­reports by induction. This answer dodges the issue: whatever “indirectly” means, induction is either a purely deductive process or not; if it were de­ductive, then theoretical knowledge would be derivable from observational reports; which is not the case.

The inductivist may try another answer: he may claim that although all factual knowledge is derived from the senses, a particular piece of knowledge may at first have to be a conjecture, and only then it has to be verified, or become a result of observation. This answer may be called verificationism, since according to it conjectures of a given kind are capable of verification by observation.

Verificationism is vague and dangerous. It is so vague that one cannot know whether it is a version of sensationalism, and it is dangerous in its raising false hopes. Let us consider its vagueness first. Verificationists often do not tell us whether or not the verified conjecture follows from the reports which have verified it, and whether or not the verifying could be secured without any prior conjecture. Does verificationism include the claim that the conjecture does follow from the reports which verify it, and the claim that in principle these reports could be procured without any previous conjecture? In case it does, verificationism is decidedly sensa­tionalist; in case it denies at least one of these two claims it is decidedly not sensationalist; in case it does not answer the question it is not de­cidable whether verificationism is or is not a version of sensationalism. Now, in case verificationism is made a version of sensationalism by adding to it a positive answer to the above question, then, according to it, verification leads to not theoretical knowledge. Sensationalist verifica­tionism only states that prior to making an observation (such as that the sun will rise tomorrow, to take Wittgenstein’s example) we may, but need not, guess correctly that observation. This is not what is usually meant by “verificationism”; so, verificationism as commonly understood is not at all sensationalist, though, unfortunately, the vagueness of it gives such an impression.

Take first the simplest case of one conjecture and a guarantee that we shall be able to verify that conjecture. As we have a guarantee that this conjecture is true, we need not verify it. Hence, this case is inconsistent with sensationalism. Even when a theory entails that possibly there is a case in which we need not verify a conjecture, that theory is inconsistant with sensationalism which asserts that we cannot ever know the truth of a conjecture prior to its verification. So let us now take different cases of verificationism and show that in each case there is a possibility of its reduction, under some possible circumstances, to the first case above. Take, again, one conjecture, and a guarantee that it be verified, but now replace the full guarantee by a partial one. Now, a partial guarantee may be of a good chance of success, which does not exclude utter failure; such a guarantee is of no avail since in case of utter failure one may claim that the guarantee is valid nonetheless. In order to exclude such a pre­posterous case, one usually claims, a partial guarantee which only assures a chance of success also assures, by the mathematical laws of chance, that on a sufficiently repeated application of a partial guarantee the partial guarantee of success becomes a perfect guarantee for success in some cases (Bernoulli’s law).

The next retreat of the inductivist would be an attempt to replace the notion of the verification of a conjecture by that of its confirmation, namely, of the verification of some of the consequences of a conjecture. However, the problem reappears. We can have no guarantee for any verification whatever, not even for a verification of a weak consequence of a conjecture. If we had, then, again, we could construct a possible case in which the guarantee will render observation unnecessary - and thus the existence of a guarantee will be inconsistent with sensationalism. Moreover, the margin between the verification of a conjecture and its mere confirmation, namely the unverified consequences of an accepted hypothesis, would be the nonsensational element of the existing body of theory.

Thus, sensationalism is incompatible with the view that informative theoretical knowledge exists, no matter how it was acquired - by verifica­tion, by confirmation, or in any other way. As Hume has already shown, inductivism (be it correct or not) fails to reconcile sensationalism with the view that theoretical knowledge exists. (Being both an inductivist and a sensationalist, Hume ended up denying the existence of theoretical knowledge.)

Historically, another limitation of sensationalism was noticed prior to Hume’s discovery of its limitation with respect to theoretical knowledge; it is the limitation with respect to theoretical concepts. Since the Middle Ages sensationalists have realized that sensationalism implies that we cannot have theoretical concepts, that all our concepts are either those derived from observations, or their combinations; that even in our wildest imagination we cannot fancy anything but new combinations of old observational material, so that all concepts are, like the concept ‘sphinx’ (to take Francis Bacon’s example), merely combinations of observational concepts. Indeed, this is Hume’s starting point. Einstein’s and Russell’s favourite argument against sensationalism is the essentially Kantian idea that mathematical concepts go very far beyond any past experience, and that some of these concepts are employed very fruitfully in science.

Yet this very criticism gives the clue to the alternative sensationalist view, namely conventionalism. Conventionalism gives great scope to the imagination, and views both mathematics and theoretical science as admirable structures produced by the imagination. But, admitting that theories go beyond experience and remaining true to sensationalism, conventionalism must empty theories of all factual or empirical content. Assuming that theories come from the imagination, and that information comes from experience, conventionalism justly concludes that theories are devoid of any information. If theory is not informative knowledge, what is it? It is, says conventionalism, our way of looking at particular facts, our way of classifying particular observed facts. Like formal mathematics, theoretical science is merely an empty structure to store information in, a way of saying things, a language. Nothing in reality strictly corresponds to abstract or imagined theoretical concepts like ‘space curvature’ or ‘atom’. These words are no more than short­hand symbols with no independent meaning (their meanings are given by implicit definitions), and statements containing them impart no more information than the information procured by sensations alone.

Although conventionalism is much clearer and more coherent a view than inductivism, it was traditionally viewed by men of science as a defeatist position, because the aim of science, it was felt, was not just to replace an unordered or an arbitrarily ordered heap of information by an elegantly ordered yet not richer stock of information. The intuitively accepted view behind the scientific tradition between 1600 and 1900 was that there exists a hidden reality (i.e. hidden from the senses) and that the aim of science is the search for it; to wit, the aim of science is to attempt to discover the laws of nature and not to invent laws of elegant and concise language. Although hardly any of the classical natural philos­ophers adhered to conventionalism consistently and persistantly, many of them used it as a second best alternative to inductivism; that is to say, when they could not present their theories as inductively based on ob­servation-reports, they chose the alternative sensationalist interpretation of their theories, and viewed them in a conventionalist manner as mere systems of classifications of knonw facts. It seems clear that traditional thinkers confined themselves to inductivism as a rule and to convention­alism as a temporary refuge, because they wished to retain their sensation­alism; they wished to retain sensationalism as they considered it the only basis for empiricism; and they wish to retain empiricism as they con­sidered it the only ground for the validity of empirical science. But in this they are mistaken.

2. Sensationalism vs. Empiricism

If we assume that informative theoretical knowledge is possible, then we may inquire into the grounds for its validity. The two traditional answers to the problem, what is the ground for the validity of theoretical knowl­edge, are apriorism and empiricism. Apriorism is the claim that the ground for the validity of informative theoretical knowledge is intellectual, and empiricism that it is sensational. Thus, empiricism is traditionally sensationalist. But sensationalism may be not empiricist at all. Obviously, conventionalism is a version of sensationalism and yet it is not a version of empiricism; indeed, it is neither apriorist nor empiricist: by including the claim that informative theoretical knowledge does not exist, con­ventionalism avoids giving rise to the problem what is the ground for such knowledge - to which problem both empiricism and apriorism are the traditional alternative solutions. Yet, as conventionalism is a sensa­tionalist view, traditional empiricists prefer conventionalism to apriorism, and were even ready to use it as a temporary refuge when their empiricism was beaten, hoping that with the increase of the amount of factual in­formation they would be able to return to their empiricism in order to find informative theoretical knowledge about the world.

Traditional natural philosophers have always emphasized the signi­ficance of empirical theoretical knowledge. The most often quoted pas­sage of Bacon’s was his parable of the ant, the spider, and the bee: the empiric or skeptic who has only reports of observed facts is like the ant which only collects; the reasoner or apriorist who has only theories is like the spider which only spins out its own material; the interpreter of nature, the true empirical theoretical philosopher, is like the bee which both collects and adds something of its own to the collected material. (This parable, says Russell, is unfair to the ant who also orders. If so, it is also unfair to the spider who also collects.) In another famous passage Bacon speaks of science as the wedding of the intellect and the senses.

These metaphors conceal a problem. Admittedly the contribution of the senses is empirical. But what is the contribution of the intellect? Is it not the case that the contribution of the intellect is non-empirical? Is not the idea of empirical theoretical science self-defeating?

Bacon must already have been aware of this problem, for he gave an answer to it (in his Preface to The Great Instauratiori). His answer is this. Just as by sensing the rays of light our eyes see things, so, by analogy, by sensing things our intellect sees the laws of nature. This answer is a traditional mystical or intuitionist view which assumes the existence of a mental eye that sees or intuits laws with complete assuredness just as the eye of the flesh sees things with complete assuredness. (The traditional mystic formula is that of the unity of the knower and the known with knowledge; it occurs in a crucial passage of Bacon.)

No wonder that no later empiricist shared this view with Bacon. The problem remains, then: how is empirical theoretical knowledge pos­sible?

The obvious substitute for Bacon’s answer is the view that the senses provide the material and the intellect the order. But this answer is the denial of empiricism. It is either conventionalist, if the order which our mind provides is claimed to be merely ours, or apriorist, if that order is claimed to coincide with the order of the world. (Kant seems to have vacillated between these two claims and clearly preferred to leave the choice between them undecided. He has stressed that the order is provided by us, but intentionally, systematically, and persistently left open the question whether this order of ours coincides with the order of the world or not. This point was overlooked by reviewers of his Critique of Pure Reason, with the result that he was incensed and wrote his Prolegomena - a much easier book to read - to set this point right, as he explains in the preface to that book. Unfortunately, some of his best modern com­mentators, such as Russell, still do him injustice on this point by ascribing to him answers to questions which he insisted on leaving unanswered, such as, is space real?)

Another empirist answer is this: when the senses make their own con­tribution they stimulate the intellect to make its own contribution. But this is not to the point: the apriorists themselves, since Bruno and Descartes, have always asserted that the senses may stimulate the in­tellect; they only denied that the senses are the source of knowledge; they declared that the intellect makes a contribution, and that we can see the independence of the validity, or the self-evidence, of this contribution. Clearly, they would argue, if any part of the contribution of the intellect is logically independent of the senses to any extent apriorism is not excluded, while if the contribution of the intellect logically depends on perceptions it cannot add to the information which can be provided by the senses. The only other alternative is to claim that the contribution of the intellect is not informative, as the conventionalists do. (Incidentally, Wittgenstein claims (Tractatus, 6.342) that theoretical knowledge is not informative, but he adds that the existence of theoretical knowledge and its simplicity are informative. This, obviously, suffices to overthrow his sensationalism and to establish him as an apriorist.)

The above discussion may explain the fact that quite a few great thinkers despaired of their empiricism and became apriorists. It is rather cheap, perhaps, to ridicule eminent apriorists (Descartes is the traditional scapegoat) whenever the validity of your own brand of empiricism is challenged. The function of the repeated sneer at apriorism is to drive home the idea that any deviation from narrow sensationalism leads towards apriorism. This idea does not solve our problem, however, but rather sharpens it. For we can state the dilemma in this way: if we do not go beyond sense experience we have no theoretical knowledge of the world, while if we do go beyond it the margin is not contained in sense experience, and is, thus, a priori.

This is the logic which led thinkers to abandon empiricism in favour of either apriorism or conventionalism. For according to both these views our present theoretical knowledge necessarily transcends our experience; they differ only as to the question of whether this knowledge is informative (apriorism) or not (conventionalism).

Yet empiricist philosophers who have studied the problem of knowledge have usually stuck to their sensationalism in spite of this refutation. Perhaps they hoped that somewhere a logical error had been committed in the refutation of sensationalist empiricism. They were unable to refute any step of the criticism, but they had a strong argument in favour of the view that a logical error could be found in the criticism. The argument is this. In our ordinary behaviour we show that we consider theoretical science as informative, for we normally rely on theoretical science as informative, for we normally rely on theoretical information. Moreover, we show that this information is indeed connected with experience, for if theoretical information clashes with the information gained by experience we prefer the latter; we accept theoretical informa­tion only when it is strongly supported by experience. Furthermore, we gain theoretical knowledge or at least theoretical hints from certain important experiments, like that of Michelson and Morley. In brief, we know that the error is there, since we know that we gain theoretical knowledge from experience.

There is no need to accept the suggestion that there is an undetected error in the criticism, and it is even dangerous to accept such a suggestion since in this act one might very well be giving up one’s rationality and hope to learn from criticism. We may even endorse the above criticism as valid while accepting the contention that we gain theoretical knowledge from experience - although it is in no way imperative to accept it. In a way this contention - we gain theoretical knowledge from experience - is indeed the core of empiricism, as it amounts to the rejection of both con­ventionalism and apriorism. But, strictly speaking, this contention is not empiricist in the traditional sense, at least in that traditional empiricism is a theory of the ground of the validity of existing knowledge, whereas this contention is a theory of learning or of scientific method. And, most important, perhaps, is that traditional empiricism is sensationalist (and thus has to be rejected), whereas this contention - that we learn about reality from experience - is not. Even if we say that all we learn about the world is derived from experience, we need not commit ourselves to sensationalism except when we identify experience with information derived from the senses. Hence, either the identification of all experience with sense experience is an error, or else any kind of empiricism is hopelessly inconsistent: only if we get rid of the identification of ex­perience with sense experience may we retain some modern version of empiricism. Let us now consider the different ways in which experience was identified with sense experience; it will turn out that this identification leads to the surprising conclusion that we cannot describe our experiences, that we never know what experience is.

3. Sense-Experience vs. Experience

The identification of experience with sense experience has been done in a naive way and in a more sophisticated way. The naive identification of experience with sense experience is a version of naive realism. It is simply the claim that (when we look carefully) we see things as they are. It was admirably criticized by many modern philosophers from Galileo and Kant to Einstein and Russell. Galileo thought the naive thesis refuted by the fact that when we walk at night we may see the moon jumping like a cat from one roof-top to another. A somewhat less picturesque but more elegant argument against it is, perhaps, Schrodinger’s argument (in Nature and the Greeks). We see the sun as being not much bigger than a cathedral. Assuming that the sun is as big as we see it, and accepting very simple, and intuitively quite obvious, trigonometrical theorems, we can calculate the distance between the eastern and western positions of the sun and find it to be no more than one day’s walking distance.

This argument does not convince the adherents of naive sensationalism. Naive sensationalism carries great force with it. Even if we do not see all things precisely as they are, we all admit that we can see this table toler­ably well (arguments from perspective notwithstanding). We admit that endless speculations and disputations will not be useful to determine some question of fact which can easily be determined by plain observation and experient - for instance, the question, what is the color of the table in the next room? Surely that much we all admit, and it is, quite obviously, the core of what the naive sensationalist wishes to assert. This explains the fact that though no serious thinker defends systematically naive sensa­tionalism, many lapse occasionally into this position.

Many historians of physics and a few physicists claim that the medieval scholars were apriorists who would consult Aristotle’s and their own reason, rather than observation, in order to determine the number of teeth a mole has. It is not known whether any medieval scholar deserves ridicule as one who never relied on his eyes but always preferred reasoning to plain observation. If such a person existed, however, he must have been most unusual. Undoubtedly, most people agree that it is often preferable to rely on one’s eyes than on one’s reasoning; and undoubtedly naive sensationalism explains or justifies this preference. (Even apriorists like Descartes and Spinoza endorsed this preference, of course; they offered most ingenious explanations or justifications of this preference, but we need not go into that here, where we take it for granted that apriorism is false.) As naive sensationalism is plainly false, the (correct) frequent preference of observation over reasoning needs a better ex­planation or justification. We do not rely on the eye of the mind because it may mislead us, and to be fair we should not rely on the eye of the flesh as it may mislead us too. This, it seems, suffices to show that the preference of perception over reasoning is not a matter of reliance. Sensationalists, however, stick to their sensationalism even after they are forced to admit that the senses can, and sometimes do, mislead us. These sensationalists claim that though our senses may mislead us, under certain conditions they do not. This is still unfair to the intellect. If we are willing to rely on the senses, even though they mislead us, after finding the conditions under which they do not, then, in all fairness, we should be also ready to rely on the intellect, even though it misleads us, after finding the conditions under which it does not. Yet according to sensationalism we should sometimes rely on our senses but never on our intellects. When the sensationalist becomes aware of the fact that ex­perience can mislead us, instead of ceasing to rely on experience and reopening the whole problem afresh, he claims that there must exist some kind of experience - pure experience, as he calls it - which cannot mislead us. This is the doctrine of sophisticated sensationalism.

There exist strong versions of sophisticated sensationalism. They specify which, or what kind of, experience cannot mislead us. The weak version is the mere assertion that pure experience exists. This may be the case. Let us examine, first, the strong versions of sophisticated sensationalism, Bacon’s and Locke’s. Bacon’s position is that the theoretical element of experience is what mislead us, not the sensational element of experience. Once we get rid of all our prejudices, of all of our preconceived ideas, we can experience things as they are (in the manner assumed by the naive sensationalist).

Very interestingly, the mere substitution of ‘class-prejudice’ for ‘pre­judice’ in the statement above yields Marx’s view, and of ‘neuroses’ in­stead yields Freud’s view. Now, Locke’s view is that the reliable ex­periences are the elements of individual sensations, which were later called sense-date; they are pure sensations like the sensations of sounds or the sight of coloured patches. Bacon’s doctrine can be shown to be in­consistent, for it is itself a preconceived idea. Locke’s view has been experimentally refuted. The identification of patches was shown to be dependent on our knowledge of geometry and perspective, the identifica­tion of a colour shown to be dependent on our language, and even the ability to distinguish between two similar sounds depends on theoretical instruction. Thus, although sophisticated sensationalism as such - the view that there exist pure sensations - is irrefutable, its two more sub­stantial versions - Bacon’s and Locke’s - are logically or empirically refutable, and they were refuted. As to the irrefutable version of sophistic­ated sensationalism - the bare claim that pure sensations exist - it does not explain the existing preference, in certain cases, of observation over theory: the preference can only be explained as the preference of the reliable if the observation in question can be identified as a pure sensation.

No one doubts that sensations are a necessary part of any experience. The fact of experience is, however, that thus far we cannot specify a type of experience which must be fully reliable; that in particular we are not aware of pure sensations; that there exists no known or reported immedi­ate or direct sense experience, just as there exists no known or reported immediate or direct experience of the electric signals which, according to modern neuro-physiology, sensations consist of. It is a fact of experience that when describing or reporting scientific experiments and observations we very rarely describe or report our sensations. Observation-reports describe facts and their having been observed - not at all the sensations of the observer while he was observing, except in rare instances such as the sensation of electric currents on the observer’s tongue.

This fact a sophisticated inductivist will readily admit, and yet he will claim in describing our scientific experiments we do report our sensations, even though indirectly. Here again our inductivist runs against our dilemma, and again he refuses to consider it, being sure that we learn from experience and that experience must be, ultimately, sense experience. Rather than resolving the dilemma he tries to purify a given report of a scientific experiment of its theoretical element and reduce the scientific report to a report about past sensations. Yet when trying to do this he soon uses theories and statements of objective facts rather than reports about sensations. He will justify his use of statements of objective facts by claiming that they were once constructed out of pure-sense- elements - thus assuming what he has set out to prove. He will also justify his speaking of facts by defending naive sensationalism. He will then retreat from naive sensationalism to a sophisticated one, and so on. It is a historical fact that very few thinkers ever tried to show how a given piece of scientific information can be decomposed into, and recomposed from, sense-perceptions; such attempts, notably Laplace’s, Mach’s and Russell’s, were complete failures because of their authors’ vacillation between naive and sophisticated views, as well as between inductivist and conventionalist views.

This argument leads inductivists to two characteristic reactions. The one is to try again. The other is to dismiss the whole debate as too sophisticated. In order to show that it is not unnecessarily over-sophistic­ated I shall take an example of the inductivist’s muddled approach to experimental errors.

It is a well-known fact that John Dalton reported having observed the atomic weight of oxygen to be, on the average, near to but slightly above 6.5 and decided that it is actually 7. Obviously, he could not get the result which we have today, namely 16, because he thought that water contains oxygen and hydrogen in equal proportions and not, as we think today, in the ratio of one to two; but better experiments, it is alleged, might have led him to the result 8 rather than to 7. It is therefore unani­mously accepted by modern historians of science that Dalton was a bad observer, Dr. Thomas Thomson’s personal testimony to the contrary notwithstanding. It is difficult to imagine that a bad observer was the inventor and improver of experimental techniques in weighing gases. If the historians who condemn Dalton were serious about the whole matter they would have tried to repeat Dalton’s experiments as his con­temporaries did. In this case they would undoubtedly get the same result as Dalton’s, just as Dalton’s contemporaries did before Davy discovered a better method which yielded the result 7.5.

It is obvious to me that Dalton’s result is respectable and yet untrue. He who doubts it will have to apply the same doubt to the results of all nineteenth-century chemical experiments. The best and most precise experiments concerning the atomic weight of chlorine then gave 35.5 as a result, and they were equally mistaken; the atomic weight of chlorine, is much nearer to either 35 or 37 than to 35.5. The naive and sophisticated inductivists alike must fail to explain all this. In order to explain why our predecessors accepted and we reject 35.5 as the atomic weight of chlorine different theories have to be referred to. It transpires, then, that contrary to all we were taught in chemistry classes and in history of science classes and in philosophy degree courses, it was the factual report which has been declared to be false in the light of modern theories. But before trying to defend this startling conclusion I wish to discuss the sophisticated con­ventionalists’ explanation of this situation. For it is the great advantage of sophisticated conventionalist that he handles this situation with great ease.

4. Sensationalism vs. Common Sense

The criticisms of naive sensationalism and of inductivism which I have presented so far do not cause the slightest difficulty to the sophisticated conventionalist. He does not claim that theoretical knowledge is derived from experience but neither does he claim that theoretical knowledge is informative. He therefore can easily reconcile the existence of unin­formative theoretical knowledge with sensationalism. He does not claim any observation report is purely sensational, so that he can stick to his reliance on the senses in spite of all the alterations which the observation reports undergo. The major defect of this position seems to lie in the fact that it sounds just too defeatist a position, defeatist both regarding theory and regarding observation. But this is an error: defeatism re­garding theory is quite sufficient. Sophisticated conventionalists can argue that even though we cannot separate the sensational element in any observation statement, this element is invariant regarding any translation of a report from one language to another. The nineteenth-century ob­servation report ‘the atomic weight of chlorine is 35.5’ is not discarded by modern chemists but is translated by them into the twentieth-century language; the translation reads: ‘the average atomic weight of terrestrial chlorine is 35.5’. The translation of the report preserves its sensational element. The sensational element has not been rejected, only the theoreti­cal element has been replaced. The twentieth-century report ‘the atomic weight of chlorine is 35 or 37’ does not contradict the nineteenth-century report ‘the atomic weight of chlorine is 35.5’: they are cast in different languages, and forgetting this fact we rashly conclude that they con­tradict each other. Before we can find out whether they are in contradic­tion or not we must state them both in one and the same language. Now we cannot easily translate the twentieth-century report into the nineteenth­century language, because the later language is better - more elegant - than the older language. So it is more convenient to translate the nine­teenth-century report into the twentieth-century language. As we have seen, the translation shows the two reports to be perfectly compatible with each other.

It is essential for this mode of thought that it is both conventionalist - in viewing theoretical science as a mere system of languages - and sophisti­cated. Had we been able to state one observational report with no theoretical overtones, then the problem which the sophisticated con­ventionalist has solved would have arisen in a very different manner and his solution to it would be obviously unacceptable. And if we wish to con­clude that it is impossible to have purely observational reports, we must assume that even though we can translate a report into many languages without losing or altering its sensational content, we shall never be able to isolate this sensational element entirely. No doubt, had we constructed all the possible languages, and had we then stated one report in all these languages, the ‘conjunction’ of all these many statements of this one ob­servation report should give us a fair idea of the observation as such. But this is merely a thought experiment: there can be infinitely many langu­ages, or theoretical systems, for any finite set of observation reports to be expressible in. Also, we have seen from the example above, using a better language is easier and more elegant than using an inferior language - even when an observation-report fits both. This may suggest that the simplest, most ideal, language, if it will ever be achieved, will enable us to state pure observation-reports. Duhem arrived at the same idea from another angle: he thought that the ideal theoretical system will also be a true informative picture of the world - from which the above follows.

This discussion seems to me to clarify a number of points. First, it explains why sophisticated conventionalism never was popular: it is somewhat too sophisticated. Second, it explains the modern search after pure observation reports. Any sensationalist alternative to Duhem’s and Meyerson’s doctrine must contain the claim that we can isolate sense im­pressions from the theoretical element with which it amalgamates when presented in a scientific report. Yet in order to be convincing one must indicate how this can be done. Now Popper has argued (1935) that, since universal names are dispositional, reports containing them contain predictions, and are thus no pure reports. E.g. the report ‘here is a glass of water’ contains predictions since the glass is breakable or else we would not call it ‘glass’, and water is decomposable, etc. Hence the im­mense literature concerning dispositions and dispositional terms which has followed Carnaps’ study (1936) of the relations between dispositions and pure observations.

Yet one should notice, perhaps, that the sophisticated view according to which we cannot separate the sense information from the theoretical element in an observation report, though unpopular amongst philos­ophers, has gained popularity amongst some schools of contemporary psychology. This is so, partly, I suppose, owing to the fact that psychol­ogists cannot evade problems concerning observation reports as easily as other scientists: such troubles are their business. Partly it is due to the influence of Kulpe’s critical realism. The full discussion of this point is beyond the scope of the present essay; yet this much can, and ought to, be said here. The sophisticated view according to which we cannot separate the sensational and the theoretical element in an observation report is not in itself intolerable; it is intolerable only when we adopt it together with an inductivist or with a conventionalist attitude. For, when we adopt it together with a critical realistic attitude, we merely admit that any observation report must contain some hypothetical element; only when we adopt it together with a conventionalist attitude it turns out that we do not quite know what we are saying. For, according to sophisticated conventionalism, only the sense element is informative, and the sense element is unisolable; hence, according to sophisticated conventionalism, the information contained in a report is unisolable. The analogous realist attitude only entails that the certain and entirely warrantable element of a report is unisolable; namely, that observation reports are never certain.

To make this clearer, consider an observation report stated in court. To say that the judge does not quite know what is the information he receives from a witness is very disquieting. Moreover, it is not at all difficult to imagine, or to draw out of history, a case in which a piece of evidence would condemn the accused when cast within one theoretical system, and acquit him when cast in another. To critical realists this causes no trouble, since they permit error in any observation report. But not so to the sophisticated conventionalist.

Duhem was not unaware of this difficulty, for he tried to solve it. He suggested that naive sensationalism should apply to commonsense situations (such as the one described above) and sophisticated sensation­alism to science. His argument in favour of this suggestion is too involved to reproduce here. Nor need it be reproduced. For this division between science and common sense cannot be maintained, especially in the light of modern perception studies, of the Kiilpe school and its derivatives. Today’s common sense, as Maxwell has already claimed, is yesterday’s frontier of science.

The sophisticated conventionalist may attempt to answer this criticsim, but the more he will do so the more he will defeat his own purpose, for he is presenting a more and more elaborate theory about science and its role in society - a theory which he must consider as informative, and which entails that it is itself uninformative. As long as we only look at scientific theories we may suggest viewing them as empty - as Duhem does. But we cannot merely suggest to a judge that he view scientific theories as empty; we have to explain to him why he should do so by providing a theory of sorts; and he will rightly apply this theory to itself in order to dismiss it as empty.

This discussion explains, I hope, why Poincare, by no means an un­sophisticated philosopher, preferred naive sensationalism to sophisticated sensationalism; it is untenable to claim that we do not quite know what we say when in ordinary circumstances we state a simple and un­problematic observation report; nor is it tenable to divorce such reports from scientific enquiry. But though Poincare’s rejection of sophisticated sensationalism is well founded in common sense, his acceptance of naive sensationalism was a serious error.

We have now come to the end of the list of traditional alternatives. Those philosophers who pin their hopes on the future success of present- day efforts to discover pure observation reports hope to erect a new in- ductivist epistemology or a less sophisticated conventionalism than Duhem’s. The rest are faced with the choice between apriorism, sophisti­cated conventionalism (a Kantian vacillation between the two), or the search for a revolutionary approach. The notorious conservatism of the bulk of philosophers (plus the unpopularity of apriorism and sophisti­cated conventionalism) is my only explanation for the popularity of the search for pure observational reports, for observable hard and fast facts, in the face of the increasing amount of evidence from modern psychology, from modern perception theory, which shows the futility of this search. Psychologists are usually unaware of the philosophical implication of their studies, of the fact that their studies give rise to the need for a new epistemology. But then they are usually not interested in this aspect.

The claim that there are no pure (or ‘neat’) observation reports is central to Ryle’s argument in his Concept of Mind (1949). By implication Ryle also rejects sophisticated conventionalism (when he denies the existence of a fundamental difference between common sense and science, pp. 288ff.). He thus faces the problem of epistemology, and he is well aware of it: he sketches a programme for a new epistemology (pp. 317-318). This need for a new epistemology is rooted not in Ryle’s central doctrine, in his proposed solution to the body-mind problem, but in his revolutionary perception theory. Popper, who dissents from Ryle’s solution to the body-mind problem, but shares Ryle’s perception theory, had outlined over a decade earlier a similar, if not the same, programme, and also proposed theories which answer the desiderata of that program­me. It is regrettable that this logic of Ryle’s argument has not been clearly seen by the general philosophical public. It is this oversight which is responsible for the popular view of Ryle’s doctrine as a version of behaviourism, even though he explicitly rejects behaviourism because it is based on the naive belief in pure observation reports. It is the same oversight which is responsible for the popular identification of Ryle’s psychological theory of knowledge as a set of dispositions with Mill’s and Schlick’s similar epistemological theory of knowledge as the proper procedure of connecting past and future events. (The two sets of theories of knowledge obviously come to solve two quite different sets of problems. The epistemological theory answers questions of status and basis of validity, the psychological theory answers questions of the seat of knowledge and its influence on behaviour. It is regrettable that Ryle’s metaphors allow for the confusion of his view with Schlick’s.)

Popper’s new theory of the status and methods of science is opposed by many philosophers because it entails the non-existence of pure observa­tion reports. This, as we have seen, is a very scanty ground for opposition. Others find it difficult to share his reasons for the acceptability of some observation reports in spite of their inherent uncertainty. It is this last point which I now wish to discuss in some detail.

5. Explanation vs. Consent

The whole literature concerning the methods of science seems to be agreed on one point, which I shall now try to criticize. It is agreed amongst philosophers that when it is said that a certain piece of information is scientifically acceptable it is meant that the piece of information in question ought to be accepted as true - to be believed. At least I have never come across any philosopher who has contested this. Popper has stressed that this acceptance must be tentative; but even he agrees that accepting a report is, for the time that it is accepted, considering it to be true. My own alternative is that observation reports ought to be accepted as a task, as something which we should try to explain, and this does not exclude the possibility that we should explain that piece of information as based on an error. This forces us to admit, I shall argue, that the problem of observation, the problem of why an observation report was made, and what is our guarantee that it is true, belongs to science and not to philosophy.

Science deals with factual information, but not with all factual in­formation and particularly not with information concerning miracles. Much has been written about the difference between scientifically acceptable and scientifically unacceptable information, and none of it seems to me satisfactory. Let me first state the difference and then discuss it. The bare facts of the matter seem to be these. In 1627 Galileo ridiculed (The Assayer) an argument from an unrepeatable ancient experiment. In 1661 Boyle published an essay ‘On the unsuccessful Experiment’ (in his Certain Physiological Essays) in which he ruled that science has nothing to do with unrepeatable experiments, that if we cannot repeat an experiment which someone claims to have performed we do not have to call him a liar or explain his claim in any other way - we can simply ignore it until it is reported to have been repeated by others. This proposal of Boyle has become a part of the scientific tradition. Although very few philosophers have discussed this situation, every physicist is well aware of it. Yet this situation should have been discussed more often, as it is problematic: the claim that any experiment is repeatable is a mere hypothesis. Boyle himself was extremely worried about this, because he thought that only factual information is certain to some degree (‘morally certain’), and that factual information is therefore always to be preferred to a hypothesis with which it clashes. Yet as the rejection of a hypothesis is based on the acceptance of an observation report and the acceptance of the observation report is based on the hypothesis that it is repeatable, it fol­lows that we reject a hypothesis not on the basis of solid facts but on the basis of another hypothesis. Is this not too arbitrary?

That the repeatability of an experiment is hypothetical can be shown by general considerations and by historical example. The general con­siderations are these: a description of an experiment is a description of the circumstances in which a certain event takes place, and a report of the experiment is the statement that at a certain time and place under the said circumstances the said event was indeed observed. Now many other circumstances were observed at the same time and place, of which there is no record and yet which may be, and sometimes are, essential to the success of the experiment. Thus, the success of the nineteenth-century experiments which made chemists think the atomic weight of chlorine to be 35.5 depended on circumstances which they did not notice but which we can vary today and thus approximate any result between 35 and 37 as we wish.

Boyle was aware of this difficulty. He demanded that we should report as many of the circumstances under which the experiment has been con­ducted as we can, and that we should vary the circumstances as much as possible. But the more circumstantial the description, the less repeatable is the described experiment; we do not know all the circumstances; we cannot vary all of them; and we cannot even report all of those which we notice. Boyle’s last and posthumous publication, Experimenta et Ob­servations Physicae, is burdened with superfluous descriptions of irrelevant circumstances. Yet in his preface to it, which he probably wrote on his death-bed, he expressed the fear that negligently he had omitted some relevant circumstances, thus rendering his own experiments unrepeatable.

The cause of this insoluble problem of Boyle is, I suggest, his rule, according to which whenever a hypothesis and a report of a repeatable experiment contradict each other it is the hypothesis which has to be thrown overboard. To my knowledge nobody has contested this rule. Even Popper, the first philosopher who has stressed the utter and in­escapable tentativity of all observation reports, has accepted Boyle’s rule. Yet the rule has to be rejected. Here is a historical example of a case in which the rule was at first correctly broken and then mistakenly ad­hered to.

In 1815—16 Prout published his celebrated hypothesis, according to which the ancient philosophers’ primordial matter is identical with hydrogen. According to this hypothesis the chemical atoms are not quite atoms, or indivisible, and their atomic weights must be multiples of the atomic weights of hydrogen atoms, namely whole numbers. Prout’s essay is full of experimental evidence, mostly not his own but compiled from the most up-to-date works of the leading chemists of his age. None of these results agreed with Prout’s hypothesis very well, and some of them did not agree with it at all. Yet he evidently considered these results as quite encouraging.

A short time later a youngster, Jean-Servais Stas, heard about this hypothesis and, to use his own words, fell in love with it. Like Prout he

hoped that with the improvement of the available experimental tech­niques the results of the measurements of atomic weights would converge towards the results predicted by Prout. Stas soon became the greatest expert in the field. His results did not agree with his expectation, and they broke his heart: he declared that his loyalty to science stood above his loves; consequently he gave up Prout’s hypothesis. One may re­member that some of the techniques by which isotopes are isolable were available to Stas. Had he insisted that the atomic weight of chlorine cannot be 35.5 he might have suggested that chlorine is a mixture of two physically different though chemically identical substances. But unlike Prout, Stas refused to stick to the hypothesis in the face of known facts in the hope that the facts will adjust themselves to theory rather than the other way round. To be more accurate, he was willing to modify his theory in one way, but one way only. Had atomic weights converged well to even whole-numbers or half-numbers, he would double the con­jectured number of hydrogen atoms in any given compound; but chlorine refused to converge even to 35.5.

This example shows that we have to improve upon Boyle’s rule. I suggest that Popper’s theory allows for a new rule. According to Popper’s view scientific theories are explanatory and testable, and the more highly explanatory and testable they are the better. This view seems to me to have gained a sufficiently wide recognition to enable me to use it without any preliminaries. My present discussion, if correct, renders Popper’s theory of the empirical basis of science superfluous; this theory (Section 29 of his Logic of Scientific Discovery) is perhaps the subtlest and most in­triguing part of his study, but it is also unsatisfactory in its very subtlety, and the cause of most of the criticisms and the misunderstandings of his views. I am glad it can be dismissed without any loss.

As Popper has argued, the demand for high testability leads to the demand to exclude the explanation of a series of successful repetitions of an experiment as due to chance. As he has also noticed, the demand for testability justifies the rule according to which unrepeatable experiments should be ignored, since repetitions are a kind of test. This led him to the tacit assumption, which I propose to reject, that results of repeatable experiments must be (tentatively) accepted as true. That he does make this assumption, though tacitly, can be seen in his acceptance of Fries’ claim that as the acceptance of observation reports should not be dogmatic it must be justified. This is a sensationalist relic in his theory. It led him to agree with Fries that the attempt at a justification leads either to an infinite regress or to a sensationalism. His own solution to the problem is that although we do not go on for ever testing observation reports by repeating the observation, we can do so when and if challenged. Hence, says Popper, there is an element of dogmatism or conventionalism in the acceptance of the report, since it may be false, as well as a sensational element, as it is causally related to sensations, as well as an element of (potential) infinite regress, since the possibilities of testing it are in­exhaustible.

All this can be ignored, I propose. We need speak neither of acceptance, nor of justification of acceptance, of any observation report. We merely have to demand that account be taken of the fact that some observation reports were made repeatedly, and that this fact be explained by some testable hypotheses. The demand to explain given observation reports by highly testable hypotheses entirely suffices. If the most testable hypothesis explains given observation reports while assuming them to be true, which is sometimes the case, we choose that hypothesis. Yet the most testable hypothesis may explain the observation reports as being the results of crude measurements, as Prout’s hypothesis did; or as results of sense illusions, as many psychological hypotheses do; or as results of specific initial conditions or specific circumstances, as Einstein’s relativity did; or as lies and propaganda - remember the totalitarian scientists! There is no empirical reason to reject such hypotheses on the basis of past experience; rather we go and test them by having recourse to new experi­ments.

To put it differently, it is not for the general theory of scientific experi­ment to explain why an experimental report was made, since the possible and even the actual explanations are varied. It is the task of a scientific hypothesis to do this. In particular, we must consider as false Boyle’s, Fries’, and Popper’s view, according to which all (repeatable) scientific observation reports are explained (tentatively or not) as true, and are therefore preferred to hypotheses which conflict with them. Whenever a report is made repeatedly, a scientific hypothesis which explains why it was made is sought for. And of all those specific hypotheses which are found, that one is preferred which is more testable than the others. Thus, when Mercury was reported to deviate from its Newtonian path, a few explanations of it were offered. One explanation of the report was based on the assumption that the observation was inaccurate, i.e. that the report was false. Another on the assumption that the initial conditions in the vicinity of the sun are more complicated than previous observers had assumed, i.e. that the report was true, but that other reports were false. Both these explanations incorporated Newton’s theory of gravity. Yet another explanation of the situation was Einstein’s theory of gravity. And the latter was preferred and tested, as it was the most easily testable. The preference for Einstein’s theory over the other two alternatives was definitely not based on the fact that the other alternatives incorporated the assumptions that some previous observation reports had been in­accurate : Einstein’s theory incorporated the assumptions that practically all observation reports had been inaccurate, of course. If this were not so, practically every observation report would refute the theory which the observation came to test: harly any observation ever fully agrees with the prediction which it comes to test.

The philosophical problem of the acceptability or otherwise of ob­servation reports can thus be entirely ignored by non-sensationalists; no philosophical problem even corresponds to it outside sensationalism. Instead, many scientific problems correspond to it: in each field of enquiry investigators have to explain all repeated observation reports, and they may explain them as true, as approximations, or as sheer fancy. These explanations are not justifications and therefore should be sus­pected, and therefore should be tested. Indeed, the assumption that a new theory contradicts an older observation report is itself a suggestion of how that theory may be tested, namely by repeating the older observation with a higher degree of accuracy. This has been recently noticed by Popper (in his ‘The Aims of Science’). But he did not notice, I think, that this amounts to the admission that observation reports may be accepted as false and that hence the problem of the empirical basis is thereby disposed of, which is my proposed view.

This proposal of mine severs the last connection between the philos­ophy of experience and sensationalism, by suggesting that philosophy should not include the attempt to discuss the causes of observation reports. Consequently my proposal sounds dangerously idealistic. I wish to argue that, on the contrary, it is the most realistic approach to experiment that has ever been proposed.

6. The Roots of Scientific Realism

The chief objection to my view would be that it is idealistic. But it is not idealistic; it leaves it to scientific hypotheses to say whether an observa­tion report is true, near to the truth, or utterly false; it leaves it to scientific hypotheses to say whether a specific observation report was stimulated by sensations emanating from things, by hallucinations, by dreams, or by the desire to achieve fame. My view may sound idealistic, but only because it trusts science to take care of realism; which science is doing very well.

But why is science realistic? The generally accepted answer is that scientists have a metaphysical faith in the existence of things physical. Following Popper I consider this answer as true but unsatisfactory: beliefs may dictate our acceptance of the scientific discipline, but the question is whether this discipline leads to realism, or whether we must add to this discipline a disposition towards realism in order to obtain science as we know it. Clearly this disposition is inessential. It is a simple fact that whether you are a realist or not, you must admit that the method of science alone already pushes you towards handling realistic hypotheses, whether you like them or not, whether you accept them or not.

The reason for this fact is very simple. Idealism is just one way of looking at our experiences, and a way whose importance was immensely exaggerated by contrasting it with all the infinitely many alternatives to it as if it were a contrast between merely two views, idealism and realism. The reason for this exaggeration is, of course, the claim, which I endorse, that sensationalism leads one irresistibly to idealism. But once we ignore sensationalism, idealism becomes one of the very many uninteresting ways in which we may try to account for our experiences. As Lewis Carroll knew, we can say not only that the world is my dream, but also that the world is his dream.

The scientific accounts of experience, then, are realistic plainly because they all differ from one historically famous though unscientific account of our experiences - idealism - an account which leads us nowhere, and which was considered significant because of its close relation to sensa­tionalism. As all versions of idealism are untestable, and scientific theories are highly testable, scientific theories are not idealistic, i.e. they are realistic. But is not our predilection for highly testable theories rooted in realism? The answer to the question as I have put it is, No. Science is realistic; more precisely, some versions of realism are scientific; but not all versions of realism are scientific. Realism alone is thus merely the rejection of idealism; it leads no more to science than to animism in any of its most primitive versions. We can be realists without wishing to explain or to test our explanations, but not vice versa. Let me show this by the following argument.

One may still feel that my attempt to ignore the general question of why observation reports are made is unrealistic; it may be unrealistic in a somewhat narrower and more naive sense than in the sense of being philosophically idealistic. One may suggest not that the whole world is my dream but merely that the scientific world is a dream. Do I allow for the possibility of a mock-science, of a situation in which some people build laboratories and state observation reports and some people try to explain them, but no one ever bothers to observe?

Let us take this possibility seriously for a moment, although it is puerile. I fear that it may have played an important role in the history of the philosophy of science even though it was never explicitly and carefully discussed (except, perhaps, by Bacon; he warned people against making reports without observing first; which, incidentally, is precisely what he himself did). Let us consider the hypothesis - call it hypothesis B- that there are very few observations and experiments going on anywhere on earth. I contend that at present almost nobody can check more than a negligible fraction of the observation reports which fill the current scienti­fic literature, and that even in one’s own field of research one must accept many reports without checking them. Thus, no one can deny hypothesis B on the basis of first-hand knowledge. But anyone can pose many awkward qeustions to those who accept hypothesis B; evading them will render hypothesis B unexplanatory, and attempts to answer them will render it more and more ad hoc, i.e. less and less testable. Hence, one who accepts Popper’s demand for explanation and high testability, will reject hypoth­esis B. All the other existing approaches will make one feel very disturbed by hypothesis B once one has taken it seriously. Sensationalism forces one to take it seriously.

The whole point of the present discussion can be summed up by stressing the unreasonableness of taking hypothesis B serioulsy on philosophical grounds together with the reasonableness of taking it seriously as a testable explanation of a picture of the situation which we may have, say as a result of a hypothetical victory of Nazism. This is nothing but Popper’s revolutionary thesis that the basis of science is social and not psychological.

But why do people observe? Why do they not simply imagine facts? My first answer is that they do, in all earnestness, try to imagine facts, but that their imagination is ludicrously less informative than the imagina­tion of experimental investigators. Not only is the imagination of the author of Arabian Nights infinitely inferior to that of Jules Verne; when a cinematic version of a science-fiction novel of Verne is done nowadays, its script-writers have to improve upon his imagination - by using what men of science present, rightly or wrongly, as observed facts! Facts are stranger than fiction. Fiction is a very poor substitute for obser­vation!

But this is only my preliminary answer. I do not wish to imply for one minute that we prefer observation to fiction because it is a better fiction than fiction; nor do I wish to belittle the significance of fiction (including that of Jules Verne) as a stimulus for observation; I only wish to argue that the fear of illusion which has ridden philosophers is rooted in an incredible overestimate of the power of our imagination. The attempt to explain an observation report, whether as a result of observation or as a result of hallucination, shows that we do not think we are so good at self-illusion: otherwise we could explain all observation reports, past and future, as a result of illusions, which is a version of idealism. It is because we do realize the limitation of our imagination that we have to ignore hypothesis B. The attempt to explain already implies that we think that we live in a world populated with humans who observe, think, and make statements, often because they think, rightly or wrongly, that they are true. In brief, we observe in order to test, though we do not always succeed. This is why I think that the problem of observation has been overrated: it has been overrated because the significance of the desire to explain or to comprehend has been underrated. The desire to explain, in its turn, has been underrated because the desire for certitude was great, and imaginative explanation is quite a different kettle of fish from certainty of any kind. As the quest for certitude or near certitude has to be abandoned anyhow, and as the demand to present highly explanatory and highly testable theories is realistic enough, we may leave it to science to explain each observation report in the most suit­able way without trying to explain, in addition, observation reports as such.

7. Conclusion

We all start from “Naive realism”, i.e. the doctrine that things are what they seem. We think that grass is green, that stones are hard, and that snow is cold. But physics assures us that the greenness of grass, the hardness of stones, and the coldness of snow, are not the greenness, hardness, and coldness that we know in our own experience, but something very different. The observer, when he seems to himself to be observing a stone, is really, if physics is to be believed, observing the effects of the stone upon himself. Thus science seems to be at war with itself; when it most means to be objective, it finds itself plunged into subjectivity against its will. Naive realism leads to physics, and physics, if true, shows that native realism is false. Therefore naive realism, if true, is false; therefore it is false.

This passage from the beginning of Russell’s An Inquiry Into Meaning and Truth (pp. 14-15), which has aroused the admiration of Einstein, is the core of Einstein’s comments on that book (in The Philosophy of Bertrand Russell). Einstein explains there how the desertion of naive realism led to sophisticated sensationalism and thus to idealism as the only alternative to apriorism. T am particularly pleased to note’, says Einstein in the conclusion of his comments, ‘that, in the last chapter of the book, it finally crops out that one can, after all, not get along without “metaphysics” [i.e. without unwarranted realism]. The only thing to which I take exception there is the bad intellectual conscience which shines through between the lines.’ This ‘bad intellectual conscience’, to sum-up, is rooted in the following implicit assumptions. First, that there exists only one picture of the world which may be properly viewed as naive realism. Second, that if science explains this naive picture of the world, it ought to accept it as true - which it does not. Third, that science explains not our naive picture, which is false, but reports about our sensations, which are true. In contrast to these tacit assumptions of Russell, and in accord with what I take Einstein’s tacit proposed to be, I propose the following view. (1) All pictures of the world which science explains are realistic. (2) All of them are naive to this or that degree. (3) Yesterday’s frontier of science is today’s rather naive realism. (4) Science is the attempt to explain the existing picture of the world, but this attempt is not based on the adoption of this picture; rather it leads to changes of the picture. (5) As Popper has suggested, science must remain at war with itself if it is to progress.

appendix: on privileged access

That everyone has some privileged access to some information is trivially true. The doctrine of privileged access is that I am the authority on all of my own experiences. Possibly this thesis was attacked by Wittgenstein (the thesis on the non-existence of private languages). The thesis was refuted by Freud (I know your dreams better than you), Duhem (I know your methods of scientific discovery better than you), Malinowski (I know your customs and habits better than you), and perception theorists (I can make you see things which are not there and describe your perceptions better than you can). The significance of this rejected thesis is that it is the basis of sensationalism and thus of all inductivist and some conventionalist philosophy.

The doctrine of privileged access has two versions; one is commonsense and trivially true; the other is somewhat more philosophical yet still fairly commonsense - and false. The true version is this. Every person has access to some information available to that person alone, and it involves one’s self, at least as an eye-witness. There is no doubt that this doctrine is true - at least in the sense that not a single author has ever put forth so much as one paragraph in an attempt to question it. The false version is that every person knows his own self best.

It is hard to say what Ludwig Wittgenstein has meant to say in his denial of the existence of a private language - or even in his very term ‘private language’. As is well-known, what he wrote on this is brief and cryptic - and subject to a whole exegetic literature. That literature seems to agree on one and only one point, and it is that Wittgenstein did try to assert the truth of the true version of the privileged access theory and deny the false version of it, and to link both with his theory of language as a social entity, as a way of life. However, in the wake of Judith Jarvis Thompson’s critique of some exegetes {American Philosophi­cal Quarterly, 1964), I shall leave this point as utterly open as I can.

The false version of the doctrine is the one which has an interesting history which I shall now sketch - and so from now on I shall entirely ignore the true and commonsense version, and identify the doctrine with its false version without fear of confusion.

The doctrine of privileged access is a consequence of a certain theory of knowledge, namely sensationalism or the doctrine that all knowledge - of the world and of one’s own self - derives from sen­sation. The clearest formulation of this doctrine was presented by Rudolf Carnap under the title ‘Methodological Solipsism’. This title, unfortunately, has caused confusion, as Carnap narrates in his Testability and Meaning (1936); people took him to be advocating solipsim proper. Yet, he insists, all he means is that everything I know about the world - at least about the world of facts - is either what I witnessed directly or what someone else has witnessed and then conveyed to my senses as sound or as the printed page, etc. From this, of course, the doctrine of privilleged access follows at once with the aid of the following premiss. I observe my own inner states directly and at any time I am conscious; I observe my outer states at any time I am conscious; contrastingly, others observe my outer states only when I am within their field of vision, and my inner states even more seldom.

So matteres have stood within philosophy for most of the time, espe­cially witin the empiricist school. But things have changed dramatically with the almost simultaneous discovery to the contrary in various fields. The discovery is due to Freud, Duhem, and Malinowski.

Freud was a sensationalist, likewise Duhem; most likely so also was Malinowski. Yet, they all showed that the auxiliary hypotheses necessary to derive the doctrine of privileged access from sensationalism are probably false. This is how the discovery failed to shake sensationalism within philosophy. This is regrettable, I suppose, since sensationalism seems to be false. I must admit that the survival of sensationalism is legitimate from the logical point of view, even though somewhat dogmatic from the methodological one. The original doctrine of sensationalism was important as it gave rise to views about the development of science, but these views included the theory of privileged access. In its modern purified version, sensationalism does not contain the old errors but, likewise, it has no relevance to theories of learning which avoid the old errors. Sensationalism, thus, survives merely as a relic of a glorious past.

Freud’s discovery within psychology is very simple. Possibly, your information was valid, you have suppressed it, and I have discovered it with the aid of (psychoanalytic) theory. Hence I know you better than you do. Of course, this assumes the theory to be true. Indeed, but this is not question-begging, because the theory has been empirically verified (claims Freud) by ample previous data. Moreover, I can some­times refresh your memory or make you release the suppressed memories you have, and thus have my information about you confirmed later by you.

A striking case is that of an analyst who hypnotizes his patient in order to reveal certain repressed data, and who instructs the patient to forget the hypnotic session in order to allow the release of repression to take place in regular analytic sessions (so as to prevent the dangerous side-effects of a sudden self-revelation). Few psychoanalysts take recourse to such a technique and I do not know with what degree of success; but success has been recorded.

More striking, I think, is the case of a mildly hysterical patient whom an analyst can often read like an open book. Why do you resent X? asks the analyst early in the encounter. Why, me? Retorts the hysterical patient with some indignation, claiming privileged access and observing that he has not reported any resentment, least of all toward X.

Most remarkable, be it true or false, is Freud’s identification of at least one such unknown object of strong resentment in most male members of the population, namely their fathers. This is part and parcel of Freud’s theory of the Oedipus complex.

Since then much more has been recorded, and, in the field of per­ception-theory proper, refuting the privileged-access theory. Any dentist remembers patient after patient feeling a toothache in the wrong place; children may feel a toothache in the stomach. Sometimes such patients, fervently believing in the theory of privileged access, make such nuisances of themselves that dentists prefer to give them total anaesthesia first so as to prevent argumentation. The dentist may know in advance that the patient has such poor knowledge of himself as to insist on the truth of the theory of privileged access.

In the field of science the theory of privileged access has made people consult scientists about their activities, especially the more successful ones; the assumption is, they could easily answer if they only would. Joseph Priestley reported his discovery of oxygen three times, finding it very instructive for future scientists, especially in view of the fact that some scientists report their discoveries incorrectly. Not that they do not know the truth - Priestley, an eighteenth-century sensationalist, fervently believed in the theory of privileged access. Had scientists reported their discoveries honestly - ‘ingenuously’ is the word he uses - they would all report the truth, namely that their discoveries had been accidental and that even they themselves were surprised at the time. All too often, however, Priestley charges, scientists invent hypotheses to explain their discoveries and then claim that the discoveries were neither accidents nor surprises, but corollaries to the hypotheses, facts sought for after the hypotheses had been invented. This, says Priestley, they do in order to lend authority to their own hypotheses.

The story of Priestley is but one instance of the application of the theory of privileged access to the history of scientific discovery. Of course, we should not object to scientists narrating their own stories, whether at their own volition, at others’ prompting, or as the ritual of the Nobel Prize Acceptance Speech requires.

The question is, How reliable and how generalizable are such reports? Pierre Duhem has argued that for logical reasons the reliance on scientists should not go as far as it all too often does. When narrating a discovery, all too often a scientist applies a theory of scientific discovery to his own case. (For example, Priestley was applying Bacon’s theory of the surprise and accident of discovery; cf. Novum Organum, I, Aph. 109.) And different scientists hold different theories of science which are not com­patible with each other.

Most scientists have believed that theory is based by induction on observation and report their activities as rooted, chronologically, in observation (in the way Priestley did and cajoled others to do). It was the fact that many scientists were inductivists and Duhem’s own op­position to inductivism which forced him to deny the privileged access of the scientist even to the scientist’s own contribution to science. All this has been put ironically by Einstein in his Herbert Spencer lecture of 1933, using the theory of induction against itself. Do not listen to what scientists say they do, he said, but look at what they do. He adds that having said this he should cancel his own lecture, but that he refuses to do so. All he meant is to deny privileged access, and so his own lecture, too, is merely deprived of that status. Hence, it is not that we should not listen to scientists, but that we need not accept their testimony as the Gospel truth; they have no privileged access.

To return to Priestley for a moment. He was in error in reporting that he discovered oxygen by mere accident. His own narrative enables us to show with no great difficulty (as I have done in my Towards an Historio­graphy of Science) that he was refuting Bayen’s hypothesis of metallic combustion, or his own improvement on it; and Bayen’s hypothesis was an extension of Joseph Black’s theory of fixed air and very much debated at the time as described in beautiful and vivid detail by A. N. Meldrum {The Eighteenth Century Revolution in Science).

Finally, to Bronislaw Malinowski. All anthropological or ethnological information about China accrued prior to the advent of Malinowski is now considered as a reflection not of the situation described, but of the naivete of the ethnologist who believed the informant, and of the in­formant’s obligation to present facts as in accord with Confucian doctrine and good sense and the like. The idea that China is classless, and that the Chinese examination system for public service is an in­strument of equality, which had so much beneficial effect on the West, is one such extreme instance.

Perhaps the discovery of the non-existence of any privileged access to one’s own social and cultural heritage should go not to Malinowski alone, but to Durkheim and Malinowski together. Be it as it may, for a time it was the fashion in British social anthropology to discard the declared reason for a custom, or an institution, as a mere veil concealing from its own practitioners its true meaning. Thus E. E. Evans-Pritchard (The Nuer) found in feuds between diverse tribes a means of keeping them in contact, quite counter-intuitively. Others found exogamy and incest taboos to be such instruments of contact with others. In his Lectures on Social Anthropology (1952), Evans-Pritchard offers a number of intriguing instances, not all from primitive society. Indeed, Durkheim already viewed religion, crime, and even suicide, as society’s means of self-preservation regardless of the individuals who happen to be the instruments of society and of their ignorance of the true meaning of their actions as revealed by Durkheim.

One curious application of all this to the history of science is the fairly recent - correct or otherwise - study of the allegedly enormous influence of Naturphilosophic on science proper. Since it was regarded as dangerous and detestable to take Naturphilosophie seriously, clearly scientists could not easily report having done so - or even know about it. Even when a scientist does, on occasion, make such a confession as Oersted’s acknowl­edgement in his doctoral dissertation of his indebtedness to Schelling, this may be viewed as an exaggeration, or aberration, and thus slowly but nicely sink into oblivion.

So much for the doctrine of privileged access and its overthrow. As I have said, the overthrow does not refute sensationalism. We may maintain that only a trained observer is qualified to observe, and deny in all cases we have thus far discussed that the persons in question were trained observers. On the contrary, the fact that the alleged privileged access has led to false observations can be revealed not by the mistaken observer but by a trained observer - be he a Freud or a Malinowski.

So be it. Query: Is Freud a trained observer of Freud? Malinowski of his own custom? If so, then Freud does have a privileged access to his own mind. Indeed, it is an important point in the study of the status of psychoanalysis even at present; generations after psychoanalysis was developed, criticized, and reformed, some opponents still find this ob­jection to it very forceful and unanswerable. A psychoanalyst can be qualified only after he is successfully analyzed and thus have a privileged access to his own self. He may need another’s help in overcoming ob­stacles, but he has to overcome the obstacles - to acquire full self- knowledge, to possess privileged access to his own self. There is the question as to how Freud did it single-handed, but I shall pass over this question here.

The most important question is not even really whether any psycho­analyst has a privileged access to himself (though, on occasion, it may be important to a patient). The question is broader: Is there a trained observer? Are trained observers really infallible? Are there types of ob­servation which are truly incorrigible and infallible and final? If so, then we may perhaps find out irrefragable facts about minds, cultures, and scientific researches, whether ours or others. If not, then all I say about myself, or about you or him, is to some measure doubtful and hence hypothetical. If so, I cannot claim that I know fully even facts about myself; hence I cannot claim I know myself best; I cannot say that I know best just what I have observed (an expert can, and sometimes does, correct others’ testimonies). But then, what is the difference between pure conjecture and a (conjectural) testimony of an eye-witness? Why eye­witnesses at all?

This terrible question pushes us to sensationalism, to the view that there does exist a definable and identifiable class of observation reports which are not in the least hypothetical. We do not know how to search for these, however, except through the avenue of privileged access - of methodological solipsisms - now closed. So we may either search for another avenue in order to give sensationalism substance, or try to offer a new theory of eye-witness testimony to replace sensationalism. This latter avenue has been explored by Popper and by myself - but only very rudimentarily. Many philosophers are still sensationalists and, having no other avenue for locating the class of incorrigible observations, they all too often fall back on the theory of privileged access - hoping to give a new and modified version of it, impregnable to the criticsim outlined thus far and defining, inter alia, the concept of trained observation. Thus far there is little or no progress to report. But as long as so many hope for success and work for it, the outcome is not very easily predictable. Some may find one avenue fruitful, others another. Only time and rational argument can tell.