appendix: the role of corroboration IN POPPER’S PHILOSOPHY

The following is a comment on D. Stove’s review of Popper’s Logic of Scientific Discovery, ‘Critical Notice’, The Australasian Journal of Philosophy 38 (1959), 173-187. I find that review a forceful expression of a very typical reading of Popper’s view, namely, reading it as yet an­other theory of justifying scientific theories by favorable evidence.

I have used the phrase “the problems at hand” twice already; I could not be more specific as I am not quite certain what are the problems which Stove has in mind. I find it somewhat bewildering that Stove does not say more explicitly what problems Popper claims to be discussing, what problems Popper’s theory of corroboration is intended to solve, or what problems Stove thinks this theory ought to solve. I wish, therefore, to supplement Stove’s interesting but seriously limited review on these points as best I can. I shall explain why I think that his criticism is en­tirely valid, and that it is in no way a criticism of the book he was re­viewing (except, perhaps, for a criticism of some ambiguity on Popper’s part).

1. Popper's Problems and Central Tenets, Popper claims to have solved two traditional philosophical problems: the problems of induction (how do we learn from experience?) and of demarcation between science and non-science (by what criterion do we decide which hypothesis is scientific?).

(This is the core of Popper’s book, as he himself claims, and as, I should think, is rather obvious anyhow. In this core there is nothing explicit about corroboration, namely about the failure to refute hypotheses; but only of success in refuting some hypotheses, which success is alleged to constitute learning from experience. And Popper’s criterion of demarca­tion does not distinguish between refuted and corroborated hypotheses. By Popper’s criterion even hypotheses which were amply refuted by experience are fully entitled to the honorary status of scientific hypotheses.)

Stove’s review relates almost entirely to Popper’s theory of corrobora­tion; he devotes two or three rather short paragraphs to the core of the book (pp. 173-174), without relating this core to the theory of corrobora­tion in a clear and explicit manner. Moreover, in these two or three paragraphs he commits some significant errors of presentation.

For example, he attributes to Popper the view that “there is no problem at all about induction” although Popper stresses the opposite, namely that Hume’s criticism, which establishes the invalidity of inductive inferences, leads to a few genuine philosophical problems, one of which is the problem of induction, namely, ‘how do we learn from experience?’. Stove, apparently, translates ‘how do we learn from experience?’ into ‘how do we know which of our hypotheses is true or at least probable?’. For Popper this traditional translation is a mistake. Stove, however, does not notice this, and having attributed the wrong question to Popper, he thereby attributes to him the wrong answer, to wit, the view that a well-corroborated theory is likely to be true.

2. The Traditional View of Scientific 'Success'. Undoubtedly, the attitude towards Einstein’s general theory of relativity changed dramatically with the result of Eddington’s observation of the total eclipse, which agreed with the theory rather well. Such a result is viewed by physicists as a great “success”, a “verification”, a “proof”, or “confirmation”. Most physicists, I contend, display great pleasure when confronted with “verifications” and give them much publicity, while (with the important modern excep­tion of the discovery of Lee and Yang) the refutation of hypotheses is displeasing to them and is toned down or even pooh-poohed.

If we accept Popper’s view, then it seems that we have to view the scientists’ traditional delight in “success” and dismissal of “failures” as their misunderstanding of their own activities, or rather as their method of advertisement which need not be taken seriously. And yet, Popper claims, there is an important element in “success”.

3. A Descriptive Theory of ‘Success’. The empirical support which a theory gains from experience, or its “success”, or “confirmation”, or “verification”, is a measure of its having stood up to severe tests, or, in Popper’s terminology, its high degree of corroboration. This is the whole of Popper’s doctrine of what “success” is. I wish to stress that the theory according to which “success” is failure to refute a hypothesis does not entail that one “success” leads to another. Nor does it entail that “success” is a Good Thing. Much less does it include the view, sometimes stated explicitly, and more often implicitly, that theoretical science is a Good Thing because it is the body of “successful” theories.

4. Corroboration and Eliminative Induction. Most men of science, and almost all philosophers, think that “success” is a Good Thing, while Popper thinks that “failure” is a Good Thing. Now, many people have stated before that the refutation of errors is a contribution to learning. But they usually agreed that this is only so because some people have preconceived ideas which are mistaken and which are obstacles to learning and therefore have to be refuted; they usually agreed that, had people been cautious and slow to advance hypotheses and to commit themselves to them, the drudgery of refutation could be greatly reduced.

This last doctrine has often been attributed to Popper.¹ And time and again he has claimed that this is not his view. In Popper’s view no degree of corroboration of a hypothesis can secure that it will not be refuted in the next test; no degree of corroboration of a hypothesis makes it even slightly more probable that it will not be refuted in the next test. Corroboration implies neither verification nor any increase of probability.

To put this somewhat differently, we can neither hope to escape error, nor to make error less likely - not even in a limited field of research. All we can hope is that we shall eliminate some errors, and replace them by other, smaller errors. This is the main point where Popper differs from the eliminationist inductivist. He views as ends what they view as means.

Strangely, Stove knows this point, but he cannot quite take it seriously. Dismissing this point as inessential to Popper’s doctrine, he (rightly) finds Popper’s doctrine to be essentially in the eliminationist inductivist tradition; but his dismissal of this point can be explained as his inability to grasp Popper’s novel and revolutionary idea.

5. On the So-Called Rational Degree of Belief. The problem of induction - how do we learn from experience? - Popper has tried to solve not by his theory of corroboration but by his theory of gradual approximation to the truth by repeatedly making explanatory hypotheses and refuting them experimentally. Another problem is, how can we avoid teaching false views or at least diminish our liability to assert mistaken views? This question, says Popper, has only one answer: the less you say the less likely you are to err. But science is an attempt to say more and more about the world, so that those engaged in science should have no fear of asserting an erroneous view, but they should do the utmost to encourage criticism. The last traditionally important question is, what theory should we believe, and why? This question, which is central in the inductivist tradition, is hardly discussed by Popper. A glance at the index to his book will reveal how little he says about beliefs, and reading these passages on belief will reveal that most of them are critical of traditional views rather than constructive.

The faith in science is a faith in a certain open-mindedness and de­tachment of belief - even to the extent of toning down beliefs and viewing them as entirely private affairs. This is the opposite of the view that science tells us what are the true objects of belief, be those the true laws of nature or the most probable among the known hypotheses in the light of the known factual information.

In the beginning of his ‘Critical Notice’ (p. 174) Stove correctly at­tributes to Popper the view that empirical science can lead to disbelief, but not to belief. From the middle of his ‘Critical Notice’ onwards, how­ever, Stove identifies Popper’s view with eliminative inductivism, whose major aim is to prescribe beliefs in probable hypotheses or in verified laws of nature. Take away from eliminative inductivism this prescription of beliefs, and you do get Popper’s view. But then the problem arises, how do we gain theoretical knowledge from experience? The eliminative inductivist’s answer to this question is, of course, that learning from experience is the same as finding which is the most probable hypothesis - the very answer which Popper rejects.

6. Knowledge and Learning. The problem of induction concerns the question of how we learn from experience, a problem which belongs to the theory of learning, or methodology. Traditionally, however, philosophers have mainly concentrated on the problem of whether or not we have knowledge based on experience, a problem which belongs to the theory of knowledge, or epistemology. There is a good reason for this: if we could show that we have knowledge based on experience, then we could thereby show that indeed we do learn from experience. But all attempts to show that we have knowledge based on experience have so far failed. One of Popper’s major revolutionary approaches was to go back from epistemology to methodology. Popper takes for granted (as an empirical fact, if you will) that we do learn from experience, and he asks by which manner, in what way, we learn. It turns out that it is easier to discuss methodological problems than epistemological problems. This is why Popper has so little to say on epistemology though he says quite a lot on methodology.

I am afraid that Stove has completely missed the point. He claims (pp. 178, 180) that the difference between having invented a hypothesis to explain a set of facts and having discovered these facts as corrobora­tions to that hypothesis is “just nil”. In other words, he claims, no matter how we have arrived at our knowledge, it is the state of knowledge that matters. And he says further that discussing this question of how we arrived at our knowledge is a psychological matter. Thus, he claims in effect, there is room only for epistemology and for psychology, but not for methodology, which is outside these two fields.

Now the situation is this. It was Popper who strongly emphasized, in opposition to many inductivist philosophers (including Bacon, Newton, and Mill), that from the point of view of assessing our present-day knowledge, the question of how we have arrived at that knowledge is entirely irrelevant. But though epistemologically irrelevant, this question is methodologically highly relevant. Here Popper argues not that it is irrelevant whether or not hypotheses are derived from facts, but rather that it is most often not the case that hypotheses are found by looking at facts; rather they are found by trying to solve concrete problems. This is not psychology. Similarly, though epistemologically it is irrelevant to ask whether we found facts by opening our eyes wide enough, methodo­logically it is most important to notice that opening one’s eyes does not lead to the discovery of new facts, and that it is easier to arrive at new facts by trying to refute our present hypotheses. Psychologists may ask whether we like to refute our pet hypotheses; methodologically what matters is that we learn by doing so. Similarly, from the epistemological viewpoint there is no difference between having discovered facts before or after explaining them. This last assertion contradicts the central doctrine of Whewell. According to Whewell a hypothesis is verified only when it explains a new fact which it was not intended to explain. Now intentions and chronology are irrelevant to the question of whether the hypothesis at hand is true. And the fact that we are often more impressed with cor­roborations than with explanations is an irrelevant psychological fact. Now that we have agreed (Stove and myself) that corroboration is irrelevant to the appraisal of our state of knowledge, and that the fact that corroborations are impressive is a psychological irrelevancy, the question to be asked is whether corroboration constitutes progress of knowledge and, if so, why.

7. Learning by Corroboration. The significance of corroborating evidence is twofold; first, it is in certain respects new evidence, and secondly, it illustrates the high explanatory power of the corroborated hypothesis. I shall not discuss the fact that high explanatory power is considered valuable, because Popper discusses this point at great length. I should only argue that the fact that we pay much attention to a corroborated hypothesis can better be explained by the desire for explanatory power than by the desire for credibility. A corroboration of a refuted hypothesis is pointless from the credibility viewpoint. And yet some corroborations to already refuted hypotheses were very important, because, I suggest, they increased their explanatory power. In this way they rendered it necessary to build future hypotheses in a manner which would yield the refuted but corroborated hypotheses as a first approximation. (This is Bohr’s correspondence principle translated into Popper’s system.) It is easy to say, after the event, that the refuted hypothesis was important because it explained much, and not because it was corroborated; the fact remains that we learned that the hypothesis explained much by testing it further and by obtaining corroborations as the results of these tests.

Thus, we want to know not only whether a hypothesis is true or false, but also to find out, as sharply as possible, the limits of its explanatory powers, both from within (corroboration) and from without (refutation).

There is a third argument in praise of corroboration which I put diffidently because, although it is quite simple and may be of some im­portance, I find it difficult to construct an example for it. Imagine first a development in which a hypothesis A is refuted by the fact a, a hypothesis B which explains both A (as an approximation) and a₉ but is refuted in the first test by the fact b₉ and a hypothesis C which explains both B (as an approximation) and b. Imagine a second process, starting with the same hypothesis A and its refutation a₉ being followed by the hypothesis C which is later corroborated by the fact b. Here, not the corroboration, but the skipping of the stage B₉ is what makes the second process more rapid than the first. The corroboration is the result of the rapidity, and not vice versa.

To conclude, the value of corroboration lies in the discovery of the corroborating facts, in the discovery of as many facts as possible which are explicable by an existing theory, and in its being a result of rapid progress.

8. Is Corroboration Really Necessary? So far I have argued that the corroboration of a theory is enlightening - though less enlightening than its refutation - even if it is not really necessary. Stove, rooted in his in- ductivist tradition, takes it for granted that factual knowledge which is acquired after inventing a hypothesis could have been acquired before­hand. This is clearly not so with refuting evidence, which is observed as a result of a long chain of deductions of a prediction from a hypothesis. But since, as Stove observes, the corroboration of a hypothesis is a refutation of a previous alternative to it, perhaps corroborating evidence could be observed before the invention of the hypothesis it corroborates. I do not wish to decide this problem, but merely to point out that while Stove knows the answer to it as a matter of course, I find it worthy of a critical discussion.

It is a famous idea, already used by Galileo against inductivism, that we see what we think we ought to see, that we interpret facts in the light of theories. This idea, that experience usually tells us only what we have already thought out for ourselves, makes it particularly difficult to see how we can learn from experience. Both Galileo and Popper have claimed that we can escape this limitation (to some extent) by being extremely critically minded. But Popper, at least, admits that this is easier said than done, and even when we are on the alert we often tend to see things as we expect them to be.

Because of this, Popper suggests, it is preferable to have a number of alternative hypotheses, and to design crucial experiments between them. In this case one hypothesis may be a useful instrument with which to refute another, and in the process of using it we may corroborate it.

There exist striking historical examples to this effect. The deviation of Mercury from the path prescribed to it by Newton’s mechanics was not viewed as a refutation of that hypothesis. After all, a few apparent deviations had occurred before, and these were later satisfactorily ac­counted for without involving overthrow of that hypothesis. Moreover, only after Einstein had explained Mercury’s deviation along new lines was a similar, though smaller, deviation discovered in other planets, including Earth, in accord with Einstein’s hypothesis. A debate is going on at present as to whether the other two corroborations of Einstein’s hypo­thesis are not in fact refutations of it; I suggest that an alternative to Einstein’s hypothesis which would be in better agreement with these facts may alter the widely accepted attitude towards them (especially if that alternative be corroborated). So at least in some instances one may doubt Stove’s assumption that corroborating evidence could be discovered prior to the hypothesis it corroborates, at least in the sense that the evidence is viewed entirely differently when it is a corroboration, in the sense that what might be considered as small deviations or even statistical errors are now viewed as important facts.

But, we should notice, there exist important refutations which were never considered as corroborations. The Michelson-Morley experiment and the Lee and Yang experiments are famous examples of this.

It is obvious, since essentially we cannot predict the result of a test, that we cannot have a satisfactory explanation of corroborations in general. Nevertheless, I wish to express my profound sense of puzzlement at the incredible corroborations which some theories have received, in the past and in our own times. I cannot escape the feeling that it is as if a deity paid us a premium for any good explanation and let it be well corroborated before it be refuted. To use the inductivist language, I am willing to bet at any odds that if a reasonably good solution should be found to any of the major problems in contemporary physics, it would be well corroborated within a short time, and with much greater ease than it would be later refuted. I cannot support this feeling, which, of course, may be totally pointless. And I do not think that Popper’s philosophy so far explains the amazing fact, if it is a fact, that so much corroboration is to be found in the history of science, although, I think, Popper’s view is the only reasonable explanation of the value of corroboration.

But, I should add here, in his later work² Popper has claimed that corroboration is important, apart from being enlightening as discovery and increasing the explanatory power of a hypothesis, and is even essential to science - as encouragement to our research. I confess that I do not quite see this point, and I would here join Stove in asking what is the essential difference between corroboration and explanation. If psycho­logical factors are ignored, and if the corroborating facts would have been found before the hypothesis which they corroborate, why, then, is corroboration essential?

9. The Indispensability of the Corroboration of Factual Reports. As Popper has pointed out, one kind of corroboration is essential for science. It is the

corroboration of factual evidence, or rather of its spatio-temporal universalization, which is often called “a general fact” or “an observable” or “a generalization”. We know that only repeatable facts are considered in science, and repeatable in the sense of corroborable, not in the sense in which the reports about flying saucers are constantly flowing in. The reason for this is that a fact is important when it contradicts a theory. And it is important to have them corroborated in order to make it simpler to accept the reports as those of general facts rather than to explain them away in some ad hoc fashion.

Popper has no wish to explain the fact that the generalizations we propose are often corroborated: it is the task of specific scientific hypo­theses to explain specific generalizations which are corroborated (and thereby to explain the fact that they are corroborated). This would not satisfy an inductivist, since such specific explanations are dubious; he wants certitude or at least high probability. But he is asking for the im­possible. All one can say in a general manner is that if we had no cor­roborable general facts we would not have science as we know it. But a world with no general facts is, perhaps, one in which even life is impossible.

10. Conclusion. Stove takes it for granted that Popper’s theory of testing is a preliminary to his theory of corroboration, and that this latter theory is a solution to the problems: how do we know and what should we be­lieve? But Popper tries to solve the problem: how do we learn? His answer is: by criticizing our errors. The idea that anything we say can be a subject for a critical examination is the core of Popper’s philosophical attitude. Stove views Popper’s recommendation of the critical attitude as a part of his theory of corroboration, and he tries to see whether it is a necessary or an eliminable part of it. He is thus putting the cart before the horse. Popper takes the critical attitude as fundamental. Corrobora­tion, according to him, is one sort of happening in the history of science which results from this attitude and to which, in turn, this attitude should be applied. Stove takes it for granted that, to Popper, a corroborated theory is corroborated because it is true or likely to be true or credible. As I understand it, Popper’s philosophy contains the idea that we should take notice of a well-corroborated theory and try to explain the fact that it was corroborated - and a variety of explanations may be available, each of which should be critically examined. Undoubtedly, Popper’s philosophy is connected with a long-standing tradition; but it is the critical tradition of Galileo and Boyle, of Kant and Whewell, rather than the inductivist tradition of Bacon, Newton, and Mill.

NOTES

¹ See von Wright, Logical Problem of Induction, second edition, Oxford 1957; H. G. Alexander, BJPS 10 (1959), 234; and S. F. Barker, Induction and Hypothesis, Cornell 1957; I. Levi, Gambling With Truth, N.Y., 1967; this is by no means an exhaustive list.

² Conjectures and Refutations, Chapter 10. See below.