At the end of 1945 Popper and his wife returned from New Zealand, where they had spent the entire war, so that he could take up the newly created post of Reader in Logic and Scientific Method at the LSE. Logic has been a respectable and respected discipline since Aristotle, of course, but the same cannot properly be said about scientific method. Popper used sometimes to joke about the discomfort of holding down a post in a non-existent subject, his thesis being that the old vision, dating at least from Bacon, of a scientific method, or the scientific method, a method that painstakingly followed leads to the unravelling of Nature's mysteries, is a dangerous illusion. Nor did he have more time for anything like a distinctively philosophical method. Yet Popper's own philosophy, the core of which is the CRITICAL RATIONALISM of my title, is in a very important respect dominated by considerations of method, and is unusual in the way in which it ruthlessly replaces traditional issues of epistemology by methodological issues.
Rationality, according to critical rationalism, is wholly a matter of method; and this rationality in no way rubs off on, or has any effect on, the outcomes of our intellectual endeavours. It is not our beliefs, or our theories, that are rational, but the manner in which we handle these beliefs and theories. To be sure, this emphasis on methodological issues is not all that there is to critical rationalism, since the word 'critical' too is critical. What critical rationalism claims is that the only way in which our theories are to be investigated is critically, negatively. But the disappearance of the epistemological dimension of rationality is what is least understood about Popper's philosophy, and it is therefore that aspect that I want to make the main topic of this lecture.
In recent years I have written much about critical rationalism, and have tried to consider in great detail the principal objections that it must face, and respond to them in a way that is fully in accordance with critical rationalism itself. As I see it, the usual objections to critical rationalism simply are not objections to critical rationalism, but objections to some hybrid view generated by an unholy alliance of it with the justificationist philosophy that it seeks to replace. Most critics simply do not realize that they are attacking the wrong target. For many years, therefore, culminating in Critical Rationalism. A Restatement and Defence (1994), I have been as interested in trying to restate critical rationalism as in defending it. I want to be able to state it in a way that makes it quite obvious that the standard objections are unsuccessful. Let me cite very briefly one or two examples of such objections. Many of the standard lines of criticism against Popper's proposed solution to the problem of induction charge this solution with itself having to take advantage of some inductive principle, so that the solution only shifts the problem, and does not resolve it. But these complaints are misguided, since they decline to take seriously the proposal that the universal statements of science are the only principles of uniformity that we ever need, and that to commit ourselves to vaguer and unfalsifiable principles of uniformity is just perverse. Other complaints, especially those against the solution to the problem of how science is rationally applied (the pragmatic problem of induction), seem to amount to no more than an exercise, not in the art of rhetoric, but in the art of asking rhetorical questions. Critical rationalists, it is said, have failed to provide any critical rationalist reasons for supposing that their methods will work better than do other methods. That, according to critical rationalism, there are no such things as reasons of this kind, is disregarded. (The matter is not made simpler by the fact that Popper sometimes spoke of 'critical reasons'.) You will see from this catalogue of misery that I almost expect critical rationalism to be misunderstood, and to be the target of inaccurately fired objections, even though there are many people who do regard themselves as critical rationalists, and see their approach to philosophy as closely following in Popper's footsteps. The truth of the matter is, I think, that there are many who do agree with Popper at a methodological level - two of whom I might mention, since I have taken some pains in Chapter 6 of (1994) to criticize their own versions of what they call 'critical rationalism', are John Watkins and Alan Musgrave - but who disagree at the epistem-ological level. Our agreement that the justification, in whole or in part, of scientific theories is impossible means that we all treat extremely lightly the question of truth within epistemology (it is a different matter within logic). Truth really causes very little heartache to those who have long abandoned the view that only what is recognisable is valuable. Actually, as we shall see, there are at least three distinct levels of disagreement possible, the least interesting one being the epistemological issue of whether certainty is attainable, the next one methodological, and the deepest and most interesting level an epistemological level again. There are many with whom I can agree at the two upper levels; few now with whom I agree deep down. It seems therefore that, even though success is not to be anticipated, yet another reformulation of critical rationalism would not go amiss.
Everyone present today will know of the origins of Popper's ideas in his disillusionment with the complacency of Alfred Adler - and to a lesser extent with Freud and his followers - and in the contrast between their cocksure-ness and the daring but humble submission of Einstein's theory of gravitation to the dictates of experience in the Eddington eclipse experiment of 1919. Popper was persuaded that empirical evidence is used in science not in order to provide endless confirmation and reconfirmation of the theory under investigation, but only as a test of the theory's correctness. Scientific theories being characteristically universal in form, and thus, as we know from Hume, not empirically verifiable, he was led to stress the asymmetry between verifiability and falsifiability. Universal hypotheses cannot be verified, but they can be falsified by a single contradictory instance. Rather surprisingly, this asymmetry has been challenged by some writers - to my dismay I saw recently that Quine (1986), p. 621 is one of them - on the grounds that whenever a hypothesis h is falsifiable its negation not-h is verifiable. But the point Popper was originally making concerned the universal hypotheses of science, and the negation of a universal hypothesis is rarely itself universal except in a degenerate way. In fairness to Quine, however, I must stress that he accepts the existence of the asymmetry when theories rather than single hypotheses are at stake, since a theory may be regarded as an infinite conjunction of its component hypotheses, falsified if one of the components is falsified, but quite beyond verification.
The asymmetry Popper identified between verifiability and falsifiability is really only a small part of the story, an important part of the story but in some respects one that is as misleading as it is helpful. It has inevitably encouraged a simpleminded depiction of two celebrated doctrines in direct opposition: the verificationism of the Vienna Circle and Popper's alternative proposal, which has come to be known as falsificationism. That verification-ism and falsificationism are opposed is not to be denied. But they are not proper mirror images of each other. Quite beyond its not being, as verificationism was, concerned with matters of meaning, falsificationism goes well beyond a simple substitution of falsity for truth in the statement of verifica-tionism. Therein lie the most important features of falsificationism and its generalization, which we now know as critical rationalism.
In the first place, since everyone now agrees that conclusive empirical verification or proof of anything interesting is out of the question, we must accept that the doctrine that should be opposed to falsificationism is not verificationism but what can be called confirmationism. More generally, standing opposed to critical rationalism is justificationism, the doctrine that the principal purpose of rational inquiry is the justification (though not the proof) of the theories and beliefs that interest us. What is at issue, that is, to repeat the words I used above, is whether empirical evidence is used in science in order to provide endless confirmation and reconfirmation of the theory under investigation, or only as a test of the theory's correctness. More generally, the question is whether reason and argument are used in intellectual inquiry in order to provide endless justification and rejustification of the theory being examined, or only as a way of testing if the theory is correct.
Phrased like this, the opposition is seen to have more than one facet. On the one hand, there is the question of what we do when we subject a theory to empirical investigation. This is plainly a methodological question, and on this front confirmationists, who think that we should look only for positive confirming evidence, or at least that confirming evidence is of value, and eliminationists, who think that only negative evidence is worth pursuing, are in direct opposition. But there is another question, often missed: the question of whether in subjecting a theory to empirical investigation we are trying to do something significant to the theory, such as confirm it or disconfirm it, or justify it or undermine it, and value evidence to the extent that it succeeds in doing this, or whether we intend only to do something with the theory, and plan to leave it in the same state after testing as it was in before. Is a theory that has undergone the ordeal by experience in any way glorified or disgraced by what has occurred, or is it unaffected? This question I take to be an epistemological one, since it is concerned with the status of the theory as an item in our corpus of knowledge, pre-eminently a concern of epistemologists from Plato to today. Critical rationalism proposes that the answer to this second question is that theories are unchanged by their interaction with experience. The opposite answer is central to nearly all forms of justificationism.
There are still some who hold something like a verificationist answer to the first question, the methodological one. What we should do, in order to investigate a theory empirically, is to look around for, and collate, examples and instances of the theory in action. We should look for successful applications and illustrations, and try to explain away apparent anomalies and counterexamples. What is expected from an observational or experimental project is positive evidence; evidence, that is, that illustrates the scope and breadth of the theory, that enables us to make a case in the theory's favour. This way of looking at experimental and observational research seems to be needlessly closely tied to extraneous psychological and social factors, and I do not in the least agree with it. But I shall not argue against it here. For many others have appreciated the force of Popper's criticism of such empirical methods, that confirmations are often easy to find (especially as they are defined in many modern theories of confirmation) and tell us little. Indeed I go further, and maintain that in reality confirmations tell us nothing. There is an almost universally widespread confidence in the value of confirming evidence - even amongst those who think that it is to be obtained by negative methods. But what this value is has never, to my knowledge, been explained. Why should an investigator care whether or not a theory that interests him is empirically confirmed or supported or not? (See Miller 1996.) Of course, he may be interested in confirmation or support for its own sake, just as some people are more excited about the trappings of wealth and power than they are about the reality of these things. But if what the investigator is primarily concerned about is the truth or falsehood of what he investigates, he will find it very difficult to see what relevance a confirmation or a supporting instance could have. There is no logical connection, for instance, between probable truth and truth simpliciter; a theory may, after all, be probably true but not true (in this regard probable truth is quite unlike certain truth). It is accordingly mysterious why, if truth is our aim, we should care about confirmation. The response of some Bayesians (for example, Howson & Urbach 1993, p. 131) to this question is quite astonishing in its simple-mindedness.
This brings us immediately to our second question, described above as epistemologlcal. What are we trying to do when we subject a theory to empirical investigation? Are we trying to find out something about the theory, or are we trying to do something to it?
According to confirmationism, we are interested not only in determining the truth value of the theory that we investigate, interesting as that truth value may be, but in changing the relation between ourselves and the theory. But according to falsificationism the purpose of scientific activity is not the glorification or humiliation of the theory under investigation, but simply a determination, if possible, that the theory is false. In consequence we should restrict our activity to methods that have the power to accomplish this, and do without the rest.
The method of empirical investigation is to investigate the theory (something that exists in its own right, just as physical objects exist in their own right) by investigating what it says about the world. We investigate the theory by meddling with the world, but not by meddling with what the theory says about the world. Sometimes our intervention may lead to the falsification of the theory under investigation - not only in the sense that the theory is shown false, but in the sense that it is made false. But that is an adjustment to the world, not to the theory. The theory is untouched. And should we decide to discard the theory, that is something that goes beyond the empirical investigation.
To put the matter briefly, and perhaps a little too slickly, falsification-ists are interested only in relations between theories and the world, most importantly correspondence and lack of correspondence, but also subsidiary properties such as explanatory power and problem-solving ability, whilst con-firmationists (and justificationists in general) are as much interested in relations between the theories and ourselves and the evidence that we have in our possession, and especially the degree of support, or degree of confirmation, that it provides.
Just as we can sometimes (but, it seems, not always; in this contrast may lie a decently restrained explanation of the idea of natural necessity) change a theory's truth value not by changing the theory but by changing the world (compare Russell, 1912, pp. 202f.), so we can sometimes increase the problemsolving ability of a theory by inventing new problems for it to solve, and increase its explanatory power by inventing new phenomena that it can explain. Note that I say here 'invent', not 'discover'. Nevertheless, these properties of the theory - truth, problem-solving ability, explanatory power - remain objective properties, not relative to the available evidence (though, as in Popper 1959, p. 399, a relativized measure of explanatory power may also be defined). We may well make a theory a better problem-solver without recognising that we are doing so. In truth, what we do when we invent such new problems is not to change the theory but to change the world.
But the usual situation is that the relation between the theory and the world is not changed by the empirical investigation, because the world is not relevantly changed. This is therefore quite different from those relations between theories and evidence that concern confirmationists, the second term of which is not the world, but the evidence that we have about the world. It is this evidence that confirmationism attempts to alter. We see here the roots of the contrast between falsificationism, which emphasizes the dynamical aspects of theory change, and confirmationism, which characteristically is more interested (but not wholly interested) in its statics. By something like a sleight of hand confirmationists move from the platitude that we can change the relation between theory and evidence by changing the evidence, to the claim that in some way the theory itself becomes changed in the course of this operation. That is to say, a theory can become justified (though not of course conclusively justified) by the accumulation of evidence.
It is worth noting in passing that there is also an increasingly large group of philosophers who deny that truth is a value at all, or that we are ever interested in the truth values of the theories that come before us. According to this group, we should restrict our attention entirely to the relation between theory and evidence, to how a theory is changed by empirical investigation. The most prominent subgroup of these nihilists, for that, I fear, is what they are, are the subjectivistic Bayesians, who repudiate entirely the traditional interest in the truth of what they investigate, and affect to be interested only in its (subjective) probability; apparently on the grounds that since we cannot know for certain whether a theory is true or not, we should not worry ourselves about this aspect of it at all. Bayesians agree with critical rationalists, I think, that investigation does not change the theory investigated; but they confine their interest to what they can know about, namely the change that takes place in the investigating agent when evidence is collected. Despite a fragment of agreement, however, this position strikes me as so unattractive, so much a counsel of despair, that I cannot properly understand how anyone can be driven to it. I think the answer is that it is often confused, even by its adherents, with a probabilistic version of confirmationism. But the mere use of probabilities, even subjective probabilities, or of Bayes's theorem, does not amount to Bayesianism, any more than the use of Pythagoras's theorem amounts to Pythagoreanism. There are indeed many writers who write as if probability, or high probability, could serve a guide to truth (see Howson & Urbach, loc.cit.}. Others, more thoughtful, realize that this simply resurrects the problem of induction.
In my opinion the epistemological question and the methodological question that I have distinguished here are far too rarely distinguished. Many writers in what may loosely be called the inductivist tradition have agreed with Popper, indeed have stressed that it was all in the writings of Bacon and of Mill, that the method of criticism or elimination is what is fundamental to empirical research, and the mindless piling up of favourable instances is of much less significance. (Some of these points have been enlarged on mercilessly by Grünbaum, 1976.) Characteristic also of the inductivist tradition, however, is the view that the eliminative method not only eliminates errors, it also bestows honour on those theories that survive the onslaught. Oddly enough, this position is usually called hypothetico-deductivism, because of its recognition of the importance of deductive logic in the analysis of the hypotheses under fire. But truly deductive hypothetico- deductivism would stop there, and would acknowledge that all that testing can do is to eliminate errors. It tells us nothing more. In other words there are many thinkers who can agree more or less completely with critical rationalism over methodological issues, but who disagree totally over the epistemological question. Here there is a crucial difference. For what critical rationalists say is that the methodological issue exhausts the matter; there is nothing more to be said. It is in this sense that critical rationalism is quintessentially a methodological thesis; not because it says more about methodology than do other positions, but because it says less about anything else.
According to critical rationalism what is of central importance in the examination of any scientific theory is whether or not it is true. That it is true is indeed not something that we can recognise, but we can guess it, and if that guess is sustained through tests, there is certainly no need to discard it. But with luck we may be able to refute the guess empirically, to show that it is not consistent with the truth of accepted test statements.
Critical rationalism is to my mind a refreshing and an inspiring thesis, allowing us to see that a great deal of the epistemological tradition, which has been concerned with how and in what way our theories are justified, if only inconclusively, by our intellectual investigations, is unnecessary. We may say goodbye to a considerable portion of epistemology, since it worries itself about something that we do not need: that is, about the justification or reasons that we have for believing what we believe. Note that this farewell to epistemology is very different from that associated with people such as Rorty (1980). Faced with the impossibility of certain truth, critical rationalists retain their interest in truth, and discard not only certainty but all its surrogates. Pragmatists and instrumentalists retain their interest in certainty and its surrogates and discard truth. Indeed, little of my discussion above has been concerned with the question of whether or not certainty is an achievable goal. This epistemological issue I have been able to bypass, since the real difference between critical rationalism and justificationism does not lie there. It can be thought to do so only if the pursuit of certainty or con-clusiveness is identified with an exclusive use of deductive logic. But critical rationalists do not use deductive logic (in contrast to some other logic) because it is conclusive, but because it supports criticism and revision: if the conclusion of a valid deductive argument is false, then so is at least one of the premises. Ampliative logic, if it existed, would play no part in the critical enterprise.
As I said earlier, there are three respects - at least three respects - in which critical rationalism goes beyond the concerns of traditional Cartesian epistemology. The first is the denial of certainty or conclusiveness. Critical rationalism is a version of fallibilism. This is an epistemological advance, but not a very great one, or even perhaps always a very wise one, as we can see from a consideration of how various is the class of fallibilist theories. Many of them, after all, have taken the adoption of fallibilism as an excuse for abandoning deductivism, in the hope that this might encourage a genuine solution to Hume's problem of induction. The second advance on traditional ways of thinking comes with the realization that fallibilism is not enough, at least not enough to solve the epistemological problems posed by theoretical science. Something is gained by use of the hypothetical method, and empiricism is better served by the critical or eliminativist approach than by any apriorism. Here the asymmetry between verifiability and falsifiability plays a crucial role, which is not seriously put in question by any acknowledgement of universal fallibilism. The emphasis becomes an emphasis on criticism and negative argument, on conjectures and refutations. This, I take it, is a methodological advance, which is independent of fallibilism. You can be an infallibilist adherent of verificationism (such as the early Carnap), a fallibilist verificationist (such as the middle Carnap), an infallibilist elimina-tionist (what Lakatos, 1970, pp. 93-103 called a dogmatic falsificationist), or a fallibilist eliminationist. But being the latter will not make you a critical rationalist. You could still think, as many do, that critical discussion can provide reasons for (though not sufficient reasons for), that is, justify, the conclusions at which it arrives. That is, that the discussion can have a (positive, but not compelling) effect on what is concluded. If you think this, then you are not a critical rationalist.
For what has almost always been missed here is the realization that the asymmetry between verifiability and falsifiability would still be there even if science were interested in existential rather than universal statements. The critical method is crucial, I agree, but not only because universal hypotheses cannot be verified. For even if induction were reduced to deduction, even then - as realized by traditional sceptics, especially Sextus Empiricus, and quite explicitly by Mill - the problem of the justification of induction would not have been solved. Deductively valid arguments are all question-begging, but at least they are valid. But if the conclusion of an inference cannot be justified by a deductive argument, what hope is there for doing it with an inference that is not even correct? This leads to the realization that it is not just induction that cannot be justified, it is everything. For making perfectly explicit this element of critical rationalist thought, which is implicit in Popper (1934) and all later writings, we are indebted to the late W.W. Bartley III. It follows that arguments are not primarily used to justify, but to criticize; not because criticism is easier, but because justification is impossible. Admittedly, there are some who conclude that arguments are not used for anything at all, but I cannot discuss them here. This is the third, I think clearly epistemological - or, if you prefer it, antiepisternological - strand in critical rationalist thought. First fallibilism, second eliminationism, third non-justificationism.
The plain fact is that we may continue as rationalists without adding any epistemological dimension to our methodological endeavours. For the use of reason resides in what we do, in the methods that we pursue, not in what effect these methods have on the theories that we investigate. The great liberating force of critical rationalism is that it permits us to be logically rigorous without, driving us along the road to irrationality. For the realization that this is possible we owe a great deal to one man: Karl Popper. It was he who first challenged the view that the role of argument is to provide us with justification. By stressing the virtue of criticism it was he who made it possible for rationalism to hold up its head in dignity again.
David Miller, Univ. Prof. Department of Philosophy, University of Warwick
[1] Extract from a lecture presented at the Congress "L'Epistomologia di Karl R. Popper" held in Cesena in October 1994
L. J. Cohen & M. B. Hesse, 1968, 'Introduction', pp. vii-xiv of L. J. Cohen &; M. B. Hesse, editors, Applications of Inductive Logic, Oxford.
A. Grünbaum, 1976, 'Is Falsifiability the Touchstone of Scientific Rationality? Karl Popper versus Inductivism', pp. 213-252 of R. S. Cohen, P. K. Feyerabend, & M. W. Wartofsky, editors, Essays in Memory of Imre Lakatos, Dordrecht.
C. Howson &: G. J. Oddie, 1979, 'Miller's So-called Paradox of Information', The British Journal for the Philosophy of Science 30, pp. 253-261.
C. Howson & P. Urbach, 1989, Scientific Reasoning: the Bayesian Approach, La Salle; 2nd edition 1993. I. Lakatos, 1970, 'Falsification & the Methodology of Scientific Research Programmes', pp. 91-195 of I. Lakatos & A. E. Musgrave, editors, Criticism & the Growth of Knowledge Cambridge.
D. K. Lewis, 1981, 'A Subjectivist's Guide to Objective Chance', pp. 263-293 of R. C. Jeffrey, editor, Studies -in Inductive Logic & Probability, Volume II, Berkeley &; Los Angeles.
D. W. Miller, 1966, 'A Paradox of Information', The British Journal for the Philosophy of Science 17, pp. 59-61.
D. W. Miller, 1979, 'What's In A Numeral?', Philosophical Studies 35, pp. 323- 344.
D. W. Miller, 1994, Critical Rationalism. A Restatement & Defence, Chicago &; La Salle.
D. W. Miller, 1995, 'Sir Karl Popper on Logic & Scientific Method', pp. 26-32 of a booklet based on the Memorial Celebration for Sir Karl Popper, December 12 1994, London School of Economics.
D. W. Miller, 1996, 'What Use is Empirical Confirmation?', Economics & Philosophy 12, pp. 197-206.
D. W. Miller, 1997, 'Sir Karl Raimund Popper C.H. F.B.A., 28 July 1902 - 17 September 1994', Biographical Memoirs of Fellows of the Royal Society of London 43, pp. 367-409.
K. R. Popper, 1934, Logik der Forschung, Wien.
K. R. Popper, 1959, The Logic of Scientific Discovery, London.
K. R. Popper, 1966, 'A Comment on Miller's New Paradox of Information', The British Journal for the Philosophy of Science 17, pp. 61-69.
K. R. Popper, 1982, Realism, & the Aim, of Science, volume I of Postscript to The Logic of Scientific Discovery, London.
K. R. Popper, 1994, Logik der Forschung, Tubingen; 10th edition of Popper, 1934.
K. R. Popper, 1998, The World of Parmenides. Essays on the Presocratic Enlightenment, London.
K. R. Popper, 1999, All Life is Problem Solving, London.
K. R. Popper & D. W. Miller, 1994, 'Contributions to the Formal Theory of Probability', pp. 3-23 of P. Humphreys, editor, Patrick Suppes: Scientific Philosopher, Volume I, Dordrecht. (The copyright in this paper belongs to the authors.)
H. Putnam, 1974, 'The "Corroboration of Theories" ', pp. 221-240 of Schilpp 1974.
W. V. O. Quine, 1986, 'Reply to Jules Vuillemin', pp. 619-622 of L. E. Hahn & P. A. Schilpp, editors, The Philosophy of W. V. Quine, La Salle.
R. Rorty, 1980, Philosophy & the Mirror of Nature, Princeton.
B. A. W. Russell, 1912, The Problems of Philosophy, London.
P. A. Schilpp, editor, 1974, The Philosophy of Karl Popper, La Salle.
J. W. N. Watkins, 1984, Science & Scepticism, London