by Alan Sokal
This problem, like many others, was very well formulated by Hume:
It is a question of fact, whether the perceptions of the senses be produced by external objects, resembling them: how shall this question be determined? By experience surely; as all other questions of a like nature. But here experience is, and must be entirely silent. The mind has never anything present to it but the perceptions, and cannot possibly reach any experience of their connexion with objects. The supposition of such a connexion is, therefore, without any foundation in reasoning.
(Hume 1988 [1748], p. 138: An Enquiry Concerning Human Understanding, Section XII, Part I)
What attitude should one adopt toward radical scepticism? The key observation is that such scepticism applies to all our knowledge: not only to the existence of atoms, electrons or genes, but also to fact that blood circulates in our veins, that the Earth is (approximately) round, and that at birth we emerged from our mother’s womb. Indeed, even the most commonplace knowledge in our everyday lives – there is a glass of water in front of me on the table – depends entirely on the supposition that our perceptions do not systematically mislead us and that they are indeed produced by external objects that, in some way, resemble those perceptions.53
The universality of Humean scepticism is also its weakness. Of course, it is irrefutable. But since no one is systematically sceptical (when he or she is sincere) with respect to ordinary knowledge, one ought to ask why scepticism is rejected in that domain and why it would nevertheless be valid when applied elsewhere, for instance, to scientific knowledge. Now, the reason why we reject systematic scepticism in everyday life is more or less obvious and is similar to the reason we reject solipsism. The best way to account for the coherence of our experience is to suppose that the outside world corresponds, at least approximately, to the image of it provided by our senses.54
Science as practice
For my part, I have no doubt that, although progressive changes are to be expected in physics, the present doctrines are likely to be nearer to the truth than any rival doctrines now before the world. Science is at no moment quite right, but it is seldom quite wrong, and has, as a rule, a better chance of being right than the theories of the unscientific. It is, therefore, rational to accept it hypothetically.
(Bertrand Russell, My Philosophical Development, 1995 [1959], p. 13)
Once the general problems of solipsism and radical scepticism have been set aside, we can get down to work. Let us suppose that we are able to obtain some more-or-less reliable knowledge of the world, at least in everyday life. We can then ask: To what extent are our senses reliable or not? To answer this question, we can compare sense impressions among themselves and vary certain parameters of our everyday experience. We can map out in this way, step by step, a practical rationality. When this is done systematically and with sufficient precision, science can begin.
For us, the scientific method is not radically different from the rational attitude in everyday life or in other domains of human knowledge. Historians, detectives and plumbers – indeed, all human beings – use the same basic methods of induction, deduction and assessment of evidence as do physicists or biochemists. Modern science tries to carry out these operations in a more careful and systematic way, by using controls and statistical tests, insisting on replication, and so forth. Moreover, scientific measurements are often much more precise than everyday observations; they allow us to discover hitherto unknown phenomena; and they often conflict with ‘common sense’. But the conflict is at the level of conclusions, not the basic approach.55, 56
The main reason for believing scientific theories (at least the best-verified ones) is that they explain the coherence of our experience. Let us be precise: here ‘experience’ refers to all our observations, including the results of laboratory experiments whose goal is to test quantitatively (sometimes to incredible precision) the predictions of scientific theories. To cite just one example, quantum electrodynamics predicts that the magnetic moment of the electron has the value57
1.001 159 652 201 ± 0.000 000 000 030,
where the ‘±’ denotes the uncertainties in the theoretical computation (which involves several approximations). A recent experiment gives the result
1.001 159 652 188 ± 0.000 000 000 004,
where the ‘±’ denotes the experimental uncertainties.58 This agreement between theory and experiment, when combined with thousands of other similar though less spectacular ones, would be a miracle if science said nothing true – or at least approximately true – about the world. The experimental confirmations of the best-established scientific theories, taken together, are evidence that we really have acquired an objective (albeit approximate and incomplete) knowledge of the natural world.59
Having reached this point in the discussion, the radical sceptic or relativist will ask what distinguishes science from other types of discourse about reality – religions or myths, for example, or pseudo-sciences such as astrology – and, above all, what criteria are used to make such a distinction. Our answer is nuanced. First of all, there are some general (but basically negative) epistemological principles, which go back at least to the seventeenth century: to be sceptical of a priori arguments, revelation, sacred texts and arguments from authority. Moreover, the experience accumulated during three centuries of scientific practice has given us a series of more-or-less general methodological principles – for example, to replicate experiments, to use controls, to test medicines in double-blind protocols – that can be justified by rational arguments. However, we do not claim that these principles can be codified in a definitive way, nor that the list is exhaustive. In other words, there does not exist (at least at present) a complete codification of scientific rationality, and we seriously doubt that one could ever exist. After all, the future is inherently unpredictable; rationality is always an adaptation to a new situation. Nevertheless – and this is the main difference between us and the radical sceptics – we think that well-developed scientific theories are in general supported by good arguments, but the rationality of those arguments must be analysed case-by-case.60
To illustrate this, let us consider an example that is in a certain sense intermediate between scientific and ordinary knowledge, namely criminal investigations.61 There are some cases in which even the hardiest sceptic will find it difficult to doubt, in practice, that the culprit has really been found: one may, after all, possess the weapon, fingerprints, DNA evidence, documents, a motive, and so forth. Nevertheless, the path leading to those discoveries can be very complicated: the investigator has to make decisions (on the leads to follow, on the evidence to seek) and draw provisional inferences, in situations of incomplete information. Nearly every investigation involves inferring the unobserved (who committed the crime) from the observed. And here, as in science, some inferences are more rational than others. The investigation could have been botched, or the ‘evidence’ might simply have been fabricated by the police. But there is no way to decide a priori, independently of the circumstances, what distinguishes a good investigation from a bad one. Nor can anyone give an absolute guarantee that a particular investigation has yielded the correct result. Moreover, no one can write a definitive treatise on The Logic of Criminal Investigation. Nevertheless – and this is the main point – no one doubts that, for some investigations at least (the best ones), the result does indeed correspond to reality. Furthermore, history has permitted us to develop certain rules for conducting an investigation: no one believes anymore in trial by fire, and we doubt the reliability of confessions obtained under torture. It is crucial to compare testimonies, to cross-examine witnesses, to search for physical evidence, etc. Even though there does not exist a methodology based on unquestionable a priori reasoning, these rules (and many others) are not arbitrary. They are rational and are based on a detailed analysis of prior experience. In our view, the ‘scientific method’ is not radically different from this kind of approach.
The absence of any ‘absolutist’ cri
teria of rationality, independent of all circumstances, implies also there is no general justification of the principle of induction (another problem going back to Hume). Quite simply, some inductions are justified and others are not; or, to be more precise, some inductions are more reasonable and others are less so. Everything depends on the case at hand: to take a classic philosophical example, the fact that we have seen the sun rise every day, together with all our astronomical knowledge, gives us good reasons to believe that it will rise tomorrow. But this does not imply that it will rise ten billion years from now (indeed, current astrophysical theories predict that it will exhaust its fuel before then).
In a sense, we always return to Hume’s problem: No statement about the real world can ever literally be proven; but to use the eminently appropriate expression from Anglo-Saxon law, it can sometimes be proven beyond any reasonable doubt. The unreasonable doubt subsists.
If we have spent so much time on these rather elementary remarks, it is because much of the relativist drift that we shall criticize has a double origin:
Part of twentieth-century epistemology (the Vienna Circle, Popper and others) has attempted to formalize the scientific method.
The partial failure of this attempt has led, in some circles, to an attitude of unreasonable scepticism.
In the rest of this chapter, we intend to show that a series of relativist arguments concerning scientific knowledge are either (a) valid critiques of some attempts to formalize the scientific method, which do not, however, in any way undermine the rationality of the scientific enterprise; or (b) mere reformulations, in one guise or another, of Humean radical scepticism.
Epistemology in crisis
Science without epistemology is – insofar as it is thinkable at all – primitive and muddled. However, no sooner has the epistemologist, who is seeking a clear system, fought his way through such a system, than he is inclined to interpret the thought-content of science in the sense of his system and to reject whatever does not fit into his system. The scientist, however, cannot afford to carry his striving for epistemological systematic that far. ... He therefore must appear to the systematic epistemologist as an unscrupulous opportunist.
(Albert Einstein, 1949, p. 684)
Much contemporary scepticism claims to find support in the writings of philosophers such as Quine, Kuhn or Feyerabend who have called into question the epistemology of the first half of the twentieth century. This epistemology is indeed in crisis. In order to understand the nature and the origin of the crisis and the impact that it may have on the philosophy of science, let us go back to Popper.62 Of course, Popper is not a relativist, quite the contrary. He is nevertheless a good starting point, first of all because many of the modern developments in epistemology (Kuhn, Feyerabend) arose in reaction to him, and secondly because, while we disagree strongly with some of the conclusions reached by Popper’s critics such as Feyerabend, it is nevertheless true that a significant part of our problems can be traced to ambiguities or inadequacies in Popper’s The Logic of Scientific Discovery.63 It is important to understand the limitations of this work in order to face more effectively the irrationalist drift created by the critiques it provoked.
Popper’s basic ideas are well known. He wants, first of all, to give a criterion for demarcating between scientific and non-scientific theories, and he thinks he has found it in the notion of falsifiability: in order to be scientific, a theory must make predictions that can, in principle, be false in the real world. For Popper, theories such as astrology or psychoanalysis avoid subjecting themselves to such a test, either by not making precise predictions or by arranging their statements in an ad hoc fashion in order to accommodate empirical results whenever they contradict the theory.64
If a theory is falsifiable, hence scientific, it may be subjected to attempts at falsification. That is, one may compare the theory’s empirical predictions with observations or experiments; and if the latter contradict the predictions, it follows that the theory is false and must be rejected. This emphasis on falsification (as opposed to verification) underlines, according to Popper, a crucial asymmetry: one can never prove that a theory is true, because it makes, in general, an infinite number of empirical predictions, of which only a finite subset can ever be tested; but one can nevertheless prove that a theory is false, because, to do that, a single (reliable) observation contradicting the theory suffices.65
The Popperian scheme – falsifiability and falsification – is not a bad one, if it is taken with a pinch of salt. But numerous difficulties spring up as soon as one tries to take falsificationist doctrine literally. It may appear attractive to abandon the uncertainty of verification in favor of the certainty of falsification. But this approach runs into two problems: by abandoning verification, one pays too high a price; and one fails to obtain what is promised, because falsification is much less certain than it seems.
The first difficulty concerns the status of scientific induction. When a theory successfully withstands an attempt at falsification, a scientist will, quite naturally, consider the theory to be partially confirmed and will accord it a greater likelihood or a higher subjective probability. The degree of likelihood depends, of course, upon the circumstances: the quality of the experiment, the unexpectedness of the result, etc. But Popper will have none of this: throughout his life, he was a stubborn opponent of any idea of ‘confirmation’ of a theory, or even of its ‘probability’. He wrote:
Are we rationally justified in reasoning from repeated instances of which we have experience to instances we have had no experience? Hume’s unrelenting answer is: No, we are not justified ... My own view is that Hume’s answer to this problem is right.
(Popper 1974, pp. 1018–19, italics in the original)66
Obviously, every induction is an inference from the observed to the unobserved, and no such inference can be justified using solely deductive logic. But, as we have seen, if this argument were to be taken seriously – if rationality were to consist only of deductive logic – it would imply also that there is no good reason to believe that the sun will rise tomorrow, and yet no one really expects the sun not to rise.
With his method of falsification, Popper thinks that he has solved Hume’s problem67, but his solution, taken literally, is a purely negative one: we can be certain that some theories are false, but never that a theory is true or even probable. Clearly, this ‘solution’ is unsatisfactory from a scientific point of view. In particular, at least one of the roles of science is to make predictions on which other people (engineers, doctors, ...) can reliably base their activities, and all such predictions rely on some form of induction.
Besides, the history of science teaches us that scientific theories come to be accepted above all because of their successes. For example, on the basis of Newtonian mechanics, physicists have been able to deduce a great number of both astronomical and terrestrial motions, in excellent agreement with observations. Moreover, the credibility of Newtonian mechanics was reinforced by correct predictions such as the return of Halley’s comet in 175968 and by spectacular discoveries such as finding Neptune in 1846 where Le Verrier and Adams predicted it should be.69 It is hard to believe that such a simple theory could predict so precisely entirely new phenomena if it were not at least approximately true.
The second difficulty with Popper’s epistemology is that falsification is much more complicated than it seems.70 To see this, let us take once again the example of Newtonian mechanics,71 understood as the combination of two laws: the law of motion, according to which force is equal to mass times acceleration; and the law of universal gravitation, according to which the force of attraction between two bodies is proportional to the product of their masses and inversely proportional to the square of the distance separating them. In what sense is this theory falsifiable? By itself, it doesn’t predict much; indeed, a great variety of motions are compatible with the laws of Newtonian mechanics and even deducible from them, if one makes appropriate assumptions about the masses of
the various celestial bodies. For example, Newton’s famous deduction of Kepler’s laws of planetary motion requires certain additional assumptions, which are logically independent of the laws of Newtonian mechanics, principally that the masses of the planets are small relative to the mass of the sun: this implies that the mutual interactions between the planets can be neglected, in a first approximation. But this hypothesis, while reasonable, is by no means self-evident: the planets could be made of a very dense material, in which case the additional hypothesis would fail. Or there could exist a large amount of invisible matter affecting the motion of the planets.72 Moreover, the interpretation of any astronomical observation depends on certain theoretical propositions, in particular on optical hypotheses concerning the functioning of telescopes and the propagation of light through space. The same is true, in fact, for any observation: for example, when one ‘measures’ an electrical current, what one really sees is the position of a needle on a screen (or numbers on a digital readout), which is interpreted, in accordance with our theories, as indicating the presence and the magnitude of a current.73
It follows that scientific propositions cannot be falsified one by one, because to deduce from them any empirical proposition whatsoever, it is necessary to make numerous additional assumptions, if only on the way measuring devices work; moreover, these hypotheses are often implicit. The American philosopher Quine has expressed this idea in a rather radical fashion:
[O]ur statements about the external world face the tribunal of sense experience not individually but only as a corporate body. ... Taken collectively, science has its double dependence upon language and experience; but this duality is not significantly traceable into the statements of science taken one by one. ...
The idea of defining a symbol in use was ... an advance over the impossible term-by-term empiricism of Locke and Hume. The statement, rather than the term, came with Bentham to be recognized as the unit accountable to an empiricist critique. But what I am now urging is that even in taking the statement as unit we have drawn our grid too finely. The unit of empirical significance is the whole of science.