by Alan Sokal
(Quine 1980 [1953], pp. 41–2)74
What can one reply to such objections? First of all, it must be emphasized that scientists, in their practice, are perfectly aware of the problem. Each time an experiment contradicts a theory, scientists ask themselves a host of questions: Is the error due to the way the experiment was performed or analysed? Is it due to the theory itself, or to some additional assumption? The experiment itself never dictates what must be done. The notion (what Quine calls the ‘empiricist dogma’) that scientific propositions can be tested one by one belongs to a fairy tale about science.
But Quine’s assertions demand serious qualifications.75 In practice, experience is not given; we do not simply contemplate the world and then interpret it. We perform specific experiments, motivated by our theories, precisely in order to test the different parts of those theories, if possible independently of one another or, at least, in different combinations. We use a set of tests, some of which serve only to check that the measuring devices indeed work as expected (by applying them to well-known situations). And, just as it is the totality of the relevant theoretical propositions that is subjected to a falsification test, so it is the totality of our empirical observations that constrains our theoretical interpretations. For example, while it is true that our astronomical knowledge depends upon hypotheses about optics, these hypotheses cannot be modified in an arbitrary way, because they can be tested, at least in part, by numerous independent experiments.
We have not, however, reached the end of our troubles. If one takes the falsificationist doctrine literally, one should declare that Newtonian mechanics was falsified already in the mid-nineteenth century by the anomalous behaviour of Mercury’s orbit.76 For a strict Popperian, the idea of putting aside certain difficulties (such as the orbit of Mercury) in the hope that they will be temporary amounts to an illegitimate strategy aimed at evading falsification. However, if one takes into account the context, one may very well maintain that it is rational to proceed in this way, at least for a certain period of time – otherwise science would be impossible. There are always experiments or observations that cannot be fully explained, or that even contradict the theory, which are put aside awaiting better days. Given the immense successes of Newtonian mechanics, it would have been unreasonable to reject it because of a single prediction (apparently) refuted by observations, since this disagreement could have all sorts of other explanations.77 Science is a rational enterprise, but difficult to codify.
Without doubt, Popper’s epistemology contains some valid insights: the emphasis on falsifiability and falsification is salutary, provided it is not taken to extremes (e.g. the blanket rejection of induction). In particular, when comparing radically different endeavors such as astronomy and astrology, it is useful, to some extent, to employ Popperian criteria. But there is no point in demanding that the pseudo-sciences follow strict rules that the scientists themselves do not follow literally (otherwise one exposes oneself to Feyerabend’s critiques, which we shall discuss later).
It is obvious that, in order to be scientific, a theory must be tested empirically in one way or another – and the more stringent the tests, the better. It is also true that predictions of unexpected phenomena often constitute the most spectacular tests. Finally, it is easier to show that a precise quantitative claim is false than to show that it is true. And it is probably a combination of these three ideas that explains, in part, Popper’s popularity among many scientists. But these ideas are not due to Popper, nor do they constitute what is original in his work. The necessity of empirical tests goes back at least to the seventeenth century, and is simply the lesson of empiricism: the rejection of a priori or revealed truths. Besides, predictions are not always the most powerful tests;78 and those tests may take relatively complex forms, which cannot be reduced to the simple falsification of hypotheses taken one by one.
All these problems would not be so serious had they not given rise to a strongly irrationalist reaction: some thinkers, notably Feyerabend, reject Popper’s epistemology for many of the reasons just discussed, and then fall into an extreme anti-scientific attitude (see below). But the rational arguments in favor of the theory of relativity or the theory of evolution are to be found in the works of Einstein, Darwin and their successors, not Popper. Thus, even if Popper’s epistemology were entirely false (which is certainly not the case), that would imply nothing concerning the validity of scientific theories.79
The Duhem–Quine thesis: Underdetermination
Another idea, often called the ‘Duhem–Quine thesis’, is that theories are underdetermined by evidence.80 The set of all our experimental data is finite; but our theories contain, at least potentially, an infinite number of empirical predictions. For example, Newtonian mechanics describes not only how the planets move, but also how a yet-to-be-launched satellite will move. How can one pass from a finite set of data to a potentially infinite set of assertions? Or, to be more precise, is there a unique way of doing this? This is rather like asking whether, given a finite set of points, there is a unique curve that passes through these points. Clearly the answer is no: there are infinitely many curves passing through any given finite set of points. Similarly, there is always a large (even infinite) number of theories compatible with the data – and this, whatever the data and whatever their number.
There are two ways to react to such a general thesis. The first approach is to apply it systematically to all our beliefs (as one is logically entitled to do). So we would conclude, for example, that, whatever the facts, there will always be just as many suspects at the end of any criminal investigation as there were at the beginning. Clearly, this looks absurd. But it is indeed what can be ‘shown’ using the underdetermination thesis: one can always invent a story (possibly a very bizarre one) in which X is guilty and Y is innocent and in which ‘the data are accounted for’ in an ad hoc fashion. We are simply back to Humean radical scepticism. The weakness of this thesis is again its generality.
Another way to deal with this problem is to consider the various concrete situations that can occur when one confronts theory with evidence:
1 One may possess evidence in favour of a given theory that is so strong that to doubt the theory would be almost as unreasonable as to believe in solipsism. For example, we have good reasons to believe that blood circulates, that biological species have evolved, that matter is composed of atoms, and a host of other things. The analogous situation, in a criminal investigation, is that in which one is sure, or almost sure, of having found the culprit.
2 One may have a number of competing theories, none of which seems totally convincing. For example, the problem of the origin of life provides (at least at present) a good example of such a situation. The analogy in criminal investigations is obviously the case in which there are several plausible suspects but it is unclear which one is really guilty. The situation may also arise in which one has just one theory, which, however, is not very convincing due to the lack of sufficiently powerful tests. In such a case, scientists implicitly apply the underdetermination thesis: since another theory, not yet conceived, might well be the right one, one confers on the sole existing theory a rather low subjective probability.
3 Finally, one may lack even a single plausible theory that accounts for all the existing data. This is probably the case today for the unification of general relativity with elementary-particle physics, as well as for many other difficult scientific problems.
Let us come back for a moment to the problem of the curve drawn through a finite number of points. What convinces us most strongly that we found the right curve is, of course, that when we perform additional experiments, the new data fit the old curve. One has to assume implicitly that there is not a cosmic conspiracy in which the real curve is very different from the curve we have drawn, but in which all our data (old and new) happen to fall on the intersection of the two. To take a phrase from Einstein, one must imagine that the Lord is subtle, but not malicious.
Kuhn and t
he incommensurability of paradigms
Much more is known now than was known fifty years ago, and much more was known then than in 1580. So there has been a great accumulation or growth of knowledge in the last four hundred years. This is an extremely well-known fact ... So a writer whose position inclined him to deny [it], or even made him at all reluctant to admit it, would inevitably seem, to the philosophers who read him, to be maintaining something extremely implausible.
(David Stove, Popper and After, 1982, p. 3)
Let us now turn our attention towards some historical analyses that have apparently provided grist for the mill of contemporary relativism. The most famous of these is undoubtedly Thomas Kuhn’s The Structure of Scientific Revolutions.81 We shall deal here exclusively with the epistemological aspect of Kuhn’s work, putting aside the details of his historical analyses.82 There is no doubt that Kuhn envisions his work as a historian as having an impact on our conception of scientific activity and thus, at least indirectly, on epistemology.83
Kuhn’s scheme is well known: The bulk of scientific activity – what Kuhn calls ‘normal science’ – takes place within ‘paradigms’, which define what kinds of problems are studied, what criteria are used to evaluate a solution, and what experimental procedures are deemed acceptable. From time to time, normal science enters into crisis – a ‘revolutionary’ period – and the paradigm changes. For instance, the birth of modern physics with Galileo and Newton constituted a rupture with Aristotle; similarly, in the twentieth century, relativity theory and quantum mechanics have overturned the Newtonian paradigm. Comparable revolutions took place in biology, during the development from a static view of species to the theory of evolution, or from Lamarck to modern genetics.
This vision of things fits so well with scientists’ perception of their own work that it is difficult to see, at first glance, what is revolutionary in this approach, much less how it could be used for anti-scientific purposes. The problem arises only when one faces the notion of the incommensurability of paradigms. On the one hand, scientists think, in general, that it is possible to decide rationally between competing theories (Newton and Einstein, for example) on the basis of observations and experiments, even if those theories are accorded the status of ‘paradigms’.84 By contrast, though one can give several meanings to the word ‘incommensurable’ and a good deal of of the debate about Kuhn’s work has centered on this question, there is at least one version of the incommensurability thesis that casts doubt on the possibility of rational comparison between competing theories, namely the idea that our experience of the world is radically conditioned by our theories, which in turn depend upon the paradigm.85 For example, Kuhn observes that chemists after Dalton reported chemical compositions as ratios of integers rather than as decimals.86 And while the atomic theory accounted for much of the data available at that time, some experiments gave conflicting results. The conclusion drawn by Kuhn is rather radical:
Chemists could not, therefore, simply accept Dalton’s theory on the evidence, for much of that was still negative. Instead, even after accepting the theory, they still had to beat nature into line, a process which, in the event, took almost another generation. When it was done, even the percentage composition of well-known compounds was different. The data themselves had changed. That is the last of the senses in which we may want to say that after a revolution scientists work in a different world.
(Kuhn 1970, p. 135)
But what exactly does Kuhn mean by ‘they still had to beat nature into line’? Is he suggesting that chemists after Dalton manipulated their data in order to make them agree with the atomic hypothesis, and that their successors keep on doing so today? And that the atomic hypothesis is false? Obviously, this is not what Kuhn thinks, but at the very least it is fair to say that he has expressed himself in an ambiguous way.87 It is likely that the measurements of chemical compositions available in the nineteenth century were rather imprecise, and it is possible that the experimenters were so strongly influenced by the atomic theory that they considered it better confirmed than it actually was. Nevertheless, we have today so much evidence in favour of atomism (much of which is independent of chemistry) that it has become irrational to doubt it.
Of course, historians have the perfect right to say that this is not what interests them: their aim is to understand what happened when the change of paradigm occurred.88 And it is interesting to see to what extent that change was based on solid empirical arguments or on extrascientific beliefs such as sun worship. In an extreme case, a correct change of paradigm may even have occurred, by fortunate accident, for completely irrational reasons. This would in no way alter the fact that the theory originally adopted for faulty reasons is today empirically established beyond any reasonable doubt. Furthermore, changes of paradigm, at least in most cases since the birth of modern science, have not occurred for completely irrational reasons. The writings of Galileo or Harvey, for instance, contain many empirical arguments and they are by no means all wrong. There is always, to be sure, a complex mixture of good and bad reasons that lead to the emergence of a new theory, and scientists’ adherence to the new paradigm may very well have taken place before the empirical evidence became totally convincing. This is not at all surprising: scientists must try to guess, as best they can, which paths to follow – life is, after all, short – and provisional decisions must often be taken in the absence of sufficient empirical evidence. This does not undermine the rationality of the scientific enterprise, but it does contribute to making the history of science so fascinating.
The basic problem is that there are, as the philosopher of science Tim Maudlin has eloquently pointed out, two Kuhns – a moderate Kuhn and his immoderate brother – jostling elbows throughout the pages of The Structure of Scientific Revolutions. The moderate Kuhn admits that the scientific debates of the past were settled correctly, but emphasizes that the evidence available at the time was weaker than is generally thought and that non-scientific considerations played a role. We have no objection of principle to the moderate Kuhn, and we leave to historians the task of investigating the extent to which these ideas are correct in concrete situations.89 By contrast, the immoderate Kuhn – who became, perhaps involuntarily, one of the founding fathers of contemporary relativism – thinks that changes of paradigm are due principally to non-empirical factors and that, once accepted, they condition our perception of the world to such an extent that they can only be confirmed by our subsequent experiences. Maudlin eloquently refutes this idea:
If presented with a moon rock, Aristotle would experience it as a rock, and as an object with a tendency to fall. He could not fail to conclude that the material of which the moon is made is not fundamentally different from terrestrial material with respect to its natural motion.90 Similarly, ever better telescopes revealed more clearly the phases of Venus, irrespective of one’s preferred cosmology,91 and even Ptolemy would have remarked the apparent rotation of a Foucault pendulum.92 The sense in which one’s paradigm may influence one’s experience of the world cannot be so strong as to guarantee that one’s experience will always accord with one’s theories, else the need to revise theories would never arise.
(Maudlin 1996, p. 442)93
Thus, while it is true that scientific experiments do not provide their own interpretation, it is also true that the theory does not determine the perception of the results.
The second objection against the radical version of Kuhn’s history of science – an objection we shall also use later against the ‘strong programme’ in the sociology of science – is that of self-refutation. Research in history, and in particular in the history of science, employs methods that are not radically different from those used in the natural sciences: studying documents, drawing the most rational inferences, making inductions based on the available data, and so forth. If arguments of this type in physics or biology did not allow us to arrive at reasonably reliable conclusions, what reason would there be to trust them in history? Why speak in a r
ealist mode about historical categories, such as paradigms, if it is an illusion to speak in a realist mode about scientific concepts (which are in fact much more precisely defined) such as electrons or DNA?94
But one may go further. It is natural to introduce a hierarchy in the degree of credence accorded to different theories, depending on the quantity and quality of the evidence supporting them.95 Every scientist – indeed, every human being – proceeds in this way and grants a higher subjective probability to the best-established theories (for instance, the evolution of species or the existence of atoms) and a lower subjective probability to more speculative theories (such as detailed theories of quantum gravity). The same reasoning applies when comparing theories in natural science with those in history or sociology. For example, the evidence of the Earth’s rotation is vastly stronger than anything Kuhn could put forward in support of his historical theories. This does not mean, of course, that physicists are more clever than historians or that they use better methods, but simply that they deal with less complex problems, involving a smaller number of variables which, moreover, are easier to measure and to control. It is impossible to avoid introducing such a hierarchy in our beliefs, and this hierarchy implies that there is no conceivable argument based on the Kuhnian view of history that could give succor to those sociologists or philosophers who wish to challenge, in a blanket way, the reliability of scientific results.
Feyerabend: ‘Anything goes’
Another famous philosopher who is often quoted in discussions about relativism is Paul Feyerabend. Let us begin by acknowledging that Feyerabend is a complicated character. His personal and political attitudes have earned him a fair amount of sympathy, and his criticisms of attempts at codifying scientific practice are often justified. Moreover, despite the title of one of his books, Farewell to Reason, he never became entirely and openly irrationalist; towards the end of his life he started to distance himself (or so it seems) from the relativist and anti-scientific attitudes of some of his followers.96 Nevertheless, Feyerabend’s writings contain numerous ambiguous or confused statements, which sometimes end in violent attacks against modern science: attacks which are simultaneously philosophical, historical and political, and in which judgments of fact are mixed with judgments of value.97