REVIEW. Patrick Enfield

Transcription

1 Brit. J. Phil. Sci. 59 (2008), REVIEW P. KYLE STANFORD Exceeding Our Grasp: Science, History, and the Problem of Unconceived Alternatives New York: Oxford University Press, 2006, (hardback) ISBN: Patrick Enfield [A]t the end of the day, the only evidence for the truth of scientific theories is the evidence that scientists [sic] use, and the only positive arguments for scientific realism are the arguments that scientists make. Peter Lipton ([2001], p. 353) Stanford s argument in this book involves a novel combination of underdetermination with the pessimistic induction, which he claims provides strong reasons for doubt about the truth of all scientific theories. In the first section of this review, I outline the overall structure of his argument, and try to show that his case against scientific realism is only as strong as the standard historical reasons for doubt about scientific truth, associated with what he calls the classical pessimistic induction (pp. 9, 19). Towards the end of that section, I offer quotations showing that Popper anticipated both the main consideration that feeds into Stanford s argument, and also the traditional realist response to such sceptical ideas, to the effect that superseded theories can continue to be accepted as at least approximately true, from the perspective of the theories that succeed them. In the rest of this review, I locate Stanford s views about the traditional or classical pessimistic induction within a broader historical context. Stanford introduces the pessimistic induction first via a quotation from Poincaré, followed by citations to Laudan s ([1981]) version of the argument (p. 7). But in the immediately ensuing elaboration of the idea, Stanford departs crucially from both the spirit and the letter of Laudan s presentation (pp. 7 11). Specifically, Stanford takes literally talk of a pessimistic induction, the idea evidently being that Laudan was committed to a historical generalisation on which he based his scepticism or pessimism. It is not entirely clear what generalisation this is supposed to be, but because Laudan discussed two questions that should be considered separately, it might be supposed that there are pessimistic generalisations associated with either or both of these issues. C The Author (2008). Published by Oxford University Press on behalf of British Society for the Philosophy of Science. All rights reserved. doi: /bjps/axn042 For Permissions, please journals.permissions@oxfordjournals.org

2 882 Review It is easy to clarify the two different issues. Consider the development of a branch of science as a sequence of theories well ordered by succession: T 1,T 2, T 3,...T n, i.e., the theory that we hold now. (Stanford offers some illustrations of such sequences: pp ) In the earlier sections of his paper, Laudan considered whether all past theories that were sufficiently successful could still be considered true from the perspective of the theory we currently accept as the best explanation of the original domain of phenomena. Many of the theories that Laudan offered as counter-examples to such an idea were from relatively early stages in the development of the branch of science to which they belong. So Laudan was considering the relations between an earlier theory T i and the current theory T n, where perhaps several theories were intermediate between the early theory and the contemporary theory we compare it with. Most of the realists whose responses to Laudan are considered by Stanford concentrated their remarks on this issue. Out of Laudan s original list of a dozen once successful theories no longer considered approximately true, because they were non-referential, attention has come to be focussed particularly on three cases concerning which realists have seen the need to offer more detailed historical accounts. Stanford devotes most attention in his later chapters to these three cases: phlogiston, caloric, and ether theories (Chapters 6 and 7). However, as the formulation just offered suggests, Stanford is mistaken in reading Laudan as offering a general reason for pessimism. Rather, Laudan claimed that realists are committed to a generalisation, and argues by counter-example against the general historical assumptions which realists seem to need to defend. In this review, I shall not consider in any detail Stanford s arguments for scepticism developed in relation to Laudan s historical examples based on loss of fundamental ontology, or reference failure. Rather, I wish to offer an overview of the earlier literature on theory change to see what lessons we can learn from it. This literature, before 1981, raised the other issue discussed by Laudan. In the section after he offered his infamous list, Laudan considered the relations between a theory and its immediate successor: between T i and T i+1.asihave already alluded, Popper held the general view that any new theory would include its immediate predecessor as a good approximation. He maintained such an account, specifically, in relation to the example of Einstein s two theories of relativity. In mitigation of his emphasis on revolutions, and despite his official rhetoric concerning the falsification or refutation of earlier theories, Popper held that Einstein s theories contained Newton s theories of motion and gravitation as limiting cases. I shall sketch how Popper seems to have taken this view over from what physicists, including Einstein himself, said about this and other examples of retention of equations from one theory into the next. However, in 1962, Kuhn and Feyerabend offered several reasons to doubt whether any theory could really survive as an approximation to its successor. Kuhn developed his sceptical arguments specifically in connection with the example that

3 The British Journal for the Philosophy of Science 883 so concerned Popper, namely Einstein s revolution. Without going into all the details of these arguments, we shall see that Kuhn s view that Newton s theory is just false anticipates the way Stanford reads Laudan as committed to a similarly general denial that any theory survives as an approximation to its successor. But Laudan was not arguing in this general way that no theory can survive as an approximation to its successor. Rather, Laudan took account of both earlier general claims about theory succession offered by Popper, and later revised suggestions offered by realists responding to Kuhn and Feyerabend in the 1960s and 1970s. Laudan contended that there are enough counter-examples to the generalisations to which realists seem committed to undermine the arguments often given to support belief in the truth of science. There is, thus, a fundamental confusion in the literature, which Stanford takes over, as to whether Laudan is offering pessimistic generalisations of his own, rather than giving counter-examples to the general claims to which realists seem to be committed. In this review I hope to clarify this issue, and also to show that traditional ways of looking at theory change found in the earlier literature offer us considerable resources for resisting the sweeping scepticism suggested by Laudan and Stanford s examples and arguments. 1 Stanford s Overall Argument Recall that the classical pessimistic induction notes simply that past successful theories have turned out to be false and suggests that we have no reason to think that present successful theories will not suffer the same fate. By contrast, I propose what I will call the new induction over the history of science: that we have, throughout the history of scientific inquiry and in virtually every scientific field, repeatedly occupied an epistemic position in which we could conceive of only one or a few theories that were well confirmed by the available evidence, while subsequent inquiry would routinely (if not invariably) reveal further, radically distinct alternatives as well confirmed by the previously available evidence as those we were inclined to accept on the strength of that evidence. (p. 19) Stanford s argument is a combination of underdetermination with the pessimistic induction. That is, there are two traditional arguments against scientific realism, and, while neither succeeds on its own, in combination they provide the basis for a new and decisive reason not to believe what scientists tell us (p. 9). Stanford might deny it, but it seems to me that what really does the work in this argument are historical claims associated with the pessimistic induction, to the effect that past theories do later turn out to be false. 1 And, 1 Perhaps I should explain the scare quotes. My point, to be developed in the rest of this review, is that the issue of whether Newton s theory approximates to Einstein s seems to be a scientific question, determined by technical results such as the classical limits. Of course, such scientific

4 884 Review certainly, the formulation of the kind of anti-realism or instrumentalism his argument leads to is familiar from discussions of the pessimistic induction. We...answer the crucial question Should we trust or believe what our own best scientific theories tell us about what things are like in otherwise inaccessible domains of the natural world? with a resounding no (pp ). We should reject the idea that even the most fundamental claims of theoretical science will persist indefinitely into the future as part of the best collection of conceptual tools we have for engaging the natural world. (p. 211) According to some versions of the pessimistic induction, each new theory T i+1 implies the falsehood of its immediate predecessor T i. But, of course, realists do not claim that T i is TRUE, only that it is approximately true. So, from now on, let us take the anti-realist claim T is false to mean T is not even approximately true. Stanford s main point seems not to be that each new theory implies the falsehood of its predecessor. Rather, the fact of theory change proves the existence of a rival theory equally well confirmed by the evidence that originally supported T i. That is, the later theory, T i+1,whichin fact did supersede T i, is at least equally well confirmed by the evidence that supported its predecessor. The two theories are not empirical equivalents, as some versions of underdetermination have either required, or tried to prove the existence of. Indeed, the earlier theory may have been refuted by some evidence that became available later in its development. Moreover, it may not be the case that the later theory can explain everything that the old theory could, and more. Stanford considers possible objections deriving from this latter suggestion of the existence of Kuhn losses (pp. 21 2). Stanford s answer seems to be analogous to the case of two students, one of whose grades were better overall, even though the other did better in one or two particular exams that they both entered. I don t think this response is adequate as it stands, but nor do I think that this issue of Kuhn losses is fundamental to questions about realist versus sceptical views of scientific progress. Stanford introduces this argument properly in the third section of Chapter 1 (the extended quotation with which we began this section of the review is three pages into this section of the book). The philosophical stage setting leading up to its introduction involves an initial sketch of the pessimistic induction (pp. 7 11), and a brief but useful overview of the underdetermination literature (pp. 8, 11 9). I have never taken underdetermination very seriously, partly because many of the rival theories, or alleged empirical equivalents, suggested, givens can be taken in more than one way. But, as I shall suggest, Kuhn s view that Newton s laws were just false seems to arise in the context of philosophical confusions about the logical connections between the concepts of approximation and is deducible from. Other anti-realists such as Stanford, Lyons and van Fraassen, who claim that Newton s theory is false, do not mention the classical limits, or any secondary scientific or philosophical literature bearing on these questions (see footnote 3). Thus, the view that Newton s theory is false seems not to be a fact. But even if it were a fact, it would not be simply or primarily a historical fact.

5 The British Journal for the Philosophy of Science 885 seem not to be genuinely different alternative scientific theories. As Stanford notes, there is a cogent objection to underdetermination as it is often argued. It seems perfectly sensible for critics of underdetermination to insist that any such equally well-confirmed alternatives actually be produced before we take them seriously or suspend judgement about the truth of the contemporary theories whose achievements they are supposedly able to replicate [as suggested by Kitcher, Leplin, and Achinstein] (p. 11). Or, as Jerry Maguire might have put it: Show me the theory! Stanford s trick is to say that, in the case of a past theory, to which one or more successor theories have since been accepted, choice of that theory is underdetermined by the original evidence, which will support these successors equally well. But is this much of an objection, in itself, to realism? Stanford tries to suggest that it might be in further philosophical stage setting in Chapter 2. The argument seems to be that the later emergence of such an alternative, previously unconsidered, undermines the assumption implicit in scientific arguments for believing T i based on elimination of all potential alternative explanations. [T]he historical record suggests that in science we are typically unable to exhaust the space of likely, plausible, or reasonable candidate theoretical explanations for a given set of phenomena before proceeding to eliminate all but a single contender, but this is just what would be required for such eliminative inferences to be reliable (p. 29). But it is not clear that, of itself, this is much of an argument against realism, unless it is also true that the new theory does imply the falsehood of the earlier one. So it would be an objection to Stanford s argument based on the inability of scientists to consider all relevant alternatives, if it were, sometimes, often, or always, possible to present the earlier theory as a good approximation to the account offered by the later theory. And, although Stanford might motivate consideration of this possibility differently from the way I have set out the objection, it is something he does quite clearly see the need to address. So, at the end of the first chapter, when outlining the form of the rest of the book, he promises that in Chapters 6 and 7 he will take up a variety of recent realist replies to the pessimistic induction that, if successful, would threaten to dispose of the problem of unconceived alternatives as well (pp. 23 4, see also pp ). He mentions here the names of the realists responding to Laudan that he will consider. In chronological order of principal relevant publication, these are Hardin and Rosenberg, Worrall, Kitcher, Leplin, and Psillos (p. 24, see also p. 142). So I think that what really does the work in Stanford s argument is the kind of historical claim associated with the pessimistic induction to the effect that all past theories have since turned out to be false, and all new theories imply the falsehood of their immediate predecessors. But Stanford defers consideration of these crucial issues because he thinks that his new argument form is substantially different from the old one. But, like the pessimistic induction, its defence

6 886 Review turns on detailed appeals to the history of science. So, three chapters offer detailed case studies in nineteenth-century biology. Devoting one chapter per biologist, Stanford explores the views about generation and heredity presented by Darwin, Galton, and Weismann (Chapters 3 5). The trouble with all this is that the historical phenomenon on which Stanford s argument turns, while not in any sense a trivial fact, is so well known as not to require detailed historical documentation (as Stanford partially concedes, pp. 51 2). As far as I can see, Stanford s argument is, in all essentials, anticipated by a way in which Popper later reconstructed the motivation for his primary theses in the philosophy of science: [My main point] is that all the observational evidence which may be claimed to support Newton s theory may also be claimed to support Einstein s very different theory. This shows, decisively, that we were simply mistaken when we thought that Newton s theory could be said to be established, or inductively proved, on the basis of the evidence. It further showed that no theory can be claimed to be inductively established. For there could be no more impressive agreement between theory and observational evidence than we had in the case of Newton s theory. If even this could not inductively establish the theory, then clearly nothing whatever could (Popper, in Magee [1971], pp. 70 1). A similar version of the argument is found in Feyerabend, who probably got it from Popper in any case. [O]ne and the same set of observational data is compatible with very different and mutually inconsistent theories (Feyerabend [1962], Section 4; emphasis in original). The reason that Newton failed to anticipate general relativity is obvious, given that non-euclidean geometries, Riemannian spaces, and general tensor calculus took another century or two to be developed. But Popper did not take such facts to undermine the successive approximation account. [It is] always possible, even for the best established scientific theory, to be superseded by a better one. All we [can] say [is] that the better one would have to contain any successful and well-tested predecessor as an approximation. In this way it would also explain why its predecessor had been successful (Popper, in Magee [1971], p. 71). So in the end, the really crucial question is whether some past theories can be construed as surviving as approximations in their successors, which issue is put back until we are two-thirds of the way into the book. And on this issue I fear that Stanford is in danger of adding to the prevalent confusions in the literature. For Stanford takes over historical claims associated with Laudan and the pessimistic induction that are so patently false that even Laudan knew better than to suggest them.

7 The British Journal for the Philosophy of Science Laudan, and the Pessimistic Induction Before Him There is a fundamental confusion in the literature about the logical form of the historical claims to which Laudan is committed. And this in turn is associated with the fundamental but generally neglected fact that Laudan is not offering considerations that undermine beliefs in science directly. Rather, Laudan claims that the onus is on realists to offer arguments for their position, and then offers historical counter-examples to the generalisations that realists either have actually stated, or that appear to be implicit in the philosophical arguments they have offered. It is hard to see how anyone could read Laudan ([1981]) as doing anything other than arguing by counter-examples. But, in fact, later realists have increasingly come to reinterpret Laudan s argument as a pessimistic (meta) induction, so that he is understood as offering generalisations in support of his pessimism. However, Hardin and Rosenberg ([1982], pp. 604, 608, and 610) were already quite clear that Laudan was arguing by counterexample(s) to realist generalisations. The issue was also brought out by Enfield ([1991], especially pp ). Hardin and Rosenberg, and Enfield, were, however, considering, respectively, two different issues tackled in Laudan s paper. The two different issues are those we clarified at the beginning of this review. Hardin and Rosenberg concentrate on the issue of whether all successful past theories T i are still considered valid approximations from the perspective of the currently accepted successor theory T n. Enfield ([1991]) focused more on the line of development of ideas from Popper to Kuhn and then Putnam in relation to methodology, progress, and the relations between a theory and its immediate successor: T i and T i+1.towards the end of this review I offer a more detailed analysis of the structure of Laudan s paper that makes nonsense of the idea that he was developing a pessimistic induction. Stanford (pp ) later offers a more thorough interpretation of Laudan s argument, but misses all of the points I document here. There can also be no doubt that Stanford attributes the pessimistic induction to Laudan, since it is the latter s works that are cited as primary sources for the argument (pp. 7, 142 6). How and why did later realists come to reinterpret Laudan as offering a pessimistic induction? This review, and (Enfield [1991]), suggest that part of the explanation derives from the fact that some such argument had been in the air between 1962 and 1981, because Kuhn and Feyerabend were indeed committed to a generally sceptical or pessimistic view of relations between successive theories. But later realists have wanted to discuss Laudan s version of the historical objection to realism without taking into account the details of Kuhn and Feyerabend s arguments, or Putnam and Boyd s role in casting realism in the precise form that Laudan considered so dubious. So, somehow, Laudan came to be read as committed to the general scepticism previously

8 888 Review expressed by Kuhn and Feyerabend, even though his arguments would not in fact take him down that path. 2 But Stanford explicitly commits, early on, to the standard idea that Laudan s argument involves a pessimism that is itself based on historical generalisations (pp. 7 11). The argument [the pessimistic induction] offers is so very straightforward (p. 9). It relies on an extremely simple form of reasoning called enumerative induction...inductive generalization or the straight rule... (p. 10). Realists cannot deny that all past theories have turned out false. Rather, their only recourse is to point to some feature of current theories that distinguishes them from all the past theories that have since been refuted. This might invalidate the pessimistic induction s projection from the fate of the latter to the prospects for the former. The lingering whiff of ad-hoc-ery or special pleading cannot dispel the fact that changed circumstances, conditions, or characteristics sometimes really do make a difference to the legitimacy of projecting an inductive generalization into the future (p. 10, emphasis added; see also pp. 11, and 43 4). If it were true that every new theory implies the falsehood of its predecessor, then this would indeed be a devastating objection to realism. On the contrary however, scientists and realists have often in the past claimed that successive approximation is a prevalent, if not universal, pattern in the historical development of the sciences. I quoted Popper holding such a view in In fact, Popper first suggested such ideas from as early as the 1930s. Although Popper ([1934], Section 85) here offered no scientific or historical examples, or citations to relevant secondary literature, there is circumstantial evidence that he got the idea from the things that physicists said about the classical limits. Certainly, we know that the two then recent relativity revolutions were central to Popper s early philosophical motivations, so he was presumably aware of the fact that 2 It is not quite true that all later realists interpreting Laudan as offering a pessimistic induction ignore Kuhn and Feyerabend completely. Psillos did not offer any account of Kuhn and Feyerabend s role in articulating the historical scepticism to which Putnam, Boyd and others were responding in the 1970s. But Worrall ([1989], p. 161, note 25) did provide a brief discussion of the Putnam Boyd argument for realism, and also (p. 147, note 10) mentioned that Kuhn and Feyerabend developed pessimistic views in connection with explanatory losses. Kitcher ([1993], p. 136, note 13) notes that reference to a pessimistic meta-induction appears in Putnam s writings where it of course was being applied to Kuhn and Feyerabend s ideas but then suggested that Laudan s version of the argument is the most sophisticated form in which the reasons for scepticism merit consideration. Kitcher ([1993], pp , 173 7) does, however, also offer a rather more detailed exegesis and critique of some of Kuhn s reasons for scepticism about realist progress. Stanford s (pp. 21 2; p. 163, note 20) understanding of Kuhn s relevance to these debates seems to derive from Kitcher s presentation. But the discussion of Kuhn in (Enfield [1991], especially pp ) brings out the crucial role, both of general considerations to do with incompatibility versus deducibility between successive theories, and also the way that Kuhn elaborated such ideas in connection with the classical limits. In this review, I explain the influence of what physicists had said about the classical limits on Popper, and how Kuhn s contrary scepticism in this connection anticipates Stanford s view that Newton s theory is not even approximately true.

9 The British Journal for the Philosophy of Science 889 scientists found limiting case relations between the modern theories and the classical theories they superseded, and held the approximate truth of the past theories on this basis. Popper ([1963], p. 32) later cited Einstein to similar effect. Einstein had said: There could be no fairer destiny for any...theory than that it should point the way to a more comprehensive theory in which it lives on, as a limiting case. In context, in this popular book first published in 1916, Einstein was rebutting the view that new theories simply involve rejection of their predecessors. He suggested that general relativity retains special relativity as a limiting case in much the same way that Maxwell s theory retains Coulomb s law for the limiting case of static charges. This kind of interpretation of the classical limits, that they show that Newton s theories (of motion in general, and gravitational forces in particular) were special cases of, limiting cases of, or approximations to, Einstein s theories (of special and general relativity), was prevalent in the physics textbook literature throughout the twentieth century. Physicists often confine such views to the particular limiting case relations they exhibit, although they do, sometimes, also generalise, and suggest that these patterns are universal in the development of physics if not of all the sciences. Philosophers, like Popper, are a bit more inclined to generalise. I am aware of relatively few other philosophers before 1962 to have made similarly explicit general claims about how the sciences grow. But, it turns out that Popper s views, both about radical theory change, and about the fact that despite such changes, new theories contain their predecessors, were amongst the most important factors behind the emergence of the historical objection to realism in the works of Kuhn and Feyerabend. Kuhn s ([1962]) writings are crucial to understanding the course of subsequent discussions of radical theory change leading up to the form in which Laudan developed the historical objection to realism. For Kuhn argues, for the first time, against the prevalent view that the classical limits show that Newton s theory, although strictly false, remains, nevertheless, approximately true. Kuhn of course knew that physicists present what look like derivations of Newton s laws from Einstein s theory. But while such a derivation has, of course, explained why Newton s Laws ever seemed to work, it has not...shown Newton s Laws to be a limiting case of Einstein s (Kuhn [1962], p. 102). Kuhn was clearly committed to the account that Putnam ([1978], pp ) later attributed to positivist instrumentalists, that a new theory will merely enable recovery of the confirmed predictions of the earlier theory, but will not also imply the approximate truth of higher level laws of that theory. Kuhn s interpretation of the classical limits strains credibility, but his arguments developed in connection with them do not entirely make sense unless they are read together with the contemporaneous arguments of his then colleague, Feyerabend. It is clear from the first sentence of Feyerabend s ([1962]) paper that his primary objection is to logical empiricism, the formalist project,

10 890 Review and the analyses of inter-theory relations in theory reductions in terms of deductive entailment. It is also clear from the original version of the paper, but less so from the reprint in Feyerabend s collected papers, just how much one of his primary arguments was indebted to a later discussion by Popper. Popper ([1957]) considered the example Galileo + Kepler Newton. Popper expanded on the point, which he notes had previously been made by Duhem, that, because Kepler s laws and Newton s theory are incompatible, Kepler s laws can neither entail, nor be entailed by, Newton s theory. Despite this, Popper endorsed standard claims made by scientists, in relation to this and numerous similar examples, that the earlier theory can still be considered approximately true, or at least an approximation to the later theory (if the truth of this later theory itself is considered subject to legitimate philosophical doubt). Thus, Popper makes three points: 1. T is inconsistent with T. 2. T is not deducible from T (even in conjunction with additional conditions AC). 3. Laws in T can be approximations to what is deducible from T. Both Kuhn and Feyerabend appear to think that we should reject (3) in the light of (1) and (2). To show how and why they thought this, and exactly why they were wrong, would take us too far away from the immediate task of understanding the book under review. Nevertheless, this discussion of the pre-history of the pessimistic induction itself is of considerable relevance. First, as I shall shortly explain, Laudan s arguments against realism take account of revised historical claims about scientific progress that appeared in the wake of Kuhn and Feyerabend s earlier historical scepticism. But, even more strikingly, Stanford derives from his discussion of Laudan and the pessimistic induction, four fundamental claims that, in fact, Laudan would reject, but Kuhn anticipated. 1. The reasons for thinking that past theories are false are general (pp. 10 1). 2. These include the fact that an earlier and a later theory are inconsistent (p. 146). 3. Newton s theory is just false (pp. 7 9, 204 6). 4. But it can still be used, because it makes sufficiently good predictions for many situations (pp. 7 9, 204 6). Because Kuhn and Stanford think that no theory survives as an approximation to its successor, they both have to deny that Newton s theory really survives as an approximation to, or a limiting case of, Einstein s theory. Einstein s theory can be accepted only with the recognition that Newton s was wrong

11 The British Journal for the Philosophy of Science 891 (Kuhn, p. 98). Newtonian mechanics is just plain false, and radically so (Stanford, p. 9). A derivation of the classical limits has, of course, explained why Newton s Laws ever seemed to work (Kuhn, p. 102). Although this sometimes invites the counterclaim that Newtonian mechanics itself is approximately true, this can only mean that its empirical predictions approximate those of its successors across a wide range of contexts (Stanford, p. 9). In other words, Newton s theory survives, although it is just false, because it makes nearly accurate predictions for many systems (Stanford, pp. 7 9; see also pp ). But Laudan (p. 129) did not deny that Newton s laws are limiting cases of Einstein s. It can, of course, be shown that some laws of classical mechanics are limiting cases of relativistic mechanics. Laudan did not deny this because he was not claiming that limiting case relations never occur between successive theories. So his argument is not general, as in point (1). Argument (2) is only needed if you want to claim that no theory ever approximates to its successor. The mere fact that new theories imply revisions and corrections of their predecessors except in trivial cases of deductive subsumption that are not genuine examples of revolutions does not, Laudan appreciates, prove failure of approximation. Laudan s argument was restricted, primarily, to cases where there is loss of fundamental ontology. Laudan (p. 129) appealed to relativity not as showing that Newton s theory is false, but rather as showing that ether theories were false. But there are other laws and general assertions made by the classical theory (e.g. claims about the density and fine structure of the ether, general laws about the character of the interaction between ether and matter, models and mechanisms detailing the compressibility of the ether) which could not conceivably be limiting cases of modern mechanics. Stanford s reconstruction of Laudan as offering a pessimistic induction simply fails to take account of three major sections of Laudan s paper. This shift has occurred in the previous realist literature that has increasingly taken Laudan s work as the most sophisticated version of the historical objection to realism. But this fails to recognise the fact that Laudan is just arguing, in turn, against what some previous philosophers had said in the course of complex philosophical arguments for realism. I take the conclusion from this entire body of literature that you should not believe what philosophers say about science. Since science is true, philosophers who argue that it isn t invariably perpetuate mistakes or fallacies. But the realist philosophers who develop arguments for realism are equally in error. It has been a mistake from the outset to argue for truth claims about science in the ways that philosophers have been doing for the last 50 years. The only sensible way to find out whether a scientific claim is true is by looking at the original scientific evidence previously offered by scientists. We can, of course, find scientific evidence wanting. But when we believe scientific claims to be true, it is because of this evidence, not on the basis of some kind of abductive argument for realism as an overarching empirical hypothesis. Or, as Putnam

12 892 Review ([1978], p. 19) had put it: [R]ealism must itself be an over-arching scientific hypothesis.and...realism is an empirical hypothesis. But Laudan s paper only makes sense when read in the context of such philosophical arguments for realism. For, in the second section (pp ), Laudan attributes two elaborate abductive arguments to realists (p. 109). Simplifying for brevity, both these arguments presuppose certain historical generalisations about science. The first presupposes that if all successful theories are approximately true, then past successful theories should still be capable of being considered approximately true in the light of current theory. Laudan devotes three sections to considering this argument, and the variant with if all replaced by if all and only (pp ). The famous list of once successful theories now considered false, because they are non-referential, appears in the third of these sections (pp ). Laudan s point is not that all theories have since turned out false, but that there are counter-examples to realist generalisations. But note that this issue is different from the one that dominated discussions in the 1960s and 1970s, concerning the relation between a theory and its immediate successor. Crystalline spheres are not the immediate predecessor of the contemporary account of the solar system given by general relativity. I don t know how far you have to go back, in the sequence Ptolemy Copernicus Kepler Newton Einstein to find someone who really believed in crystalline spheres. So in these earlier sections of his paper, Laudan is raising the issue of relations between T i and T n. The second argument is addressed in Section 6 of Laudan s paper (pp ). The most obvious source for this idea is the argument that Putnam ([1978], pp ) attributed to Boyd. Roughly, if scientists are realists, and believe theories to be approximately true, then, when they feel the need to look for a new theory, they will seek such new theories as retain their predecessors as approximations. They do adopt such methodological strategies, and it works, so realism in turn gets justified by this abductive argument. Very few philosophers have ever argued for realism in quite this way. What is true, however, is that in earlier discussions of theory change and progress, many more or less general claims had been made both about the relations between T i and T i+1, and also about the heuristic role of methodological strategies similar to that which Putnam Boyd describe. Post and Koertge made similar claims about the General Correspondence Principle, and its heuristic role. But they talked less frequently of truth and approximate truth, and more of correspondences between successive theories. They were also less obviously committed to anything like the kind of argument that Putnam Boyd presented. Again, Laudan s point is not to deny that these things ever happen. Rather, the point is that realists have asserted, explicitly, or their arguments appear to commit them implicitly to, generalisations. And there are counter-examples to such generalisations.

13 The British Journal for the Philosophy of Science 893 The prevalent reconstruction of Laudan s argument as a pessimistic induction ignores the role of these considerations in his paper. Similarly, Laudan s claim in his final section before the conclusions (Section 7, pp ) only makes sense if the objection is to a philosophical argument that realists have offered for belief in science. Realists can only be guilty of begging the question or making a Petitio Principii if they are offering an argument, at the philosophical level, that involves the abductive argument form that anti-realists already doubt is valid at the scientific level. In his discussion of realist responses to the pessimistic induction, Stanford devotes most attention to what realists have said about three of the disputed cases on Laudan s original list of a dozen false theories: phlogiston, caloric, and ether theories. I do not want to suggest that there are no problems for realists to address here. What I want to draw attention to is the way the literature has shifted from the original examples from physics, on which physicists throughout the twentieth century based their claims of successive approximation, to these disputed cases. But the end result is that contemporary anti-realists, such as Stanford, seem to be not only insufficiently aware of the technical results in physics, but also not to appreciate the crucial role of these examples in discussions of realism versus scepticism in all the literature before Laudan s paper. 3 That is, before Laudan, most attention focussed on relations between successive theories: between T i and T i+1. And Laudan discussed this literature at some length in a later section of his paper (originally numbered 6, pp in the reprint). Laudan offered a useful table of seven different claims that realists have made about such relations, with fairly plausible attributions of particular views to specific cited authors. Laudan s ([1981], p. 125) full list of philosophers offering such views is: Whewell (1840s); Popper ([1934]); Margenau (1950s); Sellars, Sklar, and Fine (1960s); and McMullin, Kordig, Post, Koertge, Boyd, Putnam, Krajewski, and Watkins (1970s). Amongst the realists that Stanford considers, Worrall s structural realism is most similar to the kind of claims about relations between successive theories found in the literature before But such claims in the 1960s and 1970s, in at least some important respects, echo what had been said by Popper and the physicists before Kuhn and Feyerabend challenged the received wisdom on such issues. In his discussion of Worrall, Stanford (pp ) illegitimately 3 Although van Fraassen does not develop a pessimistic induction type argument anywhere, he has claimed on more than one occasion that Newton s theory is just false. (van Fraassen [1985], p. 267; [1989], p. 145). Lyons ([2006], pp ), in the context of a critique of realist responses to the pessimistic induction, claims that Newton s theory is just false. His interpretation is similar to the Kuhn Stanford account. Like Stanford, neither of these other authors cites any pertinent secondary literature. (Lyons appeal to what Worrall had said about the example is inadequate, and seems not to appreciate Worrall s later endorsement of the correspondence principle in this context.) So, they do not mention the classical limits as scientific givens. Nor are they aware of the centrality of this example in earlier philosophical discussions of theory change versus realist progress.

14 894 Review changes the subject here, offering Galton s Ancestral Law of Inheritance as an equation that has not survived in any meaningful way. But since Worrall ([1989], p. 160) only claimed that the correspondence principle rule applied in the history of physics, it simply does not follow that physicists were wrong in what they said about the examples they were talking about. I also think that Worrall is wrong in saying that it is only the equations that survive, and not the ontology. Because, contrary to the impression given by the examples that Laudan and Stanford prefer to discuss, there are other cases in the history of science where new theories tell us more about the old things, rather than just denying their reality. If one takes off the blinkers that all these philosophical arguments for and against realism put on us, there is an immediate and obvious example in the way quantum theories tell us more about what atoms are made of, not that they don t exist. And if you think a bit more, you will realise that there are many other such examples. We are not restricted, in responding to Laudan, to countering what he said about his choice of examples. There are other cases in the history of science, and it is far from obvious that all the people who discussed these examples in the earlier literature are so radically mistaken as Stanford s arguments against realism suggest. References Enfield, P. [1991]: Realism, Empiricism and Scientific Revolutions, Philosophy of Science, 58, pp Feyerabend, P. K. [1962]: Explanation, Reduction, and Empiricism, in L. Sklar (ed.), 2000, Theory Reduction and Theory Change, New York: Garland Publishing, pp Hardin, C. and Rosenberg, A. [1982]: In Defence of Convergent Realism, Philosophy of Science, 49, pp Kitcher, P. [1993]: The Advancement of Science, Oxford: Oxford University Press. Kuhn, T. S. [1962]: The Structure of Scientific Revolutions, Chicago: University of Chicago Press. Laudan, L. [1981]: A Confutation of Convergent Realism, in D. Papineau (ed.), 1996, The Philosophy of Science, Oxford: Oxford University Press, pp Lipton, P. [2001]: Quests of a Realist (Review symposium on Psillos s Scientific Realism), Metascience, 10, pp Lyons, T. D. [2006]: Scientific Realism and the Stratagema de Divide et Impera, British Journal for the Philosophy of Science, 57, pp Magee, B. [1971]: Modern British Philosophy, London: Secker & Warburg. Popper, K. R. [1934]: The Logic of Scientific Discovery, 1959 translation of German original, London: Hutchinson. Popper, K. R. [1957]: The Aim of Science, in K. R. Popper, 1972, Objective Knowledge: An Evolutionary Approach, Oxford: Oxford University Press, pp

15 The British Journal for the Philosophy of Science 895 Popper, K. R. [1963]: Conjectures and Refutations: The Growth of Scientific Knowledge, London: Routledge & Kegan Paul. Putnam, H. [1978]: Meaning and the Moral Sciences, London: Routledge & Kegan Paul. van Fraassen, B. C. [1985]: Empiricism in the Philosophy of Science, in P. M. Churchland and C. A. Hooker (eds), Images of Science: Essays on Realism and Empiricism, Chicago: Chicago University Press, pp van Fraassen, B. C. [1989]: Laws and Symmetry, Oxford: Oxford University Press. Worrall, J. [1989]: Structural Realism: The Best of Both Worlds?, in D. Papineau (ed.), 1996, The Philosophy of Science, Oxford: Oxford University Press, pp