J Gen Philos Sci (2011) 42:295–316 DOI 10.1007/s10838-011-9167-8 ARTICLE
From Standard Scientific Realism and Structural Realism to Best Current Theory Realism Gerald D. Doppelt
Published online: 25 August 2011 Ó Springer Science+Business Media B.V. 2011
Abstract I defend a realist commitment to the truth of our most empirically successful current scientific theories—on the ground that it provides the best explanation of their success and the success of their falsified predecessors. I argue that this Best Current Theory Realism (BCTR) is superior to preservative realism (PR) and the structural realism (SR). I show that PR and SR rest on the implausible assumption that the success of outdated theories requires the realist to hold that these theories possessed truthful components. PR is undone by the fact that past theories succeeded even though their ontological claims about unobservables are false. SR backpeddles to argue that the realist is only committed to the truth about the structure of relations implied by the outdated theory, in order to explain its success. I argue that the structural component of theories is too bare-bones thin to explain the predictive/explanatory success of outdated theories. I conclude that BCTR can meet these objections to PR and SR, and also overcome the pessimistic meta-induction. Keywords Preservative realism Structural realism Inference-to-the-best-explanation The empirical success of science Pessimistic meta-induction Best-current-theory realism
This essay proposes six tests that should be met by any sound version of ‘‘inference-to-thebest-explanation’’ (IBE) scientific realism. I give arguments to defend my most controversial proposed test—namely, that scientific realists should be able to explain the explanatory success of theories, as well as their success with novel predictions. My essay then explains and evaluates the structural realism defended by Worrall, Carrier, Ladyman, and others—who argue effectively that structural realism overcomes the inability of standard preservative realists (e.g., Boyd and Psillos) to meet the test of providing plausible realist accounts of successful-but-false theories and thus the test of overcoming the pessimistic meta-induction. In turn, I argue that structural realists’ bare-bones notion of successful theories’ accurate models of unobservable structures of relations (e.g., Fresnel’s G. D. Doppelt (&) Department of Philosophy, 0119, University of San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0119, USA e-mail:
[email protected]
123
296
G. D. Doppelt
or Newton’s mathematical laws) is too thin and vacuous to explain the explanatory and predictive success of these theories. I examine structural realists’ account of the success of phlogiston theory and argue that the whole notion of accurate structure to which they appeal assumes the truth of current chemistry’s ontological claims concerning oxygenation and reduction. This assumption violates structuralists’ key claim that the truth of such theoretical posits cannot be known or used to explain the success of scientific theories. These arguments set the stage for my own version of IBE realism—‘‘Best Current Theory Realism’’ (BCTR). BCTR holds that the best realist explanation of the empirical success of both (1) superceded theories and (2) our best current theories, is one which is committed exclusively to the truth of our best current theories, and the epistemic virtues and standards they alone maximize. I then argue that BCTR can best meet the six tests proposed above. It can explain the empirical success (and failures) of superceded theories without the untenable assumption shared by standard and structural realists that some of these theories’ claims concerning unobservables (entities, processes, or structures of relations) were true and necessarily preserved by our best current theories and adequate to explain their empirical success.
1 Introduction to the Realism Debate By far, the most influential and powerful scientific realism remains ‘‘inference-to-the-bestexplanation’’ (or IBE) realism, which in its standard form holds that the only or best explanation of the empirical success of scientific theories is the realist claim that these theories are approximately true and that their key theoretical terms successfully refer to real unobservable entities and their properties. But many severe criticisms are advanced against IBE realism, and in response, several variations and rivals of the standard version have developed to overcome these criticisms. In particular, many philosophers are convinced that the only or best way to overcome these criticisms—without abandoning realism in favor constructive empiricism, or instrumentalism—is to embrace a view now known as structural realism (or SR). SR is a form of IBE realism in that it holds to the aim of providing a realist explanation of the success of scientific theories. But SR rejects the standard view that we can explain the empirical success of these theories by the realist view that they are approximately true, and that their key theoretical terms genuinely refer to unobservable entities. The criticisms of this view have not been met, according to structural realists, making SR itself the only viable realist alternative to empiricism and instrumentalism. Indeed, this is one of the central arguments for structural realism. On the other hand, structural realists also argue that we can explain the empirical success of the relevant scientific theories by the hypotheses that they capture the modal structure of nature, and that the mathematical or formal relations posited by successful theories are approximately true of, or accurately model, the structure of nature. Against standard realism, SR holds that structure itself can explain a theory’s empirical success, quite independently of the truth or falsity of the theory’s claims about, or references to, the unobservable entities it posits, or their causal powers. So for example, the success of Fresnel’s theory concerning the propagation of light is explained for SR by the mathematical laws which captured the ‘‘structure’’ of the process (retained in Maxwell’s theory) and not Fresnel’s hypotheses concerning the nature of light as an ethereal medium in which molecules of ether are the mechanical carriers of light waves (Worrall 1989a–d).
123
From Standard Scientific Realism
297
My aim in this paper is to evaluate the ability of SR to explain the empirical success of scientific theories, and thus its credentials as a reasonable alternative to standard IBE scientific realism. My key claim is that neither standard IBE realism nor structural realism can plausibly explain the explanatory success of scientific theories—which I will argue is a necessary condition of theories’ empirical success and at least as important for realists to explain as theories’ predictive success. Of course, if SR is the only or most defensible alternative to more standard versions of realism, then my argument will prove cold comfort to realists who see empiricists or instrumentalists waiting in the wings to declare victory. The second stage of my argument is that there is a third version of IBE realism, which provides a better account of scientific achievements than standard and structural realism. I will call this alternative ‘‘Best Current Theory Realism’’ (or ‘‘BCT’’) because it holds that realists’ commitment to the referential success and approximate truth of scientific claims about unobservables should be (1) strictly limited to our best, most successful current theories, and (2) withheld from superceded theories that are rejected by current science. On my argument, standard and structural realists share two flawed assumptions, which they take to constitute the hard core of realism: namely, (1) the assumption that the empirical success of scientific theories is primarily due to, or best explained by, the fact that such theories were saying some true things concerning the unobservable entities, processes, or structural relations in nature; (2) the assumption that these true components, or accurate representations, of superceded theories are identifiable as precisely those features which are preserved and retained in more successful, successor theories, including our best current theories. To be clear, I am not arguing that there are no continuities across theorychange in highly successful sciences; or that our best current theories do not reveal what outdated theories ‘‘got right’’ and contributed to the cognitive progress of science. Rather I hope to show that such continuities—whether at the level of theoretical reference, component hypotheses, or formal structures of relations, do not provide adequate explanations of the empirical success of outdated theories, nor do they explain the special empirical success attained by our best current theories. I attempt to provide an alternative explanation, which exploits a realist commitment to the approximate truth of our best current theories in order to explain how their predecessors could gain some empirical success, independently of successful reference, true components, or accurate structures of relations (mathematical or otherwise). I begin my argument in the next section with a proposal concerning the tests or criteria that any well-confirmed version of IBE realism should satisfy. Some of these criteria are controversial and are defended at various points in this essay. Throughout, I am mindful of the fact that IBE scientific realism is supposed to be evaluated as a scientific hypotheses and as such, should be judged by the same criteria of explanatory power, which realists identify as the traits that give scientific theories this power: unification, consilience, internal consistency, coherence with background knowledge, intuitive plausibility, predictive accuracy, completeness, novel predictions and explanations, etc.
2 What Tests Should Any Version of IBE Realism Pass? I propose the following tests or criteria for evaluating the merits of any version of IBE scientific realism, including SR. T(1) The Explanatory-Success Test: IBE realism is conventionally tested by its ability to explain the predictive success of theories, and in particular, their success in making
123
298
G. D. Doppelt
‘‘novel predictions’’. I propose that it is also tested by its ability to explain the explanatory success of theories, and in particular, their success in giving ‘‘novel explanations’’. This test is explained and justified below. T(2) The Explanation-of-Failure Test: I propose that any version of IBE realism is tested by its ability to explain the predictive and explanatory failures of an otherwise successful theory. In defense of T(1) and T(2), note that any version of IBE realism which meets these tests possesses greater unification, consilience, explanatory breadth, and completeness than any version that fails to meet them, and for example, only explains predictive success [one component of T(1)]. T(3) The Pessimistic Induction Test: Any version of IBE realism is tested by its ability to answer or rebut the pessimistic induction: the argument from (a) the many successful scientific theories which have turned out to the be false and non-referring in their central ontological claims about unobservable to (b) the likelihood that our currently most successful theories (and all future ones as well) are false and non-referring in their claims concerning unobservables. T(3) is required if a version of IBE realism is to possess intuitive plausibility, coherence with background knowledge, and unifying account of the success of outdated and current scientific theories. T(4) The Empirical Adequacy Test: Any version of IBE realism is tested by its empirical adequacy—its ability to accommodate or circumvent any particular cases of scientific theories, which are or were empirically successful, but to the best of our knowledge, false and non-referring. Even if T(3)—the pessimistic induction—commits the base rate fallacy or otherwise misjudges the number of successful-but-false theories, any such cases constitute putative counter-examples or disconfirming evidence for the realist account of success and weakens the empirical adequacy of any version that admits such counter-examples. If the truth of a theory X or Y cannot explain its success, why should it be invoked as an explanation of successful theories in general? T(5) The Cognitive Progress Test: Any version of IBE realism is tested by its ability to explain the cognitive progress of science—towards theories with ever greater explanatory power and predictive success, and towards theories that are true, or approximately so. T(5) is important so that the growing success of theories is not a miracle. Versions of IBE realism that meet T(5) have greater consilience, unification, completeness, intuitive plausibility than versions that do not. T(6) The Rational Acceptance Test: Any version of IBE realism is tested by its ability to explain the reasons that lead scientists to accept a theory as true, so that it is not a miracle that the theories which realists regard as successful and candidates for truth are embraced by scientists for similar reasons. Among these tests, T(3), (4), and (5) are undoubtedly widely accepted by realists. Over the last few decades, standard IBE realists have devoted most of their efforts to the development of versions of realism that satisfy precisely these tests. They have attempted to accommodate the putative fact that there are well-known cases in the history of science
123
From Standard Scientific Realism
299
of successful theories that are replaced and to the best of our knowledge false and nonreferring in their claims concerning unobservables; e.g., the caloric theory (ies) of heat, the ether theory (ies) of light, etc. (Laudan 1981, 1984). Indeed, structural realists are in the forefront of those who argue that standard IBE realism cannot meet these tests., but that SR can (Worrall 1989a–d; Ladyman and Ross 2007; Ladyman 2008; Carrier 2004). By explaining the empirical success of theories and cognitive progress in terms of continuities of accurate mathematical structure, SR promises to avoid the difficulties posed for standard realism by discontinuity and replacement of theories’ ontological claims concerning unobservable entities and mechanisms. Clearly, the project of IBE scientific realism depends on which theories in the history of science are taken to exhibit empirical success and how it is characterized. Standard and structural realists have agreed, for the most part, that genuinely successful theories are those and only those that exhibit striking predictive success, in particular with respect to ‘‘novel predictions’’, predictions of ‘‘new’’ kinds of phenomena that a theory is not antecedently designed to accommodate and/or ones that are unknown or improbable or mysterious until the theory predicts them (Psillos 1999). The outstanding paradigms of successful theories such as Newtonian mechanics, Fresnel’s wave theory of light propogation, Einstein’s theory of relativity all made startling novel predictions which were confirmed (Worrall 1989a–d; Carrier 2004). If genuine empirical success requires novel prediction success, IBE realists could reasonably limit the realist commitment to theories with this trait and seek to satisfy T(3), (4), and (5) for this set of ‘‘genuinely’’ successful theories alone. For this reason, my proposal T(1) that test IBE versions of realism by their ability to explain the explanatory success of theories, not just their predictive success, is controversial and requires justification. My proposal raises the stakes for scientific realists, in at least two ways. First, what IBE realists appeal to in order to explain theories’ predictive success—whether it is some of these theories’ claims about unobservable entities, mechanisms, or structure by itself—may be insufficient to explain theories’ explanatory success. Secondly, if explanatory power is a necessary condition of empirical success, then it may alter which theories are taken by realists to be successful, and wholly, or in some part and aspect, approximately true.
3 Do IBE Scientific Realists Need to Explain the Explanatory Success of Scientific Theories? There are several powerful reasons for taking the explanatory success of theories to be an essential feature of their empirical success, and thus (1) a feature of theories that IBE realists need to explain and (2) a feature which is linked to the IBE realists’ inference to the approximate truth of theories’ claims concerning unobservables. Explanation is one of the main goals of scientific inquiry—the effort to attain an understanding of empirical regularities among observable phenomena by discovering the unobservable entities, processes, and structures of relations responsible for, or at least manifest in empirical regularities. Scientists often accept theories and take them to be true precisely because they exhibit the virtues that realists identify as the grounds of explanatory power: unification, consilience, internal consistency, coherence with background knowledge, simplicity, intuitive plausibility, and empirical adequacy or accuracy in their implications for the phenomena they explain. Indeed, some theories such as Darwin’s original theory of evolution by natural selection were taken by scientists as rational to believe, on the basis of what they explain—
123
300
G. D. Doppelt
even in the absence of predictive power or novel predictions (Ladyman and Ross 2007, 78). In cases such as this, ‘‘novel explanations’’ play a key role in generating the empirical success and rational acceptance of the theory: the theory’s ability to explain kinds of phenomena it was not originally designed to explain. Indeed, what many realists call ‘‘novel predictions’’ are in fact either ‘‘novel explanations’’ or imply them. Newtonian mechanical laws ultimately accounted for many kinds of phenomena they were not first designed to explain; but some of these were well known, and not in need of prediction (Worrall 1989a–d; Carrier 2004). Or, if we choose to call these ‘‘novel predictions’’, the point is that Newtonian principles were able to explain these phenomena, and enhance the unifying power of the theory—thus its explanatory power. In the case of Fresnel’s famous startling prediction of the white spot at the center of the shadow of a circular screen illuminated by a point of light, his wave theory of light also explained why it would occur—a novel ‘‘prediction cum explanation!’’. In other cases, new theories with explanatory virtues gain scientific credibility and acceptance because background theories (assumed by them) with a high degree of predictive success also posses these explanatory virtues—linking explanation and prediction as inter-related components of empirical success. Many scientific realists have stressed the central role of explanatory power in scientific inquiry and the warrant for belief in the truth of theories (Salmon 1970, 1985, 1990; Boyd 1973, 1981, 1984a, b; McMullin 1987; Psillos 1999). From this standpoint alone, realists ought to be able to explain what they take to be a key dimension of the aims, credibility, and success of scientific theories—their explanatory power, including their power to generate ‘‘novel’’ explanations. In any case, any version of scientific realism that can explain theories’ explanatory success, as well as their predictive success, would on that score exhibit greater unification and completeness. To this extent, that version of scientific realism would possess greater confirmation. But the reasons for T(1)—the Explanatory-Success Test of IBE realism—go deeper. Constructive empiricists do not need to account for the explanatory power of theories— because they can treat it as a pragmatic goal or interest of scientific inquiry and not evidence for the truth of theories. On the other hand, IBE realists tie evidence and the confirmation of theories itself to the presence of explanatory virtues in theories (Psillos 1999, 171). IBE realists take ‘‘inference-to-the-best-explanation’’ to be the paradigm of scientific inference and confirmation of theories. On the realist view, phenomena constitute evidence for one theory over its rivals only if the theory provides a better, or the best explanation of the phenomena. This IBE conception of evidence provides the classic realist response to one of the central anti-realist arguments—the argument from the underdetermination of theory by observation. Rival or contradictory theories can have all and only the same observational consequences (or predictive success). But they cannot both be true, or confirmed as such. A classic realist response is that equivalence of observational implications is not evidential equivalence because evidence and degree of confirmation depend on which theory possesses greater explanatory virtues, e.g., unification, consilience, internal consistency, etc. (Psillos 1999, 162–174). If IBE realists tie the confirmation and truth of a theory to its explanatory success, how could scientific realism itself be confirmed if it failed to explain the explanatory success of theories? Such a failure would make the explanatory success of theories a miracle. Furthermore it would generate an incoherent version of IBE realism because it would sever any link between the ‘‘confirmation’’ and ‘‘empirical success’’ of theories. So the IBE realist’s conception of evidence and confirmation provide another powerful reason for taking T(1)—the explanatory-success-criterion—as a crucial test of the adequacy of any version of scientific realism.
123
From Standard Scientific Realism
301
Indeed, the failure of some version of IBE realism to meet T(1) creates another internal incoherence for the position. IBE realists identify themselves as naturalists and insist that they defend a ‘‘scientific’’ realism (Boyd 1981). It is a scientific realism not simply because it provides a realist view of scientific theories. It is a scientific realism because it treats realism as a scientific theory whose confirmation or disconfirmation is no different from that of any other scientific theory. Put differently, for a scientific realist, the empirical success necessary to support the truth of the realists’ theory is no different than the empirical success required to support the truth of any other scientific theory. If there is a difference, scientific realism will confront a dilemma that challenges its coherence, consistency, and plausibility. If there is a significant difference between the empirical success realists take to support the truth of scientific theories and what they take to support realism itself, then either (1) realism itself as a scientific theory lacks support or (2) realism has mischaracterized the empirical success it takes to support the truth of scientific theories in general. Either outcome undermines the plausibility of scientific realism. In fact, IBE scientific realism is routinely taken by realists to gain confirmation and ‘‘empirical success’’ exclusively through what it can explain or best explain—the ‘‘empirical success’’ of scientific theories. What predictions or novel predictions is scientific realism supposed to make? None, to the best of my knowledge! If novel predictions are the ‘‘sine quo non’’ of empirical success and crucial to the realist inference to the truth of scientific theories, then realism itself—as a scientific theory—lacks empirical success and thus confirmation. On the other hand, if scientific realists revise their current conception of empirical success as novel predictions, there is a way out of the dilemma. Allow that empirical success requires explanatory success and may be in some cases sufficient for it (without novel predictions), and the way is open for IBE scientific realism to be confirmed by what it explains, or may best explain. This provides another powerful reason for including T(1)—the Explanatory-Success-Criterion as a crucial test for IBE Scientific Realism. Without it, the dilemma I pose above undermines the internal coherence, consistency and plausibility of a scientific realism. There are as least two other ways out of the dilemma. First, IBE realists may abandon the claim that their realism is scientific and allow that some form of philosophical or metaphysical or ‘‘meta-scientific’’ argument is involved in justifying realism. This is unattractive, and in any case, would radically alter the terms of the realism debate. Secondly, IBE realists could argue that scientific realism does indeed make novel predictions and can meet the ‘‘novel prediction’’ standard of empirical success. How might this go? For example, structural realists might claim that SR—while an explanatory theory—also makes the novel prediction that any future theory in a highly successful field such as relativity physics will preserve the mathematical equations and structural relations of our best current theories. Standard IBE realists might make a similar novel prediction to the effect that any very successful future theory in a highly successful field will preserve some of the central theoretical claims concerning unobservables or central posited entities and mechanisms in our best current theories. Of course, these sorts of claims may be too vague and imprecise to count as scientific predictions. But in any case, they are unconfirmed and leave scientific realism without the novel predictive success it would require to resolve the dilemma. Here, I am identifying novel predictions as predictions concerning kinds of phenomena, which are presently unknown (i.e. the structure or content of future theories). Scientific realists might develop realism to make novel predictions concerning kinds of phenomena known to us, but different from the kind of phenomena it is designed to explain (or predict). But the main point is that the dilemma I pose arises because scientific realists
123
302
G. D. Doppelt
defend realism on the basis of its explanatory success and yet identify the empirical success of scientific theories with novel predictions. As I suggest above, the best solution to the dilemma is to accept T(1) above, require scientific realism to explain the explanatory success of scientific theories, and then develop a version of IBE realism which can explain the explanatory success of theories. Indeed, I will argue that neither standard realism nor structural realism contains the resources to explain theories’ explanatory success, and also meet the other criteria for IBE realism set out above in T(1)–T(6). This argument sets the stage for the defense of ‘‘Best Current Theory’’ scientific realism, which I argue can satisfy all of these criteria. In the next section, I open the way to my argument by examining the notion of the explanatory success of scientific theories, and the components of an explanation of their explanatory success.
4 Explanatory Success and its Explanation In science and philosophy of science, there is no consensus on the nature of scientific explanation. Nonetheless, among scientific realists, there is agreement on the standards of explanatory success and the virtues of scientific theories which give them explanatory power. As I indicate above, realists identify the explanatory power of theories with their unification, consilience, coherence with background knowledge, internal consistency, completeness, intuitive plausibility, simplicity, and empirical accuracy. These standards do not rest on any a priori analysis of scientific explanations. For a scientific realist, these standards are discovered through scientific inquiry itself and made evident in the history of science—where ever more powerful scientific theories raise and meet ever more demanding standards of explanatory and predictive success. From this standpoint, the great Newtonian achievement is not simply providing a unified account of celestial and terrestrial motion—which up to that point were thought to be essentially different natural kinds of phenomena requiring different kinds of explanation. Newtonian mechanics raises the standard of unification and consilience in physics—which all subsequent theories in the field are expected to meet. So, the empirical success of Newtonian mechanics is partly explained by the standard of explanatory unification the theory achieved and the way this standard informed the aims and cognitive progress of all physical inquiry up to our current physics of relativity. This example speaks to the issues surrounding our notion of explanatory success and its explanation. First, we note that explanatory success does not in itself imply the truth of the theory which possesses explanatory success. Though literally false, Newton’s laws provide a virtual paradigm of explanatory and predictive success—because they set and satisfied higher standards of unification, consilience, empirical accuracy, completeness, simplicity, and coherence with background knowledge (Galileo and Kepler) than any theory known at the time. IBE realism recognizes that theories which are false provided good or even the best explanations, given the standards, evidence, and theories available in the context. IBE realism sanctions an inductive ‘‘inference’’ to the best explanation which implies that the explanatory power of a theory is logically distinct from its truth. Indeed, as I argue below, inference-to-the-best explanation realism exclusively justifies an inference to the truth of our best, most well-established current theories and not any inference to the truth of predecessor theories, their components, structures, or entities—regardless of how successful they were in their time. Inference-to-the-best explanation justifies an inference to the truth of our best current theories because they provide the best explanation of
123
From Standard Scientific Realism
303
phenomena. The explanatory and predictive success of outdated theories can be explained without the assumption that they or any of their components are true, or so I will argue below. The example of the great Newtonian achievement reveals another feature of the standards governing explanatory success. Margaret Morrison has argued that the unifying power of scientific theories is distinct from their explanatory power and may even imply a loss of explanatory power (Morrison 2000). Her argument targets realist philosophers of science who identify explanatory success with unification, and take unifying power to be the central ground of an inference to the truth of theories (Friedman 1974; Glymour 1980; Kitcher 1981, 1993). I do not identify explanatory power with unification. Nonetheless, unification and consilience are among the central standards of explanatory success— though they can be defeated by a theory’s failure to meet standards of internal consistency, empirical accuracy, coherence with background knowledge, etc. Morrison’s argument is that the great unifying theories such as Newtonian mechanics and Maxwell’s theory of light and electricity achieve a mathematical unification at the price of physical-causal explanation (Morrison 2000, 1–109). From this standpoint, Newtonian mechanics unified many phenomena of motion under the inverse square law of gravitation—but failed to provide a (contact-action) physical account of gravitational forces themselves. Similarly, Maxwell’s theory of the electro-magnetic field offered a unifying mathematical account of optical, magnetic, and electrical phenomena—but ultimately abandoned the attempt to provide any physical explanation of what causes the propagation of electromagnetic waves—when the medium of the ether as the carrier of waves is deleted from the theory (Morrison 2000, 62–64). Does Morrison’s argument establish that unification sacrifices explanatory power in these cases? As I read these cases, the unifying mathematical laws of Newton and Maxwell succeed in explaining the phenomena they unify—though they do not provide mechanisms which explain why or how the laws operate. Newtonian mechanics ultimately accepts action-at-a-distance as an ‘‘unexplained explainer’’ and abandons the paradigm of contactaction central to Cartesian mechanics. No theory explains everything—lest an infinite regress of ‘‘why questions’’ would undermine all theoretical explanations. Thus unification can provide a general standard of theories’ explanatory success, even though theories alter the substantive assumptions concerning what can function as an ‘‘unexplained explainer’’. Morrison would appear to agree, though it cuts against her attempt to sever explanatory power from unification: ‘‘Maxwell’s theory was explanatory in the sense of being able to account for optical and electromagnetic phenomena using the same laws, but not in the sense of imparting an understanding how or why field-theoretic processes took place’’ (Morrison 2000, 63). The upshot of this discussion is that a theory such as Newtonian mechanics need not be true or meet every local standard (such as Cartesian insistence on contact-action) in order to attain extraordinary explanatory success, in part by unification. How then can we explain its explanatory success, which is one test—T(1) above—of scientific realism? My proposal—defended below—is as follows. We can explain the success of Newtonian mechanics by enriching what realists allow into the explanans, to include the virtues and methods of the theory—e.g. its unique unifying power at the time—and how these enable it to raise the standards of empirical success for all subsequent theories in physics. At this point, the scientific realist may balk. What role is left for reality and theoretical truths about it in this explanation of Newton’s success? On the view I defend below—‘‘Best Current Theory Realism’’—our best current theories, in this case relativity physics—provide theoretical truths concerning nature—which are also part of the best explanations of why
123
304
G. D. Doppelt
Newtonian mechanics works well for many empirical phenomena and fail for others, despite the fact (and because) his laws are false. This remains a robust form of scientific realism because it relies on the approximate truth of our best current theories as part of the best explanation of (1) their special empirical success and (2) the mixture of empirical success and failure exhibited by outdated and superceded theories. But why does this ‘‘Best Current Theory’’ realism promise a better way to meet above tests T(1)–T(5) than standard and structural versions of IBE scientific realism?
5 Standard and Structural IBE Realism My argument in this section is, simply put, that standard and structural realism share a structural defect: they both confine the realist’s explanation of the empirical success of scientific theories to components of the theories (claims about unobservable entities, processes, or structures of relations) that need to be (1) taken as true (or accurate, in the case of mathematical structures of relation) and (2) preserved, or retained in the best successor theories and thus current ones. As a result, standard realism is stuck with the difficulty that their explanations of success require that too much of superceded theories need to be retained and taken as true—including components that are non-referring and false, to the best of our knowledge. The unhappy result for structural realism is that its explanation of success requires too little of superceded theories to be retained and taken as accurate—namely, its bare-bones formal structures of relations (represented by their mathematical equations) that are by themselves, too thin to account for the explanatory and empirical success of any scientific theories. The difficulties with standard IBE realism are well-documented—so my focus here is the evaluation of structural realism. I will focus on two scientific examples which structural realists have exploited to establish the defects of standard realism and the superiority of the structural view: Fresnel’s theory of the transmission of light and the phlogiston theory of combustion and calcination. In order to meet tests T(4)–T(6) above, standard realists have been hard pressed to avoid putative cases of genuinely successful theories whose claims concerning unobservables are false and non-referring, to the best of our knowledge. Fresnel’s theory of the propagation of light in an ethereal medium succeeded in explaining and predicting many phenomena involving the diffusion of light. It provided novel predictions/explanations (e.g. the appearance of an otherwise improbable white spot at the center of the shadow of a circular screen illuminated by a light source). On stringent realist standards (e.g. novel predictions), it is an authentic empirical success. Yet Fresnel’s account depended on the hypotheses that a luminiferous ether of molecules is the carrier of light-waves; that the amplitude of light waves correlates with the velocity of displacements of ether molecules; that the transversal vibrations of light rays is proportional to the oscillations of ether molecules; etc. The existence of a luminiferous ether, and these hypotheses concerning its behavior, are ultimately rejected by Maxwell’s theory of the electromagnetic field and are entirely false, to the best of our knowledge. Yet we need to bring in these false hypotheses if we want to explain the explanatory and predictive success of Fresnel’s ether theory. The success of the theory is also explained by the fact that it meets IBE realist standards of unification, consilience, empirical accuracy, intuitive plausibility, novel predictions/explanations, etc. Structural realists reject my explanation of the empirical success of Fresnel’s ether theory. Their reason is compelling! If such non-referring, false theories can provide genuine empirical successes, standard IBE realism fails the above tests T(1)–T(6). It falls prey
123
From Standard Scientific Realism
305
to the pessimistic induction, empirical inadequacy, a failure to explain theories’ success by truthful components or referents, etc. Yet, the structural realists retain the assumption that a realist account of theories’ success must identify some aspect of theories which (1) are accurate and (2) are preserved by their successors. They have established that this aspect cannot be theories’ ontological claims about unobservable entities and mechanisms (Worrall 1989a–d; Carrier 1991, 2004; Ladyman and Ross 2007). But the case of Fresnel shows that this aspect can be the mathematical structure of relations posited by Fresnel’s equations and retained by Maxwell’s theory of the electromagnetic field—which generalizes these equations to apply to electric and magnetic phenomena, as well as the propagation of light. Though the ontology of the luminiferous ether gives way to that of an electromagnetic field, ‘‘there is nonetheless a structural, mathematical continuity between the two theories’’ (Worrall 1990, 21). For the structural realist, Fresnel’s equations (1) accurately represent the unobservable structure of relations underlying electromagnetic phenomena and (2) are preserved by Maxwell and other successor theories of the field. Thus theories’ ability to accurately represent or model the unobservable structure of relations underlying domains of phenomena may provide the best explanation of their empirical success. The advantages of structural over standard IBE realism are evident. It promises a better realist explanation because it provides a way of meeting tests T(3)–T(5). Assuming that successful theories’ representations of structure are neither falsified nor abandoned by successors (and our best current theories), structural realism may circumvent the pessimistic induction, the problem of empirical inadequacy, and the need to account for cognitive progress across theory-change. There may be some additional advantages of structural realism. If structure is preserved across theory-change, how do structural realists account for cognitive progress? Presumably, when new theories provide greater empirical success than predecessors, the structural realist will argue that the new theories don’t just preserve structure, but contain it in more accurate mathematical equations and representations of structure. Thus Maxwell’s laws are applied more generally than Fresnel’s, and the mathematical laws of relativity physics subsume Newtonian laws as a limiting case. The question here is whether this account works for progressive theory-change in general. To this end, I will examine structural realist accounts of the transition from phlogiston theory to Lavoisier’s oxygen theory. Another possible advantage of structural realism is that it has a ready-at-hand way to meet T(2)—the explanation of theory’s empirical failures. These failures can be neatly explained as the result of otherwise successful theories’ false ontological hypotheses, which limit or distort the way they applied their accurate equations and structural laws. Whatever its merits, structural realism is most severely tested by its power to meet key test T(1)—the need to explain the explanatory and predictive success of theories. By stripping down a theory’s source of success to its bare-bones, uninterpreted mathematical equations, structural realism deprives itself of the resources required to provide a plausible explanation of empirical success. Let’s begin with predictive success because the difficulty here for structural realism only becomes more severe when we turn to explanatory success. Does the structural realist have a plausible account of its favorable case of Fresnel’s predictive success, supposedly due to the accuracy of his mathematical equations? I begin with some difficulties for this view raised by Psillos and then argue that they are symptomatic of more fundamental weaknesses. Psillos observes that uninterpreted equations alone cannot explain a theory’s predictive success because its ability to make predictions at all requires other components of the theory: theoretical principles, auxiliary assumptions, background knowledge, etc. In the case at hand, so Psillos argues, Fresnel’s mathematical
123
306
G. D. Doppelt
laws and the predictions they yield depend on theoretical principles concerning the conservation of energy, the geometric arrangement of light rays where two media meet, and the relation of the amplitude of light waves to the velocity of the displacement of ether molecules. Uninterpreted mathematical equations, by themselves, in the absence of a theory’s substantive theoretical principles and mechanisms do not yield any predictions, successful or unsuccessful (Psillos 1999, 153–159). The structural realist could broaden the conception of structure to include such theoretical principles. But in that case, we begin to lose the structural realists’ crucial distinction between (1) the ontological ‘‘content’’ of Fresnel’s theory concerning the nature of light and (2) its account of the ‘‘structure’’ of relations underlying the behavior of light waves. In other words, Fresnel’s several hypotheses concerning the structure of relations exhibited by the propagation of light waves were in effect hypotheses concerning the nature of light itself, its properties and attributes (Psillos 1999, 159). And, in any case, wherever one tries to draw a line between the substantive and structural hypotheses of a theory concerning unobservables, there is no guarantee that the structural hypotheses alone generate the theory’s predictions and thus its predictive success. The problem here for SR is that it operates with a naı¨ve criterion of predictive success, on which the notion is reduced to the simplistic test of whether or not a theory’s (novel) predictions are confirmed. Psillos’s insight is that we cannot explain a theory’s predictive success unless we explain what features of the theory are responsible for its ability to make any predictions whatsoever (novel or otherwise) and these features go beyond its bare bones uninterpreted equations. But, the notion of a theory’s predictive success is yet more complicated than Psillos’s critique recognizes. Predictive success is a function of the standards of success operative in a field such as optics where theories are evaluated by their ability to predict well-known empirical regularities concerning the propagation of light—refraction, diffusion, reflection, etc. Apart from its explanatory power, a theory of optics such as Fresnel’s only counts as a predictive success if it possesses the epistemic virtues that enable it to meet standards of unification, consilency, accuracy, completeness, novelty in the range and kinds of phenomena it is expected to predict. But any explanation of the presence of such virtues in a theory takes us well beyond its mathematical or formal structure. Furthermore, there is typically a gap between the predictions of a theory and the measured values of the key variables. Thus the explanation of predictive success also needs to bring in operative standards of approximation and margins of experimental error to differentiate predictive success from failure. Given all these components of a theory’s predictive success, its explanation will require much more than the theory’s mathematical structure alone. This difficulty becomes more severe when we examine structural realists’ ability to fully meet test T(1) and explain the explanatory success of Fresnel’s theory. As I argue above, its explanation of phenomena of light transmission (reflection, refraction, etc.) clearly required the luminiferous ether and the various properties attributed to it by the theory. Fresnel’s theory, as far it went, provided explanatory power in optics because, though false, it possessed the virtues of unification, consilience, empirical accuracy, coherence with background knowledge, internal consistency, etc. that enabled it to meet and raise standards of explanation operative in the field, then and now. Maxwell’s theory raised and met these standards to a far greater degree by unifying the explanation of optical, electric, and magnetic phenomena. This explanatory achievement, however, required a very different ontological framework—an abandonment of the luminiferous ether and the adoption of a commitment to the existence of electromagnetic fields, electric charge, field vectors, etc. The mathematical and more substantive ‘‘structural’’ continuities between the two theories (e.g., a principle of conservation of energy) are insufficient to explain the explanatory
123
From Standard Scientific Realism
307
power of each of these theories. Moreover, as I argue in Sect. 3 above, scientific realism needs to explain the explanatory success of theories lest it fail to meet its own standard of internal coherence and provide the best explanation of empirical success. In the next section, I turn to the case of phlogiston theory and its replacement by the oxygen theory of Lavoisier. Structural realists treat the cases of Fresnel and Maxwell as paradigmatic for their realism because mathematical structure is so evidently preserved in this transition and does play ‘‘a’’ crucial role in the success of these theories. Can they accommodate theories in which mathematical equations and structure play no such role?
6 The Case of Phlogiston Theory The phlogiston theory of combustion provides a test case for any form of IBE scientific realism. Despite the fact that the theory does exhibit explanatory and predictive power, to my knowledge no scientific realist holds that the term phlogiston successfully refers to anything or has any of the causal powers attributed to it by the theory. The move made by some standard realists to account for the success of Fresnel’s ether theory of light—namely that the term ‘‘luminiferous ether’’ succeeded in referring to the electromagnetic field— however implausible—is clearly an even less plausible move for the case of phlogiston (Psillos 1999, 280–300). Several historians and philosophers provide excellent accounts of phlogiston theory—its achievements, anomalies, and replacement by oxygen theory (Musgrave 1976; Pyle 2000). More to the point, some structural realists have boldly addressed this case and sought to show how the structural account they offer succeeds where standard realism fails (Carrier 2004; Ladyman 2008). On the accounts provided by Ladyman and Carrier, phlogiston chemists established certain empirical regularities concerning combustion, calcination, respiration, and the changes in the qualities and weight of metals, calxes, wood, etc. which result from these processes (which we now know as oxidation and reduction). Furthermore, phlogiston theory provided a unifying explanation of these phenomena. The calcination of metals and combustion (e.g. of wood) were both explained as the emission of phlogiston— the dephlogistification of the objects and the phlogistification of the surrounding air. Thus, the theory could explain two kinds of physical processes as involving the same sort of unobservable events and effects. Phlogiston theory could further explain the qualitative differences between all metals and all calxes, on the hypotheses that the metals possessed phlogiston, absent in calxes and responsible for metals’ metallic qualities. The theory also explained the fact that wood, coal, and some other substances lose weight as a result of combustion—the weight loss results from the loss of phlogiston. Phlogiston chemists (Priestly) could also predict and explain the fact that animals in an enclosed space bring about the end of combustion, while plants do so to a much lesser degree. The explanation of these phenomena: plants rob the air of phlogiston, making combustion with the release of phlogiston more probable; while animal respiration saturates the surrounding air with phlogiston, making combustion less likely. Similarly, combustion of an object in an enclosed space will cease because as the air is filled with phlogiston, it cannot take it up any further. In their accounts of phlogiston theory, it is evident that structural realists (and other scholars) emphasize its virtues as an explanatory theory, providing a simple, unifying, consilient, coherent, accurate, and plausible account of various well-known empirical phenomena. In effect, they treat explanatory success as ‘‘a’’ or even ‘‘the’’ key to its empirical success and rational acceptance by phlogiston chemists (Becher, Stahl, Priestly,
123
308
G. D. Doppelt
Scheel, etc.). In this instance, structural realists accept my key argument in Sect. 2 above that a theory’s explanatory power is an essential component of its empirical success, and a feature that realists need explain. As one structural realist puts it, ‘‘I restrict attribution of reality to theoretical elements connected by strongly successful explanation. Not any structural continuity counts. All depends on the explanatory scores gained by the pertinent theory’’ (Carrier 2004, 157). At the same time, they observe that phlogiston also made ‘‘novel’’ predictions, thus satisfying this dimension of empirical success as well. Priestly successfully predicted that when certain calxes are heated in inflammable (phlogisticated) air, a pure metal would result (Carrier 2004, 150–151). Scheel advanced the novel prediction that new acids would be found, e.g., formic acid, lactic acid, etc. (Ladyman 2008). Thus, IBE realists confront a genuinely successful theory whose success needs to be explained. But standard realism falters because its success cannot be explained by the theory’s approximate truth; there is no phlogiston, thus the theory’s explanatory and predictive success results from false ontological hypotheses and non-referring terms. On the other hand, phlogiston theory ‘‘got something important right’’ concerning the structure of chemical reactions, the linchpin of the structural realists’ argument. In particular, phlogiston theory posited a common unobservable structure of relations underlying combustion, calcification, and respiration—making them all the same kind of process—known in current chemistry as the inverse processes of oxygenation and reduction. In effect, phlogiston chemistry posited an unobservable structural process that succeeded in unifying three different kinds of physical processes. Thus phlogiston theory ‘‘captured one great truth retained by Lavoisier in his oxygen theory, namely that combustion, respiration, and calcification are all the same kind of reaction (viz. ‘‘oxidation’’) and that these reactions have an inverse, namely reduction’’ (Ladyman 2008). Put differently, SR is committed to a ‘‘natural-kind realism’’ because it explains the strong success of theories as a result of the fact that they posit some unobservable mechanism(s) which shows that apparently different sorts of phenomena are really ‘‘equal in kind’’ (Carrier 2004, 153–159). Structural realists’ explanation of this case provides an important insight concerning the empirical success of phlogiston chemistry—but unfortunately it is not an insight that supports structural realism. As I read their account, phlogiston theory established several empirical regularities concerning chemical reactions and successfully sought a unifying theoretical explanation of these distinct kinds of reactions: an explanation which also exhibited intuitive plausibility, coherence, consistency, accuracy, and the other virtues which give a theory its explanatory power, on standard realist criteria. It is these features of phlogiston chemistry which explain its empirical success. Lavoisier preserved the standard of unifying this domain of chemical reactions and extended it to account for phenomena, which were anomalies for phlogiston chemists (e.g. weight gain in the combustion of metals) or unknown to them. We can always reformulate the attainment of unification by saying that the unifying theory establishes a structure of relations between different kinds of phenomena, showing that they constitute one generic kind and are the result of a common underlying and unobservable set of processes or entities. In this case, the structure of relations posited by phlogiston theory and retained by oxygen theory consists in the relations between three kinds of observable phenomena—combustion, calcinations, and respiration. What the structural realists need, however is some unobservable underlying structure of relations which is supposed to explain the explanatory and predictive success of phlogiston chemistry and oxygen theory. What is it in this case? What truths describe it? In the above quote from Ladyman ‘‘the one great truth’’ of phlogiston theory, retained by Lavoisier, is ‘‘that combustion, respiration, and calcification are all the same kind of reaction’’. But that
123
From Standard Scientific Realism
309
truth does not describe any underlying unobservable structure of relations, except to say ‘‘that there is some such structure’’ commonly responsible for, or manifest in, three kinds of observable phenomena (combustion, respiration, and calcification). Of course, with the benefit of current science and hindsight, we can correctly describe the structure making full use of the ontology and entities established by subsequent chemistry. Once the theory (ies) of chemical reactions successfully posits the existence of the entities and mechanisms embodied in our theoretical notions of oxygenation and reduction, on this basis alone, we can know that there is a unitary structure underlying the phenomena of combustion, respiration, and calcification. Thus, the ontological content of post-Lavosier chemistry is required to know that there is an unobservable structure of relations underlying these phenomena and to know what it is. But structural realists cannot help themselves to the ontological content of any theory including current chemical theory—because structure, all by itself, is supposed to provide a full explanation of theories’ empirical success. Besides, according to SR, we cannot know that the ontological claims made by current chemical theory concerning oxygenation, reduction, etc. are true. And if we can’t know that these processes occur, how can we know that combustion, calcification, and respiration are all the same kind of phenomena? So structural realists add a new miracle for realists to explain. How can the bare fact that there is some unobservable structure of relations underlying chemical reactions of certain kinds, and the bare fact that phlogiston chemists assumed so, possibly explain the explanatory and predictive successes of phlogiston theory? Or, for that matter the success of oxygen theory? Clearly, it is (1) the claims made by phlogiston chemists about this structure, and (2) the way they exploited these claims to attain unifying, consilient, coherent, plausible, and accurate explanations of some well-established phenomena, that best explains the empirical success of the theory. Realists can recognize that false theories which otherwise satisfy exacting standards of empirical success may be well-confirmed and successful. Scientific realism is only threatened if this situation is taken to justify the skeptical conclusion that no scientific theory can ever be known to be true or approximately true. I hope to show that my proposal below—‘‘Best Current Theory Realism’’ avoids such skepticism, although it rejects the IBE realist attempt to explain all theories’ success in terms of the truth of their central claims concerning unobservable entities, processes, or structures of relations.
7 Best Current Theory Realism My argument so far generates an impasse for IBE scientific realism. Standard realists cannot provide plausible explanations of theories’ empirical success based on their truthful theoretical components because many such theories are false in their central ontological claims concerning unobservables. Thus, standard realism fails tests T(1)–T(6); it falls prey to the pessimistic induction, empirical inadequacy, and T(1)—the provision of a coherent, consistent, plausible account of theories’ empirical success. Structural realism promises to circumvent these difficulties. It abandons the need for truth and continuity at the level of theories’ ontologies, and identifies their structural features alone as the source of continuity in theory—change and the empirical success of theories. But, if my argument is correct, the structural component is far too thin and vacuous to provide an adequate explanation of theories’ explanatory and predictive success; so it cannot meet T(1). And in any case, structural realists need to tacitly appeal to the ontological claims concerning entities and mechanisms made by modern chemistry in order to establish the existence of the structures to which they appeal. But on their own view, we cannot know that such substantive
123
310
G. D. Doppelt
ontological claims are true and without these claims, any recognizable notion of structure disappears from view. This state-of-affairs generates an impasse for IBE scientific realism unless there is another version that better satisfies test T(1)–T(6). I will argue that ‘Best Current Theory Realism’ (or ‘BCTR’) fills the bill. BCTR holds that we can know that those current scientific theories (relativity physics, molecular biology, chemical theory, plate tectonics, etc.) which meet the highest standards of explanatory and predictive success in the entire history of their respective domains of inquiry are true, or approximately so. Several features of this view make it a very natural, intuitively plausible version of IBE scientific realism. Realists are committed to an inference to the truth of theories, which provide ‘the’ best explanation of phenomena—the most unifying, consilient, accurate, coherent, complete, plausible explanations/predictions. That is precisely what our best current theories do! So IBE realism, strictly speaking, is committed to the truth of our best current theories, and NOT to the truth of their predecessors, despite the fact that they provided good, or even the best explanations in their time and context of available evidence. Secondly, IBE realists standardly assume the truth of our best current theories and could not get their arguments off the ground without it. Standard and structural realists have to rely on the truth of our best current theories in order to identify what outdated theories got right and what they got wrong. Assuming the truth of relativity physics, we can explain why Newtonian laws, though false, work or hold under a limited set of conditions and fail to hold universally. On the realist assumption that our best current theory- relativity physics—is true, we get a natural and realist explanation of the success of Newtonian mechanics, whether or not we identify true components or accurate mathematical structure in it. Put differently, because relativity physics provides part of the best explanation of the empirical success of Newtonian mechanics, a realist can naturally take this as evidence of relativity physics’ explanatory power and thus its truth (though of course it explains much more, e.g. the failures of Newtonian mechanics). In any case, the whole debate among realists concerning what enabled outdated or refuted theories to succeed is premised on the hypothesis that our best current theories are true. As I argue here, that commitment is all a realist needs. For in that case, the truth of our best current theories will constitute an essential component of the best explanation not just of their success, but also of the empirical success of outdated theories we now know to be false. Thirdly, standard and structural IBE realists are driven by the need to overcome the pessimistic induction. There is a paradox for realists concerning this demand. The induction proceeds from the putative fact that many successful theories are false, to the likelihood that our most successful theories are also false. The paradox is that the putative fact of many past theories that are ‘‘false but successful’’ depends on the assumption that our best current theories are true—otherwise we have no way of knowing that past theories are false. Without this assumption, the only induction that remains is one from the fact that past theories were successful but rejected to the likelihood that our best current theories will also be ultimately rejected. This induction does not bear on the truth or falsity of any of these theories and so does not speak to the viability of scientific realism. Put differently, the putative conclusion of the pessimistic induction that our best current theory is most likely false tacitly depends on the realist assumption that it will be superceded by a new theory that we will know to be true and will show that all its predecessors including our best current theory are false. All these features of IBE scientific realism rest on the tacit assumption that our best current theories are true and provide the necessary Archimedean point from which we can evaluate and explain the success and failure of superceded theories. But of course IBE
123
From Standard Scientific Realism
311
scientific realism is an attempt to provide an independent scientific argument which establishes that our best current theories are true. In its general forms, the ‘no miracles’ argument makes the global claim that the success of scientific theories would be a miracle were they not true. My version of realism rejects this global claim. As I indicate above, we can explain the success of superceded theories [phlogiston theory, Newtonian mechanics, the ether theory of light] without the assumption that they or any components of them concerning unobservables are true. On the other hand, ‘best current theory realism’ does hold that the truth or approximate truth of our best current theories is an indispensable part of the best explanation of their great empirical success. This is what makes BCTR a realist position. But if the success of superceded theories can be explained without the assumption that they were wholly or partly true, why can’t the empirical success of our best current theories be explained without the realist claim that they are true? Doesn’t this thought lead back to the pessimistic induction or the empirical inadequacy problem? Up to a point, BCTR will provide the same sort of explanation for our best current theory’s success as for their predecessor’s success. The truth of a theory, by itself, never provides a sufficient explanation of a theory’s success, on my argument. The explanatory and predictive success of any theory is always in part the results of its methods and virtues, and the ways they enable it to realize epistemic standards of explanation and prediction established in scientific inquiry. Indeed, highly successful theories, e.g. Newtonian mechanics, typically raise such standards by attaining a new level of unification, consilience, accuracy, etc. that reset the benchmark of empirical success for all future theories in the domain. Thus, for BCTR, the success of any theory is partly explained by its epistemic traits, virtues, and standards of empirical success. This hold as well for the explanations of our best current theories. What then justifies a BCT realist from invoking the truth of our best current theories as an indispensible part of the explanation of their success, while denying that the truth of superceded theories comes into account for their success? For BCTR, the empirical success of our best current theories possess a singularly unique feature that differentiates them from all of their predecessors and justifies the realist explanation in these cases alone. What is it? Our best current theories are unique in that they alone actualize the highest standards of empirical success, and thus confirmation, in their respective fields of scientific inquiry as a whole. This is a fact about our best current theories that calls out for explanation. If we ask what gives our best current theories this unique position at the pinnacle of explanatory and predictive success, the realist hypotheses that it is because they are true, or approximately true, provides a plausible and perhaps the best explanation. In that case, Best Current Theory Realism breaks the pessimistic induction and the problem of empirical adequacy, meeting tests T(3) and (4) above. If there is indeed a distinctive feature of (1) our best current theories that separates their success from (2) that of their predecessors, then that justifies a different explanation for (1) than for (2). And, once this difference is established, it breaks the induction from (2) to (1). Once one can establish that the best explanation of (1) current theories’ unique success is justifiably different from the best explanation of (2) their predecessors’ success, then ‘‘inference-to-the-best-explanation’’ trumps the (pessimistic) inductive argument. BCTR also circumvents the problem of empirical inadequacy because the existence of ‘‘successful-but-false’’ theories will no longer constitute counter-examples or anomalies for the realist account of our best current theory’s success. Indeed these anomalies for standard IBE realism are converted to positive evidence for BCTR; because, as I argue above, the truth of our best current theories can provide part of the best explanation of how and why outdated theories could be ‘‘successful but false.’’ Once we assume that our best current theories inform us about the way the
123
312
G. D. Doppelt
world is, we can use this knowledge to explain the empirical successes and failures of superceded theories. Standard and structural realists also assume the truth of our best current theories, in order to explain what past theories got right and wrong. But for BCT realism, these explanations can be successful and realist without the implausible assumption shared by standard and structural realism that we must attribute literal truth to some of the outdated theories’ claims or models of unobservable entities, processes, or structures of relations. For those impressed by the pessimistic induction, there is a natural reply to BCT realism’s argument that our best current theories possess a unique empirical success at the pinnacle of confirmational virtues. Couldn’t the same be said about superceded theories in their time and place? True enough, but the scientific realist is committed to explain the status of theories now, given what we now know about our best current theories and their predecessors. The fact to be explained in our context is how and why our best current theories succeed in meeting the highest standards of unification, accuracy, consilience, breadth of scope, completeness, etc. in the whole history of the field. Treated as a scientific hypothesis, the realist’s claims about our best current theories are fallible, just as these theories themselves are fallible. The inductivist may want to say that there is a good chance or high probability that in the future our best current theories will be displaced by others which in turn will meet yet higher standards of empirical success, or meet them to a much higher degree. This new situation, whether one sees it as possible or probable, does not refute best current theory realism. Rather, it would alter what the BCT realist needs to explain and which theory such realists will regard as true. Put differently, this situation would disconfirm the BCT realist’s account of which particular theories are reasonable to regard as true, but not the realist inference to the truth of whatever scientific theories meet the highest standards of empirical success in the field. This is a useful context in which to consider test T(2), the ability of some version of scientific realism to explain the failures, anomalies, or counter evidence exhibited by otherwise successful theories. Meeting this test clearly enhances the unification, consilience, plausibility, and completeness of any account of scientific realism. But it also provides a possible way to bring the standard of ‘‘novel predictions’’ into the assessment of a ‘‘scientific’’ realism. BCT realism can easily explain the failures of otherwise successful superceded theories. They fail because they are false, and our best current theories can explain why and in what respects they were false. But BCT realism can plausibly provide a different explanation of the failures of our best current theories, on the presumption that they are true. Their failures will be explained as shortfalls in the successful application or articulation of these theories—which can be remedied by future research (or found to lie outside the proper domain of the theory; or based on erroneous measurements, etc.). BCT realists’ explanation of current theory’s failures might be represented as ‘‘novel predictions’’. IBE realism, as I stress in this essay, is normally taken to stand or fall on its explanatory power concerning the success of theories. But BCT realism can be extended to make the novel prediction that our best current theory’s failures will eventually become successes, or shown to be something other than its failures. In this way, BCT realism may indeed gain a measure of added confirmation or disconfirmation, depending on the degree to which the failures of our best current theories are or are not in fact converted to successes through future research. The fact that BCT realism implies novel predictions underscore its fallibilism—in so far as its explanation of our best current theories’ success may be disconfirmed, like any other scientific theory. Nonetheless, this possibility does not undermine the claim that scientific
123
From Standard Scientific Realism
313
realism may constitute the best explanation of the special success of our best current theories, given what we now know.
8 Explaining the Cognitive Progress of Scientific Theories Towards True Theories I have proposed that any version of scientific realism is tested by T(5) above—its ability to provide a plausible explanation of the cognitive progress exhibited by scientific theories. Obviously, this is a salient feature of scientific inquiry, which demands explanation. Scientific realists have assumed that we cannot provide a genuinely realist account unless we can identify a cumulative growth in the true components of successful theories: that is, increasing accuracy in such theories’ claims or models concerning unobservable entities, properties, processes, or structures of relations. Scientific realists have taken cumulativity of theoretical truths (reference, etc.) as a necessary condition of a realist account of the cognitive progress of science. Standard and structural realists disagree over the identification of which aspects of successful theories need to exhibit such cumulativity of true or accurate representations. Best current theory realism rejects this assumption and holds that we can provide a realist account of the cognitive progress of theories without continuity or cumulativity of truthful theoretical representations between them. As I have argued here, standard realists require too much, and structural realists too little, in the way of truthful components of successful theories, to provide plausible explanations of their empirical success. But how then can Best Current Theory Realism account for the cognitive progress of scientific theories? Is it a miracle that successful scientific theories with no true or accurate component claims concerning unobservables just happen to lead to our best current theories, which are true (according to BCTR)? It is no miracle provided that Best Current Theory Realism can supply a good explanation of cognitive progress without the assumption of cumulativity of truthful components of theories rejected here. Clearly, the progress of science is most evident in the fact that its theories attain ever greater empirical success in explaining and predicting ever wider ranges of empirical regularities, which are discovered through scientific inquiry itself. So, part of the growing empirical success of theories is explained by the fact that these theories generate an enormous growth in our knowledge of the observational phenomena and problems that they seek to, and often can, predict and explain. But this itself is no miracle, because it results from the development in scientific inquiry of powerful new instruments, tools, experimental techniques, methods of inquiry, and modes of reasoning, which dramatically expand the power of its theories: Galileo’s telescope, the differential calculus, abductive reasoning, the method of unification, testability by novel explanations/predictions, the electron microscope, etc. While much of this is obvious and familiar, I would stress one essential component in BCTR’s account of cognitive progress that IBE realists ignore: namely, that scientific inquiry generates dramatic progress in the standards and epistemic goals by which its theories are tested and evaluated. The growing empirical success of scientific theories is partly explained by the fact that they raise the standards of empirical success and thus set the stage for the development of theories, which try and succeed in attaining far higher criteria or goals of accuracy, unification, consilience, completeness, simplicity, coherence, novel explanations/predictions, etc. than their predecessors. Because scientific realists desire to stress the essential role of reality, and the attainment of theoretical truth about reality, in accounting for empirical success, they tend to
123
314
G. D. Doppelt
downplay the fact that empirical success is first and foremost a function of the epistemic virtues and standards that scientific theories attain. This evident feature of theories’ empirical success does not threaten scientific realism, if we keep in mind that it is, after all, the extraordinary virtues and standards satisfied by our best theories that justify the inference to their truth, and their truth alone. Thus, for example, the extraordinary standard of unification achieved by Maxwell or Einstein—virtues of their theories and theirs alone—is what justifies the conclusion that nature itself possesses the unity, order, simplicity, coherence, etc. implied by their theories. For BCTR, this inference to the truth of their theories is justified because it remains the best explanation of how and why they alone satisfy the highest standards and virtues of empirical success known to us in the life of science. Nonetheless, BCTR provides a good realist account of the cognitive progress of scientific theories towards truth because it stresses that Einstein’s singular attainment grows out of epistemic virtues, methods, and standards in physics prepared by the whole history of successful theories leading up to it.
9 A Realist View of Scientific Rationality At the beginning of this essay, I proposed 6 tests for a defensible version of scientific realism—including T(6)—the Rational Acceptance Test—which I have not yet discussed. On this test, any version of IBE realism is evaluated by its ability to explain the reasons that lead scientists to accept a theory as true. We want to avoid the ‘‘miracle’’ that the theories, which scientific realists regard as successful and true are based on reasons that bear no similarity to the reasons that motivate scientists to regard theories as successful and true. Here, I want to argue that BCT realism meets this test far better than standard or structural realism. Among these rival versions of realism, BCT realism is the only version that argues that it is the explanatory virtues and standards satisfied by successful theories that are always an essential part of the explanation of their success. These virtues include ‘‘novel’’ explanations and predictions—but include much more besides, as I argue above. I submit that it is precisely the presence of such virtues in successful theories that motivate scientists to regard them as successful and take them to be true. This view is corroborated by the above accounts of the reasons scientists accepted Newtonian mechanics, Fresnel’s ether theory, Maxwell’s electromagnetic field, the phlogiston theory, the oxygen chemistry, in the authors cited above. Standard and structural realists stress novel predictions—which are indeed one sort of reason that motivate scientists’ acceptance of a theory. But novel predictions often entail novel explanations. In any case, novel predictions and novel explanations always increase the unification and consilience of theories across broader kinds of phenomena, which provide a larger more accurate view of scientific rationality—than novel prediction by itself. Furthermore, BCT realism stresses that the presence of sufficient explanatory virtues in a theory motivate scientists to take the theory to be true—its ontological claims about unobservable entities, processes, and structures. Standard and structural realism imply a fractured account of scientific rationality—because they disconnect their account of the empirical success of theories (i.e. they are true or possess truthful components) from the reasons that move scientists to decide which theories are true. As I argue above, realists should not generate the miracle that there is no connection between the reasons that lead scientists in practice to accept theories as true and the reasons that motivate realists to regard theories as successful and true. BCT realism bridges this connection by explaining
123
From Standard Scientific Realism
315
the success of theories, and belief in their truth, in terms of the full range of epistemic virtues and standards, which govern the actual decisions of scientists. No version of scientific realism should be accepted if it cannot account for the naturalistic grounds of scientific realism today in a way that also illuminates the grounds of practicing scientists for being realists concerning the scientific theories of their place and time.
10 How a Scientific Realist Should Identify Our Best Current Theories? Best Current Theory Realism may be challenged on the grounds that it uncritically adopts current scientific theories as the measure of scientific truth and reality. But my argument in this essay does not defend realism concerning any scientific theory merely because it is ‘‘current’’ or the best so far. We are mindful of van Fraasen’s argument that ‘‘the best’’ may be ‘‘the best of a bad lot’’. I have argued that BCT Realism affirms the truth of scientific theories, which succeed in meeting the highest standards of explanatory and predictive success in the entire history of the field. But, admittedly, this may not be sufficient for a realist commitment and is a matter of judgment, not the application of a formulaic algorithm of proof. Best current theories realism limits the commitment to the truth of our best current theories, to those which are well-established by our highest standards of explanatory and predictive success. A good test case for this view is current quantum mechanics—which enjoys an extraordinary level of predictive success. This case fuels some powerful antirealist arguments in contemporary science/philosophy of science. Barrett has argued that quantum physics is a paradigm of empirical success but against the scientific realist, cannot be true (Barrett 2002). It is internally inconsistent, and inconsistent with relativity physics. Besides, one might add, there is still considerable disagreement among quantum physicists concerning what the theory claims about the entitites and processes operative at the quantum level. For BCT realism, the situation of quantum physics does not refute scientific realism. Because the theory exhibits internal inconsistency, incoherence of interpretations, and inconsistency with background knowledge (i.e., relativity physics), its explanatory power is weak. Thus, the realist can withhold judgments concerning its truth. Barrett concludes that the theory is empirically adequate, instrumentally reliable, and that is all one needs to characterize its success. He is correct—but not in his further conclusion that the case refutes scientific realism. The realist can explain the success of many theories in terms of their empirical adequacy, while denying that they meet the standards required to warrant a belief in their truth. So for BCT realism, a commitment to the truth of a current scientific theory is restricted to theories which meet the highest standards of explanatory and predictive success—and this implies that not every best current theory is one whose truth the scientific realist needs to affirm. This observation protects the empirical adequacy of scientific realism—to the extent that there are theories with some great degree of predictive success—but conceptual difficulties that preclude a realist commitment.
References Barrett, J. (2002). Are our best physical theories (probably and/or approximately) true? In S. D. Mitchell (Ed.), PSA 2002: Proceedings of the 2002 biennial meeting of the philosophy of science association (Vol. 1, pp. 1206–1218). East Lansing, MI: Philosophy of Science Association.
123
316
G. D. Doppelt
Boyd, R. (1973). Realism, underdetermination, and the causal theory of evidence. Nous, 7, 1–12. Boyd, R. (1981). Scientific realism and naturalistic epistemology. In P. D. Asquith & T. Nickles (Eds.), PSA 1980 (Vol. 2). East Lansing, MI: Philosophy of Science Association. Boyd, R. (1984a). The current status of the realism debate. In J. Leplin (Ed.), Scientific realism. Berkeley: University of California Press. Boyd, R. (1984b). Lex Orandi est Lex Credendi. In P. M. Churchland & C. A. Hooker (Eds.), Images of science. Chicago: University of Chicago Press. Carrier, M. (1991). What is wrong with the miracle argument? Studies in History and Philosophy of Science, 22, 23–36. Carrier, M. (2004). Experimental success and the revelation of reality: The miracle argument for scientific realism. In M. Carrier, J. Roggenhofer, G. Ku¨ppers, & P. Blanchard (Eds.), Knowledge and the world: Challenges beyond the science wars. Berlin: Springer. Friedman, M. (1974). Explanation and scientific understanding. Journal of Philosophy, 71, 5–19. Glymour, C. (1980). Theory and evidence. Princeton, NJ: Princeton University Press. Kitcher, P. (1981). Explanatory unification. Philosophy of Science, 48, 207–231. Kitcher, P. (1993). The advancement of science. Oxford: Oxford University Press. Ladyman, J. (2008). Structural realism versus standard scientific realism: The case of phlogiston and dephlogisticated air (forthcoming in Synthese 2009). Theoretical frameworks and empirical underdetermination workshop at University of Du¨sseldorf, April 2008, University of Pittsburgh e-copy. Ladyman, J., Ross, D. (2007). Scientific realism, constructive empiricism, and structuralism. Chapter 2 of Everything must go: Metaphysics naturalized (pp. 66–129). Oxford: Oxford University Press. Laudan, L. (1981). A confutation of convergent realism. Philosophy of Science, 48, 19–49. Laudan, L. (1984). Explaning the success of science. In J. Cushing, et al. (Eds.), Science and reality. Notre Dame: Notre Dame University Press. McMullin, E. (1987). Explanatory success and the truth of theory. In N. Rescher (Ed.), Scientific inquiry in philosophical perspective. Lanham: University Press of America. Morrison, M. (2000). Unifying scientific theories: Physical concepts and mathematical structures. Cambridge: Cambridge University Press. Musgrave, A. (1976). Why did oxygen supplant phlogiston? Research programmes in the chemical revolution. In C. Howson (Ed.), Method and appraisal in the physical sciences. Cambridge: Cambridge University Press. Psillos, S. (1999). Scientific realism: How science tracks truth. London: Routledge. Pyle, A. (2000). The rationality of the chemical revolution. In R. Nola & H. Sankey (Eds.), After Popper, Kuhn, and Feyeraband: Recent issues in theories of scientific method. Dordrecht: Kluwer. Salmon, W. (1970). Baye’s theorem and the history of science. In R. Stuewer (Ed.), Historical and philosophical perspectives of science. Minnesota studies in the philosophy of science (Vol. 5). Minneapolis: University of Minnesota Press. Salmon, W. (1985). Empiricism: The key question. In N. Rescher (Ed.), The heritage of logical positivism. Lanham, MD: University Press of America. Salmon, W. (1990). Rationality and objectivity in science, or Tom Kuhn Meets Tom Bayes. In C.W. Savage (Ed.), Scientific theories. Minnesota studies in the philosophy of science (Vol. 14). Minneapolis: University of Minnesota Press. Worrall, J. (1989a). Structural realism: The best of both worlds? Dialectica, 43, 99–124. Worrall, J. (1989b). Fix it and be damned: A reply to Laudan. British Journal for the Philosophy of Science, 40, 376–388. Worrall, J. (1989c). Why both Popper and Watkins fail to solve the problem of induction. In F. D’Agostino & I. C. Jarvie (Eds.), Freedom and rationality: Essays in honour of John Watkins. Dordrecht: Kluwer. Worrall, J. (1989d). Fresnel, poisson, and the white spot: The role of successful predictions in the acceptance of scientific theories. In G. Gooding, et al. (Eds.), The uses of experiment. Cambridge: Cambridge University Press. Worrall, J. (1990). Scientific realism and the luminiferous ether: Resisting the ‘pessimistic meta-induction,’. Unpublished manuscript.
123