8  Science without Realism?

We should be careful to get out of an experience only the wisdom that is in it—and stop there; lest we be like the cat that sits down on a hot stove lid. She will never sit down on a hot stove lid again—and that is well; but also she will never sit down on a cold one any more.

—Samuel Langhorne Clemens (Mark Twain),

Pudd'nhead Wilson's New Calendar

The first five chapters of this book suggested not only that the problem of unconceived alternatives constitutes the most serious prima facie challenge for scientific realism but also that we have strong historical evidence of our repeated vulnerability to this challenge. That is, we have repeatedly failed even to conceive of equally well‐confirmed alternatives to our best theories that were sufficiently scientifically serious as to be actually accepted by later scientists and scientific communities. The last two chapters argued that realist responses to challenges grounded in the historical record offer little reason to resist even the original pessimistic induction, much less the problem of unconceived alternatives itself. That is, we have found no prospectively applicable way to distinguish particular present theories (or parts thereof) from their rejected predecessors in a way that should lead us to expect the ultimate fate of the former to differ from that of the latter, nor any reason to think it less likely that there are serious, well‐confirmed alternatives to our own best scientific theories that remain unconceived by contemporary theorists. Moreover, when we simply consider the specific evidence we have in support of some particular theory, both our brute intuitions and our considered reflective judgments that the available evidence renders the possibility of such alternatives remote in some particular case are demonstrably unreliable. Perhaps, then, it is time to ask whether it is possible to live without scientific realism and what the scientific enterprise might look like if we tried. That is, we might well wonder if any coherent positive image of scientific inquiry and its products remains open to us if we abandon what we described in chapter 1 as educated common sense about science: theview that our best scientific theories simply tell us (probably and approximately) how things really stand in otherwise inaccessible domains of nature.

We need not approach such a question empty‐handed. A longstanding minority tradition among scientists and philosophers of science alike has sought not only to explore the various challenges for scientific realism, but also to develop some alternative positive conception of the scientific enterprise itself. Most commonly, proposals for such alternatives have pursued some version of the idea that even our best scientific theories are simply tools or instruments for making empirical predictions and achieving other practical ends, rather than literal and /or accurate descriptions of otherwise inaccessible domains of nature, and such sentiments concerning scientific theories have a remarkably long intellectual pedigree. Indeed, Karl Popper's (1963 chap. 3) famous critique of this “instrumentalist” position as intellectually sterile counts Andreas Osiander (author of the unsigned Preface to Copernicus's On the Revolutions of the Celestial Spheres), Cardinal Bellarmino, and Bishop Berkeley as notable defenders of the view, even while resisting Duhem's claim to find its historical antecedents in classical Greek thinkers (for criticism of Popper's somewhat simplistic historiography, however, see Fine 2001). But even those who find themselves attracted to such an instrumentalist conception of science have remained deeply divided concerning just what status we are to ascribe to the claims of our best theories and how best to understand the thesis that they are no more than tools or instruments for accomplishing our practical objectives.

The nineteenth‐century physicist Ernst Mach, for example, grounded his distinctive and influential version of this instrumentalist position in a radical phenomenalism. He insisted that “[w]hat we represent to ourselves behind the appearances exists only in our understanding, and has for us only the value of a memoria technica or formula” (1911 49); thus the sole object of science is its “economical office” of “replac[ing], or sav[ing], experiences, by the reproduction and anticipation of facts in thought” (1893 577) and with “the least possible expenditure of thought” (1893 586). Mach argued, for instance, that “[i]n nature, there is no law of refraction, only different cases of refraction,” and therefore Snel's “law of refraction is a concise, compendious rule, devised by us for the mental reconstruction of” innumerable individual experiences (1893 582). Similarly, he regarded theoretical hypotheses simply as devices for the systematic classification, summary, organization and coordinated expression and prediction of innumerable particular appearances. Although Mach recognizes the existence of imperceptible degrees or instances of observed phenomena, he argues that in atomic theory, “atoms are invested with properties that absolutely contradict the attributes hitherto observed in bodies,” and thus that the atomic hypothesis is only “a mathematical model for facilitating the mental reproduction of facts”(1893 589). Theoretical concepts like ‘atoms’, Mach insists, are merely “provisional helps,” ultimately to be discarded in favor of more direct descriptions of phenomenological appearances—not because they seek unsuccessfully to describe a reality beyond appearances but rather because they successfully but only indirectly describe collections of coordinated and systematized experiences themselves.

By contrast, thinkers like Poincaré and Duhem were moved to instrumentalist sympathies by the same sorts of developments in physics around the turn of the twentieth century that provoked their concerns about underdetermination and the pessimistic induction (for more details, see Worrall 1982). By this time, the progress of physical science had begun to suggest that there might be quite genuine cases of differences between actual competing scientific theories that could not possibly be adjudicated by any straightforward appeal to empirical tests or observations. To use a famous example of Poincaré's (though not a case of actual competing theories), any set of measurements of the angles in a triangle marked out by appropriately oriented perfectly rigid rods can be accommodated by the assignment of any number of different combinations of underlying spatial geometries and compensating ‘congruence relations’ for the rods in question; if the sum of the angles differs from 180 degrees, for instance, we could either interpret the underlying geometry as Euclidean and conclude that the distance marked out by each rod varies with its position and/or orientation, or assume that the distance marked out by each rod remains constant and conclude that the underlying geometry of the relevant space is non‐Euclidean. Poincaré's response to this threatened form of underdetermination was to embrace conventionalism; that is, he regarded such theoretical matters as the assignment of a particular physical geometry to space as matters of choice or convention to be decided on grounds of greatest convenience. And this in turn implied, he suggested, that the quite useful ascription of a particular geometry to space by a theory should not be construed as literally attributing anything (truly or falsely) to nature itself: “[T]he question: Is Euclidean geometry true? … has no meaning. We might as well ask if the metric system is true, and if the old weights and measures are false … One geometry cannot be more true than another; it can only be more convenient” (Poincaré [1905] 1952 50).

Duhem sometimes drew a similar moral from his own distinctive worries about various forms of underdetermination, as well as from concerns about the pessimistic induction and the role played by idealization in physical theories: at various points he suggests that “hypotheses [are] not judgments about the nature of things, only premises intended to provide consequences conforming to experimental laws” ([1914] 1954 39), and that “propositions introduced by a theory … are neither true nor false; they are only convenient or inconvenient” ([1914] 1954 334). On the other hand, both Duhem and Poincaré insisted explicitly that theskeptical attitude they grounded in the repeated and radical discontinuities between the theoretical conceptions of nature accepted at different times across the history of scientific theorizing was of limited scope. It applied only to our efforts “to penetrate beyond the teachings of experiment or … surmise realities hidden under data observable by the senses” (Duhem [1914] 1954 274) when in fact these “merely nam[e] … the images we substituted for the real objects which Nature will hide for ever from our eyes” (Poincaré [1905] 1952 161). That is, both thinkers retained full confidence in the “experimental laws” or generalizations about observable phenomena uncovered by scientific investigations, even as each denied that such investigations were able to disclose (or that “mathematical theories” described) the actual underlying constitution of nature itself. Poincaré concludes that “we must not ask from [mathematical theories] what [they] cannot give us” and that it would be “an unreasonable demand” to expect our theories “to reveal to us the real nature of things”; instead “[t]heir only object is to co‐ordinate the physical laws with which experiment makes us acquainted … ” ([1905] 1952 211). Duhem argues likewise that the only legitimate aim of science is to “represent as simply, as completely, and as exactly as possible a set of experimental laws” ([1914] 1954 19), and famously goes so far as to insist that our attempts to explain the empirical regularities we uncover are regarded as the proper purview of speculative metaphysics rather than empirical science. It is only when we try to “strip reality from the appearances covering it like a veil, in order to see the bare reality itself” ([1914] 1954 7), he insists, that neither our own nor any genuinely scientific theories can do what we ask of them.

Even this whirlwind tour through some instrumentalist themes in the writings of Mach, Duhem, and Poincare suffices to illustrate that it has been no easy matter to parlay the intuitive idea that scientific theories are simply tools or instruments for achieving practical aims into a single clear and compelling vision of the scientific enterprise. It is quite striking, however, that even some of these early efforts to articulate an instrumentalist account of scientific theories share an explicitly semantic or linguistic character: faced with scientific claims that seem to concern events, entities, or phenomena beyond direct experience or observation, instrumentalists have repeatedly insisted either that such claims have a meaning very different from the one they seem to have or, if not, that claims of this sort can be eliminated from science altogether. On what we might call Mach's “reductive” variety of instrumentalism, for instance, the very meaning of theoretical discourse is exhausted by the implications it has concerning our sense experiences, and we may therefore simply believe the central claims of our best scientific theories after all, once they are correctly interpreted as only “really” saying something about collections of phenomenological experiences or appearances in the first place. Duhem and Poincaré seem instead to regard many scientifictheories as aspiring or attempting to describe an underlying, inaccessible reality and/or explain observable events or experiences by appeal to it and simply reject these ambitions as unsatisfiable or even unscientific in some way. Each proposes at some point a form of what we might call “syntactic” instrumentalism, on which theoretical claims are neither true nor false, nor even “really about” the world itself, but are instead simply convenient syntactic devices guiding our inferences from some observations or observable states to others. But each also at least suggests a quite different “eliminative” form of instrumentalism, on which such theoretical claims really do make just the descriptive claims about inaccessible domains of nature that they seem to but can nonetheless be eliminated altogether from our scientific descriptions of the world. Thus, instrumentalist sympathies have produced a wide variety of importantly divergent attitudes (sometimes within the works of a single author) toward the cognitive, semantic, and epistemic status of theories, including the view that extant scientific theories do not in fact make claims about inaccessible realities behind observable phenomena, that the scientific enterprise need not do so, and that it should not.¹

Perhaps it is less surprising that such a generally linguistic or semantic strategy of analysis appealed to logical positivist and logical empiricist thinkers through the middle decades of the twentieth century. The sort of reductive instrumentalism favored by Mach would engage the sympathies of those positivists and early empiricists who sought to carry out a reduction of all scientific language to a privileged phenomenological or observational basis, a project pursued most notably by the early Carnap, but also influentially by Bridgman. And when this reductive enterprise encountered the apparently insurmountable technical obstacles that led it to be abandoned even by its original architects, some positivist and empiricist thinkers would retreat to syntactic instrumentalism's distinctive claim that theoretical propositions are devoid of any meaning at all beyond the license they provide to infer from one observable state to another. That is, in the spirit of some of Duhem and Poincaré's remarks, these thinkers argued that despite appearances to the contrary the claims of theoretical science are nonassertoric: they are not claims about what the world is like, they carry no straightforward ontological commitments regarding unobservable entities, and they do not possess truth values at all.

Of course, this syntactic form of instrumentalism demands an extremely counterintuitive view of the semantics of theoretical discourse. Our scientific theories certainly seem to be making actual descriptive claims about what the world is like and to be put forward with the intention of so doing: if they make the claim that material objects like tables and toasters are made up of atoms constituted by electrons orbiting nuclei, this seems for all the world to say something about the actual constitution of tables and toasters and not simply that someobservable states of affairs may be safely inferred from others. Thus, many positivists would evade syntactic instrumentalism's implausible account of the semantics of theoretical claims by embracing the eliminative version of instrumentalism instead: equally in the spirit of suggestions we have noted from Duhem and Poincaré, these thinkers insisted that while theoretical discourse is both meaningful and irreducible, it can nonetheless be eliminated from science altogether. This eliminative form of instrumentalism gained further currency among logical positivists and logical empiricists following such technical developments as the invention of the Ramsey sentence and the proof of Craig's Theorem. The import of the latter seemed to be, as Hempel explained in introducing his famous “Theoretician's Dilemma,” that “any chain of laws and interpretive statements establishing [definite connections among observable phenomena] should then be replaceable by a law which directly links observational antecedents to observational consequents” ([1958] 1965 186). But the significance of Craig's theorem was immediately controversial: perhaps most influentially, Earnest Nagel (1961 136–137) pointed out several features of the theorem ensuring that it would be of quite limited relevance to the eliminability of actual theoretical discourse from science. And more recently, the profound differences between actual scientific theories and the sorts of artificial formal systems to which tools such as Craig's Theorem and Ramsey's technique apply have led these results to be regarded as having dubious relevance for the genuine prospects of instrumentalism.²

Thus, recent history has not offered much encouragement to those who would escape realist commitments either by reinterpreting the central claims of our scientific theories or trying to remake those theories so as to eliminate such claims. Philosophers of science have encountered what seem to be insuperable obstacles at every turn in trying to show either how the claims of theoretical science can be reduced or interpreted away or how science can be sanitized so as to remove claims about theoretical and/or unobservable entities from it. The challenges facing any such linguistic or semantic strategy for developing instrumentalism now seem sufficiently serious and its compensating attractions sufficiently slight that most contemporary philosophers, not to mention scientists themselves, regard such an enterprise as somewhere between hopeless and bankrupt.

Despite the historical influence they have enjoyed, however, such linguistic or semantic approaches to instrumentalism simply do not exhaust the available options for trying to develop the idea that even our best scientific theories are tools or instruments for guiding our practical engagement with the world rather than literal and/or accurate descriptions of otherwise inaccessible parts of that world. And we will here explore an alternative approach that is of quite a different character, although it is equally in the spirit of suggestive remarks offered byDuhem and Poincaré as well as others since. On this alternative, we might allow that our scientific theories really do make just the descriptive claims about inaccessible reaches of nature that they seem to make, and even that the fruitful investigation of nature by human beings typically requires that we propose and explore claims of this character, but nonetheless insist that we can make perfectly good practical use of the claims of such theories without believing what they say about the natural world. This approach might be called “epistemic” by contrast with earlier “semantic” or “linguistic” approaches to instrumentalism, because it restricts the set of beliefs to which we regard ourselves as entitled by the dramatic empirical successes of our best scientific theories, rather than restricting or reinterpreting the very claims that those theories are understood to be making in the first place. And the fundamental commitment of this distinctively epistemic form of instrumentalism would seem to be shared (although without this suggestive unifying label), by any number of more recently influential alternatives to scientific realism, including those developed by Thomas Kuhn, Bas Van Fraassen, and Larry Laudan.

Arthur Fine offers a useful general characterization of this epistemic variety of instrumentalism in describing how its commitment to scientific claims falls short of endorsing their truth:

Instrumentalism satisfices; it goes for something less than truth. In general, instrumentalism would be satisfied with the instrumental reliability (in the broad sense) of a theoretical story, and it treats reference in this same reliabilist way. Thus, we might say that instrumentalism treats scientific stories as though they were true, just in so far as it relies on them, and their postulated entities, as useful guides for whatever practical and theoretical jobs may arise. … One sometimes hears this expressed by saying that instrumentalism treats theoretical entities as fictions. But, if genes (for example) are to be called fictions, then one must be clear that they are not mere fictions, nor is genetic theory to be assimilated to some amusing piece of scientific fantasy. (1986 157)

The fundamental idea at work here would seem to be that we are to make use of our best scientific theories for practical purposes, but not to believe what those same theories tell us about the natural world. Although this suggestion might seem intuitively clear enough, closer scrutiny reveals that it will not suffice as an adequate general characterization of the epistemic instrumentalist approach. The central problem is that to use a theory for prediction, intervention, and other practical ends just is to believe at least some of the things it tells us about the world—for example, that the shuttle really will fall back to Earth if the booster rockets fail to ignite, or that a certain drug really will alleviate the symptoms of a given disease, or that we really should expect to find fossils of a particular kind in a particular geographic location. So the simple distinctionbetween making use of a theory and believing what it says cannot deliver what the instrumentalist needs.

If the instrumentalist must believe at least some of what a theory says about the world in order to make effective instrumental use of it, it might seem natural enough to try to find some principled way to distinguish those claims she must believe in order to make use of a theory from those that she need not. And it would seem that our commitment to the instrumental reliability of a theory means (at least) that we trust the guidance it offers to our practical or pragmatic engagement with the world around us. That is, perhaps what it means to be an instrumentalist about any particular theory is to believe the empirical predictions and recipes for intervention that the theory offers, but not the description of some part of nature in which those pragmatic recommendations are grounded. This suggestion seems to reflect at least the spirit of some of the most influential recent proposed alternatives to scientific realism. Kuhn, for example, famously claims that a later scientific theory is typically “a better instrument for discovering and solving puzzles” than its predecessors, offering more impressive “puzzle‐solutions and. …  concrete predictions,” but denies that this is because such later theories are “somehow a better representation of what nature is really like” ([1962] 1996 206). Likewise, Laudan denies that we should view successive theories as achieving greater truthlikeness or more closely approximating the truth about nature, but insists that the scientific enterprise is nonetheless progressive because successive theories typically improve over time in their ability to solve the various empirical and conceptual problems that we face (1977, 1996). To just the extent that Kuhn's “puzzles” and Laudan's “problems” involve practical contexts of prediction and intervention, these sentiments would seem to incorporate the approach to epistemic instrumentalism sketched above.

While this proposal shares the intuitive appeal of the idea that we might use our theories without believing them, it is ultimately no better able to provide a satisfying general characterization of the epistemic instrumentalist's position. For one thing, many of the predictions a theory makes just are descriptive claims about remote and inaccessible aspects of nature, and are thus part of its theoretical description of the world. Howard Stein poses this problem in a particularly elegant way, paraphrasing a remark of Eugene Wigner's pointing out that one also “uses quantum theory, for example, to calculate the density of aluminum” (1989 49). To take another example, among the most important predictions of our present cosmological theory are the characteristics it specifies for the field of so‐called “dark energy” posited to explain the present expansion of the universe. Still more important, however, is the fact that a theory's predictions of even the most mundane and familiar events (like earthquakes, supernovae, or tumors) will be shot through with just the same theoretical apparatus it uses to describe the worldquite generally: what the theory actually predicts is the shifting of the Earth's tectonic plates, the violent explosion of a distant star, or the rapid production and accumulation of genetically abnormal cells. That is, a theory's predictions of events, entities, or phenomena and its recipes for intervention with respect to them are typically offered in precisely those terms the theory uses to describe the world generally, and so cannot be more epistemically secure than the theory's own description of nature: we can hardly say that we believe a theory's prediction that a “supernova” will occur at a given time and place, for instance, but not what it says supernovae are. And it is far too late in the day to suggest that we might try to retreat into some kind of pure observation language that involves no theoretical commitments at all, using the theory to predict and intervene with respect to needle‐readings, sensations, or colored patches in the visual field rather than earthquakes, supernovae, or tumors. There is no such pure observation language, and even if there were we could not use it to describe the sorts of perfectly familiar phenomena, entities, and events with respect to which even the instrumentalist thinks our theories really do allow us to predict and intervene successfully.

We might usefully return here, however, to Kitcher's point about the diverse ways in which the reference of various tokens of a natural kind term can be fixed on different occasions of use. Recall that Kitcher hoped to use this distinction to try to show how some tokens of a natural kind term could be referential while other tokens of the very same term were not. Some of Priestly's tokens of “dephlogisticated air,” for example, had their reference fixed by his intention to refer to air once the substance emitted in combustion had been removed from it: since there is no substance emitted in combustion in the way he imagined, these tokens of the term simply failed to refer. On other occasions, however, Priestly's dominant linguistic intentions were quite different: Priestly used the term ‘dephlogisticated air’ to refer to oxygen, Kitcher suggests, when he conceived of it instead as the substance he had isolated whose inhalation was rendering his breathing particularly light and easy or the substance he ‘exploded together’ with ‘inflammable air’ to produce either water or nitric acid. Similarly, some of Fresnel's tokens of the term ‘light wave’ could refer to electromagnetic waves of high frequency, Kitcher suggests, while those whose reference was fixed by Fresnel's intention to talk about oscillations of the molecules of the ether simply did not refer at all.

In chapter 6 we noted that Kitcher's distinction offers cold comfort to the scientific realist because it runs afoul of the “trust” argument. That is, it allows Priestly to refer successfully using the term ‘dephlogisticated air’ while having a radically mistaken theoretical description of this substance, which simply divorces the question of successful reference from the real issue concerning realism: whether or not we shouldtrust the fundamental descriptions of nature offered by our best scientific theories. But Kitcher is nonetheless quite right to point out that the reference of many tokens of ‘dephlogisticated air’ simply didn't depend on the descriptive accuracy of Priestly's phlogiston theory in any significant way. When Priestly noted that the dephlogisticated air he had isolated rendered his breathing particularly light and easy, nothing in this claim depended on the accuracy of his chemical theory for the successful reference of its central terms and the claim is itself perfectly intelligible in a way that is independent of the theoretical descriptions Priestly would have associated with them. What Kitcher's distinction points out, then, is that our theoretical descriptions of the natural world, although ubiquitous in their influence, are not always or uniformly involved in the same way when we characterize events, entities, and phenomena in the world around us and the relationships between them.

These considerations suggest a natural way to characterize at least somewhat more precisely the suggestion that we might use our theories for prediction, intervention, and other pragmatic purposes without believing the theoretical descriptions they offer of the natural world. Though we cannot make use of our scientific theories for the pursuit of our practical endeavors without believing some part of what they say, perhaps it is open to us to believe the claims about entities, events, and phenomena that they make as those very claims can be understood independently of the theory or theories toward which we are adopting an instrumentalist stance. Take, for example, the claim that “the bithorax phenotype in Drosophila melanogaster is caused by a single mutation in the HOM complex of homeobox genes.” This highly theory‐laden claim carries a specific set of implications concerning entities, events, and phenomena we would expect to find in particular contexts as we understand them independently of contemporary genetic theory: that when particular materials from the bodies of organisms exhibiting a distinctive anatomical characteristic are subjected to any one of a specified set of laboratory procedures they will also exhibit a distinctive pattern on the resulting autoradiograph or electrophoresis gel, that a particular complex pattern of statistical dispositions will govern the appearance of that trait in the organism's descendants, and so on. This is certainly not to make the operationalist suggestion that this is what it “really means” to say that bithorax is caused by a single mutation in the HOM complex. Nor is it to suggest that we could ever hope to give any account either of what the various laboratory procedures for identifying the presence of a particular gene have in common or of the open‐ended list of possible exceptions to usual patterns of Mendelian inheritance without making use of the theory itself. To construe theoretical claims only in terms of their theory‐independent practical consequences in particular concrete contexts is simply not an effective way to try to think about nature; it is instead a terrible way to do so, and this is part of why the instrumentalistinsists no less than the realist that we should be using our best scientific theories for this task instead! Identifying the various ways in which a theory's concrete consequences can be grasped independently of the theory itself is important only when we seek to specify the beliefs about the world to which we remain committed once we recognize that we are not epistemically entitled to belief in a particular theory even though we know of no better way to pragmatically engage a given aspect or domain of nature than by means of it.

Of course, if a theory makes claims about entities or events to which we have no access independently of the theories that describe them or claims for which there is no sense to be made in a way that is independent of those descriptions, the instrumentalist will not believe them at all, but will simply make use of them in the process of generating further predictions and recipes for intervention whose content does not depend exhaustively on the theory's own descriptive apparatus in this way. Contemporary particle physics, for example, does not permit quarks to be isolated and so posits gluons to bind quarks within a proton, but we have no point of contact with gluons that is independent of the theory that posits them in the way that Priestly did with his dephlogisticated air or contemporary molecular geneticists do with the Drosophila HOM cluster. Thus, the instrumentalist will not believe any of the claims of contemporary particle physics regarding gluons, but will nonetheless be perfectly happy to make use of such claims in the course of generating further predictions and recipes for intervention concerning phenomena with which she does have some contact or commerce in a way that is independent of the theory itself.

This conception of instrumentalism seems to me quite closely connected to a point Fine makes in emphasizing why the instrumentalist need not accord any special epistemic significance to the distinction between observables and unobservables in articulating her position. As he notes, the distinctive fundamental commitment of instrumentalism is simply to the reliability of a given causal story, whether that story concerns observable or unobservable entities:

For according to instrumentalism what we want from our theories, posits, ideas, etc. in all the various contexts of inquiry is instrumental reliability. That is, we want them to be useful in getting things to work for the practical and theoretical purposes for which we might put them to use. This is the guiding pragmatic ideal of instrumentalism, and it treats all entities (observable or not) perfectly on par. … Of course if the cause happens to be observable, then the reliability of the story leads me to expect to observe it (other things being equal). If I make the observation, I then have independent grounds for thinking the cause to be real. If I do not make the observation or if the cause is not observable, then my commitment is just to the reliability of the causal story, and not to the reality of the cause. (1991 86)

The attraction of this picture is that it offers the instrumentalist a single systematic and principled commitment to the mere reliability of all of the claims made by a given scientific theory: it allows differences in our further beliefs to emerge simply from the fact that such reliability will be manifested differently (and we will therefore be able to accumulate different sorts of further evidence) for claims about observable rather than unobservable entities. On Fine's account, if the distinction between observables and unobservables is itself vague or indeterminate or a matter of degree, this might well impact the further beliefs the instrumentalist (or any of us) is able to form on the evidence of our senses, but it will not even be relevant to the fundamental commitment of her instrumentalism, which is simply to the reliability of the scientific theory in question across the board. In a like manner, we may resist the natural but ultimately misguided temptation we noted earlier to divide up the theory's own claims into those we need and need not believe in order to make effective practical use of it. What we want to say instead is that our fundamental commitment to the reliability of a given theory in turn commits us to the truth of whatever implications it may have for entities, events, and phenomena as they are conceived of outside of the theory itself (and indeed outside of all those theories toward which we are adopting an instrumentalist attitude). But of course, this is simply a generalization of Fine's approach to observability: when our theories carry implications concerning features or aspects of entities, events, and phenomena to which we have some independent route of epistemic access, our belief in the instrumental reliability of those theories leads us to expect these implications to turn out to be strictly and literally true. We will trust the remainder of a theory's implications to be reliable as well, but this will not mean believing any specific empirical claim about the world itself; it will mean only that they can be effectively used in generating further implications concerning matters (like the lightness of our breathing or the pattern on an electrophoresis gel) that can be ascertained in a way that is independent of the theory in question. The difference is not one of the epistemic attitudes we take toward the various claims of the theory itself, but rather of what other routes of epistemic access we may have to the furniture of the natural world about which the theory makes its presumptively reliable claims.³

This is emphatically not to say that the instrumentalist must or even could characterize the claims about successful prediction and intervention that her best theories lead her to believe in a way that is independent of any theory whatsoever, much less in the terms of a mythical “observation language.” Suppose, as Quine suggested long ago, that the middle‐sized objects of our everyday experience are no less “theoretical” entities hypothesized to make sense of the ongoing stream of experience than are atoms and genes, notwithstanding their greater familiarity. Then, it is by means of theories in this broadest sense that we come tohave any picture at all of what our world is like, what entities make it up, and how they are related to one another. But recall that only some of these theories will be open to the distinctive challenge posed by the problem of unconceived alternatives that I have suggested lies at the heart of any serious objection to scientific realism. As we have noted from the outset, what Quine calls the hypothesis of the bodies of common sense does not appear to be vulnerable to this challenge and its descriptive apparatus is therefore available to the instrumentalist to use in characterizing the earthquakes, supernovae, and tumors with respect to which she believes her best scientific theories allow her to predict and intervene successfully. It is ultimately because Priestly's claim that inhaling the “dephlogisticated air” he isolated rendered his breathing light and easy for several hours can be quite naturally interpreted in terms of these bodies of common sense that we can and do continue to count it true long after we have rejected his phlogiston theory of chemistry as profoundly mistaken: notice that if a radical revision of contemporary chemistry were to deliver the verdict that that the substance Priestly inhaled was not what we think of as oxygen either (and indeed that there is no such thing), we would still have no trouble understanding Priestly's claim. Furthermore, it seems that this would be true even if we ultimately came to see this description of the events as radically mistaken, for instance by rejecting the idea that the “dephlogisticated air” Priestly isolated was even a substance. Likewise, if a radically different successor theory ultimately replaces contemporary genetics, we might find ourselves unable to endorse or even know how to evaluate the perfectly general claim that bithorax in D. melanogaster is caused by a mutation in the HOM complex of homeobox genes, but we will have no trouble understanding or endorsing our earlier claims about the banding patterns displayed on autoradiographs and gels and about statistical dispositions governing the reappearance of bithorax in the offspring of organisms exhibiting the trait even if we do not continue to speak in these terms.

Although this account of the matter rejects as unhelpful the very idea of a pure observation language or a foundational epistemic role for observability as such, it does help to explain why observation and other perceptual processes will typically play an important role for instrumentalists in fixing beliefs about the natural world. This is because our various sensory modalities will characteristically be among the routes most commonly used by us to secure a grasp of entities, events, and phenomena in ways that are independent of the theories toward which we adopt instrumentalist attitudes. Indeed, this role for our perceptual abilities is itself a part of the everyday hypothesis of the bodies of common sense as we invoke it to account for the experiences we have, and this is surely at least part of why scientists sometimes work quite hard to devise ways to render the entities posited by their successful theories accessible to more familiar processes of observation or detection. But there is no brightline to be found here between “observable” and “unobservable” entities or “observational” and “theoretical” terms and no foundational epistemic distinction between observational and other modes of access to the phenomena. Indeed, scientific investigation itself has revealed a wide variety of ways in which our own perceptual faculties can be misled or deceived and circumstances in which our ordinary experience of the world seems to tell us things that are simply not to be trusted. These failings of our perceptual abilities can be demonstrated in ways that do not depend on the theories that we use to discover them, just as a microscope's ability to increase optical resolution can be demonstrated independently of the theories we use to construct it: human beings can be shown experimentally to be poor judges of distance, time, and speed, for example,⁴ and we have learned to trust our demonstrably reliable instruments over our demonstrably unreliable phenomenological experience to judge these matters in the world given to us by our (theoretical) description of the bodies of common sense itself. Thus, the only epistemic privilege here attached to sense experience is the broadly empiricist insistence that the evidence of our senses is where we must begin in finding out what the world is like.

It is also worth noting that the development of any such “hypothesis of the bodies of common sense” is itself an ongoing enterprise. It is by no means equivalent to a “folk physics” or any other folk science, if such there be. Such “folk” theories are manifestly false accounts of the workings of some part of the natural world and they stand refuted by the empirical evidence available in the domain of the theory long before we concern ourselves with the problem of unconceived alternatives—but of course we think that the hypothesis of the bodies of common sense is true! Particular people can and surely do believe false things, but the hypothesis of the bodies of common sense neither deifies commonplace collective wisdom about the external world nor simply replaces it with the descriptions found in our best scientific theories: it is rather common sense refined and sophisticated to reflect everything we know about the world of everyday objects as we experience and interact with them. This hypothesis begins with the fundamental supposition that our sense experience is produced by an external world full of enduring objects that we perceive and with which we causally interact, and it is this supposition that seems to escape any serious historical challenge from the possibility of radically distinct unconceived alternatives. But this commonsense grasp of the familiar objects in the world around us is continuously revised in response to discoveries about it prompted by developments in the sciences themselves: that objects dropped from moving vehicles will in fact fall along parabolic trajectories rather than directly toward the earth, that whales share important anatomical similarities with other mammals and not with fish, that the people around us constantly reason in ways that do not respect basic principles of rationalinference.⁵ Many such discoveries about the objects and interactions in the world of our everyday experience would almost certainly not have been possible but for the pursuit and development of self‐consciously theoretical natural science, but we do not depend upon the sciences in order to understand the facts they report about the world of our everyday experience. Thus it is perfectly open to the instrumentalist to allow that our scientific theories are extremely powerful tools for successfully guiding our predictions and interventions with respect to earthquakes, supernovae, tumors, and the like as those entities are grasped through our everyday experience of the world, even while recognizing that our systematic knowledge of that everyday experience is itself refined, improved, sophisticated, and corrected by the further scientific investigations we have undertaken.

If we allow that our grasp of the world is, in this sense, “theories all the way down,” it might well be impossible to coherently adopt instrumentalism as an epistemic attitude toward any and all theories whatsoever. That is, if we grant that all of our efforts to understand the world around us and our place in it are fundamentally theoretical in the broadest sense of the term, then it may well be that there is no coherent way to adopt an epistemic instrumentalist attitude toward all of our theories simultaneously.⁶ It is fortunate indeed, then, that our epistemic position in no way demands this. But it is also far from clear that a suitably sophisticated version of the hypothesis of the bodies of common sense is the only theoretical resource that will remain available to the thoughtful instrumentalist in trying to characterize what she strictly and literally believes about the external world. The preceding chapters have been concerned to argue that the problem of unconceived alternatives is a genuine challenge for scientific realism and to expose both the breadth and depth of that challenge, but they have certainly not established that the challenge can never be overcome or that every single scientific theory must automatically fall victim to it. Though we have yet to encounter them, perhaps there are specific reasons to believe that some particular theories do not remain vulnerable to the same problem of unconceived alternatives that has plagued so many of their predecessors and contemporaries. If so, these theories will count among the descriptive resources to which a thoughtful instrumentalist can appeal in characterizing the claims about the world that she believes to be true literally and without qualification. If not, the instrumentalist will have to frame what she actually believes (when called upon to do so) in terms of the entities, events, and phenomena familiar from our everyday experience of the middle‐sized bodies of common sense. Indeed, it is in some sense only now that the really difficult work begins: the work of trying to decide whether there are cases or contexts of inquiry in which the available evidence is sufficient to rebut the historically established presumption in favor of the existence of serious unconceived alternatives to even ourmost successful scientific theories.⁷ This is to some extent an explicit invitation to establish realism piecemeal, by showing that our most general reason for rejecting scientific realism about theories can be defeated in particular cases. Although the two preceding chapters have tried to show why the sorts of evidence to which realists have traditionally appealed to distinguish the expected fates of present theories from those of their failed predecessors are simply not up to this job, the naturalistic spirit of our inquiry requires that we remain open to the possibility that we will find special reasons to doubt that some particular scientific theories remain vulnerable to the problem of unconceived alternatives that seems to attend our scientific theorizing quite generally. I remain cautiously optimistic that there will be something general to be said about our systematic vulnerability to the problem of unconceived alternatives (and we had better hope that there is, since our intuitions in particular cases concerning whether we are vulnerable to the problem or not have turned out to be so spectacularly unreliable), but this strikes me as important, outstanding, and much‐neglected work facing the contemporary philosophy of science.

What the instrumentalist must be able to do, then, is connect the descriptive and inferential machinery of the theory toward which she adopts an instrumentalist attitude to entities, events, and phenomena as they can be characterized independently in the terms of whatever theories she strictly and literally believes, so that the former can be used to guide our pragmatic engagement with the latter. To adopt an instrumentalist attitude toward the phlogiston theory of chemistry, for example, we must know how the phlogistonian chemist would characterize parts of the world that are otherwise familiar to us in terms of her preferred theoretical descriptions, how she would use the theory's descriptive and inferential machinery to move from its characterization of one concrete situation to another, and how the resulting theoretical characterization could again be connected to a description of some part of the world that we are able to strictly and literally believe. This is how the instrumentalist phlogistonian arrives at the belief that heating the reddish powder in the jar by the door will force air into a connected enclosed vessel filled with water and that this air will support subsequent respiration and combustion longer and/or better than ordinary atmospheric air: such claims can be embraced without believing any part of the phlogiston theory itself or accepting any of the distinctive theoretical characterizations it offers of the central notions involved. Thus conceived, a merely instrumentally useful theory is not a device for taking us from sensations to sensations or observables to observables, as instrumentalists have traditionally supposed, but instead from states of affairs characterized in terms we can strictly and literally believe to other such states of affairs. And it is in this precise sense that we can recognize that the intervening theoretical machinery is neither reducible nor eliminablewithout believing that the claims it makes about nature are true. If I can say how the belief in a theory would lead me to predict and intervene with respect to entities, events, and phenomena as they can be characterized independently of that theory (and in terms of one or more theories that I do strictly and literally believe), then I know what further beliefs I am committed to by thinking that the theory is instrumentally useful but not true.

If this view of the matter is right, the characteristic epistemic attitude at the heart of such instrumentalism is not only already familiar to us, but is also almost strictly identical to the attitude the realist herself adopts toward some particular scientific theories. Thus, we might ask what the scientific realist means when she allows that we can perfectly well make use of Newtonian mechanics to send rockets to the moon without believing the theoretical description of nature that it gives. She will, of course, completely reject the account of nature that Newton offered—gravity is not a force exerted by massive bodies on one another, there is no absolute space or time, and so on. But she nonetheless knows perfectly well how a Newtonian would apply the theory to make predictions about and intervene in the natural world of entities, events, and phenomena as they are given to her by other theories that she does believe. That is, the relativity theorist has an independent account of the cannonballs, inclined planes, and rockets to which she could apply the theoretical mechanics she believes to be literally true, but she also knows how a Newtonian would apply her own distinctive theory to characterize those very cannonballs, inclined planes, and rockets in terms of the masses, forces, collisions, etc. that would allow her to predict and intervene with respect to them. This has nothing to do with the fact that cannonballs, inclined planes, and rockets are observable entities: she knows equally well how a Newtonian would identify forces and masses so as to make predictions about the gravitational motions of subatomic particles. And over whatever domain she thinks the theory is (more or less) instrumentally reliable she can make use of it because she knows how to apply it like a Newtonian would to entities, events, and phenomena whose existence she countenances independently of the theory. But of course the instrumentalist, too, knows how the Newtonian (or the relativity theorist, or the molecular geneticist) will connect the terms of her favored theory with the entities, events, and phenomena that the instrumentalist herself independently recognizes and accepts: at least those familiar from her everyday experience of the world mediated by the hypothesis of the bodies of common sense, and perhaps more besides. Thus, it is precisely because the instrumentalist knows how a true believer would connect up the atoms, genes, and spacetime curvature of her scientific theories to the rockets, rabbits, and autoradiographs of her own everyday experience in innumerable concrete ways that she knows precisely what it is that she believes when she adopts an instrumentalistattitude toward those same theories. And she will agree that molecular genetics, or general relativity, or whatever, is an extremely reliable tool for guiding her engagement with those (independently‐characterized) objects without believing that it is therefore true. But of course this is just what the realist does in connecting the distinctive descriptive and inferential apparatus of Newtonian mechanics to entities, events, and phenomena she thinks are truly described by some other theory or theories and using Newtonian mechanics to predict and intervene with respect to them. That is, the instrumentalist will take precisely the same attitude that the realist applies in particular contexts to theories she has specific reasons to disbelieve (like Newtonian mechanics) or to theories that she does not know how to interpret realistically (like quantum mechanics) and simply apply that familiar attitude much more broadly than the realist does.⁸ Indeed, she takes the historical evidence for the problem of unconceived alternatives as a sufficient reason to apply this same familiar attitude to virtually every theory concerning the fundamental constitution of and dynamical principles at work in nature, absent a showing of some specific reason to refrain from doing so.

As this comparison suggests, it was perhaps always a mistake to think that our task was to identify some crucial difference between the epistemic attitudes of realists and instrumentalists toward theories or theoretical claims or theoretical knowledge as such. The characters traditionally identified as the realist and the instrumentalist both recognize theories that they strictly and literally believe to be true and theories that they think are merely instrumentally useful over a wider or narrower domain of nature. The instrumentalist simply assigns a much larger set of the theories we actually have to the latter category. Furthermore, she is willing to assign a particular theory to the latter category even when she has no competing account of the same natural domain that she assigns to the former instead: the instrumental utility of the theory requires simply that she be able to connect its descriptive and inferential machinery to some set of entities, events, and phenomena in which she strictly and literally believes as they can be characterized outside of that theory itself. The difference between the realist and instrumentalist, then, is not one of a global epistemic attitude toward “theories,” but rather a local difference in the specific theories each is willing to believe on the strength of the total evidence available. This contrast also makes clear that there is no privileged or foundational role for the hypothesis of the bodies of common sense or any specific hypothesis built into the instrumentalist's account of the world and our knowledge of it. Instead, it simply turns out that this particular theoretical account can survive the challenge posed quite generally by the problem of unconceived alternatives, and it is therefore available to the instrumentalist (perhaps along with other particular theories as well) for characterizing what she strictly and literally believes, including theconsequences or implications of those theories toward which she adopts a merely instrumentalist attitude instead.

While I have suggested that the characteristic instrumentalist attitude is familiar from the realist's own approach to theories like Newtonian mechanics, there will still be important differences between the epistemic position of a person who adopts such an instrumentalist attitude toward most or all of our leading scientific theories and that of one who does not. In most cases, the realist believes that the best‐confirmed account we currently have of a given natural domain is (probably, approximately) true, so she reserves the instrumentalist attitude for theories that have useful ranges of practical application despite the fact that they conflict with a better‐confirmed account and so are presumed to be false. When the realist uses her theories to try to discover new phenomena or make novel predictions in a given scientific domain, then, she will typically make use of a theory she thinks is true and not merely instrumentally useful for this purpose. By contrast, a person who adopts the instrumentalist attitude concerning most or all of our scientific theories must be willing to accept this attitude toward a particular theory even when she has no better‐confirmed account of the same natural domain to offer in its place. She will thus find herself discovering new phenomena and making novel predictions about the natural world using a theory that she does not think is true. But it does not seem that any subtle incoherence lurks in applying the sort of instrumentalist attitude the realist herself embraces in some cases to a theory of a given natural domain when we have no competing account of that domain that we strictly and literally believe to be true. It seems perfectly coherent to deny that our own best scientific theories offer us even approximately true accounts of how things stand in otherwise inaccessible domains of nature, but insist that they are nonetheless the best presently available places to look for guidance in planning our own engagement with the natural world: in trying to produce drought‐resistant crop plants, in trying to anticipate the distribution and characteristics of undiscovered fossils, in trying to send rockets to the moon.⁹

Furthermore, the epistemic instrumentalist will often be justified not only in exploring but also in believing the implications of our best existing scientific theory in such novel applications. After all, in typical cases such a theory has already proven to be a reliable guide to prediction and intervention across a wide range of phenomena in a given natural domain that are systematically related in intricate detail, and it is this track record that grounds whatever degree of confidence the instrumentalist has in the continued reliability of the theory with respect to novel applications and predictions of further phenomena in that same systematically related domain (for further discussion, see Stanford 2000). Although she suspects that the theory will ultimately be replaced by a still more instrumentally powerful alternative that is presently unconceived,she has no specific reason to think that any particular novel application of a theory will be one in which its demonstrated reliability with respect to the phenomena in its purview will founder.¹⁰ And, of course, she knows of no other ground for her expectation about the outcome of some novel case with even nearly as much to recommend it as that offered by her most instrumentally powerful theory.

Furthermore, the instrumentalist knows that the most effective strategy available for trying to actually reach any presently unconceived instrumentally superior alternative will typically be articulating, exploring, extending, and testing the best theory she currently has. Indeed, we must not make the mistake of imagining that a theory is simply a tool for moving inferentially from some states of affairs we strictly and literally believe to others: this would be, as Stein insists, to make “a false estimate of the scope of a theory as an instrument” (1989 49). We have already noted that the instrumentalist recognizes our leading scientific theories as the best conceptual tools we have for thinking about nature, and this recognition encompasses the fact that they are the most powerful resources available to us in extending, pursuing, and expanding our further inquiry into the natural world itself:

The instrumentalist himself, if he wishes to do justice to the role actually played by theories in science, had better extend still further his conception of what theories are instruments for, to include their role as resources for inquiry; especially as sources of clues in what Peirce called “abduction”: the search for good hypotheses. (Stein 1989 52)

That is, the sophisticated instrumentalist recognizes that our leading scientific theories are the best foundation and starting point we have not only for uncovering new and unexpected phenomena, but also for opening up new areas and paths of inquiry, and in guiding ourselves to the even more powerful conceptions of natural domains that will ultimately replace the ones we now have.

We might illustrate this point in connection with the instrumentalist's more general convictions by returning to the case of Weismann's theory of the germ‐plasm, for there is no question that systematically exploring, challenging, and defending this view turned out to be critical to developing many features of the subsequent theories of inheritance and ontogeny that would ultimately come to displace it. Of course it seems quite right to endorse what would have been the instrumentalist's judgment at the time: that Weismann was in possession of a powerful conceptual tool for conceiving of and guiding our practical engagement with a particular domain of natural phenomena. This tool directed our efforts at prediction and intervention effectively with respect to a wide variety of the phenomena of inheritance and generation, from patterns of individual variation and reversion to the formation of sex cells (including Weismann's famous early prediction of the need for reduction division)to the course of embryological development. In this respect it offered some genuine advantages over the existing alternatives; it alone, for example, seemed able to allow simultaneously for both nuclear control of the cell and cellular differentiation. These advantages and the theory's systematic practical successes surely justified Weismann's willingness to make use of the theory in pragmatic contexts of prediction and intervention with respect to further phenomena of inheritance and development: for example, to insist (rightly) that acquired characteristics cannot be inherited and to predict (mistakenly) the existence of a special accessory germ‐plasm in cells capable of regeneration.¹¹ But for all that, an instrumentalist contemporary would have been equally right to think that Weismann was not in possession of even an approximately accurate description of the domain whose constitution and mechanics he sought to describe. The theory of the germ‐plasm was systematically mistaken in any number of its most fundamental features, not the least of which were those respects in which we have noted Weismann failed to conceive of important later alternatives: that the germ‐plasm is an expendable rather than productive resource for the cell which must be disintegrated over the course of ontogeny and used up in effecting cellular differentiation, that cells with different ontogenetic fates must contain different controlling germinal materials, that germ‐plasm must be reserved in a special inactivated state not only for inclusion in the sex cells but also to sustain the permanent possibilities of regeneration and dimorphism, and so on. And one of the most important constructive roles for the theory was its suggestive implications regarding these further phenomena and further directions of research that would prove fruitful, not only for refining the theory and improving our practical abilities of prediction and intervention, but also for opening up the conceptual space of theoretical alternatives that would ultimately lead to the replacement of the theory itself.

Likewise, the epistemic instrumentalist of the present day will suggest that we should certainly allow that our own best theories are effective guides to our practical projects of prediction and intervention and that we are justified in using them rather than known inferior alternatives to structure our engagement with a wide variety of natural phenomena, but she will remain convinced that they can accomplish all of this (just as Weismann's germ‐plasm theory did) while being profoundly misguided about fundamental aspects of the constitution and dynamics of the domains of nature they describe (just as Weismann's theory was). She will insist that our inability to exhaust the space of serious and well‐confirmed theoretical possibilities has left us with many theories that are neither literally nor even approximately true, but that are nonetheless able to serve us extremely well in our practical engagement with the world even as they guide the further research that will enable us to expand our view of the theoretical possibilities andultimately lead us to replace existing accounts of nature with even more powerful successors.

Indeed, there is a crucial further difference between instrumentalism and realism in that we should expect the instrumentalist to be far more willing to devote serious time and effort to the exploration of fundamentally distinct alternatives to our current theoretical conceptions of nature, as opposed to simply working to sophisticate or supplement our best current theories. For the realist, such exploration will, in general, have quite limited interest or motivation, for its rationale is found only in the relatively small chance she recognizes that contemporary theories might ultimately turn out to be fundamentally mistaken. By contrast, the instrumentalist fully expects even the best current theories to ultimately be replaced by even more powerful and fundamentally distinct conceptual tools for thinking about nature, and so she has every reason to invest serious time and effort in searching for these more powerful successors. In this sense, the instrumentalist is substantially more committed to the open‐ended character of scientific inquiry than her realist counterpart.

When instrumentalism is conceived in this robust manner, there would seem to be little reason left to endorse Popper's claim that it is a:

narrow and defensive creed according to which we cannot and need not learn or understand more about the world than we know already. A creed, moreover, which is incompatible with the appreciation of science as one of the greatest achievements of the human spirit. (1963 103)

The form of epistemic instrumentalism I have suggested does not propose that we simply follow along after our realist aspirations have given us powerful theories and timidly restrict our beliefs to their immediate practical utility in a lawyerly fashion. Its conception of inquiry is neither intellectually stultifying nor dismissive of the very real achievements that scientific theorizing has unquestionably delivered to us. Instead, the epistemic instrumentalism we have considered recognizes that boldly proposing hypotheses about the inaccessible workings of nature and then exploring and testing their implications has served as the preeminent engine of intellectual progress in the modern age. It acknowledges that scientific theorizing has produced and shows every promise of continuing to produce increasingly powerful and sophisticated tools for guiding our engagement with nature, and it regards us as eminently justified in continuing to explore the natural world by searching for more and better conceptual tools of just this same sort in just this same way. The epistemic instrumentalist will insist no less than the realist that we continue to challenge our best scientific theories by uncovering and testing further empirical implications they have, that we use them to unearth new phenomena and new ways to predict and intervene in the course of events around us, that these theories serve as the appropriate starting point in trying to determine how they themselves can be refined, improved, anddeveloped, and perhaps even that we work to unify our various scientific theories with each other and with whatever else we believe to produce a single coherent, consistent, and systematic account of the natural world as a whole.¹² She will not demand that a theory meet some looser standard of evidence in order to be accepted, or pursue the further implications of our theories less doggedly, or invest those implications with less significance, than the realist, for she is after the most powerful, most stringently tested, most broadly applicable, and most predictively accurate set of conceptual tools that are to be had. In short, the instrumentalist is in a position to take the claims of our best scientific theories about nature every bit as seriously as the realist does, even as she declines to believe everything about the world that the realist believes.

At the end of the day, then, perhaps we should worry instead that taking the claims of our scientific theories so seriously threatens to turn epistemic instrumentalism into a position that is only verbally distinct from scientific realism itself, as Nagel (1961 139) once famously suggested concerning instrumentalism of the distinctively semantic or linguistic variety. Such a claim will appeal only to those who share the mistaken conviction that an embrace of the realist label is the only possible way to express their enthusiasm for and commitment to the importance of the scientific enterprise, rather than the belief that the present accounts of nature produced by that enterprise are probably and/or approximately true. For any realist who embraces the substantive features of the view developed here will find that she has given up the central commitments animating the realist account of scientific inquiry and its products. To be sure, the realist and instrumentalist agree in embracing the theoretical proposition that there is an external world full of rocks, spiders, and distant stars as well as in any number of further beliefs about the entities, events, and phenomena in that world whose truth do not depend on the accuracy of our more detailed scientific theories: that the rocks are hard, that the spiders spin webs, and that the distant stars shine at night. They also agree that our scientific theories are the very best tools we have for predicting and intervening with respect to these rocks, spiders, and distant stars, for discovering new phenomena whose existence we did not previously suspect, and for extending our investigation of nature. But in the same straightforward sense in which they share these beliefs, the realist holds many further beliefs about those entities, events, and phenomena that the instrumentalist declines to share: that the rocks are made up of atoms with a specific internal composition, that the spiders share a common ancestor with her in the distant past, that the path of the light from the distant stars can be bent by a gravitational field. Moreover, the scientific realist believes that there are entities in the world answering to the detailed descriptions given by her best scientific theories of entities like gluons, genes, and dark energy; by contrast, the instrumentalist believes that although these theoreticaldescriptions are intimately involved in the most powerful conceptual resources we have for thinking about particular natural domains, this does not give her a compelling reason to believe that there are entities in the world answering to those descriptions.

Furthermore, the realist and instrumentalist differ perhaps most radically and most fundamentally in their conceptions of and expectations concerning the future of scientific inquiry itself. The realist supposes that our further inquiry into the natural world will continue to bear out at least in broad strokes the various conceptions of domains of nature that are articulated by our current scientific theories. But the instrumentalist offers us a very different picture of the future of human scientific inquiry. She judges it quite likely that even the most genuinely impressive and instrumentally accomplished theories of contemporary science will ultimately be replaced by more powerful conceptual tools offering fundamentally different conceptions of nature that have presently not yet even been conceived.¹³ She rejects 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. And among sensible people who are rightly impressed by the dramatic empirical successes of the best scientific theories we have, what greater difference could there be?