Why and How to be a Scientific Realist
Every Sunday at 11:30 in the morning, children all over Germany tune their TV sets to public television’s first program in order to watch Die Sendung mit der Maus—The Program with the Mouse—and, when I’m in Germany, so do I. The Program with the Mouse has many virtues, but only one of them concerns us here: Its human participants, Armin and Christopher, regularly explain things to their many curious young viewers, both familiar objects and puzzling phenomena. They have demonstrated, for instance, how items as diverse as marshmallows, paper clips, blue jeans, church bells, and jumbo jets are manufactured. They have explained how the holes get into Swiss cheese and why black soap makes white foam. And, much to my delight, just when I was thinking about scientific realism, I tuned in one Sunday and there was Armin explaining photoluminescence, that is, why the moon- and star-shaped stickers on my bedroom ceiling continue to glow in the dark for a while after I’ve turned off the lights even though they’re not connected to any source of electricity.
Armin’s account of the photoluminescent stickers included considerable detail about their manufacture—the moon and star shapes are printed on a plastic sheet, then coated with a photoluminescent chemical powder mixed with an adhesive material, and finally punched out by a sort of cookie-cutter. He also spelled out in some detail the mode of production and chemical composition of the photoluminescent powder itself—first, hydrogen sulfide gas is passed under pressure through liquid zinc chloride, resulting in a precipitate of zinc sulfide; this precipitate is then dried, mixed with copper, chlorine, and salt, powdered, and finally heated to produce the desired substance. But, as interesting as it was to learn all this, it ultimately brought little in the way of satisfaction. What I really wanted to know, of course, is why the powder glows in the dark, and, with typical Mouse program thoroughness and ingenuity, Armin proceeded to explain that too.
The explanation ran essentially as follows: Like all chemical substances, the luminescent powder consists of innumerable tiny little atoms, each of which has a central nucleus around which swarm electrons. When light shines on the powder, it imparts excess energy to the electrons. They are moved up to higher energy levels. When the lights are switched off, the electrons drop back down to their original energy levels, releasing their excess energy in the form of light. The powder glows only until the electrons no longer deliver any excess energy.
This
explanation was illustrated with an elegant model in which the energy
produced by photoelectric cells first ran an electric motor that was used to
spin up a pair of hanging weights representing the electrons. When the photoelectric cells were
disconnected, the still-spinning weights drove the motor, which then functioned
in reverse as a dynamo, for a time generating enough current to light a small
lamp representing the glowing powder.
Light energy was thus converted into angular momentum, which was then
gradually converted back into light energy.[1]
I want to make a number of claims related to this example: First, one of the things we want natural science to do for us is to explain, for instance, why certain stickers glow in the dark or, more generally, to explain the phenomenon of photoluminescence per se. That is a task that we can legitimately both demand and expect our scientific investigation of the world to ultimately discharge. Second, the account offered by The Program with the Mouse satisfies this demand, but only if we believe it. If we don’t think that there are any electrons, then the Mouse account, as I shall call it, is useless. It explains nothing. Third, the fact that the Mouse account satisfactorily explains photoluminescence is a good reason for believing that there are electrons. There are many such good reasons for believing that there are electrons. Fourth, the many good reasons for believing that there are electrons are, of course, neither severally nor collectively conclusive reasons for believing that there are electrons. None of them logically compels us to believe that there are electrons, and it could in principle turn out that there are even better reasons for believing that there aren’t any electrons. In that case we would need two new explanations: first, a different explanation of photoluminescence and, second, an explanation of why the Mouse account worked as well as it did, i.e., of why there seemed to be electrons.[2]
As I understand scientific realism, a scientific realist is someone who accepts generalizations of these four claims, namely:
(1) A central task of scientific inquiry is to provide satisfactory explanatory accounts of phenomena.
(2) Such explanatory accounts sometimes posit the existence of theoretical entities.
(3) The fact that an account which posits the existence of theoretical entities satisfactorily explains particular phenomena is a good reason for believing that the posited entities exist.
(4) All such explanatory accounts are always empirically defeasible.
As I understand scientific realism, then, a scientific realist needn’t accept either of the following claims:
(a) “The terms of a mature scientific theory typically refer.” (Richard Boyd)
(b) “The theories accepted in a mature science are typically approximately true.” (Hilary Putnam)
For reasons that I shall explain shortly, I would resist any characterization of scientific realism that made use of unexplicated notions of reference or truth. My own characterization is, I think, consistent with Wilfrid Sellars’ claim that “To have good reason to accept a theory is to have good reason to believe that the entities it postulates are real”, but since I think that ‘real’ is ultimately in the same boat as ‘reference’ and ‘truth’, I wouldn’t want to put it that way either. Sellars’ scientia mensura is better: “In the dimension of describing and explaining the world, science is the measure of all things, of what is that it is, and of what is not that it is not.” [EPM, §42] This may first remind you of Protagoras, but I hope that it also reminds you of Aristotle: Truth is “to say of what is that it is, or of what is not that it is not”.[3]
Robert Almeder’s “blind” scientific realism is an interesting case. Almeder and I agree about so many things that I am sometimes tempted to think of our views as notational variants of each other. The essential claims of his Blind Realism are captured in two theses:
T2: At any given
time, some of our presently completely authorized beliefs about the external
world, including our theoretical beliefs about unobservable entities, must as a
contingent matter of fact, and in some important measure, correctly describe
the external world.
T3: We cannot justifiably say,
justifiably pick out, or otherwise establish by appeal to any reliable decision
procedure, which of our presently completely authorized beliefs do
correctly describe the external world. (BR 144, 225)[4]
What we are to make of this clearly turns on how we understand the notion of “correctly describing the world”. Isn’t that just another way of saying ‘true’?
Not for Almeder. His bête noire is the correspondence theory of truth, that is, the notion that truth consists in “a word-world relationship of correspondence between language and extralinguistic fact”. (BR 5) Rather, on his account, truth is a “word-world relationship of semantic authorization”. That is,
an assertion’s being true principally consists in its being authorized for acceptance by some rule of acceptance embedded in our current conceptual framework. The point of calling the relationship designated by ‘is true’ a word-world relationship of semantic authorization is simply to emphasize that an assertion’s being true consists in its being authorized for acceptance in virtue of some other statement or rule in one’s language, and that its being true does not legitimize any inference to the effect that there is some nonlinguistic state of affairs uniquely described by, and corresponding to, what is asserted by the true statement. (BR 120-1)
A Blind Realist, then, will hold that “while some sentences may well succeed in describing the external world, it would be a serious mistake to regard them for that very reason as ‘true’.” (BR, 175) For such a realist, it consequently “makes perfectly good sense to assert that some true statements will, and some will not, succeed in correctly describing the external world”. (BR, 175) This sounds, at best, paradoxical, but Thesis T2 tells us that we should not despair. Some of our true beliefs—indeed, some of our true theoretical beliefs—in fact do (“in some important measure”) correctly describe the world, even though—as T3 reminds us—we can’t justifiably identify which beliefs those are.
Let us for the moment set aside the question of what it is for a belief or assertion to “correctly describe the world”. It’s clear in any event that Almeder’s argument for T2 can only have the form of a “non-constructive existence proof”, i.e., in the last analysis, a reductio. For, given T3, it clearly cannot proceed by inductively generalizing from examples of such beliefs or assertions that have been successfully identified as such. The argument that Almeder in fact offers, however, is prima facie “abductive”. More precisely, it is a version of J. J. C. Smart’s classic argument for scientific realism: “The predictive success of a scientific theory that asserts the existence of theoretical or unobservable entities suitably justifies us in believing in the existence of such entities.” Only a scientific realist can explain such predictive success.
One would [otherwise] have to suppose that there were innumerable lucky accidents about the behavior mentioned in the observational vocabulary, so that they behaved miraculously as if they were brought about by the non-existent things ostensibly talked about in the theoretical vocabulary.[5]
Although Almeder is clearly sympathetic with this line of thought, he regards Smart’s actual argument as defective. “It cannot counter the plausible suggestion that the predictive success of the theory could just as easily be a function of the truth of the non-theoretical claims made by the theory.” (BR 159)
To avoid explanation by appeal to the miraculous or the coincidental, … we need not assume the truth of the theoretical claims of the theory. We need only assume that some of the claims in the theory are true (or say the way the world is). If this is so, then Smart’s argument … requires of us only that we believe that some of the claims made by the theory say the way the world is. It does not follow that those claims will be the designated theoretical claims, because in urging that the predictive success of the theory may well be borne by the non-theoretical claims (or unsuspected theoretical claims) assumed or implied by the theory we are not appealing to the purely coincidental or to the miraculous. (BR 160)
As the oscillation between “being true” and “saying the way the world is” suggests, Almeder reads Smart’s argument as a defense of classical scientific realism, which “adopts the correspondence theory of truth as a way of explicating the truth condition for knowledge and then regards the truth condition as satisfied when the evidence condition is satisfied”. (BR 184)
Thus, for the classical scientific realist, the knowability
of the external world entails that we be able to pick out the propositions that
succeed in describing the world. …
For the blind realist, knowledge can never be more than a matter of what we are
completely justified or authorized in believing, but that of itself does not
provide us with any criterion for picking out which of our completely
justified beliefs are saying the way the world is. (BR 185)
Almeder’s version of Smart’s line of thought consequently decouples the notion of being true from that of saying how the world is. His “reformulation” of Smart’s argument deliberately excludes the words ‘true’ and ‘truth’. Curiously, however, it takes the form of a question:
On the assumption that there must be a reason why some scientific theories or hypotheses are successful as predictive instruments, what could be a better, or equally strong, explanation or reason for the fact of such success than that the successful theories or hypotheses are successful because at least some of the assertions explicitly made by the theories or hypotheses either say in some important measure how the world is, or assume, or imply, statements that say in some important measure just how the world is? (BR 161)
I want to put Almeder’s question on hold for a minute, and take a quick look at some of the distinctions that a “constructive empiricist”—Bas van Fraassen, to be precise—wants to stress. One central distinction is between believing a theory and accepting a theory. It is absolutely crucial to constructive empiricism that acceptance is something other than belief. How does the distinction go?
To believe a theory is to believe that one of its models correctly represents the world. A theory may have isomorphic models; that redundancy is easily removed. If it has been removed, then to believe the theory is to believe that exactly one of its models correctly represents the world. Therefore, if we believe of a family of theories that all are empirically adequate, but each goes beyond the phenomena, then we are still free to believe that each is false and, hence, their common part is false. For that common part is phrasable as: one of the models of one of those theories correctly represents the world. (TSP, 252)[6]
To believe a theory, in short, is to believe that it is true—all of it, including its entirely theoretical part, where being true, in turn, is a matter of “correctly representing the world”. “Acceptance of a theory,” however, “may properly involve something less (or other) than belief that it is true”. (TSP, 250). Van Fraassen’s elucidation thus invokes a second distinction, namely, between a theory’s correctly representing the world and its being “empirically adequate”.
The scientific theories that we’re interested in here tell stories about the world that go beyond the limits of observation. What is observable, van Fraassen argues, is necessarily limited in various ways. “The most general limit is that experience discloses to us no more than what has actually happened to us so far.” (EPS, 253)[7] Other limits derive from the de facto constitution of members the scientific community, e.g., our specific sensory capacities. What they are in detail is itself an empirical matter. A theory is empirically adequate if it correctly represents the observable phenomena. More technically:
A theory provides, among other things, a specification (more or less complete) of the parts of its models that are to be direct images of the structures described in measurement reports. … [Let] us call them ‘empirical substructures.’ The structures described in measurement reports we may … call ‘appearances.’ A theory is empirically adequate exactly if all appearances are isomorphic to empirical substructures. (TSP, 257)
From the epistemic perspective, then, we can treat a theory in two different ways: we can accept it as empirically adequate or believe it to be true. (TSP, 258) Constructive empiricism adopts the first of these attitudes. A theory is acceptable if it “saves the phenomena”, and “the phenomena are saved when they are exhibited as fragments of a larger unity”. (TSP, 256) This larger, theoretical, unity might also have prima facie consequences which go beyond the observable phenomena, but “our practical commitments require no credence in the more theoretical parts of the theory, and our commitments to research directions or to a direction for further theoretical exploration also need not involve that”. (EPS, 295)
Back to Almeder. There must be some reason, we are to assume, why some scientific theories are “successful as predictive instruments”. In what does such predictive success consist? What gets successfully predicted? The obvious answer is: observations, including measurements, that is, in van Fraassen’s terms, appearances. A predictively successful theory, in other words, just is an empirically adequate theory. What, then, explains the fact that some scientific theories are empirically adequate?
From van Fraassen’s constructive empiricist perspective, the question is surely an odd one. Empirical adequacy is a constraint—indeed, the only constraint—on the acceptability of theories. So it is a matter of course that the theories that we in fact accept, at least in general, will be empirically adequate.[8] His key point is that such acceptance does not require that we believe any of the theory’s explicitly theoretical assertions, in particular, its claims about theoretical or unobservable entities. It is enough that the theory “saves the phenomena”, i.e., that what its theoretical claims imply about measurements and other observations turns out to be correct. But, as far as I can see, this also satisfies Almeder’s desideratum, viz., that “at least some of the assertions explicitly made by the theories or hypotheses either say in some important measure how the world is, or assume, or imply, statements that say in some important measure just how the world is”. (BR, 161) For don’t the correct statements about the outcomes of measurements and other observations that are ex hypothesi implied by the theoretical parts of an empirically adequate theory already “say in some important measure just how the world is”?
Now Almeder might insist that the predictive success of the theories that interest us here can be explained only if some of their purely theoretical assertions “say in some important measure just how the world is”. But that needs an argument which, as far as I can see, is nowhere forthcoming. What is more, it seems to me that such a claim would seriously compromise the “blindness” of his blind realism. Among the theses that Almeder explicitly endorses in connection with his “semantic authorization” account of truth is the claim that “conceptual frameworks” can change and evolve over time
in such a way that whatever is authorized as true under one conceptual framework need not continue to be so authorized under subsequent and more adequate conceptual frameworks, (BR, 121)
where increasing “adequacy” is a matter of enhanced “systematizing power”. I take one implication of this to be that it could turn out that none of the statements about theoretical or unobservable entities authorized by a given theory are authorized by any successor to that theory. But then I am hard-pressed to understand how they might nevertheless still “say in some important measure just how the world is”—except, of course, by implying true statements about observations and measurements.
If this is right, however, it follows that Almeder’s blind realism is entirely compatible with van Fraassen’s constructive empiricism. Since van Fraassen explicitly rejects “scientific realism” and advances and defends his constructive empiricism as an anti-realist alternative, I find myself worrying about just how much of a “scientific realism” blind realism ultimately is. Since Almeder and I agree about so many things that I am sometimes tempted to think of our views as notational variants of each other, this is troublesome. I suspect that the root of the trouble lies in one of the few things that we probably don’t agree about. Almeder evidently thinks that the notion of “how the world is” (and so of “saying just how the world is”) makes intuitive sense independent of an account of the epistemology of scientific inquiry. I don’t.
One of my most fundamental disagreements with van Fraassen also lies in this neighborhood, but before I explain why I reject constructive empiricism, it will be useful to look at why van Fraassen accepts it. His basic reason seems to be that, when the chips are down, only constructive empiricism conscientiously remains true to the essential constitutive principle of empiricism as such. One can be a scientific realist, he suggests, only if one is willing to forego empiricism, strictly and properly construed.
[The] core doctrine of empiricism is that experience is the sole source of information about the world and that its limits are very strict. For an example of such a limit we may take our temporal finitude: it is not possible to have a guarantee about the future on the basis of our experience so far—otherwise we could indeed have opinions today which are not liable to modification in the course of future experience. I explicate the general limits as follows: experience can give us information only about what is both observable and actual. We may be rational in our opinions about other matters … but any defense of such opinions must be by appeal to information about matters falling within the limits of the deliverances of experience. (EPS, 253)[9]
Wilfrid Sellars once characterized ‘empiricism’ as an “accordion word”—one whose sense expands and contracts and, in the process, makes a great deal of philosophical music. Van Fraassen’s explication of empiricism belongs among the austere versions. Mine is more expansive. I would certainly not quarrel with the claim that experience is the sole source of information about the world—but ‘experience’ and ‘information’ are also accordion words, and my take on them tends to be considerably more accommodating than van Fraassen’s. Crucially, however, we will also need to say something about ‘the world’.
We have been here before, of course. When “empiricism” is what is at stake, the history of philosophy can be very instructive. Hume, for instance, adopted an austere version of the empiricist maxim: All legitimate ideas must be derivable from original impressions. Correlatively, Hume treated experience as consisting merely in having sense impressions, but that is an assumption that one can reject. Kant, for instance, rejected it. The Kantian notion of “experience” was not that of a mere sensation or sense impression but rather subsumed ostensible perceptual encounters which incorporated cognitive-epistemic episodes of “taking something as something” or “taking something to be the case”. For Kant, in the last analysis, “there is only one experience, in which all perceptions are represented as in thoroughgoing and lawlike connection”. (A110)
Was Kant also an empiricist? No naïve yes-or-no answer suffices. “Although all our cognition begins with experience,” he notoriously wrote, “it does not on that account all arise from experience” (B1), and that surely suggests that he was not. And yet the limits of cognition turn out to be the limits of possible experience, and that surely suggests that he was. In any event, Kant was demonstrably not what we might call a “tertium quid rationalist”, i.e., someone who supposes that there must be, alongside logical and evidential grounds for belief, some third primitive form of epistemic warrant. His “transcendental deduction” presupposes no such mysterious non-logical and non-evidential epistemic grounding relation, but rather illustrates new and more complex ways of bringing reason to bear on questions of epistemic legitimacy.
Van Fraassen knows very well, of course, that the notion of observability can coherently be construed in ways which allow for it to be literally correct to say that we observe, for instance, electrons, neutrinos, and black holes. To the best of my knowledge, no-one has yet claimed to have observed a free quark, but the project of trying to do so is evidently one that makes sense to the relevant practitioners. In the present debate, however, the operative notion of observability is a function of certain empirically-determinable limitations of human beings qua biological organisms.[10] It is in this restricted sense of ‘observability’ that scientific theories tell stories that go beyond the limits of observability. The crux of the debate, according to van Fraassen, precisely concerns the role of such limits:
Why should its limits be given any epistemological role at all? Why should the range of accessible possible evidence play a role, as well as what the actual evidence is? (EPS, 254)
His answer is, in essence, that he cannot imagine a reasonable alternative.
If we choose an epistemic policy to govern under what conditions, and how far, we will go beyond the evidence in our beliefs, I could not envisage a nonextreme rational policy that would make these boundaries independent of our opinions about the range of possible additional evidence. (EPS, 254)
Suppose that nothing except evidence can give justification for belief. However flexibly this is construed, it means that we can have evidence for the truth of a theory only via evidential support for its empirical adequacy. The evidence then still provides some reason for believing in the truth … but the additional belief is supererogatory. However, not everyone shares my view that what I called the pragmatic virtues provide no independent reason for belief …. (EPS, 255)
Although the terminology is contentious, I have already indicated that I find myself in the latter company. (‘Evidence’ and ‘justification’, parenthetically, are two more of those accordion words.) That makes my empiricism more expansive than van Fraassen’s. Does it also make it epistemically suspect?
The “pragmatic virtues” of a theory, van Fraassen writes, “are sometimes summarized as simplicity, sometimes as explanatoriness”. (EPS, 285) He also calls them ‘superempirical virtues’, but that is now just linguistic imperialism, and henceforth I am having none of it. The problem with pragmatic virtues, as he sees it, is that they give us no reasons for believing in the truth of virtuous theories. If simplicity, for instance, is supposed to be a reason for belief, “we must ask which attribute of simplicity, of the theory or of the depicted world, is at issue.” If the former, the thesis appears to be that the simpler products of our theorizing are “more likely to correspond to the facts”, a claim that van Fraassen simply finds groundless. So perhaps the thesis rests on the conviction that the world is simple. But even if the world is simple, he insists, there must still be lots of simple false theories, and the true theories may not predominate among the simple ones. We can indeed use simplicity as a basis of theory choice, van Fraassen concludes, “but not on the grounds that this will increase our chances of metaphysical or even empirical success”. (EPS, 285-6)
What about explanatoriness? Although I in fact think that there is perhaps something (evolutionary) to be said on behalf of simplicity, my sympathies clearly lie here. Not to put too fine a point on it, I don’t think that constructive empiricists—or, more precisely, that natural science as constructive empiricists understand it—can explain why the stickers on my bedroom ceiling continue to glow in the dark for a while after I’ve turned off the lights. Indeed, I don’t think that natural science as constructive empiricists understand it can explain much of anything at all. But I am convinced that explaining things is an essential part of what natural science is for, what it is about, and what it is up to. Van Fraassen demurs, but he evidently agrees, at least, that this is where the dialectic ultimately leads.
In philosophical practice, the dividing line between [constructive] empiricists and others does indeed appear in connection with explanation. … When it seems that two theories fit the phenomena equally well, what criteria shall we use to choose between them? The answer will have to cite properties and relations going beyond considerations of empirical adequacy. Here explanation takes center stage. (EPS, 286)
Since such considerations, “beyond evidence directly bearing on empirical adequacy,” in fact do play a role in theory assessment and acceptance, he concludes, we face a dilemma:
On the one hand, we can profess to have reasons for believing these other considerations to yield reliable indications of truth. … On the other hand, we can attempt to remain [constructive] empiricists and conclude that these extra considerations which legitimately bear on theory acceptance are not legitimate reasons for belief. This includes the corollary that acceptance is not to be identified with belief. (EPS, 287)
Now van Fraassen concedes that the explanatory reach of a theory is indeed a virtue, i.e., that it can a reason for accepting the theory.
[Ceteris paribus], theories are better if they are more explanatory, which in turn requires them to be more informative. But there are other reasons, besides the weight given to explanation, to conclude that, ceteris paribus, the better theories are the more informative. We go to science to have our questions answered about the empirical world, for many purposes—not just explanation, but practical control, mere factual curiosity (“thirst for knowledge,” if you like), suggestions for new directions of research, and perhaps others. (EPS, 280)
But, he argues, precisely because explanatory power varies directly with the informativeness of a theory, it cannot serve as a reliable indicator of truth. That is why acceptance is not belief.
At the same
time, we worry that the information given by the theory may be false.[11] So we also compare theories with an eye to
the possibility of falsehood, their credibility.[12] However this comparison is carried out, and
whether the criteria are objective or context-dependent or subjective, credibility
varies inversely with informativeness. This is most obviously so in the
paradigm case in which one theory is an extension of another: clearly the
extension has more ways of being false.
So our two models of evaluation are in radical tension, conflict—as I
put it [elsewhere[13]]: informational virtues are not
confirmational virtues. Indeed, the
two desiderata cannot be jointly maximized.
They conflict; they detract from each other.
If some reasons for acceptance
are not reasons for belief, then acceptance is not belief. And indeed some reasons for acceptance hinge
crucially on the audacity and informativeness of the theory. So acceptance is not belief. (EPS, 280-1)
Now I would indeed agree that we compare theories with respect to their credibility—and that brings us to the crux of the matter. One of the basic convictions on which my disagreements with van Fraassen rests is that the only way that one could do this, even in principle, is by comparing theories with respect to their acceptability. Credibility varies directly with acceptability because reasons for acceptance are reasons for belief.
The second horn of van Fraassen’s dilemma is constructive empiricism, which I reject. That suggests that I must accept the first horn. Do I then “profess to have reasons for believing” that explanatory considerations yield “reliable indications of truth”? Not as van Fraassen understands the claim.
My view is that explanatory considerations—what van Fraassen calls “reasons for acceptance”—are reasons for belief, and, of course, I have no quarrel with the platitude that “to believe is to believe true”, that is, as Peirce put it, “we think each one of our beliefs to be true, and, indeed, it is a mere tautology to say so”.[14] There is nothing troublesome about such a trivial immanent conception of “truth”. The Tarski biconditionals satisfactorily fix its sense, and it can generally be interpreted along “minimalist” and “deflationary” lines, e.g., “disquotationally” or “prosententially”. But for just this reason, it is not the sort of thing for which there can be “reliable indications”. For that to make sense, we need an epistemically transcendent conception of “truth”.
Van Fraassen evidently thinks of truth in such epistemically transcendent terms, as a “metaphysical” relation of “correspondence with the facts” or “correctly representing the world”. But I do not see how a constructive empiricist can make sense of such a notion, inter alia, because I cannot see what a constructive empiricist could mean by ‘the facts’ or ‘the world’. Almeder and I are clearly kindred spirits here—not surprisingly, since we have both learned a great deal from Peirce. In the last analysis, however, I don’t think that he is as conscientious about the matter as I want to be, since, although he interprets truth as an epistemically immanent relation of semantic authorization, he apparently still thinks of “saying in some important measure how the world is”, and so of “the world”, in epistemically transcendent terms.
It won’t do to say only that a theory (a claim, a hypothesis, etc.) is true in the epistemically transcendent sense just in case it corresponds to the facts or correctly describes the world or says how the world is, for that is only to endorse the biconditional
TC. A theory T is true if and only if it corresponds to the facts (correctly describes the world, says how the world is).
But what such a biconditional tells us is only that any considerations which entitle us to either side of it will be considerations which entitle us to the other. It does not tell us what those considerations are—or even if there are any such considerations at all. That’s the moral of Wittgenstein’s elegant example:
W. It’s five o’clock on the surface of the sun if and only if it’s five o’clock here and the same time on the surface of the sun.
In other words, we can know that two concepts must have the same conditions of application if any without knowing whether they do have any empirical conditions of application.[15]
As far as I can see, constructive empiricism implies that epistemically transcendent concepts of “the world” and “the facts” can’t have any empirical conditions of application. More precisely, the claim that there is a world which our theories might or might not correctly describe (that there are facts to which our theories might or might not correspond) is at best a claim that, in accordance with Constructive Empiricist principles, we could have reasons to accept, but not one that we could have reasons to believe. But if we don’t believe that there is such a world (that there are such facts), then we have no reason to distinguish between reasons for acceptance and reasons for belief. That makes Constructive Empiricism, if not actually internally incoherent, at least groundless.
Like any good Peircean pragmatist, I understand such notions as “the world” and “the facts”, and hence the notion of truth operative in the biconditional TC, in terms of the practices of scientific inquiry. Peirce’s own formulation of their relationships is both notorious and problematic:
Different minds may set out with the most antagonistic views, but the progress of [scientific] investigation carries them by a force outside of themselves to one and the same conclusion. … The opinion which is fated to be ultimately agreed to by all who investigate, is what we mean by the truth, and the object represented in this opinion is the real.[16]
Today, however, I want to bracket considerations regarding any such supposed “convergence”.[17] But, in the same vein, Peirce also claims that the cardinal virtue of sound scientific inquiry is that it invokes a method “by which our beliefs may be determined by nothing human, but by some external permanency—by something upon which our thinking has no effect”.[18] And this is more useful.
The idea that I want to take from Peirce here is that what a scientific realist needs isn’t an epistemically transcendent notion of truth, but rather an empirically acceptable conception of objectivity. That is what is fundamentally correct about van Fraassen’s insistence that the crux of empiricism consists in the conviction that our beliefs should be responsive only to experience, for what we experience is not a matter of arbitrary choice or decision. The perceptual judgments which embody our observations are spontaneous. We don’t reach them by cognitive processes lying within our deliberative control; we find ourselves with them. Perceptual judgments have a unique sort of de facto non-procedural immediacy. That makes them data—but it does not make them infallible.
From the epistemological perspective, that is, a spontaneous perceptual judgment about how things are is no less an explanatory hypothesis than is an explicit theoretical posit. As I put it elsewhere, such a spontaneous perceptual judgment expresses “a hypothesis regarding the character of the item which, qua stimulus, has (causally) evoked it”.[19] On this point, Peirce is especially clear:
Abductive inference shades into perceptual judgment without any sharp line of demarcation between them; or, in other words, our first premisses, the perceptual judgments, are to be regarded as an extreme case of abductive inferences …. (5.181)
But if this is right, then there is no in principle epistemological difference between what van Fraassen calls “empirical” and “pragmatic” virtues and, hence, none between observation and theory. That is the deepest reason why constructive empiricism ultimately cannot be sustained. In the last analysis, the acceptability of specific spontaneous perceptual judgments regarding observable things turns on considerations of overall explanatory coherence just as much as the acceptability of specific inferentially-based theoretical claims regarding unobservable things does. Both sorts of claims are consequently equally intersubjectively defeasible in the light of further experiences—and that is enough to secure their objectivity, which is all that a scientific realist needs.
At least, it’s all that my sort of scientific realist needs. Not everyone who calls himself a “scientific realist” will be satisfied with this, of course, and there is also a good deal more to be said about both Almeder’s blind realism and van Fraassen’s constructive empiricism. But there’s a definite moral to be extracted from what I have argued here that I would insist needs to be respected in any discussion of scientific realism: We first need to get our general epistemology in order, and we cannot do so if we simply take for granted such notions as “truth”, “reference”, and “the facts” or “the world”. The views I have defended here at least make it possible for science to explain why the stickers on my bedroom ceiling glow in the dark after I’ve turned off the lights, and that will perhaps be enough to satisfy those scientific realists who, like me, are also “Mouse fans”.[20]
 BACK |
NOTES