13

Taking the Metaphysical Bull by the Horns

Completing the Job

Paul Teller

This note is a follow-on to van Fraassen’s “Let’s Take the Metaphysical Bull by the Horns,” commenting on my contribution to this volume. I shall give my own interpretation of what I think van Fraassen has in mind. I will then explain why I differ from him—and this, in two respects. I will explain why I am inclined not to follow the additional step that van Fraassen wants to take, and then I will consider van Fraassen’s at least apparent characterization of my pragmatism as a kind of fictionalism. I do not think that it has to be taken that way. Explaining why will give me the opportunity to expand just a little on the last section of my paper.

Interpreting van Fraassen’s Comments

A clear picture requires a brief synopsis of van Fraassen’s constructive empiricism.[1] Van Fraassen draws a distinction between belief and acceptance as attitudes toward theories. “Acceptance of a theory involves as belief only that it is empirically adequate” (van Fraassen 1980, 12). But acceptance involves more than such belief:

To accept a theory is to make a commitment, a commitment to the further confrontation of new phenomena within the framework of that theory, a commitment to a research programme, and a wager that all relevant [observable] phenomena can be accounted for without giving up that theory. (van Fraassen 1980, 88)

Belief, by way of contrast, is an unqualified epistemic attitude. To believe a theory is to take it to be true without qualification:

For belief . . . all but the desire for truth must be “ulterior motives.” Since therefore there are reasons for acceptances [e.g., the commitments mentioned previously] which are not reasons for belief, I conclude that acceptance is not belief. (van Fraassen 1989, 192, italics added)

Next we need van Fraassen’s notion of what counts as observable: that of which oneself or anyone in one’s epistemic community can become perceptually aware without the use of instruments. The fact that the test refers to our whole epistemic community avoids the usual problems with the observational/nonobservational distinction. The relativization to one’s epistemic community is well motivated by the rationale for the whole position, as explained in my work (Teller 2001, 125–26, 129–31).

Van Fraassen provides a statement of constructive empiricism in The Scientific Image:

Science aims to give us theories which are empirically adequate, and acceptance of a theory involves belief only that it is empirically adequate. (1980, 12)[2]

If beliefs that arise from an accepted theory are to be restricted to unaided observation, what role will instruments have? Rather than providing “windows” on a world that we cannot directly observe, van Fraassen regards instruments as producing new observable phenomena. Thus, a spectrometer produces observable spectrographs. A voltmeter produces observable phenomena such as its pointer pointing to “5” on the output display (cf. Teller 2001, 130).

Van Fraassen takes from Duhem the idea of thinking of theories as producing useful ways of classifying observable phenomena. Thus, the observable phenomenon of the pointer on the voltmeter pointing to “5” may get classified as “current with voltage of 5 volts.” A visible streak on a photograph of a cloud chamber may get classified as “passage of an alpha particle.” No belief in voltages or alpha particles is involved. Instead, to accept a theory is, among other things, to accept the usefulness of the theoretical classifications in accounting for the observable.[3]

Van Fraassen’s Reinterpretation of My Pragmatist Critique of Traditional Measurement Accuracy Realism

I can now quickly apply this sketch of constructive empiricism to how van Fraassen seems to understand my pragmatist critique, what he wants to add, and why. Consider a voltmeter hooked up to a circuit. Observable characteristics include the pointer pointing to “5” and a big greasy fingerprint smudge. At the crudest level our theories will classify the pointer reading as important relative to certain interests and the smudge as uninteresting (or will not classify it at all). More specifically, and usefully, theories may classify the pointer reading as in the theoretical category of a voltage reading, with the value of five volts. Various of our theories may yield further classifications, such as the five-volt classifications being relevant to such and such, as leading to expectations of this and that, as recommending certain manipulations to achieve various ends, and so forth.

It is to be emphasized that constructive empiricists will regard these classifications as just that—with no reference to a quantity, a voltage, a value of five volts, or the like. Constructive empiricists may treat all of this with much the same affect of involvement one expects from realists, thinking in terms of voltages and so on but not as things believed, but as useful ways of thinking about and classifying observable things in application of relevant theories (cf. Teller 2001, 127).

My gloss so far may leave one feeling that van Fraassen’s appeal to theoretical classification and its uses sounds somewhat like what he has characterized me as offering—“a practically useful fiction or idealization.” What is distinctive about van Fraassen’s view is the felt need for something more. Acceptance of the relevant theories and their classificatory categories must be filled out with beliefs in the empirical implications that we draw from our application of the theories and their classifications. These beliefs provide “[a] way to make sense of truth and reference” in ways that show “the form that assessment of success in science takes.” The observable end products of the application of theoretical classifications provide the kind of referents and truth-evaluable judgments that constructive empiricists are ready to embrace as full-fledged beliefs.

Why I Do Not Take This Additional Empiricist Step

To explain my attitude I will need to use a notion that has things in common with van Fraassen’s acceptance but also differs in important respects. So I need a different term: adoption. To adopt a statement is to treat it as true for certain purposes. These purposes may be very specific and narrow. For instance, we can treat water as an incompressible, continuous medium for describing the fluid properties of water. The limiting case would be to assume a statement for the purposes of a reductio argument. Or the purposes for which one adopts a statement may be extremely broad, may be part of a background set of adopted statements presumed to apply unproblematically for anything that might reasonably come up: New York City is located in the United States.

Like adoption, acceptance has a pragmatic side; however, as I understand it, acceptance is limited to treating a theory as a basis for investigating a certain range of observable empirical phenomena. Adoption is pragmatic through and through: an instrumental attitude toward any statements that are entertained. Unlike acceptance, adoption does not require belief in empirical adequacy. And unlike acceptance, adoption of a statement presupposes that the statement can be refined, either by being made more precise (in the sense of being less vague), or more accurate, or both. (The reason for this additional clause is to facilitate integration with my larger view on treating truth, that I will mention later.[4])

In other words, I do not add van Fraassen’s additional empiricist gloss to my treatment of accuracy because I adopt much but—in van Fraassen’s sense of belief in unqualified truth—believe nothing.[5]

I can explain why most clearly where the disagreement is most salient, at the level of perception. Unlike van Fraassen, I take perception to be so much more modeling, not (when it is successful) a veridical grasp of a perceived world. This is a fundamental difference between myself and van Fraassen, who rejects representational theories of perception and indeed rejects mental representation of any kind (van Fraassen 2008, 24). I have written about perception as modeling in many places (in most detail in Teller 2008, section 7). But briefly, color vision provides an exemplar.[6] Naïvely, to see something as red is to become perceptually aware of the thing as having the intrinsic color property red. At least since the early modern period, people have appreciated that there are no intrinsic color properties, that they are “secondary properties” that involve the perceiver and, we now appreciate, also the immediate environment. In particular, a color cannot be understood as the surface spectral reflectance of an object—that is, the pattern of just what wavelengths of light an object will reflect.[7] Furthermore, our current theoretical understanding of color perception is itself incomplete, is not completely accurate, and is idealized in many ways.

The upshot is that any understanding we have of color, from the simple naïve idea of intrinsic properties to the best theory we now have, is not the identification of some property or relation in nature but a simplified model in terms of which we can usefully manage our color experience. I submit that it is going to be this way for all of perception. The world is too complicated for the brain to be able to afford to get it exactly right. Rather, the brain cuts corners and employs heuristic simplifications. In short, our perceptual experience of the world represents the world that we perceive in terms of simplified models. This includes our perceptual experience of discrete perceived objects. The problem of indeterminate spatial and temporal boundaries of objects is best understood as an idealized simplification that our perceptual apparatus makes for us that provides an excellent compromise between effectiveness and falling short of complete accuracy in representing a world that is unmanageably complex.

It is not such a big leap from perception to our representation of any aspect of the world, whether through perception, science, or other indirect means. All of it involves shortcuts, simplifications, and useful distortions. Any statement that we entertain is, at best, a representation that stands to be corrected and/or refined. Thus, the attitude of adoption is much more deeply qualified than taking one’s judgments, generally, to be fallible.[8]

I should note that the forgoing is really a simplified presentation of my views. I also interpret the vagueness of language to be the flip side of our inability to get things both perfectly precisely and perfectly accurately. One can always eliminate inaccuracy of a very precise statement by making the statement less precise. One can get a little more sense of this from the final section of my paper. In (2017) I work this out in a lot more detail, including a way of thinking about truth that makes all of this work smoothly together.

Fictionalism?

At one point van Fraassen suggests that my attitude toward all the contents of what I adopt counts as “practically useful fiction or idealization.” Idealization, yes, but I take exception that this counts as a kind of embracing of fictions. That a representation gets some things wrong does not automatically make it a fiction. Consider the opening line from the old time radio show Dragnet: “Ladies and Gentlemen. The story you are about to hear is true. Only the names have been changed to protect the innocent.”[9] Medical and especially psychiatric reports will falsify irrelevant personal details to ensure the anonymity of the subject. But that does not make such a report a work of fiction.

Engaging with the world through telling but imperfect idealizations still constitutes knowledge of the world. For a first pass at understanding how this works, consider your efforts to become familiar with the physical features of someone’s face by using photographs taken from various angles, none of which are in perfect focus. Still, you can use the photographs to build up an increasingly informative composite understanding of the face’s physical features. One easily accepts this way of developing understanding through imperfect representations because one supposes that there is the objective face “out there” with which one could, at least in principle, make comparisons. One supposes that the objective face sets the respects in which the imperfect representations are and are not accurate.

I think that the foregoing is a good summary of how we proceed, but not a good gloss on how to understand what we are doing. The perceptions with which one checks the accuracy of the imperfect photographs are themselves only more representations. And we know that any perceptions to which one might appeal as setting the referent are themselves far from completely determinate and far from perfectly accurate. We can make sense of this without appealing to an objective face and without supposing an ultimately perfect representation. We do this by appreciating that each and every one of our representations is subject to refinement, in the sense that it could be superseded by another representation that better performs for us in all the ways than did the first, and from the point of view of which the prior representation was less precise and/or less accurate.[10]

When considering the face in question, there is nothing of which we can know exactly what it is. This might be because, when it comes to the face, there is nothing “out there in nature” that is in question. Or it might be because there is something, but something of which, because it is too complicated, we can have access to no exact representations—that is, representations that do not admit, at least in principle, the sort of refinement described in the last paragraph. This is a contingent circumstance, due to the fact that the world is just too complicated for us limited beings to produce representations that could not be refined.

The best we can do is to build up an ever-sharpened understanding by considering a collection of ever increasingly refined (in the sense sketched two paragraphs back) and increasingly integrated collection of imperfect representations.[11] Given that we are limited, contingently, to this kind of way of knowing about the world, the issue of the targets of the representations “out there in nature” becomes moot. (Dare I say, it becomes a matter of metaphysics?)

Given the complexity of the world, which limits our representational powers, this is the only possible kind of knowledge of the world, of the way the world is, really. Many have responded to this view by thinking of it as some kind of instrumentalism, or perhaps idealism. As with the epithet of fictionalism, I take such characterizations to be at best misleading. Our perceptual knowledge of the perceptible world is the exemplar, the gold standard of what it is to know about the world—we can say, redundantly, about what the world is really. When it turns out that all such perceptual knowledge is always limited and imperfect, always refinable, that is no reason to conclude that, after all, it is not knowledge of the world. The further knowledge that goes beyond perception is of the same kind in its limitations and in its refinability. There is no reason that it should not likewise count as knowledge of the world, although imperfect knowledge (we can say, redundantly, of the way the world is really). So this view should count as realism, in particular about both perception and science. Indeed, given the limitations on human knowledge, this is the only sensible way of understanding realism about any kind of human factual knowledge of the world.

Notes

References