a. Analogical Reasoning

One of the most straightforward arguments in favor of scientific realism is the intuition toward realism in general. When a detective examines a murder scene, she tends to assume the hypothesized criminal actually exists. Further analysis may or may not bear out the initial hypothesis about who committed a crime—perhaps it was a psychotic killer escaped from a mental ward, not an acquaintance of the deceased—but her normal faculties reliably point toward real entities in the world. Moreover, the process of analysis generally leads an (honest) investigator toward an increasingly accurate mental picture of the perpetrator. Missteps may occur in the analysis, but are amenable to critical reassessment, so that even when the revision from the original hypothesis is substantial, the trend-line of the mental model is toward increasing accuracy. The basic picture was refined, not destroyed: the victim was not after all killed by a wizard from Alpha Centauri (and even that would be a malicious, person-like agent).

Science, on a realist view, operates similarly. Initial hypotheses may prove to have more or less in common with the real world, and not only may but must be subject to revision, even radical revision along the way. However, theories should be treated realistically, both ontologically (there really is something out there to know) and epistemically (science grants access to reality, albeit in approximate and limited fashion).

The primary argument against this analogy toward scientific realism is the idea of in-principle unobservables.² The crime scene investigator has direct experience of things like criminals (other human persons, at least) as well as of the evidence itself. No one has direct experience of electrons. More than this, the postulates of quantum mechanics (are taken to) suggest that no one can have such direct experience. Then, the antirealist argues, the scientific claim is at best unknowable: electrons may exist, but scientists have no way of knowing. They may observe a trail in a gas cloud, but whether the thing leaving the trail is really like an electron is unknowable.

Any such attempt to distinguish in principle between observables and unobservables fails.³ First, antirealists use “unobservable” to mean anything not directly observable by human senses. However, declaring such objects in-principle unobservable is question-begging: the point under debate is just whether things detected indirectly are being observed or simply hypothesized. Antirealists must first establish such an in-principle distinction between indirect and direct observation. Realists note that indirect observation is rightly considered reliable in the realm of ordinary experience, and can rationally be extended analogically. It is reasonable to conclude that the wind is blowing from seeing its effects on trees even if sitting inside a closed room (with no direct experience of the wind). This would be so even for a person who had spent his entire life locked up inside and therefore had never had direct experiential access to wind or trees. Indirect observations still count as observations.

b. Whole-Cloth Reasoning

Moreover, human reasoning seems to be “whole cloth.” It is difficult (if not impossible) to find any particular line, even a blurry one, where this kind of inferential reasoning becomes suspect.⁴ Granted that electrons are quite different from anything of which humans have direct experience, this is not finally telling against reasoning about them. Humans are quite capable of making the requisite fine distinctions required for analogical reasoning to be valid; indeed, scientific models of electrons expressly include both similarities to and sharp discontinuities with human experience.

Second, a realist view rightly makes sense of the way “observability” lies along a spectrum. On one end are those phenomena which can be detected through (normal, healthy) human senses such as hearing or vision. Moving down toward the not-directly-observable spectrum, scientists use telescopes and microscopes to enhance their ability to observe things inaccessible to ordinary human sight because of distance and size respectively. In both cases, however, there is no clear line distinguishing “observable” from “unobservable”—only a spectrum of more- or less-directly-observable phenomena. Insects may be observed with the unaided human eye, bacteria with optical microscopes, individual molecules of materials with electron microscopes. In each case, the same principle is at work in the observation: light bouncing off an object and being received.⁵

The anti-realist may object: the fact that the distinction between observables and unobservables is blurry does not mean it does not exist. As Samir Okasha notes, the line between “bald” and “hirsute” may be fuzzy (pun intended), but it is still possible to identify a bald man.⁶ Even granting the in-principle distinction between “observable” and “unobservable,” however, the objection fails to establish the strong claim made by antirealists. It does not establish that the items in question are in fact unobservable (especially when confronted with the realist argument for indirect observation). Nor does it establish that scientists would be in principle incapable of correctly modeling unobservables. At most, it establishes that there may be unobservable phenomena.⁷

Thus, it is reasonable to suppose that though current models of astronomical behavior, or of quantum mechanics, or gravity, or any other only-indirectly-observable phenomena may be incomplete or partial, they nonetheless represent something actual. There really is an electron leaving a trail in the gas chamber.

c. No Miracles

The substantial success of science in both predicting and providing coherent explanations for empirical phenomena, it seems prima facie strange that the hypotheses in question would have nothing whatsoever to do with reality. It would be a miracle of sorts if a set of theories which do not directly rely on each other would all work effectively, and effectively in conjunction, while not actually saying anything true about the world. These predictions are more than merely theoretically interesting; they have real-world application used every day: “laser technology is based on a theory about what happens when electrons in an atom go from higher to lower energy-states. And lasers work…”⁸ It seems reasonable that one reason scientific theories have been so successful is that they are approximately accurate, and increasingly so over time (if in fits and sometimes with false starts).⁹

Many critics think the evidence of change in adoption of given theories, and varying degrees of empirical success over time, is hard evidence against the notion that scientific theories generally do refer in this way. Similarly, some theories once held to be true are now thought to be neither approximately true, nor even subsets of what is true. Larry Laudan, citing a number of changes in this vein, sniffs: “The realist’s claim that we should expect referring theories to be empirically successful is simply false.”¹⁰

First, the fact that science is an iterative and approximative process does not tell against the realism asserted. Any given theory may be substantially false. This does not mean all theories are likely not to refer, only that realists must be critical, aware of the possibility of failure, and willing to hold their claims tentatively, especially at first. No scientist with a knowledge of the history of the discipline should think otherwise. Electrons might not exist; but at least for the present, there is good reason to treat them not only as useful but as real. The longer the predictions of the model are (not dis)confirmed by the evidence, the more reason scientists have for holding the conclusion confidently. Likewise, as other theories developed without direct reference provide further corroboration, it is increasingly rational to affirm a theory as true, as well as empirically useful.

Second, even if Laudan’s point is granted, it does not disprove realism. It simply proves naïve realism false—naïve, even. True: empirical success does not guarantee the truthlikeness of a theory, and vice versa. It does not follow that there is no correspondence between the two. Laudan insists that the truthlikeness of a given theory has zero bearing on its empirical fruitfulness, but it is at least possible that it is instead a contributory factor.¹¹

Finally, what about later theories which do not make special cases of the earlier theory? When Laudan, Kuhn, van Fraassen, and others examine e.g. the fluid and aether theories and compare them to the atomic theories now held by most physicists, they note that science has been empirically useful without being true in the past; the same may be true today. However, later views, currently held by realists to be accurate, can often explain the success of the earlier theory even without claiming it as a subset.¹² The atomic theory lets scientists explain the appearance of the aether. Newton’s mechanics can be applied to show why Ptolemy’s epicycles worked, even though they were false. The later theory not only has greater empirical success; its explanatory power also includes the now disconfirmed theory—whether as a special case or not. To be sure, the idea that all science proceeds linearly and that every theory is this kind of empirical superset of the former is incorrect.¹³ Still, the fact that later theories often can do so is further evidence that truthlikeness is a contributory factor in the success of scientific theories.

a. Larry Laudan’s Instrumentalism

Instrumentalism takes the effectiveness of science as indicating only its effectiveness in delivering more reliable predictions.¹⁶ Philosophers like Larry Laudan reject as both irrelevant and false the idea of “partial truth” or “verisimilitude” as incoherent: in a semantic model, statements are either true or false. This is partly fair: scientists claim their models have a semantic content corresponding approximately to the world. Yet the critique fails to account for the way scientists use the notion of “approximate-ness.” Models may be incomplete, yet yield correct results when the approximation is appropriate. There are thus three important distinctions to be made. The first is between accuracy (how near the actual value a given experiment is) and precision (the repeatability of a measurement). An archer who consistently shoots four inches to the right of the bullseye is precise but not accurate; one who is always within two inches of the target but all around it is accurate, but not precise. The second is between accuracy and the philosophical binary of truth. A scientific model might in the strict sense be “false”—that is, even its predictions are not empirically true, only close to correct, while the model may be more accurate than any previous model. An arrow is either within the bullseye or not, but there is more to scoring an archery contest—and more to evaluating the accuracy of scientific claims.

The third distinction, and the most serious, is between models’ empirical content and their semantic content. A theory might be said to be “more true,” i.e. more accurate, than any predecessor, if it is more consistent with the empirical data or capable of explaining a broader set of such data. However, the semantic claim is more likely to be binary. Either there are electrons, or there are not. If not, there may be nothing very much like electrons at all. Even so, this distinction is less sharp than Laudan supposes. Increasing approximation of reality is possible. Models may have more or less fidelity. A perfectly round globe may strictly be a false representation of the earth, which is slightly squashed due to its spin; but it is nearer correct than a piece of paper or a donut. The earth is much more like a sphere than a plane or a toroid. If reality is more like the Standard Model of quantum mechanics’ description of electrons and quarks than it is like a pre-atomic view of reality, the Standard Model is a better approximation to reality than the pre-atomic view. Voiding the distinction between accuracy and binary truth is simply a category mistake.

Finally Laudan shows how the scientific methodologies developed over the past several hundred years work, but not why—especially in cases quite different from his example of double-blind clinical studies.¹⁷ In scenarios where technological control over the environment is at stake, science clearly has an edge on other methods (so his critique of relativism stands).¹⁸ It is less clear why drug trials should be amenable to this sort of testing if the human body and drug interactions are not more like the scientific models than the non-scientific models. It was precisely the belief in the accuracy of such models which gave rise to the instrumentally effective methodology! One might posit, then, that realism itself is instrumentally useful in the pursuit of these technological gains. The problem, of course, is that such an epistemic stance is difficult to maintain at best. It is one thing (and not a very controversial one) to say that the scientific method is instrumentally effective. It is another to say that belief in scientific realism is instrumentally effective and false, because no one can sustain a belief they think false.

b. Bas C. van Fraassen and Constructive Empiricism

In Bas C. van Fraassen’s constructive empiricist model, science is (or at least should be) concerned solely with explaining the empirical data available to the human senses. Theories are useful but not true; they “do much besides answer the factual questions about regularities in the observable phenomena which, according to empiricism, are the scientist’s basic topic of concern.”¹⁹

At least three major problems confront van Fraassen’s view. One is van Fraassen’s prior commitment to materialist empiricism. He asserts, for example, that criteria such as mathematical elegance, simplicity, scope, completion, unification, and explanatory “provide reasons for using a theory… and cannot rationally guide our epistemic attitudes and decisions.”²⁰ Why? Because he regards human observational powers as the sole grounds of knowledge. As such, he also rejects any critique of the limitations of “sense data” or problems with human perception as “armchair psychology” which can safely be ignored.²¹ The sole point of interest, for van Fraassen, is explaining those regularities in observable phenomena. As Grover Maxwell comments, though:

Surely the main concerns of, say, a theoretical physicist involve such things as the actual properties and varieties of subatomic particles rather than mere predictions about where and how intense a certain spectral line will be.²²

Indeed, even in conducting their ordinary research, theoretical physicists and chemists are concerned with things they think actually exist. However much van Fraassen thinks science should only concern itself with those regularities, realism seems basic to the scientific endeavor as actually practiced.²³

The second problem for van Fraassen is closely related to the first: “empirical adequacy” is a necessary but insufficient condition for the acceptance of a scientific theory. Scientists do (and must) embrace those other conditions for accepting theories as true, and rightly so. Nothing about using those criteria to decide between theories is irrational in the least apart from van Fraassen’s prior commitment to empiricism.

Finally, van Fraassen simply reasserts the distinction between observable and unobservable entities. He begins with the idea of in-principle unobservability, but on encountering Maxwell’s argument from technological improvements, he moves the goalposts. He has in mind, he says, the limitations of human beings “qua human beings;” technological aids actual or imaginable are irrelevant.²⁴ As shown above, though, the lack of any such clear distinction is the major problem for the antirealist interested in maintaining this distinction.

c. Moreland’s Selective Antirealism

J. P. Moreland suggests an alternative from within a Christian epistemology and ontology. He fully grants a generally realist epistemology and affirms a correspondence view of truth. However, he also approves a cautiously eclectic approach to scientific realism.²⁵ He offers two major criteria for adopting antirealist stances. The first is “in those cases where the phenomena described by that theory lie outside the appropriate domain of science, or the scientific aspect of some phenomenon is inappropriately taken to be the whole phenomenon itself.”²⁶ For example: scientific explanations of mind currently tend toward reductive physicalism. But if there are good reasons to suspect that the mind is not reducible to the brain (say: the whole history of dualist arguments), any scientific theory which purports to explain mind solely in terms of material function should be treated antirealistically—as instrumentally useful, but not true. But a critically realist claim about physical-mental state correlation does not entail physicalism. Rather, the physicalist claim is not scientific but philosophical. This should certain chasten the materialists, but it is not proof against realism—not generally, and not even in the case of this specific theory. It is proof only against unwarranted extension from the claim.

Moreland also argues that “an antirealist stances should be taken toward some scientific theory in those cases where a realist view conflicts with a rationally well-established conceptual problem for that theory but an antirealist view does not.”²⁷ He suggests that the logically problematic claims of some interpretations of quantum mechanics suggest a non-realist interpretation: “the indeterminacy in nature is due to the uncertainty of our knowledge of nature or our inability to measure nature and is not a feature of nature itself.”²⁸ Similarly, if there are good philosophical reasons for rejecting infinite regress, one should be confident rejecting even an instrumentally-useful cosmological model of an infinite universe.

In these latter examples, however, as well as in the further criteria he briefly mentions,²⁹ Moreland lumps together two distinct phenomena under the same label of “anti-realism.” One is an in-principle denial that certain kinds of data are available to us; the other a denial that a specific model comports with reality. One might think that all theories about electrons are permanently subject to antirealist critiques and that we can have no true knowledge of such allegedly unobservable entities. This is quite different from thinking that a given cosmological model is wrong. Both entail prior commitments, and the warrant in both cases arises outside science. They are nonetheless different kinds of claims: one about what is knowable, and one about what actually is. Even if Moreland’s criteria were granted, this is very different from van Fraassen’s empiricism or Laudan’s instrumentalism. Rejecting a given theory is quite different from rejecting the possibility of a theory being true.

d. Non-Rational Antirealism

The strongest form of Thomas Kuhn’s description of scientific history emphasizes that the theories on either side of a scientific revolution are totally incommensurable, actually operating in “different worlds.”³⁰ The shift away from Aristotelian is one where Galileo saw differently than he had before.³¹ Kuhn’s protests that he understands the problem with the phrase notwithstanding, he regularly reverts to the idea that scientists before and after revolutions lived in different worlds.³² This cannot be right: even if the theories are distinct in important ways, and even if are not totally commensurable, there are certainly data on which they can agree. Even if what the planets were seen as shifted between the Ptolemaic and the Keplerian eras, the basic observational data were agreeable to both, for example. Likewise, though mass has different semantic content in the Newtonian and Einsteinian conceptions, both agree on many of the empirical values—including problematic measurements which prompted to the transition between them.

Taken as a chastening of overconfident or naïve realism Kuhn’s historical critique is to be appreciated. His view of scientific revolutions leads him (or at least many of his readers) further, though—to conclude that observation is so theory-laden that, though the objective world is “out there” to describe, humans have no true access to it. The data itself is always subject to the flux of another revolution. Such a view, whatever its attractions, is first of all suspect simply on the reality that scientific approaches have accomplished goals (like predicting and controlling floods) which cut across scientific and non-scientific cultures.³³ Second, any such non-rational antirealism stands clearly opposed to the Biblical idea that truth is knowable and that both moral and empirical data may be true or not—that the final standard is not human knowledge but the God-created reality outside the human knower. Third, such a view is self-defeating, as it claims objective knowledge about scientific theories while purporting that no such objective knowledge is possible.³⁴