THE EXPLANATORY ROLE OF REALISM
THE EXPLANATORYROLE
OFREALISM
JOHN WRIGHT
Introductory Remarks The debate between Realism and anti-Realism is an on-going one in philosophy. One potential way of moving the debate forward would be to find facts that could be explained by one side of the debate but not the other; since a form of evidence for any doctrine is its ability to explain things its rivals cannot. What, if anything, does the hypothesis of Realism explain? An answer sometimes given by Realists is that it explains the success of science. But Hilary Putnam and Simon Blackburn have claimed that this is something that can also be explained without Metaphysical Realism? In this paper it is argued that they are wrong: certain forms of predictive success, termed "novel" predictive success, can only be explained by Realism. 2 Putative explanations of novel success are of two kinds: those that employ the notion of truth ("alethic" explanations) and those that do not (non-alethic explanations). In this paper it is argued that alethic explanations explain novel success only if"true" is given an interpretation available only to the Realist. Non-alethic explanations either fail to explain novel success, or else are implicitly Realist. The conclusion drawn from the paper is that only if we, with the Realist, see truth as non-epistemic can we explain novel success. Some things can only be explained by Realism. 1. What is Realism, and what might it explain? Intuitively, Realism concerning a realm of entities 3 is the doctrine that those entities exist, and have their features, independently of human knowledge or thought, and independently of being represented by any
35
JOHN WRIGHT
linguistic or other symbolic system. Thus expressed, Realism is a doctrine that stands in need of explication. And so philosophers have standardly taken more precise criteria to indicate a Realist or anti-Realist position. For example, it is widely assumed that if a philosopher believes that an epistemically ideal theory might be false, then that philosopher is a Realist? Or: if a philosopher sees truth as an epistemic concept - for example, as epistemic ideality, or provability, or maximal coherence, then they are an anti-Realist. In this paper I will assume that commitment to any one of these positions is sufficient for commitment to Realism and anti-Realism respectively. Typically, Realists do not merely assert, e,g., that an ideal theory might be false. Usually, they will also want to at least tentatively identify those classes of entities towards which a Realist attitude is appropriate. One area that is frequently thus identified is science. And in this paper we will focus on the question of whether Realism might explain certain features of science, specifically, whether it can explain the phenomenon of novel empirical success; that is, the ability of a theory to successfully predict observable phenomena of a type quite different from those on the basis of which the theory was initially formulated?
2. Realism and the explanation of the success of science It is a familiar idea that Realism might be able to explain the success of science. But the phrase "the success of science" can be construed in many ways. Typically, what is claimed is that Realism can explain observational success, or the ability of science to correctly predict observable phenomena. 6 At first, it is not immediately obvious how Realism might have a role in explaining this. It is easy to see that the bare hypothesis of Realism by itself does not explain it. Realism about a particular class of entities, such as electrons; that is, the doctrine that: Electrons exist, and have their properties, independently of human knowledge or thought, and independently of being represented linguisticaliy or by any other symbolic system.
(1) plainly does not explain the predictive success of electron theory. In
36
THE EXPLANATORY ROLE OF REALISM
order to be able to do that we need to add the claim that electron theory is (approximately) true. This provides us with (an instance of) what we will call the Realist's Typk~d E.~planation ~f the Success of Science: Our theory of electrons is successful beca,se our theory of electrons is at least approximately true. 7 (2) But what has (2) to do with Realism about electrons, as exemplified by (1)? To be sure, (2) implies that electrons exist and that they have certain features or properties (the ones our theory says they do). But (2) does not seem to do anything to capture the notion of independence, which is an essential component of the idea of Realism as stated in (1). Moreover, (2), on the face of it. at least, would not appear to be available only to the Metaphysical Realist. A Pumamian "'Internal Realist", for example, could allow that our theory of electrons is at least approximately true. So: is it the case that what we have called the Realist's Typical Explanation of the Success of Science really is the excl,sive property of the Realist'? One of the main claims of this paper is that, despite appearances to the contrary, it is available only to the Realist. It should be observed that the Realist's Typical E,wlanation is certainly not the only possible explanation that the Realist could give. It seems prima facie possible that there could be other explanations that use resources exclusively available to the Realist. While it will be argued here that there are certain forms of success that can one, be explained by the Realist, it will not be maintained that the Realist's Typical Explanation is the only possible explanation: there are possible other explanations, but all of them must employ resources that are available only to the
Realist. 3. Preliminary Questions Before we consider how the success o f science is to be explained, we must consider a point of view which, if correct, would show the question tobe futile. There have been philosophers who have said that it is not possible to explain the success of science. One such philosopher was Sir Karl Popper. According to Popper, "although science has been
37
JOHN WRIGHT
miraculously successful ...... this strange fact cannot be explained." ~ And it is easy to see why it should seem to be a consequence of his views that this should be so. According to Popper, even those theories that have stood up to the most rigorous testing we have been able to provide remain merely conjectures. We are not, according to Popper, justified in asserting any such theory to be true, or even probably true. But explanations of the success of science such as the Realist's Typical Explanation do assert that scientific theories are true. However, even if we accept Popper's strongly fallibilistic views, it does not follow that we are unable to offer explanations of the success o f science. Popper would have no quarrel with the scientist who offered some theory T as an explanation of some observable phenomena: he would merely insist that T be accepted in a tentative and conjectural spirit. Similarly, there is nothing stopping a Popperean fallibilist accepting statements such as "'T is true" or "'Mature scientific theories are typically approximately true." as explanations of the success of science, so long as those explanations are accepted in the same tentative spirit that Popper recommends actual scientific theories be accepted. Another preliminary question to be considered is, "Does the success of science require explanation?". Although the view that it does not is occasionally encountered, I know of no philosopher who has argued for this. 9 And, on the face of it, it is very implausible. The predictive success of science is very remarkable. Popper even calls it "miraculous"; the onus would surely lie very heavily on the philosopher who said it did not require explanation. Of course, if it were impossible to explain the success of science we would have a reason why we need not attempt to give it an explanation: since "ought" implies "can", there is no obligation on us to do something which cannot be done. But we have just observed that it does not follow from Popperean fallibilism that the success of science cannot be explained.
4. Van Fraassen's Evolutionary Explanation of Success One possible explanation of the success o f science has been advanced by B. van Fraassen? ~ According to van Fraassen, it is a part of scientific practice to incorporate in to science o n l y those theories that are
38
THE EXPLANATORY ROLE OF REALISM
empirically adequate. If a theory is suggested which turns out not to be empirically adequate, then it is rejected. And. if we grant that scientists are continually suggesting a wide variety of scientific theories, it is only to be expected that some of them will be empirically adequate. Van Fraassen asserts that the empirically adequate ones are retained, while those that are not are eliminated. In this way science comes to contain only theories that are empirically adequate. This process has obvious similarities to Darwinian natural selection. But this evolutionary explanation makes no mention of Realism. Moreover. it is arguable that it is a better explanation than any Realist one. A Realist explanation would require us to say that Realism is, at least within a certain domain. true. But this "'evolutionary explanation" does not require us to make any assumption that is not eminently reasonable: It only requires us to postulate that scientists are advancing a wide variety of different hypotheses - which is obviously true - and that at least some of those hypotheses will, at least for a while, prove to be empirically adequate. It also requires us to make the assumption that scientists retain in science those hypotheses that are empirically adequate and remove those that are not. This certainly seems like an extremely reasonable hypothesis, although it is not without those, for example, Kuhnians, who doubt it. But it is surely the case that these hypotheses are all of them less controversial than the hypothesis of Realism itself. So, there is at least a primafacie case for saying that the evolutionary explanation of success is more acceptable than any alternative Realist explanation. However, the evolutionary explanation has a significant shortcoming. Although it may be able to explain some forms of empirical adequacy in our theories, it does not explain them all. We can here introduce a distinction between two types of empirical success. I will call them "Familiar empirical success" and "Novel empirical success". A theory enjoys merely familiar empirical success if it successfully predicts only additional instances of observational regularities of the same type as those on the basis of which the theory was, as a matter of historical fact, initially formulated? ~ It enjoys novel predictive success if it successfully predicts instances of observational regularities different from any of those on the basis of which it was initially formulated. The distinction between
39
JOHN WRIGHT
the familiar empirical success, and the novel empirical success of a theory T will, therefore, depend on the facts of the History of Science; it cannot be "'read off" from the nature of the theory itself. The evolutionary explanation perhaps does satisfactorily explain familiar empirical success.'2 But it prov ides no explanation at all of novel empirical success. The Realist, on the other hand, is able to provide a possible explanation of this type of success. One Realist explanation would be as follows: Electrons exist and have their properties independently of human language, knowledge etc. Our theories of electrons correctly describe these independently obtaining features of electrons. Therefore, any predictions, including novel predictions, derived from this true theory will themselves be true. Hence, our theory of electrons will enjoy novel predictive success. While this is an explanation of novel success, it is not evident that it is the exclusive property of the Realist. The explanation does not appear to make any essential use of the notion of independence. The only feature of truth that appears essentially in the explanation is its preservation under logical entailment: that is, that if a proposition (or conjunction of propositions) is true, then any proposition that logically follows from it will also be true. And this feature of truth is not likely to be available only to the Realist. So, if we are to find an explanatory role for Realism, we must do more than produce the above explanation of novel predictive success. However, we can say that the Realist is able to provide one possible explanation of novel predictive success, while van Fraassen's constructive empiricism is not.
5. Dummett's response to the phenomenon of novel predictive success Michael Dummett acknowledges that if a scientific theory is able to successfully predict novel observational regularities then we may be tempted to adopt a Realist interpretation of the theory.~3 But he goes on to say that the phenomenon of novel success need not "'essentially"
40
THE EXPLANATORY ROLE OF REALISM
conflict with the anti-Realist position. Dummett says that the anti-Realist about science can allow that the successful theory has. "'besides its actual content, a certain suggestive power. ''j4 Dummett is asserting that, even though a scientist might appear to derive a novel observational prediction O from a theory T, the theory does not really entail O. it merely "'suggests" it. But this clearly provides no explanation of why the "'suggested" observational prediction should turn out to be true. So Dummett has not explained novel predictive success. Clearly, the satisfactoriness of his response will depend on to just what 6xtent novel success requires explanation. If it stands strongly in need of explanation, then Dummett's failure to provide one is a serious shortcoming. Moreover. it is clear that there are many cases of novel success which stand very much in need of explanation. Consider. for example, our theory of the transuranium artificial elements. These are elements, with nuclei larger than those found in nature, but which can be manufactured in particle accelerators. Scientists are able to correctly predict the ways in which these elements will chemically combine with thousands of other substances. It is beyond the bounds of plausibility that such a large number of successes should be merely a lucky fluke. They "'cry out" for explanation. The failure of the Dummettian anti-Realist to provide such an explanation is therefore a serious shortcoming.
6. Fine's Strict Empiricist Explanation Arthur Fine has said that we need not say that theories are true, or even close to the truth, in order to explain the success of science? ~ Fine asserts that a simpler and better explanation o f the success of a theory is that it is empirically adequate. But it is plain that Fine's "explanation" is not at all satisfactory if what we want to account for is the novel predictive success of a theory. Let us say that T is a theory that was formulated on the basis of observational regularities R~. R 2 and R 3 , but that T also successfully predicted the novel observational regularity R 4 . How are we to explain the ability of T to successfully predict R 4 .9 One possible explanation is the Realist one: that T is true, and hence that the entities postulated by T exist and behave in the way that T says they do. On this Realist view it is only to be expected that the observational consequences
41
JOHN WRIGHT
of T, even the novel observational consequences, will be correct. But Fine does not explain T's novel predictive success: we want to know why T successfully predicted a new regularity, and the "explanation" we are given is that T is empirically adequate. What we want to know is why T is empirically adequate in areas other than those which formed the basis on which T was advanced as a hypothesis. Fine's suggestion does nothing to explain this:
7. The Objection from the Underdetermination of Theory by Data An important objection to the thesis that the novel success of a theory T needs to be explained by asserting that T is true comes from the thesis of the underdetermination of theory by data. IfT is a theory that explains a body of data D, then by the underdetermination thesis, there will be at least one other theory T* that also explains D. The empirical success of T could, therefore, also be explained by asserting that T* is true (or, by asserting that T* is close to the truth, or, with the deflationist, simply by asserting T*). But it is n o t necessary to explain the success ofT by making any mention of T itself. This argument is, of course, vulnerable to the objection that is based on the assumption that the thesis of the underdetermination of theory by data is true. But there is a variant of the argument which is not vulnerable to this objection, since it provides us with a way of constructing an empirically equivalent rival. The variant appeals to the idea of the Ramsey sentence of a theory. Let T R be the Ramsey sentence of T. Then it will always be possible to explain the empirical success of T by citing T R. This argument only requires us to assume that we can, for any theory that enjoys predictive success, construct its Ramsey sentence. Yet another variant of the argument appeals to the History of Science. Instead of merely appealing to the abstract possibility of a range of theories, all of which are capable of explaining a given body of data, the argument in this form points to the historical fact that there have actually been advanced rival theories that could explain common phenomena: for example, observed positions of heavenly bodies could be explained by both Copernican and Ptolemaic systems. The argument from the underdetermination of theory by data provides
42
THE EXPLANATORY ROLE OF REALISM
us with an explanation that is different from the Realist's typical one but, it will be argued, it is nevertheless a Realist explanation. Suppose T is an epistemically ideal theory. Such a theory will pass all tests to which it is subjected, provided it is subjected to those tests in "'epistemically ideal conditions. Since it will pass all such tests it will count as a highly successful theory, it will, in fact, have the maximum possible degree of success. Let T Abe empirically equivalent to T. According to the underdetermination objection it is possible to explain T's success by asserting T A. But ifT A is offered as apos~ible explanation of T's success, it is admitted that T A might be true. And if it is asserted that TAoffers a rival explanation for T's success to that given by T itself, it is thereby admitted that it is possible that T A is true and T false. That is, to admit that T Aoffers an alternative explanation of T's success is to admit that the ideal theory T might be false, which is a hallmark of Realism. Therefore, the underdetermination objection implicitly assumes a position which is tantamount to Realism. 8. The predictive success of false theories Of course, some theories, now believed to be false, have enjoyed novel predictive success? 6 Obviously, it would be wrong to use the Realist's Typical Explanation to explain their success. But how is the novel success of a false theory to he explained? Plainly, van Fraassen's evolutionary explanation will not do: it at most provides an explanation of familiar predictive success and does not explain the novel predictive success of either true or false theories. Likewise, the accounts of neither Fine nor Dummett satisfactorily explain the novel success of theories, whether true or false. One possible explanation of the novel success of a false theory is suggested by our discussion of the thesis of the underdetermination of theory by data. An advocate of that thesis objects against the Realist's Typical Explanation of Success that it is possible to explain the success of one theory T by appealing to another, empirically equivalent theory T*. It is similarly possible to explain the novel success of a false theory by appealing to some other theory which may yet be true and is able to explain the novel phenomena predicted by the false theory. And this is how we in fact explain the success of theories believed
43
JOHN WRIGHT
to be false. For example, we explain the success of Newtonian theory by pointing out that, from the point of view of Einsteinian theory, it is, except at high velocities, close to the truth. Or: we explain the success of classical optics with quantum theory. More generally, we explain the novel success of a past, falsified theory by using our best current theory. Now, this explanation may be of two kinds: We may assert that our current theory is true (or close to the truth), o r we may merely use our current theory without making the additional claim that it is true (or close to the truth). We will call the first type of explanation "'alethic" and the second a"redundancy-style" explanation. Are either of these types of explanation available to the opponent of Realism, or are they implicitly Realist? In the remainder of this paper it is argued that both alethic and redundancystyle explanations are available only to the Realist.
9. Anti-Realist interpretation of the Typical Realist Explanation: Epistemic conceptions of truth and alethic explanations of success So far in this paper it has been argued that alternatives to the Realist's Typical Explanation of Success either do not enable us to explain all forms of success, or else are implicitly Realist. There is, however, another powerful and simple move available to the opponent of Realism: accept what has here been called the Realist's Typical Explanation of Success, but give it an anti-Realist interpretation; specifically, interpret the term "truth" that appears in it in an anti-Realist manner. However, in this section it will be argued that this strategy does not work: if "true" is given an anti-Realist interpretation in the Realist's Typical Explanation, it thereby ceases to provide us with an e x p l a n a t i o n of certain forms of Success.
We may distinguish three conceptions of truth that are plausibly available to an opponent of Realism. The first of these is the epistemic conception, on which truth is identified with some epistemic property. On this conception, or family of conceptions, truth may be identified with verifiability, with "epistcmic ideality" or "'warranted assertibility under ideal conditions", with superassertibility, and so on. Another family of conceptions of truth that are natural allies with anti-Realism identify truth with coherence. A third type of view of truth which some authors
44
THE EXPLANATORY ROLE OF REALISM
have seen as allied to anti-Realism is the redundancy view and its close cousin, the minimalist view. ~7 In this section we wilt focus on the epistemic conceptions of truth: coherence and redundancy views are discussed later. An advocate of an epistemic conception of truth identifies truth with some epistemic property E. We can say they advance an identity claim of the form, "'Truth is E.". What is the status of this identity claim? There seem to be only two candidates available: either it is an analytic, conceptual truth, claiming that a correct analysis of our concept of truth reveals it to be the very same concept as E. or else it is a non-analytic, a posteriori identity statement which is metaphysically necessary. We will call the latter a Kripkean identity. It is not too difficult to see that whether "'Truth is E" is seen as an analytic identity, or as an a posteriori but metaphysically necessary Kripkean identity, the link between the theory's truth and at least some instances of its success will be necessary. Obviously, whether "'Truth is E" is claimed to be analytic or a posteriori but necessary, it is claimed that it is a necessary truth. So, an advocate of an epistemic conception of truth will subscribe to: Necessarily, if a scientific theory is true, then it has E.
(3) Now, since E is an epistemic property which is claimed by an advocate of the epistemic conception to be identical with truth, the property E will be a property of maximum rational acceptability. This concept of "'maximum rational acceptability" may be explicated in various ways: for example, as assertability under ideal conditions, as superassertibility and so on. But however it is to be explicated specifically, it is surely the case that one factor that increases the rational acceptability of a scientific theory is its success in passing experimental tests. A theory that possesses maximum rational acceptability will, all other things being equal, pass more tests than its rivals. Let us call the tests passed by a theory that has property E (that is, the property of maximal rational acceptability), and in virtue of which it does possess E, the tests that are constitutive o f it possessing E. We may then assert:
45
JOHN WRIGHT
Necessarily, if theory T has E, then it passes those tests that are constitutive of it possessing E.
(4) From (3) and (4) it follows, for an advocate of an epistemic conception o f truth: Necessarily, if a theory T is true, then T passes those tests that are constitutive of it possessing E.
(5) That is, for an advocate of an epistemic conception of truth, the relation between a theory's being true and its passing of certain tests will be a necessary rather than contingent relation. Let us now consider: what sort of explanation are we giving when we explain the success of a theory? A success of a theory is its passing a test to which we have subjected it; therefore, it is an explanation o f why a particular event occmTed. Hence, it is a causal explanation. The explanation under consideration is that this event took place because (or at least in part because) the theory tested was true. So, on the explanation under discussion, the truth of a theory is a cause of its passing the test. Therefore: If success is explained by a theory's being true, then the link between truth and success is causal, with truth as cause and success as effect.
(6) There is a long tradition in philosophy which sees the links between cause and effect as contingent, that is, that it is possible for an event which is actually a cause to take place without being followed by that event which is actually its effect and conversely that it is also possible for an event which is actually an effect to take place without having been preceded by that which is actually its cause. This is not to deny that there may be some species o f " n o m i c " necessity whereby cause and effect are linked. But the type of necessity, if any, which links cause and effect is different from the type of necessity which is a consequence of
46
THE EXPLANATORY ROLE OF REALISM
conceptual or analytic relations, and is also different from metaphysical necessity. We may therefore say: The relations between cause and effect are contingent rather than necessary. (7) It is now easy to see how an advocate of an epistemic conception of truth will be unable to use the notion of truth to explain at least some cases of the success of science. If we are'to causally explain success by saying that scientific theories are true. it follows from (6) and (7) that the relation between the truth of our theories and their success is contingent. But from (5) it follows that. for the advocate of the epistemic conception of truth, in at least some cases, the relation between a theory's truth and its success is necessary. So, in those cases of success an advocate of an epistemic conception of truth is precluded from saying that the link between truth and success is contingent, and so is precluded from using truth to causally explain those cases of success. And since a causal explanation of success is required, there are at least some cases of success that the advocate of the epistemic conception of truth is unable to explain. This argument for the explanatory incompleteness of epistemic theories of truth makes essential use of the contingent nature of causal connections; we will therefore refer to this argument as the Contingency of Causal Connections, or CCC argument. The CCC argument does not, of course, apply against the Realist, since, for the Realist, the link between the truth of a theory and its success in passing tests, is not necessary. For the Realist, it is possible that our experimental techniques, even under ideal conditions, fail to reveal the truth: a theory might pass all our tests and yet be, from the Realist perspective, in reality false, while a theory that is in reality true might fail to pass some of our tests. The idea that theoretical error is always possible is part of the "epistemological modesty" of the Realist. It means that the link between the truth of a theory and that which we regard as evidence for its truth, including tests it successfully passes, remains contingent. And a consequence of this is that a causal explanation of a theory's success remains available for the Realist.
47
JOHN WRIGHT
In summary, the manouevre of using the Realist's Typical Explanation of success but giving "true" an anti-Realist, epistemic interpretation, deprives us of a causal explanation of successes of a theory which, on the epistemic conception of truth, are constitutive of its truth. It does this because it makes the link between the truth of a theory and those successes necessary, whereas the links must be contingent if the former is to causally explain the latter. 10. The D e f l a t i o n a r y E x p l a n a t i o n of Success The deflationary explanation of success is extremely simple. The Realist's Typical Explanation of, for example, the success of the theory that electrons have negative charge is: "'Because that theory is at least approximately true." The deflationary explanation simply replaces the claim .... Electrons have negative charge" is true" with "'Electrons have negative charge." The deflationist says that the reason why the theory is successful is because electrons do, in fact, have negative charge. Although this explanation differs from what we are here calling the Realist's Typical explanation, it will be argued that it is nevertheless implicitly Realist. If the deflationist's position is to be viable it must apply to all theories; that is, it must be possible to explain the success of any theory T by simply asserting T itself, rather than saying that "T" is true(or close to the truth). Hence, if the deflationist position is to be viable, it must apply to both epistemically ideal and to epistemically non-ideal theories. But it is not too difficult to demonstrate that if the deflationist gives a causal explanation of the success of an ideal theory, then their explanation is implicitly Realist. Let us suppose that "Electrons have negative charge." is ideal. Such a theory will pass any test to which it is subjected under epistemically ideal conditions. We will call those tests passed the "ideal successes" of the theory. The deflationist will explain these ideal successes by saying, "'Because electrons have negative charge." This is a causal explanation; the deflationist is asserting that electrons having negative charge is a cause of the theory having those ideal successes. But, by the doctrine of the" contingency of the relations between cause and effect, it follows that it is possible for this effect - ( the ideal success of the theory) - to take place while its cause - (electrons having negative
48
THE EXPLANATORY ROLE OF REALISM
charge) does not: that is. that it is possible for the theory "'Electrons have negative charge." to be false even if that theory enjoys ideal successes. But to admit this is to admit that a theory might be false even though epistemically ideal, and that is a position available only to the Realist. Hence the deflationist's explanation of the ideal success of a theory is implicitly Realist. Only a philosopher who accepts the thesis that a theory might be false although ideal can consistently offer a causal explanation of ideal success. Of course, a deflationist does not have to explain the success of ideal theory T b y asserting T itself. It would still be in the spirit of deflationism to explain the ideal success of T by asserting another theory T*. which was empirically equivalent to T and therefore successful in exactly the same places as T. Such an explanation is still deflationary in spirit because the success of a theory is explained not by asserting that some (other) theory has the property of truth, but just by asserting a theory itself. But again this explanation is implicitly Realist. If T* is offered as an alternative explanation of the ideal success ofT. then it is claimed that it is at least possible that T* is true. And since T* is incompatible with T, it follows that it is also claimed that it is possible that T. although ideal, is false. So we are again led to the Realist notion that T might be false even though ideal.
11. Realism, Success and Coherence We have not yet exhausted all the theories of truth that might be available to the anti-Realist. We also need to consider the coherence theory of truth. Could it perform the explanatory tasks that deflationism and the epistemic view of truth could not.'? The idea of the coherence theory of truth is that truth consists in a relation of coherence between the elements of a system. We may distinguish between pure and impure coherence theories. For a pure coherence theory, any system of beliefs will count as true if it exhibits the requisite degree of internal coherence. But of course the standard objection to pure coherence theories is that they fail to distinguish between truth and sufficiently coherent fantasy. So there have been introduced impure coherence theories. For an impure coherence theory, truth is not a property of just any sufficiently internally
49
JOHN WRIGHT
coherent system of belief; rather, it is only a property of certain special systems, usually those systems that contain within them, for example, the set of reports of experience, or observation sentences that have, as a matter of historical fact, actually been accepted by science. It seems most implausible that apure coherence theory could explain the success of our theories. Any agreement between the predictions of a system of belief that exhibited merely internal coherence and observations of the actual world would be purely accidental. Therefore, any agreement between the predictions of such a set of beliefs and actually obtained observations would not be e.xplained by saying that the set of beliefs was true, in the sense of "'true" of the pure coherence theory. It is also easy to see that an impure coherence theory would be equally unable to explain some forms of success. Consider the form of success that we called novel predictive success. A theory enjoys novel predictive success if it successfully predicts a new type of observable phenomenon distinct from those on the basis of which the theory was, as a matter of historical fact, initially formulated. But how would this be explained by saying that the theory is true, in the sense of the impure coherence theory'? The fact that a system of beliefs contains within it the actually accepted observation statements of science does not explain why it should be that the system entails observation sentences that subsequently turn out to be correct. In order to explain this, it would appear we would need to add some additional hypothesis such as: the integrated nature of the system of beliefs ensured it was isomorphic with reality and hence yielded observational predictions that subsequently turned out to be correct. But by itself, and without any additional such hypothesis, the impure coherence theory would not explain the novel predictive success of a theory. Moreover, with the addition of such a hypothesis, the explanation appropriates features usually associated with some other form of truth, such as the correspondence theory. Further; if it is supplemented with such a hypothesis, then it becomes vulnerable to the CCC argument. Suppose it is asserted that a theory T is true in virtue of corresponding to some aspect A of the world. Then, by the CCC argument, the fact that T corresponds to A can causally explain the ideal success o f T only if it is possible for the ideal success of T to occur without T bearing the
50
THE EXPLANATORY ROLE OF REALISM
correspondence relation to A, and hence, only if it is possible for T to be false, even though ideal. In summary, neither the pure nor impure coherence theories of truth are able, by themselves, to explain novel success. If an impure coherence theory is supplemented by, for example, the hypothesis that the coherence of a system ensures its correspondence with the world, then the offered explanation becomes vulnerable to the CCC argument. The coherence theory of truth, therefore, seems to be no more able to explain all forms of success than the epistemic conception 'of truth. 12. Explaining the success of false theories We will now returu to the topic of the explanation of theories that are false. The History of Science provides us with many examples of theories. now accepted to be false, that have nevertheless enjoyed predictive success. There have even been cases of false theories that enjoyed novel predictive success. In Section 8 we noted that we explain the success of a past, falsified theory by using some hitherto unfalsified theory. In this section it will be considered whether this explanation is available to the anti-Realist. It follows directly from theses argued for in preceding sections that, in some cases, the explanation must be implicitly Realist. Let us consider what is involved in using some unfalsified theory (T ) to explain the success of a falsified theory (Tf). First, any theory passes a test successfully iff (a) it makes the prediction that (fsome experimental situation E A obtains then outcome E o will occur and (b) it is actually observed to be the case that when E a obtains, E o also occurs. We explain how a false theory Tf passed such a test by showing that unfalsified theory T can explain why it was observed that E A was followed by E oThe explanation given could be alethic: "Because T is true." or it could be a deflationary, redundancy-style explanation which simply used T without making the additional claim that it was true. That is, the explanation could be:
or:
E o was observed to follow E o was observed to follow
E A because E A because
"'T'" is true T.
51
JOHN WR1GHT
But in these explanations the truth of T (or, the state of affairs described by T) is used in a causal explanation of why a particular event occurred. The event explained is the observation that E A was followed by E o. Now, let us consider the special case in which T is ideal, and is ideal in virtue of it passing the test, or set of tests, the outcomes of which it is being used to explain. Since it is used to causally explain those outcomes, the doctrine of the contingency of causal connections tells us that it is possible for those outcomes to occur even when T is false, or the situation described by T fails to obtain. That is, i f T is used to causally explain those outcomes, it is possible for T to be false even though ideal, which is a position constitutive of Realism. Consequently, if an unfalsified theory is used to explain the success of some past falsified theory then. if those successes are also the successes in virtue of which the explaining theory is ideal, then the explanation is available only to the Realist. Of course, none of this is to deny that some successes of false theories can be explained by the anti-Realist. It is no part of the thesis of this paper that there are no forms of scientific success that can be explained by the anti-Realist; it is only maintained that there are some that can only be explained by the Realist.
Conclusions In this paper it has been argued that there are forms of success of scientific theories that can only be explained by the hypothesis of metaphysical Realism. Epistemic, and other anti-Realist conceptions of truth either fail to explain novel predictive success or remove the contingent nature of the link between truth and success which must hold if the former is to causally explain the latter. Only if, with the Realist, we see truth as non-epistemic can all forms of success be explained. Moreover, in construing truth as non-epistemic, the Realist is able to capture the independence dimension of Realism, as it makes what is true independent of our capacity to know what it is true. Since the ability of a hypothesis to provide the best, or the only, explanation of phenomena counts as evidence in its fagour, the conclusions presented here constitute evidence in favour of Realism. The position defended also shows that Hilary P u t n a m ' s Internal R e a l i s m does not furnish us with an
52
THE EXPLANATORY ROLE OF REALISM
e x p l a n a t o r i l y c o m p l e t e i n t e r p r e t a t i o n o f s c i e n c e , a n d that S i m o n B l a c k b u r n ' s project of Quasi-Realism cannot be extended to all areas of scientific discourse. There are some, restricted, areas of science in which Metaphysical Realism remains the only viable option. ~s UNIVERSITY OF NEWCASTLE AUSTRALIA NOTES An earlier version of this paper was given in the Department of Philosophy at St David's University College, gampeter. Wales. I would particularly like to thank David Cockbum, Rowan Rockingham-Gill, David Wafford and John Daniel for helpful comments and criticisms.
2
Even in his later, anti-Metaphysical Realist phase Putnam has continued to hold that (a form of) realism is the only philosophy of science that does not make the success of science a miracle, but on his view the success of science only justifies us in asserting the correctness of Internal Realism. Blackburn says that a Pumamian Internal Realist would be able to account for the success of science in his Essays in Quasi-Realism (Blackburn, 1993), especially p.30. A well worked out account of the notion of a novel predictive success can be found in J. Watkins Science and Scepticism. (Watkins, (1984) especially p.295. An alternative account is developed in my Realism and Explanatory Priority (Kluwer, (1997) especially pp.172-181. The notion, and its importance, is also discussed in A. Musgrave's "'The Ultimate Argument for Scientific Realism" in Nola, R. (1980). Here I have defined Realism as a doctrine concerning entities, although it is quite frequently defined in terms of sentences or other linguistic entities. Although there are real advantages to explicating Realism as a doctrine about classes of sentences, the basic, intuitive idea of Realism surely concerns the world and the objects in it. This definition of Realism has been used by Hilary Putnam in his writings on the topic since his Meaning and the Moral Sciences. See, for example, Part Four "'Realism and Reason", especially p. 125. Again, it is worthwhile noting precise definitions of this notion can be found in Watkins, (1984) Loc cit. And in Wright, (1997) Loc cir.
53
JOHN W R I G H T
6
v
~
,o H
,2
'~ ~4 ~5 ~6
54
This is used, for example, by Alan Musgrave in his paper "'The Ultimate Argument for Scientific Realism.". in R. Nola's Relativism and Realism in Science. (Nola, (1980)) In its more general form, the Realist's Typical Explanation of success is: "'Theory T is successful because T is at least approximately true." See Popper. (1972), p.204. An argument sometimes encountered in conversation is that the success of science does not require explanation because it is a lucky fluke. The plausibility of this suggestion clearly depends on just how likely it is that science should enjoy such a run of good luck. When we consider the predictive success of a branch of science such as physical chemistry it is very hard to believe that the success is merely a lucky fluke. The onus lies very heavily on someone who believes that the success of science is a gigantic fluke to show that such luck is not implausibly improbable. Another view is that science is not really successful, that its supposed success is some kind of an artefact of the social condition of our society. I will not comment on this view here, except to note that advocates of this point of view sometimes seem to conflate socially determined criteria for the application of a concept C with lack of risk for prediction for C. Suppose that a theory T predicts that taking drug X will cause patients to go mad. Now, it may very well be the case that the criteria lbr correct use o f " H e is mad."are in some pretty strong sense socially determined. But it does not follow from that that it will not be a risky prediction to say that if a patient takes X they will go mad. And if that risky prediction is successful then the success of theory T would, prima facie, be s o m e t h i n g that requires explanation. See van Fraassen, The Scientific Image, (O. U. P., 1980), especially pp.3940. This way of defining the notion of novel predictive success presupposes a way of individuating observational regularities. Ways of individuating these regularities are discussed in Wright (1997), pp. 172-181. Although in order to explain why a theory continues to enjoy familiar empirical success after its formulation, we would need to include in the explanation some additional premise such as "Nature is uniform." See Dummett (1978), p.152. See Dummett, Ioccit. See Fine, (1986) A good list of false but empirically successful theories is given in L. Laudan's "A Confutation of Scientific Realism." in J. Leplin (1984). An example of a
THE EXPLANATORY ROLE OF REALISM
false theory that enjoyed novel predictive success is the ether theo~" of light which was shown by Fresnel to lead to the prediction that there would be a tiny bright spot in the exact centre of a circular shadow. Putnam (1981 ) has argued that the minimalist conception of truth leads to relativism. For a discussion of the relations that can be discerned between Realism and the minimalist or redundancy conceptions of truth, see Horwich (1990). especially pp. 54-62. Of course, if ReaNsm is to be viable, it must be possible for the Realist to reply to the arguments against Realism that have been developed by Dummett and others. It would take us too far beyond our present concerns to give a complete reply to those arguments here. Instead, a very brief sketch of a reply will be given. Dummett has argued that the Realist's conception of truth is not viable because it attributes to speakers a knowledge of truthconditions they could never manifest in their behaviour. Dummett's argument is sound only if it is true that: A speaker is able to make a true statement S if and only if the speaker is able to recognise as such the conditions that would confirm the truth of that statement. But there is evidence from the History of Science that this is not so. For example, in 1931 the scientist Wolfgang Pauli postulated the existence of particles called "'neutrinos". In 1931 Pauli was unable to recognise as such any experimental results that would have confirmed "'There are neutrinos". It is natural for a Dummettian to say that this means that when Pauli uttered "'There are neutrinos.", he was not referring to the particles we now call by that name. But if this position is adopted, the subsequent construction of apparatus that led to the discovery of neutrinos becomes inexplicable. It would make the subsequent success of the theory of neutrinos "'a miracle" So, a crucial premise in Dummett's argument fails. (A more developed account of this argument can be found in Wright (1997), pp.202-217 and pp.248-255.)
REFERENCES Blackburn, S. (1993) Essays in Quasi-Realism., New York: Oxford University Press. Dummett, M. (1978), Truth and Other Enigmas, London: Duckworth. Fine, A. (1986), "'Unnatural Attitudes: Realist and Instrumentalist Attachments to Science." in Mind XCV, pp. 149-179.
55
JOHN WRIGHT
Goodman, N (1977), Ways of Worldmaking, Indianapolis: Hackett Publishing Publishing Company. Horwich, P. (1990) Truth, Oxtbrd: Basil Blackwell. Leplin, J. (ed.) (1984) Scientific Realism, Berkeley: University of California Press. Nola, R. (ed.) (1980), Relativism and Realism in Science, Dordrecht: Kluwer Academic Publishers. Popper, K. (1972). Objective Knowledge, Oxford: Clarendon Press. Putnam, H. (1978). Meaning and the Moral Sciences, London: Routledge and Kegan Paul. Putnam, H. ( 1981 ). Reason. Truth and History, Cambridge: Cambridge University Press. Putnam, H. (1983), Realism and Reason: Collected Philosophical Papers, vol 3, New York: Cambridge University Press. Van Fraassen, B. (1980) The Scientific hnage, Oxford: Clarendon Press. Watkins, J. (1984), Science and Scepticism, London: Hutchinson. Wright, J. (1997). Realism and Explanatory Priority, Dordrecht: Kluwer Academic Publishers.
56