A Novel Defense of Scientific Realism. Jarrett Leplin

Book Reviews 777

currently currently leading leadingthe therace, race,and and allow allow that that sometimes sometimes participants participants maymay pass pass each each others. others.Therefore, Therefore,totodefend defend hishis thesis thesis that that thethe evaluation evaluation method method is 'a is 'a refined refined and andcorrected correctedversion' version' ofof Lakatos's Lakatos's methodology, methodology, Kuipers Kuipers should should fur- further ther elaborate elaboratethe thedynamic dynamicand and dialectical dialectical aspects aspects of his of his comparative comparative account account of theory theory change. change.

Department of Philosophy ILKKA NIINILUOTO P.O. Box 24

00014 University of Helsinki Finland

A Novel Defense of Scientific Realism. by Jarrett Leplin. New York: Oxford University Press USA, 1997. Pp. xii + 204. H/b ?35.00.

That the successful prediction of novel phenomena yields especially strong support for a theory is a widely shared, though occasionally disputed, intuition. Leplin's project is to explicate and justify this intuition so he can use it to support a modest form of scientific realism. His idea is that while there may be

alternative explanations of the success of some of a theory's predictions, the only plausible explanation of how a theory is able successfully to predict novel

phenomena is that there is some truth to it. So the book's title promises to characterize not only its general content but also its central analytic tool.

In one sense any successful prediction may be said to be novel. Successful predictions may be distinguished from other empirical truths by the fact that,

when stated, they are not yet known to be true. But Leplin is concerned less with the novelty of knowledge of a prediction's success than with the novelty of its derivation from a theory, whether or not it is then known to be true. A pair

of conditions suffices to make the derivation from a theory of a statement asserting the occurrence of some observable phenomenon into a novel prediction of that theory in his sense. One, uniqueness, basically requires that the prediction not then be derivable from any other theory. The other, independence, is (roughly) that the provenance of the theory be suitably independent of

the prediction.

Leplin does not advance these conditions as a conceptual analysis of a common notion of a novel prediction. Nor does he claim that it is only those predictions that satisfy both these conditions whose success warrants belief in the theory from which they are derived. What he does argue is that when a prediction from a scientific theory meets both these conditions, then the success of that prediction provides some warrant for believing the theory to be not merely empirically adequate but at least partially true. This would justify a realistic interpretation of any theory with enough successful predictions of this kind. Leplin applies his analysis to notable predictions of several physical theories, arguing that it qualifies some but not all of these as novel in the epistemi-

cally relevant sense. But rather than arguing explicitly for a realistic

This content downloaded from 150.135.239.99 on Tue, 21 Mar 2017 19:05:38 UTC All use subject to http://about.jstor.org/terms

778 Book Reviews

interpretation of any of these theories, he rests content with what he calls Min-

imal Epistemic Realism (MER)-'the thesis that there are possible empirical

conditions that would warrant attributing some measure of truth to theories-not merely to their observational consequences, but to theories themselves' (p.102). Aside from detailed objections to his discussion of particular theories, I foresee three main lines of criticism here, indicated by the following questions.

Is Leplin's analysis of novelty adequate? Can it bear the epistemic weight he wants it to? Is MER any more than a sop for the would-be scientific realist?

Leplin motivates his own analysis by comparing it favourably to those of other philosophers of science. As against (one interpretation of) Lakatos, he argues that whether the occurrence of a phenomenon is derived from a theory before or after the phenomenon is observed need be of no epistemic significance. Nor need it matter whether the author of a theory knew of the phenom-

enon in advance, even if (s)he deliberately intended to come up with a theory that would account for that very phenomenon. What is crucial, according to Leplin, is that theory and observed phenomenon meet his Independence Condition: that the observational result O be independent of the theory Tin so far as

'there is a minimally adequate reconstruction of the reasoning leading to T that

does not cite any qualitative generalization of O' (p. 77). While it would be inappropriate to fault this conclusion because it conflicts with what is ordinarily intended by calling an observational result novel, the proposed independence condition warrants further scrutiny on its own terms.

By resting epistemic weight on a theory's provenance, this condition deliberately mixes the contexts of discovery and justification. It assumes that T is arrived at by a process of reasoning, but that a suitable reconstruction of this reasoning need not remain faithful to the actual thought processes of T's author(s). An adequate reconstruction in Leplin's sense is 'a valid deduction of the basic identifying hypotheses of T from independently warranted background assumptions' (p. 77). Leplin's idea is that the premisses of the deduction include not only statements generalized from known empirical results, but also general principles of simplicity or parsimony as well as more domainspecific methodological prescriptions. It seems highly implausible to this reviewer to suppose that the sorts of heu-

ristic arguments used to motivate a theory, either by its authors or by subse-

quent expositors, can be turned into a valid deductive argument from premisses that have some objective form of support. Even the best theories are often stumbled upon by scientists whose prejudices and intuitive leaps appear quite unjustified in retrospect; while subsequent expositors typically give short shrift to arguments for the theory available prior to its formulation, concentrating instead on the latest and most accurate supporting experimental evidence.

Leplin's qualification that an adequate reconstruction be minimal leads to further difficulties. The point of this qualification is to get at those empirical

This content downloaded from 150.135.239.99 on Tue, 21 Mar 2017 19:05:38 UTC All use subject to http://about.jstor.org/terms

Book Reviews 779

results that are logically essential to arriving at the theory along the lines of the

theorist's reasoning. Thus superfluous empirical premisses would be deleted

from a minimally adequate reconstruction; while an alternative result to empirical premiss P, available to but not contemplated by the theorist, would appear disjoined with P in such a reconstruction. Leplin's statement of his

Independence Condition proscribed a minimally adequate reconstruction from 'citing' a qualitative generalization of 0, but left it unclear what would be

involved in such a citation. Now we learn that a statement may be 'cited' even though it is neither included in nor implied by the premisses of that reconstruction, but merely appears as one disjunct in a premiss. Suppose that ON is an observational result unknown to anyone when a theorist first formulated T, and then predicted by no rival theory, but whose subse-

quent verification constituted strong support for T. Then ON forms no part of

the actual provenance of T, and should count as a novel prediction of T in Leplin's sense. Stellar parallax looks like a good example in the context of Copernicus's theory of planetary astronomy, and Leplin claims it is a novel prediction of that theory (p. 73). But this example illustrates a weakness of his analysis.

Let the actual empirical premisses figuring in an adequate reconstruction of

the provenance of Copernicus's theory C be 0, ....,,,,. Suppose further that if these premisses are logically weakened by omitting one of them, or by disjoin-

ing another observational result present in background knowledge in Copernicus's day to their conjunction, then the deduction of the basic identifying hypotheses of C ceases to be valid. This reconstruction will still almost certainly fail to be minimally adequate. For 0N may be disjoined with one or more

of the Oi (i=I, ....m) to yield a weakened set of empirical premisses still sufficient validly to deduce the basic identifying hypotheses of C. The only obvious way to secure the novelty of ON's prediction in the face of this objection is to insist that ON did not form part of background knowledge in Copernicus's day. But to argue that O cannot be 'cited' in a minimally adequate reconstruction of the reasoning leading to T if O is not known to be true until after T has been

proposed is to require that novel predictions be temporally novel after all. Copernicus's prediction of stellar parallax did come long before Bessel's first observation of this phenomenon (in 1838, not, as Leplin asserts, in the second half of the 19th century). But suppose that the stars had been much closerclose enough for the observation of stellar parallax to have formed part of Copernicus's background knowledge. Then Copernicus might have arrived at his theory in just the way he actually did and then argued that this observation

constituted a powerful further reason to believe that it was true. Leplin would

surely wish to endorse this argument. But in these circumstances his analysis would not count stellar parallax (ON) as a novel prediction. For Copernicus's theory would then have had an alternative provenance based on ON; and it follows that ON would be 'cited' in a minimally adequate reconstruction of the reasoning that led Copernicus to that theory. I pass over other objections to Leplin's analysis of novelty to raise a more

This content downloaded from 150.135.239.99 on Tue, 21 Mar 2017 19:05:38 UTC All use subject to http://about.jstor.org/terms

780 Book Reviews

general worry about his project. His basic strategy is explanationist: it is to argue that one should infer the (partial) truth of a theory in so far as it provides not just the best, but the only explanation of a theory's ability to predict novel observational results. Antirealists will likely object to this as yet another application of the misguided principle of inference to the best explanation. Leplin's initial response is naturalistic. He argues that science itself seeks expla-

nations of the success of its theories in order to generalize and improve upon them, and philosophy of science should acknowledge and emulate this aspect of scientific practice. It must therefore satisfy the demand for an explanation of a theory's success in predicting and explaining novel phenomena. He concludes that such success cannot be explained without assuming the theory is at least partially true.

Now while explanation is certainly a by-product and arguably also a central

goal of scientific practice, it is doubtful that there is any general mandate within science to explain the success of its theories. A successor theory does typically explain why its predecessor was successful in predicting what were at that time novel phenomena. But this is a condition on the acceptability of any successor theory, not a reason for looking beyond an existing theory. And when a new theory is successfully sought for other reasons, the explanation it gives does not always certify its predecessor's account of the underlying mechanisms and processes that give rise to the formerly novel phenomena. Consider, for example, an explanation in terms of quantum electrodynamics of the success of Fresnel's wave theory in predicting a bright spot at the centre of the shadow of a circular disc.

Despite its resourcefulness and sophistication, Leplin's strategy will likely prove unsatisfying also to many scientific realists. On the one hand his thesis of Minimal Epistemic Realism seems weak enough to leave it open whether in fact water is composed of H20 molecules or the DNA molecule has the structure of a double helix. On the other hand, the prediction and explanation of

novel observational results is only one way in which a scientific theory can receive the kind of support that most realists would take to warrant belief in its

partial truth. Scientists themselves value extreme quantitative accuracy of predictions at least as much as the prediction and explanation of qualitatively novel observational phenomena. A more accurate defence of scientific realism would surely make more of this aspect of scientific practice.

Philosophy Department RICHARD HEALEY University of Arizona,

213 Social Sciences, Tucson, AZ 85721-0027 USA

This content downloaded from 150.135.239.99 on Tue, 21 Mar 2017 19:05:38 UTC All use subject to http://about.jstor.org/terms