A Historically Informed Modus Ponens Against Scientific Realism: Articulation, Critique, and Restoration

Abstract

There are two primary arguments against scientific realism, one pertaining to underdetermination, the other to the history of science. While these arguments are usually treated as altogether distinct, P. Kyle Stanford's ‘problem of unconceived alternatives’ constitutes one kind of synthesis: I propose that Stanford's argument is best understood as a broad modus ponens underdetermination argument, into which he has inserted a unique (and doubly inductive) variant of the historical pessimistic induction. After articulating three criticisms against Stanford's argument and the evidence that he offers, I contend that, as it stands, Stanford's argument poses no threat to contemporary scientific realism. Nonetheless, upon identifying two useful insights present in Stanford's general strategy, I offer an alternative variant of the modus ponens underdetermination argument, one that, although historically informed by science, requires no inductive premises. I contend that this non-inductive but historically informed variant of the modus ponens clarifies and considerably strengthens the case against scientific realism.

Acknowledgements

For thoughtful commentary on earlier drafts of this article, I am indebted to two anonymous referees and the editor of this journal, James W. McAllister. For conversations and/or correspondence regarding various topics that pertain to this article, I am grateful to Stephen Ames, Kristian Camilleri, Anjan Chakravartty, Gerald Doppelt, Peter Lipton, David Papineau, Howard Sankey, Neil Thomason, John Tilley, and John Worrall.

Notes

1 As with Stanford's appeal to Duhem, this phrase should not be taken to mean ‘Duhemian underdetermination’. The latter is distinct: it pertains to how alternatives can be generated and retained and in fact stands as one way by which the antecedent of the Mill–Duhem conditional might be grounded.

2 With Stanford's emphasis on scientific seriousness, realists will be relieved that he is not invoking ‘Cartesian fantasies’ (Stanford 2006b, 13) as alternatives. Nonetheless, since he is concerned with unconceived alternatives, ‘well-confirmed’ pertains not to anything scientists have ‘done with’ the theories, but to a relation theories have to data, etc.

3 Moreover, it is implausible that anyone can point to a litany of failures in the history of cases in which the scientific community genuinely did dedicate its time, energy, and resources towards generating alternatives to what was available.

4 On this reading, every individual, scientist or not, continually fails at indefinitely many possible tasks he or she never genuinely sets out to achieve (e.g. parachuting into Jupiter's Red Spot).

5 As above, Stanford deems the failure thesis, a thesis about ‘theorists rather than … theories’ (Stanford 2006a, 123), the central feature of his argument. Yet at the end of nearly 80 pages of case studies dedicated to that thesis in particular, Stanford himself, surprisingly and counter to his own efforts, concedes to the irrelevance of whether an alternative did or did not occur to some specific past scientist (Stanford 2006b, 128–129).

6 Since then, Saatsi and Psillos also and rightly have made the point, and Stanford has responded, in a review symposium on his book (Saatsi et al. 2009). Here I will first unfold the importance of the point; and in note 11 below, I will critique Stanford's response.

7 Stanford himself emphasises the failure of Darwin's theory to account for transfusion experiments, which Galton ‘seems to have regarded as a decisive refutation of pangenesis’; and he notes that alternatives to Galton's theory were ‘accepted by the scientific community soon after Galton's own account was developed’ (Stanford 2006b, 81, 86, 100).

8 This is so irrespective of the first problem I pointed to—that his failure thesis is superfluous if not false—and despite his desire to focus on theorists in (2a) and (3).

9 When noting the one instance of novelty (attained by Weismann's theory), he writes,Nor, it would seem, does the ability to make successful novel predictions in a given domain of theorising indicate that [realism is] beyond the reach of the problem of unconceived alternatives, despite the currency of this notion in much recent philosophy of science. (Stanford 2006b, 130)

On the contrary, I am arguing that contemporary realism is beyond the reach of the argument and evidence Stanford has offered.

10 This is clearly not to insist that, descriptively, every philosopher who fancies him or herself a realist has explicitly invoked novelty; my point does not require an exhaustive consensus among all self-proclaimed realists. (Of course there can be exceptions, for instance, those who have not taken it upon themselves to address Laudan's original historical critique of realism and/or those who, for instance, when not explicitly engaged in the realism debate, have argued that novel success does not afford greater evidential weight than accommodation.) The point is rather, first, that Stanford has offered no challenge to those realists who do embrace the nearly three-decade-long tradition of invoking novelty (e.g. a number of those with whom Stanford engages: Worrall 1989; Leplin 1997; Psillos 1999). Second, those realists who do not explicitly require novel success of the theories to which they commit themselves need only invoke this readily available response to wholly answer Stanford's challenge. Third, and finally, as is made clear at the beginning of section 2.2, realists—even those who, despite Laudan's historical argument, might hope to allow for accommodation—do not (and cannot plausibly) embrace realism as Stanford construes it, where the justification for believing a theory is merely that ‘it offers the best available explanation for the empirical evidence we have … ’ (Stanford 2006a, 122). Since realists cannot plausibly claim justification for believing T irrespective of T's other properties (‘best’ does not entail ‘good’), Stanford's premise (1) must still be replaced by my premise (1a) (at the end of the present section), even if (1a) contains some well-motivated and testable restriction (following the ‘e.g.’) other than novel success.

11 As pointed out in note 6 above, Stanford (in Saatsi et al. 2009) has offered a response to this problem of novel success. Stanford first emphasises his failure thesis (2a and 3 in the schema above) and its distinctness, saying that the realist appeal to novel success ‘does nothing to show that the [failed] attempts of past theorists are relevantly unlike the attempts of present theorists to’ (Saatsi et al. 2009, 382) conceive of alternatives. By itself, this particular formulation of his response will not help him since, as we have seen in section 2.1, his failure thesis is false or superfluous. However, he also phrases his response in terms of (2b) (in the schema above), saying that novel success ‘does nothing to show … that there cannot be … unconceived alternatives to’ theories achieving it (Saatsi et al. 2009, 382; my italics); likewise, reinvoking Fresnel's white spot (Saatsi et al. 2009, 383) and his own two instances of possible novel success (Saatsi et al. 2009, 384) that I identified above, he writes, ‘These examples further illustrate that novel predictive success is no proof of … the absence of fundamentally distinct unconceived alternatives’ (Saatsi et al. 2009, 384; my italics). Unfortunately, each version of this defence—whether it is expressed in terms of his failure thesis, (2a) and (3), or in terms of (2b)—is seriously flawed as a defence of Stanford's inductive argument and the evidence he puts forward in support of it: the realist does not need to ‘show’, let alone ‘prove’, ‘that there cannot be’ or ‘the absence of’ unconceived alternatives. Rather the burden of providing the proper kind of evidence to support Stanford's induction(s) against the realist falls on Stanford himself. Stanford says that ‘novel success’ does not ‘allow us to simply dismiss our independently motivated worries about the possibility and significance of unconceived alternatives’ (Saatsi et al. 2009, 383; my italics). Yet, that which is ‘independently motivated’ here is Stanford's own argument, which does not arrive at the mere possibility of unconceived alternatives. His argument in its full form, as we have seen, rests on two inductive premises and claims that ‘we have every reason to believe that there are similar alternatives to our own contemporary scientific theories’ (Stanford 2006a, 123; my italics). Contrary, then, to Stanford's desires, the appeal to novel success does allow most contemporary realists to ‘simply dismiss’ (Saatsi et al. 2009, 383) Stanford's induction and the chapters of evidence that Stanford offers on its behalf. See also my third point in note 10. (Another component of Saatsi et al. 2009 defence makes appeal to a different historical argument altogether, what I have elsewhere called, favoured, and defended as the ‘pessimistic meta–modus tollens’ against realism: Lyons 2002, 2003, 2006, 2009a, 2012. Although dismissing the idea that his ‘new induction’ rests on the standard pessimistic induction, Stanford effectively admits that his argument does depend on the meta-modus tollens: Saatsi et al. 2009, 383. Unfortunately, for Stanford that logically valid argument, by itself, does not justify his induction(s).)

12 I articulate other relations between these two arguments in Lyons (2009a), where the focus is more emphatically on realist responses to the historical argument. This section of the present article has favourable and significant bearing on the argument made there—as well as on distinct ‘underconsideration’ arguments, as defended for instance by Wray (2008) and Khalifa (2010).

13 I originally introduced this Kepler–Newton case (Lyons 2006) as offering a set of counterinstances to the hypothesis the deployment realist (Psillos 1999) says we can justifiably believe: ‘those theoretical constituents that are genuinely deployed toward novel predictive successes are at least approximately true’. (There I contend that the same, if not more, credit must be attributed to those theoretical constituents deployed by scientists towards accepted generalizations, especially when the latter have, in turn, been deployed in successful specific predictions.) Beyond the set of examples in that paper and other sets introduced in Lyons (2002; see also Vickers 2013), I suggest here that the following additional examples call for investigation, none of whose particulars, as far as I know, have been introduced as bearing historically on the scientific realism debate: Thales's false posit that water is fundamental along with additional rejected constituents pertaining to, for instance, divinity, seminal principles, etc., were genuinely deployed towards, and culminated in, van Helmont's successful five-year willow tree experiment. Descartes deployed his conception of God, who created extension and instilled motion into the world, to arrive at his law of the conservation of momentum. Lagrange and Leverrier made predictions about Venus that were in close accord, but they did so by deploying dramatically divergent constituents regarding Venus's orbit and mass. The false constituent that light particles have mass was deployed by von Soldner to predict, more than a century before Einstein, that light will bend around massive objects such as the sun. This false posit was also deployed by Michell and Laplace in the pre-relativistic prediction of black holes, objects whose escape velocity is too high to allow even light—when light particles are assumed to have mass—to escape from it. Lamark, deploying the false posit that catastrophes and mass extinctions have not occurred in Earth's history, arrived at the prediction that species change over time. Scheele, arguably, did not merely co-discover oxygen, he predicted it, and he did so by deploying a set of patently false constituents of his phlogiston theory. Deploying a set of utterly false posits regarding vortices and idle wheels in his mechanical model of the ether, Maxwell successfully predicted, not only the displacement current, but also that electromagnetic waves travel at the speed of light. Deploying the posit now seen to be false that where there is optical asymmetry there is life, Pasteur was lead towards his extended successful research programme, including, for instance, his successful novel prediction that fermentation is a biological phenomenon. Premises about the early earth now taken to be patently false were deployed towards the novel and successful Miller–Urey experiment. Einstein not only predicted the existence of what we now call ‘dark energy’, he arrived at that conclusion by centrally deploying his now wholly rejected posit that the universe is static. (Another relevant question is the extent to which false posits have been deployed in thought-experiments that result, ultimately, in novel successes.) Although I am only indicating their significance here, I suggest that, along with other examples, each of these supports a logically valid modus tollens (see the point in parentheses in note 11) against the deployment realist's hypothesis. Similarly, such examples have the potential to play a role in historically grounded articulations of the kind of competitor analyses emphasised in this article.

14 For clarification, it should be noted that my proposal does not require that our favoured theories—which might account for new phenomena—have past counterparts. Rather what I am arguing is that new theories, including those theories accounting for new phenomena that have never been accounted for, will have the kind of competitors that I am showing any theory will have. The historical example simply extracts revealing relations from historical predecessor–successor instances, relations that can be instantiated with respect to new theories, where the competitors are to our favoured theories as the successors are to their predecessors in the historical examples.

15 On this point I am indebted to Peter Lipton and Nicholas Maxwell.

16 For a distinct example of another such relation, a relation that has the potential to be historically applied in this way, see Lyons (2011). Beyond my own examples, Schurz (2009) proposes that a specific logical theorem holds between present and past theories. While Schurz's broad thesis purports to establish a very specific logical continuity between such theories, and I am, in contrast, emphasising the use of contemporary science to articulate their divergence as competitors, given the nature of Schurz's theorem, I anticipate that the relation it captures could also be fruitfully directed towards the present concern to generate further kinds of competitors.