1. Hudson on Mesosomes, WIMPs, Atoms and Molecules, and Dark Matter

It is now a common procedure in contemporary philosophy of science to cherry-pick and/or heavily reinterpret—for illustrative purposes—historical cases that support one’s preferred view of a given controversy. Hudson’s book contains four such cases, two of them from contemporary astrophysical cosmology: the mesosome (a reputedly observable organelle of bacteria that does not exist), WIMPs (weakly interacting massive particles, an “exotic” form of mass that might exist in humongous amounts but has not yet been detected), Perrin’s take on the reality of atoms and molecules (they are real, not artifacts of microscientific procedures), and so-called dark matter (mass-energy that does not directly radiate or emit detectable evidence of itself—contra Clowe et. al. [Reference Clowe, Bradac, Gonzalez, Markevitch, Randall, Jones and Zaritsky2006] and Hudson himself in the book under review—but which seems to be required to exist in large galactic-scale amounts by the standard cosmological model of the universe).

The four cases are analyzed by Hudson as cases of methodologically sound technical science in which “robust” convergence reasoning (Hudson calls it “robustness reasoning”) was not used by the scientists involved, or else it was used by some of the scientists involved to justify false claims. Hudson argues that both Nancy Cartwright and Wesley Salmon were wrong in supposing Jean Perrin to have used robustness reasoning in arguing for the reality of atoms and molecules—indeed, Perrin himself was wrong in claiming to have used such reasoning, Hudson claims to show (xv, 195).

Hudson’s basic method is as follows. Whenever a scientist claims, for example, to have used robustness reasoning, Hudson argues instead that she was using some other kind of reasoning that could be mistaken for robustness reasoning but is not such reasoning. If necessary, Hudson creatively coins the distinctive style of reasoning himself. Thus, Hudson introduces us to “reliable process reasoning” (97–102), “calibration reasoning” (121), and so forth. These are terms used by Hudson with confidence that, all the standard histories and memoirs to the contrary, he understands the real reasoning that was used much better than do the scientists who made the discovery in question. Thus, Robert Kirshner, an astrophysicist who was in on the recent discovery that the universe’s space-time expansion rate is increasing, is said by Hudson to have “misrepresented” that discovery in a 2004 book about it (174–75).

Hudson is sure that these cases ought to embarrass realists committed to a robustness defense of realism (52). However, it is not obvious what is supposed to follow from these four cases even if they are granted as cases in which no success/robustness reasoning was used. Let us stipulate that Hudson has found four cases in the history of science in which specific scientists reached realist conclusions by means of argumentation that was not based on robustness and/or convergence reasoning. Put slightly differently, let us stipulate that field scientists tend not to use NMAs in their daily work. Surely neither stipulation ought to be shocking, for NMAs were invented by philosophers, mostly for philosophers, to buttress specific philosophical views about modern science. There are philosophers of science who remain partial to the notion that robustness/no-miracles arguments are second-order arguments—arguments about science and not within science—and that therefore we should not expect to find workaday scientists using such arguments very often in their daily cogitations. If this view has anything going for it, then the trouble to which Hudson goes to show, at length, how this scientist or that scientist eschewed robustness reasoning is needless trouble.

However, Hudson’s cases are richly articulated by him, and some of them do suggest a possibility that might really disturb some scientific realists who are fans of robustness reasoning. For example, in the mesosome case—mesosomes were observationally reproducible (i.e., “robust” in Hudson’s sense of the term) reputed organelles inside bacteria that turned out to be mere artifacts of the preparation process used to render the bacterial innards visible in microscopes—Hudson argues that robustness reasoning, or convergence reasoning, was the very reasoning used by scientists to support the false claim between roughly 1950 and 1970 that the mesosome was real rather than an artifact of the preparation process. Let us again grant this last claim of Hudson’s. It follows at most that robustness reasoning is fallible. Should this result surprise us? Frankly, it is hard to see how it should. All deductive reasoning with empirical premises is fallible by dint of the threat of unsoundness. So what? It is assumed in second-order NMAs for realism that the convergent data are nonartifactual, genuine, something about which the scientists have not been fooled. If the data are misleading, then all bets are off. It is hard to see how this means that no-miracles reasoning is the thing to blame or to see as problematic.

2. Hudson’s Humbler Form of Realism: Methodological Preservationism

Professor Hudson does harbor some wan realist sympathies. Hudson labels his form of realism ‘methodological preservationism’. Hudson tries to show that robustness reasoning (which he seems to treat as more or less coextensive with no-miracles reasoning) as a historically informed practice licenses realist conclusions only when it is applied to standardized, well-tested, and accurate observational methods that involve physically distinct causal processes. Hudson argues that it is not sufficient for obtaining robust convergence of outcomes that the independence in the various observational procedures be a theoretical independence—rather, he argues that it is “the physical processes” themselves that must be independent (170–74, 240–43).

One of Hudson’s main complaints about the standard theory-based NMA for realism is that realists fail to provide criteria for when two empirical procedures are theoretically independent—that is, for when they presuppose sufficiently different theoretical auxiliary hypotheses (36ff.). Fair enough, but it surely strains belief to suppose, as Hudson does, that the same problem does not bedevil providing criteria for when two empirical procedures involve sufficiently different physical processes. At least in the theoretical case we might have strong referential or semantic differences between the two theories that underwrite convergent observational procedures, but how does one even begin to try to fit a distinctness measure on physical processes? One worry is that no objective criteria may be available when it comes to constructing a distinctness measure on “the physical processes themselves,” yet Hudson in his book seems unworried about that.

Hudson argues that “preserved” (because reliable) experimental methods issue in observational results to which a realist interpretation is in order (240–42). If those preserved observational methods issue in claims about such entities, attributes, and structures as intracellular organelles, cell nucleus shapes and densities, and blood cell morphologies, then we may be realists about such things. Yet, Hudson insists, nothing theoretical is involved in any essential way in our realist confidence about such entities, attributes, and structures (240–41). Is the latter a plausible claim? The answer surely depends on whether—and to what degree—one is a Quinean-style holist or a general nonholist about the empirical methods of mature science. Hudson is clearly a nonholist about these issues, one who believes in the reality of the “smoking-gun” single-case observational result that settles outright a theoretical scientific dispute. This is illustrated in the book during Hudson’s discussion of the “Bullet Cluster” case in astrophysics that, on his retelling of it, provided “direct evidence” for the existence of dark matter—indeed, the cover of his book contains an astrophotograph of the alleged smoking-gun evidence (although the astrophoto is not directly taken from Clowe et al. Reference Clowe, Bradac, Gonzalez, Markevitch, Randall, Jones and Zaritsky2006). As Hudson retells the 2006 research by Clowe and his colleagues that produced their claim of having found direct observational evidence for the presence of dark matter, a reader will note the many theoretical auxiliary hypotheses that Hudson simply skates over without comment, taking them as given, as correct, and as nondistorting. For example, one auxiliary hypothesis is that the galaxies in two colliding galaxy clusters do not in fact collide, but rather only their “intracluster [mostly gaseous] medium” collides, becomes hot, and emits X-rays in copious amounts (see 145ff.). That single auxiliary hypothesis—about celestial entities, we should note, that are approximately 3.4 billion light-years distant from Earth (about 1.9 x 10²¹ miles)—renders the Clowe result nonobservational in my view (as does the gravitational lensing involved in our observing of the Bullet Cluster), but Hudson is nevertheless convinced that Clowe and his colleagues more or less took a digital picture of dark matter effects. It is curious that instead of seeing epistemic hubris in the touting of smoking-gun, single-case, controversy-ending observational evidence, Hudson sees epistemic hubris in the asserting of second-order philosophical NMAs.

3. The Perils of Palm Reading

In a particularly partisan part of the book Hudson argues that robustness reasoning is popular—indeed, he thinks that it is too popular—among both scientists and philosophers of science precisely because one need not know anything about the many alternative methods that converge on a common result beyond the fact that they meet a minimal reliability constraint (198–200). Hudson then cites Ian Hacking’s example of a convergent outcome from two logically independent experimental processes (electron transmission and fluorescent reemission) to show the existence of “dense bodies” in red blood cells (191), only to argue that it would not have mattered if the second procedure had been replaced with “palm reading” (199). That is to say, for all the logical weakness of commonly used robustness reasoning in Hudson’s eyes, the second method might as well have been palm reading. But palm reading does not meet any minimal reliability constraint, nor is it historically a well-explored experimental method in science. We get this snarky section in the book because Hudson seeks to explain why a type of argument he sees as bogus is taken not to be bogus, but substantive and dispositive, by so many other smart and well-trained people. According to Hudson, the explanation is that so many of those other smart people are lazy and shallow, and so in their popular writings they use a “cheap thrills” convergence argument that they know will impress the unwashed masses (198–200). Again, this utterly misconstrues the seriousness of the NMA, never mind its ad hominem aspect.

There is a cognitive disconnect that suffuses this debate. The unexpected stability of a test result under various significant changes of testing procedures/devices can be empirically exciting to field scientists of course, but that empirical excitement has little to do with the time, years ago, when J. J. C. Smart, Grover Maxwell, and Hilary Putnam looked across the whole of modern science, as if from a high hill—a specifically situated high hill, mind you. They looked from working weather satellites to effective cobalt-60 cancer treatments, from hardy winter wheat to electron microscopes, from the moving continental plates of Earth to the relatively precise forecasting of a hurricane landfall time, from the discovery of E-class antibodies in the mammalian immune system to the synthesizing of an entire family of 1,4-substituted benzodiazepine ring agents as effective and reasonably safe antianxiety drugs, from the discovery of the Oort Cloud surrounding our solar system to the advent of CAT scans, from the discovery that the Sun’s luminosity depends on the eerie process of quantum tunneling to the physicists on Mount Washington in New Hampshire who in 1962 demonstrated the physical reality of the theory of special relativity, from the invention of the smaller-than-a-car digital computing machine to the Edison System of wiring buildings using a 120/240-volt three-wire distribution with a grounded neutral. And after they were done considering these myriad, deep, and detailed successes across the whole spectrum of the sciences, they said to themselves something to the effect of, “H, E, double hockey sticks, if we are wildly wrong in the scientific details of what we think we know about all these things, then their working as well as they do is a mind-numbing and bleeding-statues miracle of staggering proportions. But appeal to miracles—of staggering proportions especially—is forbidden in philosophy of science because it amounts to ‘giving up’ in the strongest sense possible. Therefore, we can be as sure as the primordial contingency of the universe allows to humans that we have at least a lion’s share of the scientific details correct.” This version of the NMA is clearly deductive—its specific logical form is modus tollens. It is also second-order—it is about science and not within science. Please friends, end the madness and keep Bayes’s theorem well away from this version of the argument at all times (Hudson’s book has four appendixes giving long proofs using Bayes’s theorem that a reader may skip without hazard, although confirmation theory enthusiasts will certainly enjoy their skillfulness). There are, of course, a number of abductive versions, as well as a number of probabilistic versions, of the NMA, most of them deriving ultimately from the work of Richard Boyd (for a classic presentation see Boyd Reference Boyd and Leplin1984). These abductive/probabilistic versions seem especially preferred these days by philosophers like Hudson over any type of deductive version like the one suggested above. Hudson’s book, however, fails to show with high confidence that the former preference constitutes a clear philosophical improvement over the latter option.