Inventing Temperature Page 18
Successful overdetermination, too, can be more reassuring when there are fewer other hypotheses involved. It is always possible to argue that a given case of successful overdetermination is a result of coincidence, with errors systematically canceling each other out. This is precisely how Dalton criticized De Luc, as I have already mentioned. In contrast, Regnault's experiment was so austere in its logical structure that it left hardly any room for that kind of criticism. Generally speaking, involving a larger number of assumptions would allow more possibilities for
end p.94
explaining away successful overdetermination. Minimalism provides one clear way of fighting this problem.
All of that may sound uncontroversial, but minimalism actually goes against the conventional wisdom, since it recognizes virtue in circularity. What I am calling the conventional wisdom here actually goes back to Duhem. He argued that the physicist had more to worry about than the physiologist regarding the theory-ladenness of observation because laboratory instruments were generally designed on the basis of the principles of physics. Hence, while the physiologist could proceed on the basis of a faith in physics, the physicist was stuck in a vicious circle in which he had to test the hypotheses of physics on the basis of those same hypotheses of physics.62 There is a widespread impulse to break out of this circle. The minimalist advice, on the contrary, is to tighten the circle.
In the case of negative test outcomes, the lesson from the success of Regnault's minimalism is very clear: we need to lose the unfounded fear that the test of a theory by observations that rely on the same theory will produce a vacuous confirmation of that theory. Whether an apparent confirmation obtained by such circular testing is worthless is an open question. What is certain is that there is no guarantee that observations enabled by a particular theory will always validate that theory. That is a point that was noted at least as early as 1960 by Adolf Grünbaum (1960, 75, 82 ). Discussing the case of physical geometry in the context of his argument against "Duhem's thesis that the falsifiability of an isolated empirical hypothesis H as an explanans is unavoidably inconclusive," Grünbaum noted: "The initial stipulational affirmation of the Euclidean geometry G 0in the physical laws P 0used to compute the corrections [for distortions in measuring rods] in no way assures that the geometry obtained by the corrected rods will be Euclidean." A similar point has also been made more recently by others, including Allan Franklin et al. (1989) and Harold Brown (1993). Therefore we can see that Karl Popper was being somewhat careless when he declared that "it is easy to obtain confirmations, or verifications, for nearly every theory—if we look for confirmations" (1969, 36). In fact it is not always so easy to obtain confirmations. And when a theory is falsified despite being tested by observations made on its basis, it will be very difficult to evade the falsification. The circularity here is a form of the minimalism that renders a negative test result more assuredly damning, as discussed earlier. Therefore there is no clear reason to wish for theory-neutral observations or seek what Peter Kosso (1988, 1989) calls "independence" between the theory of the instrument and the theory to be tested by the observation produced by the instrument.63
Even in the case of positive test-outcomes, the comfort provided by independence is illusory. Duhem's physiologist relying on the laws of physics can be comforted only as far as those laws of physics are reliable. Taking observations away
62. See Duhem [1906] (1962), part 2, ch. 6, sec. 1 (pp. 180-183).
63. Rottschaefer (1976, 499), more than a decade earlier than Kosso's publications, already identified a similar doctrine of "theory-neutrality" as a centerpiece of a "new orthodoxy": "Thus the view that theories are tested by theory-free observations is being replaced by the view that theories are tested by theory-laden observations, but observations laden with theories neutral to the theory being tested."
end p.95
from the theory being tested is a good policy only if there are other good theories that can support relevant observations. Minimalism is a plausible strategy in the absence of such alternative theories. As we have seen in the case of De Luc, confirmation is devalued if there is suspicion that it could be a result of unexpected coincidences; minimalism reduces that kind of suspicion, by cutting out as many possible sources of uncertainty in the testing procedure.
Before closing this discussion, I must mention some clear limitations of Regnault's minimalism, as a reminder that I am only admiring it as a creative and effective solution to particular types of problems, not as a panacea. There is no guarantee that a clear winner would emerge through minimalist testing. Fortunately for Regnault, the air thermometer turned out to be the only usable thermometer that survived the test of comparability. But we can easily imagine a situation in which a few different types of thermometers would all pass the comparability test, yet still disagree from each other. It is also imaginable that there might be no thermometers at all that pass the test very well. Minimalism can create a stronger assurance about the verdict of a test when there is a verdict, but it cannot ensure the existence of a clear verdict. Like all strategies, Regnault's strategy worked only because it was applied in appropriate and fortunate circumstances.
Regnault and Post-Laplacian Empiricism
Regnault's empiricism was forged in the context of the empiricist trend dominant in post-Laplacian French science. In order to reach a deeper understanding of Regnault's work, it is important to examine his context further. The direction taken by French physics in the period directly following the end of Laplacian dominance is an important illustration of how science can cope with the failure of ambitious theorizing. The post-Laplacian phase had two major preoccupations: phenomenalistic analysis in theory and precision measurement in experiment. Let us take a closer look at each preoccupation.
The phenomenalist trend, at least in the field of thermal physics, seems to have been a direct reaction against Laplace; more generally, it constituted a loss of nerve in theorizing about unobservable entities. Very symptomatic here was the rise of Jean Baptiste Joseph Fourier (1768-1830). According to Robert Fox (1974, 120, 110 ), Fourier became "a benign, influential, but rather detached patron of the new generation" of anti-Laplacian rebels including Pierre Dulong (1785-1838), Alexis-Thérèse Petit (1791-1820), François Arago (1786-1853), and Augustin Fresnel (1788-1827). In contrast to the Laplacian dream of the one Newtonian method applied to all of the universe, the power and attraction of Fourier's work lay in a conscious and explicit narrowing of focus. The theory of heat would only deal with what was not reducible to the laws of mechanics: "whatever may be the range of mechanical theories, they do not apply to the effects of heat. These make up a special order of phenomena, which cannot be explained by the principles of motion and equilibrium" (Fourier [1822] 1955, 2; see also p. 23 ).
Fourier remained noncommittal about the ultimate metaphysical nature of heat, and in theorizing he did not focus on considerations of "deep" causes. The starting point of his analysis was simply that there be some initial distribution of heat, and
end p.96
some specified temperatures on the boundaries of the body being considered; by what mechanisms these initial and boundary conditions would be produced and maintained were not his concerns. Then he produced equations that would predict the observed diffusion of the initial distribution over time, and he hardly made any attempt at a metaphysical justification of his equations. The antimetaphysical bias in Fourier's work had a good deal of affinity to positivist philosophy. As documented by Fox, Fourier attended the lectures of Auguste Comte (1798-1851) on positivism in 1829; Comte for his part admired Fourier's work, so much so that he dedicated his Course of Positive Philosophy to Fourier.64 The affinity of Fourier's work to positivism was also emphasized in Ernst Mach's retrospective appraisal: "Fourier's theory of the conduction of heat may be characterized as an ideal physical theory. … The entire theory of Fourier really consists only in a consistent, quantitatively exact, abstract conception of the facts of conduction of heat—in an eas
ily surveyed and systematically arranged inventory of facts" (Mach [1900] 1986, 113).
Fourier represented only one section of heat theory emerging in the twilight of Laplacianism, but the phenomenalistic trend away from microphysics was a broader one, though by no means unanimous. Another important instance of phenomenalism was the work of the engineer and army officer Sadi Carnot (1796-1832), which will be discussed further in "William Thomson's Move to the Abstract" in chapter 4. Carnot's Reflections on the Motive Power of Fire (1824) was based on a provisional acceptance of the caloric theory, but it steered away from microphysical reasoning. His analysis of the ideal heat engine only sought to find relations holding between the macroscopic parameters pertaining to a body of gas: temperature, pressure, volume, and the amount of heat contained in the gas; all but the last of these variables were also directly measurable. When the civil engineer Émile Clapeyron (1799-1864) revived Carnot's work in 1834, and even when William Thomson initially took up the Carnot-Clapeyron theory in the late 1840s, it was still in this macroscopic-phenomenalistic vein, though Thomson's later work was by no means all phenomenalistic.
In addition to phenomenalism, experimental precision was the other major preoccupation in the empiricism that increasingly came to dominate nineteenth-century French physics. In itself, the quest for experimental precision was quite compatible with Laplacianism, though it became more prominent as the Laplacian emphasis on microphysical theorizing waned. According to many historians, the drive toward precision measurement was a trend that originated in the "quantifying spirit" originating in the Enlightenment,65 which continued to develop through and beyond the heyday of Laplace. The highest acclaim for precision in the early nineteenth century went to Dulong and Petit, both identified by Fox as leading rebels against Laplacian physics. The Dulong-Petit collaboration is perhaps best known now for their controversial "law of atomic heat" announced in 1819 (the observation that the product of atomic weight and specific heat is constant for all
64. The dedication was shared with Henri Marie Ducrotay de Blainville, the anatomist and zoologist. See Fox 1971, 265-266, for further discussion of Fourier's relationship with Comte.
65. See, for example, Frängsmyr et al. 1990 and Wise 1995.
end p.97
elements, which was taken to imply that all individual atoms had the same heat capacity). However, it was their two earlier joint articles on thermal expansion, the laws of cooling and thermometry (1816 and 1817) that won them undisputed respect at home and abroad, a glimpse of which can be had in Lamé's statement quoted in "Regnault: Austerity and Comparability."66
These trends formed the style of science in which Regnault was educated and to which he contributed decisively. Duhem (1899, 392) credited Regnault with effecting "a true revolution" in experimental physics. In Edmond Bouty's estimation (1915, 139): "For at least twenty-five years, the methods and the authority of Regnault dominated all of physics and became imperative in all research and teaching. Scruples for previously unknown degrees of precision became the dominant preoccupation of the young school." But what was so distinctive and powerful about Regnault's work, compared to the works of his important empiricist predecessors such as Fourier, Carnot, Clapeyron, Dulong, Petit, and Lamé?
One point is clear, and banal: Regnault's revolution was a revolution in experimental physics, not in theoretical physics. Regnault made few contributions to theory, and theory could not help the revolution that Regnault was trying to launch. Contrast that to the work of other phenomenalists. Although Fourier and Carnot were empiricists, they did not contribute very much to empirical work; that is not a paradox, only a play on words. Theorizing about observable properties does not necessarily have anything to do with producing actual observations. Fourier's and Carnot's theories did nothing immediately to improve observations. Take, again, the case of thermometry. Because Fourier declined to deal with any mechanical effects of heat (including the thermal expansion of matter), the tradition of heat theory established by him could give no help in elucidating the workings of the thermometer. In fact Fourier displayed a remarkable degree of complacency about thermometry ([1822] 1955, 26-27). Carnot's theory of heat engines only made use of the presumably known relations regarding the thermal expansion of gases and could not make any contributions toward a justification of those relations. Fourier and Carnot were at best consumers of empirical data, and the best that consumers can do is to stimulate production by the demand they create.
Phenomenalists in the tradition of Fourier and Carnot were only antimetaphysical; Regnault was antitheoretical. That is to say, even phenomenological theory was a target of skepticism for Regnault. His experiments subjected well-known empirical laws to minute scrutiny with corrosive effect. Regnault had initially come to physics from chemistry through the study of specific heats in relation to Dulong and Petit's law.67 After finding that law to be only approximately true (as many had suspected in any case), he turned to the more trusted regularities
66. The latter paper won a prize competition of the Paris Academy. Fox (1971, 238) lists Comte, Poisson, Lamé, and Whewell as some of the leading authorities who admired this work as a model of experimental method.
67. See Dumas 1885, 162. The results from this investigation were published in Regnault 1840, as well as two subsequent articles.
end p.98
regarding the behavior of gases. As Regnault cranked up the precision of his tests, even these laws were shown up as tattered approximations. Already by 1842 Regnault had collected enough data to refute two laws that had been regarded as fundamental truths regarding gases: (1) all types of gases expand to the same extent between the same limits of temperature, which was the conclusion of Gay-Lussac's and Dalton's experiments forty years earlier; (2) a given type of gas expands to the same extent between the same limits of temperature regardless of its initial density, which had been generally believed since Amontons's work in 1702.68 Regnault's memoir of 1847 repeated the refutation of these laws with further details and also gave results that showed Mariotte's (Boyle's) law to be only approximately and erratically true.69
These experiences disillusioned Regnault about the presumed progress that experimental physics had made up to his time. If some of the most trusted empirical regularities were shown to be false, then none of them could be trusted without further assurance. From that point on he eschewed any reliance on presumed laws and set himself the task of establishing true regularities by thorough data-taking through precision measurements. While he was engaged in this enterprise, it is understandable that he did not find himself excited by the new theoretical speculations issuing from the seemingly fickle brains of the likes of Faraday, Ørsted, Joule, and Mayer, whom posterity has praised for their bold and penetrating insights.70 When De Luc said that "the moral and physical Microscope are equally fit to render men cautious in their theories" (1779, 20), he could not have anticipated the spirit of Regnault's work any better. In Berthelot's estimation, Regnault was "devoted to the search for pure truth, but that search he envisioned as consisting above all in the measurement of numerical constants. He was hostile to all theories, keen to emphasize their weaknesses and contradictions" (Berthelot quoted in Langevin 1911, 44-45). For Regnault, to search for truth meant "to replace the axioms of theoreticians by precise data" (Dumas 1885, 194).
It may seem that the task Regnault set for himself was a laborious yet straightforward one. However, to someone with his intellectual integrity, it was all too obvious that the existing measurement methods relied on theoretical regularities, exactly the kind that he was hoping to test conclusively by measurements. Thus, Regnault came face to face with the fundamental circularity of empiricist theory testing. A complete avoidance of theory would have paralyzed experiment altogether. With the recognition that each experiment had to take some assumptions for granted, Regnault's conscience forced him to engage in further experiments to
68. See Regnault 1842a, 1842b. Part 2 of each memoir deals with the former law, and part
1 of the first memoir deals with the latter.
69. On the two expansion laws, see Regnault 1847, 91, 119-120. Regarding Mariotte's law, Regnault's results (1847, 148-150, 367-401) showed that it held for carbonic acid at 100°C but not at 0°C, even at low densities; for atmospheric air and nitrogen it was generally not true; the behavior of hydrogen also departed from it, but in the opposite direction from air and nitrogen. Regnault had entertained the belief that the gas laws would be true "at the limit" (that is, when the density of the gas approaches zero), in the conclusion of his earlier investigation (Regnault 1842b). By 1847 he seems to have abandoned even such limited hope.
70. See Dumas 1885, 191.
end p.99
test those assumptions. There was no end to this process, and Regnault got caught up in what Matthias Dörries has characterized as a never-ending circle of "experimental virtuosity." It seems that Regnault's original intention was to start with observations cleansed of theory, then to move on to careful theorizing on the basis of indisputable data. However, the task of obtaining indisputable data turned out to have no end, which meant that theoretical activity had to be postponed indefinitely. Regnault himself seems to have felt some frustration at this aspect of his work; in 1862 he referred to a circle that he was not able to get out of: "[T]he more I advanced in my studies, the more the circle kept growing continually …"71 Such sense of frustration is probably behind much of the lukewarm appraisals that Regnault's work has received from later generations of scientists and historians.72 For example, Robert Fox is unequivocal in acknowledging Regnault's "monumental achievements," but at the same time he judges that Regnault's "preoccupation with the tedious accumulation of results" was unfortunate especially in view of "the momentous developments in physics taking place outside France during the 1840s" (1971, 295, 299-300).