Inventing Temperature Read online

  The later memoir showed no sign that he had become any more optimistic about this issue (Regnault 1847, 165-166). If anything, he expanded further on the scornful remark just quoted about the careless earlier advocates of the gas thermometer. Regnault had no patience with the theoretical arguments trying to show that the thermal expansion of air was uniform, and he was all too aware of the circularity involved in trying to demonstrate such a proposition experimentally. Even when he noted the comparability between the air, hydrogen, and carbonic acid gas thermometers and the deviation of the sulfuric acid gas thermometer from all of them, he was careful not to say that the former were right and the latter was wrong: "Sulfuric acid gas departs notably from the law of expansion which the preceding gases show. Its coefficient of expansion decreases with temperature as taken by the air thermometer" (Regnault 1847, 190; emphasis added). He never strayed from the recognition that comparability did not imply truth. In the end, what Regnault managed to secure was only a rather grim judgment that everything else was worse than the air thermometer. Still, that was far from a meaningless achievement. This was the first time ever that anyone had advanced an argument for the choice of the correct thermometric fluid that was based on undisputed principles and unequivocal experimental results.

  Regnault's work on thermometry, like most of his experimental work, gained rapid and wide acceptance.56 His reasoning was impeccable, his technique unmatched, his thoroughness overwhelming. He did not back up his work theoretically, but he succeeded in avoiding theory so skillfully that he left no place open to any significant theoretical criticism. An important phase of the development of thermometry was completed with Regnault's publication on the comparability of gas thermometers in 1847. Ironically, just one year later the basic terms of debate would begin to shift radically and irreversibly, starting with the new theoretical definition of absolute temperature by the same young William Thomson who

  56. Statements to that effect are too numerous to cite exhaustively but see, for example, Forbes 1860, 958, and W. Thomson 1880, 40-41. As I will discuss further in chapter 4, Thomson (Lord Kelvin) always relied on Regnault's data in his articles on thermodynamics, with continuing admiration.

  end p.83

  had just made his humble pilgrimage to Regnault's laboratory. The conceptual landscape would become unrecognizably altered by the mid-1850s through the promulgation and acceptance of the principle of energy conservation, and subsequently by the powerful revival of the molecular-kinetic theory of heat. What happened to the definition and measurement of temperature through that theoretical upheaval is a story for chapter 4.

  Analysis: Measurement and Theory in the Context of Empiricism

  It is not that we propose a theory and Nature may shout NO. Rather, we propose a maze of theories, and Nature may shout INCONSISTENT.

  Imre Lakatos, "Criticism and the Methodology of Scientific Research Programmes," 1968-69

  In chapter 1 there were no particular heroes in the narrative. In this chapter there is one, and it is Victor Regnault. Hagiography is uninteresting only if it keeps celebrating the tired old saints in the same old way. Regnault's achievement deserves to be highlighted because it has been ignored for no good reason. Today most people who come to learn of Regnault's work tend to find it quite pedestrian, if not outright boring. I hope that the narrative was sufficient to show that Regnault's solution of the thermometric fluid problem was no ordinary success. It is rare to witness such an impeccable and convincing solution to a scientific problem that had plagued the best minds for such a long time. The qualities shown in Regnault's handling of this problem also pervaded his work in general. I will now attempt to elucidate the nature and value of Regnault's achievement further, by analyzing it from various angles: as a step in the extension of observability; as a responsible use of metaphysics; as a solution to the problem of "holism" in theory testing; and as the culmination of post-Laplacian empiricism in French physics.

  The Achievement of Observability, by Stages

  The improvement of measurement standards is a process contributing to the general expansion and refinement of human knowledge from the narrow and crude world of bodily sensations. The challenge for the empiricist is to make such improvement of knowledge ultimately on the basis of sense experience, since empiricism does not recognize any other ultimate authority. In the end I will argue that strict empiricism is not sufficient for the building of scientific knowledge, but it is worthwhile to see just how far it can take us. Regnault is the best guide on that path that we could ever hope for. His rigorous empiricism comes down to an insistence that empirical data

  end p.84

  will not be acquired by means of measurement procedures that themselves rely on hypotheses that have not been verified by observation.

  In order to see whether Regnault was really successful in his aim, and more generally whether and to what extent strict empiricism is viable, we must start with a careful examination of what it means to make observations. This has been a crucial point of debate in philosophical arguments regarding empiricism, especially concerning the viability of scientific realism within an empiricist epistemology. In my view, the key question is just how much we can help ourselves to, in good conscience, in constructing the empirical basis of scientific knowledge. The standard empiricist answer is that we can only use what is observable, but that does not say very much until we specify further what we mean by "observable." Among prominent contemporary commentators, Bas van Fraassen dictates the strictest limitations on what we can count as observable. What van Fraassen (1980, 8-21) means by "observability" is an in-principle perceivability by unaided human senses; this notion forms a cornerstone of his "constructive empiricism," which insists that all we can know about with any certainty are observable phenomena and science should not engage in fruitless attempts to attain truth about unobservable things.

  Some realists have attempted to invalidate the observable-unobservable distinction altogether, but I believe that van Fraassen has done enough to show that his concept of observability is coherent and meaningful, despite some acknowledged gray areas. However, I think that his critics are correct when they argue that van Fraassen's notion of observability does not have all that much relevance for scientific practice. This point was perhaps made most effectively by Grover Maxwell, although his arguments were aimed toward an earlier generation of antirealists, namely the logical positivists. Maxwell (1962, 4-6) argued that any line that may exist between the observable and the unobservable was moveable through scientific progress. In order to make this point he gave a fictional example that was essentially not so different from actual history: "In the days before the advent of microscopes, there lived a Pasteur-like scientist whom, following the usual custom, I shall call Jones." In his attempt to understand the workings of contagious diseases, Jones postulated the existence of unobservable "bugs" as the mechanism of transmission and called them "crobes." His theory gained great recognition as it led to some very effective means of disinfection and quarantine, but reasonable doubt remained regarding the real existence of crobes. However, "Jones had the good fortune to live to see the invention of the compound microscope. His crobes were 'observed' in great detail, and it became possible to identify the specific kind of microbe (for so they began to be called) which was responsible for each different disease." At that point only the most pigheaded of philosophers refused to believe the real existence of microbes.

  Although Maxwell was writing without claiming any deep knowledge of the history of bacteriology or microscopy, his main point stands. For all relevant scientific purposes, in this day and age the bacteria we observe under microscopes are treated as observable entities. That was not the case in the days before microscopes and in the early days of microscopes before they became well-established instruments of visual observation. Ian Hacking cites a most instructive case, in his well-known groundbreaking philosophical study of microscopes:

  end p.85

  We often regard Xavier Bichat as
the founder of histology, the study of living tissues. In 1800 he would not allow a microscope in his lab. In the introduction to his General Anatomy he wrote that: 'When people observe in conditions of obscurity each sees in his own way and according as he is affected. It is, therefore, observation of the vital properties that must guide us', rather than the blurred imaged provided by the best of microscopes. (Hacking 1983, 193)

  But, as Hacking notes, we do not live in Bichat's world any more. Today E. coli bacteria are much more like the Moon or ocean currents than they are like quarks or black holes. Without denying the validity of van Fraassen's concept of observability, I believe we can also profitably adopt a different notion of observability that takes into account historical contingency and scientific progress.

  The new concept of observability I propose can be put into a slogan: observability is an achievement. The relevant distinction we need to make is not between what is observable and what is not observable to the abstract category of "humans," but between what we can and cannot observe well. Although any basic commitment to empiricism will place human sensation at the core of the notion of observation, it is not difficult to acknowledge that most scientific observations consist in drawing inferences from what we sense (even if we set aside the background assumptions that might influence sensation itself).57 But we do not count just any inference made from sensations as results of "observation." The inference must be reasonably credible, or, made by a reliable process. (Therefore, this definition of observability is inextricably tied to the notion of reliability. Usually reliability is conceived as aptness to produce correct results, but my notion of observability is compatible with various notions of reliability.) All observation must be based on sensation, but what matters most is what we can infer safely from sensation, not how purely or directly the content of observation derives from the sensation. To summarize, I would define observation as reliable determination from sensation. This leaves an arbitrary decision as to just how reliable the inference has to be, but it is not so important to have a definite line. What is more important is a comparative judgment, so that we can recognize an enhancement of observability when it happens.

  These considerations of "observation" and "observability" give us a new informative angle on Regnault's achievement. Regnault's contribution to thermometry was to enhance the observability of temperature as a numerical quantity, and to do so without relying on theories. In the philosophical discussions of observability I have just referred to, a very common move is to allow the inferences involved in observation to be validated by scientific theories. For reasons that will be discussed in detail in "Minimalism against Duhemian Holism" and "Regnault and Post-Laplacian Empiricism," Regnault chose a stricter empiricist strategy for the validation of temperature standards. Most of all he wanted to avoid reliance on any quantitative theories of heat, since those theories required verification through readings of

  57. See, for instance, Shapere 1982, Kosso 1988, and Kosso 1989 for elaborations on the view that observation consists of a causal chain of interactions conveying information from the observed object to the observer, and a reverse chain of inferences by which the observer traces the flow of information.

  end p.86

  an already established numerical thermometer. In establishing the observability of the numerical concept of temperature, he could not use any presumed observations of that very quantity. How, then, did he go about establishing the observability of numerical temperature?

  Let us briefly review the overall evolution of temperature standards, as discussed in "The Iterative Improvement of Standards" in chapter 1. If we only have recourse to unaided senses (the "stage 1" standard), temperature is an observable property only in a very crude and limited sense. The invention of the thermoscope (the "stage 2" standard) created a different kind of temperature as an observable property. Numerical thermometers (the "stage 3" standard) established yet another kind of temperature concept that was observable. The single word "temperature" obscures the existence of separate layers of concepts. Now, some people certainly had a theoretical concept of temperature as a numerical quantity before numerical thermometers were actually established. At that point, temperature—stage-3 numerical temperature—existed as an unobservable property. It became observable later. Observability is neither dichotomous nor completely continuous; it progresses, improving continuously in some ways, but also in distinct stages with the successive establishment of distinctly different kinds of standards.

  Regnault's air thermometer was the best stage-3 temperature standard ever, to date. In order to establish its reliability (that is to say, to establish the observability of the numerical temperature concept), Regnault used comparability as a nontheoretical criterion of reliability, as I will discuss further in the next two sections. But he also needed other concepts whose observability was already well established, including the ordinal (thermoscope-based) temperature concept. That becomes clearer if we examine more clearly the actual construction and use of Regnault's instrument. Since his air thermometer was the constant-volume type, what it allowed was the determination of temperature from pressure. Such an instrument requires at least a qualitative assurance that the pressure of air varies smoothly with its temperature, which can be verified by means of stage-2 instruments. But how could temperature be determined from pressure in good conscience, when numerical pressure is no more sensory than numerical temperature, and no less theoretical than numerical temperature? That is only thanks to the historical fact that numerical barometers and manometers had already been established to a sufficient degree; therefore, for Regnault pressure was observable as a numerical quantity.58

  There was another important aid to Regnault's air thermometer, and that was actually the numerical mercury thermometer. This is a difficult matter, which should be considered carefully. Regnault used numerical mercury thermometers in order to measure the temperatures of the air in the tubes connecting the main

  58. In a manometer, pressure was determined from the length of the mercury column. The length of the mercury column had been established as an observable quantity before this experiment and prior to the observability of numerical pressure; that is actually not a trivial point, since the precision measurement of length in situations like this was an advanced-stage operation, often done by means of a telescope and a micrometer, in order to prevent disturbances from the observer's own body.

  end p.87

  Figure 2.2. Regnault's constant-volume air thermometer, illustration from Regnault 1847, 168-171, figs. 13 and 14. Courtesy of the British Library.

  section of the air thermometer (the large flask A in Figure 2.2) to the manometer (between point a and the mercury-level α in the right-hand figure).59 That was necessary because the air in the tubes could not be kept at the same temperature as the air in the large flask in the heat bath, and for practical reasons it was impossible to apply any kind of air thermometer to the tubes. But how was the use of the mercury thermometer allowable, when it had not been validated (worse yet, when Regnault himself had discredited it)? It might seem that even Regnault could not be completely free of unprincipled shortcuts. But the use of the mercury thermometer for this purpose was quite legitimate, for a few different reasons.

  59. Regnault's method of obtaining the air-thermometer reading was in fact quite complex. He calculated the temperature in question by equating the expressions for the weight of the air in its "initial" state (at 0°C) and its heated state (at the temperature being measured). For further details, see Chang 2001b, 279-281.

  end p.88

  First of all, the amount of air contained in those thin tubes was quite small, so any discrepancy arising from slight errors in the estimate of their temperatures would have been very small; that is a judgment we can reach without consulting any thermometers. Second, although the mercury thermometer was shown to lack comparability generally, the failure of comparability was less severe in lower temperatures (see data in table 2.4); that is helpful since we
can expect that the hot air would have cooled down quite a bit in the connecting tubes. Third, earlier comparisons between the mercury thermometer and the air thermometer had shown nearly complete agreement between 0°C and 100°C, and quite good agreement when the temperature did not exceed 100°C too far. Therefore, in the everyday range of temperatures, the reliability of the mercury thermometer stands or falls with the reliability of the air thermometer itself, and the use of the mercury thermometer does not introduce an additional point of uncertainty. Finally, there is a point that ultimately overrides all the previous ones. In the end Regnault did not, and did not need to, justify the details of the design of his thermometers; the test was done by checking comparability in the final temperature readings, not by justifying exactly how those readings were obtained. The use of mercury thermometers in Regnault's air thermometer was only heuristic in the end and did not interfere with his establishment of numerical temperature as an observable property by means of the air thermometer.