1. Introduction

Caloric theory, phlogiston chemistry, the ether-based theories of light, all once thought to be sufficiently corroborated, have nevertheless come to be rejected. Time and again in the history of science, theories that enjoyed spectacular success have proven to be false. So, according to Larry Laudan and others, it is a mistake to think that a theory's success attests to its truth-likeness or that the latter provides the best explanation for its empirical success. A major lineage of realists, from Worrall (Reference Kitcher1989a, Reference Helmholtz1989b) to Kitcher (Reference Chakravartty1993), Leplin (Reference Einstein1997), Psillos (Reference Fresnel1999), and Bain and Norton (Reference Bain, Norton, Buchwald and Warwick2004), among others, reject this skeptical reading of the history of science. They agree that past theories taken as whole are false but insist that theories that succeed in terms of risky predictions get right at least some of their distinct accounts. For example, it is noted that, although Fresnel's theory turned out to be false, substantial parts of its properly theoretical claims remain nevertheless hard to question to this day—for example, that light is made of transversal waves that (to a very high approximation) obey Fresnel's laws of polarization, reflection, and refraction. A more intricate but analogous report can be made regarding the transition to Maxwell's theory.

Realists agree that theory successions over an empirical domain of phenomena display conceptual discontinuities but deny that the latter are either as widespread or as radical as Laudan and his followers suggest, at least in mature disciplines. In particular, Kitcher, Leplin, and Psillos (KLP) have cogently stressed that successful theories are not something one takes or rejects in toto: a false theory can still be approximately true in significant ways. According to these realists, through modern science there has emerged a fairly stable and well-supported network of theory parts, amounting to clusters of theoretical assertions and posits, which is our best account of what the world is like. Leplin voices this claim well: “Where past theories have met the standards imposed for warranting theoretical belief, their eventual failure is not a total failure; those of their theoretical mechanisms implicated in achieving that warrant are recoverable from current theory” (Reference Einstein1997, 145). In their analyses of Fresnel's theory of diffraction, KLP variously conclude that the theory did not invoke the ether essentially at points relevant to Fresnel's derivation of diffraction patterns. Fresnel, they claim, did not need the ether in order to derive the experimental tests that made his theories successful. Unfortunately, this line of response has proven problematic.

What exactly, in Fresnel's derivation of diffraction patterns, made his use of the ether concept “inessential” or “idle”? Psillos (Reference Fresnel1999) offers an initially promising criterion in response to this question: “Suppose that a hypothesis H together with another set of hypotheses BH (and some auxiliaries A) entail a prediction P. Then H indispensably contributes to the generation of P if BH and A alone cannot yield P and no other available hypothesis H* which is consistent with BH and A can replace H without loss in the relevant derivation of P” (110). Two crucial issues here are what H actually amounts to and to whom the alternative hypothesis H* is available. Critics have variously pointed out that, in the ether case and other major cases, the scientists involved took extremely seriously theory components that KLP say were “idle” or “inessential.” From Fresnel on, none of the nineteenth-century theorists challenged the existence of some sort of ether for light.

Psillos's criterion counts on synchronic tracking being discerning enough to single out theoretical constituents that get right some deep aspect of the domain at hand, but the arguments and case studies KLP have provided for this seem flawed. Objectors, including many with realist leanings, have raised specific complaints along a variety of fronts. For example, Chang (Reference Chakravartty2003) rejects KLP's analyses of the case of caloric theory, arguing that many features of caloric that were rejected by subsequent physics were central to the success of caloric theory; Carrier (Reference Carrier and Carrier2004) further questions the idea that the theoretical terms of successful science typically refer to real objects, arguing that the history of science rules out realist positions except about natural kinds. Chakravartty (Reference Chakravartty2003, Reference Carrier2007) reacts to the problems of full realism about selected theory parts by advocating an intermediate stance between realism and nonrealism according to which we learn about structures and entities in the world (their dispositions, capabilities, and contexts) by sorting out the kinds of relations between things that structural realists emphasize. Lyons (Reference Fleming2006, 543) stresses that “credit will have to be attributed to all responsible constituents, including mere heuristics (such as mystical beliefs), weak analogies, mistaken calculations, logically invalid reasoning, etc.” I much agree with these critics but think that more needs to and can be said on the KLP approach to realism and the case studies they use.

In mature scientific disciplines, theories are tight constructs; breaking them into certain parts is often not feasible in their early stages. I am particularly troubled by the kind of options KLP assume past theoreticians had. Fresnel, Arago, Maxwell, Hertz, Lorentz, and so forth, all invoked the ether of light, a posit now rejected. The derivations each of these thinkers made, KLP believe, could have followed alternative paths that bypassed the now discarded posit. But this presumption is problematic. One major issue concerns the extent to which the envisaged alternative paths were (if at all) accessible to the past theorists in question (as opposed to just us). If, as KLP assume, in most salient historical cases “alternatives paths” were indeed available to theorists at the time the theory first showed predictive power, then realists have a promising “synchronic” version of the divide et impera strategy for any markedly successful theory: (a) identify those parts for which there were no “options” when the theory's winning predictions were first made; (b) show that the parts in question, far from being characteristically false, have been retained in subsequent theories of the same domain; and (c) recognize these parts as sound components of the theory, ones actually responsible for its success. We would have a neat way of selecting theory parts for realist commitment. However, if in many major cases the alternative paths invoked by KLP happened to be available only to us, then it would be difficult not to charge KLP with indulging in arbitrary retrospective projection of current theories. In the following sections, I consider this issue by looking deeper into the ether case study.

2. Fresnel and Whewell

Confidence in the wave theory grew with its success in terms of novel predictions and surprising convergences and connections. Fresnel showed convergences of theoretical results drawn from different starting points, which reassured him about the ether posit. His explorations included different derivations for important predictions. Fresnel (Reference Fresnel1821/1868) even deploys an “alternative,” geometrical approach that does not explicitly posit the ether and yet arrives at an equation “of the right sort.” Having done this, however, Fresnel thinks it crucial to devise an ether-based dynamical scheme to accommodate it and goes on until he reaches the same wave front to which he had been originally led by purely geometrical considerations. As far as Fresnel's understanding of his theory's success was concerned, his geometrical conjecture could now be regarded as substantiated by a study of the dynamics of the medium (Whittaker Reference Ganot and Atkinson1953). At the end of the memoir, Fresnel says: “The theory which I have adopted, and the simple constructions which I have deduced from it, have this remarkable character, that all the unknown quantities are determined together by the solution of the problem. We find at the same time the velocities of the ordinary ray and of the extraordinary ray, and their planes of polarization. Physicists who have studied attentively the laws of nature will feel that such simplicity and such close relation between the different elements of the phenomenon are conclusive in favor of the hypothesis on which they are based” (Fresnel Reference Fresnel1822/1868, 593; quoted in Whittacker 1953/1910, 130).

Epistemological insights soon to become central to science are already discernible in the above quotation. In the 1830s and 1840s, William Whewell, who followed with great interest the success of the wave theory and whose work on initially unexpected convergences drew in part from the sense of epistemological worth voiced by Fresnel's quote, gave those insights emblematic expression. Whewell was also a staunch optimist about the growth of theoretical knowledge. He credits past theories for specific (partial) discoveries and goes as far as presenting “undesigned scope” as a criterion of reality, stressing that theories that contributed undesigned success have never been subsequently abandoned in “all respects” (Reference Whewell1847, pt. 2, bk. 11). Fresnel's theory had been good in many regards. Confidence in the ether gained strength from this kind of scientific epistemology.

3. After Fresnel

By 1860, alternative (Lagrangian) derivations of Fresnel's equations were available, but top scientists of the period continued to think the ether “essential.” Scientific commitment to the posit did not much change as the century progressed (Whittaker Reference Ganot and Atkinson1953; Swenson Reference Fresnel1972). As Maxwell famously put it, “Whatever difficulties we may have in forming a consistent idea of the constitution of the ether, there can be no doubt that the interplanetary and interstellar spaces are not empty, but are occupied by a material substance or body, which is certainly the largest, and probably the most uniform body of which we have any knowledge” (Encyclopedia Britannica, 9th ed., vol. 8 [1893], 572; this article first appeared in 1878). Influential textbooks of the time exude scientific confidence in the existence of the ether, while admitting that the dynamical theory of the posit remains incomplete. “A motion of a particular kind communicated to the ether can give rise to the phenomenon of heat; a motion of the same kind, but of greater frequency, produces light; and it may be that a motion different in form or in character is the cause of electricity” (Ganot Reference Ganot and Atkinson1860/1893, 3). “[The question is not whether there is an ether but rather] how is the ether affected by the gross matter which it invests and permeates? Does it move when they move? If not, does the relative motion between the ether and other matter change the length of the undulation or the time of oscillation?” (Lovering Reference Lovering1874, 301). “You can imagine particles of something, the thing whose motion constitutes light. This thing we call the luminiferous ether. That is the only substance we are confident of in dynamics. One thing we are sure of, and that is the reality and substantiality of the luminiferous ether” (Thomson Reference Thomson1889, 310; quoted in Thompson Reference Whittaker1910, 1035; Thomson was later Lord Kelvin). “There can no longer be any doubt that light waves consist of electric vibrations in the all-pervading ether, and that the latter possesses the properties of an insulator and a magnetic medium” (Helmholtz Reference Helmholtz1899, xvi).

It will not do to point to the bizarre character of the postulated ether. Leading light theorists did not single out the ether as a seriously weak proposal until very late in the nineteenth century. Before that time they did not consider the ether too “speculative” or “unsupportive.” Its weird physical characteristics and peculiarities were appreciated as such, but theorists responded to them by appealing to analogies that seemed relevant at the time (e.g., with some viscous liquids, also with precursor notions of resonant transmission for high frequencies).

No less telling about confidence in the reality of the ether are the wealth of end-of-century reports on the “state of physics” around 1900 by respected commentators who go out of their way to claim that physics could not do without the ether. Negative results like those of the Michelson and Morley experiment (Reference Michelson and Morley1886) did not diminish their confidence, let alone the perceived intellectual need for the posit. To T. C. Mendenhall, for example, the demand for a medium to carry light waves was satisfied by “what is known as the ethereal medium, at a first purely imaginary substance, but whose real existence is practically established” (Reference Williams1901, 322). Numerous other reports agree: “[Regarding the ether] its discovery may well be looked upon as the most important feat of our century” (Williams Reference Williams1901/2007, 230). “Among the concepts which have come to stay in scientific thinking that of the ether must now be included. … It is as real as the concept of ‘atom’ or ‘molecule’ but hardly more so” (Thomson Reference Thomson1903/1906, 176). “[It is] a fact deduced by reasoning from experiment and observation. … There is abundant proof that it is not merely a convenient scientific fiction, but is as much an actuality as ordinary gross, tangible, and ponderable substances. It is, so to speak, matter of a higher order, and occupies a rank in the hierarchy of created things which places it above the materials we can see and touch” (Fleming Reference Fleming1902, 192; quoted in Swenson Reference Fresnel1972, 138). “[I am] practically certain that there must be a medium whose proper function is to transmit light waves” (Michelson Reference Michelson1903, 159).

And the ether lingered on. As late as 1919, we find Oliver Lodge urging that “the aether is needed for any clear conception of potential energy, for any explanation of elasticity, for any physical idea of the forces which unite and hold together the discrete particles of matter, whether by gravitation or cohesion or electric or magnetic attraction, as well as for any reasonable understanding of what is meant by the velocity of light” (Reference Lodge1919, 16–17). Lodge was even more dramatic in a speech he delivered in Edinburgh in 1921: “An intelligent deep-sea fish would disbelieve in water. … Such is our own condition in regard to the aether of space. … Let us grant, then, that the ether impinges on us only through our imagination; that does not mean that it is unreal. To me, it is the most real thing in the material universe” (Reference Lodge1921, 88 and 89; my italics).

None of the above “prove” that the ether was a “necessary” posit, of course. The passages simply document the strong commitment that physicists high in the hierarchy had for the ether part of the theory. Why were they so attached to the ether? The quoted expressions, together with the detailed deployment of the ether posit in theoretical analyses of the period, jointly suggest that more than empirical success and theoretical success were involved in the persistence of the ether hypothesis.

4. Conceptual Entrenchments

The assurances the cited scientists give also draw strength from an entrenched conception according to which to understand a physical system is to understand it mechanically. Consider William Thomson's expression of the view:

Equally relevant is the background inductive knowledge theorists had regarding waves: the prototypes for wave talk were water waves and sound waves, both of which “clearly” require a medium. Ever since waves entered the gaze of natural philosophers, the concept of “being a wave” had been understood as “being a propagating perturbation in a medium.”

More was at stake, however. The way of thinking appealed to in the previous quotations—the confident words used—had been firmly in place for a long time in physics. They express a way of thinking that had already been in place in modern science 200 years earlier and in which ideas remained embedded until relatively recently. Indeed I suggest that a substantial piece of “received metaphysics” was at work here. It comprised a hierarchy of being in which at the lowest level of what there is are “modes” or “states” of finite substance that depend for their existence on those finite substances. This view, which held a long grip on Western thought, was revitalized in early modern philosophy, particularly by Descartes (e.g., in Meditations III). The point is that, until the 1920s and even the 1930s, a wave remained regarded as a “mode of being”—that is, something completely dependent for its being on something else existing simultaneously (the material substratum whose propagated perturbation the wave was). As such, a wave ranked at the lowest level of being, along with shadows and smiles—a wave, it was thought, simply could not exist without some material substratum that was perturbed. This explains why the ether remained widely accepted as an empirically successful proposal in mainstream physics until so late in the story, its existence regarded as something beyond reasonable doubt.

So, to Fresnel (and Maxwell) the theoretical derivations that “succeeded” had conceptual links that were not up for grabs. Physics was embedded in a metaphysical framework that made the ether “necessary.” The physicists of the nineteenth century were not twentieth-century scientists. In particular, they were not prepared, let alone willing, to question conceptual links deemed “necessary” nearly as radically as subsequent generations of physicists would, beginning a few decades later.

Separating the ideas of wave and medium needed a level of conceptual atomization that would be encouraged only many decades after Fresnel, partly by the rise of positivist interpretations of science and partly by null-result experiments like those by Michelson and Morley. Full separation of the ideas of ‘wave’ and ‘requiring a physical medium’ would be at the heart of Einstein's revolutionary move at the dawn of the twentieth century and of much of the rage that it initially aroused in the scientific establishment.

Returning to the main point, if the above suggestions are correct, then contrary to KLP the optical ether was “essential” not just to Fresnel's theories such as they were but also to subsequent wave theories in the century. Denying synchronic centrality to the ether is an artifact of retrospective projection. In the nineteenth century, the ether hypothesis could not have properly been “replaced” by an independently motivated, non ad hoc, potentially explanatory alternative. Even though seemingly alternative theories were available (notably in terms of the Lagrangian formulation), to the mind of the scientists involved, being a wave implied having a medium, a view they held regardless of the formulation employed.

It might seem that the ether case works badly not just for KLP's synchronic approach but for the project of scientific realism in general. Not so, I suggest. All the above reservations notwithstanding, I think the case presented gestures toward a more natural, “diachronic” version of the divide et impera strategy, one that seizes on the gradual development of conceptual partitions, retention of theory parts, and epistemic warrant as these took place in the case at hand.

5. Dispensing with the Ether Took Time

Maxwell had struggled to connect his theory with mechanical ideas about the light ether. In 1884 Hertz rederived Maxwell's equations by a new method, casting them in their “modern form.” Although a supporter of the ether, Hertz's formulation bypassed all explicit assumptions regarding the ether; to subsequent generations, he “cleaned up” the theory, keeping just those parts that could not be left behind without incurring a serious empirical loss. Under Hertz's influence, physicists gradually came to take electric and magnetic field strengths as fundamental concepts that did not require mechanical explanation, and Maxwell's theory began to be regarded as simply “Maxwell's equations.” As already explained, however, virtually all the top figures of the age remained loyal to the old wider framework. An interesting player in this regard is Lorentz, who distanced the ether from ordinary matter by presenting it as something that did not transmit mechanical forces, was everywhere at rest, and completely penetrated matter. Now, if the ether was not dragged, then the earth should have an absolute velocity with respect to it. But the effects of the expected ether wind were not observed, and it was thus essential to explain why. Lorentz accounted for the null result of the Michelson and Morley experiment by introducing the idea that an adequate form of “length contraction” occurred as the effect of motion through the ether on intermolecular forces. In addition, in order to “save the appearances” Lorentz also had to change the time variable into what he named “local time.” Although Lorentz expected that mechanics would be replaced by electrodynamics as the basis of physics, he never abandoned his belief that the ether provided a unique privileged frame of reference at rest, in which Maxwell's equations obtained. Lorentz, as Einstein would playfully note, simply deprived the ether of all mechanical properties but immobility (Einstein Reference Einstein1920, 11–12).

By 1905 a different take on the fundamental issues at hand was in full flight in Einstein's work, now leaning toward a more radical stance on the ether issue. Einstein's rational motivations have been traced to many sources. One is his studies of philosophy (particularly the work of Kant and then the work of such empiricist interpreters of science as J. S. Mill, Karl Pearson, Richard Avenarius, and Ernest Mach). Another is the epistemological stance on novel prediction that prospered at the Zurich Polytechnic. There were also (if perhaps less so) the null results yielded by Michelson and Morley and other experimenters who tried to measure the ether's relative motion.

To Einstein, Lorentz's contraction hypothesis was put together just to save the idea of a stationary ether. Lorentz's “local time,” Einstein proposed, was just “time.” Then, on the matter of space, Einstein deprived the ether of its last mechanical quality—its immobility. Between 1905 and 1910 Einstein made clear that, in his view, the principle of the constancy of light together with the principle of relativity rendered the ether posit superfluous. Initially Einstein's ideas about space, time, and the ether enraged almost all his elders in physics. Soon, however, theoreticians of stature, notably Hermann Minkowski, began to appreciate the possibilities of special relativity as a theory of space-time. As the twentieth century progressed, other “categories of understanding” long presumed firm in physics would be challenged as well, starting a metaphysically revisionary trend that has gained force ever since.

6. Concluding Thoughts

Developments like the ether case strongly suggest that synchronic approaches to identification of sound theory parts are bound to fail. “Diachronic” versions of the divide et impera strategy might be a different matter, however. In optical theory the ether part eventually came to be singled out as an off-the-mark component and rejected accordingly. The required conceptual disentanglement took place, even if 200 years after the posit's introduction by Fresnel. However, other parts became identified as components that were both approximately correct and seemingly indispensable to current and future theories of light. None of these unveilings followed a synchronic strategy, but they did become available, many while the classical theories were still in full flight.

Realists might thus draw hope from the diachronic segregation of parts displayed by the history of Fresnel's theory and other cases. In the history reviewed here, in addition to doubts converging on the ether posit, a robust yield of putative truths became gradually highlighted (e.g., the vibrational features of light, its transversal character, basic propagation laws), amounting to arrays of thickly textured partial descriptions and narratives that were (and remain) recognized as being as safe and beyond reasonable doubt as the most compelling talk about “observable” objects. In many specifiable respects, light is largely as Fresnel said it was. Something similar holds for other past successful theories: in many respects atoms are largely as classical physics portrayed them, many chemical structures and transformations are as classical chemistry said, the natural histories of many species (e.g., whales) are in very many respects as Darwin's original theory said they were, among many other examples. The cases in question gesture toward a diachronic version of the divide et impera strategy, and so developing such a version seems a reasonable goal for realists. But, admittedly, a great deal more yet needs to be said to make the project compelling. The ghost of retrospective projection looms large against any attempt to generalize the strategy. Important questions seemingly stand in the way. In particular, prospective diachronic realists are under pressure to specify which components in current theories are the ones that can be expected to persist in successor theories and be judged approximately true from the point of view of future generations of theories and when the “success” and “gradual entrenchment” of a theory component is sufficient enough to ensure its probable, approximate truth.