Daniel Kennefick is a working physicist at the University of Arkansas who has for a couple of decades been writing well-informed historical accounts of different episodes in relativistic physics—most prominently, experimental research into gravitational waves and Einstein’s theory of general relativity. No Shadow of a Doubt systematically brings into one volume Kennefick’s research on the so-called Eddington expedition organized by the Royal Society in 1919 to test Einstein’s general relativity. The book was published on the centenary of the expedition and met with a warm reception from historians, sociologists, and practicing physicists alike. It was generally praised as a lively and accessible yet rigorous account of the intricate history. Philosophers have tended to be less attentive, and this motivates the current review. The book rewards the knowledgeable philosopher of spacetime physics with a rich tapestry of both historical detail and philosophical argument and it may appropriately be seen as part of the recent trend in integrated history and philosophy of science.

In terms of narrative style, the book is riveting—it often reads as a gripping detective story—and contains fifteen short and concise chapters (plus a brief prologue and epilogue), each to be read in one sitting. The main novelty in Kennefick’s account is that it focuses not on the celebrated Plumian Professor of Astronomy and Experimental Philosophy Arthur Eddington (1882–1944), whom the expedition is conventionally named after, but on the lesser-known Astronomer Royal Frank Watson Dyson (1868–1939), whose contribution to the expedition turns out to be of greater historical and philosophical relevance. The two men were lifelong friends and held each other in high esteem (Eddington’s ([1940]) obituary of Dyson is lavish in its praise). Kennefick’s book reveals that, ultimately, the determination of the evidence for general relativity rested mainly with Dyson and his remarkable assistants at the Greenwich Observatory. It was they who oversaw the analysis of the data coming out of the Brazilian branch of the expedition. Eddington, by contrast, led the African branch—heading off to the island of Principe—and both his data and results were less conclusive. Thus, we really ought to speak about ‘Dyson’s expedition’ or, at least, ‘the Dyson–Eddington 1919 expedition’. This change in the conventional narrative carries significant philosophical implications.

The experiment that prompted the expedition involved the measurement of the deflection of the light from distant stars as that light passed near the Sun. This required a total solar eclipse in order to observe, in the visual range of the electromagnetic spectrum, the apparent displacement of the positions of distant stars as they cross the Sun’s visual arch. The story of the expedition is recounted in multiple sources, including Eddington’s ([1920]) celebrated Space, Time, and Gravitation published just a year after the event, and I have myself introduced this historical episode to Spanish philosophical audiences (Suárez [2019]).

Kennefick makes a good point that it has become a convention to emphasize Eddington’s desire to see Einstein’s theory overthrow Newtonian mechanics. Eddington was indeed acutely aware of the shortcomings of the latter theory and is said to have been ‘one of less than three men’ at the time to fully understand the theory of relativity (the other one obviously meant to be Einstein himself). Even worse, Kennefick points to a popular misconception that Eddington faked his data so that Einstein’s theory would stand triumphant. Was this scepticism towards his work the result of suspicion about Eddington’s pacifism during the First World War? Both Einstein and Eddington were highly unusual and heterodox internationalists who resolutely condemned the war efforts on both sides. Kennefick relates with panache and some irony the troubles that this stance caused each of them in their local milieus in Berlin and Cambridge. It is striking how similar the two men’s political experiences were during the war, and how singularly they were each thrown into the profile of the lone outcast. Undoubtedly, there must have been much mutual empathy running between them.

However, the fact that it was ultimately down to Dyson and his assistants to assess the weight of the evidence against the Newtonian prediction of the starlight deflection is one significant correction to this proposed misconception (Dyson et al. [1920]). Another correction involves reassessing Eddington’s handling of his Principe data, to show that Eddington was more equanimous and fair minded than he is usually credited as being. Dyson was neither a pacifist nor known for any anti-war sentiment during or after the war, although he was certainly a believer in the international character of science in general. He was regarded as a patriot, and he credibly carried the diplomatic tasks inherent to the position of Astronomer Royal at Greenwich. There was no sense in which Dyson was invested in a post-war effort to bring the two warring nations together. Yet it was Dyson, not Eddington, who forcefully pushed for the interpretation of the data that most clearly supported Einstein’s theory over Newton’s.

The story has interesting philosophical implications for both the debate in the 1980s over the so-called experimenter’s regress, as well as ongoing efforts to develop a comprehensive philosophy of scientific data. In an influential article, philosophers John Earman and Clark Glymour ([1980]) jointly analysed the methodological background behind this famous test of Einstein’s theory of general relativity. Kennefick rightly praises their ‘carefully phrased’ scholarship (p. 186), even though he views Earman and Glymour’s work as the origin of the popular misconception that runs to the present day concerning the quality of the evidence for general relativity. Kennefick’s book is largely an attempt to redress this misconception, by taking a second look at the methodology of data analysis and collection during the Principe and Sobral expeditions, particularly chapter 11, intriguingly entitled ‘Not Only Because of Theory’. This is a phrase extracted from Eddington’s correspondence, which expresses how he was always inclined toward Einstein’s theory, but also makes clear that his acceptance of general relativity was ultimately down to the experimental results obtained.

One critical issue for assessing this episode concerns the decision to disregard one of two sets of data from the Sobral expedition. The expedition from the Greenwich observatory led by Dyson’s assistants had made sure to bring two telescopes with adjustable mirrors so as to produce two distinct sets of photographic plates. In Principe, however, Eddington only had one telescope and produced only one set of photographic plates. He also had to struggle with conditions that turned out to be less than ideal (‘Through Cloud, Hopefully’ is another chapter title inspired by a quip from Eddington’s correspondence).

Earman and Glymour point out that the set disregarded is the one prima facie less favourable to Einstein’s theory, and they suggest this implies some implicit bias for Einstein’s theory may have influenced the decision. Kennefick counters this with his careful vindication of Dyson’s central role in the phase of data analysis and provides textual evidence that the decision to discard the second set of Sobral data was ultimately Dyson’s, not Eddington’s. Dyson was, moreover, justified by antecedent doubts among the Greenwich team about whether the astrographic lens that yielded the second dataset had functioned properly. Contemporary philosophers have of course become aware that data needs to be reliably and independently arrived at if it is to play a role as evidence (Woodward [2020]), and it is a bit disappointing not to find this literature discussed in the book.

The second critical issue concerns the range of theoretical predictions under test. Dyson and Eddington endorsed a notorious ‘trichotomy’ of possible results, with the Einsteinian prediction for the deflection roughly doubling the Newtonian one, beyond the possibility of a null result. Given this trichotomy, the results from the sets of measurements that were retained clearly supported the Einsteinian prediction on balance. But how is the trichotomy justified? Here, again, Kennefick is rather convincing that Eddington, in particular, was careful and balanced in the judgement of what the theoretical options were—and his intuitions reflected not just Einstein’s views but those of other physicists working on this question at that time. Far from being biased, argues Kennefick, Eddington’s trichotomy gave Newton’s theory its best possible chance.

Nevertheless, ever since Earman and Glymour’s article, the thought persists in the literature that the theory under test was somehow assumed to fix the range of theoretical predictions allowed, the measurement procedure, and the selection and analysis of the data. This thought then gives rise, in Collins and Pinch’s ([1985]) much discussed Golem, to the experimenter’s regress: the notion that there is a circularity at the heart of any confirmation of theory by evidence that can only be overcome by means of antecedent social convention.

There are important implications here for ongoing discussions regarding the nature of scientific data, and how and under what conditions it is turned into evidence for or against any theory. Earman and Glymour rightly argue that data that lacks corroboration independent from the theory that it is meant to support cannot in fact provide evidence for or against it. Collins and Pinch go further with their experimenter’s regress, stating that the only way to turn data into genuine evidence for or against a theory involves antecedent social convention or agreement precisely upon the theory under test. The evidence for a theory is thus in a sense ‘socially constructed’. In contrast to this, Kennefick defends the orthodox view that the 1919 data that supports Einstein’s theory was reliably and faithfully obtained independently of any theoretical considerations. Of course, a theoretical background must be assumed (and certainly Eddington, at least, had a clear and distinct conception of general relativity), but there is no reason why this entails any dangerous circularity. On the contrary, as long at the theory is merely employed as a reliable tool in the analysis of the data—without taking the data to be a faithful representation of its causes—the data yields genuine and legitimate evidence for or against the theory. This conception of theories as tools is not without precedent in the philosophical literature (Suárez and Cartwright [2008]) and, since it is so clearly apposite, it is a pity not to find this discussed in No Shadow of a Doubt.

Kennefick wraps up his elegant defence of orthodoxy with an alternative methodological proposal of ‘experimental progress’ (pp. 315–17), the expectation by practicing scientists that measurement precision will improve with time until experimental evidence gets settled—an idea familiar from Peter Galison ([1987]). Kennefick goes as far as to propose it as a plausible demarcation criterion (between science and non-science). It is a lofty and daring philosophical suggestion, worthy of Dyson, Eddington, or Einstein, and this book provides some independent, reliable evidence for it.

Mauricio Suárez
Complutense University of Madrid
msuarez@filos.ucm.es

References

Collins, H. and Pinch, T. [1985]: The Golem: What Everyone Should Know about Science, Cambridge: Cambridge University Press.

Dyson, F. W., Eddington, A. and Davidson, C. R. [1920]: ‘A Determination of the Deflection of Light by the Sun’s Gravitational Field, from Observations Made at the Total Solar Eclipse of May 29, 1919’, Philosophical Transactions of the Royal Society Series A, 220, pp. 291–330.

Earman, J. and Glymour, C. [1980]: ‘Relativity and Eclipses: The British Eclipse Expeditions of 1919 and Their Predecessors’, Historical Studies in the Physical Sciences, 11, pp. 49–85.

Eddington, A. [1920]: Space, Time, and Gravitation, Cambridge: Cambridge University Press.

Eddington, A. [1940]: ‘Sir Frank Watson Dyson, 1868-1939’, Obituary Notices of Fellows of the Royal Society, 3, pp. 159–72.

Galison, P. [1987]: How Experiments End, Chicago, IL: University of Chicago Press.

Suárez, M. [2019]: Filosofía de la Ciencia: Historia y Práctica, Madrid: Tecnos.

Suárez, M. and Cartwright, N. [2008]: ‘Theories: Tools versus Models’, Studies in History and Philosophy of Science B, 39, pp. 62–81.

Woodward, J. [2000]: ‘Data, Phenomena, and Reliability’, Philosophy of Science, 67, pp. S163–79.