The Eclipse Data From 1919: The Greatest Hoax in 20th Century Science
| Scientific Paper | |
|---|---|
| Title | The Eclipse Data From 1919: The Greatest Hoax in 20th Century Science |
| Read in full | Link to paper |
| Author(s) | Richard Moody |
| Keywords | relativity, eclipse, einstein |
| Published | 2009 |
| No. of pages | 26 |
Read the full paper here
Abstract
Prior to 1919, general relativity was an obscure theory by a rising star in physics, Albert Einstein. Based on the perceived need to test this complex and intriguing concept, it was held as gospel that the sunlight passing by the sun should be bent by the gravitational attraction of the sun, something known to Sir Isaac Newton and modified by Einstein. According to prevailing wisdom, this should be observable during a total solar eclipse when the shielding of the sun's light permitted the observation of light from distant stars being "bent" around the sun. In an effort to play the role of peacemaker and kingmaker, Arthur Eddington traveled to Principe in Africa with the express purpose of proving Einstein right. Prior to that, he was an advocate for Einstein, due, in part, to the fact that both men shared the same political beliefs, Pacifism. In his zeal to be both peacemaker and kingmaker (Eddington wanted to be known as the man who discovered Einstein), Eddington engaged in corruption and derogation of the scientific data, the scientific method, and much of the scientific community. To this day, this completely manufactured data set is quoted by two prominent scientists and the organs of publication. It surpasses the Piltdown Fraud as the greatest hoax of 20th and 21st Century science.
Overview
Richard Moody Jr., a geologist by training, argues that the light-deflection measurements made during the total solar eclipse of 29 May 1919 — the observations that made Albert Einstein world-famous overnight — could not possibly have carried the precision claimed for them, and that Arthur Eddington obtained the "confirmation" of General Relativity by selecting the plates that agreed with the prediction and discarding those that did not. The paper's thesis is not that Eddington made an honest error of judgement but that the episode was, in Moody's word, a "hoax": a manufactured data set that has been repeated as settled fact for ninety years and that, on his reckoning, "surpasses the Piltdown Fraud as the greatest hoax of 20th and 21st Century science."
The paper is polemical by design. Moody announces in a preface that it is "predicated on the principle of overwhelming scientific superiority" — he piles up sources rather than building a single narrow technical case. Its departure from the mainstream account is therefore twofold. Technically, it denies that the 1919 expeditions had the instrumental or atmospheric resolution to separate the Newtonian deflection from the Einsteinian one. Sociologically, it treats the eclipse announcement not as a triumph of prediction but, borrowing Ian McCausland's phrase, as "an obstacle to objective consideration of alternatives" — the moment at which one theory of gravity acquired institutional protection from competitors.
The argument
Strong models and the corruption of data
Moody frames the episode with a general thesis about what he calls, after Graf, "strong models": dominant theories that act "like crude filters, readily admitting data consistent with the theory and systematically rejecting data inconsistent with the theory," producing a feedback loop in which corrupted data reinforce the model that selected them. He likens a strong model to the queen bee that kills off potential rivals on emerging. On this view, the eclipse data and general relativity have propped each other up for a century, and where "wealth, power and prestige come into play" the model becomes "a club to beat back promising alternatives."
The expedition was compromised before it began
The paper stresses that Eddington was not a disinterested tester. He and Einstein were both pacifists; Eddington's wartime deferment carried, on the account Moody quotes from Paul Marmet, "the express stipulation that if the war should end by May 1919, then Eddington should undertake to lead an expedition for the purpose of verifying Einstein's predictions" — verifying, Moody notes, not testing. He also quotes Chandrasekhar's remark that "had he been left to himself, he would not have planned the expeditions since he was fully convinced of the truth of the general theory of relativity," and Eddington's own admission that although the material "was very meagre compared with what we had hoped for, the writer (who it must be admitted was not altogether unbiased) believed it convincing."
The precision claim
The technical core of the paper is an argument about significant figures. Eddington's claimed reading precision was 1/100th of an arc second — corresponding, on McCausland's estimate, to about 1/100th of a millimetre at the plate edge. Against this Moody sets the standard limits of ground-based astronomy: a ten-inch telescope with perfect optics resolves about 1 arc second, and atmospheric "seeing" typically limits practical resolution to 0.5–2 arc seconds, "more typical at ordinary locations is 2–3 arc second seeing, or worse." Conditions at Principe were far from ordinary: the temperature reached 97 °F on eclipse day after 75 °F the night before, outside the instruments' working range, and the focal length of one telescope shifted so that a backup had to be used. The sudden cut-off of insolation as the Moon covered the Sun, he argues, set ground and vegetation radiating and stirred the air, making the star images dance on the plates. He reproduces the five error sources listed by Bertotti, Brill and Krotkov: coronal and atmospheric refraction, thermal distortion of the optics during the eclipse, scale changes between eclipse and comparison plates, distortion of the emulsion while drying, and measurement error on the images. His conclusion is that the allowable precision was of the order of arc minutes, not hundredths of an arc second — an overstatement by three or four orders of magnitude.
Selection of the data
The most concrete charge is drawn from Charles Lane Poor: "In the South American eclipse of 1919, less than 15% of the actual measured data was used in obtaining the announced result… All non-radial components of the actual measures were discarded as 'accidental errors'." Moody's comment is the paper's sharpest line of reasoning: if you discard every measurement that differs from the prediction, what remains will agree with the prediction by construction. He adds Poor's further complaints that the raw displacements "do not agree in direction, in size, or in the rate of decrease with distance from the sun," that the residuals look more like a refraction effect of the Earth's atmosphere (a combination of the Courvoisier effect, daylight refraction and eclipse-shadow temperature effects), and that Trumpler's reductions assume the very law under test. For the 1922 eclipse he cites a plot of 92 stars whose shifts run "in all directions, many of them going the wrong way by as large a deflection as those shifted in the predicted direction."
The popularisers
A long middle section audits how the episode is retailed to the public: Scientific American defending the accuracy of the plates, Time and the "Person of the Century" issue, the Skeptical Inquirer issue morphing Einstein into a haloed figure, the Mensa Bulletin declining to print Moody's rebuttal, and the astronomer David Levy's television account in which a single photograph "proved Einstein was correct." He reserves particular attention for the inconsistency between Stephen Hawking's A Brief History of Time, which calls the 1919 measurement "sheer luck, or a case of knowing the result they wanted to get," and Hawking's later Time article calling it a confirmation "in spectacular fashion." Moody also restates, from his earlier work, the claim that Einstein neither originated nor derived E = mc2, crediting the mass–energy relation m = E/c2 to Henri Poincaré.
Assessment
What is genuinely strong in the paper is its central arithmetical point, and it is one that historians of science have taken seriously independently of any view about relativity. The Sobral astrographic plates were degraded by focus change, the Principe plates were cloud-affected and few in number, and the published result did rest on a choice among three data sets — a choice made, as the surviving correspondence shows, by people who expected a particular answer. Poor's observation that the announced value emerged after most of the measured material had been set aside is a matter of record and not an invention of the author. Moody is also right that the claimed formal errors were small compared with plausible systematic errors, and his insistence that stated precision must be justified against instrument, atmosphere and emulsion is ordinary good practice rather than a fringe demand.
The difficulties are equally clear. First, the paper is an argument from testimony rather than from analysis: almost every technical point is a quotation from Marmet, Poor, McCausland or a seeing-limit textbook, and Moody nowhere carries out his own reduction of the plates or an explicit error budget with numbers. His central figure — that the allowable precision was "probably on the order of several arc minutes" — is asserted, not derived, and it is implausible on its face, since differential astrometry between an eclipse plate and a comparison plate of the same field is not limited by single-image seeing in the way the quoted 1–2 arc second figures describe; averaging over many star images beats the seeing limit, which is precisely why the technique was attempted at all. Second, the paper does not engage the subsequent measurement record, and this is its most serious omission. If the 1919 result were a fabrication, the deflection would still have to be explained away in the radio-interferometric measurements of quasar positions near the Sun, which since the 1970s have confirmed the light-bending coefficient to better than a part in a thousand without photographic plates, without emulsion shrinkage and without an eclipse, and in the Shapiro delay measured by the Cassini spacecraft. A demonstration that the 1919 evidence was worthless would establish that general relativity was accepted prematurely; it would not establish that its deflection prediction is wrong. Third, the sociological material — hot fusion budgets, Time magazine circulation, Hebrew University merchandising, speculation about Hawking's motives, the digression on Mileva Marić — occupies a large fraction of the text and weakens rather than supports the technical case. Imputed motive is not evidence about a photographic plate.
Read as history and as a critique of how a result becomes canonical, the paper makes a real and defensible point about 1919 and about the hazards of testing a theory one is convinced of. Read as a refutation of gravitational light bending, it does not reach that conclusion, because it never confronts the modern measurements that would have to be wrong for the conclusion to follow.