Eddington and the first test of General Relativity

Thanks to Steven Mosher on Twitter, I came across an article that discusses Arthur Eddington’s attempt to test Einstein’s Theory of General Relativity. The basic story is that Newton’s Theory of Universal Gravitation assumed that gravity was a force that acted between two masses and that this force acted instantaneously. In 1915, however, Albert Einstein proposed that rather than gravity being a force that acts instantaneously across distance, what actually happens is that masses curve spacetime and that this then influences the behaviour of all other masses in the universe; gravity is then a manifestation of this curvature of spacetime

Credit: ESA/Hubble & NASA

Credit: ESA/Hubble & NASA

One consequences of General Relativity is that if light passes close to a massive body, it will be deflected. The massive body will essentially act like a lense and, hence, this is often referred to as gravitational lensing, an example of which is shown in the image on the right. The bright orange object in the centre of the image is a massive elliptical galaxy. The blue horseshoe is an image of a much more distant galaxy that has passed almost directly behind the elliptical galaxy and the light from which has been deflected to produce a horseshoe-like image.

When Einstein first proposed his theory of General Relativity it was not easy to test, as the large telescopes we have today didn’t exist at that time. One way to do so, however, was to observe stars close to the limb of the Sun. The light from these stars would pass very close to the Sun and be deflected. Doing this, however, required making observations during a Solar eclipse and then making comparison observations at a different time to see if the light from the these stars did indeed deflect when it passed close to the limb of the Sun. To be clear, if you treat light as a particle, Newtonian gravity would also predict a deflection, but it is smaller than that predicted by General Relativity.

In 1919, two groups went to try and test General Relativity during a solar eclipse. One, including Arthur Eddington, went to Principe off the coast of Africa, and another went to Sobral, in Brazil. The story of these expeditions, and the results of their observations, is what the article I mentioned earlier is about. What makes it interesting is that the claim is that Arthur Eddington was a supporter of General Relativity and essentially massaged the analysis so as to produce a result that was consistent with General Relativity. The suggestion is that if he was being honest he would have presented an inconclusive result, rather than one that was seen as confirming Einstein’s theory.

I had heard something like this before, so seeing this made me think this might be an interesting thing to discuss. How do you judge someone in such a circumstance? It’s a long time ago, and they’ve been proven correct, but they potentially fiddled their results so as to appear to have confirmed what is now clearly one of the greatest scientific breakthrough’s of the 20th century. They were just lucky that their intuition turned out to be correct. However, before doing so I thought I would just look into this a bit more, and came across a paper called Not Only Because of Theory: Dyson, Eddington and the Competing Myths of the 1919 Eclipse Expedition. It argues that

a close examination of the views of the expedition’s organizers, and of their data analysis, suggests that they had good grounds for acting as they did, and that the key people involved, in particular the astronomer Frank Watson Dyson, were not biased in favor of Einstein.

Dyson, the Astronomer Royal at the time, being principal organiser and director of the two expeditions.

The claims against Eddington include that the results from his observations taken in Africa – which were consistent with General Relativity – were biased, and that he unjustifiably argued against some of the results obtained using the observations from Brazil, which were more consistent with the Newtonian prediction than the prediction from General Relativity. It turns out, however, that in 1979 (60 years after the original expeditions) some of the observations were reanalysed, the results of which were published in this paper. All of the observations that were reanalysed produced results consistent with General Relativity, even those that originally were more consistent with the Newtonian predicition, and produced an average that was within one standard deviation of the General Relativity prediction.

So, it seems that maybe the analyses that produced results consistent with General Relativity were not necessarily biased and that there was some justification for discounting some of those that were not consistent with General Relativity. Of course, I don’t know if there are valid criticisms of the 1919 analysis, or not, but I think the big picture issue here is more subtle than simply completely right, versus flawed and wrong. I think it’s worth bearing in mind that any form of cutting edge research is difficult. Researchers may be using methods that are new and not fully tested. They may have to make decisions about the analysis that will require some amount of subjective judgement. It’s particularly difficult when dealing with a primarily observational area, like astronomy, where there may be factors that will be beyond their control, and so even very careful planning doesn’t guarantee that they won’t later encounter unexpected problems.

It’s possible that it may later become clear that some of the judgements were poor, or that the analysis method wasn’t optimal. However, it’s far easier to recognise this in retrospect, than in advance. Science is a process in which we learn both from our mistakes and from our successes, and in which we develop our understanding over time; we don’t regard something as confirmed after a single study, even if it is by some of the leading researchers of the time. Of course there are some things, like outright fraud and plagiarism, that are clear indicators of scientific misconduct; making a judgement that others might later disagree with doesn’t, however, typically qualify.

So, it seems to me that the main message in this story is how messy science can be. It can involve making risky measurements/observations to test new hypotheses. It can involve making somewhat subjective judgements by people who cannot be completely free from bias. It can involve developing new, or modifying old, methods in order to carry out observations, or to analyse what is observed. It’s not perfect, and probably can’t be. However, over time, we can develop an ever increasing confidence in our understanding of something, even if not all steps in the process are perfect. We don’t trust Einstein’s Theory of General Relativity because of what Eddington did in 1919; we trust it because it has continued to pass tests, the most recent of which was the first detection of gravitational waves. That doesn’t mean, however, that the work done in 1919 didn’t make a significant contribution to the overall process.

Update: I originally credited the highlighting of the article to Willard, but it was actually Steven Mosher. Now corrected.

This entry was posted in Research, Science, The philosophy of science, The scientific method and tagged , , , , , , . Bookmark the permalink.

69 Responses to Eddington and the first test of General Relativity

  1. I was in a meeting last week and Mark Krumholz (who wrote this lengthy review paper about star formation) gave a talk in which and he mentioned an interesting story about Eddington and James Jeans, who is famous for coming up with the Jeans Instability. In 1926, Eddington gave the Bakerian Lecture of the Royal Society. James Jeans was not very happy with the lecture and so wrote a letter to Eddington. However, it was not directly to Eddington, it was published in the Observatory. His conclusion includes

    So much work has been done on isothermal equilibrium that it is difficult to understand how Prof. Eddington can harbour the illusion that he is doing pioneer work in unexplored territory, yet his complete absence of reference to other theoreticl workers ….. suggests that such is actually the case.

    Eddington responded and says, about Jeans’ work,

    I had long admired his ingenious paper; but, having come to the conclusion that the masses of the stars are determined by radiation pressure independently of the primitive density, I ceased to accept it.

    Jeans then posts a final letter which concludes with

    May I conclude by assuring Prof. Eddington it would give me a great pleasure if he could remove the long-standing source of friction between us by abstaining in future from making wild attacks on my work which he cannot substantiate, and by making the usual acknowledgements whenever he finds that my previous work is of use to him.

    I have often suggested that scientists prefer to fight in the literature, but had never quite intended it quite this literally. You could also tell the era by the letters starting with Gentlemen and ending with I am, Gentlemen, Yours faithfully,.

  2. I guess something I didn’t address in the post is what one should conclude if someone does indeed obviously take what turns out to be an unacceptable risk in order to present what appears to groundbreaking result. If they turn out to be right, they may get away with it, but recently we had the faster than light neutrino anomaly, and the BICEP2 gravitational wave detection that turned out to be wrong. The former case turned out – IIRC – to be a faulty connecter, and in the latter case they had significantly under-estimated the foreground dust emission, which was mainly because they had taken the data from a heavily smoothed conference paper figure from another group. In both cases, the groundbreaking results were debunked pretty quickly and it was (I think) extremely embarassing for those involved.

  3. Willard says:

  4. Magma says:

    I suggest that there is often a largely innocuous explanation for well-known examples of apparent fudging such as Eddington’s Mercury data or Milliken’s use of an erroneous value for the viscosity of air in his charged oil-drop experiment.

    The preliminary results of an experiment are often used as a basic check of the experimental set-up or calculations. Unexpected or anomalous results, large or small, are usually checked against experimental procedure (did you use the correct voltage, concentration, temperature, units, constants, calibration factors, etc., etc.?)

    Naturally enough, such sanity checks usually end when the expected results are obtained thanks to unglamorous explanations. (“Found what was wrong, professor. The Q-36 explosive space modulator had a blown fuse.”)

    In the vast majority of cases this non-blind ‘test and fix’ approach works fine. In a few others, including some now famous ones, researchers got the ‘right’ answer for the wrong reasons. In the physical sciences this is probably OK, and the cost:benefit ratio is probably quite favorable. And such errors are generally discovered* quite quickly.
    *(Well, maybe. It’s hard to count as-yet undiscovered errors.)

    In other fields, such as health, medicine, psychology, and so on, the risks that researchers’ expectations may bias the results of experiments may be much less acceptable, particularly in costly or hard to replicate studies.

  5. Magma says:

    Corrections: I mixed up Eddington’s 1919 stellar measurements with the precession of Mercury’s orbit. Both early tests of general relativity, but not the same.

    And the oil drop experiment was carried out by Robert Millikan, not Milliken.

  6. I was just looking up the precession of Mercury issue. It seems that they were aware of the discrepancy between its actual precession and what would be predicted from Newtonian physics in the 19th century and it seems that Einstein showed that General Relativity could explain it.

  7. The precision of Eddington’s results were perhaps not as important as the correct prediction of Mercury’s perihelon advance. It was enough to check that light is bent at all. The touchstone during the development and the first test was Mercury.

    Here is an updated history of the genesis of General Relativity: http://www.nature.com/news/history-einstein-was-no-lone-genius-1.18793

    “Several new theories had been proposed in which gravity, like electromagnetism, was represented by a field in the flat space-time of special relativity. A particularly promising one came from the young Finnish physicist Gunnar Nordström. In his Vienna lecture, Einstein compared his own Entwurf theory to Nordström’s theory. Einstein worked on both theories between May and late August 1913, (…)

    In the summer of 1913, Nordström visited Einstein in Zurich. Einstein convinced him that the source of the gravitational field in both their theories should be constructed out of the ‘energy–momentum tensor’ (…)

    The tensor took centre stage in the new relativistic mechanics presented in the first textbook on special relativity, Das Relativitätsprinzip, written by Max Laue in 1911. In 1912, a young Viennese physicist, Friedrich Kottler, generalized Laue’s formalism from flat to curved space-time. Einstein and Grossmann relied on this generalization in their formulation of the Entwurf theory. (…)

    Einstein also worked with Besso that summer to investigate whether the Entwurf theory could account for the missing 43˝ per century for Mercury’s perihelion. Unfortunately, they found that it could only explain 18˝. Nordström’s theory, Besso checked later, gave 7˝ in the wrong direction. (…)

    In his lecture in Vienna in September 1913, Einstein concluded his comparison of the two theories with a call for experiment to decide. The Entwurf theory predicts that gravity bends light, whereas Nordström’s does not. (…)

    (…) Worried that Hilbert might beat him to the punch, Einstein rushed new equations into print in early November 1915, modifying them the following week and again two weeks later in subsequent papers submitted to the Prussian Academy. The field equations were generally covariant at last.

    “Congratulations on conquering the perihelion motion,” Hilbert wrote to him on 19 November. “If I could calculate as fast as you can,” he quipped, “the hydrogen atom would have to bring a note from home to be excused for not radiating.”

    Einstein kept quiet on why he had been able to do the calculations so fast. They were minor variations on the ones he had done with Besso in 1913. (…)

  8. Florifulgurator,
    Thanks, I think this is very interesting, and almost always the case,

    Lesser-known and junior colleagues helped the great physicist to piece together his general theory of relativity

    Also, as that article indicates, Einstein had already shown that GR could explain the descrepancy in Mercury’s precession, so the 1919 expedition might have been the first independent attempt to test GR, but wasn’t really the first test that it had passed.

  9. This, however, isn’t quite correct

    It was enough to check that light is bent at all.

    Newtonian physics also predicts a deflection if you treat light as a particle. The deflection is, however, smaller than due to GR.

  10. Steven Mosher says:

    ‘So, it seems to me that the main message in this story is how messy science can be. “

  11. David Appell says:

    …and Then There’s Physics says:
    “Newtonian physics also predicts a deflection if you treat light as a particle. The deflection is, however, smaller than due to GR.”

    Yes, the Newtonian prediction is 1/2 Einstein’s.

  12. David Appell says:

    I once saw a picture of a conference slide with this quote:

    “Experimentalists will be surprised to learn that we will not accept any evidence that is not confirmed by theory.” — Arthur Eddington

  13. “Experimentalists will be surprised to learn that we will not accept any evidence that is not confirmed by theory.” — Arthur Eddington

    That is good.

  14. guthrie says:

    Another important point from your post is that in history of science, as in climatology, it’s a good idea to go back to both original sources and proper professional publications about the topic, rather than, as unfortunately so many people do, rely on random stuff on the internet or what they learnt at school 30 years ago.

  15. anoilman says:

    Steven Mosher says: January 20, 2017 at 10:41 pm
    ‘So, it seems to me that the main message in this story is how messy science can be. “

    Yeah… The answer fairy is terribly inconvenient that way.

  16. Bernard J. says:

    It’s probably worth giving Gregor Mendel an honorable mention for the controversy that raged for a while over the results of his experiments with flower colour inheritance in peas:


  17. Bernard J. says:

    That was “flower colour (and other characteristics) inheritance in peas”… I was going to reword it to be less clumsy and inadvertently clicked send before finishing the edit.

  18. angech says:

    Hm, he threw out inconvenient data, used limited amounts of poor quality data with gigantic standard deviation errors and steamrolled through his “findings”.
    And because the theory was right this excuses him?
    I’m sorry, I thought we were better than this?

  19. Marco says:

    Hmmm, that article simply cannot just stand on its own:

  20. angech,

    And because the theory was right this excuses him?
    I’m sorry, I thought we were better than this?

    You didn’t read it, did you?

    I don’t think you did either 😉

  21. izen says:

    ” Hm, he threw out inconvenient data, used limited amounts of poor quality data … And because the theory was right this excuses him?
    I’m sorry, I thought we were better than this?”

    No, that is exactly how good we are, it is a feature not a flaw. Mendel had obviously figured out what was going on genetically long before he did the experiment. What he wrote up was a demonstration of the underlying theory, not raw random observations from which he derived the theory.

    As with the Eddington observations it was not a ‘proof’ of the theory, but a demonstration that observations were for the theory rather than against it.

    “About thirty years ago there was much talk that geologists ought only to observe and not theorise; and I well remember some one saying that at this rate a man might as well go into a gravel-pit and count the pebbles and describe the colours. How odd it is that anyone should not see that all observation must be for or against some view if it is to be of any service! ”

  22. Marco says:

    Nope, didn’t read all of it. TL;DR when you’re battling a very, very nasty cold.

  23. Marco,
    To be fair, I did realise it was rather long. Hope you feel better soon.

  24. verytallguy says:

    the Newtonian prediction is 1/2 Einstein’s

    Can anyone point me to a simple explanation of this?

    I can cope with equations, but not tensor calculus.

  25. vtg,
    Technically you just assume a massless test particle approaching the Sun along a path that would just graze the surface. Such a particle will feel an acceleration of GM_{Sun}/r^2, and so you can simply integrate along the path and it will deflect slightly. There is a simple calculation to do this, but I can’t seem to find it at the moment.

  26. verytallguy says:


    that’s the solution for the Newtonian case, right?

    What’s the equivalent for General Relativity?

  27. Yes, thay is the Newtonian solution. The solution for the Newtonian is

    \dfrac{2GM_{Sun}}{R_{Sun} c^2}.

    The GR solution is

    \dfrac{4G M_{Sun}}{R_{Sun} c^2}.

    There’s a paper that works through it, and I have nice simple Newtonian calculation that gives me a 2GM_{Sun}{R c^2}, but I’m just trying to convince myself that that is the deflection angle.

  28. verytallguy says:

    Thanks AT.

    Is there an equivalent for the relativity calculation you can point at?

  29. JCH says:

    However, in recent decades it has been increasingly often alleged that the data-analysis of the expedition’s leaders was faulty and biased in favor of Einstein’s theory. Arthur Stanley Eddington is particularly alleged to have been prejudiced in favor of general relativity. Specifically it is claimed that some of the data, which would have favored the so-called Newtonian prediction, was thrown out on dubious grounds. …

    There is a subset of humanity who simply cannot negotiate this sort of scenario without falling head over heels into conspiracy rot.

  30. vtg,
    This post seems to do a similar calculation for GR.

  31. angech says:

    You didn’t read it, did you?
    Of course I did, hence my points.

    he threw out inconvenient data
    “Having discarded 18 plates on very specious grounds” Page 58 of “Right for the wrong reasons”

    used limited amounts of poor quality data
    “He routinely referred to only two sets of prints”
    “and his own very poor two photographs”
    “each set of photographs involved very large standard errors” page 56

    “and steamrolled through his “findings”.
    “With the three line whip imposed by a seeming holy alliance”
    “battling against what can fairly be called a cultural consensus” Page 62

    Being right for the wrong reasons is never being right at all.
    You said ” they potentially fiddled their results” but in reading the article the word should be actually.
    Marco , post fact assessments claiming that with the benefit of hindsight they might have got it right after all does not wash.
    With the information they had and the techniques they could use at the time they were wrong.
    And they knew they were wrong but persisted because the theory seemed right.
    The third guy did not disagree with Eddington so he got it through.

  32. Except the second article argues that his analysis was not biased and that he did not simply ignore inconvenient data. There is also a paper that redoes the analysis and is consistent with that argument.

  33. Marco says:

    Angech, despite having a very nasty cold, consequently a headache, and thus hardly on my best form, I have absolutely no problem seeing that the other article I cited (and ATTP, too) strongly suggests the first one isn’t quite right. You solely base your viewpoint on the first article, and ignore the second. You may want to read the second article with an open mind, and then compare with the first. Be sure to make notes about the discrepancies.

  34. izen says:

    The Waller article on Eddington is part of a book (Einstein’s Luck: The Truth Behind Some of the Greatest Scientific Discoveries.) that constructs a narrative of crucial experiments that confirmed a theory, – but the scientist actually ‘fiddled’ or at least cherry-picked the results.

    The rather more subtle reality that theories are sometimes tested by observational measures that hindsight shows were not as good as those using them thought. Pasteur did not conclusively refute the spontaneous generation of life, in retrospect we can see the experiments were not sufficient to establish that. But that sort of Whig history ignores the Bayesian impact of the observations had at the time on the perception of the credibility of a theory.

    That subsequent work both validates the original conclusions, and shows that the evidence at the time was just a part of a larger consilience of evidence is abandoned in favour of the narrative of the science misled by the confirmation bias of one special individual, not the outcome of a communal effort.

    The assertions angtech highlights from the article -“and steamrolled through his “findings”.
    “With the three line whip imposed by a seeming holy alliance”
    seem to be made with little supporting evidence and are biased in favour of the conspiratorial narrative the book is promoting.
    The fact that analysis of the Eddington results in the peer reviewed literature fails to confirm the Waller narrative of intentional distortion of data, or intentional suppression of uncertainties makes me suspect that it was adopted as a more commercially attractive perspective than a more accurate assessment of how messy science can be.

  35. But that sort of Whig history ignores the Bayesian impact of the observations had at the time on the perception of the credibility of a theory.

    I thought the same. If you have a theory that has alread potentially solved a long-standing problem (the precession of Mercury, for example) then you might have a prior that gives more weight to observations that support the theory than those that do not, especially if you are aware of issues with the observations.

  36. Andy Skuce says:

    Thanks for this piece. I wasn’t previously aware of the story. I wrote a blog article a few months ago about my own limited encounters with the philosophy of science. I mentioned that Paul Feyerabend had been instructive in his account of Galileo’s efforts to promote his science. This involved some fudging and ad-hockery on GG’s part that is not usually detailed in histories of science. Because data are messy and because scientists are people, the growth of knowledge does not always follow the smooth trajectory that you might expect if you believed the Popperian just-so stories.

    Of course, this doesn’t mean we have to go the whole hog (as Feyerabend does, or at least argues for) and consider science to be an anarchic activity indistinguishable from voodoo. But I’m sure that most working scientists are very well aware of the chaos in the research process.

  37. Anytime you’re dealing with data collected at the edge of your technical or methodological capabilities there is a high probability that you will have conflicting results. The uncertainties are often larger than the signal you’re looking for.

    I have collected data and after analysis just figuratively thrown it all away in disgust because it isn’t telling a consistent story much less the one sought after. This is pretty much what science is all about – you stop and ask yourself questions; is my expected story wrong? am I making some mistake in method? is there an error source I haven’t considered? is there a variable I’m not controlling properly? am I doing the math right (program or spreadsheet errors)?

    Every now and then, despite your best efforts (and repeated testing) you get results *consistently* opposite what you expect. Now you’re in a quandary: How can this be right? The CERN FTL neutrinos results fall into this category for me. My recollection was that even they didn’t fully believe the results, but what could they do? Ignore the results? No, they published and basically said, tell us what we did wrong. It’s a perfect case of science at work.

  38. David Appell says:

    oneillsinwisconsin says:
    “The CERN FTL neutrinos results fall into this category for me. My recollection was that even they didn’t fully believe the results, but what could they do? Ignore the results? No, they published and basically said, tell us what we did wrong. It’s a perfect case of science at work.”

    Perhaps — but perhaps they also rushed out their result for the sake of making scientific history.

    In the end their result was due to faulty equipment. So there wasn’t much anyone else could do to check that.

  39. David,
    Indeed, the FTL neutrinos was interesting because they presented it acknowledging that there may still be something wrong. However, given that they were probably the only ones who could check, one might argue that they should just have checked again.

  40. David Appell says:

    ATTP: Fair enough. But maybe a bit of trying to have it both ways, credit if it were true, an out if there turned out to be a mistake.

    It’s interesting how many instant theory papers it spawned trying to explain it. It makes me think of all those climate papers written 2005-2015 explaining the “pause.” What to make of them now that, post Karl et al 2015, better data show the pause never existed?

  41. Bernard J. says:

    As an aside, does anyone know if there’s a recontruction of the appearance of the particular lensed galaxy in the photo above?

  42. David,
    Actually, I thought I was mostly agreeing with you 🙂

  43. David Appell says:

    [Duh, it’s still early here in Oregon!]

  44. Bernard J. says:

    What to make of them now that, post Karl et al 2015, better data show the pause never existed?

    I guess one could consider it in two broad ways: did the authors attempt to approach it from the perspective of trying to determine/refine physical explanations for currently unquantified climatic components leading to the short-term fluctuations, or were they simply speculating at a broader level that heat accumulation really had somehow paused, based on the authors’ flawed appreciation of ‘noise’ superimposing over a signal? Papers in the vein of the former alternative may well have been credible forays to further refine the understanding of heat movement around the globe; papers of the second type are just likely just faff that should have been rejected by reviewers for the inadequate acknowledgement of fundamental statistical principles.

  45. Bernard J. says:

    “…just likely just…”?!

    It’s even earlier here – bordering on dawn…

  46. Bernard,
    I think this paper is doing something like that, but I haven’t had a chance to read it in any detail.

  47. Steven Mosher says:

    “But I’m sure that most working scientists are very well aware of the chaos in the research process.”

    That is why I find skeptics pointing to Old Feynman “lectures” to be particularly funny. Feynman’s description of how scientists work is utterly idealized , even his own career is littered with examples of keeping theories when the experimental data indicated otherwise.

  48. Steven Mosher says:

    ” You solely base your viewpoint on the first article, and ignore the second. ”

    This is angech’s version of tossing out bad data.

  49. In the German media it was very clear at the time that the researchers presenting the faster than light neutrinos expected their results to be wrong. The way I understood it, they were forced to hold a press conference because the news was spreading in the blogs. Other groups could have helped: they could have repeated the experiment with their own instruments. Like we all know replication is much stronger science than stupid auditing.

    Most of the “hiatus” papers were okay and interesting for our understanding of natural variability. In most cases only the framing as a “hiatus” paper, to make it more “relevant”, was wrong.

  50. In most cases only the framing as a “hiatus” paper, to make it more “relevant”, was wrong.

    Maybe, but this just seems to be an illustration of the messiness of science. You’re writing a paper. You know there’s a framing that you can use that will get your paper noticed. You also know the framing isn’t ideal, but it’s not wrong. Do you use it, or do you write something that is more careful, but means your paper won’t get much attention. Sometimes the answer is obvious (do the latter) but sometimes it is not.

  51. Willard says:

    > [T]hat article simply cannot just stand on its own: […]

    No article can. Here’s Earman & Glymour 1980:

    Click to access Earman_Glymour_1980.pdf

    Kennefick kinda forgot to quote Earman & Glymour’s alleged “thesis that Eddington’s personal bias played a significant role in the data analysis.” So here’s the paragraph where that “thesis” is introduced:

    [W]hile some aspects of Eddington’s handling of the deflection of light were in the finest tradition of science, others were not. As he confessed in Space, time and gravitation, he was “not altogether unbiased.”70 The bias showed in his treatment of the evidence: he repeatedly posed a false trichotomy for the deflection results, claimed the superiority of the qualitatively inferior Principe data, and suppressed reference to the negative Sobral results. (His discussion of the red-shift was sometimes no better; in Space, time and gravitation, for example, he held that observed shifts different from those Einstein predicted did not tell against the general theory as long as they were not absolutely nil.) Dyson was more judicious; he refrained from adducing the Principe results as evidence for general relativity and called for measurements of the deflection at the next two eclipses. But Dyson, too, joined in advocating the trichotomy and certainly did not emphasize the discrepancy between the observations and Einstein’s prediction.

    Instead of attacking Earman & Glymour’s argument directly, Kennefick rather attacks strawmen built on Earman & Glymour’s argument by (a) showing an example of such strawmen in the reviews of Waller’s book, and by (b) reciting some testimonies by Eddington, which leads him to conclude that:

    There remain many issues raised by Earman and Glymour and others which I will not address, such as the role which Eddington’s personal and scientific views did play, (especially in his own reduction of the Principe data) and the matter of Eddington’s and Dyson’s presentation of their results to the public. But one may safely conclude, on the basis of the documentary evidence, that there are no grounds to believe that the most critical decision of the eclipse data reduction was taken from a biased standpoint. Whether that decision was right or wrong, and whether it had an unwarranted influence in the reception of one of the leading physical theories of the 20th century, are of course quite separate questions, but I will offer some arguments, in the next section suggesting that the decision in favor of Einstein over Newtonian gravity was justified.

    This refutes neither Earman & Glymour nor Waller’s thesis, which is based on the “treatment of the evidence” more than anything else.

    Kennefick is more into the it’s not science, but it’s important kind of mood.

  52. angech says:

    “It’s not Science, but It’s Important”
    “Then in the bathroom brushing our teeth That’s all part of the foreplay, I love foreplay
    Then you go sort out the recycling That isn’t part of the foreplay,
    But it’s still very important”
    Flight Of The Conchords – Business Time Lyrics.

    “The assertions angtech highlights from the article seem to be made with little supporting evidence and are biased in favour of the conspiratorial narrative the book is promoting.”
    The book highlights 5 reasons why Eddington’s conclusions were palpably false.
    I guess conspiracy theories have been around for a long time, it is funny how they always crop up when describing a viewpoint one is uncomfortable with.

    ” You solely base your viewpoint on the first article, and ignore the second. ”
    The book highlights 5 reasons why Eddington’s conclusions were palpably false

  53. David B. Benson says:

    Bernard J. — Well, the lensed galaxy is galaxy shaped but I doubt the ability to reconstruct a more precise shape.

  54. angech says:

    Marco says:
    “Angech, I have absolutely no problem seeing that the other article I cited (and ATTP, too) strongly suggests the first one isn’t quite right.”

    We all agree.

    “You solely base your viewpoint on the first article, and ignore the second.”
    Whereas you do not ” solely base your viewpoint on the second article, and ignore the first.”?

    “You may want to read the second article with an open mind, and then compare with the first. Be sure to make notes about the discrepancies..”
    We all could and should. With an open mind one reads of an apologist trying to support a failing argument.
    He had 12 astrolabe plates showing Newton’s result right which he threw out in favour of 1 and a half plates taken in dodgy cloudy conditions which gave the “right” wrong answer [it was too high], which lacked enough coordinates, which had to be compared to plates taken 6 months later in England from a different time, place, position and hemisphere. Confirmation bias, cherry picking, you choose the terms.
    The second paper admitted the data was dodgy, tried to use the astrolabe plates that were thrown out. excluded 3 of them so the results are not at all comparable, used one that had a figure so different to the others that BEST would have thrown it out, hand entered estimates of the data as the fields were not spherical enough for the computers they wished to use then used those computers on the human amended data [is that called a redux?].
    And still could not fudge a figure within cooee [Aaustralianism, means close] of the expected figure though it did now negate Newton.
    The term making a silk purse out of a sow’s ear comes to mind.
    Eddington is to be greatly admired, his work here was slipshod and should be acknowledged so.

  55. I bow to your authority in matters of slipshodness, Doc.

    Your “With an open mind one reads” looks like SpeedoScience.

  56. izen says:

    “Eddington is to be greatly admired, his work here was slipshod and should be acknowledged so.”

    As the Kennefick article shows it was Dyson who analysed the plates that showed results closer to Newton than Einstein.
    The error correction applied to those plates however was known to be dubious, and using reasonable other assumptions Dyson informed Eddington that they gave results closer to GR.

    It might be interesting to reflect on why a narrative of scientists being at least slipshod, or at worst intentionally deceitful, has gained traction since sociology started doing ‘History of science’ (Feyerabend?). And why you seem to prefer that narrative.
    To most experimentalists the way in which Pasteur, Millikan, Eddington/Dyson dealt with the messy reality of uncertain data looks reasonable. The accusation from the sociologists that hard science looks for, or constructs data to support its pre-existing biases may be projection.

    At the risk of invoking Willard, while some of this fashion for bashing scientists may be due to marketing and the sociology – pigeon equivalence, it is also rooted in the epistemology of causation and the dubious duality of Humean supervenience.

  57. angech says:

    Climate Change New Year resolutions yet to kick in fully.
    One personal going well since before the New Year.
    Second is to “be nice”, not succeeding very well but trying and will improve. Speedos off.
    izen is correct, “As the Kennefick article shows it was Dyson who analyzed the plates that showed results closer to Newton than Einstein.”
    This is a lovely trail to follow with multiple sites, investigators and instruments and very long articles.
    To the best of my understanding Dyson analyzed [supervised] both the plates from Sobral from the 4 inch telescope that showed results closer to Einstein than Newton and also the astrolabe plates from Sobral that showed results closer to Newton than Einstein.
    “This instrument [the Astrographic] supports the Newtonian shift, the element of which is 0.87”
    and promptly refused to use them at all.
    Dyson et al discarded these results. They did reanalyze them trying to work out the discrepancy and with a lot of assumptions did get them up closer to Einstein than Newton but never used them, the assumptions were too debatable.

  58. Dan Riley says:

    It was readily apparent from the OPERA preprint that they hadn’t done all the crosschecks they should have to validate the result. Finally checking their clock against one of the other Gran Sasso experiments (using cosmic muons that went through both detectors) eventually helped identify the timing problem.

    Other groups did check the results too, but in most cases it took a few months for them to reconfigure their apparatus (most of the neutrino beam experiments weren’t setup for absolute velocity measurements, clock sync for that is very complicated) and get more beam time from CERN. By the time OPERA found their cable problem, there were 5 or 6 other groups with results out that disagreed with the OPERA FTL result.

    So both replication and (meaningful) auditing worked. The problem with the climate “auditors” is the lack of perspective making mountains out of molehills while (intentionally) ignoring meaningful validity checks.

  59. verytallguy says:

    So, here’s the thing.

    I would expect in the relativity case that for large R ( or small M) the equations would collapse to give the same result as for Newton. Hence Newtonian gravity is a good approximation

    That doesn’t seem (at least on my fairly cursory examination) to be the case.

    Have I misunderstood something?

    (Feel free to tell me to go away and work through the maths properly)

  60. Dan Riley says:

    In the limit, the potential goes like 2(1+\gamma)M/R, so you need low velocity (or more technically low Lorentz factor) for the test particle to recover the Newtonian limit, hence the limits for massless test particles are different. Misner, Thorne and Wheeler has a discussion of this starting at page 1099, using the PPN formalism so no tensor calculus needed.

  61. verytallguy says:


    Thank you, that makes sense.

  62. I’m not particularlty expert at GR, but here is a description that is similar to what Dan has said

    The most interesting constant is the 4 for general relativity, which is twice the Newtonian value because light moves at the speed of light. The extra bending is a consequence of Einstein’s theory of special relativity putting space and time on the same level. The theory of general relativity then formulates gravity in terms of the curvature of spacetime. Newton’s theory is the limit of general relativity that considers only time curvature; general relativity itself also calculates the space curvature. Since most objects move much slower than the speed of light, meaning that they travel much farther in time than in space, they feel mostly the time curvature. The Newtonian analysis is fine for those objects. Since light moves at the speed of light, it sees equal amounts of space and time curvature, so it bends twice as far as the Newtonian theory would predict.

  63. ngongocduchuy says:

    Really? So, we will see a bright light from some massive galaxy behind, rather than a black hole? How mind-blowing!

  64. ngongocduchuy,
    What do you mean by “rather than a black hole”? It’s certainly the case that if a very distant galaxy passes behind a closer galaxy, the closer galaxy (or sometimes cluster of closer galaxies) can act as a gravitational lens and can focus the light from the more distant galaxy, produce both distorted images, but also amplifying the light from those more distant galaxies. It is one way that we are able to study the earliest stages of galaxy formation.

    On a similar note, it is also used to detect planets around other stars in our galaxy.

  65. verytallguy says:

    Thanks again AT.

    My understanding:

    The referenced equations for GE incorporate the light speed velocity of photons. The GR solution for a particle moving at lower velocity would approximate to Newtonian mechanics resulting in halving of the deflection angle.

  66. vtg,
    Yes, that’s my understanding too (although I haven’t actually worked through the calculation).

  67. Willard says:

    > It might be interesting to reflect on why a narrative of scientists being at least slipshod, or at worst intentionally deceitful, has gained traction since sociology started doing ‘History of science’ (Feyerabend?).

    Speaking of whom:

  68. I read about this in detail a decade or two ago, describing the efforts of the 1962 expedition to reproduce, and offering a critical eye on Eddington’s findings, using statistical arguments. Having been drawn into science with a keen interest in Astronomy as a lad, and having been taught a bit of experimental and lab technique by my late dad who was a professor of Chemistry, I found the story interesting, and was initially disappointed in Eddington. I thought about the business for a few years, and then concluded I was too harsh on him, and, while I cannot comment on the matter in any professional way, not having examined the plates or being familiar with the details of the process, I think the methods revisionists might be missing something.

    My dad taught me to take polarimetry measurements when I was young. This was done on a mechanical device, peering through a pair of eyepieces, and turning a dial on a vertical axis until both sides of the orange circle before me were the right tone. A side which was mismatched was darker. I did a bunch for him, over a couple of summers, and learned about recording data, and redoing measurements independently. One thing he taught me was how when turning a mechanical dial, you always wanted to approach the point of balance from the same direction. If you overshot, you had to go back and approach from that direction again. I saw this in my dad’s techniques elsewhere in the lab, too, interacting with apparatus which is no doubt today computerized, but at the time was mechanical, electromechanical, and sometimes simply stuff like fluid levels, paying attention to the meniscus, and so on.

    A lot of these adjustments of technique do not yield results which are Gaussian distributed. Sometimes the adjustments go badly, and there are outliers introduced because of the peculiar manipulations, and sometimes there are multiple modes. There was a learning process with each kind of instrument and even each copy of an instrument, and when I started a polarimetry run, the first 100 or so were known and blinded blanks, purely for calibration. But all this made me wonder about Eddington and company, and whether or not a decades later revisionist, bringing their sensibilities of modern methods and statistics, only armed with Eddington and companies’ published papers, and plates, really could reproduced what was going on.

    So, for polarimetry, of course everything is recorded, but this is done in a lab data notebook. Clearly not all that survives to publication. It has mistakes recorded, things like flipping digits in tube identifiers, and the like … All the usual stuff of experimental data collection. But I wonder about Eddington’s notebooks, and whether or not he spent hours and hours ahead of data collection getting familiar with his instrument. If he did, it is plausible that someone, given the report of final results, might say Eddington or his other team could not possibly reproduce results as precisely as they claimed. On the other hand, when using such basic instruments, and getting familiar with them, it is possible that Eddington’s teams had greater confidence in their instruments and their ability to measure than a much later observer would, simply based upon this kind of experience. This is a change in world view, and of instrument.

    I recall how, during the Apollo program, engineers would work late into the night whenever the mockups of equipment were not being used for training, experimenting with it, trying to learn what it could do beyond the technical spec limits, and what biases the equipment had. Instruments today are much more complicated, and it’s difficult to see that being done, especially when the instruments are expensive and highly shared. But I wonder about this with Eddington, and I wonder if we are not shortchanging his craft.

  69. Pingback: Overhype much? | …and Then There's Physics

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.