On 6 November 1919, at a joint meeting of the Royal Society and the Royal Astronomical Society in London, the results of two eclipse expeditions were announced. The following morning The Times ran the headline “Revolution in Science”, and within days Albert Einstein, previously known to a few hundred physicists, was the most famous scientist alive.
What had been measured was small: a shift in the apparent position of a few stars by about one two thousandth of a degree. What it demonstrated was that Newton’s account of gravity, which had stood for two hundred and thirty two years, was an approximation.
Three Possible Answers
The question was whether gravity bends light, and by how much. There were three positions.
Zero. If light is a wave in a medium and has no mass, gravity has nothing to act on, and a ray passing the Sun continues straight.
0.87 arcseconds. If light consists of corpuscles with mass, as Newton supposed in the Opticks, then it should fall towards the Sun like any other projectile. Johann Georg von Soldner calculated the deflection on that basis in 1801, using nothing but the mechanics of the Principia, and obtained a value just under nine tenths of an arcsecond for a ray grazing the solar limb.
1.75 arcseconds. Einstein’s general theory of relativity, completed in 1915, treats gravity as curvature of spacetime rather than a force. Light follows the straightest available path through curved geometry, and because the curvature affects both time and space, the total deflection is exactly twice the Newtonian value.
Einstein had originally calculated the smaller figure in 1911, before the theory was finished, and only corrected it to the doubled value in 1915. A German expedition sent to Crimea to test the 1914 eclipse was interned at the outbreak of war and never made the measurement, which was fortunate for him, since they would have been testing a prediction he was about to withdraw.
The factor of two is what makes the experiment decisive. Distinguishing a real effect from no effect is one thing. Distinguishing between two specific non zero predictions, one exactly double the other, leaves very little room for interpretation.
Why 1919
Stars cannot be photographed next to the Sun except during a total eclipse, and not every eclipse is suitable. The eclipse of 29 May 1919 was unusually well placed on two counts.
Totality lasted six minutes and fifty one seconds, near the maximum possible, giving time for repeated exposures. And on that date the Sun sat in front of the Hyades, a rich cluster of bright stars in Taurus. Most eclipses put the Sun against a sparse patch of sky with too few stars close enough to the limb to be useful.
Frank Dyson, the Astronomer Royal, identified the opportunity in 1917 and organised two expeditions to guard against weather: one to the island of Príncipe in the Gulf of Guinea, the other to Sobral in northern Brazil.
Arthur Eddington led the Príncipe party. He was a Quaker and a conscientious objector, and in 1918 was facing possible imprisonment for refusing military service. Dyson argued to the authorities that Eddington’s participation in the eclipse expedition served the national interest, and he was granted exemption on that basis. The expedition that confirmed a German physicist’s theory was partly a device for keeping an English pacifist out of prison.
The Method
The procedure is a comparison. Photograph the star field around the Sun during totality. Then photograph the same star field months earlier or later, when the Sun is elsewhere and the stars can be seen at night. Measure the positions on both plates and look for a systematic outward displacement of the stars nearest the solar disc.
The difficulty is that everything else must be identical. The same telescope, the same plates, the same scale, the same temperature if possible, because a change of a fraction of a millimetre in the focal length imitates exactly the effect being sought.
What Happened at Each Site
Príncipe was clouded and rained heavily on the morning of the eclipse. Eddington exposed sixteen plates, mostly through cloud, and only two contained enough usable star images to measure. He and Cottingham developed them on the island and measured them there, in case the plates were damaged in transit.
Sobral had clear skies and two instruments. The four inch telescope produced excellent plates. The larger astrographic telescope did not: the mirror feeding it appears to have distorted in the heat, and its star images came out blurred and changed in scale.
The three sets of results were:
- Sobral four inch telescope: 1.98 arcseconds, with an uncertainty of about 0.12
- Príncipe: 1.61 arcseconds, with an uncertainty of about 0.30
- Sobral astrographic: 0.86 arcseconds, close to the Newtonian value
Dyson and Eddington discarded the astrographic result on the grounds that the instrument had demonstrably malfunctioned, and reported the other two as confirming Einstein’s 1.75.
The Charge of Bias
That decision has been questioned repeatedly, most influentially in the 1980s, with the suggestion that Eddington threw out the data that disagreed with a theory he already believed.
The objection deserves to be taken seriously and appears not to hold. The astrographic plates show a measurable change of scale between the eclipse and comparison exposures, consistent with a focus shift, which is an instrumental defect identifiable without reference to the answer. A reanalysis of the original Sobral plates carried out at Greenwich in 1979, using modern measuring machines, produced a value of 1.90 arcseconds with a small uncertainty, confirming the general relativistic prediction and supporting the original decision to reject the faulty instrument.
What is fair to say is that the 1919 result was less conclusive than the public reception implied. The Príncipe error bars in particular were large. The confirmation was real, and it was not as clean as the headlines.
Why It Became News
Scientific results rarely reach front pages. This one did, and the timing explains a good deal.
The announcement came six months after the Treaty of Versailles, in a Europe still counting its dead. An English expedition had travelled to two remote locations to test the theory of a German physicist, and had confirmed it. Newspapers presented it as science standing above national hatred, and the story was irresistible.
The physics was also genuinely dramatic. J.J. Thomson, chairing the meeting, described the result as one of the greatest achievements in the history of human thought, and noted that it was not the discovery of an outlying island but of a whole continent of new scientific ideas.
Einstein himself was in Berlin and learned of the confirmation by telegram. Asked later what he would have done if the expedition had contradicted him, he is reported to have said that he would have been sorry for the dear Lord, because the theory is correct. It is a good line and it also reflects something real about how he worked: general relativity was built from principles rather than assembled to fit data, and there was very little in it that could be adjusted.
How Well It Holds Up
Light bending has since been measured with far greater precision, and the measurements no longer require eclipses.
Radio interferometry, tracking quasars as the Sun passes in front of them, gives agreement with general relativity to better than one part in ten thousand. The Cassini spacecraft measured the related time delay of signals passing near the Sun to comparable precision. The Hipparcos and Gaia astrometry missions have to include the deflection routinely, since it affects star positions across the whole sky, not just near the Sun.
The effect also became a tool. Gravitational lensing, in which a massive galaxy bends the light of something behind it into arcs or multiple images, is now standard equipment in cosmology, used to map dark matter and to magnify objects too distant to observe directly. What Eddington measured as a one and a half arcsecond wobble is the same phenomenon that produces the ring of light around a black hole’s shadow.
The everyday consequence is even more mundane. Satellite navigation requires both special and general relativistic corrections, and without them positions would drift by kilometres within a day.
What the Expedition Settled
Two hundred and thirty two years separate the Principia from the Príncipe plates. In that time Newtonian gravity had accounted for the tides, the return of Halley’s comet, the shape of the Earth and the discovery of Neptune, and had failed in exactly one respect that anyone could measure: a small unexplained drift in the orbit of Mercury.
General relativity accounted for Mercury as well, but that was a known discrepancy explained after the fact. The eclipse was different. It was a number, published in advance, that nobody had ever observed, and it was twice what the older theory allowed.
Newtonian gravity remains perfectly adequate for launching spacecraft and predicting eclipses. Since 1919 it has been understood as the limiting case of something deeper, and the evidence for that was a handful of blurred photographic plates taken through cloud on an island off West Africa.
No products in the cart. 