The 1919 Eclipse Expedition: The Experiment That Made Einstein Famous
Photo: cottonbro studio / Pexels
On the morning of May 29, 1919, the moon's shadow swept across the Atlantic. For six minutes and fifty-one seconds over the island of Príncipe, off the West African coast, and a slightly shorter window over Sobral in northern Brazil, the sun went dark. Two teams of British astronomers had positioned themselves at these locations months in advance, carrying heavy glass photographic plates and precision telescopes across wartime-disrupted supply chains, with a single objective: to photograph a cluster of stars that would normally be invisible near the sun's disk, and to measure whether their apparent positions had shifted from where they sat in the nighttime sky. That shift — if it existed at all — would test a prediction made by a German theoretical physicist in 1915, four years after the war in which Britain and Germany had been enemies. Arthur Eddington and Frank Dyson were attempting to weigh the fabric of spacetime with starlight.
What Einstein Predicted — and Why It Was Testable Only During an Eclipse
Albert Einstein's general theory of relativity, published in November 1915, proposed that massive objects curve the geometry of spacetime around them, and that this curvature deflects the paths of everything passing nearby — including light. The sun, being the nearest massive object available to observation, would bend passing starlight by a calculable amount.
Source: Eddington experiment — Wikipedia
Newton's theory of gravity also predicted a deflection of starlight. Light, after all, has energy, and energy interacts with gravity in Newtonian physics. But the two predictions gave different numbers. In Newtonian gravity, the deflection of starlight grazing the sun's limb was predicted at approximately 0.87 arcseconds. Einstein's general relativity predicted exactly twice that: 1.75 arcseconds.
Source: The Eclipse of 1919 — Putting Relativity to the Test — Classic History
An arcsecond is 1/3600 of a degree — a vanishingly small angle, roughly the apparent size of a golf ball seen from a mile away. Measuring a deflection of 1.75 arcseconds required extraordinary precision and a specific observational opportunity: you can only see stars near the sun's disk when the sun's own light is blocked, which means a total solar eclipse. And for the stars to be at their most useful, they needed to be a bright cluster that would appear near the eclipsed sun at the right time of year. In May 1919, the Hyades star cluster — one of the richest near-sun star fields in the sky — would appear close to the eclipsed sun's disk. The geometry would not recur as favorably for years.
Definition — arcsecond: One arcsecond equals 1/3600 of one degree of angular measure, or 1/1,296,000 of a full circle. At the distance of the sun, an angular deflection of 1.75 arcseconds corresponds to a physical displacement of approximately 1,750 kilometers at the sun's limb.
Source: Shadow of the Moon and General Relativity — arxiv.org
Frank Watson Dyson, the British Astronomer Royal, recognized the opportunity in 1917. He secured funding and began organizing two expeditions: one to Príncipe, led by Eddington, and one to Sobral in Brazil, led by Andrew Crommelin and Charles Davidson of the Royal Greenwich Observatory.
The Setup: Two Expeditions, One Eclipse
The Príncipe expedition departed in February 1919 — a journey made possible by the Armistice of November 1918, which ended World War I just in time for international scientific cooperation to resume. Eddington, a committed Quaker and pacifist who had resisted military conscription during the war, had particular motivation to demonstrate the universality of physical law across national boundaries.
Source: A Total Solar Eclipse 100 Years Ago Proved Einstein's General Relativity — Smithsonian Magazine
Eddington and astronomer Edwin Cottingham arrived on Príncipe with a 16-inch coelostat-fed telescope designed to produce photographs of sufficient scale to measure the tiny star position shifts. They spent months setting up the equipment and taking calibration plates of the Hyades cluster in the nighttime sky — the reference images against which the eclipse plates would be compared.
Source: Einstein and Eddington and the Eclipse in Príncipe — arxiv.org
The Sobral team carried two instruments: a 4-inch astrograph and a larger astrographic telescope. The astrograph was smaller but had a good lens; the astrographic telescope had a larger aperture but proved problematic.
| Location | Team | Equipment | Pre-eclipse concern |
|---|---|---|---|
| Príncipe, West Africa | Eddington + Cottingham | 16-inch coelostat telescope | Cloud cover forecast uncertain |
| Sobral, Brazil | Crommelin + Davidson | 4-inch astrograph + larger astrographic telescope | Heat-induced focus problems |
Source: The 1919 Eclipse Results That Verified General Relativity — Royal Society Notes and Records
The Sobral site offered better weather prospects than Príncipe. Brazil's dry season in late May made clear skies likely. Príncipe's equatorial climate was less predictable, and observers on the island noted that the morning of the eclipse began overcast.
The Protocol: Photographing Stars Near the Sun
The observational procedure was carefully planned. During totality — the brief period when the moon completely covers the sun — each team would expose as many photographic plates as possible, tracking the position of the sun-adjacent Hyades stars. After the eclipse, these plates would be measured to determine the apparent positions of the stars. Those positions would be compared to the reference plates taken when the stars appeared in the nighttime sky, far from the sun.
Source: General Relativity and the 1919 Solar Eclipse — Royal Observatory Greenwich
If starlight was bent by the sun's gravity, stars appearing near the sun's disk during the eclipse would seem displaced outward — shifted slightly away from the sun's center — compared to their reference positions. The amount of that displacement, converted to an angular measure, could then be compared against the Newtonian prediction of 0.87 arcseconds and Einstein's prediction of 1.75 arcseconds.
On Príncipe, the eclipse morning was cloudy. Eddington's team could see the sun through the overcast for much of the eclipse but only managed to expose 16 plates before the window closed. Many of those plates showed the Hyades stars only faintly or not at all through the clouds. When Eddington developed the plates on the island itself, examining them that same night, he found that only two plates had enough clearly defined stars to be scientifically useful.
Those two plates gave a deflection of 1.61 ± 0.30 arcseconds. The large uncertainty (0.30 arcseconds) reflected the small number of usable stars on those plates. But 1.61, within the stated uncertainty, was consistent with Einstein's prediction of 1.75 and inconsistent with the Newtonian prediction of 0.87.
At Sobral, the larger astrographic telescope produced plates that, when measured, gave a deflection of 0.93 arcseconds — close to the Newtonian prediction and inconsistent with Einstein. The smaller 4-inch astrograph at Sobral gave a deflection of 1.98 ± 0.12 arcseconds — the most precise measurement of the entire expedition, and the one closest to Einstein's prediction.
Source: Eddington experiment — Wikipedia
The researchers now faced a decision: three sets of plates, giving three different results. Which did they trust?
The Results: Three Datasets, One Conclusion
When Dyson, Eddington, and their colleagues returned to England and completed the plate measurements, they had to decide how to weight the three datasets:
| Dataset | Measured deflection | Uncertainty | Verdict |
|---|---|---|---|
| Príncipe 16-inch (2 usable plates) | 1.61 arcsec | ±0.30 arcsec | Consistent with Einstein |
| Sobral 4-inch astrograph | 1.98 arcsec | ±0.12 arcsec | Consistent with Einstein |
| Sobral astrographic telescope | 0.93 arcsec | ±0.05 arcsec | Consistent with Newton |
Source: The 1919 Eclipse Results — Royal Society Notes and Records
Dyson — not Eddington — made the call to downweight the Sobral astrographic telescope data. His reason: the instrument had developed a focus problem in the heat of the Brazilian summer. Before the eclipse, the team had noticed that the image quality from the astrographic telescope was degrading, and they had been uncertain whether it would produce reliable results. Post-expedition, when the plates were measured, the systematic error in the astrographic telescope's results was consistent with a focus shift of the kind they had feared.
Source: Daniel Kennefick on No Shadow of a Doubt — Princeton University Press
The conclusion, announced at a joint meeting of the Royal Society and the Royal Astronomical Society on November 6, 1919: the observations were consistent with Einstein's theory of general relativity. The deflection of light by the sun's gravity matched the relativistic prediction, not the Newtonian one.
The announcement made global headlines the following morning. The London Times carried the headline "Revolution in Science — New Theory of the Universe — Newtonian Ideas Overthrown." The New York Times followed with "Lights All Askew in the Heavens." Within weeks, Albert Einstein had been transformed from a well-regarded theoretical physicist known within his field into a global celebrity.
Source: A Total Solar Eclipse 100 Years Ago Proved Einstein's General Relativity — Smithsonian Magazine
The Controversy: Did Eddington Cherry-Pick the Plates?
The accusation that Eddington committed scientific fraud — selecting the data that confirmed Einstein and discarding data that contradicted him — has circulated in the history-of-science literature for decades.
The most explicit version of this charge appeared in a 1980 paper by historians John Earman and Clark Glymour, who argued that the decision to set aside the Sobral astrographic telescope data was scientifically unjustified and that the remaining data had uncertainties large enough to make the result inconclusive in any rigorous sense. They suggested that Eddington's Quaker pacifism and personal enthusiasm for Einstein's theory may have colored his judgment.
Source: Testing Relativity from the 1919 Eclipse — Physics Today / Kennefick
The accusation grew in visibility through the 1990s and 2000s, appearing in popular histories of physics and occasionally cited as an example of confirmation bias in scientific practice.
Daniel Kennefick, a physicist and science historian at the University of Arkansas, undertook the most thorough examination of the historical record in his 2009 Physics Today article and subsequently in his 2019 book No Shadow of a Doubt. His conclusions addressed the accusation systematically.
First, the decision to downweight the Sobral astrographic telescope data was made by Dyson, not Eddington. Dyson was Astronomer Royal, skeptical of Einstein's theory at the outset, and had antecedent technical concerns about the astrographic instrument — concerns documented before the eclipse. The decision was not ex post selection of favorable data; it was a judgment about instrument reliability based on pre-existing doubts.
Second, Eddington was responsible for the Príncipe expedition, not the Sobral expedition. Attributing the Sobral data selection decision to Eddington conflates two different people making decisions about two different instruments at two different locations.
Third, a reanalysis of the original 1919 plates was conducted in 1978 by astronomers at the Royal Greenwich Observatory using modern plate-measuring equipment and computational methods unavailable in 1919. Their results confirmed the original conclusions.
Source: The 1919 Eclipse Results Which Verified General Relativity and Their Later Detractors — arxiv.org
The cherry-picking narrative had another structural problem: if Eddington was willing to fabricate results, why did the Príncipe plates — his own expedition's data — show a deflection of only 1.61 arcseconds, significantly less than Einstein's prediction of 1.75? A fraudster trying to confirm Einstein would have engineered a result closer to 1.75. Instead, the messiest, least certain data came from Eddington's own expedition, and the clearest, most precise confirmation came from the Sobral 4-inch astrograph — a dataset that neither Eddington nor Dyson was responsible for collecting.
Source: The 1919 Eclipse Results That Verified General Relativity — Royal Society Notes and Records
Replication: How Subsequent Experiments Confirmed the Result
The 1919 expedition was the first test of general relativity's light-bending prediction, but it was not the last. Several subsequent solar eclipse expeditions through the 1920s and 1930s produced measurements broadly consistent with Einstein's prediction, with improving precision as equipment improved.
| Year | Measurement | Uncertainty |
|---|---|---|
| 1919 (Sobral 4-inch) | 1.98 arcsec | ±0.12 |
| 1919 (Príncipe) | 1.61 arcsec | ±0.30 |
| 1922 (Lick Observatory, Wallal) | 1.72 arcsec | ±0.11 |
| 1929 (Sumatra) | 1.70 arcsec | ±0.10 |
| 1952 (Sudan) | 1.70 arcsec | ±0.10 |
Source: Eddington experiment — Wikipedia
The advent of radio astronomy and, particularly, Very Long Baseline Interferometry (VLBI) transformed the precision available for this measurement. VLBI uses networks of radio telescopes separated by thousands of kilometers to act as a single enormous instrument, achieving angular resolution far beyond what optical telescopes can achieve.
Modern VLBI measurements of the deflection of radio waves from distant quasars by the sun's gravity agree with Einstein's prediction with an accuracy of approximately 0.02 percent — four orders of magnitude better than the 1919 expedition achieved. The measured parameter is expressed as γ (gamma) in the parameterized post-Newtonian formalism: general relativity predicts γ = 1.00000. Modern VLBI measurements give γ = 0.99983 ± 0.00045.
Source: Testing General Relativity with Geodetic VLBI — Astronomy and Astrophysics
The question the 1919 experiment answered in principle — does starlight bend by the amount general relativity predicts — has been answered in practice to exquisite precision. General relativity is correct. The 1919 expedition pointed in the right direction, with the accuracy available at the time.
Where It Stands Today: General Relativity Confirmed at Every Scale
The 1919 eclipse experiment was a test of one prediction of one theory. General relativity has since been tested by dozens of independent methods, across an enormous range of scales and regimes.
Gravitational time dilation — the slowing of clocks in gravitational fields — has been measured in terrestrial laboratories using atomic clocks and confirmed to parts-per-trillion precision. GPS satellites require relativistic corrections to maintain accuracy; without accounting for both special and general relativistic effects, GPS positions would drift by kilometers per day. The precession of Mercury's orbit, which Newton's gravity could not fully explain, matches general relativity's prediction precisely. Gravitational lensing — the bending of light from distant galaxies by intervening mass — is now a standard tool in observational cosmology.
The detection of gravitational waves by LIGO in 2015 provided a fundamentally different confirmation: not light bending around mass, but spacetime itself rippling from the collision of two black holes 1.3 billion light-years away, with the wave properties matching general relativistic predictions in detail.
The 1919 expedition sits at the beginning of this century of tests, not at its end. It was decisive enough, and publicly visible enough, to establish the theory's credibility when it was contested. What subsequent tests have done is refine and confirm the picture that Eddington and Dyson's photographs first sketched.
Source: A Record-Setting Eddington Experiment — Physics Today
It is also worth noting what made the 1919 result historically unique. The announcement was made in November 1919, less than a year after the Armistice that ended the bloodiest war in European history. A British scientist confirming the theory of a German scientist, in the hall of the Royal Society, was a deliberate act of international reconciliation as much as a scientific event. Eddington, who had refused military service, understood the announcement as evidence that science transcended the tribalism that had driven the war.
FAQs
Q: What exactly was measured in the 1919 eclipse experiment? A: Astronomers measured the apparent positions of stars in the Hyades cluster that appeared close to the eclipsed sun during totality. These positions were compared to reference measurements of the same stars taken at night, when the sun was not nearby. Any difference in apparent position — a displacement of the stars away from the sun — would indicate that the sun's gravity was bending the light from those stars.
Q: What were the two competing predictions, and what was the difference? A: Newton's theory of gravity, applied to light with energy, predicted a deflection of approximately 0.87 arcseconds for light grazing the sun's limb. Einstein's general relativity predicted 1.75 arcseconds — exactly double the Newtonian value. The factor of two arises because general relativity adds a spacetime curvature contribution that is absent in Newtonian gravity.
Q: Why was the Sobral astrographic telescope data rejected? A: The Sobral astrographic telescope had developed a focus problem in the Brazilian summer heat, a concern the team had documented before the eclipse. Dyson — not Eddington — made the decision to downweight this dataset because of the instrument's technical failure. Subsequent analyses supported this judgment.
Q: Did Eddington cherry-pick the data to confirm Einstein? A: The historical evidence, examined in detail by physicist-historian Daniel Kennefick in his 2009 Physics Today article and 2019 book No Shadow of a Doubt, does not support this conclusion. The data rejection decision was Dyson's, based on antecedent technical concerns. A 1978 reanalysis using modern plate-measuring technology confirmed the original results.
Q: How precise is the modern measurement compared to 1919? A: Modern VLBI measurements agree with Einstein's prediction at the level of approximately 0.02 percent — about 15 times more precisely than the best 1919 plate (which had about 6 percent uncertainty). VLBI measures the bending of radio waves from distant quasars rather than the visual bending of starlight, allowing year-round measurement without needing a solar eclipse.
Q: Why did Einstein become world-famous specifically because of this experiment? A: The experiment made the news because it offered a clear, binary choice between Newton and Einstein — two predictions, one observation, one theory confirmed and one overturned. Coming just a year after World War I, with a British scientist confirming a German theory, the symbolic dimensions were enormous. Newspaper editors found a story in it that could be told simply. The physics, paradoxically, was accessible precisely because the measurement was so dramatic.
The Takeaway
The 1919 eclipse expedition succeeded in doing what it set out to do: measure whether light bends around the sun by the amount Einstein's general relativity predicted, and distinguish that amount from Newton's smaller prediction. The measurements were imprecise by modern standards, the sample of usable plates was small, and the decisions about which data to trust were made by humans with instruments that misbehaved in the heat. But the experiment pointed in the right direction, and every subsequent test of the same prediction — from the 1922 eclipse observations through modern VLBI at 0.02 percent precision — has confirmed that direction. The cherry-picking accusation, examined carefully, does not survive contact with the historical record. What survives is a measurement made under difficult conditions, with flawed instruments, by scientists working at the edge of what was technically achievable — and getting it right.
Read next:
Weekly digest
Stay ahead of the curve
Get the best science, AI, and tech stories delivered to your inbox every week. No spam — unsubscribe any time.
About the Author
SciHub101 is an independent science and technology publication founded and edited by Buğra Sözeri, a software engineer and app developer based in Türkiye. Articles are drafted against peer-reviewed literature and primary sources, fact-checked, and edited by a human before publication. Where we compare products, anything built by our own team is labelled as such — see the editorial guidelines and About page for how corrections and disclosures work.
Related Articles
Eratosthenes Measures Earth: The Shadow Experiment
Around 240 BCE, Eratosthenes measured Earth's circumference using two sticks and shadows. Here is exactly how he did it — and why the result was nearly perfect.
The Dunning-Kruger Experiment: What the 1999 Study Actually Found
The Dunning-Kruger effect is widely misquoted. The 1999 Cornell study found something real but narrower than the famous 'peak of Mount Stupid' curve suggests.
The Bohr-Einstein Debates: The Thought Experiments That Shaped Quantum Physics
Einstein spent years designing thought experiments to break quantum mechanics. Bohr defeated every one. The Bohr-Einstein debates, from Solvay 1927 to the EPR paradox.
The Michelson-Morley Experiment: The Failure That Changed Physics
In 1887, Michelson and Morley tried to measure Earth's motion through the ether and found nothing. That null result demolished a century of physics and paved the way for Einstein.
Why Time Moves Slower Near Gravity
Time moves slower near gravity because mass curves spacetime, and general relativity says clocks deeper in that curvature tick at a measurably slower rate.
The Fabric of Space-Time: Einstein's Greatest Idea Explained Simply
Space and time are not a fixed stage on which events play out — they are dynamic, bendable, and woven together into a single fabric.