The Photograph That Proved Einstein Right: The 1919 Eclipse Plates

For about five minutes on May 29, 1919, the Moon blotted out the Sun over an island in the Gulf of Guinea and a hard-baked plain in northern Brazil, and two teams of astronomers pointed their cameras at a patch of sky no one could normally see. They were not photographing the eclipse for its beauty. They were trying to measure whether a handful of stars had moved.

The stars had not moved, of course. Their light had. If Albert Einstein was right, the Sun's gravity should bend the path of starlight grazing its edge by about 1.75 arcseconds, a shift so small it is roughly the width of a coin seen from two miles away. Newton's physics allowed for a bend too, but only about half as much. The difference between those two numbers was the difference between two universes, and it was hiding on a few panes of coated glass.

When the plates were finally measured and the result announced that November, Einstein went from a physicist known mostly to other physicists to the most famous scientist alive. This is the story of the pictures that did it.

Why Starlight Should Bend

Einstein published the general theory of relativity in 1915, in the middle of a war that made it nearly impossible for a German scientist's work to reach Britain. The theory recast gravity. Instead of a force reaching invisibly across space, gravity became geometry: mass curves spacetime, and everything moving through that region, including light, follows the curve.

That last part was the testable prize. Classical physics never settled on a single answer for what gravity does to light. Treating light as Newtonian corpuscles falling in the Sun's gravitational field gives a deflection of about 0.87 arcseconds for a ray skimming the Sun, while a classical wave picture offered no natural way to run the calculation at all. Einstein's earlier work had landed near that same Newtonian figure. His completed 1915 theory doubled it. The full curvature of spacetime near the Sun should bend a grazing ray by roughly 1.75 arcseconds, twice the Newtonian value.

Here was a clean experiment. Photograph stars whose light passes close to the Sun, measure how far their positions appear shifted from where they sit in the night sky, and read off which number nature had chosen. The catch is a tricky one. You cannot see stars next to the Sun. The only time you can is during the few minutes of a total eclipse, when the Moon covers the disk and the sky near the Sun goes dark enough for faint stars to register on a plate.

The Man Who Picked the Date

The eclipse of May 29, 1919 was, for this one purpose, almost too good to be true. Frank Dyson, the Astronomer Royal and director of the Royal Observatory at Greenwich, had flagged it years earlier. Totality that day was among the longest of the century, close to seven minutes at its peak over the mid-Atlantic and roughly five minutes at the two expedition sites, still long enough for the extended exposures that faint stars demand. More important, the eclipsed Sun would sit squarely in front of the Hyades, a bright open star cluster in the constellation Taurus near the ruddy star Aldebaran. A rich field of reasonably bright stars was exactly what the measurement needed. A sparse one would have given too few reference points to trust.

Dyson had a personal history with the problem. Back in 1917, he had looked at old eclipse plates from 1905 and failed to pull a convincing deflection out of the thin scattering of stars they showed. He argued in print that 1919 was the chance to settle the matter, that the coming eclipse should serve for an ample verification, or the contrary, of Einstein's theory. He organized two expeditions to raise the odds that at least one would come home with usable plates.

One went to Sobral, in the dry interior of the Brazilian state of Ceará, run by Andrew Crommelin and Charles Davidson of Greenwich. The other went to the island of Príncipe, off the west coast of Africa, led by Arthur Eddington of Cambridge with the clockmaker Edwin Cottingham along to keep the instruments running. Eddington, a Quaker and a conscientious objector, had spent part of the war years mastering Einstein's theory when almost no one in Britain would read it. He wanted very much for it to be true, a fact his critics would later seize on.

How You Weigh Light on a Glass Plate

The method was old-fashioned photographic astrometry, and it was brutally unforgiving. During totality, each expedition exposed a series of glass photographic plates through their telescopes, capturing the Sun's darkened disk ringed by whatever stars the sky would give up. Those were the eclipse plates. Months earlier, or months later, each team photographed the same star field at night, when the Sun was nowhere near it, using the same lenses where possible, although not always in the same mounting or location. Those were the comparison plates. Eddington also used separate "check plates" of another star field entirely to calibrate changes in the telescope's scale.

The whole experiment lived in the difference between the two sets. If gravity bent the starlight, each star near the Sun would appear pushed slightly outward, away from the disk, on the eclipse plate compared to its undisturbed nighttime position. Measure that outward shift for every star, account for the way it should shrink with distance from the Sun, and you get the deflection at the Sun's edge.

Measuring it meant sliding the plates under a micrometer and reading star positions to a precision of thousandths of a millimeter. Then came the corrections, and the corrections were the hard part. A telescope's scale can drift with temperature. The Sun's heat can warp a mirror between exposures. The two plate sets were taken on different nights, at different temperatures, sometimes with different instruments, and every one of those differences could fake or erase a shift far larger than the 1.75 arcseconds they were hunting. Teasing the real signal out of that noise took the rest of the summer and most of the fall.

Clouds Over Príncipe

Eddington nearly got nothing. The morning of the eclipse, Príncipe was buried under a tropical storm. Rain fell into the early afternoon, and the sky only began to break minutes before totality. Eddington exposed sixteen plates through the thinning cloud, mostly on faith, unable to see whether any stars were registering. Most came back useless, fogged by the halo of scattered sunlight or blocked by drifting cloud. A couple of plates held a few usable star images, and that was all he had.

He could not wait for the ship home. Eddington made a preliminary measurement on the island, and one good plate appeared to support Einstein; he later spoke of that moment as the greatest of his life. The full reduction of the two usable plates, finished back in England against comparison plates taken at Oxford, yielded a deflection of about 1.61 arcseconds, with a generous probable error of roughly 0.30 arcseconds. Probable error was the statistical convention of the day, and it runs smaller than a modern standard deviation, so the true uncertainty was wider still. The result sat closer to Einstein than to Newton, but it rested on a thin handful of stars, and on its own it would not have convinced a skeptic.

The Plates That Carried the Result

Sobral had far better weather. There was some cloud, but conditions were good enough to produce numerous usable plates, and the expedition ran two instruments side by side. One was a four-inch telescope. The other was a larger astrographic lens borrowed for the trip, and it betrayed them: the instrument lost focus, probably because the heat affected the telescope or its coelostat mirror, smearing the star images into blurred, unreliable ovals.

The two Sobral instruments gave two very different answers, and this is where the story earns its footnotes. The good four-inch plates delivered the sharpest, most trustworthy result of the entire enterprise, a deflection of about 1.98 arcseconds with a reported probable error of only 0.12. That number is squarely on Einstein's side and nowhere near Newton's. The defective astrographic plates gave no single answer at all: depending on how the severe focus and scale changes were modeled, the reduction produced roughly 0.93 arcseconds, close to the Newtonian half-value, or roughly 1.52, much nearer Einstein's, both with substantial uncertainty. Dyson and his colleagues judged the dataset too compromised to carry weight and set it aside, leaning on the four-inch Sobral plates and the Príncipe pair.

Decades later that decision drew fire. In 1980 two philosophers of science argued that the team had thrown out good data that disagreed with Einstein and kept the data that agreed, that Eddington had put his thumb on the scale to hand relativity a victory. It became a popular story, the tidy kind that says a famous result was really a fudge. The record is thinner than the accusation. Eddington did not run the Sobral reduction; Dyson's Greenwich team did, and the astrographic plates had a documented, physical defect visible in the star images themselves, not merely an inconvenient number. A remeasurement reported in 1979, using modern measuring machines, recovered about 1.55 arcseconds from the astrographic plates, suggesting that their original Newtonian-looking result arose largely from difficulties determining the instrument's scale. The reanalysis reinforced the expedition's overall support for Einstein. The plates that were kept say what they were said to say.

The Announcement That Made a Legend

On November 6, 1919, the Royal Society and the Royal Astronomical Society held a joint meeting in London, under a portrait of Isaac Newton hanging on the wall. Dyson presented the results. The eclipse plates, he told the room, agreed with Einstein and not with Newton. Joseph John Thomson, the discoverer of the electron and president of the Royal Society, called it the most important result obtained in connection with the theory of gravitation since Newton's day, and said Einstein's work was one of the highest achievements of human thought.

The press did the rest. The next morning the Times of London ran the headline "Revolution in Science," with "New Theory of the Universe" and "Newtonian Ideas Overthrown" stacked beneath it. Days later the New York Times piled clause on clause: lights all askew in the heavens, men of science more or less agog. The image of a lone genius who had overturned Newton, confirmed by a British expedition photographing a German's theory the year after the war, was irresistible. Einstein woke up famous and stayed that way for the rest of his life.

One anecdote from that period has outlived almost everything else about the meeting, though it comes with a caveat. As the story goes, the physicist Ludwik Silberstein told Eddington he must be one of only three people in the world who truly understood general relativity. Eddington paused. Silberstein pressed him not to be modest, and Eddington replied that he was trying to think who the third person was. It is a wonderful line. It is also the kind of line that improves with every retelling, and it should be enjoyed as a good story rather than filed as a transcript.

What the Glass Actually Settled

Strip away the headlines and what remains is a measurement, and measurements can be redone. The 1919 error bars were large by any modern standard, and honest observers at the time said so. What turned a promising result into settled fact was a century of repetition. Astronomers tried the same eclipse experiment through the 1920s, 1930s, and beyond, improving on the numbers. Radio astronomers eventually made the whole thing cleaner by tracking quasars whose signals pass near the Sun, no eclipse required, and confirmed the deflection to a fraction of a percent.

The bending of starlight is no longer an exotic prediction. It is a working tool. Gravitational lensing, the same effect scaled up to entire galaxies bending the light of things behind them, is how astronomers now weigh dark matter and photograph objects near the edge of the observable universe. Every time a survey telescope catches a distant galaxy smeared into an arc by the gravity of a nearer cluster, it is running the 1919 experiment again, on a stage a billion times larger.

Shooting the Same Sky Today

You do not need to be testing relativity to point a camera at a total eclipse, and the modern version is far kinder than sixteen plates exposed through a storm. A total solar eclipse still demands the same two things it demanded of Eddington: a long enough lens to make the corona and any nearby stars worth capturing, and rigorous protection for your sensor and your eyes during every second that is not totality. A telephoto gets you a prominent Sun at the long end, though truly filling the frame with the disk takes far more reach, and a purpose-built solar filter over the front element is non-negotiable for the partial phases. Only in the brief window of totality does the filter come off.

The stars Eddington fought for are the hard part, then as now. Catching faint points of light around a bright corona is astrophotography under adversarial conditions, and if you want to build the skills that make a dark sky cooperate, Elia Locardi's cityscape and astrophotography course covers the exposure and stacking discipline that a scene like this rewards. The physics you would be recording is exactly what those two teams recorded on glass in 1919, only now it fits on a memory card and you already know the answer.

That is the strange gift of the 1919 plates. They captured something no one had ever seen and almost no one could see even then, a few stars shoved a hair's width sideways by the weight of the Sun, and in doing so they redrew the shape of space itself. The original eclipse plates appear to have been lost, although glass copies of at least one Sobral image survive in several observatory collections. The universe they revealed is the one we have lived in ever since.

Lead image - Albert Einstein at the blackboard during a lecture in Vienna in 1921, two years after the eclipse plates made him the most famous scientist alive. Ferdinand Schmutzer, Public domain. Source