#116 1919 · Dyson, Eddington and Davidson (1919 eclipse expedition) · Science / experimental physics
Eddington tested Einstein's theory by waiting for the Moon to do the engineering for him
the problem
Testing whether the Sun bends starlight meant photographing stars the Sun's own glare erased
background
Einstein's 1915 general theory of relativity predicted that the Sun's gravity would measurably bend the path of starlight passing close to it — a testable, falsifiable claim that would distinguish his theory from Newtonian gravity, which predicted exactly half the deflection. But testing it meant photographing stars that appear, from Earth, extremely close to the Sun in the sky — precisely the position where the Sun's own overwhelming brightness makes any nearby star impossible to see or photograph on an ordinary day.
Building equipment powerful enough to filter out the Sun's glare while still resolving faint background stars right next to it was far beyond what any telescope of the era could do, and no artificial shield could replicate blocking a light source that bright at that scale. What the experiment needed was something to physically block the Sun's disc itself, precisely and briefly, without also blocking the stars around it.
what everyone would do
The obvious path to testing Einstein's prediction was to build better instruments — a telescope or filter powerful enough to resolve faint stars right next to the Sun despite its glare. That was beyond any achievable engineering of the era: no artificial shield could block a light source as overwhelming as the Sun at that angular scale while leaving the surrounding starlight intact.
what they saw
Dyson and Eddington saw that the obstacle — the Sun's own light drowning out nearby stars — was a problem nature already solved on a predictable schedule: a total solar eclipse blocks the Sun's disc precisely and completely for a few minutes, using the Moon as a shield no engineer could improve on. The experiment didn't need to be invented from scratch; it needed to be scheduled around an event that already did the hardest part for free.
the move
Astronomers Arthur Eddington and Edwin Cottingham traveled to the island of Príncipe off West Africa, while Andrew Crommelin and Charles Davidson traveled to Sobral, Brazil, to observe the total solar eclipse of 29 May 1919 — using the Moon's own disc, precisely aligned by nature, as the free shield that made the stars near the Sun visible for the few minutes of totality. They photographed the star field around the eclipsed Sun and compared the stars' apparent positions to reference photographs of the same field taken months earlier at night, without the Sun anywhere nearby.
why it works
During totality the Moon blocks the Sun's overwhelming brightness while leaving the surrounding star field visible, which is exactly the condition the experiment needed and precisely the condition no artificial apparatus could replicate at the required scale. Photographing that star field during the brief eclipse window and comparing star positions against reference photos of the same field taken at night, months earlier without the Sun nearby, isolates exactly one variable — whether starlight passing near the Sun bent — because everything else about the stars' true positions is already known from the reference plates. The predicted deflection (1.75 arcseconds under general relativity versus 0.875 under Newtonian gravity) was large enough to distinguish the two theories cleanly, so once the glare problem was solved for free by the eclipse, the measurement itself just had to be precise, not clever.
the payoff
The measured deflection matched Einstein's predicted 1.75 arcseconds rather than the 0.875 arcseconds Newtonian gravity predicted, and the results were announced to a joint meeting of the Royal Society and Royal Astronomical Society on 6 November 1919, instantly confirming general relativity and making Einstein a global celebrity almost overnight.
where it breaks
This substitution only works when the natural event is both rare enough to be worth waiting for and precisely predictable enough to plan an expensive, logistically demanding expedition around it — eclipses are calculable years in advance, which is what made this feasible at all. It fails whenever the natural condition doesn't fully replicate the artificial one needed (partial eclipses don't fully remove the glare, clouds can ruin the single opportunity with no immediate retry, as nearly happened at both 1919 sites), and it fails when there's no natural analog at all for the obstacle being worked around, forcing genuine engineering rather than clever scheduling.
what came after
The 1919 eclipse expeditions are recognized as one of the most consequential experimental confirmations in the history of physics, and the episode remains a standard teaching example of designing an experiment around a naturally occurring, precisely timed event rather than trying to engineer an equivalent artificial condition from scratch.
references
- [1]Eddington experimentWikipedia, 2025en.wikipedia.org
- [2]General Relativity and the 1919 Solar EclipseRoyal Observatory Greenwich, 2019royalobservatorygreenwich.org
- [3]Observing general relativityThe Royal Society, 2024royalsociety.org