sciencebriefs
13:00in productionCh. 1 · Two predictions, one testable difference/ 13:00 · ceiling 15 min
Physics · Astronomy & space

Eddington experiment

1919

In 1919 two eclipse expeditions checked whether starlight bends around the Sun by Newton's amount or Einstein's larger one, and the tidy textbook story leaves out how much judgment went into choosing which data to trust.

Einstein's general relativity predicted starlight grazing the Sun would bend by about 1.75 arcseconds, roughly twice the 0.83 arcseconds a Newtonian calculation gave, a clean enough difference to settle with a single measurement during a total eclipse. Arthur Eddington and Frank Dyson organised expeditions to Príncipe and Sobral for the eclipse of 29 May 1919, photographing stars near the darkened Sun to compare against earlier night photographs of the same stars. Cloud at Príncipe and blurred images from Sobral's main telescope complicated the results, leaving Eddington to rely on the Príncipe plates and Sobral's smaller backup telescope, both of which favoured Einstein's figure. The announcement in November 1919 made Einstein an international celebrity, though later critics questioned whether the excluded data reflected confirmation bias. A 1979 re-examination of the original plates largely supported Eddington's judgment, and far more precise radio measurements confirmed the full relativistic value decades later.

Chapters & takeaways6
  1. 0:08
    Two predictions, one testable difference

    Einstein's figure was almost exactly double the Newtonian prediction for light bending.

  2. 2:10
    Two expeditions, one eclipse

    Teams travelled to Príncipe and Sobral to photograph stars near the darkened Sun.

  3. 4:20
    Clouds, blur, and a backup telescope

    Poor conditions at both sites complicated getting a clean measurement.

  4. 6:30
    A result announced to the world

    The November 1919 announcement made Einstein an international celebrity almost overnight.

  5. 8:40
    Whose data counted

    Eddington set aside the telescope readings closer to the Newtonian value.

  6. 10:50
    Vindicated, mostly

    Later re-analysis and independent radio measurements largely bore out the original conclusion.

Worth your time?

Yes. Study the whole thing.

4.5/ 5
What works
  • the core prediction being tested was specific and decisive rather than a subtle statistical trend
  • a 1979 re-examination of the original plates largely supported Eddington's original judgment
  • far more precise radio measurements later confirmed the same result independently
What does not
  • conditions at both observing sites were poor enough to complicate a clean measurement
  • the decision to set aside some telescope readings has drawn genuine, still-discussed criticism
Study it if
  • anyone who wants the real story behind a famous, oversimplified scientific triumph
  • readers interested in how judgment calls about data selection shape a result's reception
  • people curious how Einstein went from a physicist to a public celebrity overnight
Skip it if
  • readers wanting the mathematics of general relativity's light-bending prediction explained in depth
  • anyone looking for a single, uncomplicated verdict rather than a genuinely debated one
The written brief3 min read

Two predictions, one testable difference

Einstein’s general theory of relativity, published in 1915, predicted that starlight passing close to the Sun would bend through an angle of about 1.75 arcseconds for a star seen right at the Sun’s edge, almost exactly double the roughly 0.83 arcseconds a straightforward Newtonian calculation of gravity’s pull on light would give. That factor-of-two gap gave astronomers an unusually clean test: rather than hunting for a subtle statistical trend across many observations, a single well-measured number, the apparent shift in a star’s position near the eclipsed Sun, could distinguish decisively between the two theories.

Two expeditions, one eclipse

Arthur Eddington, secretary of the Royal Astronomical Society and one of the few British astronomers who fully understood Einstein’s mathematics, worked with the Astronomer Royal, Frank Dyson, to organise two expeditions to observe the total solar eclipse of 29 May 1919, when stars near the Sun would briefly become visible against a darkened sky. Eddington and Edwin Cottingham travelled to the island of Príncipe off West Africa, while Andrew Crommelin and Charles Davidson of the Greenwich Observatory travelled to Sobral in Brazil, arriving in March 1919. Both teams aimed to photograph stars, including some in the Hyades cluster, appearing close to the eclipsed Sun, then compare their positions against earlier night photographs of the same stars taken with the Sun nowhere nearby.

Clouds, blur, and a backup telescope

Conditions were difficult at both sites. Heavy morning rain and cloud at Príncipe compromised the sixteen photographs Eddington managed to take there, leaving fewer usable star images than the team had hoped for going into the observation. At Sobral, the expedition’s main 13-inch telescope produced blurred images, apparently because of unexpected heating effects on the instrument during the exposures, and only a smaller 4-inch backup telescope at the same site produced results clear enough to measure with real confidence. That backup telescope’s readings, together with the usable Príncipe plates, pointed toward Einstein’s predicted value of light deflection rather than toward the smaller Newtonian figure.

A result announced to the world

The results were announced at a joint meeting of the Royal Society and Royal Astronomical Society on 6 November 1919 and formally published the following year, presented as confirming Einstein’s prediction over the Newtonian figure. J.J. Thomson, president of the Royal Society, endorsed the conclusion at that meeting, and the story became major international news within days, an unusually fast transition for what was, on its face, a fairly technical astronomical measurement. That announcement is generally credited with turning Einstein from a respected physicist known mainly within his field into an international public figure almost overnight.

Whose data counted

The tidy public account leaves out a genuine complication: Eddington gave little weight to the blurred images from Sobral’s main 13-inch telescope, which on their own suggested a value closer to the Newtonian prediction, and relied instead on the Príncipe plates and Sobral’s smaller backup instrument, both of which favoured Einstein’s figure. By the 1970s, some historians and scientists had raised the concern that this selective weighting amounted to confirmation bias rather than a purely technical judgment, since setting aside the larger telescope’s readings, whatever the justification involving its known heating problem, meant the celebrated result rested partly on a discretionary choice rather than a single unambiguous measurement.

Vindicated, mostly

A 1979 re-examination of the original Sobral photographic plates, using measuring equipment far more advanced than what was available in 1919, largely supported Eddington’s original handling of the data and found his analysis less affected by bias than some of the later criticism suggested, though the episode remains a genuinely debated case rather than a fully closed one. Independent confirmation eventually came from an entirely different method: radio-astronomical measurements in the late 1960s directly confirmed the full relativistic deflection value with far greater precision than eclipse photography ever could. The story is worth an hour both for the elegance of the original test and for how much more complicated the real process behind its famous conclusion turns out to have been.

Same field · Physics4 of 183
Up next in Science

Edward Drinker Cope

· 10:41

Cope didn’t found American palaeontology—he flooded it with names, many of which still stand only because no one has cleared the backlog.

10:41