Henrietta Leavitt and the 25 Stars That Measured the Universe

On October 12, 2026, a short paper went up on the arXiv preprint server with an old-fashioned title: “Henrietta Leavitt’s classical Cepheid relation.” Its author, Daniel Majaess of Mount Saint Vincent University in Halifax, had done something unusual for a working astrophysicist. He took the numbers a woman wrote by hand in Cambridge, Massachusetts, more than a century ago and tested them against data from the European Space Agency’s Gaia satellite. The question was simple. When her measurements wandered off a straight line, was that just the noise of an old method, or was it something real in the sky?

Part of the answer was surprising. The scatter in her data lines up with modern measurements of the same stars, with a correlation of about 0.76. Some of what looked like error turns out to be the real shape of a small galaxy, stretched out along our line of sight, and the dust drifting through it. The same paper estimates that at least 870 of the variable stars she listed in 1908, more than nine out of ten, are the kind of pulsing star now called a classical Cepheid. A year and a half earlier, a team that included the Nobel laureate Adam Riess had rebuilt her light curves from her own notebooks and reached a similar verdict: her results are “in excellent agreement with contemporary measurements.”

That two groups would spend this kind of effort on a 1912 graph says a lot. Henrietta Swan Leavitt was paid 30 cents an hour to look at glass photographs of stars. She never ran a telescope, because the women who did her job weren’t allowed to. Out of that work came the first reliable way to measure the distance to another galaxy. It’s called Leavitt’s law, and a century later it still underpins the measurement at the center of the biggest argument in cosmology.

The women who read the sky

In the late 1800s, the Harvard College Observatory was turning into a factory for starlight. Its director, Edward Charles Pickering, had bet early on photography. A camera on a telescope could record thousands of stars on one glass plate in one night, and the plate could be studied for years afterward. The trouble was that somebody had to study it. Each plate had to be examined under magnification, and every star had to be identified, measured and written down.

Pickering’s answer was to hire women. They were educated, they were available, and they were cheap. Working as a “computer,” the old word for a person who computes, a woman at the observatory typically earned 25 cents an hour, seven hours a day, six days a week. That was less than a man would have been paid, and the arrangement suited the observatory’s budget. Women were not allowed to operate the big telescopes. Men took the photographs at night, and the women took the measurements by day.

The first of them came in through the kitchen. Williamina Fleming, a Scottish immigrant whose husband had walked out on her, was working as a maid in Pickering’s house when he hired her for part-time work at the observatory in 1879. She became one of the best plate readers in the world. She classified most of the 10,000-plus stars in the first Henry Draper Catalogue, found the Horsehead Nebula on a plate in 1888, and in 1898 became Harvard’s Curator of Astronomical Photographs, the first woman to hold the post. According to the observatory archive, she is the woman standing in the middle of the photograph at the top of this page, and Antonia Maury is the third from the left, holding a magnifying glass. Leavitt hadn’t arrived yet.

Others followed. Maury devised a way to sort stars by the sharpness of their spectral lines that Pickering refused to use. The Danish astronomer Ejnar Hertzsprung later saw that her system separated giant stars from dwarfs, which became part of the famous Hertzsprung–Russell diagram. Annie Jump Cannon, nearly deaf after a bout of scarlet fever, built the sequence of spectral types still taught today, O, B, A, F, G, K, M. Over her career she classified around 350,000 stars by hand. The money for much of this came from Mary Anna Draper, the widow of an amateur astronomer, who paid for a catalog of stellar spectra in her husband’s memory. Spectroscopy, the craft of reading a star’s makeup from its light, was barely a generation old at the time. It had only arrived with discoveries like the element first found in the Sun during the 1868 eclipse.

1868 was also the year Henrietta Swan Leavitt was born.

A minister’s daughter with a magnifier

Leavitt was born on July 4, 1868, in Lancaster, Massachusetts, the daughter of a Congregational minister. She studied at Oberlin College and then at the Society for the Collegiate Instruction of Women in Cambridge, the school that became Radcliffe College. Her courses ran from classical Greek and fine arts to analytic geometry and calculus. In her last year she took an astronomy course, and it stuck. After she graduated in 1892, she started volunteering at the observatory.

Her early years there were interrupted again and again. She traveled in Europe, worked for a time as an art assistant at Beloit College in Wisconsin, and fell seriously ill. Her hearing, which had started failing when she was a teenager, kept getting worse for the rest of her life. In 1903 she returned to Harvard for good. Because her family had money, Pickering at first didn’t need to pay her. Later he did: 30 cents an hour, $10.50 a week. That was a nickel an hour more than most of the computers made, a small sign of how much he valued her.

Colleagues described her as serious, hard-working and kind. Solon Bailey, who worked alongside her for years, later wrote that she had “the happy, joyful, faculty of appreciating all that was worthy and lovable in others.” Pickering gave her an assignment that suited her patience and her eye. She was to find variable stars, stars that change in brightness, and measure how they changed.

The method was slow and very physical. Leavitt would take two plates of the same patch of sky, shot on different nights, and compare them. Sometimes she laid a positive over a negative so that a star that hadn’t changed seemed to cancel out, while a star that had brightened or faded stood out. Each plate held thousands of stars. She went through them one by one with a magnifying eyepiece, looking for tiny differences in the size and darkness of a dot of silver.

Plates from Peru

The stars that made her famous can’t be seen from Massachusetts at all. The Large and Small Magellanic Clouds sit deep in the southern sky. To a person standing in the Andes or the Australian outback, they look like two pieces of the Milky Way that broke off and drifted away. Harvard photographed them from its Boyden Station, a southern outpost the observatory ran at Arequipa, Peru, from 1890 until it moved to South Africa in 1927.

The main camera there was the Bruce telescope, a 24-inch photographic refractor built by Alvan Clark & Sons in 1893 and paid for by the philanthropist Catherine Wolfe Bruce. Its plates were packed in crates and sent north by ship and rail to Cambridge, where they joined a collection that would eventually pass half a million glass photographs. Leavitt never saw the Magellanic Clouds with her own eyes. She knew them as patterns of black specks on glass.

What made the clouds special, though nobody quite realized it yet, was that every star in one of them is at roughly the same distance from us. The Small Magellanic Cloud is now known to be a dwarf galaxy about 200,000 light-years away. A star at its near edge and a star at its far edge differ in distance by only a few percent. So if one star in the cloud looks brighter than another, it really is brighter. Distance has been taken out of the comparison. For an astronomer trying to understand what stars actually do, that is a gift, a whole population of stars lined up at the same starting line.

1,777 variable stars

Leavitt hunted variables in the clouds with a thoroughness that startled her colleagues. In 1908 she published a paper in the Annals of the Harvard College Observatory with the plain title “1777 Variables in the Magellanic Clouds.” It was a huge haul. For most of the stars she gave only positions and rough brightness ranges. For 16 of them, all in the Small Magellanic Cloud, she had enough plates to work out a period, the number of days it takes the star to fade and brighten and come back to where it started.

In the middle of a long, dry paper she put one sentence that turned out to be the most important thing in it:

“It is worthy of notice that in Table VI the brighter variables have the longer periods.”

Most readers passed over it. Leavitt didn’t. She kept collecting plates, kept timing her stars, and four years later came back with more.

Twenty-five stars and a straight line

On March 3, 1912, Harvard College Observatory Circular 173 came out under the title “Periods of 25 Variable Stars in the Small Magellanic Cloud.” Pickering signed it, as he did all the circulars. The first sentence says it was “prepared by Miss Leavitt.” The 25 stars had periods ranging from just over a day to 127 days. When she plotted their brightness against their periods, the points made a curve. When she plotted brightness against the logarithm of the period instead, the curve straightened out.

“A straight line can be readily drawn among each of the two series of points corresponding to maxima and minima, thus showing that there is a simple relation between the brightness of the Cepheid variables and their periods.”

That is Leavitt’s law in her own words. The two lines in her figure trace each star’s brightest and faintest points. They run parallel, and the slope tells you how much brighter a star gets for each step up in period. Since all 25 stars sat at the same distance, the relation wasn’t a quirk of where they happened to be. It was a property of the stars themselves. Leavitt saw what that meant. She wrote that she hoped the parallaxes of a few Cepheids closer to home could be measured, so the relation could be tied to real distances.

The stars were Cepheids, named after Delta Cephei, a star in the constellation Cepheus that brightens and fades every 5.37 days. Its changes were first noticed in 1784 by John Goodricke, a young amateur astronomer in York, England, working with his neighbor Edward Pigott. Goodricke had been deaf since early childhood. He died in 1786 at 21, four days after he was elected to the Royal Society, without ever hearing the news. More than a century later, the most important work on the stars he found was done by another astronomer who could barely hear.

Why one line could measure the universe

The point of Leavitt’s law is easy to state. A light bulb across a dark field looks fainter than the same bulb on your desk. If you know how bright it really is, you can work out how far away it is from how faint it looks, because light spreads out and dims with the square of the distance. Astronomers call an object of known true brightness a “standard candle.” Before 1912 they didn’t have a good one.

Cepheids fixed that. Time a Cepheid’s flicker and the period tells you its true brightness. Compare that with how bright it looks and you have its distance. A star with a 30-day period anywhere in the universe should be about as luminous as one with a 30-day period in the Small Magellanic Cloud. Cepheids are also giants, thousands to tens of thousands of times brighter than the Sun, so they can be picked out in other galaxies. A distance measured this way is usually given in light-years, a unit that only works because people managed to pin down the speed of light.

What Leavitt’s law lacked at first was a zero point. Her line showed how much brighter one Cepheid was than another. It didn’t say how bright any one of them really was. Getting that number meant measuring the distance to at least a few Cepheids some other way, and they’re rare enough that none sits close to the Sun. It’s the same kind of problem as a ruler with evenly spaced marks but no numbers on them. For the length of a kilogram or a meter, scientists eventually tied their standards to constants of nature. For Cepheids, the first try came within a year.

In 1913 Ejnar Hertzsprung, the same Dane who had seen the value of Antonia Maury’s work, used statistical methods on the motions of a handful of nearby Cepheids to estimate how bright they really were. Then he used Leavitt’s line to estimate how far away the Small Magellanic Cloud was. His published figure put the cloud about ten times too close, possibly because of a slip of the pen, and it ignored the dust that dims starlight on its way to us. But the method worked. For the first time someone had put a number on the distance to something outside the solar neighborhood using a single class of star.

In 1917 the English astrophysicist Arthur Eddington explained why Cepheids behave the way they do. They pulsate. The whole star swells and shrinks, its outer layers heating and cooling, like an engine with a valve that traps heat and then lets it go. Bigger, brighter Cepheids are puffier and pulsate more slowly, the way a large bell rings at a lower pitch than a small one. Leavitt’s line was the outward sign of that physics.

The letter that destroyed a universe

The first person to push Leavitt’s law hard was Harlow Shapley, a young astronomer at Mount Wilson Observatory in California. Around 1918 he used variable stars in globular clusters, the dense balls of old stars that swarm around the Milky Way, to map where the clusters are. They turned out to be grouped around a point far off in the direction of Sagittarius. Shapley concluded that the Sun is not at the center of the galaxy. That much was right. He also concluded that the Milky Way was enormous, about 300,000 light-years across, and that it was essentially the whole universe.

On April 26, 1920, Shapley argued the point in Washington, D.C., against Heber Curtis of Lick Observatory, in what is now called the Great Debate. Curtis said the Milky Way was smaller and that the faint “spiral nebulae” scattered across the sky were other galaxies, “island universes” in their own right. Shapley said the spirals were nearby wisps inside our own system. Nobody won outright, because nobody could yet measure how far away a spiral nebula was. A year later Shapley became director of the Harvard College Observatory, Leavitt’s boss, and in 1921 he made her head of stellar photometry.

The question was settled on a mountain above Pasadena. On the night of October 5–6, 1923, in poor seeing, Edwin Hubble made a 45-minute exposure of the Andromeda Nebula with the 100-inch Hooker telescope at Mount Wilson, then the largest telescope in the world. On the plate he marked three new spots with an “N,” for nova. When he compared it with older plates, one of them didn’t act like a nova. It brightened and faded again and again, on a schedule. Hubble crossed out the N and wrote, in a rare show of excitement, “VAR!”

The star, now called V1, had a period of 31.4 days. Leavitt’s law said a Cepheid with that period must be very luminous. To look as faint as it did, it had to be roughly a million light-years away, far outside even Shapley’s giant Milky Way. Andromeda wasn’t a cloud in our galaxy. It was a galaxy of its own. Hubble sent Shapley the light curve. Shapley, holding the letter, reportedly told a colleague, “Here is the letter that destroyed my universe.” The finding reached the public through The New York Times in November 1924 and was read out at a meeting of the American Astronomical Society on January 1, 1925.

By then Leavitt was dead. She had died of stomach cancer in Cambridge on December 12, 1921, at the age of 53. According to the Harvard-Smithsonian Center for Astrophysics, heavy rain was falling that night, so the stars she had spent her life on were hidden by clouds.

Hubble kept going. He found Cepheids in other spirals and, together with Milton Humason and the redshift measurements of Vesto Slipher, showed in 1929 that the farther away a galaxy is, the faster it is moving away from us. That relation, now called the Hubble–Lemaître law, was the first hard evidence that the universe is expanding. Every step of it rested on distances, and every distance rested on Leavitt’s line. Hubble often said she deserved a Nobel Prize. Today the sky is full of the galaxies those Cepheid distances opened up, including the ones astronomers have named after sombreros, owls and other everyday things.

The yardstick that was off by half

Leavitt’s law had a hidden flaw, and it took 30 years to find. Hubble’s distances were all too small, some of them by a large factor. His first value for the expansion rate of the universe, now called the Hubble constant, was around 500 kilometers per second per megaparsec. Today’s figures are near 70, which means his universe was far too small and much too young. Taken at face value it was younger than the Earth.

The fix came from Walter Baade, a German astronomer working at Mount Wilson during the Second World War. With Los Angeles under wartime blackout, the sky over the mountain was darker than usual, and Baade managed to resolve individual stars in the center of Andromeda for the first time. He found that stars come in two broad populations, young and metal-rich in the spiral arms, old and metal-poor in the halo and the globular clusters. Cepheids came in the same two varieties. The classical Cepheids that Leavitt had measured in the Small Magellanic Cloud were young, massive stars. Many of the variables Shapley and others had used to calibrate the relation were older, fainter stars now called type II Cepheids. They had been mixed together on one scale.

In 1952, at a meeting of the International Astronomical Union in Rome, Baade announced that the distance to Andromeda, and with it the size of the known universe, had to be roughly doubled. The room was astonished. The law itself was fine. It had been calibrated with the wrong stars.

Recognition that came too late

While she was alive, Leavitt got little credit for any of this. The 1912 circular went out under Pickering’s name. Pickering assigned her other work, and she spent years building the Harvard Standard, a scale for measuring the brightness of stars on photographs, which was adopted by an international committee in 1913. That same year she found T Pyxidis, a star that flares up every few decades and is one of the most frequent recurrent novae known. Her Cepheid work was hers to discover but, as a computer, not hers to develop. The theoretical work that turned her line into a measuring tool was left to men like Hertzsprung, Shapley and Hubble.

A few years after her death, the Swedish mathematician Gösta Mittag-Leffler wrote to Harvard about her discovery. He was a member of the Royal Swedish Academy of Sciences and was interested in nominating her for the Nobel Prize in Physics. Shapley had to tell him she had died. The prize is not awarded after death.

She was buried in the family plot at Cambridge Cemetery, under a hexagonal monument topped by a globe. Since then the honors have come slowly. The asteroid 5383 Leavitt and the Leavitt crater on the Moon are named for her, in part to honor deaf astronomers. In 2016 the writer Dava Sobel told the computers’ story in a best-selling book, The Glass Universe, and the art world has taken her up too. In 2018 the Radcliffe Institute exhibited quilts by the artist Anna Von Mertens showing the night sky over Lancaster the morning Leavitt was born and over Cambridge the evening she died.

Checking her homework a century later

The renewed interest has a practical side. For the past two decades Harvard has been digitizing its plate collection through a project called DASCH, short for Digital Access to a Sky Century @ Harvard. It began in 2004 under the astrophysicist Josh Grindlay, with a custom scanner that handles two plates about every 80 seconds. Scanning finished in March 2024 after more than 435,000 plates. The final data release, at the end of 2024, holds about 23.6 billion brightness measurements of some 252 million sources, spanning roughly a hundred years of sky from the 1880s to 1990. The plates Leavitt squinted at are now images anyone can download. Her handwritten notebooks were transcribed by volunteers through the Smithsonian Transcription Center and Harvard’s Project PHaEDRA.

That made it possible to check her work star by star. In a paper published in April 2025 in the Publications of the Astronomical Society of the Pacific, Louise Breuval of the European Space Agency, Caroline Huang of the Center for Astrophysics and Adam Riess of Johns Hopkins and the Space Telescope Science Institute did exactly that. Working only from the measurements on pages 4 to 40 of Leavitt’s 1905 notebook, they rebuilt light curves for all 25 of her 1912 Cepheids, with an average of 34 observations per star. Then they set those beside modern light curves from the OGLE survey and data from Gaia.

For 17 of the 25 stars, her old light curves show a clear period and the same shape as the modern ones, down to a small extra bump that some Cepheids show on the way down from peak brightness. All 25 turn out to be classical Cepheids pulsing in their basic mode. Her periods mostly match modern values very closely, even though they were worked out by hand from a few dozen plates spread over years. Modern data cut the scatter around her line roughly in half, as you’d expect. The team also found that her faintest, shortest-period stars came out a little too bright, because photographic plates don’t respond evenly to light and because faint stars in crowded fields blend with their neighbors. That flattened her slope a bit. It’s the kind of bias modern Cepheid hunters still have to fight. The paper’s conclusion: her results are “a testament to the enduring quality of her work.”

Majaess’s October 2026 preprint goes a step further. It argues that her leftover scatter is not just error. Her residuals track the modern ones, which suggests her plates partly picked up real differences between stars: the depth of the Small Magellanic Cloud, which is stretched out toward and away from us; the width of the band of temperatures in which Cepheids pulsate; and patchy dust inside the cloud. He also looked back at the 16 stars in her 1908 Table VI and found that they already define a very tight straight line, with a correlation of about −0.98. In other words, her famous sentence about brighter stars having longer periods rested on a stronger relation than she could claim at the time.

The yardstick at the center of today’s fight

Leavitt’s law is not a museum piece. Cepheids are still the first rung of the “distance ladder” astronomers use to measure how fast the universe is expanding. Parallax, the small shift in a star’s position as the Earth goes around the Sun, gives direct distances to nearby Cepheids in the Milky Way. Those calibrate Leavitt’s law. The law then gives distances to galaxies that host both Cepheids and a certain kind of exploding star, type Ia supernovae, and those supernovae carry the measurement out to galaxies far enough away that cosmic expansion dominates their motion.

The star in the image above, RS Puppis, shows how far Cepheid work has come. It’s a Cepheid about 6,500 light-years away that brightens and fades every 41 days or so, and it happens to sit inside a cloud of dust. As each pulse of light spreads out through the dust, it lights up different parts of the cloud, and astronomers have used that echo to measure the star’s distance directly.

The trouble is that the ladder and the early universe disagree. Measurements based on the cosmic microwave background, the afterglow of the Big Bang, together with the large-scale pattern of galaxies, imply an expansion rate of about 67 to 68 kilometers per second per megaparsec. The Cepheid-based ladder keeps coming out near 73. That gap, around 8 percent, is known as the Hubble tension. If it’s real, something is missing from the standard picture of the cosmos.

For years the obvious suspect was Leavitt’s own old problem, crowding. Cepheids in distant galaxies sit among many other stars, and if their light is blended with their neighbors’, they look brighter and closer than they are. The James Webb Space Telescope has sharp enough infrared vision to settle that. In 2024 the team led by Riess reported that Webb’s images of Cepheids agreed with Hubble’s and ruled out crowding as the explanation with high confidence. In April 2026 an international group called the H0 Distance Network, pooling many different methods into what it calls the Local Distance Network, reported a value of 73.50 plus or minus 0.81, a precision of just over 1 percent. The Center for Astrophysics described it as the most precise direct measurement yet of the universe’s current expansion rate. It sits about 5 to 7 standard deviations away from the early-universe value. Other teams, including one led by Wendy Freedman at the University of Chicago, use different stars and have argued the gap may be smaller than it looks. The argument isn’t settled, but everyone involved is arguing about refinements of a method that began with one woman, a stack of glass and a straight line.

What the glass still holds

There’s a nice symmetry in the way the story has come back around. Leavitt found her law by comparing old plates with new ones. A century later, astronomers found new things in her data by comparing her old numbers with new measurements. The Harvard plates are still being mined for stars that changed slowly over the 20th century, and the observatory’s notebooks are still being read.

It’s also a story about measurement, the slow, unglamorous work that the big ideas depend on. Like John Harrison’s clocks, which solved longitude through years of patient craft, Leavitt’s discovery came from someone paying closer attention than anyone else. She had no telescope time, no theory to test and very little hearing. She had a magnifier, thousands of plates and a habit of writing down what she saw.

You can still look at her stars. From the Southern Hemisphere, the Small Magellanic Cloud is easy to see without a telescope on a dark night, a faint smudge near the south celestial pole. Even a good pair of binoculars or one of the home telescopes we’ve ranked will turn it into a grainy field of stars. A few of those specks are Cepheids, flickering on schedules that a deaf woman in Cambridge timed on glass more than a hundred years ago, and that astronomers are still using to measure the universe.

Further reading