The story
Settle for a moment beneath an imagined night sky, where the dark is not empty but deep with faint, patient light. Some of those small, steady glows are stars in our own Milky Way, close enough in cosmic terms to belong almost to the same neighborhood. Others are whole galaxies, each containing billions of stars, their light arriving only after crossing distances so great that the space between you and them has changed while the light was on its way.
On the largest scales, those galaxies are receding from one another. Not because the Milky Way occupies a privileged central place, and not because everything is simply flying outward through an already empty room, but because space itself has expanded. A useful, imperfect image is a rising loaf of dough with raisins scattered through it. As the dough swells, every raisin finds the others farther away. No raisin needs to be the center for the distances between them to grow.
The name for the present-day rate of this broad expansion is the Hubble constant, written as H-zero. It is not a speed limit, and it is not a number that has remained identical throughout the life of the universe. The expansion history changes with time; cosmologists describe that changing history with the Hubble parameter, H of time. The Hubble constant is the value of that parameter now, in this particular long, late moment of cosmic history.
Its units sound unusual at first: kilometers per second per megaparsec. A megaparsec is about 3.26 million light-years. If the constant were 70 kilometers per second per megaparsec, then a galaxy one megaparsec farther away would, on average, recede 70 kilometers per second faster. At ten megaparsecs, the difference would be about 700 kilometers per second. The rule becomes clearer as the distances become vast.
And yet this small number, so carefully written in notebooks and computer files, is not quite agreed upon. One method, beginning with nearby stars and reaching outward from measured distances, finds a value around 73 kilometers per second per megaparsec. Another begins with the oldest light we can observe and, using a well-tested model of the universe, infers a value closer to 67. The difference is only about 5.6 kilometers per second per megaparsec, roughly 8.4 percent. In ordinary life it may sound slight. In a science built on extraordinary precision, it is large enough to invite long attention.
The universe has become easier to see, and harder to measure. To understand why, you can drift back to a time when astronomers did not yet know whether the Milky Way was the whole universe.
In the early twentieth century, faint spirals seen through telescopes were called nebulae. They were hazy whorls, beautiful but uncertain. Perhaps they were nearby clouds of gas within the Milky Way. Perhaps they were separate stellar systems, unimaginably distant, what some called island universes. On April 26, 1920, at the Smithsonian’s National Museum of Natural History in Washington, D.C., Harlow Shapley and Heber Curtis publicly presented opposing arguments. Shapley held that the Milky Way comprised the universe and that spiral nebulae belonged within it. Curtis argued that the spirals were distant systems of stars in their own right.
Neither man stood outside the limits of his era. Evidence was incomplete, and some observations that appeared persuasive later proved mistaken. Claimed rotation rates for the spirals, for example, had been affected by error. The question was not settled by confidence alone. It waited for a new way to measure distance.
That way had been prepared quietly by Henrietta Swan Leavitt at Harvard College Observatory. In 1912, while studying variable stars in the Small Magellanic Cloud, she recognized a remarkable relationship among Cepheid variables. These stars brighten and fade in a regular pulse. Leavitt found that Cepheids with longer pulsation periods were intrinsically brighter than those with shorter periods.
Because the Cepheids she compared were all in the same small companion galaxy, they were approximately the same distance from Earth. Their differing apparent brightnesses could therefore reveal real differences in luminosity. From this came an extraordinary possibility. If you observed the rhythm of a Cepheid’s pulse, you could estimate how luminous it truly was. If you then compared that true luminosity with how faint it appeared from Earth, you could estimate its distance.
A stellar heartbeat had become a measuring tool.
At Mount Wilson Observatory above Los Angeles, the 100-inch Hooker telescope, completed in 1917 and then the largest in the world, offered a way to use that tool across immense distances. Edwin Hubble worked there, along with Milton Humason, an exceptionally skilled observational astronomer. On the night of October 5, 1923, Hubble photographed a star in the Andromeda nebula. At first he marked it with an N, for nova, a star that suddenly flares in brightness. But when he checked archival photographs, he found that the object had brightened and faded repeatedly, with a period of 31.4 days. It was a Cepheid. In red ink, he crossed out the N and wrote, with understandable excitement, “VAR!”
The Cepheid showed that Andromeda was far beyond the Milky Way. Hubble’s first estimate placed it at about 930,000 light-years away. That was much too small by modern understanding; Andromeda is about 2.5 million light-years distant. Still, the measurement was large enough to change the human map of existence. The spiral nebula was not a nearby cloud. It was another galaxy.
There is something restful in the fact that this turning point was itself imperfect. The universe did not surrender its scale all at once. The measurement that expanded humanity’s idea of the cosmos would later be recalibrated, corrected, and placed inside a larger story.
At the same time, another clue had been gathering in starlight. When a source of light recedes, its spectral features shift toward longer, redder wavelengths. Vesto Slipher had measured redshifts for many spiral nebulae before Hubble’s famous work on their distances. In 1927, Georges Lemaître connected galaxy redshifts and distances with solutions to Einstein’s general relativity that described an expanding universe. The relation now often bears the name Hubble–Lemaître, a gentler reminder that science is usually a woven fabric of observation, theory, instruments, and many patient minds.
In 1929, Hubble published an observed relation between the distances of galaxies and their radial velocities. More distant galaxies, on average, appeared to recede faster. The underlying relation endured. His numerical estimate of the constant did not. It was near 500 kilometers per second per megaparsec, far above the modern range between about 67 and 73.
This was not a simple blunder. The distance ladder beneath the calculation was still unfinished. Some objects treated as brilliant individual stars were actually glowing regions where many stars were forming. Cepheids themselves came in distinct stellar populations with different luminosities. As these and other matters became clearer, astronomers lowered the estimated expansion rate. A recalibration associated with Allan Sandage, Milton Humason, and Nicholas Mayall in 1956 brought a value near 530 down to about 180. Later work lowered it further.
For decades, astronomers spoke of a quieter dispute, sometimes called the 50-or-100 question. Was the Hubble constant closer to 50 or 100 kilometers per second per megaparsec? The answer mattered because expansion rate is bound to cosmic history. A larger present-day rate corresponds to a shorter rough reciprocal timescale, called the Hubble time, and a smaller rate to a longer one. But even this is only a guide, not the age of the universe itself, because the rate of expansion has not been constant.
Gradually, the question narrowed. The modern nearby measurement is often called the cosmic distance ladder. The word ladder can make it sound precarious, as though the entire result balances on one narrow rung. In practice, its strength comes from overlapping methods, repeated calibrations, and ways of checking one distance against another.
The first rung is geometry. As Earth moves around the Sun, a nearby star appears to shift very slightly against much more distant stars. This is parallax. The angles are tiny. Measuring Cepheid parallax is like seeing a grain of sand from about a hundred miles away. Yet from these shifts, astronomers can establish distances without assuming how bright a star ought to be.
There are other geometric anchors as well. Detached eclipsing binary stars in the Large Magellanic Cloud reveal their distances through orbital geometry and stellar physics. In the galaxy NGC 4258, a disk of water vapor circles a central black hole, producing microwave emission known as a megamaser. The shape, motion, and timing of that orbiting disk offer another geometric distance. These are not merely preliminary details. They are the firm ground from which the farther steps are taken.
With nearby Cepheids calibrated by such anchors, astronomers can use their periods and apparent brightnesses to measure distances to other galaxies where individual Cepheids can still be resolved. This requires care around dust, which dims and reddens starlight; around chemical composition, often called metallicity; around the exact calibration of a telescope’s detectors; and around neighboring unresolved stars whose blended light may make a Cepheid appear brighter than it truly is. In this story, such details are not distractions from the grand question. They are the grand question made tangible.
Some of the galaxies containing Cepheids have also hosted Type Ia supernovae. These are stellar explosions with light curves that can be standardized. They are not perfectly identical candles, but their brightness can be corrected using the shape and color of their fading light. Once a Type Ia supernova has occurred in a galaxy whose Cepheid distance is known, astronomers can determine the supernova’s true luminosity. Then similar supernovae can carry the calibration vastly farther outward, into the Hubble flow, where the broad expansion of space outweighs the smaller local motions caused by nearby gravity.
The Hubble Space Telescope, launched in 1990, was designed in part to sharpen this work. From above Earth’s blurring atmosphere, it could resolve Cepheids in more distant galaxies. Hubble’s Key Project studied about 800 Cepheids in 18 galaxies and, in its final late-1990s result, found a Hubble constant around 70 kilometers per second per megaparsec, with an uncertainty then near 10 percent. That did not end the work, but it transformed a broad argument into a precision measurement problem.
A major modern Cepheid–supernova analysis by the SH0ES collaboration reported in 2022 a value of 73.04 plus or minus 1.04 kilometers per second per megaparsec, including its estimate of systematic uncertainties. The analysis used 42 calibrator Type Ia supernovae and tested many variations: different geometric anchors, treatments of dust and metallicity, models for Cepheid behavior, choices of supernova samples, and local galaxy flows. Their result is not a casual number. It is the endpoint of a long chain of stars, galaxies, explosions, careful comparisons, and deliberate attempts to find where a measurement might be bent.
Then, with a long and gradual widening of the view, there is another route altogether.
Instead of beginning with nearby stars, this route begins with light released when the universe was only about 380,000 years old. Before that time, the cosmos was a hot plasma of free electrons, protons, and photons. Light scattered constantly from the free electrons, and the universe was opaque. As expansion and cooling continued, electrons joined with nuclei to form neutral atoms. Photons could at last travel long distances without repeated scattering.
That ancient light remains around us as the cosmic microwave background, cooled by expansion to a temperature of about 2.7 kelvin. It is nearly uniform in every direction, yet not perfectly so. Its subtle variations trace tiny differences in density in the young universe, differences that over billions of years would help seed galaxies, galaxy clusters, and the filamentary structure of the cosmos.
The young plasma also carried pressure waves, often called sound waves, passing through matter and radiation. Their characteristic scale became imprinted in the microwave background and, much later, in the large-scale distribution of galaxies. These patterns offer a standard ruler, not carved from wood or metal, but established by the physics of a very young universe.
The Planck satellite measured this microwave sky with extraordinary precision. Its final cosmological analysis measured the angular scale of the ancient acoustic pattern to 0.03 percent. Under the standard spatially flat six-parameter model known as Lambda Cold Dark Matter, or Lambda CDM, Planck inferred that the Hubble constant today should be 67.4 plus or minus 0.5 kilometers per second per megaparsec.
The word inferred matters. Planck did not watch galaxies today and directly read off their recession rate. It measured the early universe in exquisite detail, then used Lambda CDM, a model that has successfully described a great deal of cosmological evidence, to carry those early conditions forward to the present. The Hubble tension is therefore not a contest between one telescope and another. It is a disagreement between a late-universe distance ladder and an early-universe, model-based extrapolation.
Other observations of the large-scale universe tend to support the lower value within the same standard assumptions. The first-year results from the Dark Energy Spectroscopic Instrument, DESI, used more than six million extragalactic objects across redshifts from 0.1 to 4.2. With specified early-universe inputs, its baryon acoustic oscillation measurements found 68.52 plus or minus 0.62 kilometers per second per megaparsec. Combined with information from Planck and another microwave-background experiment called ACT, the result was 67.97 plus or minus 0.38.
So the two careful routes remain separated. At 100 megaparsecs, about 326 million light-years, a rate of 73.04 corresponds to a recession speed near 7,304 kilometers per second. A rate of 67.4 gives about 6,740. Both describe an expanding universe. The question is which precise rate best tells the story of today, and whether the bridge between the early and late universe is complete.
There are several calm possibilities. A remaining systematic effect may lie somewhere in the local ladder: blended starlight around Cepheids, dust, telescope calibration, stellar populations, the standardization of supernovae, or subtle local motions of galaxies. These possibilities are not neglected; they are examined repeatedly because precision requires skepticism that is patient rather than suspicious.
Or the early-universe inference may be limited by the assumed model. If some unrecognized physics in the young universe changed the scale of the primordial sound horizon, the inferred value of today’s Hubble constant could shift. Early dark energy, additional relativistic particles, altered recombination physics, and modified gravity have all been explored. They remain hypotheses, not discoveries, and simple extensions that partially ease the discrepancy have not generally been favored by Planck’s data.
There may also be a more subtle mixture of assumptions, statistical fluctuations, and correlations among data sets. The difference has sometimes been described as about five sigma in comparisons between the SH0ES result and Planck’s Lambda CDM prediction. That phrase means the gap is much larger than the quoted random uncertainties would ordinarily suggest, if the assumptions and estimates of systematic error are all correct. It does not mean that the universe has announced a verdict.
The James Webb Space Telescope has entered this patient investigation as a sharper infrared eye. In 2024, a SH0ES-led study compared more than 1,000 Cepheids in NGC 4258 and five supernova-host galaxies, using Hubble and Webb observations. It found no significant average distance difference between the two, reporting minus 0.01 plus or minus 0.03 magnitude. This strongly tested the concern that distance-dependent crowding in Hubble images could account for the tension. It did not prove that every possible systematic effect in the local ladder has disappeared, but it closed off one important path.
Webb has also made clearer comparisons among different stellar indicators. The Chicago–Carnegie Hubble Program has used Cepheids, the tip of the red giant branch, and a class of carbon-rich stars called J-region asymptotic giant branch stars, or JAGB stars. In a 2024 status report, its values ranged from about 67.96 for JAGB stars to 72.05 for Cepheids, with the red giant branch result near 69.85 and uncertainties broader than those of the flagship SH0ES measurement. Its combined estimate lay near 69.96, with statistical and systematic uncertainties of a little over one kilometer per second per megaparsec each.
This is not one group silencing another. Different indicators, samples, anchors, and treatments of uncertainty are being compared with increasing clarity. The purpose is not to protect a favored number. It is to discover how confidently each rung of every path can bear weight.
There are paths that do not use the traditional ladder at all. A distant quasar can be gravitationally lensed into multiple images by a massive galaxy between it and us. Since the light paths have different lengths and pass through different gravitational terrain, their variations arrive at slightly different times. Those time delays can reveal cosmic distances, although the mass distribution of the lens must be modeled carefully.
Gravitational-wave mergers offer another possibility. Their waveforms can provide a distance directly, making them standard sirens rather than standard candles. If a host galaxy’s redshift is known, or if galaxy information can be used statistically, the distance and recession information can help estimate the Hubble constant. The neutron-star merger GW170817, observed on August 17, 2017, gave a broad result near 70. More recently, a LIGO–Virgo–KAGRA analysis based on 236 gravitational-wave detections reported 71.0, with an uncertainty still wide enough to remain compatible with both the higher and lower sides.
Perhaps that is the deepest comfort in this small numerical tension. It is not a single ominous crack across the sky. It is a carefully mapped border between what can be measured nearby, what can be inferred from the first light of the universe, and what remains quietly uncertain between them.
The Cepheids still pulse in distant galaxies, keeping time with the regularity that Henrietta Leavitt learned to read. Supernovae brighten and fade. Ancient microwave light continues to arrive from a universe only 380,000 years old. Galaxies trace the old sound waves across immense reaches of space. Every method carries its own kind of memory, and every measurement asks to be checked by another.
Somewhere between the heartbeat of a star and the afterglow of the beginning sits one small number that humanity is still learning how to read. You do not need to resolve it tonight. The universe has had billions of years to unfold, and its remaining questions can rest, for now, in the dark beyond the window, expanding gently into stillness.