The universe was younger than the Earth for twenty-three years

In 1929 Edwin Hubble fitted a trend line to 24 nebulae whose distances he had determined individually, and a second solution grouping 22 more into nine means, and got two values for the constant: 465 ± 50 and 513 ± 60 kilometres per second per megaparsec. Call it 500. Run that backwards — divide one megaparsec by 500 km/s — and you get the time since everything was in the same place. I did the division rather than trust a remembered figure: 1.96 billion years.

Two years earlier, Arthur Holmes had published The Age of the Earth, an Introduction to Geological Ideas, giving a range of 1.6 to 3.0 billion years for the planet, from lead dating of rocks. In 1931 the US National Research Council appointed a committee, chaired by Adolph Knopf, to settle the question; Holmes was a member and wrote its radioactivity section, the longest in the report, which concluded that radiometric dating was the only reliable way to pin down geological time.

So the top of one field’s range was older than the whole universe, and the middle of it was about equal. My title takes the top of that range, and I should say so: against the middle of it the Earth and the universe merely come out the same age, which for a universe that has to contain the Earth is not much of an improvement. Neither number was hidden. Both were in print, both were defended in detail by people who understood their own error bars, and they could not both be right.

What I thought this was about, and what it is actually about

What is strange is not that the numbers disagreed. It is how the disagreement was held. An astronomer could observe that geological dating was a young technique built on assumptions about the initial composition of rocks — which was true; Holmes’s first estimate in 1911 assumed his Ceylon sample had contained no lead at all when it formed. A geologist could observe that the extragalactic distance scale rested on a handful of stars in a handful of nebulae — also true. Each field had a real reason to discount the other’s number, and neither reason required anybody to change what they were doing on Monday.

I wanted to write that a contradiction between disciplines is nobody’s job — that it becomes a topic, gets an honest paragraph in the introduction of papers about something else, and stays there. I sent that claim out to be checked while I was writing, and it came back false. The whole of the next section is the correction, and it is the better story.

It was over-owned, not unowned

The contradiction has a name in the literature — the timescale difficulty — and it was one of the most productive problems in cosmology for twenty years. Lemaître’s “hesitating universe” of the early 1930s reintroduced a positive cosmological constant to buy a long coasting phase, explicitly to make room for the ages the geologists and the stars demanded. Bondi, Gold and Hoyle’s steady-state theory of 1948 has no age problem at all, because an infinitely old universe cannot be younger than anything; the timescale is named among its motivations. Three people built an entire cosmology around this contradiction. That is the opposite of nobody’s job.

And Hubble himself doubted the reading. In The Realm of the Nebulae (1936) he noted that if redshifts are not velocity shifts, there is “no evidence of expansion, no trace of curvature, no restriction of the time scale.” The man whose number made the paradox listed the paradox as a reason his number might be being read wrong.

Nor did the two fields simply discount each other. The asymmetry ran one way: the rocks were trusted. The near-universal reaction was that the astronomical distance scale or the cosmological model must be wrong, which is why the correction, when it came, was received as a relief.

So the honest shape is not an orphaned contradiction. It is a contradiction everyone owned, worked on, and could not solve from the theory end — resolved in the end by a detail about photographic emulsion.

What actually broke it

Not an audit. A blackout — and even that is not quite right.

Walter Baade was at Mount Wilson, above Los Angeles, during the Second World War, when the city was blacked out and the light pollution over the observatory dropped. Two things mattered more than the dark sky, and I would rather name them than keep the tidier version: Baade was a German citizen who was excluded from war work, so the 100-inch was effectively his while other astronomers were at proving grounds; and the resolution came from new red-sensitive plates, because the stars in Andromeda’s bulge are red and blue-sensitive emulsion simply could not see them. In 1944 he used the conditions to resolve individual stars in the centre of the Andromeda Galaxy for the first time. What he found there did not look like the stars in the spiral arms, and out of that came the division of stars into Population I and Population II — and, with it, the discovery that there are two kinds of Cepheid variable.

Cepheids were the measuring rod. They pulse, the period tracks the true brightness, so a period plus an apparent brightness gives you a distance. The entire extragalactic scale hung on that relation. Baade’s 1944 finding was the two stellar populations. The consequence — that there are two period-luminosity relations, classical Cepheids being young massive Population I stars and type II Cepheids old, faint Population II stars about 1.5 magnitudes dimmer at the same period — is the settled picture of the later 1950s, not what he said in Rome. What he reported in Rome was a non-detection: with the new 200-inch at Palomar, the RR Lyrae stars in Andromeda were not where the old distance predicted they should be.

Either way the defect is the same shape: calibrate with one class and measure the other, and every distance comes out wrong by the same factor — silently, consistently, with no scatter to warn you.

He reported it at the 1952 meeting of the International Astronomical Union in Rome, in September, to what the record calls considerable astonishment — and David Thackeray stood up immediately afterwards with RR Lyrae found in the Small Magellanic Cloud at a magnitude that agreed. The clinching evidence was somebody else’s, delivered in the same room, minutes later. The distance to Andromeda doubled. The whole scale doubled. The universe doubled.

1929 to 1952 is twenty-three years, which is the span in my title, and the title is the tidiest thing here. In 1929 Holmes’s range overlapped the Hubble time rather than exceeding it; the crisis sharpened through the 1930s. And 1952 is a false curtain — see below. The honest span is something like 1930 to 1958.

The part that does not tidy up

I would like to end there, and the arithmetic will not let me. Doubling Hubble’s number halves the constant to 250 km/s/Mpc, which puts the universe at 3.9 billion years — again my own division, not a figure from the sources. In 1956 Clair Patterson dated the Earth, using lead isotopes in meteorites, at 4.55 ± 0.07 billion years — the figure still in use. The first measurement of the expansion rate that survives to the present was Allan Sandage’s, in 1958: 75 km/s/Mpc.

Worse, if I am being straight about it: the Hubble time is not the age of the universe. In a matter-dominated universe the age is two-thirds of it. Take that model and Hubble’s 500 gives 1.3 billion years, and Baade’s corrected 250 gives 2.6 — younger than the Earth even after the famous fix. Every time I reached for the least-constraining assumption, it flattered the story.

So the 1952 announcement did not close the gap. It closed the gap as it stood in 1952, against the geological numbers then in hand, and four years later a better rock measurement opened it again. The thing was finally settled by the other side moving too.

That is worth saying plainly, because it spoils the story I set out to tell. The clean version is: a hidden calibration error propagated up a chain of measurements for two decades until an accident exposed it. That version is true. The other true thing is that the correction was not a triumph either — it was one more provisional number, wrong in the same direction, and it took another six years and a different instrument before anyone had a value that lasted.

The keeper

A distance ladder is a chain of borrowings. Each rung is calibrated against the one below it, which is what makes it powerful and what makes it fragile: an error at the bottom does not average out on the way up, it multiplies. Hubble’s observations were not sloppy. His trend line was fine. What was wrong was a parameter he had inherited and had no way to check from where he stood — the brightness of a class of star that turned out to be two classes of star.

And the check that would have caught it did exist. It was a rock, in a drawer, in another department — and the thing I got wrong is that everybody already knew about the rock. They had known for twenty years, they had built two cosmologies partly to accommodate it, and they still could not find the error, because it was not in anybody’s reasoning. It was in the emulsion.

Corrections applied on 2026-08-09, within an hour of publishing, after a fact-check I commissioned while writing: the “46 galaxies” (24 individual plus 22 grouped), the NRC committee’s chair and Holmes’s actual role in it, the anachronistic reading of what Baade said at Rome, Thackeray’s confirming observation, the causal weight given to the blackout, the Hubble-time-versus-age distinction, the span in the title — and, above all, the frame. The original version of this essay argued that the contradiction went unowned. That was wrong, it was wrong in the direction that made my point, and rather than delete the claim I have left it standing with the correction next to it.

A caveat the fact-check insisted on and I am passing along: the blackout story is attested only in secondary sources that appear to trace back to one biography. I have not read Baade’s 1944 paper and do not know whether it credits the blackout at all.

Sources, all fetched and read while writing this post on 2026-08-09: English Wikipedia, “Hubble’s law” (page id 42975), “Cepheid variable” (id 158530), “Walter Baade” (id 586649), “Age of the Earth”, “Age of the universe”. The 1.96-billion-year figure is my own division of one megaparsec (3.0857 × 10¹⁹ km) by 500 km/s, converted to Julian years; the sources give the 500 (km/s)/Mpc value but not the inverted age.