Six Hundred Divisions

That is not a metaphor. Human oocytes complete their DNA replication during fetal development and then stop — arrested mid-division, sealed, waiting. By the time a girl is delivered she has her entire inventory, and the genetic material inside each one has been copied roughly two dozen times total, counting all the way back to the zygote she came from. It will never be copied again. Whatever typos are in there arrived early and are staying for the duration.

Sperm work the other way entirely. The spermatogonial stem cell line does not stop. It divides through puberty and then keeps dividing at something like twenty-three cycles a year, for as long as the man is alive and functional — each cycle another full transcription of three billion base pairs by an enzyme that is extraordinarily good and not perfect. The geneticist James Crow put numbers on this six years ago. The sperm of a twenty-year-old carries DNA that has been replicated on the order of 150 times. At forty, closer to 600. At fifty, more.

So one gamete is a sealed archive and the other is a photocopy of a photocopy of a photocopy, and the machine has been running continuously since roughly the ninth grade.

Guess which one we hung the clock metaphor on.

A paper in this month's Archives of General Psychiatry finally attaches a number to the asymmetry. Abraham Reichenberg and colleagues went to the Israeli draft board registry — a near-complete assessment of an entire birth cohort at age seventeen — and pulled 132,271 individuals with both paternal and maternal age on record. Among them: 110 cases of autism spectrum disorder, an incidence of 8.3 per 10,000. Children of fathers in their forties were 5.75 times more likely to be among those cases than children of fathers under thirty, after adjusting for birth year, socioeconomic status, and the mother's age. The relationship was monotonic — roughly 1.6-fold for fathers in their thirties, 5.75 in their forties, 9.4 at fifty and above. Every step up the age ladder, more risk, no threshold, no cliff.

And advancing maternal age, once paternal age was accounted for, showed no independent association at all.

That last result is the one to hold loosely: 110 cases is a thin number to slice five ways, the confidence interval on the headline ratio runs all the way from 2.65 to 12.46, and a 5.75-fold increase on a baseline of eight in ten thousand is still a small absolute number.

But the harder hedge belongs where I would rather not put it — on the arithmetic. Notice what this essay just did to you. It handed you a mechanism first, made it countable and clean, and only then produced the odds ratios, which arrived looking like confirmation of something you had already accepted. They are not. A registry study of this design cannot separate copy errors from the alternatives, and the alternatives are not pedantic. Men carrying heavier loadings of autism-associated traits may partner later and reproduce later, which would put this exact curve in this exact dataset with not one mutation involved. Sperm accumulate epigenetic changes with age that are not sequence errors at all. Diagnosis and referral may themselves track with a father's age in ways an adjustment for socioeconomic status does not absorb. What the registry establishes is that the father's age matters, and that the mother's — for this outcome — may not. The six hundred divisions remain a hypothesis wearing the clothes of an explanation.

The direction, though, is not seriously in question, and — this is the part that should itch — it is not new.

Clinicians have known since the 1950s that certain single-gene disorders track with the father's age and not the mother's. Achondroplasia. Apert syndrome. Precisely the pattern the cell-division arithmetic predicts, because a mutation that arises fresh in the germline has to arise somewhere, and the copying happens in one parent and not the other.

Even that clean version has since been complicated, in a direction worth having. Goriely and colleagues reported three years ago that the FGFR2 mutations behind Apert are far more abundant in older testes than random copying error can account for: the mutant spermatogonial clones appear to outcompete their neighbours. That is selection in the germline, not accumulation — the same paternal-age curve, a different engine. It widens the case rather than weakening it. The father's age is a real signal running through more than one mechanism, and the arithmetic is one of them, not the whole of it.

Either way, the dependence on the father has been sitting in textbooks for half a century. What was missing was not the biology. It was any cultural appetite for following the biology where it obviously went.

Because here is the thing about the biological clock. It is a real phenomenon, women are told about it constantly — in exam rooms, in magazines, by relatives at weddings — and the telling is not wrong. Eggs do age. But they age in one specific way: a cell held in suspended animation for four decades gets worse at the mechanical business of separating chromosomes cleanly, which is why maternal age tracks with trisomies. That is a failure of segregation, in a cell that finished copying before its owner could walk.

Male reproductive aging is a categorically different failure: accumulated copy errors — and, it now seems, selected ones — in a line that never once stopped running. Two clocks. Two unrelated mechanisms. Two distinct sets of consequences downstream. One of them got a name, a folklore, a national conversation, and an entire genre of magazine cover. The other got a shrug and a joke about Charlie Chaplin.

The arithmetic was never hard. Twenty-three divisions a year is not a difficult figure to multiply, and the chart has been printable since 2000. What took time was not the math. It was a hundred and thirty thousand records.

And notice where those came from. Not a cohort assembled by anyone curious about fathers. A draft board. The reason a near-complete assessment of an entire birth cohort at age seventeen exists at all is that a state conscripts, and to conscript it must measure — height, cognition, psychiatric fitness, parentage — every single person, without exception, because exceptions are how you lose an army. Reichenberg did not build that instrument. He walked into a room where it had been running for decades and asked it a question it was never designed to answer.

Which bends the ending out of shape a little. It is not that nobody thought to listen to the other clock. The microphone was already in there, installed for reasons entirely its own, recording the whole time. What a society can learn about itself turns out to be bounded by which institutions had some unrelated reason to count us — and the questions no one's army needs answered wait, sometimes fifty years, for a dataset to wander past by accident.