The Mutation Stopwatch: Dating a Split

The question we ended on

Last tutorial handed you a clock — and then, in the final line, told you it was broken for half the jobs you'll want to do.

Recombination, remember, is a metronome that chops. When two populations mix, the first child's genome is a few enormous unbroken blocks of pure-A and pure-B DNA. Every generation after, meiosis makes a few cuts, and those blocks get shorter and shorter at a steady rate. So the length of the surviving blocks tells you when the mixing happened. Long blocks, recent mix. Confetti, ancient mix. A clock.

Now ask the opposite question. Not "when did these two groups mix?" but "when did these two groups split apart and stop sharing DNA entirely?" Say, humans and Neanderthals. At some point, deep in the past, one population of ancient humans divided into two, the two halves wandered off, and for hundreds of thousands of years they had nothing to do with each other genetically.

Try to point the recombination clock at that. It just... shrugs. Recombination only chops blocks that exist because two ancestries got mixed together. But a split is the absence of mixing. There are no A/B blocks to measure, because after the split nobody's combining anything — the two lineages are just sitting in separate rooms, drifting apart. The chopping clock needs a mixture to chop. A split gives it nothing to bite on.

So we need a second clock. A completely different one — one that ticks not when populations mix, but when they're apart. One that runs during the long silence between two lineages.

That clock is mutation. And it's the second hand of the whole genetic timepiece — the thing under every "these two split roughly 500,000 years ago" number Reich ever says.

Two clocks, two jobs — don't mix them up

Hold both in your head at once, because keeping them straight is half the battle:

• Recombination clock (last tutorial): measures block length. Dates when two populations MIXED. Runs after a merger.
• Mutation clock (this tutorial): measures number of differences. Dates when two lineages SPLIT. Runs during a separation.

One clock times the smoothie; the other times the silence. Reich needs both, because his whole picture of the past is populations repeatedly splitting apart and crashing back together — a tree that keeps turning into a web. Recombination reads the reunions. Mutation reads the divorces. This chapter is the divorce clock.

Remember the copy-slip

To build it we need one ingredient, and you already own it — from tutorial three.

Every time a cell copies its DNA, it copies about three billion letters, and it's astonishingly accurate. But not perfect. Every so often it fumbles a single letter — a C where there should have been a T, a dropped rung, a typo. That's a mutation: a rare, random copy-slip that then gets passed down to every cell descended from that one. We met these as the origin of variation — the reason no two genomes are identical. The typos are why you're not a clone of anyone.

Here's the number that turns typos into a clock: when a parent passes their genome to a child, the child is born carrying roughly a few dozen brand-new mutations — letters that appear in the child that were in neither parent. A few dozen fresh typos, every generation, forever. Rare per letter, but relentless across the whole genome and across deep time.

Now — a few dozen random slips per generation. Random. So how on earth could something random keep steady time? That's the objection everyone raises, and answering it is the whole trick.

Why random still keeps time (the radioactive-decay trick)

Here's the resolution, and it's one of the most beautiful ideas in all of science, borrowed straight from physics.

Take a lump of a radioactive element — uranium, say. Ask: which specific atom will decay next, and when? Nobody can tell you. It's fundamentally random; a single atom might decay in a second or sit unchanged for a billion years. Utterly unpredictable.

Now ask a different question: in this lump of a trillion trillion atoms, what fraction will decay in the next hour? That, you can predict to staggering precision. Because when you have an astronomical number of independent random events, the individual randomness washes out and a rock-steady average rate emerges. This is exactly why carbon dating works: you can't predict one carbon atom, but a steady fraction of any large sample decays per millennium, so the amount left is a clock. Randomness in the small, iron regularity in the large.

Mutations are the same story:

You cannot predict which letter of the genome will mutate, or when — each slip is as random as a single atom's decay. But across three billion letters and thousands of generations, the rate of new mutations is remarkably steady. The genome is a big enough lump that the average holds.

So the "few dozen per generation" isn't a wobbly guess that swings from 3 to 300. It's a stable, measurable tempo — because it's an average over an enormous number of independent tiny lotteries. That steadiness is the tick of the clock. Random events, steady rate, exactly like the atoms in the rock.

*(One honest caveat the pros track: the tempo isn't perfectly identical everywhere or every era — it drifts a little with generation length, parental age, and genomic neighborhood. The molecular clock is a very good clock, not an atomic one. Reich's teams spend real effort calibrating it, which is the next piece.)

*

The clock itself: two lineages, drifting apart

Now assemble the stopwatch. It takes exactly three moves.

Move 1 — start them identical. The instant two populations split from one common ancestral group, their genomes are essentially the same. Zero differences that arose after the split. The clock reads zero.

Move 2 — let them drift independently. The two lineages are in separate rooms now. Each one keeps copying its DNA generation after generation, and each one independently collects its own few-dozen-per-generation typos. Crucially, lineage A's new typos land in different random spots than lineage B's — they're not sharing, so they never coordinate. The set of differences between the two genomes grows and grows.

Move 3 — count the differences. Take one genome from lineage A and one from lineage B, line them up letter by letter, and count the positions where they disagree. That pile of disagreements is nothing but accumulated mutations since the split — the sand that's dropped through the hourglass since the two rooms were sealed.

And because mutations drip at a steady rate, the count is proportional to time:

More differences → longer since the split. Fewer differences → more recent split. Count the letters where two lineages disagree, divide by the mutation rate, and you get the number of generations since they had a common ancestor. Multiply by years-per-generation, and you get a date.

That's the entire mutation clock. It's almost embarrassingly simple once you see it: differences ÷ rate = time. Where recombination reads the length of blocks, mutation reads the number of typos — and both are just steady processes whose accumulation you can run backwards into a date.

But how fast does it tick? (calibration)

There's one gap left, and it's the gap that separates a real science from arm-waving. "Divide by the mutation rate" — fine, but how do you know the rate? A clock is useless if you don't know how fast the hand sweeps. Getting that number is called calibration, and there are two beautiful ways to do it.

Way one — watch it directly (the trio method). This is the modern, brute-force answer, and it only became possible when sequencing got cheap. Take a mother, a father, and their child, and sequence all three genomes completely. Now scan the child for letters that appear in neither parent. Every such letter is a brand-new mutation, born in this one generation, caught in the act. Count them, and you've measured the per-generation mutation rate directly, with your own eyes — no assumptions, no fossils. When people say "roughly a few dozen new mutations per generation," that number came from literally counting them in family trios.

Way two — anchor to a known date (fossil calibration). The older, clever answer. Suppose you already know from fossils and geology that two species split at some particular time — the classic anchor is humans and chimpanzees, whose split is dated by fossils to roughly six to seven million years ago. Now count the genetic differences between a human and a chimp. You know the differences, and you know the time — so you can solve for the rate: differences ÷ time = rate. Once you've pinned the rate against that one fossil-dated anchor, you can turn it around and use it to date any other split for which you have no fossils at all.

Two independent roads — counting live mutations in families, and anchoring to fossil-dated splits — and the fact that they roughly agree is a big part of why anyone trusts the clock. Historical aside worth savoring: the whole idea that molecules keep time like this was dreamed up in the early 1960s by Emile Zuckerkandl and Linus Pauling, who noticed that the number of differences in the hemoglobin protein between species lined up neatly with how long ago those species had diverged. They coined the phrase "molecular clock," and half a century later it's the backbone of every date on the tree of life.

The payoff: dating the Neanderthal split

Now point the finished stopwatch at the question we opened with.

We have a complete Neanderthal genome — recovered from a ~40,000-year-old fossil bone, sequenced letter by letter. We have plenty of living human genomes. Line them up, count the letters where Neanderthals and living humans systematically disagree, divide by the mutation rate we calibrated above, convert generations to years — and out drops a number: humans and Neanderthals last shared a common ancestral population somewhere around 500,000 to 700,000 years ago.

Sit with how astonishing that is. Nobody dug up a labeled receipt. Nobody watched it happen. We took two genomes, counted the typos that had accumulated in each since they parted ways, divided by a tempo we'd measured in modern families and chimp fossils, and reconstructed a moment more than half a million years deep — the moment one ancient population became two.

And it doesn't stop at Neanderthals. Every branch point on the tree of human populations gets its date this same way. When did Denisovans split from Neanderthals? Count differences, divide, done. When did the ancestors of one population diverge from another? Same clock. The mutation stopwatch is what turns a bare shape — this group is related to that one — into a dated history with real years on every fork. It builds the timeline that the recombination clock then annotates with all the later reunions.

The fine print (one honest subtlety)

One wrinkle, because you're ready for it and because it shows up constantly on the podcast — and skipping it would leave you with a slightly-too-clean picture.

When you count differences and get a date, that date is technically the age of the most recent common ancestor of the two DNA sequences — not exactly the moment the two populations split. And those two things are subtly different, because the ancestral population already carried variation before it split. Two copies of a gene can trace back to a shared ancestor that lived comfortably before the population itself divided.

So the genetic date is always a little bit older than the true population split. The genes diverge first; the population catches up later. (The technical name for "when two sequences trace back to their common ancestor" is coalescence — worth knowing, because when Reich distinguishes "when the genes split" from "when the populations split," this gap is exactly what he means.) You don't need to compute it — just know the clock reads a hair early, and that the pros correct for it. File it; it's a load-bearing distinction later.

Cashing it against the podcast

Go back to the podcast with both clocks in hand, and listen to how naturally the two kinds of dates now sort themselves:

• "Neanderthals and modern humans split around half a million years ago" — mutation clock. Count the differences between the two genomes, divide by the rate. A split, dated by accumulated typos.
• "South Asians are a mixture from 4000 to 2000 years ago" — recombination clock. Measure the length of the ancestry blocks. A mixture, dated by how far the chopping has gone.

Every date Reich gives is one of these two clocks, and you can now hear which is which just from the verb. "Split," "diverged," "separated" → the mutation stopwatch, counting differences. "Mixed," "admixed," "merged" → the recombination metronome, measuring block length. Two clocks, two verbs, two jobs. With both, a genome stops being a static string of letters and becomes a dated diary — every split has a year, every reunion has a year, and you can read the whole itinerary of a people out of the DNA of one living person.

The challenge

Three questions, escalating:

1. Warm-up. In one sentence: why is the recombination clock (block length) useless for dating the human–Neanderthal split, forcing us to reach for the mutation clock instead?

2. The real one. Mutations happen at random spots in the genome — you can never predict which letter will slip next. So how can something so random be used as a precise clock? Give the reasoning, and name the physics phenomenon it's exactly analogous to. (Hint: think about a lump of uranium versus a single atom.)

3. Think ahead. Two populations, X and Y. You count the differences between an X genome and a Y genome and get a certain number. Your colleague reminds you that this number dates the split of the DNA sequences, which is older than the split of the populations. Why are those two dates different — where does the extra age come from? (You've just met coalescence, and you're standing at the door of the next idea: what happens to all this variation when a population stays small and isolated for a long time — the slow, powerful force of genetic drift.)

Next: a short, deliberate detour before we push on. You've built genetics molecule-first for six chapters — now meet **Mendel, the monk who deduced the rules of inheritance in 1865 with no molecules at all. It's the chapter that names the classical-vs-molecular divide and finally cashes the "allele" IOU from tutorial five. Then, right after, the force that quietly governs both clocks: **genetic drift* — how, in a small or isolated population, pure chance can make some variants vanish and others take over, no natural selection required. Drift is how populations become distinct in the first place — and it's what finally lets "these two groups are genetically different" mean something precise. That's the road ahead.*