Meiosis & Recombination: The Shuffle

The question we ended on

Last tutorial closed with a puzzle I want you holding the whole way through. You got 23 chromosomes from your mother and 23 from your father — that's how you ended up with 46, kept as 23 pairs. Fine.

But your mother has 46 chromosomes of her own. So when she made the single egg cell that became you, she had to somehow pack exactly 23 of them into it — not 46, not 24, exactly half. And here's the strange part I slipped in at the very end: each of those 23 chromosomes she handed you isn't a clean copy of one of her chromosomes. It's a spliced-together mosaic of the chromosome she got from her mother and the one she got from her father.

Two questions, then, and this whole chapter is the answer:

1. How does a parent with 46 make a cell with exactly 23?
2. Why is each chromosome that gets passed down a patchwork, not a clean copy?

The answers are one process called meiosis and one event inside it called recombination. And they are not molecular trivia — recombination turns out to be the single most important tool David Reich uses to read history out of a genome. By the end of this you'll understand how he dates the moment two ancient populations mixed — not roughly, but to within a few thousand years. Let's earn it.

First, why halving is not optional

Start with a counting problem, because it makes the whole thing feel inevitable.

Suppose parents just handed down all their chromosomes. Mum gives 46, dad gives 46 — the child has 92. That child grows up, partners with someone who also has 92, and their kid has 184. The next generation: 368. Within a handful of generations the number explodes into nonsense, and cells that are mostly made of chromosomes have no room to actually do anything.

That's obviously not what happens — humans have had 46 chromosomes for a very long time and the number stays put. So there must be a rule, and the rule is forced by simple arithmetic:

The cells that make babies — eggs and sperm — must carry exactly half the normal number. 23, not 46. So that when egg meets sperm, 23 + 23 = 46, and the count resets perfectly every generation.

Those half-sized cells have a name: gametes (egg and sperm). Every other cell in your body has the full 46 and is called diploid — "two sets." Gametes have 23 and are haploid — "one set." The entire job of the special division we're about to meet is to turn one diploid cell (46) into gametes (23) without botching the count.

That division is meiosis. (Don't confuse it with ordinary cell division — mitosis — which copies everything and makes two identical 46-chromosome cells, the kind of division you did when you grew from a single cell. Meiosis is the rarer, weirder one, reserved for making gametes, and it does something mitosis never does.)

Meiosis: how you get from 46 to 23

Here's the trick, and it has a counterintuitive shape: to halve the chromosomes, the cell first doubles them, then divides twice.

Walk through it slowly, because the "double first" step trips everyone up:

1. Copy everything. Just before dividing, the cell duplicates all 46 chromosomes — exactly the "copy the whole library before splitting" step from last tutorial. Each chromosome is now its familiar tight X: two identical copies joined at the waist.
2. Pair up the partners. Now the magic: each chromosome finds its matching partner — mum's chromosome-1 lines up right alongside dad's chromosome-1, mum's 2 beside dad's 2, all 23 pairs standing in formation. (This lining-up is what mitosis never does, and it's what makes everything else possible.)
3. First division — split the pairs. The cell pulls the pairs apart: mum's chromosome-1 goes one way, dad's chromosome-1 the other. Now each daughter cell has 23 chromosomes (still doubled — still X-shaped). The count is already halved.
4. Second division — split the X's. Each of those cells divides again, this time pulling apart the two halves of each X. The result: up to four gametes, each with a clean set of 23 single chromosomes.

One diploid cell (46) → four haploid gametes (23). Count preserved for the next generation. That's meiosis.

The first source of shuffle: 23 coin-flips

Now look closely at step 3, because something quietly momentous happens there.

When the 23 pairs get pulled apart, which member of each pair ends up in a given gamete — mum's or dad's — is independent for every pair. Pair 1 might send mum's copy; pair 2 might send dad's; pair 3 mum's again — it's like flipping 23 separate coins, one per chromosome.

How many different gametes can that produce? Two choices, 23 times over: 2²³ ≈ 8.4 million distinct combinations — and that's before the second shuffle we haven't met yet. This is the first reason no two siblings (bar identical twins) are the same: each of your parents can produce millions of genetically different eggs or sperm, and you are one specific egg meeting one specific sperm out of an astronomical lottery.

But notice what this shuffle does not do. It deals out whole, intact chromosomes — you get mum's entire chromosome-1 or dad's entire chromosome-1, never a blend of the two. If coin-flips were the whole story, every chromosome you pass down would be a pristine copy of one you got from a single parent. And that would leave our second question — why is each passed-down chromosome a mosaic? — completely unanswered. Something else has to be going on.

The second source of shuffle: crossing-over

Rewind to step 2 — the moment the partners line up side by side. While they're pressed together, something physical happens: the paired chromosomes swap chunks. Mum's chromosome-7 and dad's chromosome-7 lie alongside each other, break at one or more matching points, and trade the pieces — a length of mum's strand splices into dad's, and vice versa. This literal cut-and-swap is called crossing-over, or recombination.

The consequence is exactly the mosaic we've been chasing:

After recombination, the chromosome that ends up in a gamete is no longer "mum's whole chromosome-7" or "dad's whole chromosome-7." It's a stitched patchwork — a stretch of mum's, then a stretch of dad's, then mum's again — spliced together at the crossover points.

So the chromosome-7 you inherited from your mother is itself a mosaic of her mother's and her father's chromosome-7. That's the answer to the puzzle we opened with. Every chromosome you carry (except two — hold that thought) is a quilt, and each patch traces back to a different grandparent, great-grandparent, and beyond. Your genome isn't a deck of cards dealt whole from your ancestors; it's a deck that gets cut and reshuffled every single generation.

Why the Y and mtDNA are special — now it's obvious

Remember the two "clean, unshuffled lineages" from last tutorial — the Y chromosome (father → son) and mitochondrial DNA (mother → child)? I promised recombination would explain why they escape the shuffle. Here it is, and it falls out for free:

Recombination needs a partner to swap with — two matched chromosomes lying side by side. But the Y has no real partner: it pairs with the X, and the two are so different (the Y is a short runt) that they barely swap at all. And mitochondrial DNA sits outside the nucleus entirely, in its own little power-plant, with no partner and no meiosis happening to it at all. No partner, no crossing-over, no shuffle. So the Y and the mtDNA ride down the generations essentially un-mixed — which is exactly what makes them clean single-line tracers of your paternal and maternal ancestry. The rest of your genome is a blender; these two are a straight copy. Same fact as last tutorial, but now you can see the mechanism that makes it true.

The payoff: recombination is a stopwatch

Here's where a molecular detail becomes a time machine — the reason this is the keystone tutorial of the whole series.

Picture two populations that have lived apart for a long time — say the ancestors of group A and group B — different enough that you could look at a chunk of DNA and tell which group it came from. Now they meet and mix. What does the very first mixed child's genome look like?

Long, unbroken blocks. That first child got a whole chromosome from an A parent and a whole chromosome from a B parent (via a gamete each). Their DNA is a few enormous stretches of pure-A and pure-B, side by side.

Now let generations pass. Every generation, meiosis runs, and recombination makes a few cuts along each chromosome — splicing A-segments and B-segments together and, crucially, chopping the long pure blocks into shorter and shorter pieces. After one generation the A/B chunks are still huge. After ten generations they're noticeably shorter. After a hundred generations they're chopped into tiny confetti. The blocks only ever get shorter, and they shorten at a roughly steady, known rate.

So flip it around, and you have a clock:

Measure the average length of the ancestry blocks in a living person's genome. Long blocks → the mixture was recent (few cuts have happened yet). Short blocks → the mixture was ancient (many generations of chopping). Segment length dates when the two populations mixed.

That's it. That's the trick. Recombination is a metronome that chops at a steady tempo, and the length of the surviving chunks is your read-out of how many beats — how many generations — have passed since the mixing event.

Cashing it against the podcast

Now go back to the things Reich says on the podcast that sounded like unsupported magic, and watch them turn into arithmetic you understand:

• "South Asians are the result of mixture between groups very different from each other — as different as Europeans and East Asians — **4000 to 2000 years ago, and then crystallizing into a relative lack of mixture since that time."
• "Europeans are the result of mixture of Yamnaya and farmers and hunter-gatherers."

How could anyone possibly know a mixture happened 4000 to 2000 years ago and not, say, 20,000? He didn't find a dated receipt. He measured the length of the ancestry segments in living South Asian genomes, saw how far recombination had chopped them down, counted backward the number of generations that much chopping takes, multiplied by ~28 years per generation — and landed on a date. The segments are still fairly long, which is exactly why the answer is "a few thousand years ago" and not "fifty thousand." A blender that runs at a known speed lets you measure how long the smoothie has been running.

Every time Reich says a population is "a mixture of X and Y from N years ago," the N almost always rests on this one idea: recombination chops ancestry blocks at a steady rate, so block length is a clock. You now own the machinery under his most quoted claims.

The challenge

Three questions, escalating:

1. Warm-up. Your mother has 46 chromosomes but put exactly 23 into the egg that became you. In one sentence, why does the number have to be halved — what would go wrong if she passed all 46?

2. The real one. Two brothers have the same two parents, yet their genomes are different (they're not identical twins). Name the two distinct shuffling mechanisms inside meiosis that make this true, and say what each one does — one deals out whole chromosomes, the other does something to the chromosomes themselves. (Hint: 2²³ is one of them.)

3. Think ahead. Population geneticists find that a certain group carries ancestry from two very different source populations, and the ancestry blocks are very short on average. Did those two populations mix recently or a long time ago? Explain your reasoning using recombination as a clock. (You've just re-derived, from scratch, how Reich dates an admixture event — and you're standing right next to the other clock we open next: not the chopping of blocks, but the steady drip of new mutations.)

Next: recombination dates mixture events by chopping blocks — but it can't tell you how long ago two whole species (say, us and Neanderthals) split, because after a split there's no mixing to chop. For that you need a different clock entirely: **mutations, the rare copy-slips from tutorial three, accumulating at a steady drip. Count the differences, divide by the rate, and you get the "500,000 years ago" dates. That's the mutation stopwatch — the second hand of the genetic clock.