Mixture, Not Trees: The Trellis

The tidy story you were just sold

Last chapter handed you a clean machine. Take one population. Split it in two — a mountain rises, a group walks off, a social wall goes up. The halves stop trading alleles. Each one drifts on its own luck. Their allele frequencies pull apart. Wait long enough and the two are genetically distinguishable.

Now do it again. And again. Split, drift, split, drift.

What have you just drawn?

A tree. One trunk, branches forking off, each fork wandering further from the others, never touching again. It's the diagram in every biology textbook you've ever seen. It's the diagram in your head right now, and you built it yourself two paragraphs ago out of parts you already owned.

It's also the single biggest thing David Reich's field had to demolish.

Here's the thing about the tree: for species, it's basically right. Lions and tigers split, and lion and tiger lineages do not, as a rule, grow back together. Real tree limbs don't fuse. Reich says so explicitly — a tree is a good analogy for species.

"...it is a dangerous analogy for human populations. The genome revolution has taught us that great mixtures of highly divergent populations have occurred repeatedly. Instead of a tree, a better metaphor may be a **trellis, branching and remixing far back into the past." — David Reich

Branching and remixing. The forks come back. Groups that drifted apart for thirty thousand years walk into each other and have children, and the branch that was supposed to keep diverging forever instead merges into another branch.

That's admixture. And this chapter is the hinge where the series stops building tools and starts using them — because everything you need to catch a mixture in the act is already in your hands.

The population that sits in the middle

Start with the clue that broke the tree open, and it's one that lands close to home: India.

When Reich's group first plotted Indian genomes against European and East Asian ones, Indian groups didn't cluster in their own tidy corner. They smeared out along a line — what his team called the Indian Cline. And critically, in Reich's words, their allele frequencies were "on average, intermediate between those in Europeans and East Asians."

Intermediate. In between. So — mixture, obviously. Case closed?

No. And this is the intellectual crux of the whole chapter, so slow down here.

There are two completely different histories that both produce a population with in-between allele frequencies:

Story A — mixture. Two populations drifted apart for tens of thousands of years, then met and had children. Their descendants carry alleles from both, so their frequencies land between the two sources. Trellis.

Story B — a middle branch. No mixture at all. Population X split off from the common trunk at a point between where Europeans and East Asians split, and just drifted there on its own. It never mixed with anybody. It's simply related to both, at a middling distance, and so it looks intermediate. Pure tree.

Both stories predict the same thing: a population sitting in the middle. So an intermediate allele frequency is not evidence of mixture. On its own, it proves nothing. If you'd stopped there, you'd have a hunch and a fight, not a finding.

You need something that Story B cannot fake.

The tiebreaker is inside your chromosomes

Here's the move, and it's beautiful because you already built the tool in T6.

Stop thinking about a mixed person as an average of two populations. That's the wrong picture entirely. A mixed person is not a blend, not a smoothie, not beige paint made from two colours.

A mixed person is a mosaic.

Run the first generation. A man from population P and a woman from population Q have a child. What does that child's genome look like? Each of their 23 chromosome pairs has one entire chromosome from P and one entire chromosome from Q. Not mixed at the letter level — mixed at the chunk level, in the biggest chunks physically possible. Perfect, unbroken, chromosome-length blocks.

Now let time run. That child grows up and makes gametes, and T6 tells you exactly what happens: meiosis, and crossing-over. The P chromosome and the Q chromosome pair up and swap segments. The chromosome that gets passed down is a quilt — a run of P, then a run of Q, then P again. Reich puts the rate at roughly one or two breakpoints per chromosome per generation.

Next generation, chopped again. And again. The blocks get shorter every single generation, forever.

So here is what mixture leaves behind, and what a tree simply cannot counterfeit:

A mixed genome is a patchwork of long ancestry blocks — stretches of DNA that are unmistakably P, alternating with stretches that are unmistakably Q, laid end to end along the chromosome.

Story B has no way to produce that. A population that just sat there drifting in the middle has no P chunks and no Q chunks; it has one uniform ancestry smeared evenly across every chromosome, the same everywhere you look. The lumpiness is the fingerprint. Mixture is visible not because of the average, but because of the structure — and structure is something a single history of quiet drift can't manufacture.

And now the part that turns a fingerprint into history.

The blocks get chopped at a steady rate. That is T6's recombination clock, and this is the exact job it was built for. Measure how long the blocks still are in people today, work out how many generations of chopping it takes to get from whole chromosomes down to that length, and you have not just "these two populations mixed" but when.

Priya Moorjani, a graduate student in Reich's lab, did precisely this for India. The answer:

All Indian groups analysed showed mixture dates between four thousand and two thousand years ago.

Not "sometime in prehistory." A date. Read off the length of the leftover chunks in living people's chromosomes. That's the clock you learned in T6, doing the most consequential work in the book.

The collision that formed India

Now the finding itself — and it's about you, so let's not be coy about it.

Reich's group concluded that essentially everyone in mainland India today descends from a mixture of two highly divergent ancestral populations, which they named carefully:

• ANI — Ancestral North Indians. Related to Europeans, central Asians, Near Easterners, people of the Caucasus.
• ASI — Ancestral South Indians. Related, but very distantly and from tens of thousands of years back, to East Asians; otherwise not close to any population living outside India.

And the line that should stop you: before they mixed, ANI and ASI were as different from each other as Europeans and East Asians are today. Those two groups met on the subcontinent and merged. That's not two neighbouring villages exchanging cousins. That's two ends of Eurasia colliding.

The proportions vary — West-Eurasian-related ancestry in Indian groups runs from about 20% to about 80% — and that gradient is the Indian Cline itself. The line on the plot wasn't a mystery to be explained; it was a mixing gradient being seen directly.

Then Reich's group went looking for anyone who'd escaped it. Every mainland population had some. The only people with none turned out to be the indigenous people of Little Andaman Island — census size under a hundred — who became the essential reference point precisely because they were unmixed. Which is a lovely bit of scientific irony: the key to a billion people's history was a group small enough to fit in a school hall.

Which brings the sentence Reich chose to publish:

"No group in India can claim genetic purity."

Worth knowing how much care went into that sentence. It was written in Hyderabad in October 2008 during what Reich calls "the tensest twenty-four hours of my scientific career" — his Indian collaborators, Singh and Thangaraj, were close to killing the project, because "West Eurasian" ancestry in Indians implied a mass migration into India, which their own mitochondrial DNA work didn't obviously support and which is politically explosive in a way that needs no explanation to you. They proposed the softer phrase "genetic sharing," which would have left open the possibility that no mixture happened at all. Reich refused it, because the data said mixture. The compromise was the neutral naming — ANI and ASI, defined by where they ended up, with no claim about homelands or migration routes baked into the words. They hammered it out on Diwali night, fireworks going off outside.

That's what a good scientific name looks like: it carries the finding and refuses to carry the politics.

The uncomfortable details, because they're in the data

Two more results, both of which you can now read for yourself:

Mixture happened along the grain of social power. Groups of traditionally higher status in the caste system carry, on average, more ANI ancestry than groups they live among — even in the same state, speaking the same language. Brahmins are the clear case. There are documented exceptions and whole groups that shifted status, but the statistical pattern holds. Indo-European-speaking groups have more ANI ancestry, Dravidian-speaking groups more ASI, which is how the genetics hints at who brought which language family.

And the mixture was sex-biased. Look at the two unshuffled lineages from T5 — Y (father→son) and mtDNA (mother→child) — and they tell different stories about the same event. Around 20–40% of Indian men carry a Y-chromosome type descending from a single male ancestor in the last ~6,800–4,800 years. But Indian mitochondrial DNA is almost entirely India-specific, looking like it came from the ASI even in the north. Different story on the paternal line than the maternal line means the incoming ANI ancestry came overwhelmingly from men.

This is why T5 bothered planting Y and mtDNA as separate tracers, by the way. If everything mixed evenly, they'd be redundant. They aren't redundant, and the gap between them is itself the data.

Reich notes the same asymmetry in African Americans (European ancestry arriving roughly 4:1 from the male side) and Colombians (about 50:1). His summary: "males from populations with more power tend to pair with females from populations with less." The genome recorded the social arrangement, and it didn't have the option of being tactful about it.

It's not just India. It's everyone.

If India were a one-off, it'd be a curiosity. It isn't. Reich's group found the same shape everywhere they pointed the method — and Europe turns out to be India's near-twin.

Europeans today descend from three highly divergent populations that came together over the last nine thousand years:

1. Indigenous European hunter-gatherers.
2. Farmers out of Anatolia / the Near East, spreading after ~9,000 years ago.
3. Steppe pastoralists related to the Yamnaya, sweeping in after ~5,000 years ago — bringing wheeled wagons, a new economy, and probably Indo-European languages.

Compare that to India: resident hunter-gatherers, then a Near Eastern farming migration after ~9,000 years ago, then a steppe migration after ~5,000 years ago. It's the same three-act structure on both ends of the continent. Same script, different cast. The two stories that feel most distinct — "European history" and "Indian history" — are, at the genetic level, two performances of one play.

And once you're looking, it doesn't stop:

"The lesson from ancient DNA and the genome revolution has been that anyone in the world is the result of recurrent mixture again and again in the past. You might think that the last 500 years are unusual... But almost every group in the world is the result of many mixture events as profound as these on many timescales." — David Reich

Near Easterners: mixtures of early Iranians, early Levantines, Anatolians — groups wildly different from each other. Native Americans: multiple distinct pulses out of Asia. East Asians, Papuans, Africans: substructure and merging everywhere you look, on every timescale.

Predict-your-confusion sidebar #1. "If literally everyone is mixed, hasn't the word 'mixed' stopped meaning anything?" Fair worry, and no. "Admixture" isn't a vibe, it's a specific, dated, measurable event: two populations that had drifted far enough apart to be distinguishable came together at an identifiable time, and left long ancestry blocks you can find and measure. Saying Indians are ANI+ASI mixed 4,000–2,000 years ago is a falsifiable claim with a number attached. It's the opposite of vague.

Predict-your-confusion sidebar #2. "If I'm '60% ANI', does that mean 60% of my ancestors were ANI?" No — and this trips up almost everyone. It means about 60% of the genome you happen to be carrying traces back to that source. Genealogy and genetics come apart fast (T6 again: you inherit half your DNA from each parent, but only a quarter-ish from each grandparent on average, and by ten generations back most of your ancestors contributed literally zero DNA to you). Ancestry percentages describe your genome, not your family tree.

The trellis, and why there is no bottom

Now the payoff, and it's a genuinely strange one.

You've been picturing mixture as two pure source populations merging into a mixed descendant. Natural picture. Also wrong — because the moment you ask where ANI came from, you find that ANI is itself a mixture. So is ASI. So are the Anatolian farmers, and the Yamnaya, who Reich shows formed from earlier populations mixing on the steppe.

There is no floor. Go back far enough looking for the pure ancestral stock and you never find it; you just find older mixtures of older mixtures. Here's Reich on how deep it runs:

"It's very plausible that people's ancestors are not all in Ethiopia 200,000 years ago. In fact some of them are maybe in North Africa. Some of them are maybe in West Africa... That braid and that trellis is coming together again and again over time. As you move further back, they'll collapse. Some will go extinct, some will reappear, some will re-merge. **At any one point, there's never a singularity." — David Reich

Never a singularity. There is no moment in the past where humanity is one clean unmixed group you could point at and call the original. Every time you zoom back hoping to find the trunk, the trunk resolves into more branches braiding together.

So the trellis isn't a decorative metaphor sitting next to the tree. It's a claim that the tree has no valid version at any zoom level.

And that reframes the phrase Reich went to war over in Hyderabad. "No group in India can claim genetic purity" is not a diplomatic gesture or a moral statement. It's a category error being pointed out. Genetic purity isn't rare, or lost, or diluted. It's a property that has never existed for any human group at any point in history — because every candidate for a "pure" source population turns out, on inspection, to be a mixture of things that mixed earlier.

That's Reich's thesis. That's the book. You now have it.

The problem this leaves you with

One thing should be nagging at you.

We just spent a chapter talking confidently about the ANI and the ASI. Their ancestry proportions. Roughly when they mixed. Which languages they likely carried.

Nobody has ever met an ANI. There is no unmixed ANI population alive today. There never will be. They were reconstructed entirely from the statistical residue they left inside living Indians' chromosomes — inferred, named, and dated without a single direct observation.

That's a ghost population, and Reich's field is full of them. The Ancient North Eurasians were proposed as a pure inference — a group nobody had ever sampled — and Reich's team could show they must have been about as distinct from everyone else alive at the time as West Eurasians, Native Americans, and East Asians are from each other today. Today, more than half the world's population carries between 5% and 40% Ancient North Eurasian ancestry. Hundreds of millions of genomes' worth of a people who exist nowhere.

Which raises the obvious, uncomfortable question: how much would you trust a population that exists only as an inference?

The answer arrived at the end of 2013, when a genome came out of the bones of a boy who died at Mal'ta in Siberia twenty-four thousand years ago — and it matched the predicted ghost. The inference had been right, and now it had a skeleton.

Which is the whole next move of the field: stop deducing the past from living people, and go sequence the dead directly.

The challenge

1. Warm-up. A friend shows you a population whose allele frequencies sit exactly halfway between two others and says "clearly a 50/50 mixture." Give the rival explanation that fits the same data, then tell them the one measurement that separates the two — and explain why one history can produce it and the other physically cannot.

2. The real one. Two populations mixed once, cleanly, N generations ago, and haven't mixed since. Describe what a descendant's chromosome looks like at generation 1, generation 10, and generation 500 — and explain why the mixture eventually becomes undatable by this method. (What is actually being lost? Then: does that make old mixtures invisible, or just undatable this particular way? Look back at T3 and T5 for what else survives.)

3. Think ahead. You're handed genomes from a living population and you find long ancestry blocks from a source that matches nothing alive on Earth. Argue that the source population really existed, rather than being a statistical artefact — then name the one kind of evidence that would settle it beyond argument. (You just re-derived the ghost-population problem and the reason the entire field pivoted to digging.)

Next: you've been reading the past out of living people, and it works — but every answer arrives as an inference about someone you've never seen. So the field did the obvious impossible thing and went after the source: DNA out of 40,000-year-old bone. Next chapter is the ancient-DNA revolution — why everyone was certain it couldn't be done, what contamination and decay actually do to a genome, and what the clean rooms changed. It's the technical fact the entire second half of Reich's work stands on.