Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Back to Podcast Digest
Why Scientists Missed Recent Human Evolution – David Reich
TL;DR
Ancient DNA crushed the history question before it could answer the biology question — David Reich says the field has been transformative on migration, mixture, and sex-biased ancestry, but for traits like skin pigmentation or milk digestion, sample sizes were too small until the last few years.
One ancient genome is rich for ancestry but almost useless for tracking a single trait variant — a single person encodes tens of thousands of ancestors for population history, but only gives one or two copies of a specific allele, which is why detecting 2-10% frequency shifts requires massive datasets.
The old consensus was that recent human natural selection was mostly quiet, and Reich argues that was misleading — because Europeans and East Asians split only 40,000-50,000 years ago and still show few near-fixed differences, many scientists inferred little directional selection in recent human evolution.
Reich's new result is that 98% of allele-frequency change is not adaptation but drift, migration, and population structure — that means natural selection signals are real but drowned out by demographic noise, especially during major replacement events.
Big migrations like the Yamnaya expansion are the worst moments to detect selection at individual genes — around 4,500 years ago in Europe, 40-80% of ancestry shifted due to steppe pastoralist migration, causing huge genome-wide frequency swings that are statistically uninformative for pinpointing adaptation.
The trick is to look at many small, quiet intervals and ask whether the same mutation keeps nudging in one direction — Reich describes Europe and the Middle East as an 'archipelago' of semi-isolated populations, where repeated directional pushes across hundreds of years reveal genuine natural selection.
The Breakdown
The dream of ancient DNA finally catches up to biology
Reich opens by saying the original dream of ancient DNA, going back 16 or 17 years, was not just reconstructing migrations but watching human biology change over time. The surprise is that the field succeeded wildly on history — mixture, replacement, and sex-biased ancestry everywhere — while mostly failing on biology, because the datasets just weren't big enough.
Why one genome tells you a lot about ancestry and almost nothing about selection
He gives a clean explanation: one genome is effectively many people when you're asking about history — parents, grandparents, great-grandparents, all the way back thousands of generations. But for a specific variant tied to skin pigmentation, milk digestion, or behavior, that same individual gives you only one or two copies, so you need huge sample sizes to see subtle frequency changes through time.
Why frequency shifts matter at all
Dwarkesh asks the obvious value question: why care about allele frequencies? Reich's answer is that environmental change — agriculture, domesticated animals, altitude, temperature — creates pressure, and the evidence of adaptation is a variant drifting in one direction more than chance would allow.
The old mainstream view: recent human selection was mostly quiet
Reich says the dominant view for decades was that natural selection had been pretty quiescent over the last several hundred thousand years. One reason was comparative: Europeans and East Asians, despite splitting 40,000-50,000 years ago, show very few mutations that are anywhere near 100% different in frequency, which made it seem like humans had mostly settled into an evolutionary optimum.
What the new study actually finds: drift dominates, selection still matters
Using the much larger ancient DNA datasets his lab has been building — with Ali Akbari leading the effort over seven years — the team tries to separate directional selection from everything else. Their headline number is stark: about 98% of frequency change across roughly 10 million variable positions is explained by other forces, especially genetic drift, migration, and population structure.
Dwarkesh pushes on replacement as a kind of selection
Dwarkesh asks a sharp question: if one population replaces another, why not call that selection at the group level? Reich agrees it could matter in some sense, but says statistically it is a whole-genome shift that may reflect culture or technology as much as genetics, so it doesn't tell you whether one specific mutation was being favored over the rest of the genome.
The 'archipelago' idea and the Yamnaya example
Reich's memorable metaphor is to treat Europe and the Middle East as an archipelago of little populations, isolated in space and time between big waves of migration. Those quiet windows are where you can see a mutation slowly 'blowing' in one direction; by contrast, the steppe migration into Europe 4,500 years ago — when 40%, 50%, even 80% of ancestry became Yamnaya in some places — creates giant swings that swamp any clean signal of gene-level selection.
Share
Keep Reading
Two Ways to Follow
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.
