Overview
David Reich describes a large ancient-DNA study that challenges the prevailing view that recent human natural selection was largely quiescent. Using approximately 16,000 ancient individuals, modern comparison samples, and a relatedness-based model spanning roughly 10 million genetic positions, the researchers attempt to separate consistent directional selection from the much larger effects of migration, admixture, and genetic drift. Although selection accounts for only about 2% of allele-frequency change, the study finds hundreds of high-confidence selected positions and evidence for thousands of weaker signals. Immune and metabolic traits show the clearest enrichment, while behavioral traits appear harder to detect because they are distributed across many variants of individually small effect. A striking result is that selection intensified or reversed for numerous traits between about 5,000 and 2,000 years ago, suggesting that Bronze Age population density, animal contact, technology, and social organization imposed unusually strong biological pressures. The interview also explores polygenic predictors, declining obesity-related risk, the absence of recent universal selective sweeps, and a speculative alternative model connecting modern humans and Neanderthals. The central lesson is methodological as much as historical: population movement can mimic adaptation, modern trait labels do not identify the ancient target of selection, and sufficiently large datasets can overturn plausible evolutionary narratives.
Sections
Higher-Level Implications
Patterns and implications that emerge across the study's results and methodological discussion.
- The study reframes recent human evolution as continuous fine-tuning rather than a sequence dominated by rare, complete selective sweeps. Strong demographic turnover can conceal many smaller adaptive movements that repeatedly point in the same direction.
- The Bronze Age signal implies that cultural intensity matters independently of technological first appearance. Farming may have created the prerequisite environment, while later density, animal exposure, trade, and social complexity generated the strongest biological pressures.
- Trait-associated variants provide external validation for the selection statistic, but modern phenotype names remain imperfect historical interpretations. The genetic signal may be real even when the proposed ancient trait is wrong.
- The interview repeatedly demonstrates that absence of a fixed difference is not absence of biological evolution. Adaptation may operate through coordinated, modest shifts across many existing variants.
Key Concepts
Terms needed to interpret the study and its limitations.
- Genetic drift: random generation-to-generation changes in variant frequency, whose fluctuations are larger in smaller populations.
- Directional selection: systematic pressure that pushes a mutation or polygenic trait toward a different adaptive set point rather than merely removing newly arising harmful mutations.
- Evolutionary mismatch: a condition in which variants shaped in an earlier environment, such as hunter-gatherer life, are poorly suited to a newly agricultural, pastoral, or urban environment.
- Background selection: the removal of newly arising harmful mutations, especially around genes; it can create patterns correlated with functional regions without representing adaptation to a new environment.
- Selection statistic: an approximately Gaussian measure of how strongly a mutation changes consistently over time relative to a null value representing no directional selection.
- Polygenic trait: a characteristic influenced by many variants, usually of individually small effect, making selection harder to detect at any single location.
Study Design and Quantitative Details
Specific data, modeling, validation, and laboratory details described in the interview.
- The analysis covers roughly 10 million variable positions across about 22,000 individuals, including approximately 16,000 ancient people distributed over 18,000 years; around 10,000 ancient individuals were newly reported.
- The model predicts each genotype from the individual's relatedness to all other samples, then tests whether adding a constant directional-selection term improves the prediction beyond ancestry, migration, bottlenecks, and drift.
- At least 479 approximately independent positions were reported with 99% confidence. The transcript gives conflicting estimates at the 50% threshold: one passage says 7,200 candidate positions with about 3,600 expected to be real, while the later methodology discussion says about 3,800 positions are each more than 50% likely to be real.
- Independent validation used genome-wide association results from the UK Biobank. Approximately 15% of random positions were associated with at least one measured trait, rising to roughly 60%-70% among positions with a selection statistic above about five.
- The enrichment plateau was used to calibrate a position-specific probability of genuine selection. Analyses stratified by background-selection exposure and allele frequency reportedly produced the same pattern.
- In-solution enrichment captures targeted human DNA fragments from samples that often contain less than 1%-10% human DNA, with the remainder largely microbial. Robotic processing now allows some laboratories to generate genome-scale data from thousands of ancient individuals per year.
Important Contrasts
Explicit alternatives and distinctions central to the interview.
- Migration can cause abrupt genome-wide frequency shifts, whereas directional selection is inferred when particular variants repeatedly move in the same direction within relatively stable local populations.
- Earlier ancient-DNA scans found 12 strong loci in 2015 and 21 in a larger 2024 study, while the new relatedness-based approach reports hundreds of high-confidence independent positions.
- Immune traits often involve fewer, stronger-effect loci, while behavioral traits involve many weak-effect loci; the former therefore produce clearer signals among the strongest selected positions.
- The standard archaic-human model treats Neanderthals and Denisovans as sister groups with later admixture patches. Reich's speculative model instead treats a modern-related expansion and local genetic swamping as the common explanation for shared technology, timing, mitochondrial DNA, and Y chromosomes.