The human genome carries surprisingly little genetic diversity for a species of eight billion. New genomic research suggests one possible reason: hundreds of thousands of years ago, our ancestors may have dwindled to a population of only around 1,300 breeding individuals, leaving a genetic bottleneck whose effects can still be detected today.

Humanity can fill megacities, cross oceans and count more than eight billion living members, yet our genomes retain the signature of a much smaller past. That contrast does not mean people are genetically identical. It means our present census grew far faster than mutation could rebuild every strand of variation lost over deep time.

One proposed loss was extraordinary. A genomic model published in 2023 inferred that an ancestral population shared by living humans shrank to an effective size of about 1,280 breeding individuals between roughly 930,000 and 813,000 years ago. Since then, other researchers have argued that the signal may be a statistical artifact. A 2026 response has defended it. The bottleneck is now an active test of how much history can be recovered from modern DNA, not a settled near-extinction story.

Eight billion people can inherit a narrow genetic archive

Population size and genetic diversity respond on different timescales. A population can grow explosively in a few thousand years while carrying variants descended from a far smaller group. New mutations accumulate slowly. Genetic drift, the random loss of variants as generations reproduce, can remove diversity much faster when a population is small.

The 1000 Genomes Project cataloged more than 88 million variants across 2,504 people from 26 populations, which sounds like enormous variation. It is enormous in absolute terms, yet any two human genomes remain overwhelmingly alike. That combination is possible because a genome contains billions of positions. A small percentage difference still amounts to millions of variants, while leaving humans less diverse than several other primate species.

The 1,280 figure is not a literal head count

The key quantity is effective population size, usually written as Ne. It is the size of an idealized population that would experience genetic drift at the rate seen in the data. Ne can be much smaller than the number of bodies actually alive. Unequal reproductive success, changing sex ratios, age structure and subdivision among groups can all reduce it.

So the claim is not that researchers found a lost valley occupied by exactly 1,280 survivors. It is that the ancestry represented in living genomes behaved, under the model, as if it had passed through a breeding population of that approximate size. The ancestors involved also predated Homo sapiens. They belonged to earlier Homo populations ancestral to modern humans and perhaps to Neanderthals and Denisovans.

FitCoal read the past through variant frequencies

The original Science study, led by Wangjie Hu, introduced a method called the fast infinitesimal time coalescent process, or FitCoal. The team analyzed modern genome sequences from 3,154 people in African and non-African populations. Instead of reading ancient DNA from 900,000 years ago, which is not available, the method examined the site frequency spectrum: how often genetic variants appear across present-day samples.

Different demographic histories leave different expected frequency patterns. FitCoal searched for the population-size history that best matched those patterns and inferred a fall from roughly 100,000 effective breeding individuals to about 1,280. The low point lasted an estimated 117,000 years. ScienceBlog’s original report on the result described why that estimate attracted so much attention.

The timing made a compelling evolutionary narrative

The proposed interval sits near the Early to Middle Pleistocene transition, when glacial cycles became longer and more intense. It also falls in a difficult stretch of the African and Eurasian fossil record. The original authors suggested climate disruption could have reduced suitable habitat and food, while a later recovery might align with changes in the human lineage.

Those correspondences are suggestive, but they cannot confirm a genomic model. Fossil gaps can reflect where researchers have looked, where sediments survived and whether remains fossilized. Climate shifts do not demonstrate that one ancestral population crashed to a particular Ne. Nor does a bottleneck automatically explain later brain expansion or chromosome evolution merely because their estimated dates overlap.

Independent analyses found reasons for doubt

Two later studies tested the claim using different assumptions and tools. One concluded that the signal was likely a statistical artifact, showing that ancient population structure could generate patterns resembling a sharp crash. Another found insufficient evidence for a bottleneck of the proposed magnitude after filtering difficult genomic regions and comparing African and non-African samples.

A separate method called PHLASH, published in Nature Genetics, estimated historical population sizes while modeling uncertainty and recombination. Across 159 population estimates, it found little sign of an effective size near 1,280 during the proposed interval. Its simulation tests indicated that a crash as strong as the FitCoal result should have produced a detectable signal. In the real data, most population estimates stayed around 10,000 to 15,000, and none fell below 5,000.

A 2026 defence says the bottleneck can be missed

The argument did not end there. A July 2026 paper in Molecular Biology and Evolution compared FitCoal with one of the methods used in the critiques across a very large set of coalescent simulations. The authors reported that FitCoal accurately reconstructed sharp ancient bottlenecks while the competing approach could smooth them away. They also argued that fossil, climate and dispersal evidence remains consistent with a severe decline.

That new defence of the severe-bottleneck model includes Wangjie Hu and Haipeng Li from the original research line. Their simulations directly address methodological criticism and should not be dismissed, but the overlap also matters. This is a rebuttal and extension by proponents of FitCoal, not a fully independent replication that settles the dispute.

What the genome can and cannot yet tell us

The human genome unquestionably records ancient changes in population size, migration and mixture. Bottlenecks associated with movements out of Africa are supported across many methods, and long-term effective population sizes have often been small relative to today’s census. The narrower question is whether one extreme, prolonged crash around 900,000 years ago is required by the data.

For now, about 1,300 should be treated as a model-dependent estimate, not a count of the last people on Earth. Human population structure can mimic a small Ne, genomic regions are shaped by selection and recombination, and different inference methods compress deep history in different ways. Evidence from living genomes also traces only the ancestral branches that survived into people now sampled.

The proposed bottleneck remains a plausible contributor to humanity’s relatively limited diversity, but it is not the only one. Repeated population contractions, founder events, ancient subdivision and recent explosive growth all helped shape the pattern. The strongest conclusion is therefore more modest and more interesting: eight billion people carry enough shared genetic history to reveal ancient scarcity, while the exact depth and shape of that scarcity remain open scientific questions.