In 1972, a geneticist named Richard Lewontin sat down with data from 17 genetic markers sampled from seven broad population groups that the science of his era still called races, categories like Caucasian, Black African, and East Asian. He wanted to know, mathematically, how much of human genetic variation existed between those groups compared to within them. Where he expected to find a clear biological basis for dividing humanity into distinct types, he instead found that 85. 4% of all human genetic diversity existed within populations, not between them.
Two people standing in the same racial category from the same continent could be more genetically distant from each other than either one was from someone on a completely different continent. The differences between the racial categories themselves accounted for only a sliver of what was left. The finding was a structural collapse of a framework that had shaped human civilization for over a century. For a hundred years, humanity had built its understanding of itself on skin color, hair texture, and facial structure.
Nations were built on it, wars were fought over it, and immigration laws were written around it. Lewontin’s numbers suggested that the entire edifice had been constructed on a biological misunderstanding. But science is never as clean as a single headline number, and in 2003, thirty years after Lewontin’s paper, a Cambridge statistician named A. W.
F. Edwards published a direct challenge. His argument was not that Lewontin’s math was wrong. The 85/15 split held up.
Edwards contended instead that Lewontin had treated each of the 17 genetic markers as if they were independent signals, when in reality, much of the information that distinguishes populations lies not in any single marker but in the correlations between many markers considered together. Look at one gene at a time and populations blur into each other. Look at thousands of genes simultaneously in combination, and you can often statistically sort someone into their region of ancestral origin with striking accuracy. Two of the most rigorous minds in population genetics looked at the same underlying biological reality and walked away with opposite interpretations.
One said race has almost no genetic basis. The other said the pattern is there if you know how to read it. The fight was never really about the data. It was about what the data was going to be used for, and that fight did not stay buried in academic journals.
By 2024, technology had changed completely. Researchers were no longer working with 17 protein markers measured by hand. They had full genomic sequencing applied to hundreds of thousands of people, cross-referenced against global ancestry databases built from every inhabited continent on Earth. That year, a paper landed in the journal Nature analyzing genomic data from participants in the All of Us research program, a massive United States database built specifically to reduce medical inequities by recruiting from historically underrepresented populations.
The paper’s technique for parsing race and ancestry data drew immediate fire from critics who argued that the method could be misconstrued to support the idea that human beings can be neatly sorted into distinct biological races, the exact conclusion that fifty years of genetics had supposedly closed the door on. This was a peer-reviewed Nature paper, and the backlash came from inside the scientific establishment itself. A different team went back into the same well of data, armed with a different method, determined to find out who was right. In June 2025, a study published in the American Journal of Human Genetics, led by researchers including Eduardo Tarazona-Santos, analyzed the DNA of more than 230,000 people enrolled in the All of Us database using principal component analysis alongside global reference panels to compare people’s self-identified race and ethnicity against their actual genetic ancestry.
It was an actual population-scale comparison between what people call themselves and what their DNA says about where their ancestors came from. The results did not just complicate the picture. They blew a hole straight through it. The researchers found that most genetic variants existed within self-identified racial and ethnic groups rather than between them, confirming, more than fifty years later and with vastly more powerful tools, the same core pattern Lewontin had first detected in 1972.
But this study went further. Instead of finding distinct genetic clusters that lined up neatly with racial categories, the team found gradients, continuous streams of genetic variation flowing across and through the boundaries that white, black, and Asian American are supposed to represent. A slope with no clean line anywhere on it marking where one race ends and another begins. The researchers also discovered significant genetic variation even among people who shared the exact same self-reported racial and ethnic identity, depending on which US state they lived in.
Among people who identified as Hispanic or Latino, those living in states like California, Texas, and Arizona carried a notably higher proportion of Native American ancestry than Hispanic or Latino people elsewhere in the country, a pattern tracing directly back to the fact that those states were historically part of Mexico. Meanwhile, Hispanic or Latino participants living in New York showed the highest proportion of African ancestry of any state in the study, a signature consistent with more recent migration patterns from the Caribbean. Same racial checkbox, same ethnic label, two almost entirely different genetic histories shaped not by biology but by centuries of ships, borders, and migration routes. The authors concluded that social constructs of race and ethnicity do not accurately reflect underlying genetic ancestry, and stated plainly that they do not recommend using race and ethnicity as proxies for ancestry in genetic research going forward.
One of the study’s own collaborators called the findings critical for developing appropriate precision medicine solutions for everyone, given how the genetic backgrounds of people in the United States turn out to be far more complex than the racial and ethnic categories used to describe them. The deeper the science digs, the further back the real revelation stretches. For decades, the standard model taught in classrooms was simple and clean. Modern humans evolved from a single ancestral population somewhere in Africa, and everyone alive today descends from that one founding group.
A sweeping genetic analysis incorporating data from diverse modern African populations alongside fossil evidence has begun dismantling that single origin story piece by piece. Researchers now suggest early humans likely evolved from multiple intermingling populations across hundreds of thousands of years, groups that stayed connected and continued exchanging genes with one another even after they began diverging somewhere between 120,000 and 135,000 years ago. Not one tribe branching outward, but multiple populations splitting apart and merging back together again and again. The deepest, most unbroken thread of that history is still alive today in the genomes of specific living people.
A study analyzing ancient genomes from Southern Africa reported roughly 9,000 years of genetic continuity in the region, and highlighted the Nama people, whose modern genetic signature carries some of the most distinct and ancient lineages found anywhere on the planet. The plot twist does not stop at Africa. Outside of Africa, nearly every non-African population on Earth carries measurable amounts of DNA inherited from Neanderthals, an entirely separate human lineage that our ancestors encountered and interbred with tens of thousands of years ago. There is no population on Earth that represents an unbroken, unmixed genetic lineage stretching back to some mythical founding group.
Every population is a mosaic. And the traits people have historically used to define race, skin tone, hair texture, facial features, are not, genetically speaking, deeply rooted markers of ancestry at all. They are surface-level adaptations shaped rapidly and repeatedly by local environments, particularly by exposure to ultraviolet light near or far from the equator. Similar pigmentation adaptations arose independently multiple times in different lineages facing similar environmental pressures, a phenomenon geneticists call convergent evolution.
Two people with similar skin tones on opposite sides of the planet may share almost nothing else in their genome. Two people with wildly different skin tones may share deep, recent common ancestry. The marker doing the most cultural work was biologically doing almost none of the actual work. For over a century, entire governments, immigration systems, and legal frameworks were constructed on the assumption that visible traits reliably tracked deep ancestry, temperament, ability, and worth.
The genetic evidence mounting for fifty years, now confirmed at a scale of hundreds of thousands of genomes, says that assumption was never supported by the biology. It was a story people told about the biology, mistaken for the biology itself. The science is not finished. The 2024 paper and the 2025 paper disagreed on methodology, and somewhere out there a future paper is already being drafted that will complicate both of them further.
New sequencing technology keeps arriving, new ancient genomes keep getting recovered from bone fragments buried for tens of thousands of years, and every year the resolution of the picture gets sharper. Every time it does, the old simple categories fracture a little further. Precision medicine is racing forward, built on the promise of tailoring treatment to individual genetic profiles rather than broad racial categories, but implementing the shift means rebuilding decades of clinical assumptions, funding structures, and diagnostic tools. Forensic science is developing techniques to predict a person’s likely ancestry or physical appearance directly from a DNA sample left at a crime scene, a technology that could revolutionize investigations or just as easily encode the same old racial assumptions into a new, more scientific-looking form.
Underneath it all sits a political question no genome sequencer can answer. If the genetic evidence increasingly shows that race is a gradient, not a category, a human invention layered over deep, tangled biology, what happens to the laws, the census forms, the medical guidelines, and the cultural identities all built using race as a fixed, countable unit? The genome does not care about census categories, national borders, or the checkboxes on a hospital intake form. It carries only what actually happened.
Every migration, every mixing, every isolated population that stayed put for 9,000 years, every population that scattered and remixed a thousand times over. The categories used to talk about human difference were built for administrative convenience, not biological accuracy, and the deeper science digs, the more that gap widens rather than closes.