Modern Humans Did Not Admix with Neanderthals during Their Range Expansion into Europe

One hundred and twenty. That's the ceiling — not a rate, not a percentage, but an absolute count of cross-population matings that the genetics of modern Europeans can tolerate without contradiction. One hundred and twenty encounters, spread across twelve thousand years of cohabitation between two human populations sharing the same continent. That number is what Currat and Excoffier arrived at in 2004, and it changes how you have to think about what it meant for modern humans and Neanderthals to have "met." The puzzle starts with a mismatch. Neanderthals vanished from Europe somewhere between forty-two thousand and thirty thousand years ago, right as anatomically modern humans were spreading in from the Near East. Genetic surveys of mitochondrial DNA, or mtDNA, the genetic material passed down exclusively through maternal lines, examined more than four thousand sequences from present-day Europeans and found exactly zero Neanderthal lineages. Seven early modern European fossils told the same story. Whatever happened during that replacement, Neanderthal mothers were not contributing to the lineage that became us. But here was the problem: the statistical tools available at the time couldn't actually close the case. The older models, which assumed a single, instantaneous mixing event in an undivided population of constant size, could only rule out very large Neanderthal contributions. Contributions smaller than twenty-five percent couldn't be statistically rejected. Twenty-five percent means roughly one in four matings producing a Neanderthal-derived lineage entering the modern human gene pool. That's not a trace — that's substantial biological mixing. So you had two things that refused to agree: zero detected Neanderthal mtDNA on one side, and statistical models that couldn't rule out massive interbreeding on the other. Currat and Excoffier set out to close that gap. Their solution was to build a model that actually looked like what happened — not a single mixing event frozen in time, but a wave of modern humans moving across Europe cell by cell, crowding out Neanderthals as they went. They represented Europe, the Near East, and North Africa as a grid of seven thousand five hundred square cells, each roughly two thousand five hundred square kilometers. Two local populations — modern humans and Neanderthals — could coexist within a cell, and competition between them was density-dependent: as modern humans filled a cell, Neanderthals were progressively squeezed out. The model included explicit life-cycle steps — admixture, competition, migration — run for one thousand six hundred generations, corresponding to about forty thousand years. On top of that demographic history, Currat and Excoffier ran coalescent simulations, working backward through recorded genealogies to estimate what ancestry four thousand sampled modern European mtDNA sequences should show under different interbreeding rates. They tested nine demographic scenarios, varying origin location, source population size, growth rates, migration speed, and carrying capacities, to make sure their conclusions weren't hostage to any single set of assumptions. The most important insight from the model is counterintuitive, and it's worth sitting with. When modern humans expand as a wave, the people at the leading edge are repeatedly founding new demes from small groups. Those founder events amplify any genes that get picked up along the way. A single Neanderthal mtDNA entering a growing modern human deme at the wavefront doesn't disappear into a large population — it enters a small, growing one, where logistic growth inflates it in absolute numbers before drift can eliminate it. Migrants leaving that deme carry the lineage forward to the next founding event. The process cascades. This means the model is inherently sensitive to even tiny amounts of admixture — built to find Neanderthal ancestry if it exists. The simulation results make this concrete. In Currat and Excoffier's basic scenario, an average of just one admixture event for every ten demes produces an expected thirty-eight percent of modern European mtDNA tracing back to Neanderthals. In a scenario where modern humans had a smaller competitive advantage, the same rate still yields fifteen percent Neanderthal contribution. Across most scenarios, two admixture events per deme over the coexistence period pushes the expected Neanderthal contribution above ninety-five percent of the modern human gene pool. The wavefront amplifier is powerful. And yet the actual data show essentially nothing. That's what makes the absence so striking — the model was designed to be sensitive, and it still finds no signal. The numbers that come out of this framework are severe. The maximum-likelihood estimate for the interbreeding rate is zero — the best fit to the observed data is no admixture at all. The ninety-five percent confidence interval places the upper bound below 0.1 percent. Depending on the scenario, the total admixture estimates range from thirty-four events, in a scenario with a more easterly modern human origin, to one hundred and twenty events in the scenario allowing longer local cohabitation times of twenty-one to thirty-seven generations. Across the core scenarios, the upper bound sits at roughly 0.015 admixture events per deme; even the more permissive scenarios cap out at around 0.03. These bounds held across the full range of demographic assumptions tested — varying origin, diffusion of the source population, colonization speed, and symmetry of gene flow. The result is not sensitive to parameter choices. Compare that to the old ceiling of twenty-five percent. The spatially explicit model doesn't just refine the estimate — it collapses it by more than two orders of magnitude, from one in four matings to fewer than one in a thousand. And it translates that rate into something you can picture: at most one hundred and twenty fertile cross-population matings, across all of Europe, over the entire twelve thousand year period these two populations shared the continent. Twelve thousand years is longer than all of recorded human history. One hundred and twenty events is fewer people than a small village. Currat and Excoffier draw the biological conclusion directly. Because mtDNA traces maternal lineages, the result is specifically sensitive to the direction of crossing — and the near-complete absence of Neanderthal mtDNA is most consistent, they argue, with an almost complete sterility between Neanderthal females and modern human males. At that level of reproductive isolation, the two populations were probably distinct biological species — not different tribes, not different cultures, but separate species that happened to occupy the same landscape for over a hundred centuries without meaningfully merging. They are careful about what the model can't do. It doesn't incorporate long-range dispersal, environmental heterogeneity, or the demographic disruption of the later Neolithic expansion. Direct evidence from fossil Y chromosomes or nuclear DNA would add constraints the mtDNA picture alone can't provide. These are real limits. But the core finding — that the wavefront amplification model tightens the interbreeding ceiling so dramatically — holds across all nine scenarios tested. That's the strongest kind of robustness a simulation study can demonstrate. What lingers is the human strangeness of it. Twelve thousand years is an almost incomprehensible span of shared geography. Modern humans and Neanderthals overlapped across the same forests, valleys, and coastlines for longer than agriculture has existed. And the best current estimate, from a model deliberately designed to detect any signal of interbreeding, is that their maternal lineages mixed at most one hundred and twenty times total. Whether that reflects active avoidance, ecological separation, or a reproductive barrier so deep it operated almost automatically — the data, on their own, can't say. What Currat and Excoffier can say, with unusual precision, is that the barrier was nearly absolute. Neighbors for millennia, and almost never kin. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.