We tend to tell the disappearance of the Neanderthals as if it ought to have one decisive cause. Europe became too cold. Modern humans arrived with better tools. Two kinds of people met, and one defeated the other.
All three versions are attractively simple. The archaeological record is not.
A 2026 study in Quaternary Science Reviews approaches the problem from a different angle. Instead of searching for a single fatal event, a team led by Université de Montréal archaeologist Ariane Burke asked how habitable regions across Europe were arranged, how that arrangement changed with the climate, and whether separated populations could plausibly remain connected.
The result is subtle. The researchers did not find climate steadily wiping suitable Neanderthal habitat from the map. Good regions survived repeated cold and mild swings. But suitable territory is not much protection if the people occupying it are thinly spread and the next viable region lies too far away. In parts of Europe, the modelled Neanderthal network looked fragile in exactly that way.
Homo sapiens, by comparison, had a more connected set of potentially suitable regions. Connections can act as demographic insurance, allowing small communities to exchange mates, information and resources, or to find refuge when one area fails. The model suggests that some Neanderthal populations had less of that insurance.
Turning scattered sites into a map of possible lives
The study covers the period from about 60,000 to 33,700 years ago, spanning ten cold stadial and milder interstadial phases of Marine Isotope Stage 3. During this interval, Homo sapiens established a lasting presence in Europe and Neanderthals disappeared from the archaeological record.
Burke and her colleagues began with dated archaeological sites. They used Mousterian sites as evidence of Neanderthal presence and Aurignacian sites as evidence of early Homo sapiens. Those associations are not perfect labels for biological species, but they are the clearest available cultural proxies. The team excluded more disputed traditions, including the Châtelperronian, Proto-Aurignacian and Initial Upper Palaeolithic, rather than quietly assigning uncertain sites to one population.
Climate simulations supplied temperature, precipitation and measures of climate variability. Elevation and slope were added to describe the terrain. The climate data, initially far coarser, were statistically downscaled to a grid of roughly 15 by 15 kilometres.
From this, the researchers built four habitat-suitability models: Neanderthals and Homo sapiens under both cold and mild conditions. They used a machine-learning method called Random Forest, averaged across 100 runs, to estimate where the environmental combination resembled that surrounding known sites.
Then came the unusual step. Drawing on ethnographic estimates for mobile hunter-gatherers, the team searched for large patches that might support regional bands of roughly 175 to 570 people. After testing maps containing five, ten, 15 and 20 regions, they used a ten-region scenario to compare the shape and connectivity of the two populations’ potential networks.
This is where careful wording matters. The study did not discover an actual Neanderthal social network. It did not recover friendship ties, marriage routes or journeys between named communities. Its links connect modelled habitat regions. They show where contact between bands may have been easier or harder if people occupied the most suitable areas.
The cold moved the refuge, but did not erase it
If climate alone had doomed the Neanderthals, we might expect the models to show their suitable habitat degrading severely as Europe lurched through glacial instability. That was not what emerged.
Optimal habitat persisted for both populations under cold and mild conditions, although its location shifted. Southwestern Europe repeatedly appeared as a core region where populations could contract during hard periods and expand again when conditions improved. Southern Iberia formed a second particularly durable Neanderthal core.
The authors interpret this persistence as evidence of spatial resilience. Climate still mattered. A habitat that moves can force a group to move with it, stretch travel routes, split neighbours apart and alter access to prey or raw materials. Climate variability may also be more disruptive than a simple change in average temperature. What the model challenges is the claim that cold spells simply removed the places in which Neanderthals could live.
That is an important distinction because extinction can happen while habitable land remains. A map can contain food, water and shelter yet still fail to support a population over the long run if communities become too small, too isolated or repeatedly unlucky.
It also fits a less caricatured view of Neanderthal adaptability. ScienceBlog recently covered the evidence that Neanderthals at Barnham may have deliberately made fire 400,000 years ago. Whatever finally happened to their lineage, it cannot sensibly be reduced to an inability to cope with difficult environments.
Why connection can matter more than the amount of land
Imagine several small communities scattered across a landscape. One loses experienced hunters in an accident. Another has several years with too few surviving children. A third finds that its usual prey has shifted. None of those shocks needs to be catastrophic if people can move, join neighbours, find partners or draw on information from farther away.
Isolation changes the arithmetic. In a small population, the death or departure of a handful of people matters more. A local decline can become permanent because there is no nearby community able to replenish it. Over generations, weak connections can also narrow the pool of potential partners and increase vulnerability to inbreeding.
The new models suggest that the optimal Neanderthal regions in western and southeastern Europe were linked only tenuously across large distances. Homo sapiens regions were generally better connected, and mild interstadial conditions opened relatively more suitable habitat for the newcomers.
This does not mean Neanderthals lacked long-distance contacts. Archaeologists have traced raw materials across considerable distances, evidence that objects and probably people moved between groups. The model’s claim is comparative and geographic: the arrangement of the best regions offered fewer easy links in some parts of the Neanderthal range.
Population history had already left Neanderthals exposed. Earlier research covered by ScienceBlog found signs that Neanderthal populations suffered a severe contraction around 110,000 years ago, long before their final disappearance. A lineage can survive a bottleneck and remain capable for tens of thousands of years, while still carrying demographic vulnerabilities into the next run of shocks.
There may not have been one European ending
One of the study’s most useful conclusions is that the balance of pressures probably differed by region.
In western Europe, the model showed substantial overlap between optimal regions for Neanderthals and Homo sapiens, including around important Neanderthal cores. Direct interaction may therefore have mattered more there. That interaction could include competition, displacement or cultural exchange. It could also include absorption into a larger population.
In southeastern Europe, by contrast, the Neanderthal network was more stretched. There, random demographic losses and weak links between bands may have done more of the damage. Southern Iberia remained a strong refuge, consistent with evidence that Neanderthals persisted comparatively late in the southwest.
Even the word “replacement” needs care. Neanderthals vanished as a distinct population, but not every part of their ancestry vanished with them. The two populations interbred, and Neanderthal DNA survives in people today. ScienceBlog has previously looked at how the genetic record of Neanderthal and modern-human interbreeding is itself uneven.
Genetic assimilation is therefore one possible part of the regional story, not a proven master explanation. Nor does this model test whether one population was cognitively, socially or technologically “superior.” It compares environmental suitability and the spatial arrangement of large habitat patches. Any stronger verdict would be imported from outside the analysis.
The overlap maps also weaken a simple resource-war story. Neanderthals and Homo sapiens showed subtle differences in habitat preference, which may have reduced direct competition for the same places. The researchers do not rule competition out. They argue that it is unlikely to have operated with the same intensity everywhere or to explain the whole disappearance by itself.
A powerful model with deliberately visible limits
All reconstructions of deep prehistory rest on incomplete evidence, and the authors are unusually clear about what their model leaves out.
Species-distribution models work best when a population has had time to settle into the full environmental range it can occupy. Neither group necessarily met that condition. Homo sapiens was expanding across Europe while Neanderthals were declining, so the archaeological points may show movement through a landscape rather than equilibrium with it. The amount of resulting uncertainty cannot currently be measured.
The models also omitted net primary productivity, a measure connected to the energy available through plants and food webs. Without it, the study cannot calculate carrying capacity or produce direct population counts. The paper discusses rough population ranges under its regional-band assumptions, but these are scenario estimates, not a census of Ice Age Europe.
Another limitation is literally hidden underwater. The analysis used modern coastlines. During colder periods, sea levels were lower and broad continental shelves were dry land. Leaving them out may distort possible connections across the Adriatic and between mainland Europe and what is now Britain.
Finally, a 15-kilometre grid and climate phases spanning centuries or millennia cannot recover the seasonal choices of one band, a brief local drought or an encounter in a particular valley. Archaeological sites are also an uneven sample. Some places preserved remains better than others, and some have been investigated far more intensively.
These limits do not make the exercise empty. They define what kind of evidence it supplies. The maps are a structured test of whether habitat loss and landscape connectivity could plausibly fit the archaeological pattern. They are not a replay of the extinction.
The last Neanderthal probably did not know they were last
In the end, the study replaces a dramatic question with a quieter one. Not “What killed the Neanderthals?” but “Why did fewer and fewer local populations recover?”
The answer suggested by the model is not that their world suddenly became unlivable. Suitable places endured. Yet some of those places were arranged in ways that made human connection harder, especially when climate shifts moved the best ground. Small, separated groups then faced environmental uncertainty, demographic chance and an expanding population of Homo sapiens whose regional geography may have been more forgiving.
A Université de Montréal account of the work describes connected populations as having access to a kind of safety valve. That is a useful image, as long as we remember that the valve is inferred from the landscape rather than observed directly.
No single cold snap, battle or invention need have ended the Neanderthals across a continent. Western groups may have faced more interaction with newcomers. Southeastern groups may have been more isolated. Iberian refuges may have held on longer. Interbreeding may have blurred some endings while local extinctions made others final.
The last Neanderthal population probably did not experience itself as the final chapter of a species. It was more likely one community among many, adapting to the next season as its ancestors had done for hundreds of thousands of years. What changed was that, eventually, no connected population remained from which another recovery could begin.