Celtic languages spread from central Europe, not the Atlantic, ancient DNA suggests

The Celtic languages that survive today in Ireland, Wales, Scotland and Brittany were probably not born on the Atlantic coast where they cling on. The largest genetic study of the question so far points instead to people who moved west out of eastern central Europe at the end of the Bronze Age, and whose ancestry turns up in every region where Celtic was later spoken.

That matters because the origin of Celtic is one of the longest-running arguments in European prehistory, and words, pottery and swords alone could never settle it. A team led from the University of Copenhagen has now tested the rival theories against the DNA of thousands of ancient Europeans, and only one of them fits both the genetic record and the clock that linguists use to date the birth of the language family.

The study, published in Nature, analysed genome-wide data from 4,587 ancient individuals. It concludes that a westward expansion linked to the Urnfield culture, roughly 3,200 to 2,800 years ago, is the most economical explanation for how Celtic reached France, the British Isles, northeastern Spain and Italy. “The question of where Celtic languages came from has been debated for more than 100 years,” said Guus Kroonen, a historical linguist at Leiden University and a co-author. “Ancient DNA gives us an entirely new way to evaluate competing theories and brings important new evidence into the discussion.”

The retreat is what makes the question so stubborn. Around 2,500 years ago, Gaulish was spoken from France to the Balkans, Celtiberian in the Spanish interior and Lepontic in the Italian Alps. Today the surviving Celtic languages have two or three million speakers, according to Nature’s news team, against more than a billion for Germanic languages such as English and German.

How they read migrations in shared DNA

The dataset includes 712 ancient genomes sequenced for the first time, combined with thousands of published ones. The hard part is that almost everyone in Bronze Age western Europe was closely related after steppe herders arrived from the east more than 4,000 years ago, so ordinary comparisons blur together.

To get around that, the team measured identity by descent: long stretches of DNA that two people share because they inherited them from a common ancestor. Each ancient individual received a “palette” describing how much of this DNA they shared with each of 386 genetic clusters in the dataset. The researchers then modelled every person as a mixture of earlier source groups, which shows when a new ancestry arrives somewhere and where it came from.

Early farmers in different parts of Europe also left slightly different genetic signatures, so a rising share of, say, the French-Iberian farmer signature in Britain works like a tracer dye for people arriving from the south. The main results were checked with three independent methods, including the Y chromosome passed from father to son.

Four theories were tested. One ties Celtic to the Bell Beaker culture of more than 4,000 years ago, “Celtic from the West” places its birth in the seafaring networks of the Atlantic Bronze Age, a third puts it in inland France, and “Celtic from the East” links it to the Urnfield culture and its successors, Hallstatt and La Tène.

Three migrations, and only one fits the clock

The genomes reveal three population movements that matter for Celtic. The first is the spread of steppe-related ancestry with Bell Beaker groups after about 4,400 years ago. The second is a large movement from southwestern Europe into Britain between about 3,800 and 3,200 years ago, which fits the Atlantic theory geographically. The third is the westward spread of ancestry from eastern central Europe between roughly 3,200 and 2,800 years ago, reaching France, Britain and Ireland, northeastern Iberia and Italy.

Timing is what separates them. Linguists estimate when Proto-Celtic, the undocumented ancestor of all Celtic languages, took shape and broke apart, and most of those estimates fall between about 3,600 and 2,800 years ago, with the strongest overlap between 3,200 and 3,000. The split between the Irish and Welsh branches is usually placed later, with the strongest overlap between about 2,800 and 2,500 years ago. The Bell Beaker expansion and the movement into Britain both come too early for most of those dates. Only the Urnfield-era movement lands inside the window.

In Iberia, the signal is clear. Ancestry associated with the Urnfield culture appears in almost all sampled individuals after about 3,200 years ago, and people buried in Urnfield contexts in the region already carry ancestry from the Hungarian and Serbian Bronze Age by about 3,150 years ago. The Urnfield culture brought new burial customs, fortified settlements, advanced metalworking and changes in warfare. “What is really exciting is that we now have genetic evidence of large-scale population movements that could previously be studied mainly through archaeological finds,” said Johan Ling, an archaeologist at the University of Gothenburg who worked on the study.

A companion study redraws the map of Germanic languages

The same team published a second paper alongside it, on the Germanic family, which later spread into many regions where Celtic had been dominant. It identifies a previously unknown population that entered Sweden, probably from the Baltic region, around 4,000 years ago and spread across Scandinavia within about five centuries. That group offers a possible vehicle for the arrival of early Germanic, about 800 years later than traditionally assumed.

By about 1,650 years ago, people from southern Scandinavia were moving south into what are now Germany, Poland and the Netherlands, areas that probably still spoke Celtic, and the same ancestry later appears among the Anglo-Saxons in Britain. A back-migration into Denmark and southern Sweden coincides with the emergence of Old Norse, the language of the Vikings. “People from northern Germany largely replaced the local population in Denmark and southern Sweden. But they also brought with them a new religion, with Odin as the principal god, and a new material culture inspired by the Huns,” said Kristian Kristiansen, an archaeologist at the University of Gothenburg. The Urnfield expansion that carried Celtic west, by contrast, left little trace in Scandinavia.

What the genomes cannot prove

DNA records ancestry, not speech. “You can easily adopt a new language without changing your genes,” Kroonen told Nature. Eske Willerslev, the senior author, makes the same point: “DNA sequences do not tell us directly which language people spoke. But by combining ancient DNA with linguistic and archaeological evidence, we can begin to trace how languages spread and changed over thousands of years.”

The key period is also the worst sampled. Urnfield communities cremated their dead, so very few genomes survive from between about 3,200 and 2,800 years ago, and the movement is largely inferred from the people who lived before and after it. The genetic impact also weakens the farther west it goes. In Iron Age England, the largest share of ancestry still traces back to the Bell Beaker population, which means a language may have spread with a minority of newcomers rather than through wholesale replacement.

The authors stress that their preferred model is the simplest of the four, not a proven one. The Bell Beaker theory also requires few separate migrations, and history does not always take the most parsimonious path. If related dialects were already spoken across western Europe, Celtic may have spread partly by diffusion between neighbours who could understand each other. A later eastward spread of Gaulish during the La Tène period shows up only as a faint genetic signal that needs confirmation. Co-author John Koch of the University of Wales Centre for Advanced Welsh and Celtic Studies co-edited the book series that gave the Atlantic theory its name, and the paper leaves room for parts of it.

Common questions about the origin of the Celtic languages

Where did the Celtic languages come from?

Nobody knows for certain, but this study favours eastern central Europe. A westward migration linked to the Urnfield culture, roughly 3,200 to 2,800 years ago, best matches both the ancient DNA and linguists’ dates for the birth of Proto-Celtic. The authors present it as the most economical explanation, not a final answer.

Did the Celtic languages start in Ireland?

Probably not. Around 2,500 years ago Celtic languages were spoken from Spain to the Balkans. Ireland and Britain are where the family held on longest, not where the genetic evidence suggests it began.

What was the Urnfield culture?

It was a Late Bronze Age culture of central Europe named after its custom of burying cremated remains in urns in large cemeteries. That cremation habit is why so few genomes survive from the period.

Which Celtic languages are still spoken today?

Six survive in two branches. The Goidelic branch includes Irish, Scottish Gaelic and Manx, and the Brittonic branch includes Welsh, Breton and Cornish. Manx and Cornish were revived after losing their last native speakers.

What comes next

The two papers appeared in Nature on 4 November 2026, after the Celtic study was first posted as a preprint in February 2025. The authors say the next step is denser sampling from Bronze Age northern Italy and central Europe, the regions where the Urnfield-related ancestry took shape. Those genomes could also explain why Celtic shares so much with the Italic languages, the family that later produced Latin.

Reference: McColl et al., “Tracing the spread of Celtic languages using ancient genomics,” Nature, 2026. DOI: 10.1038/s41586-026-11198-9

Tags: archaeology, ancient DNA, Bronze Age, Celtic languages, Eske Willerslev, Germanic languages