On 25 July 2026, Robert John Langdon published a post arguing that the Bell Beaker migration model rests on a statistical illusion. Its central evidence is a table of 43 pre-Bell Beaker R1b individuals compiled, he says, from the Allen Ancient DNA Resource, together with a Mesolithic R1b individual from Aveline's Hole in Somerset said to move the starting point of British population history back by several millennia.
I have re-run the query against the source file. The table can be checked, and it does not hold. The Aveline's Hole claim has a specific and traceable explanation, and it is not the one offered.
What follows is the audit. The dataset is the AADR 1240k annotation file, release v66.p1, which anyone can download and reproduce.
What the post gets right
It is worth being clear about this first, because the post's opening three sections are sound and the rebuttal does not depend on disputing them.
Ancient DNA is fragmentary. Genomes are partially reconstructed rather than read. Coverage varies by orders of magnitude between individuals who appear side by side in a published table. Radiocarbon results are probability distributions, not dates. Bayesian chronological models are conditional on their priors, and a published range can be more precise than the underlying measurements alone would justify. Databases are samples, not censuses, shaped by preservation, excavation history and research priorities.
All of that is true. None of it is news — these caveats appear routinely in the supplementary material of the papers the post is criticising — but stating them for a general readership is a service rather than a fault.
The post is also correct that R1b existed in Europe before the Bell Beaker period. That has been published, mainstream, uncontroversial data since 2017.
The difficulty is what is built on top.
The category error
"R1b" is used throughout the post as though it named a population. It does not. It names a macro-haplogroup roughly eighteen to twenty thousand years deep, with two primary branches, and the distinction between its subclades is the entire substance of the question at issue.
The individual from Villabruna in northern Italy, dated to about 14,000 years ago, is R1b1a-L754. The Iron Gates Mesolithic individuals from Serbia and Romania are L754-derived. The Balkan Chalcolithic individuals from Pietrele, which supply eight rows of the post's own table, are basal or V88-type lineages. Blätterhöhle in Germany is V88.
None of these are ancestral to R1b-M269 > L51 > P312, the clade that expands across north-western Europe after 2500 BCE and that accounts for the overwhelming majority of modern British paternal ancestry. They are collateral branches. Their presence in Mesolithic and Neolithic Europe is a well-known fact that no researcher has ever disputed, and it bears on the Beaker question in roughly the way that the presence of wolves in Pleistocene Europe bears on the origin of the Labrador retriever.
The post's argument requires the reader not to notice this. Once noticed, the argument does not survive it.
There is a nomenclature trap here that catches honest readers too. AADR haplogroup strings have shifted across ISOGG releases, and "R1b1b" has denoted different nodes at different times. Anyone working from the strings rather than the primary publications will find them genuinely confusing. That is a reason for care, not a defence of the conclusion.
The chronological conflation
The post treats "pre-Bell Beaker" as though it meant "pre-steppe." It does not, and the gap between the two is where a fifth of its evidence sits.
Nine of the 43 are Bohemian: Plotiště nad Labem, Obříství, Vliněves, Stadice, Konobrže, dated between roughly 2919 and 2636 calBCE and carrying P312-derived subclades. These are Corded Ware burials.
Papac and colleagues, publishing in Science Advances in 2021, reported that Corded Ware appeared in Bohemia by about 2900 BCE and that R1b-L151 was already the most common Y-lineage among the earliest Corded Ware males there — six of eleven — with P312 most likely diversifying from within that pool. The same study documents the subsequent volatility with some precision: P312 rises to complete fixation in late Bell Beaker Bohemia and then falls to twenty per cent in preclassical Únětice, implying a minimum eighty per cent influx of new Y-lineages at the onset of the Early Bronze Age.
The Bohemian individuals in the post's table are therefore not counter-evidence to the steppe migration. They are its evidence, dated and labelled, filed under the wrong heading.
Aveline's Hole
Chapter 4 of the post rests on the identification of an R1b lineage from Aveline's Hole dating to the Mesolithic. No such individual appears in the post's own appendix. The only two British entries are Neolithic.
The sample exists. It is I3004, from Aveline's Hole, Burrington Combe, directly dated to 8606–8300 calBCE (OxA-34339), published in Brace et al. 2019. Its Y-haplogroup in AADR is R1b1a1b1a1a2a5a~, terminal SNP R-Y11281.
Its assessment in AADR is CRITICAL.
The hapConX X-chromosome contamination estimate is [0.212, 0.364] — between twenty-one and thirty-six per cent. The documented threshold at which AADR marks a sample CRITICAL or FAIL is a lower bound above 0.03. This sample exceeds that by a factor of seven. The second warning on the record is high.popgen.heterozygosity, the other standard signature of contamination. The individual carries 112,612 SNPs on 1240k targets.
The Y-call itself is the tell. R-Y11281 sits deep within P312 — which is what the modern British population overwhelmingly carries. A low-coverage British Mesolithic sample with a fifth to a third of its X-chromosome reads coming from somewhere else, returning a deeply derived P312 subclade, is not a discovery about the Mesolithic. It is a description of laboratory contamination, and the curators recorded it as such.
For completeness, the other Aveline's Hole individuals in AADR are I3006 (Mesolithic, female), I3005 and I3010 (both Early Neolithic, both female). This matches Brace et al. 2019 and the subsequent site-specific reassessment in the Proceedings of the University of Bristol Spelaeological Society: four individuals yielding genomic data, two Mesolithic with the expected Western Hunter-Gatherer signature, two Early Neolithic with Aegean farmer ancestry, separated by nearly five millennia and identified only when the crania were directly dated.
That last detail deserves emphasis, because the post has the moral of the story backwards. Aveline's Hole is the textbook case of undated cave material proving to be thousands of years younger than assumed. It is an argument for stringent chronological control, not against it.
The count
The post states that it examined every published prehistoric male dated before 2500 BC in the AADR, and reports 1,351 individuals of whom 43 are R1b, an observed frequency of 3.18 per cent.
I cannot reproduce those figures. Deduplicating on Individual ID as the AADR README instructs — the same individual can appear under several Genetic IDs, and counting rows inflates everything — and taking ancient males with a mean date at or before 2500 BCE:
|
|---|
The fourth row is the closest approximation I can construct to the post's denominator. Restricting to non-steppe Europe gives 1,452 males, near enough 1,351 that something of this kind was probably done.
That same filter contains 111 R1b individuals, not 43.
Sixty-nine are missing from a table captioned as listing all confirmed pre-Bell Beaker R1b individuals: six Iron Gates Mesolithic individuals from Serbia, eight from Latvia across the Kunda and Narva groups, twelve from Bulgaria including four from Varna, nine from Denmark, three further Czech Corded Ware, and others from Spain, Italy, Germany, Hungary, Slovakia, Sweden, Switzerland, Poland and Estonia.
The observed frequency is therefore not 3.18 per cent under any reading I can construct. It is 7.6 per cent on the post's own apparent geography and 16.3 per cent for Europe as a whole.
This does not support the post's argument. It damages it. The bulk of pre-2500 BCE European R1b sits in Yamnaya burials from Samara, Kalmykia, Orenburg, Rostov and Moldova, in Khvalynsk and Ekaterinovka, in Afanasievo, and in Corded Ware. Making the number larger makes the steppe expansion more visible, not less. The post's central statistic understates its own database by a factor of two and a half, and correcting the error strengthens the model it was assembled to refute.
The arithmetic that follows
Even taking 3.18 per cent at face value, the extrapolation to "approximately 8,000 to 16,000 R1b individuals" does not work.
A Y-haplogroup frequency is a frequency among males. Applying it to a total population of 250,000 to 500,000 counts everybody twice; the internally consistent figure would be roughly 4,000 to 8,000. The denominator spans about 7000 to 2500 BCE, mostly Neolithic and Chalcolithic, while the population estimate is Mesolithic — and European population grew by an order of magnitude across that interval, so combining them is not conservative but incoherent. Those population estimates themselves span more than a factor of two in the published literature and rest on sparse ethnographic analogy and climate-envelope modelling.
More fundamentally, multiplying any non-zero frequency by a large population yields a large number. That establishes nothing about descent. The Beaker model is not a claim that no R1b existed before 2500 BCE. It is a claim about which subclade expanded and left descendants. Since most of the 43 belong to lineages with negligible modern Western European paternal descent, the headcount is irrelevant to the proposition it is deployed against.
The post also invokes sampling bias in one direction only. If the database undercounts R1b males, it equally undercounts I2a, G2a and everyone else. A frequency is a ratio, and bias in the denominator does not preferentially inflate the numerator.
The 43
All 43 identifiers exist in v66.p1. The table is not fabricated, and it is worth saying so.
On AADR's own annotations, however:
- Eight carry
ASSESSMENT = Questionable: I6912, PNL001, I18101, JK2804, ATP3, I3035, I2611, OC. - Thirteen have contextual dates only, with no direct radiocarbon determination. The AADR file distinguishes these explicitly; the post's table preserves the distinction in its "cal BCE" versus "BCE" notation without remarking on it.
- Two carry the outlier group label
England_N-o— both British entries. - VLI011 and VLI015 are father and son, recorded as a first-degree pair. Two of the 43 are one observation. KON003 has two second-degree relatives at the same site; I14176 has a 2.5-degree relative.
- Twenty-five both pass AADR quality control and are directly dated. Twenty-four carry the literal
Passvalue; the twenty-fifth, I1590 from Blätterhöhle Cave, isMERGE_PASS.
"Forty-three confirmed" is not a description the source file supports. The point matters because the post's rhetorical structure depends on accumulation — that independent discoveries across many countries and laboratories cannot all be anomalies. Kinship, QC flags and undated contexts all reduce the number of independent observations.
The two British entries deserve individual attention, since they are the ones bearing on Britain.
I2611, from Summerhill, Blaydon, directly dated 3092–2905 calBCE and called R-L21, is the single most striking item in the table. Its AADR warning field reads: carbon date is unexpected for the archaeological context and genetic profile, with hapConX at [0.016, 0.038]. It is grouped as England_N-o, an outlier, and published in Patterson et al. 2021. The curators identified the anomaly before the post did, and recorded their view of it.
I3035, from Fox Holes Cave, has a contextual date only of 4000–3500 BCE and a U106-derived call. Its warnings record ANGSD contamination at [0.015, 0.028], hapConX at [0.014, 0.022], a library with technical problems, and that the observed and expected predicted dates differ by more than a thousand years. Another cave assemblage, another chronological mismatch.
What the evidence actually shows
The Beaker model does not rest on Y-haplogroups. It rests on genome-wide ancestry proportions, and the post does not mention them once.
Olalde and colleagues, in Nature in 2018, estimated that around ninety per cent of Britain's gene pool was replaced within a few centuries of the Beaker horizon, on the basis of autosomal ancestry. The alternatives the post proposes in its seventh section — gradual admixture, regional survival, cultural diffusion, multiple episodes — are not alternatives to the mainstream picture; they are components of it. Papac's Bohemian study documents Corded Ware males assimilating females of diverse local backgrounds. Work on Bronze Age Orkney has shown Neolithic I2a lineages persisting locally long after Beaker-derived ancestry arrives elsewhere in the genome.
But none of the four alternatives explains the observation the model was built to explain, and one of them cannot in principle. Cultural diffusion does not transmit autosomes. Pottery styles travel without people; steppe ancestry does not.
There is a final irony in the post's framing. It presents the Beaker migration model as an entrenched orthodoxy of more than two decades, self-reinforcing and resistant to revision. The model dates from 2015 and 2018. Before roughly 2010 the dominant view held R1b to be a Palaeolithic or Mesolithic Western European lineage — close to the position the post presents as suppressed heterodoxy. The consensus moved because the evidence moved, which is precisely the process the post claims does not occur.
Method, and where I might be wrong
The audit uses the AADR 1240k annotation file, release v66.p1, dated 8 June 2026. The post cites "AADR v66.1," which is not a version string the project has issued; the sequence runs v66.0 (12 April 2026), subsequently decommissioned, then v66.p1. The difference between those two releases concerns 161 modern Papuan individuals and cannot affect any count of prehistoric European males, so nothing substantive turns on it.
Two filter judgements are mine and are contestable. I deduplicated on Individual ID rather than counting rows, and I used the mean-BP date field rather than range endpoints. Both follow the README's guidance, but neither is the only defensible choice, which is why the sensitivity table above shows several variants rather than a single number. The gap between 43 and 111 survives all of them.
I could not reproduce 1,351 exactly, and it is possible I have misread the filter that produced it. The remedy is straightforward: publish the query. A count presented as a direct read of a public database, in a post whose thesis is that archaeologists hide their assumptions inside models, should be reproducible by anyone who downloads the file.
Sources
- Brace, S. et al. (2019) Ancient genomes indicate population replacement in Early Neolithic Britain. Nature Ecology & Evolution 3, 765–771.
- Schulting, R., Booth, T., Brace, S. et al. (2019) Aveline's Hole: an unexpected twist in the tale. Proceedings of the University of Bristol Spelaeological Society.
- Olalde, I. et al. (2018) The Beaker phenomenon and the genomic transformation of northwest Europe. Nature 555, 190–196.
- Papac, L. et al. (2021) Dynamic changes in genomic and social structures in third millennium BCE central Europe. Science Advances 7, eabi6941.
- Patterson, N. et al. (2022) Large-scale migration into Britain during the Middle to Late Bronze Age. Nature 601, 588–594.
- Mathieson, I. et al. (2018) The genomic history of southeastern Europe. Nature 555, 197–203.
- González-Fortes, G. et al. (2017) Paleogenomic evidence for multi-generational mixing between Neolithic farmers and Mesolithic hunter-gatherers in the Lower Danube basin. Current Biology 27, 1801–1810.
- Fu, Q. et al. (2016) The genetic history of Ice Age Europe. Nature 534, 200–205.
- Haak, W. et al. (2015) Massive migration from the steppe was a source for Indo-European languages in Europe. Nature 522, 207–211.
- Dulias, K. et al. (2022) Ancient DNA at the edge of the world: continental immigration and the persistence of Neolithic male lineages in Bronze Age Orkney. PNAS 119, e2108001119.
- Allen Ancient DNA Resource, v66.p1 (8 June 2026), 1240k annotation file. Harvard Dataverse.
No comments:
Post a Comment
Comments welcome on fresh posts - you just need a Google account to do so.