Substack cometh, and lo it is good. (Pricing)

The Anglo-Saxonization of England happened through a mass migration

The Anglo-Saxon migration and the formation of the early English gene pool:

The history of the British Isles and Ireland is characterized by multiple periods of major cultural change, including the influential transformation after the end of Roman rule, which precipitated shifts in language, settlement patterns and material culture…The extent to which migration from continental Europe mediated these transitions is a matter of long-standing debate…Here we study genome-wide ancient DNA from 460 medieval northwestern Europeans—including 278 individuals from England—alongside archaeological data, to infer contemporary population dynamics. We identify a substantial increase of continental northern European ancestry in early medieval England, which is closely related to the early medieval and present-day inhabitants of Germany and Denmark, implying large-scale substantial migration across the North Sea into Britain during the Early Middle Ages. As a result, the individuals who we analysed from eastern England derived up to 76% of their ancestry from the continental North Sea zone, albeit with substantial regional variation and heterogeneity within sites. We show that women with immigrant ancestry were more often furnished with grave goods than women with local ancestry, whereas men with weapons were as likely not to be of immigrant ancestry. A comparison with present-day Britain indicates that subsequent demographic events reduced the fraction of continental northern European ancestry while introducing further ancestry components into the English gene pool, including substantial southwestern European ancestry most closely related to that seen in Iron Age France

1) More migration than earlier papers. Looks like increasing ancient DNA coverage helped
2) 75% Y chromosomal turnover in eastern England
3) A third component, detected in the PoBI paper, is confirmed, and seems related to continuous later gene flow from northern France. This is ubiquitous across England, and I do wonder now what the Norman Conquest and the unification of large regions of northern France with England did in the early medieval period

11 thoughts on “The Anglo-Saxonization of England happened through a mass migration

  1. Finally out!

    As the paper gave dates for all the samples (albeit typically not radiocarbon dates it seems), and Davidski at Eurogenes gave Global25 preliminary coordinates for them (may change slightly after paper released), I did a plot comparing the distance on G25 (should correlate to real genetic distance) to present-day England over time for ancient samples: https://imgur.com/a/bj9D5Z8

    It really does look like over time there’s a general fall of distance of ancient people to present-day people in the same place (with a punctuated event at the LBA-IA transition that leads to a quicker fall in distance)… until we get to the Saxon Era and Viking samples and the distance pushes up again. The Saxons look more differentiated from present-day English than the Iron Age people do, on this measure! And there’s no apparent bounceback between the Anglo-Saxon samples and the Viking Age samples.

    I think the contention in this paper is that the sort of bounce-back after this towards a more Iron Age like profile – though more EEF rich – is predominantly due to migration of people associated to French_IA like ancestry. That seems possible though I have caveats:

    1) French_IA people may not have been around in quantity by this time; on their f3 statistic derived plot that looks to break out these trends (https://imgur.com/a/guW2ijF), the present-day French are south and east (Germanic) shifted compared to the IA groups. The French_EMA is somewhat uncharacterised, except for the preprint of Isabel Alves (https://www.biorxiv.org/content/10.1101/2022.02.03.478491v1).

    2) Their samples here are vastly predominantly from the early Middle Ages (see Fig 1 from the paper). This is great as it shows the pulse really clearly, but does mean that the samples we have from England from after about 750 and before 1066 (the ones I used in my graphic) are predominantly from the sites that Margaryan sampled to characterise the genetics of the Vikings.

    That does have the problem that these sites might be enriched for recent Scandinavian ancestry. (Which they note in their sample set that “Already during the Early Middle Ages, several individuals from multiple sites exhibit modest degrees of excess affinity (5.4%) to present-day individuals from the Scandinavian peninsula (Supplementary Fig. 6.2a), indicating additional sources. Although close cultural contacts to the Scandinavian peninsula are attested in the archaeological record, we did not find this genetic variation to be geographically stratified within early medieval England (Supplementary Fig. 6.2b). This Scandinavian Peninsula-related ancestry increases substantially (to 30.6%) only during the Viking period (Supplementary Note 6).”).

    Also, I would note that the highest figure of replacement in the EMA Anglo-Saxons (85%) comes from choosing the Lower Saxon groups with more EEF ancestry as the reference. In their modelling in the supplement, you can see that groups from Mecklenburg in North Germany and from Denmark can model with figures more like 68-65% replacement. (Supplementary Figure 3.9). These are all working models. (Some models can get >40% continuity even).

    There’s actually a reason to prefer the sort of figure around 66% IMO because when they look for the England EMA samples that cluster closest to Central-North Europeans (look most unadmixed), in the f3 analysis they seem to indicate that those samples are closest to the samples from Mecklenburg (who would model as about 68% of the ancestors of England_EMA).

    So if the migration was more “Anglo” and less “Saxon” (and I would guess it was from the above), with more Jutes and Angles and fewer Saxons, British survival and contribution in the EMA was probably a little bit higher than the headline figure. I think they fixate a bit too much more on the lower Saxons in their models than seems warranted by the actual most CNE rich of their samples from England_EMA, who potentially seem more to resemble what Angles would have been like.

    They also find that the samples from Lower Saxony has a different relationship to the Latvia_BA samples that are heavy on “Balto-Slavic drift” in Davidski’s tools, compared to England_EMA_CNE (I also found this caused problems when using the Lower Saxons as ancestors for England_EMA in the models I tried in Global 25).

    Nevertheless, although I don’t think this paper totally closes the door on the exact degree of initial replacement and whether there was more integration in the later Anglo-Saxon period (genomes from Wessex in later AS period might be interesting?), I do think it’s pretty clear that the initial EMA population was at least 2/3 derived from this CNE heartland!

  2. Another comment on the broad y-dna patterns here.

    A recent paper on the UKBiobank gave y-dna haplogroups for all 169,477 White British samples, across the whole UK. Then this paper and Patterson and Isakov’s paper give us data for ~100-140 males each from Iron Age England and EMA England.

    If we then compare and contrast these, on the assumption that the 17% of the UK that is not from England would be expected to be continuous with Iron Age England and the 83% from England would be continuous with England_EMA: https://imgur.com/a/fhlS6tg

    Broadly what we’d find is that I1a is lower in present day Britain than we’d expect from the EMA and IA sample and the respective assumptions of continuity, while R1b and the non-R1b/I1a clades are higher than expected from that.

    That kind of supports the paper finding some discontinuity between England_EMA and present day England in the direction of more enrichment of R1b and rarer haplogroups, and a decline of the distinctively Germanic I1a. (I.e. it’s what we’d expect from some enrichment from the direction of France/Western British Isles). Albeit sample sizes of 100 are still pretty small!

  3. So, earlier reports of Germanic ancestry being around 50% at most in coastal Saxon Shore/Danelaw areas and fading away toward Wales and Cornwall to very little were less accurate?

    So now the traditional historiography (the Saxons displacing the Brythonic speakers) appears more correct?

    Things keep changing!

    What are the implications of intrusive females appearing to be high status (burial goods), but males with weapons being locals?

  4. @Matt, some good caveats there (and impressive you already looked it over so well).

    An obvious one I had in mind before the paper itself was sorta addressed by them I notice with a first look: “One potential caveat in this analysis is our relatively sparse Roman sample from England, where we particularly lack samples from the south, which might have pre-existing France IA-related ancestry. We, therefore, turned to one of our early medieval sites, the post-Roman cemetery of Worth Matravers at the southern coast of Dorset, whose individuals have nearly no CNE ancestry (less than 6% on average), and thus may serve as a more temporally close proxy for post-Roman Britain before the arrival of CNEs. When used as a source in our model, we found that the estimates of France IA-related ancestry in present-day England changed by less than 3% on average across the regions (Fig. 5b), suggesting that France IA-related ancestry entered England to a substantial amount after the Roman period.” though we might want some more samples, still, for a clearer picture.

    The general model that has been looking to be the case for a while (IA/Roman + Anglo-Saxon-Scandinavian + something continental) is shown in the paper using some really nice ways, though the *very* specifics/proportions are still I think quite arguable as you laid out so well. Some interesting shifts in Scandinavia, Germany and the Netherlands visible in their Supplementary Figure 3.6 too. I havent’t noticed them commenting on those so far, to read what they think, but it isn’t the topic of the paper anyway.

    Re: the Latvia_BA affinity, what do you make of “Noticeably, this is not the case for (…) later populations from Latvia or contemporaneous Bronze Age populations from Lithuania and Estonia”? Does seem to be the only population comparison with somewhat significant Z there but why not the other Bronze Age Baltic groups too (maybe the lower coverage for the similar Estonian BA set that doesn’t impact G25 as much in finding that affinity?)?

  5. Thanks Forgetful; in fairness I had some time to prepare and know what I’m looking for and it’s also entirely possible I might have skimmed things. I also think this one when looking at samples may benefit from the y sub-clade knowledge that I don’t have.

    Re; Latvia_BA, yeah, that’s a good point; I re-examined it, and you’re right, you would expect that to hit those other populations if it’s about the Latvia_BA drift. The Lithuania_BA is in the same direction but of a lesser magnitude and beneath significance (https://imgur.com/a/Oos0Iw3). The Latvia_BA is generally particularly more displaced in that direction in stats/PCA than Lithuania_BA (more drift and/or HG ancestry), but I don’t know if that’s enough to justify it.

    Another thing that may interest you; I took the proportions of CNE ancestry in the England_EMA samples from the supplement, separated them into groups according to the proportion, and then made averages of their Global_25 coordinates. This is the results: https://imgur.com/a/ql9wee2

    Generally it does seem like the G25 averages pretty closely follow their estimates, where i.e. the average from the ones assigned to low CNE ancestry group are matching closest to England_IA, and the ones from the 90-100% or 100% averages I made are very close to DEU_Mecklenburg_LA_EMA (the Haven samples, who should represent the earliest and perhaps most viable unadmixed source in this new set of CNE samples). The highest CNE ancestry England_EMA in their average placement do look very close to the Haven samples, so unless they’re using a subset of the Lower_Saxony samples or something went wrong with the Global25 data, it does look like that’s a more viable solution still, to me.

    On the other hand, looking at them as individuals, the placement is a total chaos on the dimensions that should split these populations apart. I don’t think you could reliably recreate their assignment to WBI / CNE ancestry from G25+Vahaduo. So it looks like either the G25 is not splitting this Celtic-Germanic stuff effectively, or the f4 stat used in the paper is somewhat confounded in some sense (that looks unlikely based on everything they’ve done though).

    Oddly the samples they assign high proportions of WBI ancestry (in the 2 way model) to seem to have slightly more Anatolian ancestry than the average for Iron Age England though, judging by the placement. This seems to lead to them placing slightly closer to Bretons than Welsh for’ex. This may have something to do with why they think some French_IA related ancestry was also entering Southern Britain, when they start to look at 3 way modelling these EMA samples.

  6. @Forgetful, may interest you, some thoughts around comparisons of a model I made for G25 and the f4 stat and proportions in the paper for the EMA_England samples: https://imgur.com/a/LdIUtZ1

    In general there seems like a pretty high correlation between the assigned DEU_EMA ancestry and Iron Age England in the G25 model, particularly at levels of averages, although this range of ancestry is compressed in the G25 model.

    However the G25 model also correlates worse with the f4 stat (CNE,WBI;X,YRI.SG) and adds little to no independent predictive value to the f4 stat when both it and the assigned CNE proportion from the paper are put into a multivariate model (i.e. its not like the G25 model is finding samples with a high f4 stat that were missed by the paper’s supervised ADMIXTURE).

    One other thing though, which seems potentially rather major, is that there seems to be zero relationship between the assigned CNE fraction or the f4 stat and the paper’s assigned steppe ancestry fraction.

    That’s slightly reassuring that the WBI vs CNE is not confounded by steppe, but it’s also slightly worrying for the paper’s conclusions!

    The paper is sort of concluding here that there’s a two-way model for England_EMA between local post-IA British, who are slightly but noticeably lower steppe, and a more steppe rich CNE population.

    But this seems like a problem if there isn’t actually a relationship between steppe ancestry and CNE proportion in their models…

    It may suggest that the CNE proportion and WBI proportion assigned by their admixture, rather than actually capturing a single CNE source population and Iron Age Brits from the area of England, are actually just covering continental vs insular sources generally.

    That doesn’t push back on the scale of the migration, but might actually suggest that it’s larger than thought, and also from a wider range of sources, from steppe-poor regions in the continent and steppe-rich regions of the Isles (Ireland?), thus leading to no relationship between steppe ancestry and CNE:WBI. That’s a “Migrationist” result still, but it’s perhaps one that would lead to “Kossina’s Frown” (because it’s finding a folk migration by a people with a homogenized level of steppe ancestry…). (I wonder if this paper would’ve benefitted from a more Patterson+Isakov approach with more consideration on small variances in steppe ancestry…)

    Alternatively, their model for assigning steppe ancestry is just not very good; of course you might get intra-individual differences but on average there should be some covariance overall, if its working at all… (Also note here that this is looking at a rather narrow range of steppe ancestry, as I’ve already excluded the samples that G25 modelled as well outside the NW European range of Steppe:WHG:EEF.) I don’t massively think this is definitely the case though, as when I use a G25 derived steppe ancestry estimate, while this correlates much better with the paper’s CNE estimate than the paper’s CNE estimate, it’s still low. I might look at this with a wider range of samples, as they’ve included f4 and ancestry estimates (WBI vs CNE) for England_IA too.

  7. @Matt

    Lots of interesting stuff from you to go through, here and on Eurogenes. Thanks a bunch!

    By the way, if you care to weigh in, an interesting question about their use of |Z|>2 in parts of the paper has come up on anthrogenica (e.g. “None of the England_EMA individuals is significantly (Z > 2) closer related to NOR than to CNE.”). The generally accepted cutoff, as far as I know, is |Z|>3 but various studies in the past have also utilized |Z|>2 at times for some results they also considered significant enough to mention from recollection. Since you seem to have a relevant background/understanding, what do you think of that difference in choice?

    Maybe Razib could weigh in on that if he catches the question and has an opinion on that issue as well?

  8. @Forgetful; Is there any more context on that one? Is it in regards to the Scandinavian affinities of the Viking Age samples? Otherwise I’m not sure which samples we would expect to have signficant Z scores comparing f4(NOR,CNE;X,YRI.SG) and why it would matter…

  9. @Matt

    My bad. I’m referring to this part in the supplement: https://postimg.cc/k296dWps

    Though my question is more general too. Results in some studies with a cutoff of |Z|>2 are considered sometimes [an example that came to mind from Lazaridis et al. Minoans-Mycenaeans: “When we apply this intuition to the best models for the Mycenaeans(Extended Data Fig. 6), we observe that none of them clearly outperforms the others as there are no statistics with |Z|>3 (Table S2.28). However, we do notice that the model 79%Minoan_Lasithi+21%Europe_LNBA tends to share more drift with Mycenaeans (at the |Z|>2 level).” though I’m pretty sure this has been the case in other ones too that don’t come to mind right now] but the more generally considered cutoff seems to be |Z|>3.

    What do you make of the difference in choice in general, if you have an opinion on it? Does it depend on some other specifics which one is a still appropriate choice or is the lower cutoff choice a more tentative, though still significant result to at least consider in general?

Comments are closed.