Tuesday, March 13, 2018

First real foray into Migration Period Europe: the Gepid, Roman, Ostrogoth and others

This is going to be our first meaningful look at the all important Migration Period, thanks to the recently published Veeramah et al. 2018 paper and accompanying dataset (see here). The Migration Period is generally regarded to have been the time when present-day Europe first began to take shape, in a rather sudden and violent way, with, you guessed it, a lot of migrations taking place.

Here's where most of the ancients from Veeramah et al. 2018 cluster in my Principal Component Analysis (PCA) of ancient West Eurasian genetic variation. Those East Germanics (the Gepid and Ostrogoth) are certainly very eastern, and indeed more exotic than I would've ever expected them to be. But I do love surprises like this. The relevant datasheet is available here.

Obviously, as per the paper, the ACD in about half of the labels stands for Artificial Cranial Deformation. I've also updated my Global25 datasheets with many of the same ancients. You can use these datasheets to plot them on 2D or 3D "genetic maps", and model their ancestry proportions. Feel free to share your findings in the comments below.

Global25 datasheet

Global25 datasheet (scaled)

Global25 pop averages

Global25 pop averages (scaled)

Here are a few of my own models for some of the more interesting of these individuals, using nMonte3 and based mainly on Iron Age (IA) reference samples. I used the same data file for all of the models; it includes scaled coordinates and is available for download here.

[1] distance%=3.7819



[1] distance%=3.6339




[1] distance%=2.5535




[1] distance%=2.9444



The Gepid and Ostrogoth show significant Scythian- and Armenian-related ancestry proportions, respectively. Should that be taken literally? Or do we have to wait for, say, Avar and Hunnic genomes to expect more realistic models?

Update 15/03/2018: This is where many of the Medieval German samples cluster in my PCA of modern-day Northern European genetic variation (see here). Obviously, I could only run the individuals with wholly or overwhelmingly North European genomes, and most of these turned out to be the males without any signs of ACD. They look very West Germanic. The relevant datasheet is available here.

See also...

Modeling genetic ancestry with Davidski: step by step


  1. Maenchen-Helfen reported (world of the huns?) that the gepid burials of pannonia contained mongolian,siberian, and even sinid skulls.

  2. Germany_Medieval_outlier:STR_300

    Sicilian_East 5.083045
    Italian_South 5.179631
    Mycenaean 5.364110
    Greek 5.470509
    Sicilian_West 5.614518
    Maltese 5.811958
    Italian_Tuscan 5.875314

    [1] "distance%=3.5662"

    "Anatolia_BA:I2683" 34.45
    "Balkans_ChL:I2425" 31.55
    "Peloponnese_N" 14.5
    "Yamnaya_Samara" 13.55
    "LBK_EN:I2032" 3.8
    "Levant_BA" 2.15

    (Another east-med individual after the ones in the Langobards paper, weird! Byzantine?)


    Spanish_Baleares Spanish_Cataluna
    2.403037 2.657032
    Spanish_Murcia Spanish_Cantabria
    2.766374 2.908718
    Spanish_Andalucia Spanish_Castilla_La_Mancha
    2.940393 3.080841
    Hungary_BA Spanish_Aragon
    3.084510 3.128334

    (Probably he was Decimus Maximus Meridius)

  3. After the hun period the step is gonna be a wild mix of pretty much every kind of Euroasian genomes.

  4. What to expect if they concentrated on outliers and deformed skulls? More interesting is the Bavarian case since the intermarriages with local Southern shifted women might have been fairly common on a more regular base.

  5. Could the similarity to southeastern Europeans in the ACD women suggest that pre-Germanic Bavarians were more closely related to southeastern Europeans?

  6. Thats doubtful because Celtoromans are not known for that practise which was a rather Turkic/Hunnic fashion.
    Of course all kind of people under Turkic cultural influence might have adopted it at some point. But I would be careful. We definitely need more samples for a solid diachronic analysis.

  7. @Steven

    No. Bavaria would have looked like Central European Bell Beaker, which is basically Celto-Germanic. The trend we are seeing is that it wasn't a problem choosing female mates outside the core tribe.

  8. ACD was adopted by several Germanic peoples, Goths obviously, Gepids, Heruli, but also Burgundians (hence ACD women found in Eastern Central France: Burgundy, Forez....).
    Alemanni seem to have also adopted this practice for a short time. And it existed within the Franks, but it's rare.
    Often, ACD women found in Northern Gaul have been found with oriental style grave goods.

  9. @Elliv J

    The males are *not* admixted. See fig.3. I think we see political mariages here.

  10. There are also two new papers on Iberia


  11. The Roman soldier was probably from Aquitaine in South France/North Spain. He clusters with modern Basque & south French.

  12. Re: Crimean Ostrogoth, I've tried another procedure, which should allow using nMonte in a semi-supervised way to get at the target's ancestry.

    Take Crimean Ostrogoth, then take present day / ancient samples, then simulate a point in a vector away further along the vector of Crimean Ostrogoth vs ancient / present day samples.

    So: https://pastebin.com/04AesCfK

    The first two of these are Ukraine_Ostrogoth_ACD:KER_1_(Non_Germanic_Ancestry_1_25%) and Ukraine_Ostrogoth_ACD:KER_1_(Non_Germanic_Ancestry_2_25%); these "subtract" 25% of Swedish:Norwegian average and Anglo_Saxon:Nordic_IA respectively. Results look fairly South Caucasian / Armenian / Assyrian, with possibly less CHG affinity (the points are fairly different in some dimensions).

    Third one is Ukraine_Ostrogoth_ACD:KER_1_(Non_Greek_Trab_Ancestry_75%), which takes away 75% Greek_Trabazon ancestry from the Crimean Goth. This seems to get a position that looks fairly close to Germanic peoples in the main West Eurasian dimensions, but looks to have a strong tilt towards having East Ancestry as well (e.g. German+East Asian, roughly). I think it's hard to understand how that combination of ancestry could've happened, but on the other hand, I believe there was known to be dense colonization of Greeks around the Crimean...? Hitting these with nMonte may tell us something (or not!), if I have time for that today.

  13. @Matt,

    Can you please do a cline intersection when you get a chance?

    I have an idea(pops) as to how we can derive some Roman-age Italians with a few assumptions.

    Let me know.

  14. @Samuel Andrews

    "The Roman soldier was probably from Aquitaine in South France/North Spain. He clusters with modern Basque & south French."

    and @AWood

    "Bavaria would have looked like Central European Bell Beaker, which is basically Celto-Germanic."

    Germany_Bronze_Age:RISE471 from MBA Bavaria is also close to the Southern French, judging from his Global 10 coords. He's not in the Global 25 sheet. But I would be very hesitant to expect German Bell Beaker-like people in MBA and IA Bavaria.

  15. Scythian ancestry in Gepidia is no surprise at all. Certainly it would be nice to know to what extent it may have been mediated by Hunnic or Avar influence, but in any case we know that Gepids at one stage lived in, and were culturally dominated by groups that had lived in, long-established Scytho-Sarmatian territory. One way or another, it's no surprise to see those Scythian genes still reflected in Gepid individuals.

    Armenian ancestry in Crimean Ostrogoths is much more exciting. It's not actually clear what "MLBA" Armenian ancestry would mean. We know Armenia seems to have been ruled by Iranian or Iranian-related peoples from the 4th century BC. Armenian itself is not attested until the 5th century - we've no idea when Armenians actually arrived in Armenia, and contrariwise we don't know who used to live in Armenia or what might have happened to them. Armenian nationalists would probably like to interpret this as "Armenian tribes had already arrived in Armenia by the time of the 'Armenian' sample used here, and sister-tribes to the Armenians were clearly still part of the Scytho-Sarmatian confederacies". But we don't really know. This is certainly a case where a lot more genetic evidence from the various Scythian, Sarmatian, Armenian, Hunnic, Avar, Khazar, and Caucasian tribes would be needed to clear up who exactly was where at what time.

  16. It could be that southern Ukraine was very much Caucasus-like before Germans and Slavs.

    Germanic migration is looking more important with each paper. Sam, Bavaria will not be very Beaker-like prior to Germanic migration. They might look like more of a French/Tuscan mix.

  17. @ Chad

    Can you check f4:

    Mbuti (or Chimp) MN/ChL Iberia Iberia EN MN Germany
    Mbuti MNChL Iberia Iberia EN France MLN
    Mbuti MNChL Iberia Iberia EN Brit Neol.
    Mbuti MnCHl Iberia Iberia EN LBK central
    Mbuti MNChL Iberia Iberia EN Hungary LN Tizsa (or similar)

  18. The most interesting thing to me is the presence of the Scythian_Zevakinochilikta + Han + Nganasan ancestral contributions in the autosomal composition of the presumably Hun-admixed individuals.

    Its true that, for the longest time, we get rumblings of some early Turkic or para-Turkic contingent from what dribs and drabs of evidence we receive from the Hunnic period. If this:

    Scythian_ZevakinoChilikta,26.4 + Han,6 + Nganassan,3.8

    which gets normalised to this:

    Scythian_ZevakinoChilikta (73%)
    Han (16%)
    Nganassan (10%)
    Total (100%)

    is representative of the Eastern, Eastern-Steppe-derived component of the population movements associated with the migration period, then we know that the earliest pulses of people from Siberia and Central Asia to West Eurasia that can putatively be associated with the Turks already resemble present-day South Siberians, and do not resemble such groups as Chuvash or the other Uralic-like Turkic populations, such as the Tatars, which probably were the result of language shifts.

    Note that Scythian_ZevakinoChilikta is in turn about 35% East Asian, and this particular Scythian has the most Southern East Asian component, resembling Mongols, relative to other Scythians, such as Scythian_Pazyryk, whose East Asian part resembles Yakuts, Nganasan and other Central and Northern Siberians.

    @ Davidski

    Could you mark on the PCA the position of Balkans_IA?

    @ Ariel

    Thanks for these distances.

    @ Andre

    Thanks for the paper. It seems post-Metal Age gene flow in Europe is greater than expected everywhere we look. In Spain, pan-Mediterranean gene flow, North African gene flow, and Central European contributions are probably all more significant than we expected.

  19. @Anyone So, regarding the Asian part of Gepid sample: can we model what modern/ancient population it is close to?

  20. Might be able to tomorrow, Rob. I have a paper due tomorrow night.

  21. @ryukendo kendow

    I can't highlight Balkans_IA on the plot now, but it's the sample right next to Beaker_Sicily. Except unlike Beaker_Sicily it does actually have steppe ancestry.


  22. The not deformed Medieval Bavarians can be modelled as:



    unscaled, nMonte3

  23. vacouswastriel

    Please speak for Yourself. If You don't know something and You are ignorant that doesn't mean others don't know. Even in Diakonoff's Mushki theory Armenians were in Armenia at LBA.

  24. Btw there is some 100 ancient mitos from Armenia (some still not unpublished ) and I haven't seen even 1 East Asian mito there. While Scythians had lots off it. And modern Ossetians descendant of Sarmatians also had ENA mitos. Btw there is no single U7 from ancient Armenia also. Which is quite surprising not? That is why Yamna probably also didn't have U7.

  25. Those Caucasian shift in Crimea is in most likelihood from Greek settlers. Trabzon Greeks could be similar to Armenia MLBA so I don't see any mystery there.

  26. @huijbregts

    Those England_Roman samples are the decapitated presumed gladiators? Martiniano did identity-by-state checks and found them comparable with Welsh, but also some samples had an affinity to Baltics. (Supplementary fig 12.) Also, they pop up as having some serious extra WHG admixture in the Portugese Bronze age paper by Martiniano (Fig 2).

    Maybe these Bavarian samples - and also the coming Longobard samples - are ideal to do similar checks.

  27. [1] "distance%=2.1071"



    As others have pointed out, it seems he might be Black Sea Greek + steppe (with the expected eastern affiliations). The general attribution and context in the paper would agree and also explain why it basically ends up looking Anatolian Turkish in the paper.

    Germany_Medieval_outlier itself seems Aegean-like:

    [1] "distance%=3.0481"



    FN_2 is Iberian-like and his mtDNA seems to point towards Western Europe? but I'd still be curious to see what e.g. the post-BA northern Balkans were generally like at the time. You can get decent models both ways in nMonte at least.

    The Gepid (VIM_2) seems like a Scandinavian-steppe-northern Balkan mix.

    Seems like the East Germanic confederations picked up a good amount of ancestry along the way (including some potentially Slavic, or "Slavic-like" at least), going from the apparently overall Scandivanian-like Wielbark to at least some looking like this. The 'steppification' that's apparent in the paper wasn't too surprising.

    Some of the other deformed and intermediate samples I played with are interesting too, at least some seem similarly like potential mixes of Scandinavian, steppe and Balkan stuff.

    The non-deformed ones are more "mainstream" per the paper, either Aegean-like or western Central-Northern European overall.

  28. @All

    The Global25 datasheets have been updated with the following new ancients from Iberia.


    As far as I know, Iberia_BA:esp005 is the first sample in the ancient DNA record with R1b-DF27. He shows about 17% steppe ancestry.

  29. @Dave, can you supply a link to the new sheets? Also, a special request if you haven't already done it: can you also add in sample I1392? She is a very old non-steppe sample from France right on the other side of the Germany border. She can likely proved a better non-steppe ancestry fit for most Bell Beaker samples and may replace Iberian Copper Age groups. Thanks!

  30. Davidski

    Are you planning to add the samples from the Longobards paper in global25?

  31. I'm guessing these eastern Germanics absorbed some Sarmatians in the Ukraine?

  32. The nMonte models for my relatives have changed significantly with the new samples:

    Maternal grandmother (3/4 Swabian, 1/4 northwest Swiss):

    [1] "distance%=0.7598 / distance=0.007598"

    "French" 28
    "Germany_Medieval" 27.4
    "Croatian_vLBA:I3313" 18.7
    "Scythian_Samara" 6.25
    "Germany_Roman" 5.45
    "Nordic_IA" 4.7
    "Hungary_BA:I1504" 3.9
    "Germany_Medieval_outlier" 3.4
    "Levant_BA" 1.2
    "Anatolia_ChL" 1
    "Germany_Medieval_ACD" 0
    "Ukraine_Ostrogoth_ACD" 0

    Looks a bit like overfitted, but all these admixures make sense more or less. Apparently she's only about 1/3 Germanic.

    Extrapolated coords of paternal grandfather (from both sides of the German/Swiss Rhine border)

    [1] "distance%=3.131 / distance=0.03131"

    "Germany_Medieval" 87.4
    "Croatian_vLBA:I3313" 6.8
    "Levant_BA" 4.25
    "Hungary_IA" 1.55
    "French" 0
    "Germany_Medieval_ACD" 0
    "Germany_Medieval_outlier" 0
    "Germany_Roman" 0
    "Ukraine_Ostrogoth_ACD" 0
    "Serbia_Gepid_ACD" 0

    Mostly Germanic with no French-like Celtic admixture. I wonder if that Levant_BA is real. Might be very ancient Jewish admixture?

    Extrapolated coords of my maternal half (which is half from Cesena, Italy):

    [1] "distance%=2.6554 / distance=0.026554"

    "French" 51.45
    "Croatian_vLBA:I3313" 17.5
    "Beaker_Northern_Italy" 9.6
    "England_Roman_outlier" 6.8
    "Germany_Roman" 6.7
    "Germany_Medieval" 2.85
    "Mozabite" 2.7
    "Croatian_MBA" 1.3
    "Hungary_BA:I1504" 1.1

    Apparently what I inherited from my maternal grandmother is heavy on the French-like side. And unfortunately it's not always clear here what is from the Italian grandfather and what not.

  33. We have some clear examples of Dark Ages Germanics picking up ancestry as they went. Paul the Deacon describes how the Longobards readily assimilated all peoples they conquered, thus becoming stronger and stronger. Jordanes describes his family as follows:


    "The Sciri, moreover, and the Sadagarii and certain of the Alani with their leader, Candac by name, received Scythia Minor and Lower Moesia. Paria, the father of my father Alanoviiamuth (that is to say, my grandfather), was secretary to this Candac as long as he lived. To his sister's son Gunthigis, also called Baza, the Master of the Soldiery, who was the son of Andag the son of Andela, who was descended from the stock of the Amali, I also, Jordanes, although an unlearned man before my conversion, was secretary."

  34. @Davidski

    "As far as I know, Iberia_BA:esp005 is the first sample in the ancient DNA record with R1b-DF27"

    Don't forget about the I0806 Beaker from Quedlinburg.

  35. @Anthro_Survey, what couple of clines are you interested in seeing the intersection of?


    Also all had a go at nMonte with the two sims for Crimean Ostrogoth's ancestry that I made earlier:

    Using all modern population averages:

    1: (Ostrogoth - 25% Swedish+Norwegian) - distance%=3.0075: Greek_Trabzon,66.6, BedouinB,10.4, Abkhasian,7.2, Maltese,7, Georgian_Imer,3.4, Algerian,2, Yakut,1.2, Quechua,1, Selkup,0.6, Surui,0.6
    2: (Ostrogoth - 75% Greek_Trabazon) - distance%=10.6061 - Norwegian,44.4, Finnish_East,32.2, Saharawi,10.8, Quechua,5.8, Basque_Spanish,4, Ket,2.2, Biaka,0.6

    Using all ancient+modern population averages:
    1: distance%=3.0576: Levant_BA,20.6, Armenia_MLBA,19.4, CHG,18.4, Minoan_Lasithi,16, Anatolia_ChL,13.8, Levant_N,8.2, Clovis,2, Karasuk_outlier,1.2, Scythian_Pazyryk,0.4
    2: distance%=10.7309: Comb_Ceramic,28.2, Wales_CA_EBA,27.8, Natufian,11.2, Mentese_N,9, Beaker_Southern_France,7.4, Clovis,5.8, England_Anglo-Saxon,5.2, Karasuk_outlier,3.8, Tanzania_Pemba_700BP,1.6

    So overall it seems like, if we are having to fit Ostrogoth with some Germanic fraction (and a level of Germanic ancestry that doesn't project the other fraction waaay far down south beyond Greek Trabazon), then Greek Trabazon+Scandinavian is a pretty good two population fit.
    Looking at the linear correlation of 75:25 Greek_Trab:Germanic vs the real Ostrogoth: https://imgur.com/Nwfl3gL. Pretty linear correlation.

    But looking at the residuals to see what is very slightly different: https://imgur.com/NJ01UVo
    Zooming in on those residuals: https://imgur.com/uC7RCBM (Below the line = closer to 75:25 Greek_Trab:Germanic than real Ostrogoth; above the line = closer to real Ostrogoth).

    It looks like as a combination, 75:25 Greek_Trab:Germanic is a little too close to CHG and Iran_N to match.

    On the contrary, the real Ostrogoth is a little closer to a) Levant_N / Levant_BA influenced populations ("East Med" populations?), b) Anatolia_N heavy populations (poss because of relationship between Anatolian and Levant Neolithic?), c) specifically German_Medieval Artificially Cranial Deformation samples and the Gepid, d) Siberian influenced populations. That suggests there's some common factor between the ACD, Gepid and Ostrogoth samples here, and that it has some connection to a combined affinity for Levant and Siberian influence that is unusual today? (This is probably completely off base).

  36. Those 'Goths' were probably Turkic.

  37. Matt,

    While I'm a bit skeptical of this particular sample having significant "Germanic" ancestry, out of curiosity, what do you get if you use Greek_Central_Anatolia instead?

    I'm a bit curious because

    [1] "distance%=2.1071"



    [1] "distance%=2.7475"




    [1] "distance%=4.6454"



  38. @Alogo, I didn't have the Greek_Central_Anatolia average in the spreadsheet I'm using, so I'd have to hunt it down.

    Also, yeah, reasonable to be skeptical of the sample having Germanic (e.g. Scandinavian like) ancestry, models are simply a kind of "If he has this ancestry at all in a significant quantity, then what is the best fitting other side....", rather than actually absolute best fit. I think I'll have to look over the haplotype/Chromopainter analysis they have in the paper, to see if that can identify any direct link at all.

    To be honest, reading the SI it does seem like this skull has pretty shady and poorly documented history.

  39. @Matt,

    Of course. :-)

    I'm using David's spreadsheets btw. KER_1 seems to have greater affinities to the low-coverage STR_502 from a quick look? STR_502 also similarly seems to be assigned to modern Anatolian Turks and get Tuscan, along with a little East and South Asian in their ADMIXTURE.

    And fully agreed but its nMonte results at first glance seem very interesting nonetheless when considering its Bosporan/Kerch context, much like a lot of the other more "mixed" individuals.

  40. @Richard Rocca

    That Beaker sample from eastern France with no steppe (or perhaps ultra low level steppe) ancestry is now in the Global25 datasheets. Look for the label and ID Beaker_Central_Europe_no_steppe:I1392.

    The links to the Global25 datasheets are in the above blog entry. These links are always the same, even after minor and major updates.


    I'll add the ancient samples from the Longobard paper when it's formally published and the genotype data are available from the authors.


    The two Sarmatians from Pokrovka, Russia, that I have in my dataset don't appear to explain the eastern and southeastern shifts for the Gepid and Ostrogoth, respectively. Heavily East Asian admixed Scythians are preferred by the Gepid, and ancient Balkan and Caucasus samples by the Ostrogoth.

  41. Outlier could be Euro-Jewish, easily.

  42. @David, can you run nMonte3 on AED_1108, who also has significant East Asian ancestry, and on ALH_3 and AED_92 who are derived for EDAR?

  43. @Matt

    Let's see if these two clines come close to meaningfully intersecting in G25 or not:
    Samaritan----->Italian Jew--->X<-----Tuscan<-----Medieval German average(custom)

    Also, is your email written in the last post of the cheddar man thread(some characters omitted) still current?

    This is the custom average: 0.122666077,0.131862538,0.060745462,0.047828846,0.039320923,0.019865692,0.011262462,0.004437615,0.001730692,-0.000967308,-0.004846769,0.002789769,-0.007638923,-0.002646692,0.015315538,0.000387615,-0.011142615,0.002718846,0.004592692,0.005300769,0.005941462,0.003329077,0.001507308,0.009435923,-0.000764462

  44. Good thing I checked in, neat new ancient genomes.

    Rerunning my parents on the data including the new samples, my dad has a strange affinity to the Ukrainian ACD Ostrogoth. His distance to Ukraine_Ostrogoth_ACD:KER_1 is 4.95% (second closest of all non-Turkish distances to my dad, the first being a Greek from Trabzon).

    Using Alberto's ancestorsMix on the pop-average data without Turks and Azeris, I immediately get these for my dad:

    > getAncestors('PopAve-pruned.txt','dad-scaled.txt',1)

    Kurdish 50%
    Ukraine_Ostrogoth_ACD 50%

    Distance 3.8243%

    > getAncestors('PopAve-pruned.txt','dad-scaled.txt',2)

    Anatolia_BA 25%
    Greek_Trabzon 25%
    Iran_IA 25%
    Serbia_Gepid_ACD 25%

    Distance 2.4722%

    That ACD stuff is even being picked out of a file that needs further massive pruning (first run on new GlobalPCA25 files).

    Seems that my Dad might be a bit of a Hun. This will make my family laugh real good since it just MAKES SENSE. Hah!

  45. @Joshua Lipson,

    "Outlier could be Euro-Jewish, easily"

    Ashkenazi Jews have an obvious Near Eastern pull while the outlier basically looks just like an Iron age genome from Bulgaria, very European in particular Neolithic European farmer-derived so not like Ashkenazi Jews.

    I'd think that in 500 AD, whatever Jews lived in Germany, they were much more Near Eastern/Hebrew than Ashkenazi Jews today.

  46. @Open Genomes

    [1] distance%=4.0083




    [1] distance%=2.6918



    AED_92 isn't in the datasheet.

  47. Incase you guys forget, Goths (Wielbark) in Poland had Danish affinity accoridng to Polish media reports. One study published their Y DNA, it was mostly I1.

    The two east Germans in this study have complex origins. But more data is needed to know for sure how mixed they were.

  48. It'll be interesting to see German Bavarians from 500ad in north Europe PCA alongside the Slavs in Bohemia dating to 500ad.

  49. The Spain paper was earlier mentioned. Thus we see R1b & steppe admixture was arriving by the tail end of the BB period, and first seen in northern Iberia, and by the Bronze Age it had homogenized (relatively) through Iberia (eg in El Agar and Cogatas cultures).
    Also noteworthy are some of the above remarks, as people are (slowly) realizing (to their "surprise") significant shifts occurred within Europe after the BB period (eg Bavarians =/= unadulaterated Beaker folk), which must have had profound yet linguistic and cultural impacts.

    So with some solid BA Iberia data now, let's have a look at modern Iberians & their neighbours.

    Iberia_BA 73 %
    Armenia_EBA 17.8 %
    Mozabite 9 %
    Bonda 0.1 %
    Globular_Amphora 0 %
    Anatolia_BA 0 %
    Remedello_BA 0 %
    Iberia_ChL 0 %
    Balkans_IA:I5769 0.1 %
    d 0.024%

    (the Mozabite level peaks in the West - as suggested by several studies, IIRC, and is lower in eastern Spaniards)

    Iberia_BA 95.6 %
    Armenia_EBA 4.2 %
    Han 0.2 %
    Mozabite 0 %
    d 0.025%

    Iberia_BA 78.3 %
    Armenia_EBA 20.1 %
    Bonda 0.9 %
    Han 0.7 %

    So the axiom that Basques are a relict population is indeed true. However, earlier postulates were off by 30, 000 years, lol- as it was not the Ice Age, but Bronze Age.
    But what does the small but sure Armenia BA represent in the Basques IE neighbours ?

    Lastly, a pet theory of some of the boys is that Basque is related to Sardinian. That's a complete negatory, as they have completely different histories.

    Remedello_BA 41.7 %
    Iberia_ChL 30.4 %
    Anatolia_BA 20.4 %
    Balkans_IA:I5769 5.2 %
    Iberia_BA 1.9 %
    Mozabite 0.4 %
    Han 0 %
    Bonda 0 %
    Globular_Amphora 0 %
    Armenia_EBA 0 %

    I think more sense will arrive with mid-late Bronze Age & Iron Age genomes which will modify the coarse -grained theories which are currently popular (mostly due to parochialist reasons rather than empirical basis)

    1. French East certainly isn't an Iberian and Armenian mix.

  50. @Rob

    I don't think these sorts of exotic women introduced new languages into Europe. At least not permanently.

    My bet is that they quickly integrated into their host populations and started speaking whatever their Beaker/Corded Ware-like men were speaking.

  51. @ David

    "I don't think these sorts of exotic women introduced new languages into Europe. At least not permanently."

    LOl maybe not, but why not ? But I do love your default anthropological explanations ("exotic women !")

    No, more likely its related to the migration of a novel group of people from the mid to late Bronze Age into, say, Iberia; which were not the original DF27 founder population, but a Bronze Age'ized group of elites carrying new patterns of already existing European haplogroups - such as the R1b-U152 seen in Spaniards but far less common in Basques (on the top of my head) from ? Halstatt - associated with an early wave of para-Celtic (ancestors of Liguarian). There are similar shifts in Netherlands (U106 replacing earlier L21), central Europe (I2c, I2a2) in Unetice. Of course, this is as predicted by linguists and archaeologists (eg Wessex culture, Nordic Bronze Age, etc). Indeed, through much of mitteleuropa, not much of the Beaker cultural heritage remains after 2000 BC.
    But as I said still early days, IMO. And the Armenia signature is an oddity, i have to say.

  52. @Rob

    LOl maybe not, but why not? But I do love your default anthropological explanations ("exotic women !")

    The Bavarian males from this paper, just like modern-day North, Central and East Europeans, resemble Central Beakers, Unetice and Trzciniec peoples in terms of genome-wide genetic structure, if not, strictly speaking, Y-haplogroups.

    So apart from some paternal founder effects, intra-regional migrations, and a bit of admixture from afar here and there, there's been remarkable genetic continuity and homogeneity across this part of Europe since the Bronze Age.

    And it looks like the really exotic ancestry came mostly with women. Sure, you're free to speculate that such female-biased migrations, clearly for the purposes of female exogamy, caused language shifts in ancient Europe, but I personally don't see it as a realistic scenario.

  53. Nice to see that we have a hint how Trabzon and Crimea Greeks were looking. Btw Crimea Greeks have y dna L1b which almost certainly they got from Trabzon.

  54. @Rob

    The problem with that model is that it implies that Basques have more steppe ancestry than other Iberians, and they most likely don't, as David showed here https://docs.google.com/spreadsheets/d/1gzxMCTBTJo0NZ5jsTXnM2N6mkbBUUTXHLRgOw2CqQb4/edit#gid=515556785
    Modern Basques could even be a rather late arrival into Iberia from behond the Pyrenees whereas Roman period Basques in Spain became romanized and now a part of the Spanish-speaking NE Iberians. Or a mix of these. We don't really know.

    As for the "Armenian" signal, it could have come from either the Roman period, or when CHG-related ancestry expanded throughout the Mediterranean, arriving in Iberia slightly later and in fewer numers than it did in the east (so analogous to BB). No idea what kind of languages they could have spoken, besides the obvious non-IE Iberian language. It could also never had a strong importance and disappearead without we ever having traces of it

    1. @André
      Another problem about modern Spaniards and Portuguese is the impact of migration linked to Reconquista. It's clear from modern distribution and structures that Reconquista is the last-to-date main even regarding Iberian populations.
      Historians have often amphized the level of "Franks" importance in Reconquista, militarily but also as migrants to settle conquered lands. At least till the loss by the king of Aragon of his Southern France lands (in Massif Central). Sadly the size of this admixture is unknown (and the last paper using modern samples is not helping).
      My conclusion would be that without some decent French samples, it could be difficult to understand how modern Iberian came to be.

  55. And few words about Iranian tribes in South Caucasus and their possible lookup. Scythian burials are well attested in Azerbaijan were they had a kingdom. They also show up sometimes in Armenia. It is quite possible that their autosomes will be similar to Armenia LBA. Maybe more steppic shifted
    But based on that one can't amalgame and declare that fortified and sedentary towns in Armenia LBA were all Scythian. It would be a nonsense.

  56. Women almost always adopted the language of the males worldwide. Why should it be any different in patriarchal and patrilocal, originally agnatic Indo-Europeans? The idea is ridiculous.

  57. @Folker

    What you say regarding the reconquista is precisely what a new study in modern Iberians concluded. I suppose that, for people who are familiar with Iberian history, that sounds unimpressive but it's good to see it confirmed nontheless:

    "Overall, the pattern of genetic differentiation we observe in Spain reflects the linguistic and geopolitical boundaries present around the end of the time of Muslim rule in Spain, suggesting this period has had a significant and long-term impact on the genetic structure observed in modern Spain, over 500 years later."

    Each kingdom/language-group homogenised as the reconquest moved south and people settled in newly-aquired territories, forming a homogenous pattern. It also explains why the differences in Iberia are mostly West-East rather than North-South.

    As for the Portuguese, also expected results:
    "We speculate that the pattern we see is driven by later internal migratory flows, such as between Portugal and Galicia, and this would also explain why Galicia and Portugal show indistinguishable ancestry sharing with non-Spanish groups more generally"
    "In a complementary analysis that included Portugal, although fewer SNPs (Methods), Portuguese individuals co-clustered with individuals in Galicia"
    "the strong genetic similarity between individuals from Portugal and Spanish individuals, especially those located in Galicia (Supplementary Figure 1a), means they are likely to share a similar admixture history"
    So they are virtually Galicians, which is also unsurprising considering history and language.

    As for non-Iberians taking part in the Reconquista or repopulations, I would also like to see it studied, but it might be a bit complicated and relatively unimportant. Besides soldiers who decided to stay, it is documented that civilians also settled in many areas, for example Burgundians in Central Portugal (if I'm not mistaken). Their numbers could have been too low to cause any significant impact

    Also, sorry everyone else for the side-tracking

  58. @ Andre

    "The problem with that model is that it implies that Basques have more steppe ancestry than other Iberians, and they most likely don't, as David showed here "

    I disagree.
    Look at look at the paper you introduced - See fig. 2
    Reconquista does not explain away that fact, because it is consistent in all Iberians: Basque have higher BA Iberian ancestry than other Iberians, & significantly so. Similar ancestry is even lower in English and French, so reconquista doesn't seem to be the culprit.
    Dave's qpAdm sheet is using some rather distant sources, anyhow, using CHG etc.
    As for the extra Armenia-EBA - sure - it could be from Roman women, as you lads like to theorize :)
    I guess we'll find out soon enough when it appeared.

  59. @André

    As for the "Armenian" signal, it could have come from either the Roman period, or when CHG-related ancestry expanded throughout the Mediterranean, arriving in Iberia slightly later and in fewer numers than it did in the east (so analogous to BB). No idea what kind of languages they could have spoken, besides the obvious non-IE Iberian language. It could also never had a strong importance and disappearead without we ever having traces of it.

    I don't think that Phoenician or Greek colonies had a significant impact in the Iberian gene pool. More important was their cultural impact, but linguistically they clearly didn't have any apart from the alphabet for writing (for obvious reasons).

    Basques, as you mention, probably moved late to their current location in Iberia (Roman times, maybe), with their core area being Aquitaine (where by Roman times the language was spoken). Clearly and Atlantic area. Basque is probably related to the other Iberian languages, though this needs further research to be clear.

    We now know that the first and most dominating Bronze Age culture from Iberia (El Argar, SE Iberia) was R1b (from a teaser in an interview with one researcher).

    We also know that Celtic languages expanded to Western Europe around the VII century BCE and replaced the languages spoken in large areas. But some of the most populated ones didn't shift to Celtic. It seems reasonable to think that those who did shift to Celtic did so from a language related to those who didn't. And indeed, there is a Vasconic substrate in Western Europe (or at least it's hypothesized, since these things are always difficult to prove with certainty).

    Basques have roughly the same amount of "Yamnaya" admixture as other Iberians (25-30%), And like them and the rest of Western Europeans a very high incidence of R1b-L51+. The difference comes from them being most unadmixed since the Bronze Age, as Rob said above.

    Now compare this situation in Western Europe with what we know from Bronze Age Greece. Mycenaean Greeks have some 5-6% real Yamnaya admixture (being much closer to the steppe) and presumably a quite low level of R1b-Z2103 (to be confirmed still, but anything else would be surprising). And they did speak an IE language. The 3 Anatolian BA samples we have show no steppe admixture and no R1b/a. We don't know if they spoke IE, but if they didn't, their neighbours did.

    There seems to be a distant connection (speculative, but there it is) between Basque and NE Caucasian languages. Today R1b-Z2103 is more common in speakers of these languages and in former Hurrian areas (Hurrian probably being related to NE Caucasian), with Armenians being more or less considered descendants (partly) of the Hurrians.

    So putting all together, I'm still struggling to see any connection between R1b and IE languages. Even in Italy the former Etruria is the place with highest incidence of R1b. We'll need ancient DNA to see how Etruscans differed from Romans, but it wouldn't be surprising for Romans to have a South Eastern pull and higher Y-DNA J.

    Let's see what the early Celtic culture from Hallstadt looks like, but unlike many other here I do expect they'll show a shift to the Balkans. And we'll see about their Y-DNA. I didn't have time to play with the latest Iberian samples, but it's interesting that even French_East show that Armenia_EBA shift that hardly appears in Basques.

  60. @André
    I've read it and found some issues (methodology and postulates). And it amphize greatly French affinity.
    Anyway, there are papers about French settlements in Northern Spain (especially in Navaria and Aragon, for obvious reasons given their kings did have also lands in France). Some fueros were exclusively French. So, probably not a minor issue.

    Sorry for this side story.

  61. @ Alberto
    Yes I see the Etruscan link too. Although Aram has attempted to expplain it away as a G2a link, i think that’s grasping at straws

    Luckily, we’ll have halstatt genomes soon enough

    I think we need to not be so indo -Eurocentric and see that other great languages expanded in the copper -Bronze Age

  62. @ Chad

    "French East certainly isn't an Iberian and Armenian mix."

    I didn't suggest that literally, Chad. Obviously there wasn't a migration from Iberia to France c. 1500 BC.

    1. Of course. Remove Iberia and see what kind of mix you get. I doubt it'll find the 40% Saxon-like admixture, but worth a peek. I can see if qpGraph picks it up better. I'll also get those stats tonight sometime.

  63. @Alberto, Rob

    When I said "Roman period" I'm not advocating 'Roman women' (in fact I'd guess it would be mostly male-mediated), and "CHG-related" not Carthaginian or Greek, but possibly something earlier, such as mid-late 2nd millennium BC. I really have no personal opinion on the matter, hopefully the upcoming Lalueza-Fox paper will answer that, and it will be what it will be.

    This new Iberian paper was interesting, but I guess it's true usefulness is to prove modern structure within Iberia (or Spain specifically, since there was no geographical study of Portuguese samples) is related to languages and medieval history, rather than admixture events that happened in the past. 'French' might not even be a good choice in the first place. Anyway, as above, the Lalueza-Fox paper will hopefully explain admixture a lot better than this paper did.

    As for R1b and IE, sorry Alberto, but I still find it very unlikely that they are not related. The ancient samples seem very adamant

    You might be right, particularly in the East, but I think that in Portugal that was less of a thing. I could be wrong as even the first Portuguese kings were Burgundian, and the Templars had a long story of operation and cooperation with the kingdom.

  64. @All

    I've updated the post with a North Euro PCA featuring many of the German Medieval samples, but obviously only those that qualified as North Europeans. Almost all of them are males with no signs of ACD.

    They're obviously very Germanic genetically, but also rather similar to the Central European Beakers from the Bronze Age, which I put on the same plot recently...


  65. @Rob

    No one here ever claimed that female exogamy was the main source of exotic admixture outside of Barbarian Northern, Central and Eastern Europe. Rome is a different thing altogether in this context.

  66. Joshua Lipson said: "Outlier could be Euro-Jewish, easily."

    As Sam also mentioned, the outlier doesn't seem to have the Levantine affinities even Ashkenazi Jews have. Unless we're positing an entirely different group of Jews that had mostly/entirely "European" affinities.

    It looks ancient Southeastern European of some kind to me.

  67. @David

    Is it me, or do these German samples seem more south-shifted on PCAs than the Lombards from https://www.biorxiv.org/content/early/2018/02/20/268250 ?

  68. @André

    I'll know more about those Lombards when I run them myself. Can't say much now.

  69. @Alberto

    Mycenaean Greeks have some 5-6% real Yamnaya admixture.

    Not sure how you got this figure. However, at the risk of stating the obvious, genes don't speak languages, although they can be used to track migrations that are likely to have spread languages.

    In other words, Yamnaya_Bulgaria was also real Yamnaya from the steppe, even if barely 50% Yamnaya_Samara-like, and Potapovka was real Potapovka from the steppe, even if only 70% Yamnaya_Samara-like.

    And if either Yamnaya_Bulgaria or Potapovka spread proto-Greek to Greece, and it does look that way based on their DNA and the DNA of the Mycenaeans, then Mycenaean Greek is from the steppe.

    No one ever said it came directly from Early Bronze Age Samara, and that unadmixed Yamnaya_Samara spread it to Mycenaean Greece.

  70. " Mycenaean Greek is from the steppe." Very doubtful for the TOTAL specifics of the language. MG differs greatly from reconstructed Late PIE (esp. as to its lexicon). MG= the result of the interplay between incoming people (whose own language was probably no longer "classic" PIE due to interplay with previous populations "on the way"...) and locals. Much the same can be said for practically all of the existing IE languages (some more some less). But the basis of development certainly seems "steppic".

  71. @André

    As for R1b and IE, sorry Alberto, but I still find it very unlikely that they are not related. The ancient samples seem very adamant

    Hhhmmm... my whole explanation is based on the available ancient samples and trying to fill the gaps with assumptions that look reasonable given those ancient and modern samples.

    Can you elaborate on which R1b samples exactly are so adamantly showing a specific connection to IE languages? As I said, I struggle to see this connection with the available data.

  72. @Davidski

    I do count Yamnaya_Bulgaria as real Yamnaya. I put the world "real" because Mycenaeans (the 4 samples we have, anyway) can be modelled as 10-12% Yamnaya without the Neolithic/Chalcolithic Balkans. Once you add those, Yamnaya admixture goes down to around half.

    And yes, it came from Bulgaria with Yamnaya migrants that left the steppe c. 3000 BC. Good that you now see this (and not the Sintashta-related hypothesis).

  73. @Davidski,

    Could you also show Nordic_LN on the Northern Europe PCA?

  74. A general observation on Global 25 / nMonte modelling: It seems it doesn't always get the ancestries right; in some cases it merges various different pulls into one intermediate pull. For instance, my East Prussian grandmother gets 22.8% Baltic_BA:Turlojiske3 and 38.9% Nordic_IA, which makes sense, given her probable Old Prussian admixture and her historical North German roots. However, her son gets 0% Baltic_BA:Turlojiske3. This is genetically impossible. So what's happened? Her son gets lots of Nordic_IA instead, which is a merged pull that arose from his North German/Nordic_IA, his Germany_Medieval and his Baltic_BA:Turlojiske3. Only when I delete the Nordic_IA from his reference pops, then the Baltic_BA:Turlojiske emerges with perfectly logical 12.4% - and all the Germanic ancestry is then packed into his Germany_Medieval.

    So in order to model my Italian ancestry, it's probably best to use the inferred coords of my Italian grandfather, which are extrapolated twice (my mother is extrapolated once, and then a second extrapolation yields my grandfather's coords). That's still better than to read everything out of my extrapolated maternal side. And this is what I get:

    [1] "distance%=5.7384 / distance=0.057384"

    "French" 29.75
    "Beaker_Northern_Italy" 25.45
    "Croatian_MBA" 12.65
    "Baden_LCA" 8.9
    "England_Roman_outlier" 8.4
    "Mozabite" 7.1
    "Croatian_vLBA:I3313" 4.75
    "Peloponnese_N_outlier:I3920" 3

    Beaker_Northern_Italy, Croatian_MBA and Peloponnese_N_outlier probably represent the different strains of the Chalcolithic layer. Baden_LCA might possibly have something to do with the Etruscans. Croatian_vLBA is still Iron Age Liburnian admixture IMO. And French the Padanian Celts of course. Quite substantial in fact. And then exotic Roman Age accretions from the Levant and North Africa complete the mixture. The Levantine admixture isn't excessively strong in this model which is a positive hint to its accuracy.

  75. @ Anthro Survey

    Initial Italian_Jew - Italian_Tuscan distance: 0.0618811481067.

    Distance: 0.0261594334055

    2.62%, so it worked pretty well, in nMonte it would be considered as a successful result.

    It seems that this slight shift between ghosts points to Mozabite.

    Custom_Italian_Tuscan_x_Italian_Jew_Samaritan_cline 57%
    Mozabite 6.6%

    Distance 0.857%

    So if the Mozabite admixture in Italian_Jew compared to Samaritan is reasonable, than it looks like the point on Custom - Tuscan line is the one you are looking for.

    The Sephardic, Italian, and Syrian Jews all showed a low level component (8%–11%) that was shared with the North African Mozabite population when K equaled 6 (Figures 2A and 2B). This component was less apparent among the Ashkenazi and Middle Eastern Jews (Figure 2B; Figure S3).

  76. Re: Ostrogoth

    Armenia_MLBA:RISE423 41.2%
    ALPc_MN:I2384 20.8%
    Abkhasian:350_R02C02 19.5%
    BedouinB:HGDP00630 9%
    Yamnaya_Samara:I0438 4.9%
    Mari:mari3 4.3%
    Yakut:878_B 0.3%

    Distance 1.6307%

    No sign of anything specifically Germanic.
    Maybe other samples in the future will change this picture, but for now it looks like Gothic went to Crimea from the Balkans.

  77. Northern Euro PCA with Baiuvars and added clusters


  78. @Arza

    "No sign of anything specifically Germanic.
    Maybe other samples in the future will change this picture, but for now it looks like Gothic went to Crimea from the Balkans.

    The burial site is the Greek colony of Pantikapaion, which was occupied by the Gothic tribe of the Tetraxits, then it was destroyed by the Huns. This burial could be Greek, Gothic, Hunnish or Sarmatian. The fact that it has skull deformation could be that someone copied Hunnic traditions. Similar things were seen with the Roman population in Gaul which copied Frankish clothing styles and names.

  79. Germany_Roman:FN2 could have been from Navarra, Spain or Guipuzcoa, Spain since that is where the Spanish Basque auxiliaries of Vasconum and Vardulorum were from.


    A book in Spanish on Basques in the Roman Empire is at http://sehn.org.es/wp-content/uploads/2017/08/80109.pdf

  80. @Dospaises

    O, you have to love wiki for these kind of lists.

    Ala 1 Hispanorum Auriana was located Biriciana (Weißenburg,Bavaria) which is roughly 20-30 km north f FN_2.


  81. @ epoch2013

    This sample is not fully Greek, Anatolian, Caucasian, steppe nor Asian. It's of mixed, local origin.

    Greek_Central_Anatolia:G25003 89.5%
    EHG:I0061 10.5%

    Distance 2.9941%

    If just one grand-grand-grand parent was a Scandinavian Goth we should see this clearly in the results (6.25%). So now you have 16 samples and still nothing.

    There is no pull towards Scandinavia in this sample at all. Where are the Gothic hordes that conquered and subdued everyone on their way? They left no trace?

  82. The Iberian paper is really quite impressive and very useful, because from the papers from this and the POBI (People of the British Isles) project and also the Irish DNA project, we see that sociologically significant aspects of human populations, such as political boundaries between regional polities, or especially boundaries between speech regions, leave significant signals on the patterns of gene flow in the post-MLBA period.

    Also populations in the post-MLBA period remained highly unstable and prone to flux. (vs this "invisible Romana submerga" type understanding where post-Metal Age gene flow patterns were never significant, anywhere, or that there is some "France-Profonde" type peasant population with romanticised millennia-deep roots; these don't exist wherever we've checked at least.)

    E.g. the more historically significant aspect of Galicians in Spain is the fact that they used to be the same speech community as Portuguese, until the 14th century, and we find extensive homogenization of gene pools in similar speech communities such that N African ancestry peaks in Galicia in Spain despite Galicia being never ruled or significantly settled by N Africans in the historical record.

    Which means that the old point about N African ancestry peaking in the West of Iberia doesn't work to date N African influence to the Paleolithic, Neolithic or even Metal Ages (but we already knew this even before this paper came out because the segments were very long in Iberians, such that the largest fraction of NAf ancestry must be from the latest admixture, which is in the 1st mil CE).

    The Balkan_IA and Germanic genomes also suggest this, though the Germanic case may be harder to determine because in terms of broad ancestry North-Central-Eastern Europeans are all similar. But I expect David's N Euro PCA will show that populations that area were relatively unstable as well.

  83. Ancient North Africans


  84. @ Alberto

    I find it very weird that you are trying to get R1b to speak Indo-European. R1b is just one genetic unit, its like saying you expect the lactase persistence allele to speak Indo-European or the blue-eye allele to have spread the Neolithic.

    IE was spoken by a community, and communities have population-specific gene pools, it makes far more sense to tie Yamnaya ancestry (since earliest Corded Ware + Yamnaya and peri-Yamnaya all had similar gene pools) to language than just a single locus. R1b just came along for the ride (and looking at earliest CW, R1a as well).

    In fact a fun excercise would be to get, say 20 linguists to analyse expanded Swadesh lists of present-day IE languages, and then try to tie IE vocab fractions with autosomal Yamnaya fractions, I'm expecting the correlation to be quite good and, by comparing with simulations, extremely statistically non-random and unlikely to be due to chance.

  85. @Arza

    That looks like a Greek with local admixture. Completely to be expected in the Gothic empire. Or even in the Hunnic empire that followed it. Gothic rule wasn't more than a generation:

    "In Late Antique times the city’s importance ceased and the settlement structure changed. Under King Rheskuporis VI. (reign: 314/15-341/42) Pantikapaion’s mintage came to an end; yet, in his regency the city became newly fortified. Around the middle of the 4th century AD Gothic Tetraxits established their rule (26) before it was overtaken by the Huns in the seventies of the same century (27, 28). After they had destroyed the city to an unclear extent, the Huns and their allies were probably also aware of its strategically important location."

    Off course the Crimea is the place where the Goths survived until the onset of the modern age. So yes, there must be Gothic samples to be found. But is this sample such a problem? Certainly not.

    Mind you, the dating of the sample is this:

    "Radiocarbon dating from a piece of the same petrous bone yielded an age of 256 – 401 cal"

  86. @Chad: Do you have the NA haplogroups?

  87. @Arza

    "So now you have 16 samples and still nothing."

    We have 16 samples of (Ostro-)goths? Have I missed something? Where are the other 15 published?

  88. Consist- ently, we find that all males with sufficient nuclear DNA preservation carry Y haplogroup E1b1b1a1 (M-78; table S16). This haplogroup occurs most frequently in present-day North and East African populations (18). The closely related E1b1b1b (M-123) haplogroup has been reported for Epipaleo- lithic Natufians and Pre-Pottery Neolithic Levantines (“Le- vant_N”) (16). Unsupervised genetic clustering also suggests a connection of Taforalt to the Near East. The three major components that comprise the Taforalt genomes are maxim- ized in early Holocene Levantines, East African hunter-gath- erer Hadza from north-central Tanzania, and West Africans (K = 10; Fig. 2B). In contrast, present-day North Africans have smaller sub-Saharan African components with minimal Hadza-related contribution (Fig. 2B).

  89. @Chad

    Also U6 is apparently from the east. If not some traces of Paleolithic Europe should have been found.

  90. [1] distance%=4.0083



    That's an amazingly high percentage of East-Asian/Siberian for an archaeological Central European.

    Let's say hypothetically that this woman was a Hun, given the early date, or from one of the Hunnic subgroups (Sabirs, Bulgars, and Khazars) who presumably would have started out very close to the their Hunnic ancestors.

    I hope everyone realizes that we do in fact have samples from two Iron Age Altaians from around 800 CE who were almost certainly Uyghurs, since they lived at the right time and place for the height of the Uyghur Kaganate.

    David, would it be possible then to compare the Siberian / East Asian components of AED_1108 with these Iron Age Altaian likley Uyghurs, to see if their Asian components have the same source?

    Is there a possibility of checking for IBD, or rare allele sharing?

    In other words, can we use these "Iron Age" individuals from nearly opposite ends of Eurasia, who are only about 300 years apart, to zero in on a aDNA-based source from this presumably Proto-Turkic expansion?

    This could be the "Siberian / East Asian" component that we are seeing in many populations today (for example, many groups from the Caucasus, some Anatolian Turks, and the SGDP Hungarian mentioned in the study) and also among later historical groups that might be sequenced such as the the early Magyars. If we ever get sequencing from the many ancient Chinese samples that have been tested in a rudimentary way, for example these individuals from the ancient Zhou Dynasty capital of Luoyang.

    Ancient DNA Reveals That the Genetic Structure of the Northern Han Chinese Was Shaped Prior to 3,000 Years Ago

    It may be that the Zhou people, who could have been Q1a2* on the Y, were a Siberian people who assimilated to Chinese culture.

    There are is also likely to be some sequencing from Medieval Magyars:
    Maternal Genetic Ancestry and Legacy of 10th Century AD Hungarians

    It would be interesting to see the differences between the Uralic-speaking Magyars and the proto-Turkic Huns and Avars.

    No doubt the emergence of the Han Dynasty played a big role in the consolidation of the proto-Hunnic Xiongnu and their migration westward.

    I think we've got just enough between these part-Asian Migration Era Europeans and the Iron Age Altaians to get the start of an answer.

  91. @All

    Please note that I've just renamed the Gepid and Ostrogothic samples in all datasheets and graphics to the following...



  92. @@Arza

    Why would the Goth speak a Germanic language (in fact Gothic is the only East Germanic language that we have a writen example for (Bible translation by Wulfila) and their leaders having Germanic names, if they dont have any relation to Germanic people at all?

  93. @ epoch2013
    16 grand-grand-grand non-Germanic parents.

    @ Fanty
    Cultural diffusion. Crimean Goths may have been language-switchers from the Balkans who came to Crimea via Greek trade network. Hopefully future samples will expose the path of Gothic migration.

  94. @Alberto

    You're choosing to be very naive. Like I said, genes don't speak languages, people do. All we can use genes for are to track migrations and thus help to validate earlier theories laid down by historical linguists.

    So, for instance, it was posited over 100 years ago based on linguistics data that during the Bronze Age there were large scale migrations from the Pontic-Caspian steppe all the way to Iberia in the west and South Asia in the east. That is now being validated by ancient and modern-day DNA data.

    It makes no difference whether people who now speak Basque, Dravidian, Uralic or Turkic carry significant levels of admixture from those Bronze Age migrations from the steppe. Why? Because genes don't speak languages, people do, and sometimes they switch languages, even back and forth, but overall, the correlations and patterns are rather clear.


    The North Euro coordinates for Nordic_LN RSIE98 were posted in this blog entry.



    It seems like you've just discovered the problem of overfitting, which is the result of high correlations between very similar reference samples. We've been discussing this issue in the comments here for a while now, and it's something that I warned about in this blog entry.



    It seems that you're horribly overfitting your models for the Ostrogoth just to get rid of any potential Scandinavian signal. A little less bias would go a long way methinks.

    Try much simpler, chronologically more parsimonious models with Nordic_IA as one of the few reference samples and show us what you get.

    @Open Genomes

    There's not much high quality data to work with between these East Germanic samples and the RISE Altaians. I can't see an IBD run being viable. But you might be able to get some interesting insights by looking more closely at the Global25 data, including in a more systematic way with nMonte than what I've done to date.


    I think you'll find that Spanish_Pais_Vasco are the Spaniards with the highest levels of steppe ancestry in Iberia, much higher than what Basque_Spanish and even Basque_France have, so they can't be explained away as former Basques-speakers. But they also totally lack that Armenian-related signal that other Spaniards have.

    So I'd say that Spanish_Pais_Vasco are the best fit for a population relict of the Iberian Bronze Age and, more importantly, the likely Indo-European-speaking Beakers with the highest levels of steppe ancestry.

    This isn't surprising, because thanks to their geography, much like Basques, they were shielded from the recent exotic gene flows that affected other Iberians, and only really mixed with their neighbors, the Basques. This is probably where Basques get their Indo-European admixture and high frequency of R1b-M269.

  95. @Davidski

    You're choosing to be very naive. Like I said, genes don't speak languages, people do. All we can use genes for are to track migrations and thus help to validate earlier theories laid down by historical linguists.

    So, for instance, it was posited over 100 years ago based on linguistics data that during the Bronze Age there were large scale migrations from the Pontic-Caspian steppe all the way to Iberia in the west and South Asia in the east. That is now being validated by ancient and modern-day DNA data.

    It makes no difference whether people who now speak Basque, Dravidian, Uralic or Turkic carry significant levels of admixture from those Bronze Age migrations from the steppe. Why? Because genes don't speak languages, people do, and sometimes they switch languages, even back and forth, but overall, the correlations and patterns are rather clear.

    I think you'll find that Spanish_Pais_Vasco are the Spaniards with the highest levels of steppe ancestry in Iberia, much higher than what Basque_Spanish and even Basque_France have, so they can't be explained away as former Basques-speakers. But they also totally lack that Armenian-related signal that other Spaniards have.

    So I'd say that Spanish_Pais_Vasco are the best fit for a population relict of the Iberian Bronze Age and, more importantly, the likely Indo-European-speaking Beakers with the highest levels of steppe ancestry.

    This isn't surprising, because thanks to their geography, much like Basques, they were shielded from the recent exotic gene flows that affected other Iberians, and only really mixed with their neighbors, the Basques. This is probably where Basques get their Indo-European admixture and high frequency of R1b-M269.

    What a great post!

  96. The Basque argument is some seriously weak BS. Oh look, a Bashkir group that is over-whelmingly R1a. So, R1a can't at all be linked to Indo European. See how dumb that sounds? Go with the rule rather than the exception. Basques are mostly MN Iberian with maybe 30-40% Beaker ancestry. What necessitates them having to be IE? Anyway, it's not 2003. Get over the Basque and move on.

    1. This comment has been removed by the author.

    2. I haven't looked at it yet and doesn't really change anything. They're an exception, like those Bashkirs. It's not personal or anaphylactic. If I ever based my claims on an outlier for a paper at school, I wouldn't be a straight A student. They're not the norm, and nothing to base an argument on.

  97. @epoch2013

    Thanks for the info on Ala I Hispanorum Auriana

    Nice to be able to find a picture of the garrison at https://www.ancient.eu/image/4710/ I haven't found info on the origin of the soldiers yet.

  98. Yo Chad, once the genomes for North Africa are out we are gonna need qpGraph. Landscape teeming w ghosts right there...

    1. Oh, I am planning on it. I've got a feeling something is there.

  99. Check this out. Perfect sorta problem for qpGraph.

  100. I'll do what I can, but there is a issues with Africans in trees where you don't use just panel 4 or 5 with the Human Origins set. My mind is not coming up with the right words to describe the issue. Too many late nights after midnight and 1am on papers and getting up at 5am for work.

  101. Yes. I seriously need some sleep, but there is too much new stuff I want to look at, between Iberian farmers, Germans, East Asian trees, North Africans... I could almost call into work to look at some stuff!

  102. Trying to fit in some more work on the African tree and a new French project too. Stretched for time here...

  103. I'm saying that the minority admixture from BB took on the language of the Chalcolithic population they mixed into. Not really a problem when you consider they are some of the least Yamnaya and Bell Beaker admixed people. Bashkirs have founder effects with different groups. Different subpops have different YDNA dominating. It is a founder effect. Same with the Basque. Almost all DF27. A few or even one DF27 guy mixed with older Iberians 4000 years ago. A little impact from Celtiberians, and other surrounding Spaniards put in a bit of other R1b. Nothing complicated or unrealistic about this.

    I know a lot of about European history from Aurignacian to present. No worries there. Who called BB origins years ago and said no Iberian input?

    1. @ Chad
      I disagree. That Iberia is the tail end of Yamnaya admixture is irrelevant
      What matters is the massive social rupture and significant genetic impact after 2400 BC. Therefore it is highly unlikely that Basque is Chalcolithic
      And yes, you know more than the average punter here
      And you’re also wrong about Northern Magdalenians. :)

  104. Unrelated, but I think the Magdalenian thing I have here is going to stick too.

  105. @Chad,
    "Who called BB origins years ago and said no Iberian input?"

    Kudos to you for that. It should be recognized this strengthens the credibility of any new hypotheses you make.

  106. Basque isn't young, nor did it likely cover much of an area more than it does today. Maybe less in Roman times, looking at tribal names there, with Vascones being only one of several. Sociolinguistics, a regionally isolated group. There is not Vasconic influence across all of Beaker territory. It is not related to Bell Beaker. I don't really care to spend more time on them.

    I am certain that my work on the Magdalenian admixture is correct, so I don't see how it would be wrong. You would have to be able to show me how it is a worse statistical fit, because I can't see how.

  107. Chad
    I’ll disprove your magdalenian theory in one line:
    They’re were I(xI2), U2-8
    Azilians are I2, U5b
    Ergo- no direct link

    1. Only a handful of samples. U5b in ElMiron. Like it or not, Magdalenian admixture is looking good.

    2. Chad
      Yes maybe southern, epigravetian admixed, Magdalenians

  108. @Rob.
    Iberia_BA 78.3 %
    Armenia_EBA 20.1 %
    Bonda 0.9 %
    Han 0.7 %

    That model might work for Spain but not for France.







  109. I've been crying about recent near eastern ancestry in western Europe for four years. That was at the time before Natufian or CHG/Iran Neo had been discovered. My only method for measuring it was ANE K8 or ADMIXTURE tests on GEDmatch so I speculated on heavy southwest Asian admixture.

    The Armenian signal being related to Celts is an interesting idea because it seems to also appears very lightly in Britain. But I super doubt that. I'm not yet convinced Beaker folk weren't proto-Celts.

    Even if Beaker folk weren't proto-Celts, there's no good reason to suspect proto-Celts weren't a standard Yamnaya/MN farmer mix.

  110. @ Sam
    I’m not proposing anything universal. Only for Iberia
    As we get more and more data we will get a finer picture - and recall you yourself noted additional shifts in Britain after B.B.
    One thing we have to keep in mind for now is that people in the past were not as monolingual as most of us are today; or at least there were always segments within a society which spoke 2 or more languages. Haplogroups like L51 were in different cultures (Vucedol, B.B., Nagyerev, etc) which actually differed and were even antagonistic. Therefore the linguistic correlates are going to be complex

  111. @Chad

    Especially check if that affinity to WHG that Natufians and Anatolians have is absent in the Iberomausurian sample. Maybe it is *only* Basal Eurasian.

  112. @Rob

    "I’ll disprove your magdalenian theory in one line:
    They’re were I(xI2), U2-8
    Azilians are I2, U5b
    Ergo- no direct link

    On account of the Y-DNA I think they basically aren't enough calls for those Magdalenians to be called anything more than I. That doesn't mean they are not I2. See Genetiker:


  113. @ Epoch
    Fair enough; but phylogeny supports the likelihood that they’re not I2. And there’s also the mtDNA; and several other facets of material evidence - notably- a 1000 year hiatus in the Rhineland between Magdalenian and Azilian-FMG, and the fact that they’re different cultural systems
    I.e. these are distinct peoples

  114. Both I1 and I2 calls in one sample, I2 also in another. There is a lot of cultural overlap with Azilian and Magdalenian. Enough that many thought Azilian to be a cruder continuation. They're sure looking 25-45% Magdalenian in each sample in France, Germany, and Belgium between 6500-12,000BCE.

  115. @epoch2013
    U6 is very intriguing. However, even more mysterious is the combination of M and U6 in palaeolithic Europe compared to M1 and U6 in Taforalt 15 kya.

    European M is the following: Aurignacian Belgium GoyetQ 116-1 c. 34 kya, Gravettian LaRochette France c. 27 kya, Gravettian Ostuni1 Italy c. 27 kya.

    Gravettian Ostuni1 from Italy is assigned to the Vĕstonice cluster and Goyet is unassigned.

    Upper Paleolithic Peştera Muierilor Muierii2 from Romania is c. 33 kya old and U6. Muierii2 is not assigned to any cluster.

    It seems that WHG is not the right reference in this case, but the fact that Taforalt do not show any relatedness to WHG and only to Natufians goes against any European influence. However, I think that there is a connecton between M and U6 and between Taforalt and Palaeolithic Europe.

  116. @Kristiina

    In the Supp Info Natufian is modeled as 1/3rd WHG, and Taforalt as 70% Natufian/30% African. Fig S19. So I reckon that they don't need *extra* European admixture.

  117. @Rob

    "a 1000 year hiatus in the Rhineland between Magdalenian and Azilian-FMG"

    That could be Laachersee related. Would also explain why the U8/2 mtDNA lineages went extinct.

    "Yes maybe southern, epigravetian admixed, Magdalenians"

    Indeed. Paglicci74 has U5b but Hervella 2012 also has a Spanish Magdalenian as U5b, although it has another as R0/HV. So it may not be reliable. Or R0 - or even HV -existed among pre-LGM Europeans..

  118. @Davidski & Azra

    "It seems that you're horribly overfitting your models for the Ostrogoth just to get rid of any potential Scandinavian signal. A little less bias would go a long way methinks.

    Try much simpler, chronologically more parsimonious models with Nordic_IA as one of the few reference samples and show us what you get.

    Ah. Off course. That 10% EHG plus firm chunk of farmer from the Greek Anatolian are basically proxy masking for northern ancestry.

  119. I'm pretty certain of this analysis. Significant contribution from Magdalenians to WHG north of the Balkans. Remember how Baalberg was P1 and R1 when they were low quality like Magdalenians. If better coverage is achieved, don't be surprised to see the I1 and/or I2 calls to be legit.

  120. Arza,
    Per above, I'm a bit skeptical too right now about the particular sample's "Germanic" ancestry** and, the reality of it aside, with current samples seems to prefer some sort of "'Greco-Anatolian-Armenian'-steppe with some eastern stuff" model more in nMonte. But I don't see how just one sample that likely comes from the Cimmerian Bosporus at the site of a Greek colony necessarily tells us anything but that those migrating confederations could end up very mixed and incorporating various ethnolinguistic groups and influences.

    I understand you likely believe that Wielbark and Cherniakhov didn't involve anything Germanic at all but the possibility that Gothic went from the Balkans to the steppe is on the same level as e.g. thinking that Slavic made the same trek. Every historian who's studied the sources agrees that the Goths appear initially raiding the Balkans and Asia Minor from north of the Danube and the Black Sea, even if we want to remain agnostic about their previous potential affiliations.

    ** but e.g.

    [1] "distance%=2.3639"




    [1] "distance%=2.3859"



    1. Remove Anatolia Chl. That's about 5000 years in the past from this time.

  121. @Alogo

    Those models could point one half blood Gothic grandparent. That fits perfectly. Even the tad Karasuk hinting at some Hunnic blood.

  122. @epoch,

    It's possible but until better tests come along from more qualified people than me (more samples from the region would help too, of course) I'll remain a tad skeptical of that specific sample harboring much of it at least. I had to torture it a bit to get Scandinavian-like stuff in Global25. Instead, the Gepid from Serbia for example comes out as a Scandinavian-steppe-Balkan-like mix (which would make sense if that's actually the case) very very easily as David showed too, e.g.:

    [1] "distance%=2.388"




    [1] "distance%=2.5923"



    We'll see!

  123. I posted a model with more proximal sources in my first comment but the fact that the later Pontic Sea area, both south and north, isn't sampled that well doesn't help and might be giving out some red herrings. You're right of course but Anatolia_Chl seems to resemble some sort of local ancestry this sample has relatively well though (and seems generally very close to modern Anatolian Greeks).

    Here's an alternative, just using a few Scythian, Nordic, Balkan, Anatolian, Armenian, Levantine and "Celtic" references and seeing what it comes out as:

    [1] "distance%=2.6454"



    Reasonably pruning a bit and seeing what pops up every time I end up with

    [1] "distance%=2.7177"



    Removing the German outlier because who knows the exact background, even though it seems mostly Balkan/Aegean-like:

    [1] "distance%=2.8935"



  124. @Alogo

    Then she might have been a woman of Greek descent, having been subjected to a Hunnic skull deforming ritual, buried in a Gothic fashion. Maybe a local woman married to a Crimean Goth. Or maybe the burial habbits were influenced by the Goths. The

    She certainly is not worth 16 samples, as Azra mentioned. Besides, radiocarbon dating of her remains could mean she lived before, during or after the Goths ruled the place or before, during or after the Huns sacked it.

  125. @Rob

    I like your Sardinia model. It's realistic as it's consistent with influence stemming from a pre-Beaker Western warrior complex associated with anthropomorphic stelae. It's also realistic with respect to the Aegean admixture.

    As for French_East:
    That Armenia_EBA signal is a geometric composite(vector sum?) of additional steppe relative to Iberia_BA+minor Roman-age ME admixture(consisting of "Greek" and Levantine introgression). Ashkenazi admixture in France is not uncommon, either: at least one of the French samples from Dave's PCA has pretty significant CM sharing with them.

    As for Extremadura, Mozabite+Armenia_EBA represents a total of North African(Roman age and post Roman),Italian, "Greek"(mainly Roman-age, I would think, but Guadalqivir may have had limited interactions with Aegean and Anatolia(e.g. Tartessos), and Levantine influences. Roman era Guadalqivir area(Baetica, etc) would have been a cosmopolitan region. Converso assimilation is also another factor to bear in mind.

  126. @Arza

    Thank you!

    Indeed, a nice intersection and produces quite a crisp fit(although Tuscan+Samaritan+Mozabite produces a passable ~2% fit, as well). If the ghost you've derived is real, it implies ~12% Germanic contribution to Tuscans and/or local individuals with higher Europe_MLBA ancestry being more important in the long run.

  127. @Davidski

    I'm not sure that what I wrote about boils down to mere overfitting. My father's mother prefers Nordic_IA from Sweden over Germany_Medieval from Bavaria, and that seems historically sensible, with ancestry from Holstein for instance. So my father ought to have some Nordic_IA as well. The problem's just that in his case this eats up all his Baltic percents. So I have to delete the Nordic_IA in his case, to allow his Baltic percents to emerge, even though he must have some Nordic_IA-like stuff from his mother.

  128. A Global 25 tour through Iberia to measure North African admixture gives very similar results to the ones on the paper about it. Not that we didn't know, but good to test it with ancient samples (Old Europe + Yamnaya + Armenia/Anatolia + Moroccan).

    Basically close to 10% Moroccan in the West, down to 1.8 in the East (Valencia), and around 5% in between, with Basques all at 0%. Canary Islanders go up to 21%.


    It can also be seen that Yamnaya is quite stable around 30%. And what came with it (Globular Amphora, some Czech_MN, Tisza_LN, Protoboleraz_LCA...). Most with Bell Beakers.

  129. @Simon_W @Andre

    Btw, been playing around with Italian_Bergamo.

    In models where it takes extra ~30% Germany_Medieval_average(I used the custom one above to chop out SE-related admixture), it does just as well with extra ~30% BB or Unetice.

    It's somewhat disheartening because we may never be able to tease out the Germanic stratum from earlier BB/Unetice-like Polada, Halstatt, etc. waves emanating from Central Europe unless we resort to rare allele, CM comparisons and other finer methods with a plethora of pre-Roman, Roman-age and post Roman sampels from the region.

    Judging by the frequency of R-u152 and the paucity of other R1b (e.g. u106) and I1,though, it's not likely that Germanic admixture makes up the bulk of that extra signal.

  130. @Anthro
    I don't believe Unetice had much of an impact in Northern Italy. Beaker_Northern_Italy was the result of the first steppe admixed wave. Polada may have had some influence from EBA southern Germany. (Perhaps similar to south German Beakers.) Terramare-Palafitte may or may not have had some Hungary_BA:I1504-like admixture. The Celtic influence may have been rather French-like. I you play around with these, and also add the Germany_Medieval, what does it prefer? I suppose this also depends on what southern/southeastern admixture you assume in pre-Lombard Bergamo.

  131. That theory connecting R1b and Celtic was popular in the internet before we had ancient DNA. But this is 2018. I think it's time to put that myth to rest. We know that R1b expanded in Western Europe almost 2000 years before the Celtic Language did. Let's try to catch up.

    And while at it, let's try once more to put another myth to rest. The one about Basques being an exception. Guys, some minimum Wikipedia:


    All the areas in red, green and yellow are the Iberian and Vasconic speakers. Whether they are related or not is not clear, because of poor understanding of the languages and because they had a much higher diversity than the very recently arrived Celtic. The Iberian part was also the most developed and populated, so in Iberia those non-IE speakers clearly outnumbered the Celtic ones. Adding France and Great Britain probably Celts outnumbered them, but still these non-IE speakers might have been a sizeable 3rd of the population.

    For the autosomal genetics of their descendants, see the post above. The Y-DNA you all know it by the book.

    And here their Bronze Age ancestors using the exact same setup as above (you know their Y-DNA too):


    The last one there is actually a possible descendant, that Roman soldier from South Germany. He does resemble modern Spaniards, but who knows. Here the distances from the 2 Iberia_BA samples and that German one to all other populations in G25:


    And some more Wikipedia to know something about Iberians and their culture:


    And now some autosomal comparison using the exact same setup as above, but with Mycenaean samples (once more):


    It's also true that NE Caucasian populations are autosomally one of the closest living people to Yamnaya. Their Y-DNA is quite subject to founder effects, but some of them certainly have a decent amount of R1b-Z2103.

    I hear that also Etruscan (or Tyrsenian in general) are possibly related to Hurrian and NE Caucasian too. Just hypotheses from linguists these things, but there's that.


    I take that being naive as a compliment, of course. In this place where everyone already knows everything I feel refreshed for having this naive and new look at the data instead of sticking to a decades old hypothesis (one among many others, BTW).

    If others were not so scared about the possibility of IE languages not being from the steppe this place would be more interesting. Really, what's so bad about IE languages coming from anywhere else? (no preference, just let them come from where ever they did. In the end we'll all agree when things are clear anyway).

  132. As for the Ostrogothic skull, I think some people have already pointed it out. But if you check the info in SI1 it really throws some serious doubts on the ethnic classification of this skull as Gothic. There is no archaeological context, apart from the info that it's from a burial chamber where a diadem was found that belonged to another skull. And this diadem may date to 400 AD and was inspired by Hunnic fashions. The city was an ancient Greek colony, hence the locals may have been a mix of Greeks and non-Greek natives. The Crimea is close to the Caucasus BTW. Southeastern CHG related admixture may be from the nearby Caucasus as well. Goths took over the city around 350 AD, and the Huns followed 20 years later. So the skull could be a mix of Hellenized locals, Goths and Huns or even without Gothic admixture.

  133. @Alberto

    When I said you were being naive, I didn't mean that you were open minded and original in your interpretation of the new data. What I was pointing out was that without your straw man theory, in which genes speak languages, you don't have any viable arguments against the steppe hypothesis.

  134. @Simon

    I'll see later later today or tomorrow.

    If memory serves me right, though, I've used only the following:
    One Beaker resembling Germany_Roman(Occitan-like)
    Minoan and Anatolia_BA
    Beaker_Northern_Italy(the one w/steppe but w/out Aegean admixture)
    Germany_Medieval_custom OR Unetice OR Beaker_Central_Europe average

    Basically, there was no significant difference in the fits.

    I'm thinking that Gaulish waves may have resembled either French or Occitans/Germany_Roman w/U152 and L21, but there was a continuous stream of "proper" Europe_MLBA stuff from Beaker times till then.

  135. @All

    I've made some major updates to the Global25 datasheets, taking out many individuals and replacing them with others. I also added a few new populations to the data, including the Welsh.

    Btw, I'll try to blog about the new North African paper later today or tomorrow, using an angle that hasn't been seen anywhere else yet.

  136. @ Anthro Survey
    I don't see how the Celtic waves around Bergamo could have resembled Occitans. The first wave, the Lepontii, came with the Canegrate culture, which shows influence from eastern France. And the second wave, the Insubres, were a pagus of the Aedui according to Livy, hence also from eastern France. Only if the pre-Celtic locals were kind of Iberian- or Basque-like, the resulting population after admixture would be Occitan-like, I suppose.

  137. Alberto,

    I wouldn't necessarily trust Wikipedia. Anyway, Celtic is not the only IE language(s) of Iberia. There is pre-Celtic Indo-European as well. In a land where maybe only 25% of incoming genes are linked back to Indo-Europeans, I see no issue as to why some isolated mountain groups would retain some non-IE speech, as the IE groups are likely not in their sphere of influence. We could even be talking R1b getting into the mountains due to some type of scouts or traders as well. Look at the R1b in NE Canadian tribes without a whole great deal of European input. Founder effects can have some dramatic effects. We don't know what the BA Iberian spoke, but the area of Central and Western Iberia is expected to be IE speaking, out away form the Pyrennes groups that are isolated and not even united in their own right.

  138. To my understanding, overfitting means you get better fits using more populations than are really in the tested individual's/pop's ancestry. In contrast the problem I noted is that in some cases you get better fits using less populations than are in the tested individual's ancestry... It's not a terrible problem, though; I'm still a fan of this Global 25 / nMonte approach.

  139. @Alogo, yeah, I think the fit David posted for the Gepid does look pretty feasible.

    Re-sampled PCA with averages for Global_25 for just fairly West Eurasian populations and just Iron Age or most recent: https://imgur.com/a/s9V6p. I think it could be useful to visualise certain things (the big black broad convex hull is the ACD samples).

    Raises whether Balkans_IA+Germanic/Nordic_IA might work well for various of the ACD samples without the East Asian ancestry? It seems like that may work; it looks like there might be some thing going on with these samples where the don't overlap modern SE Europeans in some G25 dimensionality where NW+SW European/NE+SE European are distinct, and Balkans_IA is intermediate. (Though Gepid is maybe not 100% in the "right" place on this dimensionality, that said).

  140. As far as Basques are concerned, you cannot ignore some facts. First of all that basques or whoever was their ancestry, was a matriarcal society, with a language showing so. That means that if by any chance, men disappear by following their sheeps, as they were agriculturalist and pastoralists, women stayed working the land. And if they had to, they took other men, foreigners to have sons. So they kept their language, their families and their traditions, and men didn´t interefere.
    Nobody knows the deseases brought by inmigrants from the East that could have affected men. Remember measles produces infertilty in men. So women could have taken men in their own terms.
    And these facts have been true, when basque men emigrated to America, and the women stayed in the country, in the XXth century.
    The subject about the Goths having no scandinavians or few scandinavians ancerstors, has been known since a while. The fact has been confirmed through genetic studies.
    There are different tales about the Goths, being Jordanes only one of them.
    If they left Scandinavia two hundred years BPA, and lived close to the Black Sea, moving through the steppe, with other tribes for five hundred years, when they entered into the Roman Empire History, we should expect these roaming people to have made alliances with other esteppe people, whatever their origins. We could even relate them to the Gutis, or to the Getas mentioned by old historians. And of course, they were all mixed up with other ethnias.
    For sure a sort of Germanic lingua franca was spoken by all of them, including Turks and Hans. There are tales written by Icelanders, talking about Uldin who became Odin , and their human sacrificies introduced in Europe from people coming from central Asia in the time of invasions.

  141. @Olga

    According to historical records the Celts were also a matriarchal society so we know that line of reasoning is a non starter.

  142. @Romulus,
    "Celts were also a matriarchal society"

    Presumably, some Celts were more egalitarian than normal but that wouldn't make them matriarchal.

  143. 2.2147



    Global25 isn't good for detecting post-Beaker waves from Europe into Britain. Modern Isles Celts seem to have a kind of low-WHG continental ancestry. I'm not sure. It could have something to with Celtic languages. But right now, with the resources available, I can only speculate.

  144. Welsh seem to be very very slightly more 'northern' than English. I wonder if there's somekind of French ancestry in England.

  145. @Alberto

    A language map based off of Roman records doesn't prove or disprove the adoption of the language or what the DNA was like of the original speakers of Euskera was like prior to the Bronze Age.

  146. @Samuel Andrews

    I agree not fully matriarchal but figures like Boudicca make a strong case for a deviation from the norm. The case there is much stronger than Olga's far fetched connections between language and the status of women in Basque society.

  147. @ Epoch

    "That could be Laachersee related. Would also explain why the U8/2 mtDNA lineages went extinct."

    That was later.

    It goes like this
    13500-12000 BC: north Magdalenian/ Hamburgian (U2/ U2-8*)
    ---------------- population change
    12000-11400 new Federmesser coming from south (Azilian) (U5b- Bichon (Azil.), Oberkassel - FMG)
    -> LacherSee: Population concentrates in Jutland - shifts to Brommean (11400 -10200 BC)
    10200-9000 BC Ahrensburgian - Swiderian. Expands from Brommean through north Euro plain
    9000 BC onward Maglemose early Mesolithic (and appearance of U5a now)

    1. I really wouldn't be too sure of that. U2 and U8 are in Neolithic Scotland, Germany, Hungary, etc.

      Again, don't be surprised to see a lot of I2 and maybe I1 with higher quality samples. The poor Baalberg R1b calls ended up being legit. There are a few I2 calls in Magdalenians. My analysis is showing Iron Gates and Magdalenian share a lot of ancestry after Vestonice, but before ElMiron, 10-15ky after I2 developed. There's absolutely no reason Magdalenian can't be mostly I2.

  148. @Simon_W

    Ok, using your suggested pops, this is what I got:
    [1] "distance%=0.68 / distance=0.0068"


    French_East 52.35
    Minoan_Lasithi 25.85
    Beaker_Northern_Italy:I2478 14.65
    Custom Germanic 4.70
    Anatolia_BA 2.00

    In this run, it took neither 1504, Unetice, Central European Beaker av, nor the steppe-less North Italian Beaker. This may or may not mean much if fits clear the convex hull using the others, but for simplicity's sake, here's what I'm thinking:

    As for Gauls being Occitan-like-- To be more precise, I meant something Occitan-shifted compared to the average BBs like the Roman soldier from Germany or Beaker I7283. Consider the possibility of there initially being an excess of EEF-like communities relative to German BB-like folks. Over the following centuries, fusion of the former and a stream of migrants from Central Europe produces Aedui et al ultimately more EEF-shifted than modern French_East.

    If this was the case, we're looking at potentially 25% Germanic input:
    [1] "distance%=1.4255 / distance=0.014255"
    custom_medieval_german 54
    Germany_Roman 46

    Note that unless this assumption is made, it's virtually impossible to determine the Germanic portion in French_East per Monte's resolution.

    Consider these two statistically equivalent models:
    distance%=1.993 / distance=0.01993"
    custom 70.40
    Beaker_Southern_France:I1388 27.05
    Unetice 2.55

    Now, if I restrict to only Unetice, we still clear the convex hull pretty well:
    [1] "distance%=2.5176 / distance=0.025176"
    Unetice 64
    Beaker_South[EASTERN]ern_France_lower_steppe:I1388 36

    Then again, it could just as likely(if not more so) that Aedui were already looking like French_East. This is probably more realistic, tbh, and a scenario I FAVOR but I'm just curious to test out different scenarios.

  149. @ Chad

    "Anyway, Celtic is not the only IE language(s) of Iberia. There is pre-Celtic Indo-European as well."

    Yes there was eg Lusitanian. Still, this is not likely to be 1500 years earlier than Celtic propper, but perhaps a couple hundred years older. Remember, it is also only attested in the Roman Era.
    If pre-proto-Celtic expanded with BB, then how did it stay uniform for so long ? This would be hard to given the post-BB collapse of the Atlantic facade network.
    To settle these questions, aDNA from more recent prehistory is needed as well as a flexible mindset ; although it has to be said that the non-IE languages were prediminantly on the Mediterranean side of Iberia

  150. I think the Bronze Age Crises which affected Myceneans Etc had knock on effects in temperate Europe, which might have propelled linguistic shifts
    Little wonder soon after this we see expansion of cultures solidly linked to Proto/para Celtic and pre-Germanic

  151. We should also keep in mind that if the current ancestral composition of the Basques owes itself to migration of different IE-speaking "adventurer" groups from central Europe, none would have the chance to assert its language and would in turn be forced to linguistically assimilate.

    I'm open minded when it comes to the idea of some early BB groups speaking Vasconic languages though.

  152. Matt,

    Thanks a bunch.

    As for your interesting question about what might be going on with medievals vs moderns there, ideally I'd have to recheck later to give you a decent enough answer but something I noticed earlier when trying to model even the samples who seem generally quite similar to modern Balkanites (and is similarly visible in your PCA if I take a simplistic approach) was this sort of thing: https://pastebin.com/j9N0BQJQ

    At face value, it seems that the admixture this sort of sample has just coincidentally makes it relatively similar to modern day Balkanites and most of the actual admixture that generally took the moderns perhaps from something close to Balkan_IA-to-Balkan_BA-like to present-day came a bit later with another, more important and somewhat different migration (but things might be even more complex than that). Probably too much hypothesizing right now though.

  153. @Alogo

    Also try using a combination of
    Germany_Medieval_outlier:STR_300 and Germany_Medieval_ACD:STR_355
    for the putative Late Roman substratum in modern-day Balkaners.

    Something interesting happens w/Albanians and Macedonians. These areas are proximal and we'd expect a relatively similar substratum at first glance, but this isn't the case when we force both of them to the same Slavic proxy(i.e.lock Baltic_BA and Slav_Czech w/chosen proportions into one custom sample). Albanians get a substrate that's more outlier300-shifted. Even when the Baltic_BA and Slav ratios are allowed to vary, the substrate is still more SE-shifted for Albanians but Macedonians get considerably more Baltic_BA in relation to Slav_Czech.

    So, what do you think ultimately played a bigger role: variability in invader groups or in the substrate?

    The substrate could be significantly different, ofc, if the Balkanic ancestors of Albanians fled the coastline(as we're told) and settled the hills where their ethnogenesis took place in the early middle ages. Coastal areas would have been more susceptible to either Roman-age or Bronze-age Aegean influence.

  154. @Rob

    One could imagine that the Younger Dryas could have caused havoc among the Magdalenians/ Hamburgians and even decimation of the population. The Younger Dryas was a return to glacial conditions which temporarily reversed the gradual climatic warming after the Last Glacial Maximum started receding around 20,000 BP. The Finnish Wikipedia gives the dates from 12,700 to 11,560 years BP). "The change was relatively sudden, taking place in decades, and it resulted in a decline of 2 to 6 degrees Celsius and advances of glaciers and drier conditions, over much of the temperate northern hemisphere."

  155. ... The Finnish Wikipedia notes that the decline in temperatures was biggest close to the Atlantic where the average temperature dropped at least 6 degrees. Therefore France could have been more affected than other areas. The cold spell lasted 1 300 years.

  156. @ Chad

    " My analysis is showing Iron Gates and Magdalenian share a lot of ancestry after Vestonice, but before ElMiron, 10-15ky after I2 developed. There's absolutely no reason Magdalenian can't be mostly I2"

    I doubt there is any real Magdalenian ancestry in Iron Gates. The Magdalenian didn;t even reach Hungary or Italy, let alone Balkans.
    Rather, what you're perhaps detecting is the shared ancestry between Magdalenian and pre-second-wave-ANE Balkans/ East-central Europe.
    In other words, there were 2 or 3 waves of movement east to west during/ after the LGM, each successively more ANE admixed (one way or another). That El Miron is the most Villabruna-like of the Magdalenians, its U5b; as well as the R1b in Iboussieres suggests to me that the post-Magdalenian wave might have expanded from SEE via Italy to Cantabria/ sth France.

    @ Kristiina

    The YD is too late to explain the demise of the Magdalenian in western Europe.
    See above .

    1. Rob,

      The shared ancestry is not ANE related or Magdalenian, but post Gravettian. It is the pre-extra ANE group with ancestry from a Boncuklu-related source. The Iron Gates samples could have some Magdalenian, but I model them with none. With them as a baseline, the southern Italy WHG is 13% Magdalenian and Villabruna is 25% Magdalenian.
      There is evidence of contact in Northern Italy, so no surprise. There also isn't anything preventing minor geneflow between Iron Gates and Magdalenians. It could've happened.
      U5b is also old enough to also be in Magdalenians. More samples are needed. ElMiron is not the most WHG related sample, but the rest are so poor quality that the stats are absurd with regards to relationships with Neanderthals, Asians, etc.. Just wait and see what more stuff brings.

  157. @Rob
    You give the end date of 12 000 BC as well as the German Wikipedia, but the English and French Wikipedia give the end date of c. 12 000 BP. Which one is correct?

  158. @ Kris

    14 kyBP is the end of the Late Magdalenian in Paris basin, which is from where the succeeding Azilian & FMG expand north.

    See 3/4 of way down:

    The issue is, an 'Epi-Magdalenian' continues in Iberia and central Europe (Poland, Bohemia) beyond that date, which lie beyond the Azilian - FMG expansion zone.

  159. I did some googling and found this article "Du Paléolithique final au Mésolithique dans les Alpes du Nord françaises: peuplements et environnements naturels" (http://docplayer.fr/41760546-Du-paleolithique-final-au-mesolithique-dans-les-alpes-du-nord-francaises-peuplements-et-environnements-naturels.html). When I look at Fig 1., I see that the Magdalenien settlements start before the oldest Dryas and continue through Bølling and Allerød interstadials more or less until the Younger Dryas. However, this paper is only about the Magdaleniens of the Northwestern Alps and I do not know if it applies to the demise of Magdaleniens in general.

  160. @Kristiina

    There is this promising article about a different region: https://www.sciencedirect.com/science/article/pii/S1040618215014688

    But it behind paywall, so i can't tell what is their conclusion about the end times. Also, it is not about the Northern region either.

  161. @Rob
    That's interesting. So does this mean that in the Northwestern Alps Magdalenien lasted longer than in the Paris basin where Federmesser started c. 13,900 years ago. One could imagine that Federmesser came from the east. However, Azilian seems to have originated in a more southern area as its core area is northern Spain and southern France. If Magdalenien started from the same area as Azilian later on, there could be a connection that way.

  162. @ Kristiina

    People used to write that the Magdalenian expanded from Franco-Cantabria. That's not entirely correct, because it too might have come via the East (Bagedoulian culture).
    Then, IMO, the Azilian expands from the Pyrenees. Either fresh immigrants or a subset of Magdalenians who were in contact with Italian Epigravettians, moved through France, and morphed into Federmesser north of the Seine. From these Federmesser groups then came Ahrensburg, Tanged points, & even the western-wing of Kunda.

    And yes, the Magdalenian might have survived in the Alpine regions- as you mentioned, and I alluded to - Bohemia. Of course, we still see U2 lingering around (Motala, Latvia, Bad-Durenburg)

  163. @Alogo

    Oh boy. Scratch most of what I said earlier. I mixed up my Late Roman Balkan proxies.

    It should be STR_310, not 355(which is just a regular Germanic sample).

    And this is how it looks like:

    When I give Monte leeway, it opts for the Trzinnec-associated Baltic_BA and the substrate ratio is almost identical in Albs and Makos.

    When I use the 50-50 Slav_Czech-Trzinec, it's somewhat, but not drastically different(1.7 vs 2.0 ratio).

    It makes more sense, tbh, considering a connection through Via Egnatia which could have "corrected" any imbalances of Aegean-related ancestry.

  164. @Alogo, cheers for running those models.

    I've run (yet more I know!), a Linear Discriminant Analysis (rather than a PCA) on a slightly more restricted set of the samples than I used in my last last. Basically still modern+Iron Age, but restricting on the South Central Asia, North African, Middle Eastern samples.

    I wanted to try and take away some clutter so I could look at each of the ACD samples and get a sense of where they sit on clines (by eye).

    Results: https://imgur.com/a/EAHvq

    Generally in line with your work, it looks like the samples can fit mostly as varying mixes of the Scythians/Sarmatians, Nordic_IA, German_Medieval_outlier.

    Looking at the most outlying ACD samples from their populatio convex hull, and their differences:

    AED_1108: By far the most East Eurasian influenced and I would guess influenced by some Scythian(AlbyBel/Pazyrnk/ZevakinoChilikta) like population. Almost a straight German_Medieval_Outler+Scythian admix?

    STR_328: looks not to have outlying East Asian influence, but a slightly outlying "east" position in West Eurasia dimensions, so I would guess influence from Sarmartian_Pokrovka / Scythian_Samara (e.g. maybe same population as AED_1108, but East Eurasian influence uneven in this population)?

    NW_54: Closest to a straightforward mix of Germany_Medieval_Outlier/Balkan_IA and Nordic_IA, but still with some offset towards probably Scythian(AlbyBel/Pazyrnk/ZevakinoChilikta).

    BIM_33: Like the ACD sample most influenced by the Baltic_BA cline, and the one that sits closest to present day Hungarians/East Germans/Austrians and Croats relative to NW Europe.

    STR_355: The most Germanic / NW European of the ACD samples, and not hugely distinguishable from present day England/Netherlands/Scottish/Irish/Welsh

    STR_228: Largely intermediate all the above

    I think my impressions are mostly in line with your runs for NW_54.
    Only real other comment I would make on your fits (which look pretty good) is that the Sarmatians may matter as it looks to me from the samples that Scythian/Sarmatian like groups (who may have spoken quite different languages from each other) who entered Europe may have been quite different from each other and uneven in East Eurasian like ancestry, so that may be important in the models. I'm not sure if you used them and they just picked up 0 or whether you didn't use them.

    I think as well, this does make me wonder (and I think these could be similar thoughts to Anthro Survey's and probably others in this thread) whether the German_Medieval_Outlier actually is representative of a fairly Balkans_IA like general cosmopolitan European Roman population.

    (Consider, as was discussed in some previous threads, similarities and differences between Balkans_IA / German_Medieval_Outlier and present day Italians, and the different streams of admixture that would be required to get to the various present day Italians from a Balkans_IA / German_Medieval_Outlier position and how they would fit with the historically described migrations to impact the late Roman Empire.)

    Which in turn makes me wonder whether the apparently Iberian like Roman solider ("Maximus" ;) ) from Germany is actually so Iberian like specifically, or he is actually a phenomenon of this generalized Roman+local German?

  165. @Rob

    That powerpoint is interesting as it considers "art" very Magdalenian. An Ahrensberg site in the Netherlands had this:


    Another interesting thing is that horse hunting is considered typically Magdalenian. We have a U5b almost contemporary to El Miron from the Paglicci cave. There are also cave painting in that cave that depicts horses, for as far as I can find considered Epigravettian.

    Just loose thoughts, mind you.


    Or maybe the odd unknown admixture in Iron Gates is actually the absence of Magdalenian.

    Also just a loose thought.

  166. There isn't anything odd or unknown. At the root, ElMiron and Iron Gates are mostly a mix of Vestonice and Boncuklu. There's something related to GoyetQ116-1 in ElMiron and MA1 into Iron Gates. Iron gates does have a Z pull>2 towards Goyet yet, so they may have very minor ~10% admixture from Magdalenians or something very closely related.
    I'd like to see epiAurignacian and see if they are like Magdalenians.


  167. @ Romulus

    The existence of Queens, is not a proof that a society is/ has been matriarchal. Proves are facts in relation to property, inheritance, capacity to buy, sell, sign contracts without the supervision of a father, brother or husband.
    You may find proves in the language, when the cognates have different names depending if the relation is maternal or paternal.
    You may find proves in the use of the last names. In Iberia, women don´t lose their family names when they get married. More than that, they keep calling them by their birth names. The status of the mother´s family is as important as the father´s.
    And all these facts happen in the middle of a patriarchal world.
    I have never used my husband's last name, and my signature is with may name an my father´s and mother´s name. Husbands come and go, family stays.

  168. @Chad

    I was refering to this from the Mathieson paper:

    "We modelled Iron Gates hunter-gatherers as a mixture of WHG- and EHG-related ancestry (Supplementary Table 3), which showed that they are intermediate between the two (WHG contributing approximately 85%, and EHG approximately 15%, of ancestry). However, this qpAdm model does not fit well (P = 0.0003, Supplementary Table 3) and the Iron Gates hunter-gatherers show an affinity towards northwestern-Anatolian-Neolithic-, relative to WHG-, ancestry populations (Supplementary Table 2). In addition, Iron Gates hunter-gatherers carry mitochondrial haplogroup K1 (7/40) as well as other subclades of haplogroups U (32/40) and H (1/40), in contrast to WHG, EHG and Scandinavian hunter-gatherers—almost all of whom carry haplogroups U5 or U2. One interpretation is that the Iron Gates hunter-gatherers have ancestry that is not present in either WHG or EHG. Possible scenarios include genetic contact between the ancestors of the Iron Gates population and a northwestern-Anatolian-Neolithic-related population, or that the Iron Gates population is related to the source population from which the WHG split during a re-expansion into Europe from the southeast after the Last Glacial Maximum"

    Combined with the fact that in a tweet (for what that's worth) Mathieson stated this:

    "Iron_Gates=WHG+EHG+AN doesn't fit though. I guess that the Greek/Southern Balkan Mesolithic is mostly like Barcin but with more CHG and maybe more WHG, and there is gene flow with Iron Gates etc."


    Although Barcin doesn't show such a strong signal as Anatolian.

  169. @Davidski:
    Looking forward to your blog entry on North Africa, sounds exciting!

    Also, sorry I didn't get back to you regarding the Turkish samples. I got sick and haven't played with nMonte since then. I'll check the updated sheet tomorrow.

  170. About the whole Iberia thing, just pointing out that a more profitable and illuminating thing to do would be to posit that some mixture of Spanish_Pais_Vasco and Basque represented an autochthonous population post-BA, and then add Mycenean, Jordan EBA and Anatolia BA on top of that, in addition to Mozabite, Medieval German and some Italian reference (I dunno which one is best to pick for this, depends on your estimate as to which aDNA sample best represents pre-Imperial Romans.)

    And exclude Armenia EBA, Armenia MLBA etc from the analysis, cos input from the Caucasus is difficult to interpret in the context of Iberia.

    Suspect the results will be very interesting, what with the different regional Spanish samples we have, which may give us regional variation that helps pinpoint the historic correlates of the contributions.

  171. Some G25 nMonte on the ACD German samples: https://imgur.com/a/xN8M3

    Contrary to my expectation, some preference for Armenia_MLBA in some samples, no preference for Sarmatians over Scythians in any of them. Essentially also, none of the samples other than STR328 and AED1108 took any Scythian; others are essentially largely NW/N European Iron Age+medieval Germany outlier with some low levels accretions of Armenia_MLBA. Some presence of Baltic_BA stuff but in different ratios to modern Balkans people.

    (Calc file: https://pastebin.com/vrcfaT8T
    Raw fits: https://pastebin.com/LGU8TV0E

    I've grouped a lot of the populations from the calc file/raw fits for the purposes of the above linked table)

  172. @ Chad
    We just need a Balkan late Paleo sample from the same time as Miron
    It’ll have Magdalenian- like affinities also, due to shared Aurignacian heritage
    Then it’ll just require some additional Boncuklu and ANE admixture

    Btw can you check
    Mbuti. MA-1. Boncuklu Barcin
    Mbuti. EHG. Boncuklu Barcin

  173. I'm not really seeing it like that in any graph. Not at all.

    MA1 and EHG stats are basically zero.

  174. @ Chad

    You mean there is zero ANE / EHG in Iron Gates ?

    1. No, I'm seeing them both coming out of Gravettian, with some Goyet in ElMiron.

  175. @ RK

    "About the whole Iberia thing, just pointing out that a more profitable and illuminating thing to do would be to posit that some mixture of Spanish_Pais_Vasco and Basque represented an autochthonous population post-BA, and then add Mycenean, Jordan EBA and Anatolia BA on top of that, in addition to Mozabite, Medieval German and some Italian reference (I dunno which one is best to pick for this, depends on your estimate as to which aDNA sample best represents pre-Imperial Romans.)

    And exclude Armenia EBA, Armenia MLBA etc from the analysis, cos input from the Caucasus is difficult to interpret in the context of Iberia"

    That is an interesting idea. I did not exclude Armenia EBA, but added your suggestions - Mycenean, BA Levant, Anatolia BA.

    The results are essentially the same:
    Spaniards are a mix of BA Iberia, Mozabite and "Armenia EBA". Perhaps some MBA Italian samples will fit the bill ?
    Myceneans, Levant BA & Anatolia BA make no impact; not surprisingly tbh (even Balkan IA was added).


    The real Chalcolithic peoples are Sardinians, with added Anatolia & Greek stuff. Makes sense given that Sardinians have heaps of M26, whilst Basques are ~ 90% DF-27.
    Also note the lack of Levantine admixture.

  176. @Rob

    What happens if you push Armenia_EBA out? I have a feeling its not real, but represents an aggregate that so happens to fall in the right direction.

    (Not sure what @Matt would make of this, it seems Global25 should be less prone to this type of "fusion" problem, e.g. it splits the Yamnaya in South Asians into Srubnaya_Outlier and Andronovo pretty well, so why fuse admixtures in Iberia?)

    Also the figures for Spanish_Pais_Vasco (91% Iberia_BA) are pretty incredible.

  177. @Rob

    Iberians have some extra Yamnaya-related input on top of what Iberia_BA and Iberian Beakers had. So add Yamnaya_Samara and see what happens to the Armenia_EBA ancestry proportions in your models.

  178. @Chad

    I wouldn't necessarily trust Wikipedia.

    LOL! I don't need to trust Wikipedia to know Iberian prehistory. If you think that map is wrong, then you have lots of academic material online to check that it's right.

    And again, Iberians, Turdetani and Aquitani/Vacones are not isolated groups from the mountains. Jesus, look at a map of Aquitaine:


    This is the core area of modern Basques. They were not hiding in the Pyrenees. That's low land in the Atlantic coast.

    And all Iberians proper and Turdetani (lato sensu) are in low land in the Mediterraean to Atlantic coast.

    So please, try to understand this, because it's important.

    You're basically proposing that:

    1 - The language of the Bell Beakrs in Western Europe went extinct in the early Iron Age, while the language of the EEFs stayed alive and well.

    2 - Those speakers of the language of the EEFs are some 25-30% Yamnaya and outnumber the Celtic speakers in Iberia. They are at least 50% Bell Beaker from Central Europe and around 79-85% R1b (consistent with all other Western Europeans populations, not some founder effect).

    3 - The language of the Bell Beakers that went extinct and therefor we have never seen, was IE for sure. Because it was.

    4 - At the same time, Mycenaean samples clearly prove that their language came from the steppe.

    5 - The Anatolian samples that we have were probably some lost Hattians who got killed by IEs, who lived right there but genetically will be very significantly different. So we can ignore those Anatolians.

    That's something I'm not buying. Too many very unlikely scenarios that all of them need to turn out to be correct at the same time.

    I prefer the simple and logical explanations that match the data as it is, without trying to fit it into what it isn't.

    When other data comes that changes this, I will adapt to it, whatever it is. For now this is what we have.

  179. Re: Lusitanian, yes, thats some evidence that I have to bear in mind. I don't like to dismiss evidence that's there.

    But you have to understand the limitations of that evidence to put it into context and know that you can't hang onto that alone.

    Lusitanian is very scarcely attested. More importantly, it is attested very late, and not in the Iberian script, as Celtiberian, but only in the Latin alphabet. This means that by the time it was written they were literate in Latin. And by the few texts that we have the language looks quite Celtic-like, but with some Italic-like features. It could be a proper pre-Celtic language. But it could just be a divergent Celtic language with late Italic influence.

    We need to know who exactly where the IEs and then get specific Lusitanian aDNA to solve that question. By itself it's not a very strong proof of anything (though I don't dismiss it, as I said). Time will tell.

  180. Alberto,

    No need for your numbered strawman crap. I never talked about most things on there.

    So, rather than Bell Beaker being NW IE that develops into Celtic, Italic, Lusitanian, etc. , which was lost in some places in Iberia, you would rather have Bell Beaker speak Vasconic and maybe 10 or more dead languages from Iberia, Aquitaine, Italy, Germany, Britain, etc. Right... A highly uniform pop that had just spread out, was ultra-multi-lingual, rather than the edges where Beaker ancestry is in the minority, having been only marginally impacted in an IE speaking network... Right, so much more logical. If languages split that much and fast into completely unrelated tongues, how can I still communicate with Brits or Aussies?

    Believe what you want though. I don't have time for 1-5 strawmans all day. Things to do.

  181. @ Dave\ RK

    Adding Yamnaya essentially removes Armenia EBA , and partly eats into BA Iberia. But it pushes up balkans IA in a major way (20-30%)
    Taken at face value, it suggests need for additional/ ongoing Central European admixture to Iberia (?halstatt, la Tène) after its EBA, and some south European (continental) input (perhaps indeed Roman) ; but very minimal “Mediterranean”.

  182. Yes, the extra Yamnaya signal is probably a proxy for Celtic and also Germanic admixture in Iberia. There is also some extra CHG-related ancestry in many parts of Iberia, but it's rather low, and I think it can be explained by admixture from Romans and Greeks.

  183. Maju and his rather limited cheer squad are at it again.



Read the rules before posting and ignore banned commentators.

Comments by people with the nick "Unknown" are no longer allowed.

New rules for comments

Banned commentators list