search this blog

Monday, December 28, 2015

Ancient genomes from Ireland point to population upheavals in Atlantic Europe during the Neolithic and Bronze Age

Open access at PNAS:

Abstract: The Neolithic and Bronze Age transitions were profound cultural shifts catalyzed in parts of Europe by migrations, first of early farmers from the Near East and then Bronze Age herders from the Pontic Steppe. However, a decades-long, unresolved controversy is whether population change or cultural adoption occurred at the Atlantic edge, within the British Isles. We address this issue by using the first whole genome data from prehistoric Irish individuals. A Neolithic woman (3343–3020 cal BC) from a megalithic burial (10.3× coverage) possessed a genome of predominantly Near Eastern origin. She had some hunter–gatherer ancestry but belonged to a population of large effective size, suggesting a substantial influx of early farmers to the island. Three Bronze Age individuals from Rathlin Island (2026–1534 cal BC), including one high coverage (10.5×) genome, showed substantial Steppe genetic heritage indicating that the European population upheavals of the third millennium manifested all of the way from southern Siberia to the western ocean. This turnover invites the possibility of accompanying introduction of Indo-European, perhaps early Celtic, language. Irish Bronze Age haplotypic similarity is strongest within modern Irish, Scottish, and Welsh populations, and several important genetic variants that today show maximal or very high frequencies in Ireland appear at this horizon. These include those coding for lactase persistence, blue eye color, Y chromosome R1b haplotypes, and the hemochromatosis C282Y allele; to our knowledge, the first detection of a known Mendelian disease variant in prehistory. These findings together suggest the establishment of central attributes of the Irish genome 4,000 y ago.

Cassidy et al., Neolithic and Bronze Age migration to Ireland and establishment of the insular Atlantic genome, PNAS, Published online before print December 28, 2015, doi: 10.1073/pnas.1518445113


Chad Rohlfsen said...

Surprise, suprise.. all BA males are R1b and steppe derived..

Paul Ó Duḃṫaiġ said...

All the BA males are more specifically R1b-L21, with one confirmed DF21+, Rathlin2 is also probably DF21+ but they only got 1x read on position so played it safe and called him as DF13+

Busby (Royal Society) had R1b-L21 at about 70% of his total sample for Ireland.

The first modern Irish full genome a number of years ago also happened to be R1b-DF21+ -- interesting enough one of the Hinxton Iron age samples from Eastern England was also R1b-DF21+

Roy King said...

Nice! All the BA Irish samples are M527 derived like those of contemporary Ireland. Should be a boon for Davidski's modeling!

Roy King said...

Sorry M529 derived (L21).

Nomen Cognomen said...

Something's not adding up here.

U5 is EHG/WHG maternal steppe lineage, R1b is supposedly paternal steppe lineage, yet only a third of their DNA is Yamna?

J2b is also a near eastern maternal lineage, not Caucus.

Davidski said...

The Irish Bronze Age genomes are mixtures of EEF (from the Near East), indigenous WHG and steppe invaders.

So what doesn't add up?

Ryan said...

Too early to be Celts, and probably (though maybe not necessarily) too early to be some other unattested branch of IE. Early IE wouldn't be enriching their WHG ancestry either. Probably a WHG (and CHG?) rich expansion that had EHG admixture too. Proto-Vasconic? Originating somewhere on the Danube?

I'd be interested in what you come back with when you get your hands on the samples David, but I'm calling right now that they'll have too much CHG and WHG to be explained by PIE migration.

Chad Rohlfsen said...


You're very wrong. You should've looked at the stats before making that statement. There is nothing CHG+WHG, without EHG. Not even close. The significant stats between MN and Irish BA, are with EHG. We have Danube samples with the Hungarian BA. They have the highest EHG-CHG ratio of any ancients. You should brush up on this stuff.

Krefter said...


Either they were Celtic speakers or the expansion of Celtic languages had little or no affect on the DNA of the British isles.

The old assumption Celts=Hallstatt or Iron age or Gauls is false. Hallstatt and relatives are just the culture *some* Celts belonged to. Hallstatt or its ancestors can't explain Celtic speakers in the British Isles and Iberia.

The oldest Celtic inscriptions are from Italy and Portugal, and both pre-date the Hallstatt culture. The one from Italy belonged to a culture that arrived in Italy before 1000 BC.

There's no valid reason why Celtic languages could not have expanded 4,000-4,300 years ago.

Nomen Cognomen said...


How did 66% non-steppe admixture enter into these samples when 2/3 of them have EHG mtdna, and supposed EHG Y-Dna. When you think about it they have probably 15-18% EHG, but both EHG Y-Dna and MtDNA.

If in Yamna the CHG component entered maternally, how did 66% non-steppe admixture enter into LBA people when 5/6 uni-parental markers are EHG. Wouldn't it make more sense if high non-steppe R1b people mixed with maternal wives of R1a-ers in Central Europe that had steppe mtDna?

Davidski said...

Isn't U5b2a2 more likely to be an WHG mtDNA haplogroup?

Anyway, you can't seriously expect strong correlations between genome-wide and uniparental DNA in a sample of three people with complex ancestry.

Considering that their recent ancestors came from Central Europe, which was a melting pot during the Early Bronze Age, even 100% steppe Y-DNA/mtDNA and 90% Middle Neolithic genome-wide dna would still make sense.

Ryan said...

You're right, looking at the paper now, this is greatly different from what I expected.

"Either they were Celtic speakers or the expansion of Celtic languages had little or no affect on the DNA of the British isles."

I'm assuming the latter. I figured the Urnfield culture was proto Italo-Celtic, and that the Atlantic Bronze Age was non-IE. I may have to re-evalute that though. I'd still like to see what David comes back with when he gets his hands on the sample.

Ryan said...

I'd note that they assume CHG is from Yamnaya. I'd like to see that assumption tested.

Nirjhar007 said...

IMO this finding is somewhat harmonious with the suggestions here, correct me if i'm wrong.
Metallurgy arrived in Ireland with new people, generally known as the Bell Beaker People, from their characteristic pottery, in the shape of an inverted bell.[9] This was quite different from the finely made, round-bottomed pottery of the Neolithic. It is found, for example, at Ross Island, and associated with copper-mining there. It is thought by some scholars to be associated with the first appearance of Indo-Europeans in Europe (possibly Proto-Celtic),[10] though this theory is not universally accepted.

The Bronze Age began once copper was alloyed with tin to produce true Bronze artefacts, and this took place around 2000 BC, when some Ballybeg flat axes and associated metalwork were produced. The period preceding this, in which Lough Ravel and most Ballybeg axes were produced, and which is known as the Copper Age or Chalcolithic, commenced about 2500 BC.

terryt said...

"Surprise, suprise.. all BA males are R1b and steppe derived".

The R1b is hardly a surprise. This paper from 2000 said exactly the same thing. Men on the west coast of Ireland were almost totally R1b:

"Too early to be Celts, and probably (though maybe not necessarily) too early to be some other unattested branch of IE".


"Either they were Celtic speakers or the expansion of Celtic languages had little or no affect on the DNA of the British isles."

That seems likely to me. Basques are heavily R1b and certainly not Celtic, although the men in that case may have adopted the pre-existing language.

Grey said...

"How did 66% non-steppe admixture enter into these samples when 2/3 of them have EHG mtdna, and supposed EHG Y-Dna. When you think about it they have probably 15-18% EHG, but both EHG Y-Dna and MtDNA."

That could be explained by disproportionately male mediated migrations over multiple hops e.g. a combination of already married older males and unmarried younger males picking up local wives after arrival.

(assuming some or all of the U mtdna was WHG)

migration from steppe in 5:3 male:female ratio (50 male, 30 female), move to EEF region, pick up (20) eef wives for excess males, result
100% steppe ydna
80% steppe mtdna
80% steppe autosomal
20% eef autosomal
20% eef mtdna

second migration to WHG region, same disproportionate male:female ratio but this time picking up 20 whg wives for the excess males, result
100% steppe ydna
64% steppe mtdna and adna
16% eef mtdna and adna
20% whg mtdna and adna

just an example but you can see how disproportionate males could skew the results over multiple hops.


Separately given the geography I tend to think if there was a population who started out around the same place then a faster "hare" part could have come by sea via Portugal while a "tortoise" part arrived later via Hallstadt for a bit of distant cousin on distant cousin violence.


"If in Yamna the CHG component entered maternally, how did 66% non-steppe admixture enter into LBA people when 5/6 uni-parental markers are EHG. Wouldn't it make more sense if high non-steppe R1b people mixed with maternal wives of R1a-ers in Central Europe that had steppe mtDna?"

The start region may have changed over time e.g. more CHG input over time, so later waves may have had more CHG than earlier ones.

Krefter said...

@BA Irish were probably Celts.

This is certainlly the direction the evidence is going. You guys have an assumption Celtic languages can't be 4,000 years ago. Why not? They existed 2,000 years ago and today. Why couldn't they have existed 2,000 years before 2,000yo?

Urnfield, Hallstatt, etc. can't explain all Celtic languages. The distribution of Celtic languages match Bell beaker and R1b-P312 very well.

Ireland's Neolithic population essentially disappeared. There was certainly language replacement. Steppe ancestry has already been linked with IE languages. Why couldn't the new language they brought Celtic and ancestral to Gealic?

Did you guys notice French, Germans, Spanish, and Tuscans are the closest genealogically relatives to Ireland_BA after British/Irish? The geneaological relationship is closer to Spanish than to Polish or BeloRussian even though overall genetic makeup is closer to the later. There's a good chance this has to do with Celtic languages.

Just because Italic likely arrived in Italy after 2000 BC, doesn't mean in Central Europe before Italics came to Italy, they must have spoken proto-Italo Celtic instead of Celtic. It's possible Italic and Celtic coexisted for 100s of years in Central Europe.

Grey said...

On the language there's also the p-celtic vs q-celtic thing which may be relevant (or maybe not).

Krefter said...

"There was no "Celts" in 2000 BC
It doesn't take much imaginationor intelligence to figure out why, and how it can be squared with the aDNA evidence"

There's no way to know this.

Grey said...


On the hare and tortoise thing I can imagine three options

1) Danube / Central Europe with the "hare" part being traders/artisans moving ahead of the main wave.

2) Danube / Central Europe as far as Hungary, blocked by LBK for a while so some going through Slovenia to northern Italy then a partial maritime journey to southern France, Iberia and the Atlantic coast.

3) Full maritime odyssey from Black Sea to island hopping through the Med to Portugal to Atlantic coast. I admit I prefer this option cos it would make such a cool movie rather than any other reason.

Rob said...


yes there is ;)
The sound laws which define Celtic came about somewhat later than 2000 BC.

Blasonario Cremonese said...

I don't know why are there some people that suggest Mt-DNA of the samples are only steppe-derived... If you take a look at the European Neolithic aDNA on Ancestral Journeys, you will se that we have some of the mt-DNA of those Irish guys in Neolithic Europe:

- Schöningen Germany Salzmünde [SALZ 12] 4100-3950 BC J2b1a Brandt 2013
- Farmer Germany Blätterhöhle [BLA 27] 3869 ± 59 BC U5b2a2 Bollongino 2013

Obviously, there are other samples with those mt-DNA.

U5b2a2 is also found in Mesolithic Germany

Gioiello said...

How many presuppositions! That the Mesolithic link with Southern Mediterraneans is due to a Middle Eastern origin, that BA Ireland comes from Eastern Europe (Kurganists?) where only old subclades of R1b has been found, why no R1a overwhelmingly in Eastern Europe etc etc. Have you ever known the theory of the "Italian Refugium", the migration 7500 years ago to Iberia, the expansion of Bell Beakers from Iberia, Southern France, Tyrrhenian Italy?

Colin Welling said...


That could be explained by disproportionately male mediated migrations over multiple hops e.g. a combination of already married older males and unmarried younger males picking up local wives after arrival.

What are you even talking about? What disproportionately male mediated migration are you thinking of? The neolithic migration? nope! The neolithic irish are the same as their neolithic forebearers who were located hundreds of miles to the southeast. How about the bronze age irish then? Again, nope! The bronze age irish are nearly identical to their bronze age forebearers from around Germany. The paper even adds that the neolithic mixture in bronze age irish probably came from germany, not ireland.

There it is. Two of the most important migrations in europe, causing huge genetic turnover, were actually forged in very particular regions (southeastern europe for EEF and central europe for BA ire/ger/sca) before they spread throughout europe barely mixing with the locals in many instances. In regards to human behavior and the migration history of europe, how have all the recent adna studies not hit you like a brick wall. You can't parrot the same armchair archeology that relies on rich men taking out the local men and marrying the women. Its just stupid at this point, and frankly no better than the people who are holding out on an iberian refuge for the r1b-m269 lineage.

Its very simple. Many of these neolithic and bronze age migrations were "complete". The men did not mix with the locals and the women came with the men. If it were no so, then the irish neolithics would be heavily hg... but they aren't.

Colin Welling said...


How did 66% non-steppe admixture enter into these samples when 2/3 of them have EHG mtdna, and supposed EHG Y-Dna. When you think about it they have probably 15-18% EHG, but both EHG Y-Dna and MtDNA.

Thats a very illustrative question and I'm glad you asked.

Founder effects and drift, along with natural selection, don't just act through gender. They happen regardless of gender. When people see a big difference between uniparental dna (usually the ydna) and autosomal dna people usually think it is due to a different in the relative sizes of the mating pools for men and women. You noticed that such an explanation wouldn't work because EEF uniparental dna is under represented on the male and female lineages. But assuming these few samples are actually representative, enormous if, simple drift can still cause things like this to happen.

Colin Welling said...

Hey! Did anybody notice the map which shows that the neolithic irish were least related to modern russians and people of the levant???

I don't think it meshes well with the idea that most of EEF came from ancient levant. If it were post neolithic SSA admixture in levant that caused such separation, then why are sicilians so related to the neolithic irish. If its the lack of WHG in levant that causes the distance of neo irish and modern levant, then why is russia so genetically distant from the neo irish. If the yamnaya in the modern russians are a big penalty for relatedness to neo irisht then shouldn't the turkish be very distant from the neo irish?

I don't think it adds up. does anybody know of a way to use the neo irish to test the hypothesis that EEF largely came from the levant?

Shaikorth said...

Colin, that's a haplotype donation analysis rather than one of overall similarity. Egyptians have more Neolithic Irish haplotypes than modern Welsh or Orcadians, but we know for sure they are not as similar to Middle Neolithics overall.

Re: Levant, Druze have extremely low haplotype sharing and that's likely because extreme endogamy is prone to destroying ancient haplotypes.

Gioiello said...

Here is the mt hg of this Old Irish woman:

17. AY738944(Italy) Achilli HV0-T195C 13-APR-2007 T72C T195C A263G 309.1C 309.2C 315.1C 523.1C 523.2A A750G A1438G A2706G A4769G T5093C C6059T C7028T G7762A A8860G G11719A A13933G A15326G T16298C
18. EF660968(Italy) Gasparre HV0-T195C 04-JUL-2007 T195C A263G 315.1C A357C A750G A1438G A2706G A4769G G6962A C7028T A8706G A8860G A15326G T16298C
19. HM765457 Zaragoza HV0-T195C 01-AUG-2010 T72C T195C A263G 315.1C A750G A1438G A2706G A4769G T6776C C7028T A8014G A8860G C15324T A15326G T16298C
20. JN794568(English) FTDNA HV0-T195C 12-OCT-2011 T72C T195C A263G 309.1C 315.1C A750G A1438G A2706G A4769G C7028T T8736C A8860G G15110A A15326G T16298C
21. KT808880 FTDNA HV0-T195C 30-SEP-2015 T72C T195C A263G 309.1C 315.1C A750G A1438G A2706G A4769G C7028T A8860G A14693G A15326G G16129A T16298C
22. KP340137(Italy) De Fanti HV0-T195C 09-DEC-2015 T72C T195C A263G A750G A1438G A2706G A4769G T6392C C7028T A8860G G13145A A15326G T16298C T16311C
23. KP340139(Italy) De Fanti HV0-T195C 09-DEC-2015 T72C C150T T195C A263G A750G A1438G A2706G T4688C A4769G C7028T T7684C A7930G A8860G C13503T A15326G C16169T T16298C
24. KP340140(Italy) De Fanti HV0-T195C 09-DEC-2015 T72C A93G T195C A263G T310C C518T A750G A1438G A2706G A4769G C7028T A8860G G13804A A14582G A15326G T16298C
25. KP340142(Italy) De Fanti HV0-T195C 09-DEC-2015 T72C T195C A263G A750G A1438G C1706T A2706G A4769G C7028T A8860G A15326G C16111T T16298C T16519C

Sonic Reducer said...

For what it's worth, the Amesbury Archer was born in the Alpine Region of Central Europe

Colin Welling said...

oh, thanks for clearing that up shaik

Alberto said...

Very interesting paper, even if it's nothing too unexpected. I'm a bit surprised, though, that being an Irish centered paper they didn't even attempt to tackle the Celtic question.

Interesting this ChromoPainter tool for some fine grained haplotype donation analysis. I think it will be become very useful when we start to get dozen of high coverage samples per study that will allow to determine quite accurately the population sources for each culture. It's looks very interesting how it can discriminate the sharing between different Neolithic samples with modern populations (Stuttgart and NE1 with Italy and Greece, Ballynahatty with Iberia and Sardinia), or with Loschbour (NW Europe to SE Europe cline, higher in Spanish than in Lithuanian or Polish), Rathlin1 (clearly high in Scotland, Ireland and Wales), etc...

And once again one very important phenomenon for understanding the genetic make up of modern Europeans that gets systematically ignored in all the papers so far: The PCA (regardless of the projection bias) shows the Neolithic Hunter-gatherers from Sweden as admixed with Neolithic farmers. And more important, the confirmation of Motala-like admixture in Esperstedt_MN (already suggested in many Haak et al. stats, but never directly as here):

Mbuti Esperstedt_MN : Western_HG Scandinavian_HG 0.0048 1.221

Reading between the lines it seems inevitable that other populations like Unetice and Corded Ware also have this type of admixture. About time they start paying attention to this.

Paul Ó Duḃṫaiġ said...

To be honest I would think 2000BC is probably too early for differenation of Proto-Celtic (deletion of Proto-IE /P/ etc.) don't think we'll ever know to be honest as the earliest dateable inscriptions in a Celtic language are from the 6th century BC.

However it's quite possible that during the bronze age you had a continuum of dialects of western Proto-IE, within which eventually both Celtic and Italic families arose.

So a potential scenario would be developments in one of these dialects gained prestige status and over time spread within the dialect chain due to status etc. (lets say via trading networks etc.)

The dating of the three Rathlin males is put down as:

So you are talking about remains from 250-700 years before "Atlantic Bronze age" is generally regarded as having started. So it's quite possible we are looking at undifferenated Proto-IE speakers (if at all)

All three are R1b-L21, one is confirmed R1b-DF21, second had 1x read for DF21 but they decided to keep him as R-DF13 (90% of all L21 is also DF13). DF21 obviously is still present in large amounts in Ireland today. The first modern Irish genome sequenced happened to belong to a R1b-DF21+ male (one of Hinxton remains in England was also DF21)

Rob said...



Matt said...

Speaking of the haplo stuff, quick PCA based on median haplotype donation stats from the paper, with extra population labeling:

A set of plots of stats:

(Colour scheme same as paper's supplement)

@Alberto, the Esperstedt stats were noticeable to me as well, and highlighted by the paper. I guess some evidence of localised HG ancestry in different MN? I think it'll be interesting to look at Polish and Belarusian MN farmers when samples crop up to see what is happening with them. I don't think these differences in HG ancestry to MN will affect populations too much (as MN populations don't seem to have too much regional continuity to present day folk themselves), but interesting.

Rob said...


Your comments are interesting in light what FrankN has been highlighting in northern Europe, and the need for samples from east-central Europe.

Grey said...

@Colin Welling

"What are you even talking about?"

I'm talking about how multiple disproportionately male migration events can lead to a population having ydna frequencies out of sync with their adna frequencies.


"You can't parrot the same armchair archeology that relies on rich men taking out the local men and marrying the women."

Straw man.

A migration event that only involves a small group from a larger population would likely involve a mixture of families and extra single men imo.


"If it were no so, then the irish neolithics would be heavily hg... but they aren't."

From the paper

"Along with other MNs, Ballynahatty displays increased levels of hunter–gatherer introgression compared with earlier farming populations."

"estimate a hunter–gatherer component of 42 ± 2% within a predominantly early farmer genome"


"How about the bronze age irish then? Again, nope! The bronze age irish are nearly identical to their bronze age forebearers from around Germany. The paper even adds that the neolithic mixture in bronze age irish probably came from germany, not ireland."

Isn't that circular?

Another explanation would be both the German and Irish result were the product of a similar type of event i.e. a disproportionately male mixture with a similar local neolithic population.

The proof then might lie in whether there are small differences between the two.

Matt said...

Btw, Alberto, another thing about that Esperstedt stat you mention is that at the same time:

D (Mbuti Esperstedt_MN Scandinavian_HG Eastern_HG): -0.0303 -7.224
D (Mbuti Esperstedt_MN Western_HG Eastern_HG): -0.0239 -4.949

Since Esper has a higher stat against EHG with SHG than WHG, that would also argue against the idea that Esper is showing a similarity to SHG only through an "early" entry of EHG rather than SHG themselves.

(Also compare -
D(Mbuti Ballynahatty Irish_Bronze Yamnaya) D: -0.0276, Z: -8.937
D(Mbuti Gok2 Irish_Bronze Yamnaya) D: -0.0268, Z: -7.293
to D (Mbuti Esperstedt_MN Irish_Bronze Yamnaya) -0.0123 -3.485

still significant, but to a reduced degree.)

Grey said...

From the paper

"It has been suggested that the high frequencies of both C282Y and H63D are due to heterozygote advantage related to nutritional advantage in Neolithic iron-poor diets (52), mitigation of celiac disease (53), and increased resistance to parasitic infection (54)."

Would another possibility be this gene was selected for alongside LP because mammal milk always lacks iron?

If correct the prevalence would correlate with LP.

Cossue said... (?)

Krefter said...


The experts suspect West Celtic languages came to Ireland in 2000 BC.

From everything we've learned over the last few years, Cultural/Linguistic change=change in genetics. This is why I think they were Celtic.

I emailed one of the narrators of a recent BBC documentary on Celts. In his docs he suggests Celts of the British Isles are descended of the Neolithic/Mesolithic people of the British Isles and had no genetic connections to other Celts.

I explained Isles Celts are essentially immigrants from Late Neolithic Central Europe. And that Celtic languages likely arrived with large movements of people in 2000 BC. I think starting today people will stop the BS, Palaeilthic refugium, Celtic=no genes stuff.

Nirjhar007 said...

all BA males are R1b and steppe derived.
But i can say why steppe derived R1b? The three individuals from the Bronze Age Irish site are R1b1a2a1a2c-L21, which is typical of British isles and absent in Eastern Europe, not to say of the steppes, but quite strong in Basque country:

I think this is related with Bell Beakers, they have some green/teal component as other Bell Beakers but they do not come from the steppe, rather from Iberia...
Maybe it is not a case that Ireland was called Hibernia. I also doubt that these people were IE speakers, rather Vasconic. See here some connections of Basque and Irish:

Colin Welling said...


I'm talking about how multiple disproportionately male migration events can lead to a population having ydna frequencies out of sync with their adna frequencies.

That could not have happened as I already explained.

"Along with other MNs, Ballynahatty displays increased levels of hunter–gatherer introgression compared with earlier farming populations."

"estimate a hunter–gatherer component of 42 ± 2% within a predominantly early farmer genome"

Admixture happened but these studies are telling us that, by and large, the admixture events only happened in a few places.

EEF was well established in southeast europe in the early neolithic. EEF then traveled to Spain and Germany with almost no further admixture. That means the overwhelming majority of men were creating offspring with other farmer women and not hunter gatherer women.

Even the difference between the early neollithic europeans and middle neolithic europeans is incredibly small, meaning that there was still an overwhelming bias for farmer men to mate with farmer women.

if you go deeper into the matter and realize that mixed individuals tend to go farmer, almost by default, then the implication is that the average absorption rate of hunter gatherer into the farmer wave was less than once over the span of thousands of miles and thousands of years. That is how unlikely it was for a farmer male to mate with a hunter gatherer woman rather than a farmer woman. let that sink in...

Another explanation would be both the German and Irish result were the product of a similar type of event i.e. a disproportionately male mixture with a similar local neolithic population.

this... doesn't even make sense. Are you suggesting that steppe herders actually made it to ireland rather than first having mixed with central europeans? That didn't happen. Nor would it even support your idea.

The fact is that the bronze age irish population was formed in central europe and barely did any mixing on its way to ireland. Once again, the men were by and large sticking to their own and not mating with the local women.

Karl_K said...


"I think this is related with Bell Beakers, they have some green/teal component as other Bell Beakers but they do not come from the steppe, rather from Iberia..."

Perhaps they came more immediately from Iberia, but these Bronze Age R1b migrations still originated in the steppe before migrating to Iberia. I would bet on them being IE, with Basque being unusual by keeping the original language of their region.

Gioiello said...


I think you are right. Please, say Krefter that I didn't speak of "Palaeolitic Refugium" but of an "Italian Refugium" and of a migration from Italy to Iberia 7500 years ago and from there to Ireland (see the mt hg HV0-196C clearly Italian, and I am waiting that they find also R-L51*) and after of an expansion of Bell Beakers from Iberia, Southern France and Tyrrhenian Italy, and it seems to me that all that is confirmed.

Alberto said...


Thanks for the graphs. It's a pity that they didn't provide these same haplotype sharing stats for Kotias and Karelia_HG to see how well a PCA would then correlate with the ones we're used to see. If the tool proves to be accurate it looks quite promising for getting all the fine details that now we have to make assumptions about.

About the SHG admixture, yes, I think it's real. Similar stats about Unetice were done here back then. So I do think that the phenomenon is going to be relevant - difficult to guess how much exactly, but when in this case with the Rathlin samples they estimate the MN ancestry at about 2/3 (using Gok2 as the best match), I think that factoring in the SHG ancestry in other MN groups from Eastern Europe might make a significant difference.

BTW, Lezgins seem to have quite an elevated haplotype sharing with Rathlin. Being high in R1b and maybe related to some BA Caucasus populations (Maykop? Kura-Araxes?... or maybe not, we'll have to see) it looks intriguing. In any case it will be interesting to see those stats with other ancient genomes. I think Davidski mentioned that he'd try to take a look at the program when he has some time.

Matt said...

@Alberto: It's a pity that they didn't provide these same haplotype sharing stats for Kotias and Karelia_HG to see how well a PCA would then correlate with the ones we're used to see. If the tool proves to be accurate it looks quite promising for getting all the fine details that now we have to make assumptions about.

And Yamnaya too, of course. Unfortunately, I'm not sure if the coverage for Kotias / Karelia_HG is good enough for that (although Kotias may be and they just didn't have it).

Haplotypes add some power, although there is some vulnerability to drift breaking down / replacing haplotypes that might play a role in enhancing some of the differences here (e.g. Orcadians are differently positioned to the rest of Northwest Europeans. Or the Slavic groups possibly may have some effect from being a mix of a recent expansion of a small essentially EBA group proto-Slavic with other EBA groups who were more widely distributed across Eastern Europe?).

Yeah, the Lezgins are outliers in their haplotype sharing with both Rathlin1 and BR2. I'd take this to indicate that they've actually got steppe related admixture from the Yamnaya at a fair clip, while this is less prominent in other Caucasians and West Asians who are still fairly close to them in unlinked analyses (the ancient composition is similar, but other populations actually less affected by Yamnaya related populations).
Maybe vice versa, I suppose (steppe ancestry populations have some Lezgin related ancestry).

Haplotypes could add some power to talking about the South Asian story (more haplotypes with Yamnaya than expected based on CHG and Yamnaya sharing on IBS / f3 / D, probably more actual steppe linked ancestry, and vice versa).

( and for versions of the same haplotype PCA which look at the relative haplotype sharing compared to the median or average. Look more like the unlinked West Eurasia autosomal PCA.)

Grey said...

@Colin Welling

"That could not have happened as I already explained."

Maybe it didn't. The point is people keep saying that ydna becoming detached from adna to such an extent is *impossible* when it obviously isn't. All you need is a reason why a migrating group would be mostly male - like mining.


"EEF was well established in southeast europe in the early neolithic. EEF then traveled to Spain and Germany with almost no further admixture."

I'm not disputing that. However at some later point that must have changed because there was a WHG resurgence for some reason.

Again, from the paper

"estimate a hunter–gatherer component of 42 ± 2% within a predominantly early farmer genome"


"Are you suggesting that steppe herders actually made it to ireland rather than first having mixed with central europeans? That didn't happen. Nor would it even support your idea."

Nope, because it wouldn't support my idea - which is simply if you have multiple male majority migrations in a sequence then the end result autosomally could be almost anything because it would depend on the sequence.

Or maybe it was an express trip from Germany.

Dospaises said...


You aren't doing a very good job of playing connect the dots.

Alberto said...


Yes, from ancient to modern the accuracy will always decrease and be subject to some biasing factors. I was thinking more about being able to use this with high coverage ancient samples, now that they have refined the methodology for extracting higher coverage from them (as long as the DNA is actually there and not definitely gone).

For example to be able to get a better idea about the genesis of CWC. If in addition to Dstats, admixture, etc... we could use this method to get haplotype sharing between Yamnaya/CW, Karelia_HG/CWC, Kotias/CWC, Motala/CWC, Espersted_MN/CWC, etc... it could give us quite reliable information about which of the possible models actually seems to be the real one. But we'll see how possible is it to get enough samples with enough coverage to do this (AFAIR Karelia_HG and Kotias do have high coverage, as does Esperstedt_MN. But I think no Yamnaya or CW sample has similar coverage to this).

About the Lezgins, yes, I thought that Yamnaya admixture could be the explanation. But seeing how they share more haplotypes with Rathlin1 than Lithuanian, Polish, Russian or Finnish do, it does not seem like specifically Yamnaya related (unless these other European populations indeed have low sharing with Yamnaya because CW didn't actually receive direct Yamnaya input?). But anyway, I don't want to read too much into something that might just be anecdotal.

Kristiina said...

Only one fourth of ancient Irish mtDNA is steppe-related.
HV0 (Ballynahatty) has been found in the following contexts: Starčevo Neolithic Vinkovci Nama Croatia HV0, Rossen Neolithic Oberweiderstedt Germany HV0, Lengyel NE Brześć Kujawski Poland HV0, Rossen Neolithic Wittmar Germany HV0, Middle Neolithic Bom Santo Portugal HV0, Late Neolithic Calden Germany HV0, Middle Neolithic Odagsen Germany HV0, Middle Neolithic Treilles Aveyron France HV0 x2, Corded Ware Oberwiederstedt 2 Germany HV0e.

According to the recent HV paper (De Fanti et al 2015), HV0a coalesces at around 12.5 kya in the BEAST tree and at 11 kya in the mtPhyl one. HV0 and HV-16311 samples were mostly coming from North/Western Europe and Northern Italy. Haplogroup V which is particularly frequent in Fennoscandia is nested within the HV0 clade. In De Fanti et al 2015, it is proposed that several European HV lineages arose in Italy: ”The resulting hypothesis of a glacial refugium in Southern Italy has implications for the understanding of late Paleolithic population movements and is discussed within the archaeological cultural shifts occurred over the entire continent.” ”Fig 2 localizes the distribution of the major HV*(xH,V) lineages in Italy, with 36 sample sites distributed all over the peninsula and major islands and more than 1,000 individuals screened (S3 Table). Major trends are distinguishable: HV0 appeared to have higher frequencies in Northern Italy, with the exception of Enna, in Sicily. Conversely, HV4, HV* and HV-73 were more frequent in Southern Italy.

Among the three Bronze Age haplogroups, U5b2a2 (Rathlin2) is originally clearly Mesolithic European haplogroup and has been found in the following contexts: Blätterhöhle [BLA 6] U5b2a2 (8796 BC!), Schöningen Neolithic Salzmünde Germany U5b2a2c, Baalberge Middle Neolithic Quedlinburg VII Germany U5b2a2, Farmer Middle Neolithic Blätterhöhle Germany U5b2a2 x2, Fisher-gatherer Middle Neolithic Blätterhöhle Germany U5b2a2, Farmer Middle Neolithic Blätterhöhle Germany U5b2a5.

J2b1a (Rathlin3) is a Neolithic European haplogroup which looks like having spread from Europe to the Steppe during the Bronze Age. It has been found in the following contexts and: (RRBP NE Gurgy Les Noisats France J2 (16193T, 16362C)), Schöningen NE Salzmünde Germany J2b1a, Regional TRB Salzmünde Germany J2b1a, Funnel Beaker/TRB Frälsegården Gokhem Sweden J2b1a, El Mirador Chalcolitic Spain J2b1a3, Corded Ware Spreitenbach-Moosweg Switzerland J2b1a, Unetice BA Plotzkau 3 Germany J2b1a, Sintashta Stepnoe 7 Russia J2b1a2a, Srubnaya Samara Barinovka I Russia J2b1a2a, Mezhovskaya Kapova Cave Russia J2b1a

Only U5a1b1 looks like a typical Eastern European (in particular Slavic) haplotype which has spread from there to Western Europe, Fennoscandia, Siberia and Pakistan/India. It has been found in the following contexts: Corded Ware Karsdorf Germany U5a1b1, Bell Beaker Benzingerode Heimburg Germany U5a1b1, Unetice BA Przeclawice Poland U5a1b1.

FrankN said...

Let's start with the Neolithic - the Celtic stuff will become endless:

Generally not much of a surprise, except for the somehow elevated WHG share. Loschbaur as closest WHG relative, Gokhem the closest MN - no surprise to me. Next things I would like to see is WHG ancestry from Brittany or surrounding, and a chunk of Michelsberg Culture aDNA sampled from the Paris Basin, the Rhine and somewhere in Lower Saxony. My guess here: The more western the Michelsberg sample, the more closely it matches the Irish one. That the Michelsberg colonists didn't come overland to NW Sweden has anyway been clear to everybody looking a bit deeper into early 4thM post-Ertebolle archeology. Some contemporary Scottish and Orkney aDNA would of course help to understand better their route.

To get the Michelsberg story completed, La Hoguette Group (Middle Rhone - West Alpine hunter-pastoralists) aDNA would also be nice. They seem to have been neolithicised from the Mediterranean, which could explain the Med appeal in Ballynahatty's EEF component.

Rob said...

@ Krefter

You're not getting it
Just read what Paul O Dubtaig wrote
It's a simple distinction between when a mass migration occured and when *proto-Celtic propper* can be dated to.
Ie anything in 2000 BC was still pre-Celtic

Krefter said...


All Celtic languages went extinct, except for in England/Wales and Ireland(Gaelic in Scotland is from Ireland). That makes me doubt there is enough linguistic data to understand the linguistic evolution of Celtic languages.

I trust genetic trends more than age estimates of a language family that barely exists anymore. Everything from farming to Indo European languages involved movements of people. Everyone so far who sides with language change but no genetic change has turned out wrong. I think Celtic languages with no genetic change will turn out to be wrong to.

Annie Mouse said...

Well what I am seeing is:

(1) A neolithic lady (5-5.5kya) who's mitochondrial haplotype (group?) tracks with central Spain with a genetic affinity that lies between Spain and modern Ireland (pointy bit of pale green). There is no way to know what her Daddies haplotype/haplogroup was. She was a contemporary with the Yamna culture but has no apparent Yamna-esque influence. Shame she was not a man.

(2) A bunch of Bronze age Irish guys with classic well developed Irish Y haplogroups but a genetic affinity sitting over Eastern Europe. Clearly Yamna-esque influenced. These guys are a long, long way away genetically from the modern autosomal Irish.

Ireland is tricky as it is long known to have gone through multiple cycles of population bust and boom. Rathlin 1 (oldest) is a circa 3,000 year old Bronze aged guy with a ~3000 year old haplogroup (R1b1a2a1a2c1g, S192/DF21 (a subclade of L21). So he might be one of the actual guys who developed the haplogroup in the British Isles. Very nice timing. Rathlin 2 and 3 are also L21, a haplogroup that looks like it might have radiated out of Ireland. The parent haplogroup P312/S116 aka R1b1a2a1a2 looks like it came out of Spain (S116* map).

So I think Irish L21 subclades came out of Spain. This could have been a Yamnaya-esque group taking a circuitous route, or something earlier. Not sure yet. Shame The Ballyhatty lady was not a man.

terryt said...

"Just read what Paul O Dubtaig wrote"

I agree. A good comment. I thought this was particularly relevant:

"However it's quite possible that during the bronze age you had a continuum of dialects of western Proto-IE, within which eventually both Celtic and Italic families arose".

Rob said...

Frank N

(From the other thread)

"In any case, when taking Czechs as Central European proxy, sizewise some 90% of their CHG was already present with Bell Beakers and did clearly not arrive via the Danube"

"Clearly"? Where did it come from then ?

Although we lack aDNA from the Balkans, I think it is already becoming apparent that there was a "bi-wave" movement of CHG into Europe. Wherever the "CHG"(s) (NB: possible plural) came from - one came directly via the Balkans c. 4000 BC (EHG -deficient), and the other via the steppe (with EHG). The meeting point was Hungary (carpathian basin).

So, barring independent spread via the north European plain, I'd say most of the CHG in central Europe came from the Danube, spreading to the Elbe, which appears to have been in turn a tertiary staging point for western Europe.

* "Many people think of Creoles only as outcomes of contact between Colonial Europeans and "Natives", but there are, e.g., purely African-based Creoles such as Sango. Analysis of Austronesian, which displays a quite homogenised grammar/morphology, but substantially differing vocabularies among the sub-families, is a kind of forerunner in applying the concept of Creoles to historic linguistics.
The link below demonstrates typological features (grammar, morphology) changing at similar rates as lexicons, but unconnected to them. This renders the standard equation "language family = shared lexicon & shared typology" questionable. Note the typological neighborhood network (Fig.1) showing, a.o.,
- French branching from German,
- Finnish (but not Hungarian!) sharing a root with the Western IE cluster,
- a Caucasian-Burushaki-Hindi-Kannada(Drav.)-Turkic cluster, from the root of which also spring Basque (Cauc. side) and Quechua (Turkic side)
- a Japanese-Korean-Burmese branch (with the root shared by KhoeKhoe),
- neighbouring Afro-Asiatic and Eastern Niger-Congo clusters, but Western NC (Yoruba, Sango) located on a distinct branch that is related to a Polynesian-Indonesian- Thai-Vietnamese cluster.
- The Greek-Albanian-Bulgarian-Romanian branch described in the Supp.Mat. is also worth noting."

My understanding is that IE cannot be a creole because creoles lack inflexional morhpology, even if they become 'complex' by becoming 'nativitized'.

The Greek-Albanian-ROmanian- South Slavic branch you're talking about is a Sprachbund, not a linguistic node, so its not really supportive of a creaolization hypothesis, unless Ive misunderstood.

mooreisbetter said...

This paper completely destroys certain folks' cherished notions of sexual selection or royal descent for R1bs, and provides the far more prosaic answer that you had a large founder population that displaced earlier populations on an island that few future conquerors bothered with.

There's a long explanation of the findings here:

Davidski said...

As far as I know, no serious commentators here ever argued that R1b rose in frequency as a result of sexual selection.

And it's pretty hard to ignore the idea that the vast majority of Eurasian R1b descends from "royal" lines, since R1b has now been found in a number of "royal" Kurgans, and the Kurgan PIE hypothesis is looking better than ever at this stage.

So what's the problem?

Aram said...

Quick and dirty visual comparison of Bell_Beacker from Czechia with modern Czechs.

Anatolian Neolithic level remains more or less the same. Maybe slight increase and homogeneization.
CHG increase from BB to modern Czech. Not much but I think measurable.
EHG decrease.
WHG also seems to decrease but in England or Germany the case could be different.
And surprise. There is a clear apparition of SWA in modern Czech 2-4% that was absent in BB.

So this SWA expansion seems to affect not only the North Near Eastern pops but also Europe.

It would be interesting to make the same comparison for Irish people.

Ryan said...

Is R1b actually over represented among European royals? And is there even a robust way to measure this, or is it simply a case of "surprise surprise - royal families with roots in western Europe tend to belong to the most common haplogroup in western Europe"?

Kristiina said...

Although mtDNA haplogroups are subject to drift – as all uniparental markers – they often correlate surprisingly well with the autosomal makeup of an ethnic group.

The recent paper estimates that following the methods described in Haak et al., which use a collection of outgroup populations, the mixture proportions of three different sources in the total Irish Bronze Age group are the following: Linearbandkeramik (Early Neolithic) 35±6%, Loschbour (WHG) 26±12%, and Yamnaya 39±8%.

In spite of a very small number of samples (only three), mtDNA haplogroups of the Bronze Age individuals reflect 33% WHG input, 33% Neolithic input and 33% Yamnaya input.

However, I agree with Rob that CHG probably came to Europe via several routes and is not only connected with IE's.

Colin Welling said...


The point is people keep saying that ydna becoming detached from adna to such an extent is *impossible* when it obviously isn't.

who said that?

The point is that it didn't happen when the farming wave spanned across europe over the course of thousands of years. That implies an incredibly strong bias for men to migrate with and mate with their own farmer kind.

However at some later point that must have changed because there was a WHG resurgence for some reason.

Yes, but over thousands of years, and many matings, the farmer populations didn't have much more WHG than the earlier farmers. Because hg people tend to be 100% WHG, we can conclude that farming men and women very rarely mated with hg.

Farming people, on average, took in much less than one hg per person, in the sum of thousands of years. There are other dynamics which can complicate the picture but the fact remains that MN farmers are much much closer to early farmers than they are to neighboring hunter gatherers.

The steppe related "migration" from ukraine to ireland is also an impressive example of a movement predominantly made up of men and women, whereby the men were far more likely to mate with other steppe related women, rather than middle neolithic women. If there was a significant gender bias of men marrying local women, the migration would have become very diluted very fast. When I say very diluted, I mean a lot more diluted than 33%. For perspective, it only takes 5 generations of mating with locals to become diluted to 3%.

In the big picture, what we are seeing is an astonishing amount of homogeneity within a given culture (bronze age triangle of germany, sweden, and ireland). Rather than seeing a strong genetic gradient within a given culture, which is what male dominated migrations would predict, we are seeing a strong genetic gradient between cultures. In other words, men and women stuck to their ethnic groups.

Colin Welling said...


Do you think you can tease out any information on whether EEF really is compatible with an origin in the levant.

Also, do you think you can make a map of (modern) european ancestry that derives from SHG?

Colin Welling said...

lastly, does anybody think they can use the new methods in this paper to distinguish with central european populations the bronze age irish descend from, i.e. was it corded ware or bell beaker or some variation?

Davidski said...

I'm putting together a haplotype analysis with around 10 relatively high coverage genomes and 2,000 present-day samples to look at the IBD relationship between Yamnaya, Corded Ware, Kotias, Georgians and others.

Alberto said...

@Colin Welling

Unfortunately we still lack the data to have enough resolution to answer those questions accurately. And even with the available samples, I lack the tools to extract any further information. Maybe if someone with the samples gives a try to ChromoPainter to test a few other samples that have enough coverage we could get some more details.

But some comments:

"Do you think you can tease out any information on whether EEF really is compatible with an origin in the levant."

Why, do you have any reason to think otherwise? I think there's ample agreement about the Near Eastern origin of EEF. We have samples from Anatolia, and populations from the Levant share a lot of this ancestry with them (from the new haplotype sharing data also Syrians, Jordanians and Egyptians do share a lot of haplotypes with NE1 and Stuttgart).

Also, do you think you can make a map of (modern) European ancestry that derives from SHG?

Not really. But I can raise some questions. For example, the Unetice samples from Haak et al. were best modeled as Motala + Spain_MN, better than using any EHG or Yamnaya. The score was not excellent, but that could be because they were lacking CHG samples, which Unetice clearly has ancestry from. So let's suppose that indeed Unetice can get a very good model with Motala + Spain_MN + CHG. This would mean that a pretty modern North European population (genetically speaking) would be consistent with having 0% steppe ancestry. And belonging to I2a haplogroup there would be no questions about that either. Now, I'm not saying this model is necessarily correct (we've not even tested it, and even if it works I'm a bit sceptic about qpAdm models), but it does mean that we might not be taking into account the different possibilities that exist. Lithuanians can be modeled as 50% Yamnaya, but it's likely that they can also be modeled as 0% Yamnaya (or anything in between).

It won't be long before we start to get a more complete and accurate picture.

Alberto said...

@Colin Welling

Regarding your comments about the intermarriage patterns, it's interesting to note that they are consistent with the data you're pointing at, but they are inconsistent with a steppe origin of the Bell Beakers and Irish BA genomes.

I mean, we supposedly have an R1b population in the steppe that starts to take so many CHG wives that they lose half of their ancestry. (And what were they doing with their daughters, killing or marginalizing them to force their sons to take foreign wives?). And then at some point this mixed population starts to migrate from somewhere around Ukraine to Germany, and on their way they picked up so many European "wives" (and killed the European G2a guys? And again, what happened with their daughters?) that they lost 2/3 of their ancestry (at this point they only had 1/6th left of their original steppe ancestry). However, once they reached Germany they completely changed this behaviour, and from Germany to Ireland they didn't take any foreign wife but instead kept their daughters so they could intermarriage among the group, and apparently killed (or got rid of somehow) both foreign men and women they encountered on their way.

I think we might be missing something here.

Grey said...

@Colin Welling

If you look at the colonization of the Americas there are three models (simplified)

1) North America
- family based colonization, follows your pattern

2) Frontier Zone - exception to the above on the frontier where you have a majority male zone of trappers, traders, miners and pioneers many of whom take local wives.

3) South America
- mostly male colonization, local wives

As far as the neolithic farmer expansion is concerned I'd say the current data suggests it was mostly (1) with exceptions in frontier zones like maybe Ireland.

An analogy might be if farming hadn't been viable in Arizona at the time so the main colonization wave stopped at the state border with the Arizonan population the descendants of gold miners and Apache women.

Matt said...

Davidski: I'm putting together a haplotype analysis with around 10 relatively high coverage genomes and 2,000 present-day samples to look at the IBD relationship between Yamnaya, Corded Ware, Kotias, Georgians and others.

Cool. Question: Out of SHG (Motala), EHG (Karelia) and remaining WHG(KO1 and La Brana), which if any are high enough coverage?

I'm thinking it might be interesting to look at the median haplotype donation between SHG, EHG and the various WHG with modern populations around the Baltic Sea / Eastern Europe. SHG is indistinguishable from a mix of EHG and WHG in outgroup tests, and not very distinguishable in unlinked shared drift (as measured via IBS, f3, d stats), so this might tell us something about whether SHG actually made contributions that last to today. (Back when Skoglund sampled the first Neolithic era SHG in 2012, there was a peak in allele sharing with Poland - [note Bel here is Belgium, not Belarus]).

Same with KO1 vs La Brana vs Loschbour - those could tell us to some degree whether there is any regional continuity of WHG ancestry at all (although likely not quantify this).

With the haplotype sharing in this paper, I think it's potentially interesting that the haplotype donation with the European Neolithic samples (Ballynahatty, NE1 and Stuttgart) is actually pretty high for Egyptian, Syrian and Jordanian in contrast to Druze, Palestinian and Bedouin. That doesn't seem like it has too much to do with levels of African admixture (and theoretically haplotypes should be less sensitive to recent admixture anyway, since long haplotypes are not especially likely to be broken up by anciently divergent ancestry compared to recently divergent), so must be due to some other factor (endogamy or something else)?

Simon_W said...

My main conclusion from the paper: These data show that the massive influx of LNBA people into the British Isles predated the cultural changes of the Middle Bronze Age. I had anticipated this, but it hadn't been quite certain, previously. Most likely the biggest change in Britain already occured with Bell Beaker migrations from the continent.

In contrast, ATP9 from El Portalon postdates the older two of these Irish Bronze Age samples by several centuries and shows much less "LNBA admixture", less than the modern Spanish and Portuguese. (It might be argued it's because she's from near the Basque country, but presumably the main thrust of IE invaders occured later in Iberia.)

It's very unfortunate that Ballynahatty, the only Neolithic sample, is female and has no yDNA. So I fear this won't convince the "R1b from the west" crowd. They will still maintain that the R1b expansion had nothing to do with the arrival of steppe admixture. ALthough we yet have to find a prehistoric R1b-M269 carrier without steppe admixture. Up to now they all had substantial amounts of it.

Regarding the question of Celticity, I'm not sure that the defining sound shifts can be that firmly and exactly dated, but it's noteworthy that Chang et al. dated the split between Latin and Insular Celtic to shortly after 2000 BC, and the split between the former two and Germanic to roughly 2200 BC, so according to these estimates the Rathlin samples may have spoken a very archaic form of Celtic.

As for the question if R1b spread with Vasconic or Celtic, it should be noted that both Basques and Insular Celts have a very high incidence of R1b, but the former are autosomally more like Middle Neolithic farmers than almost all other modern European peoples, while the latter carry substantial steppe admixture. So the answer depends on the origin of west European R1b, i.e. if it came together with the steppe admixture or if it's a relic of Middle Neolithic farmers. And I think the former hypothesis is much more likely.

Krefter said...

"As for the question if R1b spread with Vasconic or Celtic"

Vasconic shouldn't even be considered. There were other non-IE languages in Iberia and Italy, I don't see what makes Vasconic special. We don't know enough about pre-Latin languages of West Europe to know whether Vasconic is likely or not. If not Celtic, we have to admit we have no idea.

Simon_W said...

@ Krefter

I tend to agree that it's at least dubious that there existed a widespread Vasconic substrate all over western and central Europe. But if you look at the modern R1b frequencies, it peaks almost only in the westernmost Celts/Romanized Celts and the Basques. Only in Catalonia, where Iberian had been spoken prior to Latin, R1b is similarly common. And at least Iberian has quite some similarities with Basque/Aquitanian. Unfortunately we don't know for sure if they arose from very longstanding contacts or are linguistically genetic. Italy is a different case, with a lower R1b peak in northern Italy and even lower incidence of R1b in Tuscany and northern Latium where Etruscan had been spoken. Sardinia and Sicily (with the probably non-IE Sican languages) aren't R1b champions either. (I leave out the north Picene language, because I'm not sure if it's real at all.)

Simon_W said...

But regarding Celtic genetics, I find it noteworthy that the Celtic peoples were apparently quite diverse. BR2 from a Hungarian (late BA) Urnfield context rather resembles the French according to many tests, like the MDLP K13. And the Middle BA Tumulus culture sample from near Augsburg in southern Germany was similar, judging from his Dodecad K7b results. So the French may be mostly Gaulish-descended after all, with little accretions from the Roman age, and so there appears to have been a rather French-like southern belt of Celts while the Insular Celts (probably also the Belgae) were quite different, NW European-like.

FrankN said...

Some takeaways on BA Rathlins from the Supp.Mat.

1. Archeological context:
"The radiocarbon dates of C122 (Sk 1) and C105 (Sk 2) revealed that it would have been impossible for them to have been buried at the same time. As such, it is possible that the cist had been re-opened at some stage to facilitate further burials or that one of the individuals was an ‘ancestor’ whose remains had been curated for a period of time prior to their burial in the cist."
Noting that all three Rathlins share the same yDNA alleles (to the extent they could be determined), I think a "family grave" scenario is quite likely. Rathlin1 could be the lineages "founding father"; In that case we may assume the immigration to have taken place around 2000 BC.
A "family grave" interpretation would imply to take the Rathlins as one rather than as three independent samples; though the changes from Rathlin1 to Rathlin3 might reveal something about the wives taken.

2. Origin of the migration:
p51: "While the Irish Bronze Age contains a substantial amount of Middle Neolithic ancestry we find no strong evidence to suggest that the Irish population from which Ballynahatty came was its source. It is also not likely to be identical to the component of Middle Neolithic ancestry found in German LN/BA samples, or indeed any continental LN/BA so far sampled.(..)The exception to this is the strong signal of introgression seen into Rathlin1 from BenzigerodeHeimburg_LN."
Note that Benzingerode lies at the northern outskirts of the Harz mountains, Europe's largest silver producer in the middle ages, with copper mining attested from at least 1500 BC onwards. It is also within the epicentre of yDNA I2a2, which apparently found its way to the British Isles in several waves between the Chalcolithic and Anglo-Saxon times.
The Nebra Sky Disc, created around 2000 BC, consists of Cornish tin and gold, and Tirolean copper, evidencing trade links between Cornwall and the Elbe-Saale region at that time. A parsimonous explanation could thus be an Elbe-Saale BB colonization of England/Cornwall (where the "missing genetic link" might eventually be found), and subsequent entry into Northern Ireland from there.

3. Further Admix:
Dstats (p. 53) show Rathlin2 differing from Rathlin1 by being more Sintashta/Andronowo like, but also having a Cardial_EN signal that is lacking with Rathlin1. Rathlin3 shows a substantial Unetice_Czech signal. Whether we are dealing with later immigration waves here, or these admixes were already present with the original colonists, and entered the Rathlin lineage via local intermarriage, remains to be seen.
Note that the Cardial_EN signal in Rathlin2, and consistently high KO1 and Loschbaur WHG signals with all three Rathlins could point at some continuity of the Irish Neolothic (Ballynahatty) population, as is also hinted at in p.50f of the Supp.Mat.

4. Continuity:
Here are the top median haplotype donations from Rathlin1 to modern populations (Tab S14.2, p.60):
- Ireland 36.313
- Scotland 36.512
- Welsh 35.745
- Germany/Austria 33.658
- French 32.299
- English 32.212
- Norwegian 31.425
- Orcadian 30.072
- Tuscan 29.202
- ..
- Lezghin 27.028
- Finnish 26.431
- ..
- Chuvash 23.865
- Basque 23.824
- Georgian 23.284

Obviously, separate Cornish data would have been nice, and having just 4 modern samples from Germany/Austria is (as usually) limiting the meaningfulness of the comparison. Still, the trajectory is obviously via Central Europe rather than Iberia, and ultimately more Caucasian than EHG/Steppe.

FrankN said...

ADDENDUM: I first thought of a typing mistake when checking the list provided under #4 in my post above. But it isn't a mistake: Rathlin1 is indeed (slightly) closer to modern Scots than Irish, which might provide a hint on where to look for the direct source of the immigration into Ireland (supposing there was a stop-over somewhere in Britain).

andrew said...


"I'm assuming the latter. I figured the Urnfield culture was proto Italo-Celtic, and that the Atlantic Bronze Age was non-IE. I may have to re-evalute that though."

My thoughts exactly. And contra Krefter:CeC

1. Basque could only get their modern genetic profile if they converted away from IE or had massive introgression from IE populations.
2. The historically attested range of Basque-like languages was considerably broader than the current range. Add cultures in close continuity with them and it is even larger indeed.
3. This was a quite rapid expansion reducing the time for language evolution - it had run its course in less than the time between Shakespeare and modern English.
4. The Vasconic subtrate corresponds quite closely with Bell Beaker territory.
5. We know that Celtic is young because of the similarity of attested languages in the family. Even if the substrate weren't Vasconic, it would have to have been something other than Celtic (compare Eastern Europe with Baltic and Balkan languages which were IE were replaced by IE Slavic languages).
6. We can estimate Celtic's age from its close linguistic ties to Italic language which are better attested in time frame due to its proximity to the literate Mediterranean world. The Vasconic substrate is likely what differentiates Celtic languages from Italic one where there was a different substrate.
7. Celtic languages were historically attested as they were emerging even though it was patchy because it was on the fringe of the literate world. Its oral histories are themselves pretty recent. And, Celtic culture has lots of other components that can be used to trace it.
8. The is cultural continuity in the Bell Beaker region which suggests a common language family and the huge demic change at the time of their appearance refutes the 20th century conventional wisdom that they were merely traders and merchants with little demic impact. And demic impact that big also suggests language shift. Bell Beaker linguistic unity is also suggested by the extent of trade and internal migration within the Bell Beaker region.
9. Cremation is a pretty good litmus test of IE v. non-IE languages in other places where their arrival is dated and that comes with Urnfield not Bell Beaker.
10. There are genetic traces of a latter, much weaker demic impact Celtic transition which by consensus of archaeologists, historians and geneticists happened much more gradually.

Krefter said...

"Urnfield context rather resembles the French according to many tests, like the MDLP K13. And the Middle BA Tumulus culture sample from near Augsburg in southern Germany was similar, judging from his Dodecad K7b results. "

I don't know of any BA sample from South Germany. Are you referring to the BA guy from Halberstedt? He had a typical East European version of R1a and had around twice as much Steppe as modern French. He was could have been straight from Corded Ware.

andrew said...


Also, to be clear, nobody is suggesting these days that Vasconic languages including Basque arrived in the Mesolithic, and predominantly people are beginning to agree that it wasn't the language of the first Cardial Pottery or LBK farmers either. It arrived with steppe autosomal genetics and Western European haplotypes of R1b, ca. 3000 BCE to 2000 BCE depending on the location, almost surely with the Bell Beaker people, and replaced the language of the first farmers which in turn had probably caused the European hunter-gatherer languages to go extinct or become moribund.

While great genetic shifts almost insure language shift, this isn't necessary and elite dominance can bring about language shift without a huge demic effect. Some of the prototypical examples are the shift to modern Hungarian, a Uralic language in a country with less than 1% Uralic genetics, the Pygmy loss of their original language with low levels of Bantu admixture, the Mittani shift from non-IE Kassite language to an IE language, and the Northern Mesopotamian shift from Semitic Akkadian to non-IE, non-Semitic Kassite, to Hittite in a fairly short span with modest genetic impact. Turkey's shift to Turkish involved similar demic impact to the estimated demic impact of the Celts in the Iron Age.

Romulus said...


#6 in your list is interesting. Etruscans are worth considering too.

If Vasconic was a farming language, you would expect to see it of all places in Sardinia.

Romulus said...

Another point is that Vasconic is an Agglutinative language which is the kind of language we would expect a people originating in Siberia to speak.

FrankN said...

Now on the Celtic stuff - time to do away with some myths:

1. (East-)Hallstatt was not Celtic: There have been decades of intensive toponymic research in South Tirol, Austria and Switzerland. They show lots of pre-German/ pre-Roman place names, hardly any of which east of the Rhine is Celtic (and the few that are can easily be attributed to the Celtic expansion after the 4th century BC). Instead, their phonological pattern has been assigned to what is called East Alpine IE Types A/B. Type A is tentatively linked to Venetic, Type B to Illyrian (not that we have much knowledge on both of them).

To give an example: Kind of a Celtic "lead toponym" is *brig "elevation, hilltop (settlement)", well attested in Celtiberia (e.g. Braga), and also occuring quite frequently between Briancon (Brigantium), Bregenz (Brigantium) and Sarrebruck (Sarabriga). Its East Alpine version is Virgl/Wörgl. Same IE root, but phonetically impossible to be derived from Celtic, nor Latin, nor German (which has "Berg"/"Burg" instead). Further examples and more background (in German):

2. Ireland was multilingual well into the 6th cent. AD
Celts couldn't pronounce the sound "p", It got lost early (presumably via “p”->”ph”->”h”->’null’), evidenced by Gaul/Oir. athair “father” (lat. pater). “P-Celts” regained that sound quite soon again, shifting “kw”->”p”, to form words like pimpe(->quinque) “five”. [Interestingly, there are parallels to Germanic here: Gothic (and Turkish) atta “father” displays the same loss of initial “p”; Germanic “five” obviously has been borrowed from a P-Celtic pimpe with subsequent “p”->”f/v” sound shift.]

In Q-Celtic, including OIr., however, the “kw” was maintained (c.f..Irish cúig “five”). Moreover, the fact that St. Patrick was known in archaic Irish as St. Cathraig demonstrates the inability to pronounce “P” continuing into the early middle ages. Old Irish, from the 7th century onwards, however, has the “P” reappearing - interestingly on words that don’t have a clear IE etymology such as partán 'crab' or petta “pet”. Since borrowing from Latin, OEngl, and also Basque is unlikely for the words in question, they should have come from a pre-Celtic substrate language. That language must still have been in use in Ireland by the 6th/7th century – earlier borrowing would have turned the “pet(ta)” into a “cet(ta)”.
P. Schrijver (2013) even goes a step further. For the similarity between all Insular Celtic languages, he postulates a split as late as around 100 AD. According to him, Celtic would only have arrived in Ireland after Roman conquest of Britain with migrants from Wales. Schrijver is one of the most renowned Celticists worldwide, so his opinion weighs heavily. However, I haven’t yet seen this theory being discussed by peers, hence acceptance or refusal by the scientific community is currently difficult to gauge.

Slumbery said...

"Romulus said...

Another point is that Vasconic is an Agglutinative language which is the kind of language we would expect a people originating in Siberia to speak.

Do you mean typical Siberian groups like Sumerians, Armenians and some Dravidian and Cartvelian languages? :) There is absolutely nothing in agglutinative languages that particularly ties them to Siberia. Not only because they are attested from very ancient times from places that are not really closely connected to Siberia, but also this is an attribute that varies even within established language families and can pretty much change with time even during the history of one language.

Romulus said...

Absoultely nothing except Altaic, Proto-Uralic, Eskimo & Native Amercian languages.

Rob said...


"to be clear, nobody is suggesting these days that Vasconic languages including Basque arrived in the Mesolithic, and predominantly people are beginning ..."

Are there any references for that ?

Davidski said...


In regards to the haplotype tests, first I'm going to look at the steppe invasions, with Corded Ware, Hungary EN and BA, LBK, Srubnaya and Yamnaya genomes.

If that goes well, and the output from the first few chromosomes looks pretty good, I'll check out the issue of hunter-gatherer survival/rebound.

The Mathieson dataset includes a new high coverage La Brana-1 genome and a new decent coverage KO1 genome, so it should be doable.

Arch Hades said...

R1b's high frequency in Western Europe is not complicated. It arose in the East but there was a huge founder effect in Western Europe some time after the Neolithic. The founder probably was significantly EEF and WHG admixed, but unlike the pre Bronze age populations of Western Europe this dude also had a ton of Steppe ancestry which was also the ultimate source of his R1b. R1b not being derived from the steppe makes little sense because EEFs and WHGs in Western Europe almost completely lacked it.

Rob said...
This comment has been removed by the author.
Kurd Dgk said...


When i saw your post about "signal of admixture in Kotias from a source basal to MA1 & EHG", I thought what a coincidence. I think we are on parallel tracks. I posted something similar at and with regards to my findings prompted by you pointing out the issue of "CHG like" admixture in Austronesians which my latest ASI calculator had shown.

Basically, I had concluded that it appeared that the Onge, due to their drift, caused a divergence of Eurasia 9's ASI cluster's from a hitherto undefined "basal KOTIAS-S Asian" cluster, to such an extent, that alleles in S Asians and the TRIBALS that would normally be associated with this as of yet undefined basal Kotias - S Asian population, latched onto Eurasia 9's CHG component, in lieu of ASI, because of the distance of Eurasia 9's ASI component, caused by the drifted Onge from this hypothetical Kotias-S Asian basal cluster.

You call it a source basal to MA 1 and EHG, and I called it a source basal to Kotias and S Indian tribals.

The only thing is I am not too optimistic that we will see good DNA sequences from the late upper paleolithic from S Asia to build on this theory, although they may not have to be from S Asia

Rob said...

@ Arch Hades

""It arose in the East but there was a huge founder effect in Western Europe some time after the Neolithic. The founder probably was significantly EEF and WHG admixed, but unlike the pre Bronze age populations of Western Europe this dude also had a ton of Steppe ancestry ...""

Needless to say, it wasn't a "dude" or a "royal clan "
It was a mass migration (relatively speaking) of a whole people- an interbreeding network widely dispersed but maintaining social and biological ties . And by "migration" we are probably looking at a relative residential and ecological shift - as far as Central Europe is concerned (a "filling in")

Davidski said...


The massive migrations from the steppe to Europe and Asia took place generations after Y-chromosome founder effects on the steppe brought the main Kurgan "royal" lines close to fixation.

So one doesn't contradict the other.


The Admixture program works in mysterious ways. I have problems reading the output more often than not.

If the output doesn't correlate with well established results from formal tests, it's usually best to move on.

I know that there are many ancient samples, including Mesolithic samples, from north India and Pakistan stored at a university in the US. I've notified a source at an ancient DNA lab about this. Let's see what happens next.

Rob said...


"The massive migrations from the steppe to Europe and Asia took place generations after Y-chromosome founder effects on the steppe brought the main Kurgan "royal" lines close to fixation."

I see. So the 'takeover" of the steppe by Z93 of the Z2103 steppe was kind of like a "War of the Roses".
Maybe they'll make a TV series about it ?

Rob said...


The only "Royal" kurgans were those of Majkop

The steppe-Yamnaya kurgans are very austere
Given their sheer numbers, they look like emulations of earlier, richer ones in the Caucasus and Cernavoda/ Usatavo (derived from Cucuteni)

Some basic but important facts worth knowing ..

Kurd said...


It would be great indeed if we can get good quality samples, assuming they are from the right time.

The funny thing is that all I had to do was remove the drifted Onge from the run. The decrease in distance from this ghost basal cluster to my ASI cluster resulting from Onge removal was just sufficient that S Indian and Austronesian alleles shifted from CHG and latched onto ASI. The net result was 0 CHG in those tribals, accompanied by a corresponding increase in ASI.

I started suspecting the existence of this ghost basal source right after acquiring CHG from Jones et al. Anytime I would do an IBS run or Dstats, I would see Kotias more aligned with MOTA, Indian tribals, and others, than some of the other ancients from the area. There is no reason to believe there are any substantial sequencing errors with Kotias

Krefter said...


I know you've made a lot of graphs from D-stats. Do you have them all saved? If so can you post your email, so I can ask you sometime in the future to see the graphs. Or maybe you send all of them to my email:

Ryan said...


"Do you mean typical Siberian groups like Sumerians, Armenians and some Dravidian and Cartvelian languages? :)"

Probably excepting Sumerians, that's a list of very CHG peoples. I'm not saying I necessarily by this linguistic argument, but there's not a lot of evidence ruling it out either.

David's Tree-mix runs had CHG as 25% ANE too, so there's a plausible reason why CHG-rich and ANE-rich groups could share some deep linguistic affinity. That's basically the premise of Dene-Caucasian too, and it can

That doesn't mean it's necessarily correct either - there's a lot of ifs and buts - but I don't think it can be firmly excluded yet either.

Ryan said...

Re: the Royal R1b stuff, doing a bit of reading, it seems like a lot of that is simply because Hugh Capet happened to be R1b, which isn't surprising given his family originates from Worms. Another important figure would be Elimar I, Count of Oldenburg, who founded the House of Oldenburg. He was from Lorraine.

The War of the Roses comment is a good one. The Yorks seem to have been G2a, and the Lancasters R1b. Which is the true haplogroup of the Plantagenets? Probably R1b given their origins in NW France, but we can't be sure. And that goes back to the previous comment - why should it be surprising that a bunch of families from very R1b rich areas turn out to be the R1b?

Not to mention the discrepancy means there's at least 1 non-paternity event in the Plantagenet line. It's not like a royal haplogroup could maintain itself over so many generations when the Queen could be shacked up with the stableboy. They didn't have paternity tests. When Christian I became King of Denmark, Norway and Sweden, it wasn't because he was R1b. Heck, it wasn't even because of his fauther - his claim came from his mother descending from from Eric V of Denmark!

FrankN said...

@Ryan: Interesting. Even more interesting would be to know why most of Armenian and Georgian nobility apparently has yDNA I2c, which supposedly should be a WHG marker.

Rob said...


What do you suppose happened to all the pre-3000 BC Neolithic 'culture groups' in Germany? Were they subsumed under the CWC rubric ?

Slumbery said...

Romulus: the fact that there are lot of agglutinative languages that do not realistically have Siberian roots is alone sufficient to disprove the claim that this grammatical attribute would be particularly Siberian or that a group with such a language should be suspected to have Siberian roots just because of this attribute. Note that even my range of examples was far from full, please look it up. Your list of possibly Siberian-related instances holds no weight as an argument against this. (And I also made two other points.)

Ryan: my point was that this feature is not a trustworthy indicative for the starting area of an ancient migration. Nevertheless, a lot of ancient Middle Eastern languages were agglutinative, not just Sumerian. According to Wikipedia: Gutian, Elamite, Hattic, Hurrian, Kassite, Lullubi, Sumerian, Urartian. Most of them almost surely were high on CHG and I would not even discard Sumerians, we do not know their genetics after all.
Still, this is not a very robust argument either, but apparently you are aware of that. :)

Kristiina said...

I agree that CHG area looks like harbouring agglutinative languages while Siberia harbours more divergent languages. Yeniseian languages are particular and Na Dené languages are particular as well, and both have been compared with Sino-Tibetan languages. Also Jukaghir is structurally unique (split intransitive alignment system, lack of subordination). Many/most American languages are polysynthetic, and Eskimo-Aleut languages also show many polysynthetic features similarly as Chukotko-Kamchatkan languages and Nivkh which is a language isolate. Moreover, in the recent linguistic paper "Support for linguistic macrofamilies from weighted sequence alignment", Japanese language was on the same branch with Ainu and Austroasiatic languages and not with Altaic languages.

Currently, I think that IE languages, Uralic, Turkic and Mongolic/Tungusic all have a Bronze Age/Iron Age Steppe origin and replaced previous languages in the areas where they spread.

FrankN said...

"What do you suppose happened to all the pre-3000 BC Neolithic 'culture groups' in Germany? Were they subsumed under the CWC rubric ?

First a technical note: There have been two major collaborative research projects under Federal funding over the last decade. One, called "Early Monumentalism & Social Differentiation" deals with the Megalithic, which in practice means mostly FB, and is coordinated by J. Müller (Kiel). The second one is a joint Franco-Belgian-German project on the Michelsberg Culture, the German component of which is coordinated by the second grandseigneur of German neolithic archeology, D. Gronenborn (Mainz). [A third, smaller but still sizeable initiative, funds cooperative research with national institutes on the EE Neolithic. A 2008 outline (FU Berlin) announced aDNA sampling of Steppe cultures, the results of which we have been seeing now.]

However, these are lighthouse projects, from which most of what I have been presenting here originates. Otherwise (except for earlier, intensive LBK research) it is still pretty dark as concerns German prehistory - a PhD here, a local initiative there, but hardly any over-regional compilation of finds, not even to speak of a comprehensive analysis. It gets better again once we approach Hallstatt/ La Tene, but even there its mostly regional patches. Thus, don't expect too much from me when it comes to CWC, Unetice etc.

CWC seems to be a "catch all" term that apparently emerged from an early 20th century desire to create comprehensive "eras", ideally encompassing all German and/or Austro-Hungarian territories of that period - Swiss-German archeology gladly joined in if they had the opportunity. CWC was combined with GAC (which is now being refuted), the Single Grave culture (refuted by part of German archeology), the Battle Axe culture (relation still unclear). In any case, until 2500 BC it coexisted with late FB/GAC-derived cultures, afterwards with BB. The Elbe-Havel and/or Schönefelder Culture (sometimes separated, sometimes treated as unit) around and NW of Berlin was never part of the CWC horizon.

The "state of the art" analysis appears to still be Furholt's 2003 paper linked below (German, but English abstract/ legends). Google doesn't find any German publication on CWC since 2010 (well, the money was with MC/FB).

Furholt distinguishes 3 phases:
1) Non-ceramic phase (2900-2700BC), when the CWC-typical burial rite (crouched, E-W oriented, gender-differentiated single burials) emerged - only in Poland already accompanied by CW ceramics
2) "Einheitshorizont" (unified horizon, 2700-2550 BC): CWC burials accompanied by similar vessel types across the whole area. These vessel types are closely connected to Globular Amphora, and also spread preferably in areas that had GAC settlement before.
3) Regional subcultures (2580-2200BC), in some regions (Elbe-Saale, Tauber Valley, Denmark) continuing at lower intensity until 2000BC.

Note that for Southern Germany, excluding the Tauber Valley, 5 (in words: five) ceramic CWC burials have been recorded for 2700-2200 BC, and 16 in NW Germany/Netherlands.
Furholt concludes a gradual, more "fashion-like" spread of CWC, and deems substantial immigration unlikely.

This analysis turns the attention back to GAC, which was clearly migratory, though generally hybridising/ overforming rather than replacing existing cultures. See this recent english-language paper on their western advance for details:

Aram said...

Imho we will find some R1b in GAC culture with CHG. If not there then in Baden / Vucedol.

For the linguistic and genetic correlations. If any real correlation exists.
CHG seems to exhibit a tendency to have high number of phonemes.

mickeydodds1 said...

Whoever the 'eastern hunter gatherers' were, those men, and it was men, surely must rank as the awesomest humans ever to have walked the planet, and probably ever will walk this planet.

Alberto said...

Since BR2 looks mostly like a mix of NE1 and Motala, I thought I'd check a BR2 - NE1 Haplotype donation to get an idea of a possible Motala-like haplotype donation. Not something necessarily accurate, though it does seem to make some sense:

Lithuanian and Polish at the top is more or less what I would expect, and a significant difference with Ireland and Scottish. But let's wait for real Motala results (or ideally a Motala-like samples from around Poland, if that actually comes at some point).

Ryan said...

@Alberto - Agreed on all points, and I'd just add that I think some samples from the Lower Danube would be nice too.

@FrankN - the royal families of Armenia and Georgia claim common descent from King David in Israel, so as you can imagine that makes finding real information about their origins difficult lol.

Based on a quick google though, I2c seems to peak in frequency (albeit still low frequency) in Crete and in the Caucasus, so maybe it's a minor CHG lineage? The break up of I2 would presumably be pretty ancient.

Krefter said...

"Since BR2 looks mostly like a mix of NE1 and Motala, I thought I'd check a BR2 - NE1 Haplotype donation to get an idea of a possible Motala-like haplotype donation."

Did you do the test yourself? Isn't this test designed for tracking recent ancestry? IMO, that's Ireland_BA was so distant from Lithuanian and so close to Isles Celts.

I think you should consider doing that test with moderns more often than with ancients, if it tracks recent ancestry. Motala shouldn't have recent shared ancestry with anyone, even if Motala-like people contributed ancestry to some moderns. And BR2 is probably close to East Europeans in this analysis because he lived in Hungary in 1000 BC.

I'd like to see stats for.
South Dutch: Recent ancesry with French.
Austria: Recent ancestry with Balkans.
East Germany: Recent ancestry with Poland.
Croatia: Recent ancestry with all Slavs.
Serbia: Recent ancestry with all Slavs.
All of Iberia: Recent ancestry within Iberia.
All of France:
All of Italy:

Another great test would be, Copper age Spanish, to see if they have recent ancestry with modern Spanish(definitely have a lot). Also, LN/BA Norse with modern Norse, should give a similar connection Ireland_BA has with Isles Celts.

There's lots of questions for Euro-genes diversity. France, Italy, and Germany are big regions/countries where we don't know if the people have lots of recent ancestry or are very similar genetically. It doesn't seem most Italians do. This analysis can help answer those questions.

Shaikorth said...

Krefter, a lot of haplotype work about modern populations has been done already.

ie. World analysis:

How to read
by row: scan each line to see which populations (columns) are the bigger donors for each row.
by column: scan each column to see which populations (rows) are the bigger recipients for each column.

High quality ancients will be interesting to see now. Lazaridis et al actually includes a run, but the pdf is too undetailed to find out much about the Chromopainter portion.

J. S. said...

"The fine-scale genetic structure of the French population" Aude Saint-Pierre, still in preprint.
Maybe it could help some of you.

Alberto said...


No, I didn't do the test myself, that's just the data from the paper. All I did was to take the numbers and calculate the diff between BR2 haplotype donation and NE1 haplotype donation (BR2 minus NE1).

David said he's taking this info from several ancient samples and many modern populations, so that's going to be cool. Though. as you say, comparing ancients to moderns might not be totally accurate, it still looks quite useful. Ideally this method will be really useful when we start getting many high coverage ancient samples with good overlap between them, because that can give additional information (sometimes decisive) to know which of the models that formal stats show as possible is the most likely to be correct.

Matt said...

@ David, thanks for the info. Good luck.

@ Krefter, I created an email at for contact if you want to email anything. With the graphs I've created, I have kept most of them somewhere on my HD - I've not really been very systematic about naming them and storing them though.

Btw, re the haplotype donation: - Graphing haplotype donation against shared drift as modelled by D(Chimp,Ancient,Mbuti,Pop). Note I graphed the Bedouin from the paper against BedouinB, which may have some effect, and I also used the closest ancient I had in the D stat.

Correlation between Loschbour shared drift and haplotype donation is pretty good with expected outliers, less good for Ballynahatty haplotype vs shared drift with LBK while NE1 and Stuttgart haplotype vs LBK shared drift has quite low correlation.

The contrast between haplotype sharing with EN farmers vs shared drift for many Near Eastern and West Asian populations is quite extreme (esp for the Levant populations likely to have higher population size) - PCA on differences between haplotype donation (same thing without Rathlin -

Rob said...

@ FrankN

Thanks. I'm actually familiar with that article, but was good to re-read
I wonder what the role the Baden culture might have had in stimulating the cattle/breeders of GAC, who in turn appeared to have pre-empted / facilitated CWC

Aram said...

Happy New Year.

For I2 in Armenia and Georgia. The bulk of that I2 is under the branch I2c2.
Formation age 15800 ybp, the real TMRCA calculated by my friend is ~5000 ybp.
His frequency is 3.5% among Armenians.

I think it was a Anatolian Neolithic marker ( I2c was present in Anatolian Farmers Y DNA ) who had a some sort of bottleneck during the transition from Neolithic to Bronze Age.
But before becoming an Anatolian Neolithic marker it could be a WHG hunter in Anatolia. His high TMRCA permits such an assumption. We don't know did WHG people existed in Anatolia before ENF. But they could live, why not?

As for the probable link of I2c2 with Bagratid ruling family that is a more complicated story.

Karl_K said...


"We don't know did WHG people existed in Anatolia before ENF. But they could live, why not?"

We actually do know that they were there.

Gioiello said...

@ Aram

Hg. I2c: “But before becoming an Anatolian Neolithic marker it could be a WHG hunter in Anatolia”. Of course, as all the other hgs found in Barcin which have anything to do with Middle East but come from Southern European hunter-gatherers, more than WHG. They come of course from the “Italian Refugium” and from Southern European Refugia more linked with the Balkans, in fact R1b hasn’t been found in aDNA from Anatolia, being more western, except the few old haplotypes migrated before to the Russian plain.

Alberto said...


Those first plots comparing halplotype donation with shared drift are very interesting. They seem to tell us quite well about the difference (limitations?) of each method. With more testing we'll see, but tentatively:

- As you pointed out above, haplotype donation seems to be quite affected by isolation, endogamy or similar. Populations that are quite specific, like Druze, Bedouin, Basque, Lithuanian, probably Sardinian, seem to have lower haplotype sharing than it would be expected. So this should be taken into account when using this method with these populations (others affected could be Kalash, Ket, Selkup,...). (As a side note, these same kind of populations might have a higher shared drift than expected, as we usually see them pretty high in formal stats, though this effect is quite less apparent, but we saw it recently in some Balochi/Brahui stats with the Kalash).

- Population with even small amounts of very divergent (from the main components compared in the populations) admixture are quite affected in formal stats (depressed shared drift). Here we see it with small amounts of Sub-Saharan admixture (Druze, Iranian, BedouinB, Syrian, Jordanian... and to a much lesser extent with Sicilian or Spanish), or with ENA admixture (Iranian, Chuvash, to a lesser degree Turkish, Adygei, Finnish...), though in the case of ENA only if compared to populations with Basal Eurasian admixture, but not if compared to Loschbour).

So this could be quite useful to deal with populations that fall under any of these 2 categories, since one method could correct for the limitation in the other (the case of Egyptians is quite clear, when with formal stats they will appear very distant from Eurasian populations due to SSA admixture, but the haplotype sharing place them high in Anatolian farmer ancestry, which should be correct).

The second graphs with subtraction also show that the method works well, I think. The results for, say, BR2 - Loschbour or Rathllin - Loschbour make sense when thinking about the admixture components. Probably with Yamnaya added (or Kotias and Karelia_HG) we could have a PCA that looks quite good.

Matt said...

Alberto: As you pointed out above, haplotype donation seems to be quite affected by isolation, endogamy or similar. Populations that are quite specific, like Druze, Bedouin, Basque, Lithuanian, probably Sardinian, seem to have lower haplotype sharing than it would be expected.

I think that seems likely to be an issue for Palestinians and Druze and also Basques and Sardinians and possibly Orcadians.

Lithuanians I'm less sure about being affected, as their gROH (mean genomic runs of homozygosity) and nROH (count of gROH>1.5 Mb per sample) for these Lithuanian samples seem quite typical for Europeans. See figure 5 in Esko et al 2013 on isolated populations in Northeast Italy - I can't see that Lithuanians are more homozygous than the Poles in that set. The blogger M Myllyla also produced a list of populations by homozygosity on his blog, and although teh Lithuanians are higher here, they are found as less homozygous than the Irish samples he has For them it might be more than their extra WHG relatedness comes through a SHG / EHG route, so doesn't predict elevated haplotype donation from Loschbour very much (we'll probably see in the future).

One other factor that might affect these, by the way, might be the time of admixture. If there were Loschbour / EN holdouts in some region who mixed with the other local pop relatively late, their haplotypes would have less time to break down, so more donation would be detectable. Haplotypes are affected by a lot of factors but are potentially very useful for fine scale (this otherwise similar population exactly?) and timing.

Shaikorth said...

Esko et al's Lithuanians are not the same individuals that are in public datasets such as Human Origins and used in most tests, and there is much more of them. They might appear less endogamous and perhaps also less homozygous because they are a more diverse sample.

Runs of homozygosity are bigger than Chromopainter chunks so they may not be directly comparable, but here is one of Anders Pålsen's runs featuring the public set Lithuanians

Their internal chunk sharing is about at the level of Finns and the more heterogenous Orcadian/British cluster. Clearly lower than the endogamous Orcadians, Saamis, Sardinians and Basques though.

Simon_W said...

One important point in the discussion about the age of the earliest Celtic language: Germanic proper is defined by the Germanic soundshift, and convincing evidence suggests that this occured remarkably late, after 500 BC. A famous example is the Skythian loanword kanba (hemp). The word (along with the product) were introduced into the Germanic area after 500 BC, and as you can see, it was affected by the Germanic soundshift, which means that this occured after that. The proto-Germanic word would be chanapis (*χanapiz).

So how is this to be reconciled with the date of the Germanic vs Italo-Celtic split inferred by Chang et al.? Quite simply: The languages started to diverge long before the defining soundshifts occured. So maybe these early Bronze Age Irish didn't yet speak Celtic proper, but something that already started diverging into this direction.

Simon_W said...

@ Krefter, re: Middle Bronze Age sample from southern Germany, I was referring to RISE471 from Allentoft et al. It's a male from Untermeitingen, in the Augsburg area. The supplementary table 1 doesn't give any exact date, it just places him into the Bronze Age, but according to Figure 1 in the main text he's Middle Bronze Age. I analyzed the sample with DIY Dodecad, and this are his results:

K7b =
72.94% Atlantic_Baltic
20.43% Southern
6.09% West_Asian

K12b =
47.9% Atlantic_Med
33.31% North_European
18.2% Caucasus
0.03% Gedrosia

For comparison, BR2 had:

K7b =
69.91% Atlantic_Baltic
19.5% Southern
10.1% West_Asian

K12b =
41.68% North_European
36.28% Atlantic_Med
16.57% Caucasus
3.48% Southwest_Asian
1.14% Gedrosia

So they look quite similar.

Simon_W said...

Judging from this he has even somewhat more MN farmer ancestry than BR2.

Simon_W said...

As for the theory that Bell Beaker was Vasconic: In my opinion there is no sensible reason to assume a common language for the entirety of the Bell Beaker complex. Initially it was just a package of some fashionable goods that spread from Iberia. And then there were several sub-complexes within the Bell Beaker world, and some of them were even antagonistic to each other, see Harrison and Heyd 2007. (A group from the Danubian Beaker area destroyed the anthropomorphic stelae of the Bell Beaker folks from Sion Le Petit Chasseur.)

Simon_W said...

HOWEVER: It does make sense to assume a bunch of related languages along the Atlantic façade of Europe during the early and middle Neolithic. After all, there were just two streams of early farmers moving into Europe (Danubian and Cardial). Although these were genetically very similar, they appear to have sprung from two different groups of early farmers, so two different founding languages for the EEF seem like a reasonable assumption. The Neolithic Irish woman shares more drift with Iberian EEF than with German ones, therefore a predominance of Cardium derived ancestry may be assumed. I don't think the gradual, slow increase in WHG ancestry, and the overall minor WHG accretion is compatible with a shift to WHG derived languages, therefore they all would have spoken Cardium derived languages. But of course not one single language. They undoubtedly diverged a lot during the millennia that followed the initial settlement. Although there were movements and contacts along the Atlantic façade, evidenced by the spread of Megalithism. And by the large effective population size indicated by the Irish Neolithic woman, they were not endogamous.

Simon_W said...

@ Krefter

I just noticed that according to the latest version of the Eupedia R1b map, Catalonia is no longer on a par with the Basques and the insular Celts, their R1b has dropped:

So the highest levels of west European R1b are associated with Basques, Insular Celts and Romanized westernmost Gaulish Celts.

Simon_W said...

There are quite a lot of parallels between Basque/Aquitanian and Iberian, and although their genetic relationship isn't yet proved beyond doubt, the evidence is compatible with it. And there are just two other west European ancient languages that have been suggested to be non-IE: Tartessian and Pictish. However, the exact affiliation of Tartessian remains uncertain (one expert has suggested it had an "Iberoid" appeal) and Pictish (or rather the language of the strangest Ogham inscriptions) may have been Celtic according to some more recent analyses.

East of western Europe, there were basically just two non-IE language families:



Minoan / Eteocretan (though it's not beyond doubt that they were related, but it would make sense historically).

One of these may have been related with the Danubian EEF languages (wild guess, but possible), the other may have arrived later with CHG-related influx.

Cyprus with Eteocypriot is already far in the east.

Simon_W said...

There is no direct linguistic evidence from the pre-Roman languages of Sardinia and Corsica, but according to some ancient authors these people were related with the Iberians, which is also compatible with the toponymy in Sardinia, I've heard. From the Sican language few fragments do exist, but hardly enough for a proper classification. According to Thucydides the Sicani were of Iberian stock, too. So the entire pre-IE West Med area appears to have been Iberian-related. And yet this part of Europe has less R1b than the Atlantic façade, so probably R1b wasn't associated with Iberian.

I think anyway, associating R1b too much with Basques isn't very fair towards the few real Basques who don't have it. R1b cannot be what makes someone a Basque, can it?

Krefter said...


Those results for RISE471 are interesting. He's from close to where Gauls lived, in South Germany in-between Austria and Switzerland. He's of pretty low coverage though.

There are so many ancient genomes now it's hard to keep track. I'm going to start gathering as much info as I can from people who can do analysis of them.

Gioiello said...

@ Simon W
“According to Thucydides the Sicani were of Iberian stock, too”.
Trying to understand the origins of Sicans from Thucydides is the same in understanding the origins of Etruscans from Erodotus (his theory was linked to the politics of Smyrne in the 6th century, with no real proof). Sicans and Sicels were Italic peoples, and the R1b proof you carry has no meaning: western Sicily is high in R1b and even in old subclades.

Krefter said...

The Kumtepe genomes are in. they address the genetic changes that occurred in Turkey after the Neolithic, and might have genomes to prove there was migration from other regions, I don't know yet.

Alberto said...

From the abstract is seems to be mostly an Anatolian farmer, but already showing some amount (probably not too big) of CHG admixture. Given the early date (4700 BC) I guess it's quite in line with what we could expect.

I hope there's some more info soon.

Roy King said...

Yay! Much CHG-like component in Kum6. More than any other Neolithic sample studied thus far.

Krefter said...

Kum6 doesn't appear to have CHG. He's a normal EEF except: Has a strangely close relationship with Otzei and West Asians.

Chad Rohlfsen said...

That's not CHG in Kumtepe. No statistical difference in the relationship to Yamnaya, versus other EN samples. Look at lower K's. Nothing in Yamnaya is in Kumptepe, on top of that, there is no CHG and EHG component. You guys really need to read and investigate stuff before flying off the handle.

Roy King said...

Sorry Chad--look at K=12 where one can easily see the connection between Kum6 and other West Asian populations, more so than the European Neolithic samples. Not being focused on Yamnaya and with no axe to grind, I would suggest bracketing your ideas about Yamnaya and study directly Kum6 and the other Neolithic samples!

Chad Rohlfsen said...


You need to look at stats and not some bullshit, K12 modern cluster stuff. It's nothing present in Yamnaya. That k12, orange component is Middle Neolithic, not EN. That green is an increased NE affinity, also present in other EN samples, which will be almost identical in Neolithic Anatolians, if they tested them. This Kumptepe sample will probably be identical to Anatolians. Not only that, the quality is so low, I wouldn't even bother with these samples. You guys put too much weight on Admixture runs in these papers, which is always based on modern clusters, and not ancients. You need to look at stats!

Chad Rohlfsen said...

Double post again from Anthrogenica, just to stress the importance of looking at stats, before you guys try judging stuff on Admixture output. It is very important to first look here.

From the paper...

* Furthermore, the Bronze Age Yamnayan component suggested to be a part of the Corded ware expansion [19, 24] is
not present in Kum6, and thus is not producing any increased affinity to the ancestors of the Yamnaya culture from north of the Caucasus (D-Denisovan, Yamnaya_RISE; Kum6, early farmer), all Z > _1.7).

Then, look here.

Denisovan Yamnaya_Kalmykia Anatolia_Neolithic LBK_EN 0.0044 1.746 23450 23246 492005

Kumtepe looks like an Anatolian to me. Please, refer to stats. Admixture outputs on these papers leave too much open to interpretation that can often be wrong.

Open Genomes said...

To help the discussion about Omrak et al. (2015), here are links to the images of the MDS Plot, and a "flipped and rotated" version that more or less replicates the "standard West Eurasian MDS plot orientation", and also, a very enlarged image of the K=12 Eurasia-wide Admixture analysis from the supplementary data.

Omrak et al. (2016) Fig. 1 PC plot of aDNA samples including Kum6
Omrak et al. (2016) Fig. 1 PC plot of aDNA samples including Kum6 (rotated)

Omrak et al. (2016) Fig. S2 K=12 Worldwide PCA

Notice that Kum6 has an additional "blue" component not found among the other Early Farmers, but found among the Druze and adjacent Levantines, also among the the Indo-Pacific-speaking Kusunda hunter-gatherers of Nepal, and the Tubalar of the Altai .. the problem is that their K=12 just doesn't correspond to components like CHG. (Notice that Yamnaya has no "dark green" even though this is modal among Northwest Caucasians. Instead, Yamnaya has something that appears to be related to ANI.)

Nirjhar007 said...

Open Genomes,
Yamnaya has something that appears to be related to ANI
Can you please elaborate a bit ? :).

Simon_W said...

@ Gioiello

"Sicans ... were Italic people"

I classify this in under "whishful thinking".

Gioiello said...

@ Simon W

I think that you’ll realize soon that mine are dreams at the light of the reason, as we say in Italian.

Balaji said...

The Cassidy et al. paper associates cultural and genetic changes with “the arrival of agriculture” and “the onset of metallurgy” - reasonable enough. They then say, “The second great wave of change starts with the appearance of copper mines, associated with Bell Beaker pottery, which are quickly followed by Bronze tool-making, weaponry, and gold working, with distinct Food Vessel pottery succeeding from the earlier beakers.” Nirjhar and Gioiello have already commented that it must be Bell Beaker people who migrated to Ireland rather than Corded Ware people. The Corded Ware people did have ancestry from the Steppe but the Bell Beakaer people were contemporaneous with Corded Ware and originated far from the Steppe in South Western Europe. In Central Europe, they eventually replaced the Corded Ware. Moreover the Corded Ware people only had stone battle axes and were unschooled in metallurgy. It is likely that Cassidy et al. are wrong when the assert, “Thus, it is clear that the great wave of genomic change which swept from above the Black Sea into Europe around 3000 BC washed all of the way to the northeast shore of its most westerly island.” Haak et al. only made the modest claim, “These results provide support for the theory of a steppe origin of at least some of the Indo‐European languages of Europe.”

The new year is likely to see the cutting down to size of the Steppe hypothesis and show that only some of the Indo-European languages of Europe originated in the Steppe.

Rokus said...

Isaac (2010) asserts on Celtic that it 'makes no sense to think of Celtic becoming what it is on the late Neolithic and the early Bronze-Age Atlantic seaboard while sharing striking grammatical innovations with Indo-Iranian, Baltic, Slavic, Greek, Tocharian and Albanian.' Likewise, Gibson & Wodtko (2011) agree that 'Celtic, thus, is not the first IE language entering an area of non-IE speech'. In Western Europe this means that Germanic and Italic, and a couple of minor extinct IE languages as well, arrived earlier than Celtic or its direct precursor. Now in Ireland it seems all but proven that there was cultural and genetic continuity straight back to Bronze Age and hence, Late Neolithic Bell Beaker times, another (?) complete linguistic turnover in the Bronze Age is less than likely. This is a strong indication that the arrival of IE languages elsewhere in Western Europe does NOT necessarily correlate with eastern European influences.

Grizzlor said...

I wonder if the I2a in western Ireland (Connaught) is an WHG /EEF remnant or some later founder effect (or just tagging along with the r1b alphas from the Steppe).

Cyffordd Dyfi said...

The case of Graham Isaac (and the late K. H. Schmidt before him) for an eastern IE background for Celtic depends on assuming that a very few common morphological features shared by Celtic and one or more eastern branch (such as Indo-Iranian) must be innovations. They could, however, be survivals from PIE that died out in the other branches. This is why Celtic as an eastern IE language has never become standard doctrine. On the other hand, the grammar and lexicon for an Italo-Celtic node can be worked out in considerable detail. In short, the case is not unambiguous and cannot stand on its own without clear confirmation from archaeology or genetics. I don’t think that Gibson and Wodtko really mean what’s been inferred here.

Rokus said...


Not so few features, and at least I now read in Cassidy a quite clear confirmation from genetics. The observations of Gibson and Wodtko are indeed more general, though nonetheless telling concerning the deepest IE layer in Western Europe commonly agreed upon.

Isaac- Origins of the Celtic Languages:

[...] the 'Italo-Celtic' postulate still has adherents (cf. Kortlandt 2007). But the counter-arguments are well documented and have not been adequately answered.

[...] it turns out that, despite the western location of Celtic in historical times, it shares no significant innovative features with the other prominent language of the Indo-European west, Italic and Germanic.

Celtic, Germanic and Italic were neighbours, because that is where we find them when they emerge into history. So of course they share many words with each other to the exclusion of other Indo-European languages.

Celtic shares the relative pronoun of the form Pro-Indo-European *ios, *ieh, *iod, with Greek, Slavic, Indo-Iranian, and Phrygian, to the exclusion of the western languages Italic and Germanic (and of others).

The old assumption that that relative pronoun must be attributed to the undifferentiated Proto-Indo-European language itself has been undermined by the recognition of chronological layers of dialectal innovations within that protolanguage, and that Greek and Indo-Iranian in particular represent a particularly late form of that language, compared with the form represented by, say, Anatolian or Italic

Of similar distribution are the forms of the future tense in the suffix *-sie-/*-sio-

[...] the reduplicateed thematic sigmatic desiderative is an extraodinary complex and innovative formation found only in Indo-Iranian and Celtic (in Old Irish future formations).

The fact that the syncretism of plain-voiced (Proto-Indo-European *bh, *dh, *gh, *g'h) stops is common to Celtic and Iranian, but not Indo-Iranian, as a whole, is indicative of differing dialectal contacts within Indo-Iranian - those of Iranian with Slavic and Baltic are hardly in question - and does not compromise the postulate that Celtic partook partially of the same areal configuration [...] both - perhaps at opposite ends - parts of a contact continuum.

The two sigmatic formations previoulsly mentioned, future and desiderative, are symptomatic of a general explosion in productivity of sigmatic verbal forms in general, represetned also, and most strikingly, by the rise and generalization of the sigmatic aorist as seen in Greek, Indo-Iranian, Slavic and Celtic. [...]it is only in Greek, Indo-Iranian, Slavic and Celtic that the sigmatic aorist became the productive formation par excellence for all new verbs, as well as being generalized in many cases to old verbs whose inherited aorist formations were moribund. [...] it cannot be emphasized strongly enough that it is the expanded productivity of this formation that constitutes in this case the significant innovation common in Greek, Indo-Iranian, Slavic and Celtic, not merely the existence of the formation.

Phonological innovations give us further insight into, and further geographical and chronological information on, the course of the development of Celtic out of Proto-Indo-European. For instance, certain inherited clusters of consonanrs, a dental stop followed by a dorsal stop, underwent a metathesis to dorsal-dental in Celtic and in Greek, and, it now seems, nowhere else.

Rokus said...

Celtic shares an important characteristic feature of phonological development with Tocharian in the rule of the treatment of certain word-initial laryngeals.
The Celtic and Tocharian treatments are structurally identical. There are so many factors relevant in these processesthat the identical Celtic and Tocharian outputs from the same inputs, contrasting with the Italic and Greek treatment, is diagnostic of this change taking place in common in contiguous languages or dialects.

A very striking waty in which Celtic differs in its historical phonology from the other apparently western languages, Italic and Germanic, is that the original distinction between the plain-voiced stops and the so-called aspirated voiced stops was neutralized in Celtic, whereas they remained distinct in Italic and Germanic. Celtic shares this syncretism of the plain-voiced and aspirated voiced stops with Beltic, Slavic, Albanian and Iranian (all other Indo-European languagesd maintaining the distinction, even if alterning the phonetics of its realization).

The loss of the distinction between the plain-voiced and the aspirated voiced stops is one of the most fundamental alterations to Iranian into its two branches. That places that change at around 2000 BC+/- <=c.300. This means that the dialect or dialects that would, several centurues later, become Common Celtic, were part of a language-contact area about 2000 BC which encompassed also the ancestors of Beltic, Slavic, Albanian and Iranian.

All this can only have been going on in eastern Europe. It makes no sense to think of Celtic becoming what it is on the late Neolithic and the early Bronze-Age Atlantic seaboard while sharing striking grammatical innovations with Indo-Iranian, Baltic, Slavic, Greek, Tocharian and Albanian.

Reconstruction of the Celtic whereabouts:
1.the innovatory morhological characteristics of the Celtic languages place them in their earliest accessible phase of becoming what they are in a dialectal complex that includes also the ancestors of Indo-Iranian, Greek, Baltic and Slavic.
2. several distinctive phonological characteristics, symptomatic of shifting proximity to other languages over time, indicate further areal relations with the ancestors of Greek, Tocharian
3. around 2000 BC, Baltic and Slavic, still together with Albanian and Iranian.

Gioiello said...

Rokus, I think that what you (or others) says has a little meaning. The fact you speak about are just “areal” as you say, i.e. that they may be due to some contact with other linguistic group in Central Europe, but what separates Celtic from the languages you speak about is just that it is a centum language and the others above all are satem ones, thus they come from different regions of Europe: Celtic from West and satem languages from East. The ancestors of Celtic may have been just those Bell Beakers who from Iberia or Southern France or Tyrrhenian Italy migrated to Central Europe and took the place of the previous CW people and perhaps there got some contact with the satem languages. I have no doubt that centum IE languages came from Italy and were linked with Cardial and after with Bell Beakers and hg. R1b (and others). When aDNA of Tyrrhenian Italy is tested, we’ll get an answer to all my hypotheses.

Rob said...

Very interesting; thanks

Simon_W said...

@ Open Genomes

That blue component found in Kum6 is also present in the earliest European farmers, albeit at a lower percentage: I see it in Starcevo_EN, LBKT_EN and in NE1.

@ Chad

I'm not so sure that the increased dark green component in Kum6 is just from early Anatolian farmers. Because some of the purest and earliest European EEF also have a light green component that's modal in Bedouin B. You can see it in Starcevo_EN and in LBKT_EN. That's also increased Near Eastern affinity, but of a different kind than the dark green component that's modal in Georgians. I'd bet that early Anatolian farmers would have that light green component, too. Without saying anything about CHG or the non-EHG stuff in Yamnaya, to me this simply looks like increased affinity to northern West Asians near the Caucasus.

Kristiina said...

This is to FrankN! If you are interested, I commented your earlier discussion with Ebizur about Korean and Burushaski but not the Burushaski part but the Korean word "hyo".

Cyffordd Dyfi said...

With reference to what Rokus said, thanks, very interesting.
It should be noted that the convergence of voiced aspirates and voiced unaspirated stops occurred in Iranian after it split with Indic. Indo-Iranian would, at the stage of common development, already have evolved far enough away from the western branches that there would not have been a high degree of mutual intelligibility. Leaving aside the question of why this would even have happened if Proto-Celtic (or the PIE that became PC) and Proto-Indo-Iranian were next to each other and in close enough contact to share innovations, it would be an effect more like bilingual contact than a dialect continuum. Many of the IE family trees have Celtic splitting off before Indo-Iranian, in which case Celtic's main defining features would probably have existed before Iranian had split off from Indo-Iranian as defined by the de-aspiration of the voiced aspirates and the other distinctive Iranian innovations.
The Italo-Celtic hypothesis is more robust than the eastern Celtic one and therefore keeps returning through different lines of inquiry by new generations of researchers.
I don't expect that many readers will see Cassidy et al. 2015 as support for the eastern Celtic hypothesis, as it shows that the steppe component reached Ireland early rather than following prolonged development alongside the region of the incipient satem languages. There as also the negative absence of evidence for a second IE wave entering (insular) Celtic territory in what remains of prehistory after the EBA.

Rob said...


The Italo-Celtic hypothesis has failed to convince for decades. Scholars now even doubt that was a Proto Celtic node itself, but rather occurred due to secondary areal convergences

Rokus said...

Satemization is indeed considered older than the syncretism of the plain-voiced and aspirated voiced stops that Celtic shares with Baltic, Slavic, Albanian and Iranian, and I agree it is quite unlikely it originated in an already satemized area altogether. Its occurrence in the east may thus rather be related for example to Allentoft's observation that Sintashta - that emerged at the beginning of the 2nd millenium in the Urals - had a closer affinity to the west, more in particular due to 'the presence of European Neolithic farmer ancestry in both the Corded Ware and the Sintashta'. In which case Iranian already separated from Indic, that does not share the feature, already within the wider Andronovo horizon. Hence, maybe the feature even originated in ancestral Celtic! Especially given the older age of Irish Beaker. However, this does not change the more eastern affinity of Celtic in relation with Germanic and Italic, on linguistic grounds and now supported by genetics.

Gioiello said...

@ Rokus
We, wrongly, speak of genetics and linguistics put together, but very likely the links are more complex. It seems to me that more and more people begin thinking that R1b-L51 and subclades come from West and not from East, from Bell Beakers expanded from Iberia, Southern France and Tyrrhenian Italy. We’ll see next if my theory of an “Italian Refugium” (which seems to me it too more and more followed) is right and if the Cardials came from Middle East or were autochthonous of Italy and nearby. After we’ll answer also to the origin of the IE languages, but don’t forget that also Mallory&Adams said that the most varied presence of IE languages (at least of the centum type) was in Italy.

Rob said...

Apparently Mesolithic - Bronze age samples from Italy are coming this year. So you're suggesting will find a whole bunch of R1b ?

Gioiello said...

@ Rob
I thank you. I understood that something was coming in that meaning from the post of Balaji (whoever he is and whoever you are). I think that that post, as I said immediately to a friend of mine, did mean that some test in that meaning were in the labs. Of course I am not sure, but we’ll see next if my theories about the “Italian Refugium” in these last pretty ten years were right or wrong. We are waiting the last test upon a friend of ours from Yseq: he is L51+, U106-, P312-, and will be or L151+ thus possibly DF100+ or the first L51* found so far, being diffculty PF7589, having DYS426=12. Of course, whichever will be the result, we are finding the oldest haplotypes both for the Y and the mt in “Tyrrhenian Italy”.