View Full Version : Little experiment with Dutch Beakers and French Neolithic

09-05-2020, 10:03 AM
This must be something most of you are already aware of but I keep hearing here and there that Bronze age Beakers had a huge impact in Western Europe, so I decided to run an experiment with the help of my compatriot Helgenes50.

We selected the average coordinates of 2 groups of ancient samples, the Dutch beakers (Bell_Beaker_NLD) and the Southern French Neolithic from Clos de Roques (FRA_MLN_Clos_de_Roque)

The selected Neolithic samples contain both WHG and farmer ancestry and are quite homogeneous


1st experiment

Helgenes50 produced 6 zombie samples that are all a mix of these 2 components in different proportions, below their coordinates:

Zombie10 contains 10% of Neolithic and 90% of Beaker, Zombie20 contains 20% of Neolithic and 80% of Beaker and so on.

Zombie10,0.1282644,0.1280245,0.0614928,0.0621371,0 .0309108,0.0187543,-7.54e-05,0.0028576,-0.0008616,-0.0025406,-0.0046762,0.0040887,-0.0103563,-0.0149413,0.021733,0.0095089,-0.0040773,-0.0016938,0.0010375,0.0064098,0.006462,0.0025317,-0.0014943,0.0026027,-0.0003298
Zombie20,0.128051,0.1331698,0.0602891,0.051034,0.0 372787,0.0132102,-0.0006501,0.0023692,0.0064833,0.0081763,-0.0042167,0.0052552,-0.0125918,-0.0152945,0.0182727,0.0090073,-0.0021079,-0.0011865,0.0008338,0.0055818,0.0070749,0.0023741,-0.0032044,-0.0009399,-0.0001955
Zombie30,0.1278376,0.1383151,0.0590855,0.0399308,0 .0436465,0.0076661,-0.0012249,0.0018808,0.0138283,0.0188933,-0.0037573,0.0064217,-0.0148273,-0.0156478,0.0148123,0.0085056,-0.0001386,-0.0006793,0.00063,0.0047538,0.0076879,0.0022164,-0.0049144,-0.0044826,-6.13e-05
Zombie40,0.1276242,0.1434604,0.0578818,0.0288277,0 .0500143,0.002122,-0.0017997,0.0013924,0.0211732,0.0296102,-0.0032979,0.0075882,-0.0170628,-0.0160011,0.0113519,0.008004,0.0018307,-0.000172,0.0004263,0.0039258,0.0083009,0.0020587,-0.0066245,-0.0080253,7.29e-05
Zombie50,0.1274108,0.1486057,0.0566782,0.0177246,0 .0563822,-0.0034221,-0.0023744,0.000904,0.0285182,0.0403272,-0.0028384,0.0087547,-0.0192982,-0.0163544,0.0078916,0.0075024,0.0038,0.0003353,0.0 002226,0.0030977,0.0089139,0.001901,-0.0083346,-0.011568,0.0002072
Zombie60,0.1271973,0.1537511,0.0554746,0.0066215,0 .06275,-0.0089663,-0.0029492,0.0004156,0.0358632,0.0510442,-0.002379,0.0099213,-0.0215337,-0.0167076,0.0044312,0.0070008,0.0057694,0.0008426, 1.89e-05,0.0022697,0.0095268,0.0017434,-0.0100447,-0.0151106,0.0003414

We then compared these samples to modern population, keeping the 5 closest population for each sample:

Distance to: Zombie60
0.02472206 Basque_Spanish
0.02967697 Spanish_La_Rioja
0.03090172 French_South
0.03103093 Basque_French
0.03616576 Spanish_Pais_Vasco

Distance to: Zombie50
0.02050831 French_South
0.02056286 Basque_Spanish
0.02396702 Basque_French
0.02884770 Spanish_Pais_Vasco
0.02888149 Spanish_Castello

Distance to: Zombie40
0.02453633 French_Occitanie
0.02737575 French_South
0.02830016 French_Auvergne
0.03238960 Basque_French
0.03246990 French_Pas-de-Calais

Distance to: Zombie30
0.02337488 French_Pas-de-Calais
0.02450708 French_Brittany
0.02747585 French_Nord
0.02749456 Belgian
0.02827739 French_Paris

Distance to: Zombie20
0.02093451 French_Brittany
0.02118766 Welsh
0.02141809 Orcadian
0.02305440 Scottish
0.02321544 English

Distance to: Zombie10
0.02379927 Icelandic
0.02480305 Orcadian
0.02533529 Norwegian
0.02614087 Irish
0.02632710 Scottish

2nde experiment

We ran a model made of these 2 averages and compare it to all the modern populations, then did some filtering of the results in Excel and only kept the populations having a distance < 0.04 from the model. My idea was to check which population would come up closer to the model first and in which proportions for each component.


Some Interesting points I noticed:

- Spanish Basques are roughly 50% Beaker / 50% Neo and the closest to the model
- All these populations seem for the most part a mix of these 2 populations
- British have on average between 15 and 20% Neolithic, Irish slightly less.
- French have between 20% and 50% of Neolithic ancestry (Basques), on average around 35%.

09-06-2020, 02:48 PM
It seems important to me that you know that this model (NL_BBC + FR_MN) actually works very badly. Examples (all the models here are calculated on Reich_1240K, qpfstats + qpAdm, under optimal conditions, that is to say by only retaining for the analysis the transversions, and each group represented by the individuals of maximum coverage):

Right pops for all models:


left pops:

best coefficients: 0.878 0.122
std. errors: 0.046 0.046

chisq tail prob 41.650 0.000254359


best coefficients: 0.775 0.225
std. errors: 0.043 0.043

chisq tail prob 39.245 0.000334213


best coefficients: 0.945 0.055
std. errors: 0.057 0.057

chisq tail prob 24.010 0.0456917

It is no doubt adventurous to dare to propose an explanation for this phenomenon before having seriously studied it, but one thing nevertheless seems clear. If many of us are now putting the Dutch BBC on the initiative of the BBC phenomenon, nobody imagines waves of Dutch BBs surging south and west (not to mention the North) to meet farmers from the Neolithic era, moreover all farmers of the same autosomal constitution. It is undoubtedly more reasonable to postulate a complex phenomenon of swarming from a Dutch center, from which the multiple poles of BBC culture derive by fusion with the indigenous populations, populations of which many studies, in particular the most recent , that of the Franco-German team of S. Brunel, showed the diversity. This diversity can essentially be summarized in the EEF / WHG balance. This problem of EEF / WHG balance is of course aggravated here by the fact that the source NL_BBC also contributes its share of EEF and WHG, in a proportion which cannot be modulated. In addition, it seems really difficult to imagine that all European groups allow themselves to be modeled properly with reference to the beginnings of the Bronze Age. That there was a certain continuity everywhere, no one doubts it. But this continuity is not exclusive of complexities, themselves resulting in disturbances of the autosomal composition. My conclusion at this point is that it will be difficult to find better than a three-reference model X + Anatolia_N + WHG, where X should ideally be chosen as close as possible to the most western groups of the Yamna culture. Nevertheless, it is clear that some binary models work much better than the proposed model:

left pops:

best coefficients: 0.468 0.532
std. errors: 0.025 0.025

chisq tail prob 20.636 0.111344

left pops:

best coefficients: 0.570 0.430
std. errors: 0.036 0.036

chisq tail prob 7.518 0.912896

The fact that the early poz81 CWC is almost 100% identical to the Yamnaya from Ukraine probably has a lot to do with this, since it partly addresses the balance problem I was talking about before. In the French model, replacing the GACs by France_MN brings down the adjustment characteristics. It is possible that some groups from Brunel's study (I am thinking of GRG or OBN) are better supplements than GACs, but that would need to be verified.

Now something that may surprise you. Look at these models, purely Scandinavian:


best coefficients: 0.983 0.017
std. errors: 0.083 0.083

chisq tail prob 2.611 0.999587

left pops:

best coefficients: 0.864 0.136
std. errors: 0.071 0.071
6.477 0.952995

left pops:

best coefficients: 0.755 0.245
std. errors: 0.067 0.067
12.996 0.526828

Don't search an explanation in the quality of the genomes involved:

Netherlands_BellBeaker_ref valid snps: 143823
Sweden_BAC valid snps: 167851

Sweden_Megalithic.SG_ans017 valid snps: 182021
France_MN valid snps: 158493

09-06-2020, 03:36 PM
Thanks for your suggestions, do you have have an idea which samples we could choose on the G25 to reproduce that kind of experimentation ?, Here my idea was to check which populations would come closer to these Zombie samples, and which populations would come closer to the selected model, and the results are still quite logic and close to what we predicted, it still gives a broad idea I think even if far from reality. I am perfectly aware there is no such continuity in all these countries and that they all experimented waves of migrations since the Bronze age, also aware that the interaction between bell beakers and Neolithic people must have been much more complex that this (we saw this with Ancestra) and that Bell Beakers themselves had their part of Neolithic ancestry too, but for the Basques maybe, we must be quite close to reality, don't you think ?

09-06-2020, 03:56 PM
The choice I made for the samples is only directed by the number of SNPs available, which is not a problem for you. As to reproduce these stats with G25+nMonte, allow me not to express once more my scepticism, I'm too scared to get lynched. :behindsofa: But if you want absolutely see what it gives, all the samples are in G25 (search Sweden_BattleAxe and Sweden_Megalithic, the Polish early CWC is poz81).