Locating Proto-Uralic

[Rober_tce]

The thread is re opened.

Remember all users stay on topic. Thank you all.

[Jaska]

The oldest Uralic region is the southern part of the Taiga zone. We also know based on the linguistic evidence that until ~200 CE, the Saami language was spoken only in Southern Finland. Therefore, any migrations in the northernmost zone of Europe before that (including the Siberian migration to the Kola Peninsula ~2000 BCE) had nothing to do with the Saami language. This is the scientific framework we all must accept.

Of course some people may call some other populations "Saami", but it does not make sense, if that population did not speak Saami and did not call themselves Saami. "Saaminess" is not a genetically determined characteristic, and therefore "Saaminess" cannot be seen from the DNA. More about that here in pages 458-460:
https://www.academia.edu/114870366/Comme...hnogenesis
(Piha, Heikkilä & Häkkinen 2023: Comment on the article Archaeology, Language, and the Question of Sámi Ethnogenesis)

[Anglesqueville]

It is quite ironic to find in the text of the two Norwegians many of the flaws which have spoiled the discussion on the various threads devoted to proto-Germanic. As you notice, they apply the adjective "Saami" to everything and anything, which allows them to see the Saamis or their ancestors in archaeological or genetic data (the latter interpreted very crudely). However, if we leave aside a claim of identity (which only has meaning today), such an adjective only has linguistic meaning, to designate a family of languages.
Concerning the Kola_BA group, I agree that the question of its possible involvement in the constitution of the Saami gene pool does not correlate with that of the propagation of the Saami languages in Finland and Scandinavia. I don't know if it can have one with the non-Uralic substrate. I doubt it, but I don't have a really convincing argument to make. That said, this question, approached solely from the angle of autosomal genetics, is not as obvious as some people want to make it out to be. The problem with Saamis is the abundance of different models, all mathematically satisfactory, without BOO and with BOO. I can write a post on this if it is of interest to this thread.

[Jaska]

That said, this question, approached solely from the angle of autosomal genetics, is not as obvious as some people want to make it out to be. The problem with Saamis is the abundance of different models, all mathematically satisfactory, without BOO and with BOO. I can write a post on this if it is of interest to this thread.

Yes, it would be very interesting.

[Anglesqueville]

About Levänluhta_IA (DA234 and DA238). I use the data imputed by Allentoft with GLIMPSE. Allentoft, with ADMIXTURE supervised gets for these two individuals the following models:
Individual Scandinavia_4000BP_3000BP Estonia_3000BP_2500BP EurasiaN_2000BP_1000BP
DA234 0,10 0,61 0,29
DA238 0,00 0,48 0,52

EurasiaN_2000BP_1000BP is a cluster that contains (among others) two individuals from BOO and the Minino_IA individual NEO538 (Russia_IronAge.ial in my files). I wondered whether qpAdm would give similar results and could discriminate between BOO and Minino.
First, using Falköping_N as pre-Germanic source, I don't find any Germanic affinity in any of the two Levänluhta, despite the uncontroversial linguistic Germanic influx in Saami. For the rest, two very satisfactory models:

left pops:
Finland_IA.imputed_allentoft_238
Estonia_IA.ial
Russia_IronAge.ial


best coefficients: 0.214 0.786
totmean: 0.214 0.786
boot mean: 0.212 0.788
std. errors: 0.061 0.061


fixed pat wt dof chisq tail prob
00 0 13 9.841 0.706905 0.214 0.786
01 1 14 239.282 0 1.000 0.000
10 1 14 20.586 0.112725 0.000 1.000
best pat: 00 0.706905 - -
best pat: 10 0.112725 chi(nested): 10.745 p-value for nested model: 0.0010455

left pops:
Finland_IA.imputed_allentoft_234
Estonia_IA.ial
Russia_IronAge.ial


best coefficients: 0.543 0.457
totmean: 0.543 0.457
boot mean: 0.542 0.458
std. errors: 0.052 0.052


fixed pat wt dof chisq tail prob
00 0 13 9.846 0.706483 0.543 0.457
01 1 14 96.383 2.33051e-14 1.000 0.000
10 1 14 77.869 7.01677e-11 0.000 1.000
best pat: 00 0.706483 - -
best pat: 10 7.01677e-11 chi(nested): 68.023 p-value for nested model: 1.61619e-16


If I replace Estonia_IA with VolgaOka_IA (imputed by me with BEAGLE5) I get very similar models:

left pops:
Finland_IA.imputed_allentoft_238
volga_oka_IA_imput
Russia_IronAge.ial


best coefficients: 0.290 0.710
totmean: 0.290 0.710
boot mean: 0.286 0.714
std. errors: 0.089 0.089

fixed pat wt dof chisq tail prob
00 0 13 10.621 0.64252 0.290 0.710
01 1 14 101.191 2.80297e-15 1.000 0.000
10 1 14 20.577 0.112977 0.000 1.000
best pat: 00 0.64252 - -
best pat: 10 0.112977 chi(nested): 9.956 p-value for nested model: 0.00160351

left pops:
Finland_IA.imputed_allentoft_234
volga_oka_IA_imput
Russia_IronAge.ial


best coefficients: 0.757 0.243
totmean: 0.757 0.243
boot mean: 0.756 0.244
std. errors: 0.077 0.077

fixed pat wt dof chisq tail prob
00 0 13 11.178 0.595909 0.757 0.243
01 1 14 22.585 0.0673728 1.000 0.000
10 1 14 77.863 7.03253e-11 0.000 1.000
best pat: 00 0.595909 - -
best pat: 01 0.0673728 chi(nested): 11.407 p-value for nested model: 0.000731816

Now what happens when I inject BOO in the game? Let's take the example of DA234. I used Estonia _IA and Allentoft's BOO (NEO60 to 62)
left pops:
Finland_IA.imputed_allentoft_234
Estonia_IA.ial
Russia_IronAge.ial
BOO.ial


best coefficients: 0.550 0.316 0.134
totmean: 0.550 0.316 0.134
boot mean: 0.560 0.250 0.190
std. errors: 0.413 2.695 2.290

fixed pat wt dof chisq tail prob
000 0 12 8.917 0.709969 0.550 0.316 0.134
001 1 13 10.166 0.680267 0.540 0.460 0.000
010 1 13 13.917 0.379709 0.600 0.000 0.400
100 1 13 37.727 0.000318392 0.000 2.802 -1.802 infeasible
011 2 14 96.671 2.05401e-14 1.000 0.000 0.000
101 2 14 77.852 7.06779e-11 0.000 1.000 0.000
110 2 14 170.677 4.9973e-29 0.000 0.000 1.000
best pat: 000 0.709969 - -
best pat: 001 0.680267 chi(nested): 1.249 p-value for nested model: 0.263727
best pat: 101 7.06779e-11 chi(nested): 67.685 p-value for nested model: 1.91802e-16

You see that the slight gain in tail prob is obtained at the cost of an explosion of standard errors. This makes this model very unpleasant, and in my opinion, such a model could hardly be used as an argument in favour of BOO participation in the constitution of the DA234 gene pool.
That said, there is something even more unpleasant. The idea that the genetics of the Saami would have drawn from two sources, one being Germanic, the other being BOO (which therefore becomes the origin of the Saami linguistic family in this part of the world), is linguistically burlesque. It turns out that qpAdm could very easily be called upon to support this thesis:
left pops:
Finland_IA.imputed_allentoft_238
Sweden_N_Falköping.imputed_Allentoft
BOO.ial

best coefficients: 0.302 0.698
totmean: 0.302 0.698
boot mean: 0.302 0.698
std. errors: 0.039 0.039


fixed pat wt dof chisq tail prob
00 0 13 10.900 0.619177 0.302 0.698
01 1 14 335.027 0 1.000 0.000
10 1 14 65.007 1.52518e-08 0.000 1.000
best pat: 00 0.619177 - -
best pat: 10 1.52518e-08 chi(nested): 54.107 p-value for nested model: 1.89847e-13



left pops:
Finland_IA.imputed_allentoft_234
Sweden_N_Falköping.imputed_Allentoft
BOO.ial


best coefficients: 0.533 0.467
totmean: 0.533 0.467
boot mean: 0.533 0.467
std. errors: 0.038 0.038


fixed pat wt dof chisq tail prob
00 0 13 7.882 0.851208 0.533 0.467
01 1 14 156.253 4.01605e-26 1.000 0.000
10 1 14 168.770 1.21308e-28 0.000 1.000
best pat: 00 0.851208 - -
best pat: 01 4.01605e-26 chi(nested): 148.371 p-value for nested model: 3.93592e-34

This shows that a qpAdm model can very well be mathematically perfectly satisfactory and have no link with reality.

[Jaska]

Thank you, Anglesqueville!
To complicate things more, it is not even certain that the Levänluhta population represents the ethno-linguistic Saami. After all, these remains are dated between 300-800 CE, the early end being very close to the expansion of Late Proto-Saami ca. 200 CE, originating in the Southernmost Finland. It would be doubtful whether the ethno-linguistic turnover would have been completed so quickly in so wide area. The spread of Kjelmöy Ceramics from the north to the south in Finland could have spread together with the BOO-related ancestry.

According to the qpAdm results by Peltola et al. 2023, the Levänluhta population could descend partly from the BOO population, partly from some Farmer ancestry rich population (in Southern Sweden such are the Megalithic Culture population and the Funnel Beaker Culture population). If we should guess the languages related to the three relevant ancestries, one could throw e.g. the following guesses:

1. The EHG ancestry: an unknown Paleo-European language of the Comb Ceramic Culture.
2. The Siberian ancestry: an unknown Paleo-Siberian language from the Late Neolithic Yakutia, or perhaps even some Yukaghir-related language.
3. The Farmer ancestry: an unknown Paleo-European/Paleo-Near-Eastern language, or perhaps even some Vasconic or Semitic language, as has also been proposed.

It could be well possible that there were not yet Indo-European or Uralic languages present north from Southern Finland in the Early Iron Age. But still, the modern Saami and Finns derive at least half of their ancestry from the population of Corded Ware Cultures, and these cultures arrived in Southwestern Finland already ca. 2800 BCE. So my next question is, can you try how some Corded Ware sample (from Estonia, perhaps) acts within the left populations?

Peltola et al. 2023 rejected the CWC-related ancestry for both the BOO and the Levänluhta populations, but as we have seen, every study gives different results. Clearly there are myriads of mathematically fitting admixtures available by the qpAdm method, and as you stated, the correct one of these probably cannot be picked up only by staring at the numerical values of the model, but an educated guess by a geneticist is also needed.

[Anglesqueville]

For all of this, I can't wait for Childebayeva-Haak et al's data to finally be published. I go to edmond every day, but so far only their demographic file is available.

[Anglesqueville]

^^ I have the greatest distrust of Peltola's models. That a genetic affinity with EEF (Peltola uses LBK_EN) could have survived in a group, for such a long time, at such a high rate, and in this region, seems to me a completely mind-blowing hypothesis. Furthermore, even though I keep in mind that I am only a modest amateur, I am unable to reproduce these models. On the contrary, the models obtained by Allentoft with other programs (ADMIXTURE in supervised mode) are clearly in line with those I obtained with qpAdm. But let's leave Levänluhta aside for now. When I have some time I will take stock of the modern Saamis

[Kale]

Qpadm models

What are the right pops in those models?

[Anglesqueville]

Mine:

right pops:
Russia_Ust_Ishim.DG
Cameroon_SMA.DG
Italy_North_Villabruna_HG
Czech_Vestonice16
Belgium_UP_GoyetQ116_1
Russia_MA1_HG.SG
Iran_GanjDareh_N
Kazakhstan_Botai_Eneolithic.SG
Russia_Kostenki14.SG
Ukraine_Mesolithic
Indian_GreatAndaman_100BP.SG
Russia_Arkhangelsk_Veretye_Mesolithic.SG
Israel_PPNB
Georgia_Kotias.SG
Turkey_N_I0707

Peltola's:

Ethiopia_4500BP.SG, CHG, Raqefet_M_Natufian, Onge, Villabruna, ANE, Mixe, Sweden_HG_Motala

[Jaska]

How about those populations which are theoretically possible sources in Northern Europe, like Motala and Veretye? Are they distant enough to be OK when placed within the right populations? Or does the result change, if they are instead included within the left populations?

[Æsir]

What is the latest available Saami language tree with datings?

[Jaska]

What is the latest available Saami language tree with datings?

What do you mean by "latest available"? The most recent presented dating? Or the latest possible dating? Also "language tree" is here difficult to understand.

For Late Proto-Saami, the most recent latest dating based on Northwest Germanic and Proto-Scandinavian loanword layers is ca. 200 CE. 
https://journal.fi/sananjalka/article/view/95727
(Page 117)

[Æsir]

What is the latest available Saami language tree with datings?

What do you mean by "latest available"? The most recent presented dating? Or the latest possible dating? Also "language tree" is here difficult to understand.

For Late Proto-Saami, the most recent latest dating based on Northwest Germanic and Proto-Scandinavian loanword layers is ca. 200 CE. 
https://journal.fi/sananjalka/article/view/95727
(Page 117)

Sorry, rally english.

The most current available presentation of the Saami languages, their relationship to each other and date of birth estimates.
Info-graphics are just nice way to represent data.

But I really wanted to ask about Southern Saami especially and you answered and delivered on that.

[Æsir]

Toisaalta relevantti kysymys on myös, miksi germaanista on lainautunut niin paljon
samaa sanastoa eteläsaameen kuin muihin saamelaiskieliin, jos se kerran on lainautu-
nut erikseen kumpaankin. Todennäköinen selitys on, että skandinaavit ovat levittäneet
samoja innovaatioita eri puolilla saamelaisten asuttamia alueita, joten tarve innovaatioi-
den nimeämiseen on ollut joka puolella jotakuinkin sama.
Paikannimistön perusteella länsikantasaame näyttää syntyneen jo Länsi-Suomessa
(Heikkilä 2012). Koska eteläkantasaame on eronnut aikaisemmin kuin muut kantamurteet,
senkin on täytynyt alkaa eriytyä Länsi-Suomessa. Siksi eteläkantasaamen leviäminen suo-
raan länteen Pohjanlahden yli Skandinaviaan näyttää hyvinkin mahdolliselta. Leviämisreitin
selvittely vaatisi kuitenkin oman tutkimuksensa; tässä artikkelissa keskitymme ajoitukseen.

It is not possible that most loans would come trough Southern Saami?