Locating Proto-Uralic

[Anglesqueville]

Siksi eteläkantasaamen leviäminen suoraan länteen Pohjanlahden yli Skandinaviaan näyttää hyvinkin mahdolliselta.

That's why the spread of Southern Sami directly westward across the Gulf of Bothnia to Scandinavia seems very possible.

This sentence particularly interests me.

[Æsir]

Siksi eteläkantasaamen leviäminen suoraan länteen Pohjanlahden yli Skandinaviaan näyttää hyvinkin mahdolliselta.

That's why the spread of Southern Sami directly westward across the Gulf of Bothnia to Scandinavia seems very possible.

This sentence particularly interests me.

[Kale]

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

There's not a whole lot there constraining the Northeast-Asian ancestry. Might I recommend adding something like Baikal_EN, DevilsCave, and/or Kra001. Kostenki, Goyet, and Vestonice are probably redundant also, since WHG-EHG are in the right also and any affinity to the former will be mediated by the latter. Yamnaya/Afanasievo might help too, there's enough drift in them to make them unique from EHG+CHG as far as right pops go.

[Anglesqueville]

^^ When I publish a model, there are often a few dozen behind it (like everyone else I imagine). A random example, adding for example samples from Siberia_Neolithic and Afanasievo to the right_list, to the left_list
Finland_IA.imputed_allentoft_234
Estonia_IA.ial
Russia_IronAge.ial
Russia_Bolshoy ...
changes the coefficients from 0.556/0.320/0.125 to 0.507/0.345/0.148, the tail prob from 0.75 to 0.63, without changing anything significant in the standard errors (which is the most unfortunate, but there was no hope). Same thing if I prune the right list, which only gives a slight gain in execution time. The question of the right list, complicated by the stories of rotating models, is a headache and often a waste of time.

[Jaska]

Æsir:

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

To the other Saami languages? No, because there are so often phonological inconsistencies, which point to parallel borrowing:

SaE [South Saami] klahtje < EKSa [Southern Proto-Saami] *klačče̮ ’paarma’  [‘horse-fly, deer-fly’]
~ LKSa [Northwestern Proto-Saami] *(s)lāvčā ’nautakiiliäinen’
← KSk *klaggjan ’paarma’


Anglesqueville

“That's why the spread of Southern Sami directly westward across the Gulf of Bothnia to Scandinavia seems very possible.”
This sentence particularly interests me.

Then I have good news for you! More about that in our recent second part:
Häkkinen, Jaakko & Piha, Minerva 2023: Kantasaamesta eteläkantasaameen, osa 2 : Äännehistorian todisteita eteläsaamen varhaisesta eriytymisestä.
[From Proto-Saami to Southern Proto-Saami, part 2 : Historical phonological evidence about an early divergence of South Saami.]
https://www.academia.edu/114961035/Kanta...sest%C3%A4

Here is a quickly corrected Google translation beginning from the page 24:

"5.2 Gävle Saami as a daughter language of Southern Proto-Saami

Is Gävle Saami a remnant of a very early unified Southern Proto-Saami speaking area that extended to the coast of southern Norland, or does it represent the language of Sami people who only later migrated from the interior to the coast and from north to south? Based on several pieces of evidence, we lean towards the first option.

First of all, the new taxonomy of the Saami languages that we have introduced, in which the Southern Proto-Saami has separated first while the other Saami languages still continue together, makes the spread that went around the northern region less likely than before. A significantly shorter route would go directly from Western Finland across the Gulf of Bothnia to the coast of Norland.

Secondly, based on Lars-Gunnar Larsson's lexical analysis, the preform of Gävle Saami must have been very south early on. The main proof of this is the plant name ‹Gosen milkie grasi› ‘Ballota nigra’ (literally “cow’s-milk-plant”). The plant grows in Southeastern Sweden, and the northern limit of its occurrence runs around Västmanland and Gästrikland. The regions further north or inland no longer correspond to the natural-geographic picture that can be obtained from the word list, as Larsson has argued (Larsson 2005; 2018, 211–213).

All parts of the name ‹Gosen milkie grasi› are ancient Proto-Scandinavian loanwords, the age of which is evidenced by their widespread distribution and regular representation in the Saami languages. However, such common loanwords do not prove the long local continuity of Gävle Saami, but the chronological value of the name lies in the fact that a plant from so southern region overall has an old name in Gävle Saami. If the language had spread to the area only later, the name of the plant would most probably have been borrowed from the Swedes who already lived in the area (in Swedish bosyska).

Thirdly, Gävle Saami seems to have ancient and perhaps even Proto-Scandinavian loanwords that do not exist in other Saami languages or that, based on phonetic features and differences in meaning, point to borrowing separately from the other languages. These words are difficult to explain in any way other than assuming the long local continuity of this language.

SaG ‹eiker lost› ’oak leaf’ < EKSa *äjke̮®
← OSc eik ’tammi’ > Sw ek (cf. Saami *(h)ājkke̮ ‘old tree’ ← PSc *aik(u) ‘oak’

SaG ‹londen lost› ’maple leaf’ < EKSa *lun-?
← PSc *hluni- ‘maple’ or OSc hlynr > Sw lönn

SaG ‹nuttå› ’hazelnut(bush)’ < EKSa *nuttɔ̄
← PSc *hnutu ‘hazelnut’ or OSc hnot > Sw nöt

These words require even more detailed analysis, but it is already possible to rule out a late borrowing from Swedish on the phonological basis. These three apparently old loanwords seem to indicate a long separate development of Gävle Saami in the southeast from the current Saami region, because the oak (Quercus robur), maple (Acer platanoides), and hazelnut (Corylus avellana) only grow in the coastal lowlands of Southern Norland. Old Scandinavian loanwords would be approximately 1000 years old, and Proto-Scandinavian loanwords more than 1500 years old.

It seems that the early form of Gävle Saami formed a regionally separate contact surface against the Scandinavians. Preliminary evidence shows that Gävle Saami did not spread to the region only later from the north, but represents a long-term local continuity, thus testifying to the Southern Proto-Saami speaking area that covered the whole of Central Scandinavia, possibly already in the first half of the first millennium, but at the latest at the turn of the second millennium."

[Kale]

The question of the right list, complicated by the stories of rotating models, is a headache and often a waste of time.

In what way is the right list a headache? Maybe my unconventional perspective derived from having stared at a spreadsheet of outgroup F3 stats for eons has blinded me to the views of others on this matter?

[Æsir]

Æsir:

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

To the other Saami languages? No, because there are so often phonological inconsistencies, which point to parallel borrowing:

SaE [South Saami] klahtje < EKSa [Southern Proto-Saami] *klačče̮ ’paarma’  [‘horse-fly, deer-fly’]
~ LKSa [Northwestern Proto-Saami] *(s)lāvčā ’nautakiiliäinen’
← KSk *klaggjan ’paarma’

The inconsistencies could not be a result of the paleo languages of the borrowing communities?

[Anglesqueville]

The question of the right list, complicated by the stories of rotating models, is a headache and often a waste of time.

In what way is the right list a headache? Maybe my unconventional perspective derived from having stared at a spreadsheet of outgroup F3 stats for eons has blinded me to the views of others on this matter?

I wrote a long post about this on AG, it may still be accessible in the Genoplot archives. When you read Patterson's founding text things seem clear. But as soon as you look at the examples he gives (and at the time I looked at them blindingly) the clarity disappears. The more you examine the concrete cases of use of qpAdm, in particular when you compare the right lists used by professionals and those chosen by bloggers, it becomes hellish, at least for a mathematician like me who always dreams of perfectly defined rules. A few years ago I spent a lot of time working on D-stat matrices from qpAdm tours differing only in their right lists, using various mathematical tools and approaches, in the hope of being able to come up with something as a minimal philosophy, rules of good practice. I gave up.

[Jaska]

Æsir:

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

To the other Saami languages? No, because there are so often phonological inconsistencies, which point to parallel borrowing:

SaE [South Saami] klahtje < EKSa [Southern Proto-Saami] *klačče̮ ’paarma’  [‘horse-fly, deer-fly’]
~ LKSa [Northwestern Proto-Saami] *(s)lāvčā ’nautakiiliäinen’
← KSk *klaggjan ’paarma’

The inconsistencies could not be a result of the paleo languages of the borrowing communities?

Well, it is theoretically possible. But that, what we know about the ancient unknown languages in Fennoscandia based on the loanwords and placenames in the Saami languages, does not explain these kind of irregularities. Therefore there is no need to assume any mediator language, and these can very well be results of parallel borrowings showing different sound substitutions in different Saami proto-dialects.

[Kale]

The question of the right list, complicated by the stories of rotating models, is a headache and often a waste of time.

In what way is the right list a headache? Maybe my unconventional perspective derived from having stared at a spreadsheet of outgroup F3 stats for eons has blinded me to the views of others on this matter?

I wrote a long post about this on AG, it may still be accessible in the Genoplot archives. When you read Patterson's founding text things seem clear. But as soon as you look at the examples he gives (and at the time I looked at them blindingly) the clarity disappears. The more you examine the concrete cases of use of qpAdm, in particular when you compare the right lists used by professionals and those chosen by bloggers, it becomes hellish, at least for a mathematician like me who always dreams of perfectly defined rules. A few years ago I spent a lot of time working on D-stat matrices from qpAdm tours differing only in their right lists, using various mathematical tools and approaches, in the hope of being able to come up with something as a minimal philosophy, rules of good practice. I gave up.

Don't fall into 'analysis paralysis'. There's no hard and fast rules, but it's not an unnavigable labyrinth.
First compile your right pop candidates. A right pop must have significant/substantial unique drift. Such as if you take a list of outgroup F3 stats, other members of the pop should be top of the list (F3: Mbuti Gravettian_KremsVestonice X)
Gravettian_Italy 0.321814
Sunghir.SG 0.304920
Italy_SanTeodoro_HG 0.300779
Gravettian_France_Ormesson 0.300235
...Or members of the pop should be significantly more related to each other than the sum of their components.
(F3: Mbuti Afanasievo_KarasukIII.SG X)
Anatolia_Barcin_N 0.275555
CHG.SG 0.280964
Ukraine_Dereivka_NHG 0.296192
RUS_Vologda_Minino_HG 0.300776
Yamnaya_Samara 0.300572

There are a finite number of major population groups like that. Worldwide among ancient samples I doubt we're dealing with more than about 100. From there it's about narrowing the list to the most appropriate with a few considerations.
- There's 99.99% chance that for example a Neolithic Moroccan will not need to be discerned for specifically Jomon or Amazonian ancestry.
- No need to include right pops whose affinity will be mediated by other right pops, for example there is no way an Iron Age European's affinity (or lack thereof) to Sunghir would be altered by some source without effecting even more it's affinity to more temporally proximate right pops, making Sunghir redundant. If Sunghir is not an asset it's a type-II error liability.
- Start general and get more specific if needed. If a sample shows no evidence of WHG ancestry, there's no need to crowd the right with a bunch of WHG pops, one is sufficient to show an account for the possibility.

Qpadm requires a bit of knowledge about the test sample, either from archaeology to select probable sources, or about the genetics of it if you are unsure about what the sources are, as well as knowledge of the major phylogenic points in aDNA. Consequently I usually explore a sample with other methods first, and use Qpadm last as a confirmation rather than a test.

[Æsir]

Æsir:

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

To the other Saami languages? No, because there are so often phonological inconsistencies, which point to parallel borrowing:

SaE [South Saami] klahtje < EKSa [Southern Proto-Saami] *klačče̮ ’paarma’  [‘horse-fly, deer-fly’]
~ LKSa [Northwestern Proto-Saami] *(s)lāvčā ’nautakiiliäinen’
← KSk *klaggjan ’paarma’

The inconsistencies could not be a result of the paleo languages of the borrowing communities?

Well, it is theoretically possible. But that, what we know about the ancient unknown languages in Fennoscandia based on the loanwords and placenames in the Saami languages, does not explain these kind of irregularities. Therefore there is no need to assume any mediator language, and these can very well be results of parallel borrowings showing different sound substitutions in different Saami proto-dialects.

What is the latest on the paleo languages? 2 or more? Geography?

[Jaska]

What is the latest on the paleo languages? 2 or more? Geography?

Sorry, I do not understand what the question is. Did you drop some words?

[kolompar]

Saami and Levanluhta form a clade against almost everything, you should be able to model Saami as 100% Levanluhta, with maybe a little recent admixture from their neighbours.

I'm not an expert with qpAdm but I could model Levanluhta with Bolshoy in the right.

Code:

> qpadm("aadr_v54.1.p1_1240K_public", c("Finland_Levanluhta_B", "Russia_Krasnoyarsk_BA.SG", "Russia_Karelia_HG"), c("Mbuti.DG", "Turkey_N", "Russia_Samara_EBA_Yamnaya", "Italy_North_Villabruna_HG", "Russia_Shamanka_Eneolithic.SG", "Russia_Bolshoy", "Kazakhstan_Botai_Eneolithic.SG", "Russia_Arkhangelsk_Veretye_Mesolithic.SG", "Ukraine_Mesolithic", "China_AmurRiver_Mesolithic", "Iran_GanjDareh_N", "Sweden_Motala_HG"), "Finland_Levanluhta")
i Reading metadata...
i Computing block lengths for 1150639 SNPs...
i Number of SNPs after excluding those with missing data: 108384
i Computing admixture weights...
i Computing standard errors...
i Computing number of admixture waves...
$weights
# A tibble: 3 x 5
  target             left                     weight     se     z
  <chr>              <chr>                     <dbl>  <dbl> <dbl>
1 Finland_Levanluhta Finland_Levanluhta_B      0.629 0.0358 17.6
2 Finland_Levanluhta Russia_Krasnoyarsk_BA.SG  0.231 0.0167 13.8
3 Finland_Levanluhta Russia_Karelia_HG         0.141 0.0348  4.03

$rankdrop
# A tibble: 3 x 7
  f4rank   dof  chisq         p dofdiff chisqdiff   p_nested
   <int> <int>  <dbl>     <dbl>   <int>     <dbl>      <dbl>
1      2     9   11.5 2.45e-  1      11      636.  3.35e-129
2      1    20  647.  3.38e-124      13     3075.  0       
3      0    33 3722.  0              NA       NA  NA       

Same thing happens with G25, Levanluhta will prefer Krasnoyarsk to Bolshoy (except for the one kra001 shifted individual sometimes).
With all averages:

Code:

Target: Finland_Levanluhta
Distance: 2.8093% / 0.02809265 | R3P
43.0    Russia_Mezhovskaya.SG
37.8    Lithuania_LN_o
19.2    Russia_Krasnoyarsk_BA.SG

Code:

Target: Finland_Levanluhta
Distance: 2.3815% / 0.02381515 | R4P
36.8    Sweden_IA_2.SG
25.6    Russia_Krasnoyarsk_BA.SG
19.0    Estonia_BA.SG
18.6    Russia_Vologda_Veretye_Mesolithic.SG

Individuals:

Code:

Target: Finland_Levanluhta
Distance: 2.0311% / 0.02031105 | R3P
25.0    Sweden_Viking.SG:VK395_noUDG.SG
24.0    Russia_Krasnoyarsk_BA.SG:kra001_noUDG.SG
18.6    Russia_Vologda_Veretye_Mesolithic.SG:KAR001_noUDG.SG
16.8    Sweden_Viking.SG:VK406_noUDG.SG
9.4    Sweden_Viking.SG:VK108_noUDG.SG
3.4    Sweden_Viking.SG:VK432_noUDG.SG
1.8    Sweden_Viking.SG:VK452_noUDG.SG
1.0    Sweden_Viking.SG:VK56_noUDG.SG

Bolshoy have this Ugric-specific WSHG/Yeniseian pull. Early Magyar adventurers, most probably.

Code:

Target: Russia_Bolshoy
Distance: 1.5445% / 0.01544504 | R3P
40.2    Russia_Karelia_HG
32.2    Russia_KusnarenkovoKarajakupovo_Medieval.SG
27.6    Russia_Krasnoyarsk_BA.SG

[Jaska]

Kolompar

Saami and Levanluhta form a clade against almost everything, you should be able to model Saami as 100% Levanluhta, with maybe a little recent admixture from their neighbours.

Thank you, interesting.

Kolompar

I'm not an expert with qpAdm but I could model Levanluhta with Bolshoy in the right.

OK. Zeng et al have no Levänluhta, but BOO has very little in common with the Saami in their qpAdm: from the first to the latter, EHG would have been reduced from 55 % to <10 %, while the Siberian ancestry only would have halved.

Kolompar

Bolshoy have this Ugric-specific WSHG/Yeniseian pull. Early Magyar adventurers, most probably.

There is no point to use later linguistic labels at that time depth. During the time when the North(west)ern Siberians moved to the Kola Peninsula ca. 2000 BCE, the Uralic speakers had not yet spread in Northwest Siberia. Sure, the migrants to Kola had West Siberian ancestry, but ethno-linguistically they were not yet the same folk as the modern Uralic-speaking West Siberians. Spatially, the label “Magyar” is not relevant anywhere north from the forest steppe in the Southern Ural Region.

[Jaska]

Here is an important point:

The result that the Siberian ancestry in the Uralic populations can be explained almost exclusively by the Yakutia ancestry does not agree with the Siberian homeland for Proto-Uralic. Here are the different Siberian ancestries in the Uralic speaking populations according to the best qpAdm results by Zeng et al. 2023:

1. The Nganasans: the Yakutia ancestry, the Mongolia ancestry, the Tyumen_HG ancestry.
2. The Enets: the Yakutia ancestry, the Mongolia ancestry.
3. All the rest: the Yakutia ancestry.

If Proto-Uralic was spreading from Southern Central Siberia, the westward migrating people should have acquired other Siberian ancestries on their way toward the west (at least the Tyumen_HG ancestry and the Altai ancestry). On the other hand, the Nganasans should have no other Siberian ancestries than the Yakutia ancestry, if they had remained in the ancient Uralic homeland in Northern Central Siberia. In reality none but the two easternmost Uralic populations have several Siberian ancestries.

If Proto-Uralic spread from the Ural Region or even from the more western location, then we should expect the situation which we see now: in the European and in the Trans-Urals Uralic populations there is only the Yakutia ancestry (which alone has spread across the Urals to the west, until only recently other ancestries have followed), and only in the easternmost Uralic populations there are several Siberian ancestries (because these populations have met several Siberian populations on their way).

The real situation does not agree with the Siberian homeland for Proto-Uralic.