The Great Macrofamily thread: Indo-Uralic, Altaic, Eurasiatic, Nostratic etc.

[Moose-tache]

I agree with a lot of your points, Sal, including the broad points about openness to criticism. But nobody has ever said "Sorry for being honest" without looking like a tosser. Yes, Weepingelf's comments got a little too personal, but your response isn't even trying to diffuse that, and actually added more personal attacks to the conversation. If Weepingelf wanted to end the feud and get back on topic, how could they even do it? They're not going to prostrate themselves and beg for forgiveness, especially when they believe you will condescendingly gloat about it, or maybe not even let it go at all.

We all feel like we're surrounded by mean dumb-dumbs on the internet sometimes. The hard part is keeping our heads and being civil, even if you can't make people come to Jesus about Macrofamilies. Hell, the number of times I've bitten my tongue when people start to talk about how their language is "the oldest in the world" ought to have given me a fully perforated tongue by now.

[Howl]

Because, in the end, if you don't acknowledge the validity of criticism, how can you hope to make "progress" in your "research"? If it's not OK to point out the implausibility of one idea, how will you ever improve your ideas? If you genuinely want to link two languages together, then you shouldn't leap to the easiest, least unfalsifiable approach, one that reshuffles the pack of one language so dramatically in order to fit another - because if that's your first line of attack, what evidence could ever disprove it? What evidence could ever cause you to reevaluate that theory, when you can retool that methodology to turn anything into anything?

If you actually want to make progress, begin with an open mind. Identify problems, don't just assume whatever solution Old Albic prefers. If, for example, Indo-Uralic is valid, then the fact that one language has a series that the other doesn't have is a problem, with several possible solutions - but instead, you state as fact, but without any stated reason or evidence, that "the phonological change", the solution is of a particular sort. Why? That's not reasoning about macrofamily relationships, that's macrofamily fan-fiction!

I may not agree with Sal about the likelihood of Indo-Uralic, but I totally agree with what he states here!

[WeepingElf]

Everything is alright, Salmoneus. I think we just have misunderstood each other, and believe that we can sort it out. First of all, I apologize for likening your behaviour to that of a plonkster, and for summarizing your position as ignoramus et ignorabimus ad aeternum. That was my impression from your posts, but now I think this impression was wrong. You are simply deeply sceptical about any paleolinguistic speculations, and that is of course a legitimate position. As I already wrote, your objections are valid.

This brings me to your first accusation, that I only wish to discuss my ideas with people who agree with me. This is not the case. Different opinions are essential to a fruitful and productive discussion. But there is something such a discussion requires, and that is a common ground of some sorts. If there is no common ground, the discussion will be fruitless, like trying to discuss evolutionary biology with a creationist. In paleolinguistics, the common ground would be that there once were languages that are lost in history (a truism which every linguist would agree with, I think), and that it is possible to catch glimpses of those lost languages (which not every linguist agrees with, but without this assumption, paleolinguistic discussions become meaningless). And I felt that with your radical scepticism, you had left this common ground here. It is not helpful to say, "No, you are probably wrong" each time someone attempts a hypothesis regardless of the contents of that hypothesis. But it is perhaps good to have a "plonkster" here who reminds us all of the fact that our speculations are just speculations from time to time. Helps us stay on the floor of facts and avoid unsound flights of fancy.

Your second accusation is that I mix fact and fiction in my hypothesis. While I don't claim that I never did and can't pledge that I'll never do, I try to avoid it. I should have worded my post about the hypothetical Pre-PIE phoneme inventory restructurings differently. The way I wrote it, it could be misinterpreted as a statement of a fact, which it was not meant as. Of course the "Great Vowel Collapse" and the associated changes are only a hypothesis of mine which may be utterly wrong! Also, I am aware of the dangers you face if you do, like me, both research into lost prehistoric languages and make conlangs based on the outcomes of that research. One must be very careful to keep the things one has "found out" about the conlangs out of the train of thought that underlies the research.

In fact, I have given up the hypothesis Old Albic was originally based on long ago. That was that the Elves, via the Bell Beaker people, descended from the Linearbandkeramik (LBK) people, whom I assumed to be relatives of the PIE speakers on the ground that both descend from refugees from the Black Sea Flood. However, it turned out that LBK was genetically and archaeologically unrelated to either Bell Beaker or Yamnaya, and that there probably was no Black Sea Flood at all, at least not at a time convenient to such a hypothesis. So I dropped this idea.

I came up instead with the "Aquan" idea, according to which a sister language of PIE, branching off before the rise of ablaut, was spoken in Late Neolithic Central Europe, and that the Elves descend from those people, again via Bell Beaker. Meanwhile, I have grown sceptical of even that! If there ever was an "Aquan" language, it may just have been a third prong on the Anatolian vs. Classical IE fork in which ablaut was levelled out by a combination of sound change and analogy. Yet, the evidence of the mere existence of "Aquan" is so tenuous (with the Old European Hydronymy perhaps just being a linguistic equivalent of ley lines) that it is perhaps best to drop it, at least when it comes to theorizing about real languages. Using such an idea in conlangs is of course another matter

[Howl]

@WeepingElf: I was hesitating about replying to your post, because for me it is mostly just a rehash of old dogmas that I find unhelpful in linking IE to Uralic. There are more possibilities than the scenario you posted. Salmoneus already pointed them out. And the only way to find out what happened is to do real research.

For one thing, I don't think the vowel collapse happened before Indo-Uralic, if it ever happened. PIE has at least *i, *u, *o, *e. Probably it also had *ə in certain zero-grade positions. And maybe even *ē, *a and *ā, though those are more controversial. You like to compare Uralic to the Siberian languages. Those languages also have smaller vowel inventories than Uralic.

Another issue is PIE's ablaut. One hypothesis of mine is that Proto-Uralic also had ablaut. Then the ablaut in Ob-Ugric would be a retention, instead of an independent innovation. Ablaut in Proto-Uralic goes against the current dogma of Uralicists, but I am definitely not the first person who is thinking about it. Steinitz had already proposed something similar some 70 years ago. And even in the current reconstruction of Proto-Uralic, I can find cases like PU *kala 'fish, fish net' ~ PU *kältä 'to fish with a net' ~ PU *kulta 'to fish with a net'.

With regards to the stop system, Tocharian also had a major reduction. So it would not be unprecedented to propose something similar for Uralic. On the other hand, there is the possibility that PIE enlarged its stop system with a consonant-gradation-like mechanism.

I don't understand what you mean with 'Mitian palatals'. Uralic has *ś, *š and *č. I think ś came from s in certain consonant clusters. For me, *š would be a second *s in Indo-Uralic, but Uralicist Ante Aikio makes a good case that *š is a loan-word phoneme [The Finnic ‘secondary e-stems’ and Proto-Uralic vocalism, Ante Aikio, 2015]. And I have a hunch that Uralic *č may also be a loan-word phoneme.

[WeepingElf]

Fair. There are many possibilities. If IE and Uralic are related, then something must have happened in one or both of them in order to explain why the phonologies of the two are so different. We do not know which of the two is the more conservative, and surely, Proto-Uralic is not "essentially unchanged" Proto-Indo-Uralic, though it seems to me as if it was closer to the other "Mitian" languages (Turkic, Mongolic, Tungusic, Yukaghir, Chukotko-Kamchtkan, Eskimo-Aleut) than IE seems to be. Hence, my assumption that Proto-Indo-Uralic was more like PU than like PIE. This may be utterly wrong!

With "Mitian palatals" I mean that most non-IE Mitian families have a fuller series of palatal consonants, including stops, a nasal and a liquid, which PIE seems to lack apart from the semivowel *y (unless the palatovelars count, but PIE has only stops in that series).

(And note that "PIE0" and "PIE1" are not what is usually called "Early PIE" (pre-Anatolian PIE) and "Late PIE" (post-Anatolian PIE). The latter two are "PIE2" and "PIE3" in my nomenclature (which is based on a nomenclature introduced by Francisco Rodríguez Adrados, who, however, has no "PIE0" and whose ideas of "PIE1" are IMHO dead wrong), but yet earlier stages from which no other descendants than PIE2 and PIE3 (and their descendants) are known. My hypothetical "Aquan" family, on which the "Hesperic" conlang family which Old Albic belongs to is based, would be a descendant of PIE1, though I am no longer sure about that, as I wrote earlier today.)

[mèþru]

It could be PIE is conservative and the rest have common innovations, whether genetic or areal

[WeepingElf]

It could be PIE is conservative and the rest have common innovations, whether genetic or areal

That can't be ruled out, but that would mean that seven families which do not seem to form a valid node all innovated in similar directions. It is IE that seems to be the "odd man out" among the Mitian bunch (besides Chukotko-Kamchatkan, which also looks weird, of course in very different ways than IE), while Uralic looks more "typically Mitian" to me. Yet, we don't even know whether Indo-Uralic is a valid node within Mitian or whatever; after all, some people think that Yukaghir is the closest living kin of Uralic (but to me, the lexical similarities between Uralic and Yukaghir rather look like loanwords from an extinct Samoyedic language - today, Samoyedic and Yukaghir are about 2,000 km apart from each other, but the languages between them, Evenki and Yakut, are later arrivals, and the two families may have met somewhere in between), and Eskimo-Aleut also looks quite similar (though that may simply be conservatism in both families).

Proto-Uralic of course can be expected to have innovated, too. For instance, its reduced vocalism in non-first syllables (just a higher and a lower vowel, basically, while there are as many as eight vowel qualities in first syllables) is unique among the Mitian languages and thus probably an innovation. Tropylium once suggested that PU *ü is also an innovation, resulting from some sort of chain shift in which back vowels were raised and old *u went out of the way by fronting. At any rate, in my hypothesis it seems as if PIE0 had no front rounded vowels, because otherwise PIE would probably have a fourth velar series (labialized and palatalized) and perhaps a high front rounded vowel *ü, which it doesn't have.

[KathTheDragon]

I still think that invoking "Mitian" in support of any hypothesis is misleading at best, circular at worst.

[Znex]

In the Blevins thread, I summarised her argument proposing a connection between Indo-European and Basque, and I thought I might highlight some pronominal parallels now between Basque and Uralic that I've been thinking about. Maybe these will be interesting as well.

Code: Select all

{1SG}
PIE *éǵ, (e)me-				PB *ekʰ-, ene- << **egi, emene				PU *(e)mi-n-
> Hit. ʔuk, ammu-, amm-			> aB eketa {1SG,0SG}, en-, ene- {1SG}			> Mansi am, ānəm-, Hung. én 
> Sans. ahám, ma-			> mB enne- {1SG.OBL}					> Nen. mań, Ngan. mənə
> Anc.Gr. egṓ, (e)mo-, (e)me		> B en- {VRB 1SG.OBL}, ni {1SG.ABS}			> Fin. minu- > Coll.Fin. mä~mie, mu-
> Lat. egō, me-											> Sami mon, mu-

{2SG}
PIE *túh₂, t(w)(e)- < **tī, tw-		PB *tʰi, tʰo						PU *ti-n-, ?tu-n-
> Hit. tsīk, tu-, twē-			> B hi {2SGFAM}, to {2SGFAM.M.VOC},			> Hung. te
> Sans. tvám, tú-, t(a)v-		ho-re {2SGFAM-GEN}, txo(txo) {boy.VOC}			> Nen. toďi, Ngan. tənə
> Anc.Gr. sý, so-, se										> Fin. sinu- > Coll.Fin. sä~sie, su-
> Lat. tū, tu-, te-										> Sami don, du-

[WeepingElf]

I still think that invoking "Mitian" in support of any hypothesis is misleading at best, circular at worst.

Fair. Outgroup comparison is an established way of telling which of two related entities has innovated where they disagree, but this of course requires that the relationships have been already established. Example: Insular Celtic and Continental Celtic differ in many ways, but as other IE languages (such as Latin) are more like Continental than Insular Celtic, one can tell that it is Insular Celtic which has innovated here, while Continental Celtic is quite conservative. At first glance, the case discussed here seems similar, with IE being the "Insular Celtic", Uralic the "Continental Celtic" and Mitian "IE". But there is of course a fundamental difference: None of the relationships have been established so far, so the "result" that IE has innovated remains guesswork, and the conclusion is fallacious (even if it may be plausible).

Indeed, this has to do with my main criticism against Nostratic and most other macrofamily proposals: They approach the problem from the wrong end. They draw up a huge family tree, and then try to reconstruct the root by comparing languages from opposing ends of the tree (such as IE and Afroasatic). They work from the top down. In my opinion, macrocomparison is better done from the bottom up, by first comparing language families that seem to be closest, and then proceeding to deeper entities. Of course, this has problems, too, as one doesn't really know beforehand which entities are closest and a promising place to start with. While IE and Uralic are geographically closer to each other than to other "Mitian" languages (if we assume, as most linguists do, that Turkic, the next closest neighbour, originated somewhere around the Altai mountains), but that still of course does not mean that they are most closely related. They may constitute two different westward movements which are not especially close in the family tree. But you have to start somewhere ...

[Nortaneous]

In the Blevins thread, I summarised her argument proposing a connection between Indo-European and Basque, and I thought I might highlight some pronominal parallels now between Basque and Uralic that I've been thinking about.

The PIE 2SG.NOM probably had /i/ rather than /u/ -- /u/ spread from the oblique cases in most descendants, but /i/ is preserved in Hittite and possibly Albanian.

It's hard to tell for Albanian, since *ī > i but *ū > *wi > y~i, and *ī wouldn't have conditioned palatalization or anything. Wiktionary's proposal that the 1SG.NOM u (extended in Standard Albanian to unë) is from *h₁éǵ with the vowel of the 2SG.NOM is a little unusual: it'd have to have occurred before laryngeal-conditioned vowel lengthening or been subject to early irregular vowel shortening, since AFAIK *ū can't give /u/. Orel instead proposes that the ego-form was lost entirely and replaced by a reflexive form like *swom.

But positing Proto-Albanian *tī instead of *tū has the benefit that ti would be regular -- you do sometimes get i from *ū, but it's not the default reflex and there's nothing here that would condition it. Orel says *ū > i word-finally, but his only other example (leaving aside 'mouse', where *ū was preceded by a labial, which he says regularly conditions *ū > i) is thi < *suH- 'pig', where the initial consonant is also irregular! And it must've been there in pre-Anatolian PIE, so why not.

[mae]

-

[KathTheDragon]

Not necessarily. Levelling of *u from the non-nominative forms seems like a pretty repeatable innovation, so it's of little value for subgrouping

[WeepingElf]

I wish to come back to the first part of my post from last Wednesday (the genetics of PIE and PU) after we got sidetracked discussing whether what I speculated about in the second part (the hypothetical sound changes from PIU to PIE) can be known at all. I have realized that I got something wrong.

First of all, Y-DNA haplogroups are not the be-all, end-all of human genetics. They of course only trace the male lines, and are just one thing out of many; but they are the easiest for non-specialists like me to get a handle on, while the picture of the mtDNA haplogroups, which trace the female lines, is so blurry that hardly anything can be made out, and autosomal DNA with all its cross-overs is very difficult.

Then, I wrote that Khvalynsk was mostly R1a and Dniepr-Donets mostly R1b, but it is apparently actually the other way round. Yamnaya seems to have been almost uniformly R1b, at least the warrior-nobility, who are what are found in the kurgans, the commoners having been buried in humbler graves which are lost in time. But Corded Ware is mostly R1a, whatever to make of that.

What regards Maykop (which I did not discuss at all so far), which one is tempted to see as Proto-Abkhaz-Adyghe, Eupedia has it as R1b, but that haplogroup apparently is not frequent in the northern Caucasus today, at least not according to the tables I found on Wikipedia. Were they ousted by newcomers later (in which case Proto-Abkhaz-Adyghe was the language of the newcomers rather than of Maykop)? Or is Eupedia just wrong there (they managed to place Khvalynsk in the wrong location after all, so anything is possible)?

[Salmoneus]

Everything is alright, Salmoneus. I think we just have misunderstood each other, and believe that we can sort it out. First of all, I apologize for likening your behaviour to that of a plonkster, and for summarizing your position as ignoramus et ignorabimus ad aeternum. That was my impression from your posts, but now I think this impression was wrong.

OK. Thank you. And I'm sorry I was overly sharp in my response.

[Salmoneus]

First of all, Y-DNA haplogroups are not the be-all, end-all of human genetics. They of course only trace the male lines, and are just one thing out of many; but they are the easiest for non-specialists like me to get a handle on, while the picture of the mtDNA haplogroups, which trace the female lines, is so blurry that hardly anything can be made out, and autosomal DNA with all its cross-overs is very difficult.

This is very true, and very important. Where people used to see a great chasm, for example, between R1a and R1b in Europe, it turns out that actually the R1b Beaker Folk who expanded from central europe somewhere and who make up most of that R1b continuum are pretty much identical genetically to the R1a Corded Ware people they left behind.

In particular, I think it's important to note that in an expansive, warrior-caste patrilineal society, probably with some sort of concubinage, like the early IE societies, Y-DNA clades can expand very quickly. High-status men can have a LOT more kids and grandkids than low-status men. And Y-DNA is subject to founder effects in small societies.

It should also be said that 'R1b' and 'R1a' were/are not monolithic blocks. Already at the time of the Yamnaya, their society would have been home to a whole bunch of related but distinct patrilineages. Basically, there were a lot of patrilineages, and most of them were either R1b or R1a - but they weren't all exactly the same R1a or R1b.

Also, we've only tested a tiny fraction of the population. There were thousands or tens of thousands of Yamnaya, and we've tested... half a dozen? The fact that Haak tested 7 Yamnaya and found all of them were R1b, and five of them exactly the same clade of R1b, tells us something - but it doesn't tell us everything. That was probably the royal lineage, or at least a noble lineage, but there were surely other lineages among the people as a whole. In some ways it actually makes sense that those who left to migrate to other areas might not have been the rulers back home...

What regards Maykop (which I did not discuss at all so far), which one is tempted to see as Proto-Abkhaz-Adyghe, Eupedia has it as R1b, but that haplogroup apparently is not frequent in the northern Caucasus today, at least not according to the tables I found on Wikipedia. Were they ousted by newcomers later (in which case Proto-Abkhaz-Adyghe was the language of the newcomers rather than of Maykop)? Or is Eupedia just wrong there (they managed to place Khvalynsk in the wrong location after all, so anything is possible)?

As I understand it, and as far as I recall, Maykop have no clear descendents today.

[Nortaneous]

Would there be any notable subgrouping/historical implications of Albanian having a more conservative form for the 2SG pronoun than almost any other IE branches? It's already taken to be an independent branch but would this mean it split off earlier than other single-language branches?

No.

Think of it like English: there's no reason a dialect couldn't preserve differences from the standard in its pronominal system -- thou in Newfoundland or some parts of Britain, or yinz or youse instead of y'all in America -- while continuing to participate in the main stream of diachronic change.

[WeepingElf]

First of all, Y-DNA haplogroups are not the be-all, end-all of human genetics. They of course only trace the male lines, and are just one thing out of many; but they are the easiest for non-specialists like me to get a handle on, while the picture of the mtDNA haplogroups, which trace the female lines, is so blurry that hardly anything can be made out, and autosomal DNA with all its cross-overs is very difficult.

This is very true, and very important. Where people used to see a great chasm, for example, between R1a and R1b in Europe, it turns out that actually the R1b Beaker Folk who expanded from central europe somewhere and who make up most of that R1b continuum are pretty much identical genetically to the R1a Corded Ware people they left behind.

Sure. And Corded Ware is genetically very close to Yamnaya, apart from the seeming Y-DNA hapologroup discrepancy which probably was a quirk of the ruling classes anyway - see below.

In particular, I think it's important to note that in an expansive, warrior-caste patrilineal society, probably with some sort of concubinage, like the early IE societies, Y-DNA clades can expand very quickly. High-status men can have a LOT more kids and grandkids than low-status men. And Y-DNA is subject to founder effects in small societies.

Yep.

It should also be said that 'R1b' and 'R1a' were/are not monolithic blocks. Already at the time of the Yamnaya, their society would have been home to a whole bunch of related but distinct patrilineages. Basically, there were a lot of patrilineages, and most of them were either R1b or R1a - but they weren't all exactly the same R1a or R1b.

Certainly!

Also, we've only tested a tiny fraction of the population. There were thousands or tens of thousands of Yamnaya, and we've tested... half a dozen? The fact that Haak tested 7 Yamnaya and found all of them were R1b, and five of them exactly the same clade of R1b, tells us something - but it doesn't tell us everything. That was probably the royal lineage, or at least a noble lineage, but there were surely other lineages among the people as a whole. In some ways it actually makes sense that those who left to migrate to other areas might not have been the rulers back home...

Are these 7 Yamnaya specimens from kurgan graves? If yes, they were chieftains - the Khvalynsk-descendant "royal lineage", as you call it, that had taken over Dniepr-Donets. Most likely, the common men were predominantly R1a, as Dniepr-Donets had been and their modern descendants in Ukraine and southern Russia still are. This would make the seeming Y-DNA haplogroup discrepancy between "R1b-dominant" Yamnaya and R1a-dominant Corded Ware a non-issue.

It may be that Bell Beaker was founded by Yamnaya nobles (perhaps younger sons of chieftains) who sought new lands to rule over in the west, while Corded Ware was founded by Yamnaya commoners who sought freedom from their old lords and ascension to nobility by conquering new lands in the north. The result would in such a scenario have been that Bell Beaker was R1b and Corded Ware R1a, which is what we observe.

I even came up with the irreverend idea last night that Late PIE wasn't spoken by Yamnaya but by Corded Ware, and Yamnaya spoke Pre-Proto-Anatolian, but I think this doesn't really make sense.

What regards Maykop (which I did not discuss at all so far), which one is tempted to see as Proto-Abkhaz-Adyghe, Eupedia has it as R1b, but that haplogroup apparently is not frequent in the northern Caucasus today, at least not according to the tables I found on Wikipedia. Were they ousted by newcomers later (in which case Proto-Abkhaz-Adyghe was the language of the newcomers rather than of Maykop)? Or is Eupedia just wrong there (they managed to place Khvalynsk in the wrong location after all, so anything is possible)?

As I understand it, and as far as I recall, Maykop have no clear descendents today.

Then, NWC (and also NEC) would have moved in later from somewhere else, probably from the south, and Maykop wouldn't have spoken Proto-NWC. The Maykop people may thus have spoken anything, maybe a sister language of Late PIE, maybe a language related to that of Dniepr-Donets. What seeming lexical resemblances are there between IE and NWC may be substratum loanwords from this lost Maykop language into NWC. The realization that the Maykop language probably wasn't related to any of the known Caucasian families of course throws up the question whether the hypothetical Paleo-Pontic substratum language of the Dniepr-Donets people was typologically anything like the modern Caucasian languages. It could have been of a completely different structure, thus knocking the bottom out of the old "IE=Indo-Uralic on Caucasian substratum" theory. (Of course, the NWC languages could have acquired the relevant parts of their typological profile - ejectives, palatalized and labialalized consonants, reduced vowel systems - from the Maykop substratum, but this is a very adventurous speculation.)

I have found this map of Mesolithic Y-DNA haplogroups, which, however, is not very accurate, I think. But I guess it is not sheer fancy. So WHG ("Paleo-Atlantic" in my paleolinguistic model) would be I1 and I2, EHG ("Paleo-Pontic") R1a (the boundary zone between these two, I think, was a good deal farther west than what that map shows - I'd draw it roughly from Riga to Odessa or something like that), Siberians N1c and Central Asians R1b. Fair.

In the late Mesolithic, N1c and R1b advanced into eastern Europe, bringing in languages unrelated to each other and to the "Paleo-Pontic" languages. The European R1b population would then, in my model, have shifted, for whichever reason, to the language of their northern N1c neighbours, altering it under the influence of their substratum (we may nickname their old language "Arwanbi" - guess how I arrived at that name). This would be the point, if IE and Uralic are indeed related to each other (which remains to be shown by means of the comparative method), where Indo-Uralic, originally the language of the N1c people of the Volga region, broke up into what were to become PIE and Proto-Uralic. So the Khvalynsk people would have spoken an Arwanbi-influenced Indo-Uralic; when they took over the Paleo-Pontic-speaking Dniepr-Donets culture, forming Sredny Stog which would later become Yamnaya, their Arwanbi-influenced Indo-Uralic would in turn be altered by a Paleo-Pontic substratum and become PIE. Meanwhile, the language of the N1c people in the north developed undisturbedly into Proto-Uralic.

This, together with the fact that Uralic shows greater resemblance than IE to such Siberian and Central Asian languages like Turkic or Yukaghir, and also to Eskimo-Aleut, is my main reason to guess that Uralic is more conservative than IE, though this of course requires independent proof by means of reconstructing Indo-Uralic, which would constitute a major research programme probably too vast for us amateurs to work out (even if historical linguistics is a field, as cases like de Saussure and Ventris show, where amateurs and students can achieve major breakthroughs). Language shifts often result in substantial restructurings of the language that takes over, as seems to have been the case, for instance, in Insular Celtic, whose radial restructuring, apparently within less than 1000 years, has for long been ascribed to the structure of the pre-Celtic substratum languages, whatever their affiliation.

ADDENDUM: I wonder whether the identification of R1b as the dominant Y-DNA haplogroup in the Maykop culture, as shown on the (notoriously imprecise) Eupedia maps, is due to the same mistake as in the case of the Yamnaya culture, where there is good reason to assume that the common men were R1a rather than R1b: samples taken from kurgan graves which only show the genetic composition of the ruling class while the common men had different haplogroups. And maybe the language replacement failed in Maykop, and the original language survived as NWC. I don't know whether this is plausible, though.

SECOND ADDENDUM: It may have been that the Yamnaya nobles spoke a more conservative PIE dialect than the commoners. If the social stratification was tight enough, the nobles may have spoken a form of PIE that was only slightly influenced by the Paleo-Pontic substratum, while the speech of the commoners was profoundly altered by it. In the scenario laid out above, in which Yamnaya (or Sredny Stog; the earliest "kurgan invasions" into Central Europe were at the time of Sredny Stog already) nobles founded Bell Beaker but Yamnaya commoners founded Corded Ware, the PIE dialects of the Bell Beaker culture would then have been more archaic than those of the Corded Ware culture. This, however, is of course sheer speculation and may be utter bullfrogs.

[Nortaneous]

Speaking of Albanian pronouns, there's early 1SG u, extended in various ways -- the standard has unë, and Italo-Albanian has uth, with what appears to be a demonstrative.

I wonder if there's final devoicing anywhere in Italo-Albanian... because if we back-project the Hittite form into early IE *uǵ, the expected Albanian descendant is udh. And we're all writing in a language where the descendant of *ǵ in the 1SG.NOM was irregularly lost, so there's that.

Wiktionary has some nonsense about copying the vowel from the 2SG, but Orel's explanation is that the ego-form was lost and replaced with a reflexive. Sure, there's a reflexive form u, but why would a reflexive replace the 1SG.NOM? Is this attested anywhere else? And is there any explanation for /u/ in the Hittite 1SG.NOM?

[mèþru]

Italo-Albanian?