Zompist Bboard Again

Moose-tache wrote: ↑Thu Feb 13, 2020 7:02 am It doesn't bode well that "North Caucasian" is already bonkers.

Some sort of macrofamily involving NEC seems possible, but I don't think enough work has been done on NWC yet. There'd have to be a firm grounding for a pre-VVS stage, unless of course the VVS is original. And given that featural transfer was not a major part of the development of the Marshallese VVS -- /tˠ tʲ/ go back to *s *t, rather than *tV[-front] *tV[+front] -- I'm a little skeptical of anything that assumes the NWC VVS developed mostly through featural transfer.

Moose-tache wrote: ↑Thu Feb 13, 2020 7:02 am It doesn't bode well that "North Caucasian" is already bonkers.

North Caucasian seems well enough like a valid family (more solid than e.g. Afrasian or Niger-Congo sensu lato, IMO), the problem is that it's Phonology Hell and it might be a while before we get reconstructions that are both usable for further comparison work and not overly kitchensinky.

The proposed connection with Basque is at least vaguely believable as being routable thru the Neolithic expansion of agriculture into Europe.

Nortaneous wrote: ↑Thu Feb 13, 2020 12:02 pmSome sort of macrofamily involving NEC seems possible, but I don't think enough work has been done on NWC yet. There'd have to be a firm grounding for a pre-VVS stage, unless of course the VVS is original. And given that featural transfer was not a major part of the development of the Marshallese VVS -- /tˠ tʲ/ go back to *s *t, rather than *tV[-front] *tV[+front] -- I'm a little skeptical of anything that assumes the NWC VVS developed mostly through featural transfer.

VVS from feature transfer is also well enough attested in Nenets. The usual proposal for NWC is by the way that palatalization comes from feature transfer, but labialization is common inheritance with NEC — plausible enough in light of languages like Khwarshi (Tsezic) or Tsakhur (Lezgic) which also have well-developed labialized alveolar and labialized palatal series, in addition to the more typologically common labialized velars and labialized uvulars.

Tropylium wrote: ↑Thu Feb 13, 2020 1:15 pm

Moose-tache wrote: ↑Thu Feb 13, 2020 7:02 am It doesn't bode well that "North Caucasian" is already bonkers.

North Caucasian seems well enough like a valid family (more solid than e.g. Afrasian or Niger-Congo sensu lato, IMO), the problem is that it's Phonology Hell and it might be a while before we get reconstructions that are both usable for further comparison work and not overly kitchensinky.

Sure, there are macrofamily proposals that are much wackier than North Caucasian, but where is the evidence for North Caucasian? Those two language families look so vastly different from each other that I find it hard to believe that they are in any way demonstrably related. I mean, they have less in common with each other than IE and Uralic. Sure, the phonologies are vaguely similar, but only vaguely so, and what similarities there are in their phonologies are almost certainly due to convergence. Also, both show ergativity and SOV word order - like many languages in many parts of the world. But everywhere else, they are almost the opposite of each other!

The proposed connection with Basque is at least vaguely believable as being routable thru the Neolithic expansion of agriculture into Europe.

Ah, Vasco-Caucasian! This calls back memories of Octaviano Basque has almost nothing in common with either NWC or NEC except some basic typological features like SOV word order and ergativity. But the Neolithic migrations may indeed account for them, so this is not entirely off the rockers. However, the Neolithic Near East was full of small language families and isolates, judging from the diversity attested there in the Bronze Age, so these lineages may date back to the Upper Paleolithic, with their common root way beyond reach for historical linguistics.

Nortaneous wrote: ↑Thu Feb 13, 2020 12:02 pmSome sort of macrofamily involving NEC seems possible, but I don't think enough work has been done on NWC yet. There'd have to be a firm grounding for a pre-VVS stage, unless of course the VVS is original. And given that featural transfer was not a major part of the development of the Marshallese VVS -- /tˠ tʲ/ go back to *s *t, rather than *tV[-front] *tV[+front] -- I'm a little skeptical of anything that assumes the NWC VVS developed mostly through featural transfer.

VVS from feature transfer is also well enough attested in Nenets. The usual proposal for NWC is by the way that palatalization comes from feature transfer, but labialization is common inheritance with NEC — plausible enough in light of languages like Khwarshi (Tsezic) or Tsakhur (Lezgic) which also have well-developed labialized alveolar and labialized palatal series, in addition to the more typologically common labialized velars and labialized uvulars.

Labialization is also found in Kartvelian, in some sense - any non-labial consonant or consonant cluster may be followed by *w, which looks as if there once was labialization which was later segmented out. So this seems to be a pan-Caucasian feature.

Johanna Nichols on Proto-NEC:

Purely phonological reconstruction, however, proves to be problematic, even for the limited task of comparing Nakh words to standard Daghestanian sets, because for many cognate sets only a single root consonant is reconstructable. Most ND roots have a structure C1VC2 or C1VRC2, but C1 is prone to be replaced by a gender prefix in various daughter languages, and the medial resonant is in a position of neutralization and, in addition, can also be incorporated into the system of gender alternations in the eastern branches. This means that only one consonant in the root - C2 - is likely to display a regular phonological correspondence, a fact which makes the search for cognates hazardous. Therefore further restrictions on potential cognates must be imposed, and most of them have to do with grammar rather than phonology. ...

Vocalism and ablaut are further good guides to secure cognate status, but they cannot be established for every cognate set. PND vowels are difficult to pinpoint, but one distinction is often fairly clear: whether a vowel was rounded or not.

AFAIK there's no serious PNEC reconstruction yet.

jal wrote: ↑Thu Feb 13, 2020 9:17 am Well, maybe it doesn't say so exactly, but here it says something about some harmony not yet occurring.


JAL

Yeah that’s talking about labial harmony - some later Mongolian languages including Khalkha turned /a ə/ into /ɔ o/ next to other /ɔ o/.

If you want a reeaaally in depth look at vowel harmony in Mongolian and its neighbours, check out this paper.

-

mae wrote: ↑Thu Feb 13, 2020 8:55 pm My impression of "reconstructions" for Austroasiatic is that they have a similar problem to what Nathan Hill claims is at issue in PTB reconstructions--scholars tend to just take as canonical what's more or less a modern attested form and then "derive" everything through semantically vacuous 'prefixes' and so on, so regular sound change sometimes seems to take the backseat.

Unreconstructable prefixes are necessary for AA. Many languages developed a bimoraicity constraint and extended monomoraic words by addition of semantically vacuous and non-corresponding prefixes. (Cf. Sinitic: if there's another branch with a Mandarin-like minimal word constraint, that branch will have correspondences with Mandarin in some syllables of words, but the words in total obviously won't correspond, since they'll have been filled out differently.)

Extant PTB reconstructions are basically Proto-Tibeto-Burman in the literal sense: produced mostly by starting with Tibetan and fiddling with it until Burmese pops out. You can't produce PIE by fiddling with Latin until it starts lining up with Old Irish. Roger Blench has even suggested in print that the familiar sesquisyllabic structure wasn't even ancestral, and instead developed in the distant ancestors of Chinese, Tibetan, and Burmese from contact with Austroasiatic! Also, the internal grouping is not at all settled -- I've seen it suggested that Qiangic isn't monophyletic, for example, and some allegedly TB languages may not be TB at all. (Blench's "Kho-Bwa", etc.) Reference works have to be taken with a mine of salt, and the Anglosphere ones are still a little better than the Chinese ones, where they apparently still think Tai-Kadai and Hmong-Mien are in, presumably on the grounds of they're tonal and they aren't anything else so people figured they were before Haudricourt and didn't have a compelling reason to update after him.

The prefixes are fiddly because they're later additions that shouldn't be projected back to PTB.

Indo-European and Austronesian have protolang reconstructions that are, as far as I can tell, pretty well worked out. Uralic is still fiddly, but there's a solid foundation. And there are various 'microfamilies' with reconstructions that are at least serviceable, like Algonquian and Semitic. But I really wouldn't trust any protolang reconstruction for a large family that isn't IE or Aus at this point. (Maybe Pama-Nyungan is decent? I have no idea.)

It's unfortunate that the epistemic status of these things isn't marked. Reference works just up and say shit. Papuan language classifications, for example, are extremely preliminary, but aren't marked as such. It's reasonable to have preliminary classifications that remain to be proven -- when you go to apply the comparative method, you first have to work out through 'unscientific' means which languages to apply it to -- but... they aren't proven.

Blench might have the right idea about the two-syllable words. That would basically sink all of the attempts to reconstruct the proto-language, but it would also explain why we're having such a hard time building a proto-language that likely wasnt spoken all that long ago. (Though PIE eludes us as well, despite being the best attested family in the entire world.)

though i see at http://enwp.org/Sino-Tibetan_languages#Homeland that he says it was c. 7000 BC, so quite a lot of time to lose those two syllable words, compared to Tai.

re PIE being solidly reconstructible: the grammar, i guess, but the phonology and the word structure are wildly divergent from one author to the next, e.g. compare the versions of Schleicher's fable ,even the newer ones:

http://enwp.org/Schleicher%27s_fable#Th ... the_Horses

Pabappa wrote: ↑Fri Feb 14, 2020 1:15 am re PIE being solidly reconstructible: the grammar, i guess, but the phonology and the word structure are wildly divergent from one author to the next, e.g. compare the versions of Schleicher's fable ,even the newer ones:

Phonology is fucked. Just transcribe them into the standard notation and ignore all the ones that don't have laryngeals. I probably did a bad job with these but w/e, close enough fordemonstration.

Adams:
[gʷr̥hₓḗy] h₂ówis, kʷésyo wlh₂néh₂ ne (h₁é) est, h₁éḱwons spéḱet, h₁oynom gʰe gʷr̥hₓúm wóǵhom wéǵhontm̥ h₁oynom-kʷe méǵh₂m bʰórom, h₁oynom-kʷe ǵʰménm̥ hₓṓḱu bʰérontm̥. h₂ówis tu h₁eḱwoibʰ(y)os wewkʷét: 'ḱḗr h₂egʰnutór moy h₁éḱwons h₂éǵontm̥ h₂nérm̥ widn̥téy. h₁éḱwōs tu wewkʷónt: 'ḱludʰí, h₂ówei, ḱḗr gʰe h₂eghnutór n̥sméy widn̥tbʰ(y)ós. h₂nḗr, pótis, h₂éwyom r̥ wl̥h₂néh₂m sebʰi kʷr̥néwti nu gʷʰérmom wéstrom néǵʰi h₂éwyom wl̥h₂néh₂ h₁ésti.' Tód ḱeḱluwṓs h₂ówis h₂éǵrom bʰugét.

Luhr: (middle with -y, augment, dat.pl -bhos)
h₂ówis, (H)yésmin h₂wl̥h₂néh₂ ne éh₁est, dedork’e (h₁)ékwons, tóm, wóǵʰom gʷérh₂um wéǵʰontm̥, tóm, bʰórom méǵoh₂m̥, tóm, dʰǵʰémonm̥ h₂oHkú bʰérontm̥. h₂ówis (h₁)ékwobʰos ewewkʷe(t): kḗrd h₂gʰnutoy moy widn̥téy dʰǵʰmónm̥ (h₁)ékwons h₂éǵontm̥. (h₁)ékwōs ewewkʷ: kludʰí, h₂ówi! kḗrd h₂gʰnutoy widn̥tbʰós: dʰǵʰémō(n), pótis, h₂wlnéh₂m h₂ówyom kʷnewti sébʰoy gʷʰérmom wéstrom; h₂éwibʰoskʷe h₂wlh₂néh₂ né h₁esti. tód k’ek’luwṓs h₂ówis h₂éǵrom ebʰuge(t).

Melchert:
h₂áwey h₁yosméy h₂wl̥h₁náh₂ né h₁ést, só h₁éḱwoms derḱt. só gʷr̥hₓúm wóǵʰom wéǵʰet; só méǵh₂m̥ bʰórom; só (dʰ)ǵʰémonm̥ h₂ṓḱu bʰéret. h₂ówis h₁éḱwoybʰ(y)os wéwk(ʷ)et: (dʰ)ǵʰémonm̥ spéḱyoh₂ h₁éḱwoms h₁yós h₂áǵeti, ḱḗr moy agʰnutór. h₁éḱwōs tu wéwkʷont: ḱludʰí, h₂owei! tód spéḱyomes/n, n̥sméi aghnutór ḱḗr: (dʰ)ǵhémō pótis sē h₂áwyōm h₂wl̥h₁nā́h₁ (sic) gʷʰérmom wéstrom (h₁)wébʰt, h₂áwibʰ(y)os tu h₂wl̥h₁náh₂ né h₁ésti. tód ḱeḱluwṓs h₂ówis h₂aǵróm bhugét.

Byrd:
h₂áwey h₁yosméy h₂wl̥h₁náh₂ né h₁ést, só h₁éḱwoms derḱt. só gʷr̥hₓúm wóǵʰom weǵʰed; só méǵh₂m̥ bʰórom; só dʰǵʰémonm̥ h₂ṓḱu bʰered. h₂ówis h₁ékʷoybʰyos wewked: “dʰǵʰémonm̥ spéḱyoh₂ h₁éḱwoms-kʷe h₂áǵeti, ḱḗr moy agʰnutor”. h₁éḱwōs tu wewkond: “ḱludʰí, h₂owey! tód spéḱyomes, n̥sméy agʰnutór ḱḗr: dʰǵʰémō, pótis, sē h₂áwyes h₂wl̥h₁náh₂ gʷʰérmom wéstrom wept, h₂áwibʰyos tu h₂wl̥h₁náh₂ né h₁esti”. tód ḱeḱluwṓs h₂ówis h₂aǵróm bʰuged.

Kortlandt 1:
h₃ewis i h₂weli nēh₁st h₁eḱums wēyt, to gʷr̥h₂ewm woǵʰom uǵʰentm̥, to m̥ǵeh₂m̥ borom, to dǵmenm̥ h₁oh₁ḱu bʰrentm̥. wēwkt h₃ewis h₁iḱwos, h₂edʰǵʰo h₁me ḱērt h₂nerm̥ widenti h₁eḱums h₂ǵentm̥. wewkn̥t h₁iḱwes, ḱludʰi h₃we, h₂edʰǵʰo n̥sme ḱērt widenti, h₂nēr potis h₃uyom h₂weli swe gʷʰermom westi kʷr̥newti, h₃wēy kʷe h₂weli neh₁sti. to ḱeḱluwus h₃ewis bʰleh₂nom bʰēwgt.

Kortlandt 2:
h₃ewis yoy h₂ulh₁neh₂ nēh₁s h₁eḱuns h₁e uēid, tom gʷrh₂ewm woǵʰom weǵʰontm̥, tom m̥ǵeh₂m̥ bʰorom, tom dʰǵʰmenm̥ h₁oh₁ḱu bʰerontm̥. h₁e wēwk h₃ewis h₁eḱumus, h₂edʰǵʰo h₁moy ḱērd h₂nerm̥ widenti h₁eḱuns h₂eǵontm̥. h₁e wewkn̥d h₁iḱwes, ḱludʰi h₃wey, h₂edʰǵʰo n̥smi ḱērd widenti, h₂nēr potis h₃uyom h₂ulh₁neh₂m subʰi gʷʰermom westi kʷr̥newti, h₃uymus kʷe h₂ulh₁neh₂ neh₁sti. tod ḱeḱluwus h₃ewis pleh₂nom bʰēwg.

WeepingElf wrote: ↑Thu Feb 13, 2020 3:51 pmbut where is the evidence for North Caucasian? Those two language families look so vastly different from each other that I find it hard to believe that they are in any way demonstrably related. I mean, they have less in common with each other than IE and Uralic.

I recommend starting with S. Starostin's A North Caucasian Etymological Dictionary, from where you can follow into the earlier literature. Of course most of it in Russian though. (Bouda, Dumézil and Trubetzkoy are notable exceptions, but going just by them is really not going to be a good foundation.)

Note that I have concerns with Starostin & co.'s reconstruction methodology (which has a clear tendency to complicate reconstructions as one goes back in time), but the data backing their NEC and NCc work is abundant and previously established, and seems mostly legit. Remember that most long-known language families started off with some pretty bad reconstructions too, but no-one thinks that e.g. Schleicher's PIE being too Sanskrit-centric in any way invalidates the family's existence. This existence of real data is what IMO makes North Caucasian fundamentally more reliable than macrofamilies like AA or Niger-Congo (even narrow Atlantic-Congo is not held up by much real data yet).

Like Nort and mae point out, really most accepted families also still lack the kind of solid reconstructions that PIE has today. They don't even need to be huge for this:
– Proto-Turkic proper has still not been worked out to general satisfaction due to vowel system problems with Chuvash
– Proto-Berber is only in the last 10–20 years starting to shape up into something resembling stability
– Proto-Dravidian so far is really just Proto-Wider South Dravidian with the small fringe languages (Northern and "Central" Dravidian) treated as just an appendix
– while Proto-Eskimo and Proto-Chukotkan are shallow enough to be easily reconstructable, their exact correspondences with Aleut and Kamchatkan have been charted out only partially
– "more obscure" families i.e. the likes of Chibchan, Iwaidjan or Maban are generally going to be doing worse yet.

So NCc seems to be well over the worldwide waterline we use for claiming a family as established. At minimum it's unprincipled to judge it for reconstruction being less advanced than its neighbors IE and Semitic, when these are really two of the best-researched families in the world.

WeepingElf wrote: ↑Thu Feb 13, 2020 3:51 pmBasque has almost nothing in common with either NWC or NEC except some basic typological features like SOV word order and ergativity.

If you look at typology, sure; but then typology is not what establishes relationships, it's shared common material with regular sound correspondences, and that has been proposed at least.

The risk is that highly detailed NCc phonology and the necessary "long-branch" status of Basque allows too many degrees of freedom to set up comparisons. However the Vasco-Caucasian proposals consistently have more substance and also more grammar than, say, the Basque–IE proposals out there. It's a pretty weak sign but enough for it to be not totally bonkers as a hypothesis (which I would think is a pretty low bar anyway, e.g. Eurasiatic / Mitian is also not totally bonkers).

Tropylium wrote: ↑Fri Feb 14, 2020 9:23 am

WeepingElf wrote: ↑Thu Feb 13, 2020 3:51 pmbut where is the evidence for North Caucasian? Those two language families look so vastly different from each other that I find it hard to believe that they are in any way demonstrably related. I mean, they have less in common with each other than IE and Uralic.

I recommend starting with S. Starostin's A North Caucasian Etymological Dictionary, from where you can follow into the earlier literature. Of course most of it in Russian though. (Bouda, Dumézil and Trubetzkoy are notable exceptions, but going just by them is really not going to be a good foundation.)

Note that I have concerns with Starostin & co.'s reconstruction methodology (which has a clear tendency to complicate reconstructions as one goes back in time), but the data backing their NEC and NCc work is abundant and previously established, and seems mostly legit. Remember that most long-known language families started off with some pretty bad reconstructions too, but no-one thinks that e.g. Schleicher's PIE being too Sanskrit-centric in any way invalidates the family's existence. This existence of real data is what IMO makes North Caucasian fundamentally more reliable than macrofamilies like AA or Niger-Congo (even narrow Atlantic-Congo is not held up by much real data yet).

I see. In fact, with the two families spoken side by side, a common origin makes sense. Yet, when two languages or families share much lexicon but no morphology, aren't loanwords the more likelier hypothesis? But I know too little about NWC and NEC, and nothing about Starostin's dictionary to make a judgment here.

So NCc seems to be well over the worldwide waterline we use for claiming a family as established. At minimum it's unprincipled to judge it for reconstruction being less advanced than its neighbors IE and Semitic, when these are really two of the best-researched families in the world.

Fair. We are standing at the beginning of the exploration of the question, so one should not expect a "finished product". As you say, IE comparative linguistics started with ideas that did not prevail, such as a decent of all IE languages from Vedic (which was believed to be much older than it actually is, due to Indian traditions that the Vedas were 5,000 years old). The history of reconstructed PIE is essentially a history of emancipation from the "Sanskrit" model, with changes made when they turned out to be inevitable. There is a good chance that the thing still resembles Old Indic more than it should!

WeepingElf wrote: ↑Thu Feb 13, 2020 3:51 pmBasque has almost nothing in common with either NWC or NEC except some basic typological features like SOV word order and ergativity.

If you look at typology, sure; but then typology is not what establishes relationships, it's shared common material with regular sound correspondences, and that has been proposed at least.

Which is precisely what I was trying to say!

The risk is that highly detailed NCc phonology and the necessary "long-branch" status of Basque allows too many degrees of freedom to set up comparisons. However the Vasco-Caucasian proposals consistently have more substance and also more grammar than, say, the Basque–IE proposals out there. It's a pretty weak sign but enough for it to be not totally bonkers as a hypothesis (which I would think is a pretty low bar anyway, e.g. Eurasiatic / Mitian is also not totally bonkers).

Sure. In my opinion, Mitian is not totally bonkers, it is a good candidate of an actually extant macrofamily. The morphological resemblances are not easy to explain otherwise, and even the genetics seem to play ball here.

Whimemsz wrote: ↑Fri Feb 14, 2020 2:03 pmThe main remaining problem is how precisely Blackfoot fits into everything. (I do think the traditional reconstruction of the vowels is sort of backwards, but...ehh.)

So why is Blackfoot tricky to fit in? And would you mind elaborating the issues you see with the vowel reconstruction? (I don’t know if this is the thread to do it)

So what state is Proto-Algic in?

We're pretty sure that proto-Algic exists, but two of its three branches are attested by a single language each that diverged from the others perhaps as much as 7000 years ago. Reconstructing the proto-language is a bunch of wild shots in the dark, really, and always will be. Right now the consonant inventory featured on Wikipedia looks like a superset of the inventories of the three branches, which is something that tends to happen when you don't, and can't, know the diachronics that led to the present-day setups. It's the same thing that happens with reconstructions of Nostratic, and it used to be true to some extent of PIE, where people were drawing tables showing four stop series, laryngeals, five vowels, length contrasts, and sometimes even tone all projected back into PIE just because each of these features appears in one or more of the descendant languages.

But now imagine that the only attested IE languages were modern Hindi, German, and French. Would we be drawing up inventories for PIE with four stop series, a dozen fricatives, nasal vowels, and rich consonant clusters? Probably not *quite* that bad because we would at least be able to see a few patterns that would help us piece out the sound changes, but ... 7000 years is a long time, and there's a good chance that the three Algic branches have many cognates that we just can't see.

There's no reason for us to reject the proto-Algic consonant inventory outright, because there is no tangible evidence against it ... but it's become apparent over time that reconstructions like this ... i think the term is "external"? .... tend to overestimate phoneme inventories because we lack the ability to figure out how new phonemes could have arisen in each branch.

WeepingElf wrote: ↑Fri Feb 14, 2020 11:50 amI see. In fact, with the two families spoken side by side, a common origin makes sense. Yet, when two languages or families share much lexicon but no morphology, aren't loanwords the more likelier hypothesis? But I know too little about NWC and NEC, and nothing about Starostin's dictionary to make a judgment here.

Supposedly they would have some shared grammatical markers too. WP:North Caucasian has a table of pronouns that looks, while not transparently obvious, as plausible if the sound correspondences check out. NW Caucasian though is so phonologically concentrated that it has probably been losing a bunch of affixal morphology too along the way in any case; broadly the same problem as with trying to find morphological evidence for most of Sino-Tibetan.

There also does not seem to have been substantial contact anytime recently that would cover either all of NEC or all of NWC. Geographically Abkhaz and Ubykh are closer to Kartvelian than anything East Caucasian. It could be possible to suppose that some EC branch like Nakh used to be spread more widely in the Forecaucasian steppe before the arrival of Turkic and IE (something has to be the third newest arrival in the Caucasus), but it would seem that that would then have to be 5000 years or more back already. So not very compareable to any kind of long-standing neighbor situation like Germanic/Finnic, Indic/Dravidian, Cushitic/Ethiosemitic, Quechua/Aymara where it's simple to explain the vast majority of similarities as loanwords.

Whimemsz wrote: ↑Sat Feb 15, 2020 6:27 pm The Wikipedia article is based on the work of Paul Proulx, whose Proto-Algic reconstructions and methodology are a joke -- basically doing exactly what Pabappa said, "finding" new correspondences for every potential cognate set even if it's only attested in the one set, and thus ending up, basically, projecting the inventories of all the daughter languages combined back to Proto-Algic -- and not accepted by other Algonquianists or Algicists. There's a little discussion here.

So how is this reconstruction inferior to Starostin's reconstruction of North Caucasian?

Tropylium wrote: ↑Sat Feb 15, 2020 4:40 pm

WeepingElf wrote: ↑Fri Feb 14, 2020 11:50 amI see. In fact, with the two families spoken side by side, a common origin makes sense. Yet, when two languages or families share much lexicon but no morphology, aren't loanwords the more likelier hypothesis? But I know too little about NWC and NEC, and nothing about Starostin's dictionary to make a judgment here.

Supposedly they would have some shared grammatical markers too. WP:North Caucasian has a table of pronouns that looks, while not transparently obvious, as plausible if the sound correspondences check out. NW Caucasian though is so phonologically concentrated that it has probably been losing a bunch of affixal morphology too along the way in any case; broadly the same problem as with trying to find morphological evidence for most of Sino-Tibetan.

Thank you. That evidence doesn't look all that bad at all; in fact, it looks almost as good as Indo-Uralic, for instance (indeed, the numeral correspondences are better - IE and Uralic totally diverge there). I wasn't aware that the two families actually have so much in common.

There also does not seem to have been substantial contact anytime recently that would cover either all of NEC or all of NWC. Geographically Abkhaz and Ubykh are closer to Kartvelian than anything East Caucasian. It could be possible to suppose that some EC branch like Nakh used to be spread more widely in the Forecaucasian steppe before the arrival of Turkic and IE (something has to be the third newest arrival in the Caucasus), but it would seem that that would then have to be 5000 years or more back already. So not very compareable to any kind of long-standing neighbor situation like Germanic/Finnic, Indic/Dravidian, Cushitic/Ethiosemitic, Quechua/Aymara where it's simple to explain the vast majority of similarities as loanwords.

I see. Many scholars assume that the Late Neolithic Maykop culture spoke Proto-NWC; that seems to make sense to me, but I don't know how good this idea actually is. Some scholars assume an NEC origin south of the Caucasus in what is now Azerbaijan, but I am not sold to that idea. I consider it more likely that NEC also originated north of the Caucasus on the Caspian coast, i.e. in what is now Daghestan (which is, of course, the place where NEC diversity is greatest now). However, these two families probably were neighbours for quite some time, given the fact that Ossetian and the various Turkic languages of North Caucasia are, of course, later intrusions. The divergence of Nakh wthin NEC may be due to a NWC substratum, though I don't know whether Nakh is structurally closer to NWC than Daghestanian, so this idea may be totally bonkers. Some of the differences between NWC and NEC are also a matter of analysis: what, for instance, is a case suffix in an agglutinating language, and what a postposition? (And postpositions could of course have become case markers.) Are those large arrays of "local cases" found in many Daghestanian languages really cases on a par with the ergative, genitive or dative? Etc.

The "Caucasian substratum" held responsible for the divergence of PIE from the "Common Mitian" type by some scholars since it was first proposed by Uhlenbeck in the 1930s may have been either the Maykop language (i.e., Proto-NWC) or a lost lineage either related to or typologically affiliated to NWC (or to NEC, or both). There are some problems with an NWC substratum, though, as PIE shows some "Caucasianisms" alien to Mitian such as the three types of stops (but which three types these were originally, we do not know for sure; the glottalic hypothesis is only a possibility) and the presence of palatalized and labialized velars, but there are some other points where PIE doesn't look "Caucasian" at all, such as the single sibilant phoneme where Caucasian languages (especially NWC) show full series of sibilant affircates and fricatives at at least two (in NWC, even four) places of articulation. The PIE ablaut system looks more like the Kartvelian one than anything in either NWC or NEC. Yet we know that the Yamnaya and Maykop cultures interacted to a great degree; they have many traits in common (such as kurgan burials), and it may have been such that the Yamnaya were Neolithicized from the south via Maykop.