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

[Whimemsz]

.

[Tropylium]

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).

A hypothesis I've sometimes entertained is that Proto-NEC could have been spoken somewhere outside the Caucasus entirely, with all the known branches being intrusions into the Caucasus at various times. So e.g. Nakh would have arrived from the steppes to the north (just as Ossetic or Balkar did), Lezgic and possibly closely related branches would have arrived from the steppes to the east of the Caspian (just as Azerbaijani and NW Iranian did). If various NEC languages were spoken in the steppes before the IE expansion, this could be a part of the Caucasic substratum in IE, as well as an explanation for supposedly Caucasic features in families like Yeniseian and Burushaski. (Even some loanwords in Uralic have been proposed, though the idea is well out of fashion.)

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.

In phonology at least Nakh seems really the farthest away from NWC, with no labialized consonants and fairly large vowel systems. But that's the situation today, I have no idea what the languages might have looked like 5000–6000 years ago in Maykop times. Most likely not exactly the same.

also Nikolaev has published some stuff on reconstructing "Algonquian-Wakashan" [= Algic-Wakashan including Nivkh] that is very bad (…)

Would you have any comments beyond this or know of any critiques by others? It's a proposal that seems noteworthy, but I don't know anything about the families involved.

[Whimemsz]

.

[Nortaneous]

apparently eight in Proto-Nivkh

I think Fortescue reconstructs Proto-Nivkh with six (implying not much in the way of development in the vowels, which makes sense since Proto-Nivkh would be very shallow), but Nikolaev is instead relying on Mudrak's unpublished reconstruction, which apparently takes into account hypothetical extinct 'Nivkhic' languages attested through loanwords into Yukaghir and Proto-Chukotko-Kamchatkan. (Fortescue thinks Nivkh instead forms a family with Chukotko-Kamchatkan.)

I don't know about Yukaghir, but Fortescue's PCK at least didn't strike me as entirely solid. Maybe Mudrak's is better, dunno, his CK stuff is in Starling.

[Whimemsz]

.

[WeepingElf]

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).

A hypothesis I've sometimes entertained is that Proto-NEC could have been spoken somewhere outside the Caucasus entirely, with all the known branches being intrusions into the Caucasus at various times. So e.g. Nakh would have arrived from the steppes to the north (just as Ossetic or Balkar did), Lezgic and possibly closely related branches would have arrived from the steppes to the east of the Caspian (just as Azerbaijani and NW Iranian did). If various NEC languages were spoken in the steppes before the IE expansion, this could be a part of the Caucasic substratum in IE, as well as an explanation for supposedly Caucasic features in families like Yeniseian and Burushaski. (Even some loanwords in Uralic have been proposed, though the idea is well out of fashion.)

I see little reason to assume an origin of Proto-NEC far away from the Caucasus, though the typological similarity between NEC and Burushaski is indeed striking (but there doesn't seem to be much resemblance in substance, which counts here). I'd say that Proto-NEC was most likely spoken on the Caspian coast in Daghestan and maybe a bit into Azerbaijan. Of course, the idea you have sometimes entertained is not totally bonkers; it could have been like that, but the null hypothesis would be a location in Daghestan or northeastern Azerbaijan. Why can't the three Caucasian families have unfolded just where they are now? There were people, and thus languages, there around 3000 BC, so why shouldn't they have spoken Proto-NWC, Proto-NEC and Proto-Kartvelian?

What regards the "Caucasic substratum" in IE, that may have been another, now completely extinct lineage distinct from (but perhaps related to) both NWC and NEC, as I wrote yesterday - if there was such a substratum at all. It is popular to assume a substratum when a language undergoes drastic changes in its structure in a rather short time (as with Insular Celtic, which seems to have warped from the conservative IE type attested in Continental Celtic to the much different type historically observed within less than 1,000 years), but we just don't know how deep the relationship between IE and Uralic actually is - it may be on the order of 5,000 years reckoned from PIE and PU (which are roughly contemporaneous), in which time a lot could have happened. IMHO the main reasons to assume a substratum are (1) that the PIE and PU speakers are genetically quite different (e.g., different Y-DNA haplogroups - with PU speakers more similar to Siberians than PIE speakers), so it seems a bit as if a language shift has happened here, probably with the later PIE speakers, i.e. the ancestors of the Yamnaya people, and (2) that PIE is pretty much an "odd man out" in the Mitian group, while PU looks much like a typical Mitian language.

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.

In phonology at least Nakh seems really the farthest away from NWC, with no labialized consonants and fairly large vowel systems. But that's the situation today, I have no idea what the languages might have looked like 5000–6000 years ago in Maykop times. Most likely not exactly the same.

Yes. The Nakh languages are among those NEC languages with the smallest consonant and the largest vowel inventories, thus phonologically farthest away from NWC; but that may have once been different. I haven't seen reconstructions of Proto-NWC and Proto-NEC yet, so I can't say anything about their former typology.

[Nortaneous]

What little I've seen of PNWC looks like current NWC phonologically. Which makes sense; NWC looks like a pretty shallow family.

[mae]

-

[Tropylium]

There were people, and thus languages, there around 3000 BC, so why shouldn't they have spoken Proto-NWC, Proto-NEC and Proto-Kartvelian?

PNWC and PKv, why not. NEC looks older than that however.

I don't know of any pressing reason to "pull away" any of the NEC branches (Burushaski etc. comparisons are probably not solid enough for that), but I think there might be a "push" argument: over the last 4000 years, there have been new languages forcing their way into the Caucasus at a rate of roughly about a lineage per millennia: Armenian, Ossetic, Kipchak, Kalmyk, Russian. Does it make sense to think that this was preceded by a multimillennial period of stability? If not, does it make sense to assume that presumable Neolithic / Chalcolithic arrivals would just have disappeared later on? (E.g. does later steppe peoples having horses make enough of a difference on either?)

if there was such a substratum at all (…) IMHO the main reasons to assume a substratum are (1) that the PIE and PU speakers are genetically quite different (e.g., different Y-DNA haplogroups - with PU speakers more similar to Siberians than PIE speakers), so it seems a bit as if a language shift has happened here, probably with the later PIE speakers, i.e. the ancestors of the Yamnaya people, and (2) that PIE is pretty much an "odd man out" in the Mitian group, while PU looks much like a typical Mitian language.

Hmm. On the other hand, I would suspect Uralic has itself some kind of a Paleosiberian substrate. This could be enough to explain (1). If there was a substrate with similar typology also further east, and/or if some features of Ural-Altaic and/or Uralo-Siberian typology might be rather recent areal features, that could explain (2). It might be mistaken entirely to use modern Caucasic as an explanation for typological features of PIE 5000 years ago. IE or pre-IE influence in Caucasic could have happened too!

I haven't seen reconstructions of Proto-NWC and Proto-NEC yet, so I can't say anything about their former typology.

For a starting point, here's what Starostin cites as the branch inventories, plus the Avar, Khinalug and Lak inventories from Wikipedia. Segments in parentheses are dubious or loanword phonemes, segments with (ː) have a geminate/fortis equivalent.

Code: Select all

LABIALS / DENTALS
Nakh     p' p    b f   t' t    d
Avar        p    b     t' t    d
Andic       p    b     t' t    d tʷ' tʷ    dʷ
Tsezic   p' p    b     t' t    d tʷ' tʷ    dʷ
Lezgic   p' p(ː) b     t' t(ː) d tʷ' tʷ(ː)
Dargwa   p' p(ː) b     t' t(ː) d tʷ'       dʷ
Lak      p' p(ː) b     t' t(ː) d
Khinalug p' p(ː) b f v t' t(ː) d

Code: Select all

LHATERALS
N                 dɮ  ɬ
Av  (tɬː')  tɬː       ɬ(ː)
An  tɬ'(ː) tɬ(ː)      ɬ(ː)
Ts  tɬ'    tɬ         ɬ
Lz  tɬ'(ː) tɬ(ː) (dɮ) ɬ(ː)   tɬʷ'(ː) tɬʷ(ː) ɬʷ(ː)
D   tɬ'    tɬ(ː)
Lak –
Kh  –

Code: Select all

SIBILANTS
N   ts'    ts    dz s
Av  ts'(ː) ts(ː)    s(ː) z
An  ts'(ː) ts(ː)    s(ː) z
Ts  ts'    ts       s      tsʷ'    tsʷ    sʷ
Lz  ts'(ː) ts(ː)    s(ː) z tsʷ'(ː) tsʷ(ː) sʷ(ː)
D   ts'    ts(ː) dz s(ː) z
Lak ts'    ts(ː)    s(ː) z
Kh  ts'    ts(ː)    s    z

Code: Select all

SHIBILANTS
N   tʃ'    tʃ     dʒ  ʃ
Av  tʃ'(ː) tʃ(ː)      ʃ(ː) ʒ
An  tʃ'(ː) tʃ(ː)      ʃ(ː) ʒ tʃʷ'(ː) tʃʷ          ʃʷ(ː) ʒʷ
Ts  tʃ'    tʃ         ʃ      tʃʷ'    tʃʷ          ʃʷ
Lz  tʃ'(ː) tʃ(ː) (dʒ) ʃ(ː) ʒ tʃʷ'    tʃʷ(ː) (dʒʷ) ʃʷ(ː) ʒʷ
D   tʃ'    tʃ(ː)      ʃ(ː) ʒ
Lak tʃ'    tʃ(ː)      ʃ(ː) ʒ tʃʷ'    tʃʷ(ː)       ʃʷ(ː) ʒʷ
Kh  tʃ'    tʃ(ː)  dʒ  ʃ    ʒ

Code: Select all

VELARS
N   k'    k    g           kʲ' kʲ gʲ
Av  k'(ː) k(ː) g     x(ː)
An  k'(ː) k(ː) g    (x)             kʷ'(ː) kʷ(ː) gʷ xʷ
Ts  k'    k    g                    kʷ'    kʷ    gʷ
Lz  k'    k(ː) g                    kʷ'    kʷ(ː)
D   k'    k(ː) g     x(ː) ɣ         kʷ'    kʷ(ː) gʷ xʷ(ː) ɣʷ
Lak k'    k(ː) g     x(ː)           kʷ'    kʷ(ː) gʷ xʷ(ː)
Kh  k'    k(ː) g kx (x)  (ɣ)

Code: Select all

UVULARS
N   q'    q    χ    ʁ
Av   q'ː   qː  χ(ː) ʁ
An  q'(ː) q(ː) χ(ː) ʁ        qʷ'(ː) qʷ(ː) χʷ(ː) ʁʷ
Ts  q'    q    χ             qʷ'    qʷ    χʷ
Lz  q'(ː) q(ː) χ(ː) ʁ        qʷ'(ː) qʷ(ː) χʷ(ː)
D   q'    q(ː) χ(ː) ʁ        qʷ'    qʷ(ː) χʷ(ː) ʁʷ
Lak q'    q(ː) χ(ː) ʁ        qʷ'    qʷ(ː) χʷ(ː)
Kh         qː  χ    ʁ qχ' qχ 

Lz  qˤ'(ː) qˤ(ː) χˤ(ː)    qʷˤ'(ː) qʷˤ(ː) χʷˤ(ː)
D   qˤ'    qˤ(ː) χˤ(ː) ʁˤ qʷˤ'    qʷˤ(ː) χʷˤ(ː) ʁʷˤ

Code: Select all

LARYNGEALS
N    ʡ  ħ ʢ      ʔ h ɦ
Av      ħ ʕ      ʔ h
An  (ʡ) ħ        ʔ h
Ts      ħ ʕ      ʔ h
Lz   ʡ  ħ   ʡʷ   ʔ h  ʔʷ    ʔˤ hˤ
D       ħ ʕ   ʕʷ ʔ h     hʷ    hˤ
Lak (ʡ) ħ        ʔ h
Kh     (ħ ʕ)       h

Code: Select all

SONORANTS
N   m    n    r l    j w
Av  m    n    r l    j w
An  m    n    r l    j w
Ts  m    n    r l    j
Lz  m(ː) n(ː) r l(ː) j w u̯
D   m    n    r l    j w
Lak m    n    r l    j w
Kh  m    n    r l    j

So Lezgic is clearly the most phonologically complex branch, while the general lack of labialized consonants in Nakh and Avar seems notable. Avar–Andic seems to be characterizable by presence of /kʼː/ and lack of /pʼ/. Lak and Dargwa seem to share /xʷː/ but not really anything else.

Starostin's PNEC inventory is larger still than any of these, pretty much an extended version of the main phonological tendencies: it has labialized versions for everything except labials and sonorants; geminates for all fricatives, affricates & back stops; a systematic voiced stop / voiceless fricative contrast (including *ɢ and variants); and a third sibilant series that he reconstructs as *tɕ, which merges either with the *ts or the *tʃ series everywhere.

Only Chechen has an unusually large vowel inventory, the other Nakh languages basically have the garden variety /a e i o u/ with a little bit of weirdness sprinkled on top. However, it's pretty obvious that the Chechen vowel inventory is due in large part to Germanic style umlaut

OK, scratch that.

[WeepingElf]

There were people, and thus languages, there around 3000 BC, so why shouldn't they have spoken Proto-NWC, Proto-NEC and Proto-Kartvelian?

PNWC and PKv, why not. NEC looks older than that however.

I don't know of any pressing reason to "pull away" any of the NEC branches (Burushaski etc. comparisons are probably not solid enough for that), but I think there might be a "push" argument: over the last 4000 years, there have been new languages forcing their way into the Caucasus at a rate of roughly about a lineage per millennia: Armenian, Ossetic, Kipchak, Kalmyk, Russian. Does it make sense to think that this was preceded by a multimillennial period of stability? If not, does it make sense to assume that presumable Neolithic / Chalcolithic arrivals would just have disappeared later on? (E.g. does later steppe peoples having horses make enough of a difference on either?)

I think horses do make enough of a difference. Without horses, IE wouldn't be as huge as it is today, after all. I can't say anything about the time depth of NEC. What concerns the vexing typological similarities between NEC and Burushaski, I have been fancying from time to time that they are the last surviving branches of an old family once spoken in the Iranian highlands, but there doesn't seem to be much to back up such a connection. Of course, it is possible, even likely, that some earlier lineages in the Caucasus have vanished completely.

if there was such a substratum at all (…) IMHO the main reasons to assume a substratum are (1) that the PIE and PU speakers are genetically quite different (e.g., different Y-DNA haplogroups - with PU speakers more similar to Siberians than PIE speakers), so it seems a bit as if a language shift has happened here, probably with the later PIE speakers, i.e. the ancestors of the Yamnaya people, and (2) that PIE is pretty much an "odd man out" in the Mitian group, while PU looks much like a typical Mitian language.

Hmm. On the other hand, I would suspect Uralic has itself some kind of a Paleosiberian substrate. This could be enough to explain (1). If there was a substrate with similar typology also further east, and/or if some features of Ural-Altaic and/or Uralo-Siberian typology might be rather recent areal features, that could explain (2). It might be mistaken entirely to use modern Caucasic as an explanation for typological features of PIE 5000 years ago. IE or pre-IE influence in Caucasic could have happened too!

Indeed, the Caucasian languages may have changed a lot since then. As for the Paleosiberian substratum in Uralic, at least Yukaghir, Chukotko-Kamchatkan and especially Eskimo-Aleut show good signs of being related to IE and Uralic in their morphology, and it seems as if the "Micro-Altaic" languages can be counted in as well. So the genetic and linguistic relationships seem to match, more or less, here, while IE is both genetically and typologically divergent, which IMHO makes a substratum in IE IMHO likelier than one in Uralic. But this throws up the question why did the ancestors of the PIE people shift to a Mitian language. Language shifts usually are motivated by the target language being associated with a more sophisticated culture, or having social prestige. We tend to assume that the Neolithic PIE speakers were more advanced than the Mesolithic PU speakers, but before the former adopted farming (mostly pastoralism, but also some agriculture in the western part of their domain), the forest people may have been more prosperous than the steppe people because they lived on richer foraging grounds.

I haven't seen reconstructions of Proto-NWC and Proto-NEC yet, so I can't say anything about their former typology.

For a starting point, here's what Starostin cites as the branch inventories, plus the Avar, Khinalug and Lak inventories from Wikipedia. Segments in parentheses are dubious or loanword phonemes, segments with (ː) have a geminate/fortis equivalent.

[NEC consonant charts snipped]

So Lezgic is clearly the most phonologically complex branch, while the general lack of labialized consonants in Nakh and Avar seems notable. Avar–Andic seems to be characterizable by presence of /kʼː/ and lack of /pʼ/. Lak and Dargwa seem to share /xʷː/ but not really anything else.

Starostin's PNEC inventory is larger still than any of these, pretty much an extended version of the main phonological tendencies: it has labialized versions for everything except labials and sonorants; geminates for all fricatives, affricates & back stops; a systematic voiced stop / voiceless fricative contrast (including *ɢ and variants); and a third sibilant series that he reconstructs as *tɕ, which merges either with the *ts or the *tʃ series everywhere.

If the protolanguage phoneme inventory is larger than any of the daughters', there is IMHO probably something wrong with it, though it may just be some phoneme splits being missed, and there is a chance that all languages have simplified in some way or other, especially if the inventory was huge to start with, though the more diverse a language family is, the more likely it is that some branches have reduced and others have expanded the inventory.

[Tropylium]

We tend to assume that the Neolithic PIE speakers were more advanced than the Mesolithic PU speakers, but before the former adopted farming (mostly pastoralism, but also some agriculture in the western part of their domain), the forest people may have been more prosperous than the steppe people because they lived on richer foraging grounds.

Interesting suggestion. Could be true on a per capita basis. Lower population densities in the north might counter much of this; in Central Asia and Siberia that could be however in turn countered by the major Siberian rivers and the West Siberian Glacial Lake before them as transport channels that can unite people across large territories. This does not work in the European steppe though. How far east would it be reasonable to draw pre-PIE?

If the protolanguage phoneme inventory is larger than any of the daughters', there is IMHO probably something wrong with it

Correct, that's what I also think is the clearest problem with Moscow School-type reconstructions. There are 131 different main consonant correspondences asserted, and then 57 more for word-medial consonant clusters. Even if there were a decent amount of cognate vocabulary from all branches, already a normal distribution of consonant frequencies will ensure that a lot of these will be in many branches based on just one example.

I should also remind that this still does not necessarily invalidate any of the data itself. If we counted between a bunch of modern IE languages, it would be easy get together 100+ different overall consonant correspondence patterns. Most of this number just would be due to deep conditional sound changes, or just pure one-off irregularities. Perhaps a case like Indo-Iranian *ȷ́ʰŕ̥da- 'heart' could end up counted as evidence for "a rare phoneme *ḱʰ", or a case like Germanic *fedwor for "a rare phoneme *pʷ". Also, good luck discovering e.g. Verner's Law without having Ancient Greek and Sanskrit stress to consult, or laryngeal aspiration in Indo-Iranian without having Anatolian (and more productive ablaut in older IE) for establishing the existence of laryngeals.

[WeepingElf]

We tend to assume that the Neolithic PIE speakers were more advanced than the Mesolithic PU speakers, but before the former adopted farming (mostly pastoralism, but also some agriculture in the western part of their domain), the forest people may have been more prosperous than the steppe people because they lived on richer foraging grounds.

Interesting suggestion. Could be true on a per capita basis. Lower population densities in the north might counter much of this; in Central Asia and Siberia that could be however in turn countered by the major Siberian rivers and the West Siberian Glacial Lake before them as transport channels that can unite people across large territories. This does not work in the European steppe though. How far east would it be reasonable to draw pre-PIE?

I'd put it in the area of the lower Volga and Ural rivers, but it could have originated further east. After all, Mitian probably came from the east; I'd guess that Proto-Mitian was spoken near Lake Baykal at the end of the ice age.

If the protolanguage phoneme inventory is larger than any of the daughters', there is IMHO probably something wrong with it

Correct, that's what I also think is the clearest problem with Moscow School-type reconstructions. There are 131 different main consonant correspondences asserted, and then 57 more for word-medial consonant clusters. Even if there were a decent amount of cognate vocabulary from all branches, already a normal distribution of consonant frequencies will ensure that a lot of these will be in many branches based on just one example.

Yes. I'd not say that these reconstructions were dead wrong like, for instance, Octaviano's attempts or the claim of some creationists that protolanguages such as PIE or Proto-Sino-Tibetan were created at the Confusion of Tongues at Babel - just that there is something wrong with them which allows for improvement.

I should also remind that this still does not necessarily invalidate any of the data itself. If we counted between a bunch of modern IE languages, it would be easy get together 100+ different overall consonant correspondence patterns. Most of this number just would be due to deep conditional sound changes, or just pure one-off irregularities. Perhaps a case like Indo-Iranian *ȷ́ʰŕ̥da- 'heart' could end up counted as evidence for "a rare phoneme *ḱʰ", or a case like Germanic *fedwor for "a rare phoneme *pʷ". Also, good luck discovering e.g. Verner's Law without having Ancient Greek and Sanskrit stress to consult, or laryngeal aspiration in Indo-Iranian without having Anatolian (and more productive ablaut in older IE) for establishing the existence of laryngeals.

Yep.

[Tropylium]

I don't know of critiques by others because I assume most Algonquianists aren't even aware of the publications, which were a few years ago in the Journal of Language Relationship (which is...... I mean it's got what look to be SOME reasonable articles among the more insane ones, but it's not like it's something a lot of non-long range comparisonists are gonna be aware of or reading). And if any of them are aware of them, they presumably looked at them and figured they had better things to do.

I read JoLR regularly; the non-long-range content — which is most of it — is perfectly sane, the review and method papers are roughy sane, and I don't see any reason to a priori ignore work appearing there on "established long-range families" like stuff on Afrasian etymologies. The most recent issue is a special issue on Old Chinese, known to be a difficult topic, but the list of authors is full of established experts and I don't think I see reason to malign any of it as "insane". This would seem to be one of the more probable places to find real achievements in long-range comparison published in.

(In terms of method papers, G. Starostin's Macro-comparative linguistics in the 21st century is a pretty good read really.)

As for why I think it's basically garbage, I'll just mention a few points. First, there's some of the usual red flags. The Proto-Algonquian-Wakashan (PAW) vowel inventory is huge compared to any of the daughters (…) A number of sets have undefined vowels (the good ol' "V") or one or more of the consonants are unclear.

I'm repeating myself, but while this is a good reason to doubt the phonological reconstruction, it's not that much of a reason to ignore etymological comparisons entirely. Discovering nontrivial etymologies tends to precede reliable reconstructions for them by a long while.

Nikolaev unfortunately relies on Proulx's reconstructions for Proto-Algic (…) just browsing through and looking at some of the correspondence sets, they're riddled with unmotivated segmentations out of whatever material is necessary to remove in order to have something that he can actually compare to the other languages, inaccurate segmentation of words, misanalysis of words with known etymologies, words known to be unrelated being compared in the same set, improper interpretation of his sources including mistaken phonemic/phonological interpretation, unmotivated appeals to "reduplication" or "analogy" that can't possibly be correct, etc. etc.

Sanity check: if anyone's knowledge of Proto-Algic is patchy so far, and if Nikolaev mostly does not directly deal with Proto-Algonquian, how would you know Proto-Algic segmentations or etymologies to be incorrect?

The first example in the first paper where there seems to be something funny going on with morphology is entry #2 (in fact the 3rd entry overall after #1a and 1b): 'ashes', given as Nivkh *phl-ɨŋg, Algic *p(el)-enekw-. This is noted to be based on a precedent with an Algic correspondence with Wiyot pł- ([pɬ]?) ~ Yurok p- ~ Alq. *p-, then observing examples with Yurok p-, Alq. *p- corresponding with Nivkh *phl-, and proposing that these correspondences both come from a single source. Looks sane enough to me so far. There is then a follow-up proposal that this would be a compound where the #pl part is cognate with Salishan *pəqʷ' 'powder', the #Vnkw part with Wakashan *ʔan-, *ʔanakʷ- 'fire' … which yeah looks like garbage. But this is a proposal for morphological segmentation in his PAW, not in Algic (let alone Algonquian). If we throw it out, we are still left with the Nivkh–Algic comparison that could be provisionally treated as monomorphemic #plenkʷ-.

More morphology is happening in entry #4a, where an Algic *taaɣ-w- is followed with a footnote: "PAlg nominal stems may include one of the 3 thematic formants: ∅, *w and *y (…)". Other entries I see such formants claimed for include #11a (*te(e)k-w-l- ~ *čeek-w-r 'heart'; no idea what the *-l-/-r- are doing or based on, and they don't seem to be needed for the PAW comparison either), #38d (*-aačk-w- ~ *-ečk- ~ *-etk- 'head'), #46 (*p(el)ak-w- 'leaf'), #50 (*ihk-w- 'louse') (all *w, I don't see any examples of alleged *y formants). In the absense of a reference it's hard to know if this is an established idea or a new proposal entirely. Even if supposing it is all-new, though, what broadly seems to be happening is that stem-final *-kʷ- is noted to correspond also to non-labialized velars/uvulars in Wakashan. Nikolaev prefers to suggests that this is intrusive labialization in Algic. Perhaps it is; would we have reason to think it cannot be? — Some cases though would also seem to be alternately accountable by loss of labialization in Wakashan in consonant clusters.

Now I agree it's pretty unpleasant to wade thru papers where every second etymology has this kind of a stack of speculation attached, without it clearly being marked off (the Starostin paper I linked also stresses this) or without properly expanding on proposed new analyses of subgroup phenomena. But I also think that some amount of this has been necessary for establishing every nontrivial language family.

i.e. what would be interesting to see is a review of proposals like this that actually engages with the core claims instead of getting stuck on how particular etymologies look bad in various superficial ways. (Obviously, Nikolaev too could do a lot more to facilitate this — for starters, if he doesn't even think they're clearly related, everything involving Salishan and Chukotko-Kamchatkan should probably be presented separately from the core W/N/A correspondences.)

[jal]

Language shifts usually are motivated by the target language being associated with a more sophisticated culture, or having social prestige.

That we know of. There's no way of telling why pre-historic people gave up a language to speak another. (Also, given the way the Y-chromosome travelled compared to the mitochondrian DNA, I wouldn't be surprised if intermerriage didn't play a part as well; after all, there weren't that many people back then to begin with.)


JAL

[WeepingElf]

Language shifts usually are motivated by the target language being associated with a more sophisticated culture, or having social prestige.

That we know of. There's no way of telling why pre-historic people gave up a language to speak another. (Also, given the way the Y-chromosome travelled compared to the mitochondrian DNA, I wouldn't be surprised if intermerriage didn't play a part as well; after all, there weren't that many people back then to begin with.)

Intermarriage surely played a role, yes. "Pure races" have never been a fact, only a racist fantasy. What regards the PIE and PU speakers, the PU speakers apparently were "Eastern Hunter Gatherers" who are a mix of "Western Hunter Gatherers" and "Ancient North Eurasians", while the PIE speakers apparently were a mix of the same (though perhaps with a higher percentage of "Western Hunter Gatherers") and "Caucasus Hunter Gatherers". Or so say the geneticists. Fact is that the Y-DNA profiles of PIE and PU speakers are quite different, with the PU speakers being quite similar to indigenous Siberians in this regard. In many other regards, of course, most Uralic speakers are more like IE speakers; perhaps we are dealing with a combination of Siberian males and European females here, or whatever. Genetic data are not easy to interpret, as is shown by the fact that different people arrive at vastly different conclusions from the same data. So one should not put too much faith in those pesky genes, especially not when considering languages.

My point was mainly a linguistic observation, though: If you compare the various Mitian protolanguages, PU looks more "typically Mitian" than PIE. Hence, it seems likely that PU is more conservative than PIE. And in some of the points in which PIE looks "odd" from a Mitian perspective, it looks like a Caucasian language - an observation already made by C. C. Uhlenbeck in the 1930s, leading him to the Caucasian substratum hypothesis. However, there are other "oddities" in PIE which don't look "Caucasian", such as the existence of only one sibilant phoneme, *s, which is so frequent that one may suspect a merger of several different phonemes here. Caucasian languages have richer s(h)ibilant inventories, especially the NWC ones.

[Tropylium]

There are 131 different main consonant correspondences asserted, and then 57 more for word-medial consonant clusters. Even if there were a decent amount of cognate vocabulary from all branches, already a normal distribution of consonant frequencies will ensure that a lot of these will be in many branches based on just one example.

So I looked over an example: the cognates to *pC, *bC cluster in Nakh. It's indeed not in good shape: there are broad generalities like *pħ corresponding to uvulars, *pχ corresponding to velars (Nakh also has *x > *χ in more general) and *ps corresponding to sibilants elsewhere, but Starostin's claim that these would come from certain very specific labialized consonants seems to be unfounded. Even per him most labialized consonants do not turn into clusters like these.

Regardless it does not seem possible to dismiss entirely the etymologies either: they include several well-distributed basic NEC words like 'dog', 'five', 'vein', 'arm', 'eye', 'arm' (Nakh *pħu, *pχi, *pχa, *pħars, *bʡa). I would not put much stock on the binary examples at least: Nakh *pħāʁi 'pot' ~ Lak /qquqqu/ 'jug'; Nakh *bʡāʁum 'pillar' ~ Akusha /qʼaˤqʼ/ 'hill'; Chechen /bʕēžū/ 'steppe spider' ~ Lezgic *qʼˤulčʼ(a) 'maggot/moth'.

For that matter, already some of the sub-branch reconstructions give impressions that they might be overengineered. Clearly not majorly, but labialization seems suspiciously unstable in all branches, and there are individual oddities like Lezgian *tɬʼ reconstructed for a /kʼ/ ~ /qʼ/ correspondence (with /kʼ/ even in Archi, which is otherwise the only Lezgian language retaining lateral fricatives).

[Nortaneous]

there are individual oddities like Lezgian *tɬʼ reconstructed for a /kʼ/ ~ /qʼ/ correspondence (with /kʼ/ even in Archi, which is otherwise the only Lezgian language retaining lateral fricatives).

This isn't too big a deal. In some sense, proto-languages are abstract notation devices; PAn *R is shorthand for a vast array of correspondence sets. (In another sense, proto-languages are approximations of real languages, but the precise phonetics may not be able to be approximated -- we have decent grounds to guess that PAn *t *d were [t̪ d], but we can't say much about *R other than that it was probably not an obstruent. And the abstract notation device has to be pretty well in place before it can start being treated as a real language, I think.) Writing *tɬʼ instead of *Kʼ is just a notational choice. A questionable one, to be sure -- it's implicitly assuming far too much with a far higher degree of confidence than is warranted -- but for whatever reason it seems that they want to write their proto-languages in IPA.

[Howl]

If the protolanguage phoneme inventory is larger than any of the daughters', there is IMHO probably something wrong with it

Correct, that's what I also think is the clearest problem with Moscow School-type reconstructions. There are 131 different main consonant correspondences asserted, and then 57 more for word-medial consonant clusters. Even if there were a decent amount of cognate vocabulary from all branches, already a normal distribution of consonant frequencies will ensure that a lot of these will be in many branches based on just one example.

I should also remind that this still does not necessarily invalidate any of the data itself. If we counted between a bunch of modern IE languages, it would be easy get together 100+ different overall consonant correspondence patterns. Most of this number just would be due to deep conditional sound changes, or just pure one-off irregularities. Perhaps a case like Indo-Iranian *ȷ́ʰŕ̥da- 'heart' could end up counted as evidence for "a rare phoneme *ḱʰ", or a case like Germanic *fedwor for "a rare phoneme *pʷ". Also, good luck discovering e.g. Verner's Law without having Ancient Greek and Sanskrit stress to consult, or laryngeal aspiration in Indo-Iranian without having Anatolian (and more productive ablaut in older IE) for establishing the existence of laryngeals.

I think it is also a methodological problem of many long-range reconstructions. In the history of real language families, we often see that multiple phonemes correspond to multiple phonemes in a complex way due to conditional sound changes. For example, Ancient Greek t corresponds to PIE *t and *kʷ, but PIE *kʷ can also correspond to Ancient Greek p and k. But that concept seems to be almost totally alien to the worst lumpers. For example, how many conditional sound changes do the various versions of Nostratic have?

My point was mainly a linguistic observation, though: If you compare the various Mitian protolanguages, PU looks more "typically Mitian" than PIE. Hence, it seems likely that PU is more conservative than PIE. And in some of the points in which PIE looks "odd" from a Mitian perspective, it looks like a Caucasian language - an observation already made by C. C. Uhlenbeck in the 1930s, leading him to the Caucasian substratum hypothesis. However, there are other "oddities" in PIE which don't look "Caucasian", such as the existence of only one sibilant phoneme, *s, which is so frequent that one may suspect a merger of several different phonemes here. Caucasian languages have richer s(h)ibilant inventories, especially the NWC ones.

However, there is one important counterpoint. PIE is the only Mitian language family that has been attested well before before the Middle Ages. All the other language families only have late attestations. And the modern PIE languages look way more "typically Mitian" than the PIE we reconstruct from Anatolian, Ancient Greek, Sanskrit etc. So some of that 'Mitian' typology may simply have been due to shared drift.

[Whimemsz]

.

[Nortaneous]

the magical back vowel ("*A") which conditions uvularization in Nivkh even though it leaves no other trace (e.g., "belly, stomach"), and in others before a front vowel ("*E") which leaves velars alone in Nivkh (e.g., "heart")

Not so unreasonable on Nivkh-internal grounds -- internal reconstruction shows that Nivkh underwent dramatic vowel loss, which probably had the effect of making uvulars phonemic. (If you ignore recent loans, the velar/uvular contrast is found mostly [entirely?] in final position.) Postulating vowels for 'PAW' that were preserved until relatively recently in Nivkh and lost without recoverable compensation everywhere else, however, is questionable.

(Can the restructuring of the Nivkh phonology -- intervocalic lenition, vowel loss, etc. -- be dated with loanword evidence?)