Nostratic, Eurasiatic, Mitian, ...

[KathTheDragon]

If you take the plain voiced stops as the last stop series that PIE gained, why couldn't you explain their distribution simply by the distribution of their originating condition?

While this is an attractive idea, this suffers from exactly the same problem from the two-dorsal hypothesis: there is no clear originating condition. All three kinds of stops can appear in the same kinds of environments (e.g. any root with only one stop), so it's difficult to posit a scenario whereby the three stop series can come from two ancestral stop series without bloating the inventory somehow; to me, and also WeepingElf, I imagine, adding a third stop series seems the least egregious option.

[WeepingElf]

I am like you on this, Kath. There were three velar obstruent series in PIE, as in the standard reconstruction (I call them "front", "back" and "labialized"). And what regards the Uralic (and probably Proto-Mitian) palatal series, I used to consider it plausible that it merged with the dental series in PIE, but it could just as well have merged with the front velars! This would be phonetically more plausible, and help explain why the front velars are more frequent in PIE than the back velars; that would of course also mean that PIE0 (the pre-GVC stage) still had a separate palatal series, but I am more and more leaning towards the idea that PIE0 and Proto-Indo-Uralic were the same thing anyway.

Alas, as I already said a few says ago, the only way to definitely answer questions of this kind is to actually establish correspondence sets, but at least, considerations of this kind may tell us where it is promising to look for those in the first place.

[Soap]

@Soap
And how would you interpret it if you assume that neither Indo-European nor early Proto-Uralic had ejectives?

If it were proven that neither family had ever had ejectives, I would admit that I was wrong and look for another explanation for the correspondence between the single stop series in Uralic and the triple stop series in PIE. I'm not sold on the glottalic theory, which posits voiceless ejectives as a part of PIE proper, but I am a strong believer in Indo-Uralic, and I think that voiceless ejectives at some stage of pre-PIE are a good way of explaining the discrepancy between the two stop inventories.

I admit that posing a unilateral collapse of /k k' g/ > /k/ in Uralic is unlikely, but such a shift could have happened slowly over time, and been polyconditional. For example, proto-Uralic may have turned some of the stops into fricatives (proto-Uralic is commonly reconstructed with a four-way s/š/ś/x contrast) or nasals.

I think Indo-Uralic is a useful first step towards Mitian and should be established first, but I think Indo-Uralic reconstruction is best done with having the state of affairs in likely further relatives in mind. Such considerations lead to the working hypothesis that the Proto-Uralic phonology is more conservative than the PIE one. Also, the PU agglutinating morphology looks more archaic than the PIE fusional one.

I agree. I think, though, that the shared parent language did have at least a voicing contrast in stops, which had been already lost by the time of proto-Uralic, and may have had an ejective series as well.

I am like you on this, Kath. There were three velar obstruent series in PIE, as in the standard reconstruction (I call them "front", "back" and "labialized").

I think the question there was about modes of articulation, rather than place. Probably all of us agree that there were three different modes of articulation at least by the time of PIE proper.

-----
Also, to continue an earlier discussion, I found this in a draft that I never finsihed writing up:

Next up is CK. Where are the m/t pronouns? Well, there M and T in Chukchi, if you don't mind them being the final consonants in /G@m/ and /G@t/.

But also word-initially in the plural forms muri "we" and turi "you", which is evidence in favor of a previously existing pair of single morphemes /m/ and /t/.

[WeepingElf]

I am like you on this, Kath. There were three velar obstruent series in PIE, as in the standard reconstruction (I call them "front", "back" and "labialized"). And what regards the Uralic (and probably Proto-Mitian) palatal series, I used to consider it plausible that it merged with the dental series in PIE, but it could just as well have merged with the front velars! This would be phonetically more plausible, and help explain why the front velars are more frequent in PIE than the back velars; that would of course also mean that PIE0 (the pre-GVC stage) still had a separate palatal series, but I am more and more leaning towards the idea that PIE0 and Proto-Indo-Uralic were the same thing anyway.

Or, though I consider this less likely, the palatal-velar merger may have happened before the PIU velar series split into three. In that case, all three velar emphatics could continue a PIU hushing affricate *č.

[KathTheDragon]

I admit that posing a unilateral collapse of /k k' g/ > /k/ in Uralic is unlikely, but such a shift could have happened slowly over time, and been polyconditional. For example, proto-Uralic may have turned some of the stops into fricatives (proto-Uralic is commonly reconstructed with a four-way s/š/ś/x contrast) or nasals.

Unlikely why? Tocharian did it (eventually)

[Pole, the]

I find it highly unlikely that we'll ever be able to reconstruct the phonological history of PIE, given that the closest cousins of Indo-European languages have probably been wiped out / replaced by Indo-Europeans themselves.

As an alternative to ejectives, we might consider implosives for PIU, yielding the PIE plain voiced stops, and the PU nasals.

I kinda like the idea of Proto-Indo-Whatever having /n/ as its only phonemic nasal with later /ɓ ɗ ɠ → m n ŋ/ in Uralic (Uralo-Siberian?) and /ɓ → m/ in Pre-Indo-European.

[Tropylium]

it's been recently shown that e.g. Proto-Mongolic, Proto-Tungusic and Proto-Korean should be reconstructed with pharyngeal harmony (*u̘ ~ *ʊ̙ etc.), as still in Khalkha, instead of palatal harmony as in Uralic (*y ~ *u etc.). Turkic has also been proposed to have shifted from pharyngeal to palatal harmony recently; there are lines of evidence such as "front vowels" triggering initial voicing in Oghuz (= Turkic-Azeri-Turkmen), which makes zero phonetic sense, but which starts working quite well with +ATR vowels triggering voicing instead. The same also happens to be attested from Armenian right nearby.

Sounds interesting. Can you give the relevant references for weekend reading?

Starting point for reading: Seongyeon Ko's thesis on pharyngeal harmony. On extending this to Turkic, Bert Vaux has several papers on this, maybe with [atr] harmony in Altaic languages as the best starting point.

[gach]

Starting point for reading: Seongyeon Ko's thesis on pharyngeal harmony. On extending this to Turkic, Bert Vaux has several papers on this, maybe with [atr] harmony in Altaic languages as the best starting point.

Thanks. I have a feeling that I may have added something like these on my to-read list once and then totally forgot about it.

[WeepingElf]

I think Indo-Uralic is a useful first step towards Mitian and should be established first, but I think Indo-Uralic reconstruction is best done with having the state of affairs in likely further relatives in mind. Such considerations lead to the working hypothesis that the Proto-Uralic phonology is more conservative than the PIE one. Also, the PU agglutinating morphology looks more archaic than the PIE fusional one.

I agree. I think, though, that the shared parent language did have at least a voicing contrast in stops, which had been already lost by the time of proto-Uralic, and may have had an ejective series as well.

I used to think that way, too; the reason why I no longer do is that language families such as Yukaghir and Eskimo-Aleut are much more like Uralic in this regard, which makes it more plausible that Uralic is more conservative than IE. An Indo-Uralic system resembling "glottalic PIE" would require either the parallel loss of the threefold stop grades in every other Mitian language family (with the possible exception of Altaic, if Starostin's reconstruction is correct), or an innovation in Proto-Indo-Uralic and a subsequent reversal of that development in Uralic. Neither is a parsimonious hypothesis.

Unless, of course, Uralic is more closely related to the far eastern Mitian languages than to IE and the Indo-Uralic node is invalid, in which case it is of course possible that Proto-Mitian was like PIE and the non-IE Mitian (or Uralo-Siberian, to set Altaic aside) languages form a valid node and have innovated; but I doubt that. Neither geography nor genetics play ball here.

I find it highly unlikely that we'll ever be able to reconstruct the phonological history of PIE, given that the closest cousins of Indo-European languages have probably been wiped out / replaced by Indo-Europeans themselves.

The closest relatives probably indeed met that fate; but some language family will be the closest living kin, and Uralic is IMHO a good candidate.

Also, there is the technique of internal reconstruction, which allows to get glimpses into the prehistory of a language without looking at relatives, from patterns found in the language itself.

As an alternative to ejectives, we might consider implosives for PIU, yielding the PIE plain voiced stops, and the PU nasals.

I kinda like the idea of Proto-Indo-Whatever having /n/ as its only phonemic nasal with later /ɓ ɗ ɠ → m n ŋ/ in Uralic (Uralo-Siberian?) and /ɓ → m/ in Pre-Indo-European.

I don't think so. The other Mitian languages are more like Uralic in terms of nasals, so proposing a Proto-Basque-like one-nasal system is, apart from its typological uncommonness, not a very plausible move. It seems as if PIE has just lost its palatal and velar nasals, probably by merging them into /n/.

[WeepingElf]

Thinking about the position of IE within Mitian I came up with an idea which I am not (yet) convinced of, but which IMHO needs to be considered.

Typologically, IE is pretty much the "odd man out" within Mitian, and I think that where IE differs from the other Mitian languages, it alone has innovated. This doesn't say anything about when it branched off from where in the whole family tree. IE seems to be a Mitian language that was at least partly drawn into a Caucasian/Near Eastern language area with emphatic/glottalized stops, ablaut, fusional morphology and the like, altering its structure from one that once was more like the other MItian languages.

So far, I have been entertaining the notion that IE is probably most closely related to Uralic, and I am still entertaining that notion now. However, it cannot be ruled out IMHO that this is actually wrong and IE is most closely related to Altaic (if that's a valid node within Mitian at all!). If you look at the Proto-Altaic reconstruction by Starostin, Dybo and Mudrak (which I am not very convinced of, but maybe it is at least partly correct), it looks like halfway between the Uralo-Siberian and the IE type. One one hand, it is similar to IE with its aspirated/tenuis/voiced triads as opposed to the Uralo-Siberian voiceless stop/voiced fricative pairs; and on the other hand, it is like Uralo-Siberian with its full-fledged palatal series and full set of nasals in all four places of articulation.

Thus, maybe Proto-Mitian split first into a northern (Uralo-Siberian) and a southern (Indo-Altaic) branch which then fanned out west and east along the different ecological zones: Uralo-Siberian in the taiga, Indo-Altaic in the steppe. In such a model, it may also be the case that Indo-Altaic is more conservative than Uralo-Siberian.

Yet, I feel that in terms of several other things such as morphology, IE is closer to Uralic than to Altaic, so I am not convinced by my own new idea yet, but it has its strong points and is IMHO worth discussion. Also, I have my doubts against the validity of the Altaic node, feeling that Turkic is in some ways (e.g., the personal endings on verbs) more like IE and Uralic than like Mongolic and Tungusic. Perhaps this is simply a case where the family tree model breaks down, and we are dealing with intersecting isoglosses of which some connect IE with Uralic and others with Altaic (or at least Turkic; Mongolic and Tungusic seem to have much less in common with IE and Uralic than Turkic has). Typology can be deceptive!

[mèþru]

Altaic is probably also a specahbund. Maybe you could have Mongolo-Tungusic-Siberian as one branch and Indo-Uralo-Turkic as another.

[KathTheDragon]

As an alternative to ejectives, we might consider implosives for PIU, yielding the PIE plain voiced stops, and the PU nasals.

I kinda like the idea of Proto-Indo-Whatever having /n/ as its only phonemic nasal with later /ɓ ɗ ɠ → m n ŋ/ in Uralic (Uralo-Siberian?) and /ɓ → m/ in Pre-Indo-European.

I don't think so. The other Mitian languages are more like Uralic in terms of nasals, so proposing a Proto-Basque-like one-nasal system is, apart from its typological uncommonness, not a very plausible move. It seems as if PIE has just lost its palatal and velar nasals, probably by merging them into /n/.

I wasn't proposing that anyhow. I was proposing at least /m n/. In any case, I am doubtful of any conclusions based on this hypothetical family (seriously, Indo-Uraliac is speculative enough!), so I reject your rejection.

[WeepingElf]

Oh, I am sorry. I did not mean to put anyone down; I just wanted to say that I did not consider the idea particularly good (and I apologize for misunderstanding you as to propose a one-nasal system). No need to be snotty, though. We are all just talking about things which we probably never could actually work out even if we work together in perfect harmony. As Tropylium has laid out here, this could be a major academic linguistics research project keeping several professionals busy. Of course, one or the other of us could be the next Michael Ventris, but there is no way knowing; at least, we can try our best to avoid becoming another bunch of Octavianos

As for the suggested hypothesis, it cannot of course be ruled out. It is indeed possible that Proto-Indo-Uralic (or Proto-Mitian, or whatever) had a set of implosive stops which gave the plain voiced stops (or whatever) in PIE and shifted to nasals in PU; yet, almost every Mitian protolanguage except PIE has a full set of nasals of the Uralic kind, so it is more parsimonious to assume that PIE innovated here and the others including Uralic preserve the Proto-Mitian situation which would also have prevailed in Proto-Indo-Uralic.

It is quite clearly PIE that is the "odd man out" in this cluster (my term for a unit which is hard to decide whether it is a family, a Sprachbund or yet something else), so you get the most economic trajectories of change if one assumes that wherever PIE deviates from the "Common Mitian" type, it innovated - perhaps due to a substratum or Sprachbund it was drawn into, perhaps a Near Eastern or Caucasian one. PIE seems to be to Mitian something like what Insular Celtic is to IE. Except that in the latter case, we of course know what happened (though we still don't really know why Insular Celtic is so weird - speculations abound here).

[Tropylium]

I wish to resume here the discussion of the origin of the PIE stop grades we had on the Great PIE Thread and which led me to start this thread. (…) Now if the PIU/Mitian inventory was like Fortescue's Uralo-Siberian one, *t' could have emerged from the affricate *c (this would have to have gone somewhere as PIE doesn't have any affricate).

As you've already mentioned, this kind of thinking seems unlikely to go anywhere much, if it's not based in actual sound correspondences. And thinking that Uralic is "generally archaic" seems like a poor starting point as well. The bisyllabic CVCCV root structure and general agglutination are maybe likely to be more archaic than the monosyllabic (s)CCVCC and fusionality in IE, but it does not mean that everything else would be archaic, too.

There is no PU *c /ts/, by the way; this is merely a typewriter-friendly alternate transcription from the 60s thru 80s, either for *ć (in which case *č gets rewritten → *z) or *č (in which case *ć gets rewritten→ *cj).

I don't think there's any actual evidence for PIE *d ~ PU *č or *ć to go around either, and I suspect the affricates in Uralic are not archaisms to begin with. *č and *ć could be from e.g. *Cs and *Cś (including *ns, *nś > *nč, *ńć). Hyllested has suggested there might be instead a correspondence with PIE *Kʷ ~ PU *č: e.g. PIE *h₃negʷʰ- ~ PU *künčə 'nail'; PIE *h₂angʷʰis 'snake' ~ PU *kunčV 'worm'. If so, this could be through either some earlier palatalization process in Uralic (*kʷe > *če), some velarization process in IE (*č > *[ʈʂ] > *[ʈʂʷ] > *kʷ), or maybe some kind of cluster simplification (*kr > *kw > *kʷ; *kr > *tr > *č). But it would be nice to find more than one example.

Related to the last possibility, I wonder what PIU phonotax could have been like. Are all word-initial consonant clusters in PIE secondary, or has Uralic secondarily simplified some of them? There's a precedent in Middle Indo-Aryan, which simplified almost all Old IA consonant clusters (e.g. pr tr kr > p t k; rp rt rk > pp tt kk; sp st sk > (p)pʰ (t)tʰ (k)kʰ).

I mean like, one would come up with an Indo-Uralic correspondence set, e.g. (I don't know yet whether this set is valid or not) PIE *dh : PU *ð. But what was the PIU sound like? Taking a look at languages such as Eskimo-Aleut, or Fortescue's Uralo-Siberian reconstruction attempt, one may guess that the Proto-Indo-Uralic sound may have been *ð.

Let's suppose there is such a correspondence. The first thing I'd look into is typology: is d > ð or ð > d more likely? The more common option would seem to be the first. The second is common only in three specific cases:
– in fortition environments (word-initially in West Germanic)
– when combined with θ > t (Ossetic, Yaghnobi, Continental Nordic, Aramaic, standard Hebrew)
– as a kind of hypercorrection (standard Finnish /d/ for earlier /ð/, introduced from non-native pronunciation by Swedish speakers).
*ð in both Uralic and Eskimo-Aleut (and Chukotko-Kamchatkan) might be then simply parallel developments from an earlier *d.

This is all theoretical anyway, though, since the actual data doesn't really have clear recurring correspondences. A few good-looking pairs are:
*dʰ ~ *ð: PIE *h₂we-dʰ- ~ PU *kuða- 'to weave'; ? PIE *Hnedʰ- ~ PU *ńiðä- 'to bind' (neither very consistently attested)
*d ~ *ð: PIE *ḱad- 'to fall' ~ PU *śaða- to rain'
*nt ~ *ð: PIE *h₂ant- ~ PU *eðə- 'front'
*rd ~ *ð: PIE *ḱerd- ~ PU *śVðämə 'heart'
*rgʰ ~ *ð(k): PIE *bʰergʰ- 'high' ~ PU *piðə 'height, length', *pið-kä 'tall, long'

This is heterogeneous enough that at least some of these comparisons are most likely wrong. The first three have also been proposed to be loans instead.

As an alternative to ejectives, we might consider implosives for PIU, yielding the PIE plain voiced stops, and the PU nasals.

This may be more likely, since as per Kümmel, we can find some decent data for this (e.g. PIE *deiḱ- 'to point' ~ PU *näkə- 'to see'). This has the benefit of providing a correspondence for PU *ŋ, which has no obvious counterpart in IE.

On the other hand, PU did not have *ŋ-, while PIE has plenty enough of *g- and *gʷ-. So that would need some kind of an explanation, too. There's *g ~ *k in e.g. PIE *gerh₂ōws 'crane' ~ PU *kurkə 'crane', but this could very easily be onomatopoetic (if inherited, we would probably expect a labiovelar in IE).

Assuming Proto-Mitian *m *n *ń *ŋ has similar problems. Why does Tungusic allow all four word-initially, Turkic none of them, and most others just the front nasals (*m *n and, if applicable, *ń)?

Again, typology suggests that the situation where *-ŋ- in particular was only allowed within a word (in Uralic, Mongolic, Yukaghir, CK, EA) probably develops from an earlier system where there was no **ŋ at all, and after which there was some kind of a word-internal cluster simplification such as *ng > *ŋ (this is how this situation developed in various languages like Germanic, Avestan or innovative parts of Dravidian, after all).

On the other hand, there do not seem to be any examples of either *ŋ or *ŋk in Uralic corresponding to an IE *nK cluster.

Here's some stuff I found on Wikipedia that can serve as a starting point:
Among the sound correspondences which Čop did assert were (1972:162) (…)

Čop's stuff is a decent starting point for data, but if you've ever actually looked at his papers (you can find scans of his compilation work Indouralica online through the usual channels), they suffer from a lot of junk analyses. For one, there's gratuitous segmentation to make things "more comparable"; e.g. he compares Uralic *uwa 'flow, stream' (reliably established) with an internally reconstructed pre-PIE *aw- 'to flow', from which *w-ed- 'water' is then supposedly a derivative. On the other, to some extent there's the usual cherry-picking problem: comparisons between isolated single-branch terms like Finnic *liiva 'sand' ~ Greek λυμα 'dirt' (the former has been rather explained as a loan from Baltic *gleiwā).

[WeepingElf]

I wish to resume here the discussion of the origin of the PIE stop grades we had on the Great PIE Thread and which led me to start this thread. (…) Now if the PIU/Mitian inventory was like Fortescue's Uralo-Siberian one, *t' could have emerged from the affricate *c (this would have to have gone somewhere as PIE doesn't have any affricate).

As you've already mentioned, this kind of thinking seems unlikely to go anywhere much, if it's not based in actual sound correspondences. And thinking that Uralic is "generally archaic" seems like a poor starting point as well. The bisyllabic CVCCV root structure and general agglutination are maybe likely to be more archaic than the monosyllabic (s)CCVCC and fusionality in IE, but it does not mean that everything else would be archaic, too.

Fair. The only way to straighten out these question marks is finding actual correspondence sets. Discussions like this one are at most useful in finding out where to look for those. And you are right that Uralic is not necessarily more conservative than IE. Sure, Fortescue's reconstruction of Uralo-Siberian looks more like PU than like PIE, but we don't know how good it is. We of course have the similarities between Uralic and Eskimo-Aleut which in light of the geographical separation are perhaps best explained as archaisms in these two language families, but we can't be sure. Most likely, there are innovations in PU and archaisms in PIE as well!

There is no PU *c /ts/, by the way; this is merely a typewriter-friendly alternate transcription from the 60s thru 80s, either for *ć (in which case *č gets rewritten → *z) or *č (in which case *ć gets rewritten→ *cj).

I don't think there's any actual evidence for PIE *d ~ PU *č or *ć to go around either, and I suspect the affricates in Uralic are not archaisms to begin with. *č and *ć could be from e.g. *Cs and *Cś (including *ns, *nś > *nč, *ńć). Hyllested has suggested there might be instead a correspondence with PIE *Kʷ ~ PU *č: e.g. PIE *h₃negʷʰ- ~ PU *künčə 'nail'; PIE *h₂angʷʰis 'snake' ~ PU *kunčV 'worm'. If so, this could be through either some earlier palatalization process in Uralic (*kʷe > *če), some velarization process in IE (*č > *[ʈʂ] > *[ʈʂʷ] > *kʷ), or maybe some kind of cluster simplification (*kr > *kw > *kʷ; *kr > *tr > *č). But it would be nice to find more than one example.

Good that you say that; this thing was beginning to turn into an idée fixe in my head, and it is time to say goodbye to it. That's the just way people like Octaviano have managed to deceive themselves!

Related to the last possibility, I wonder what PIU phonotax could have been like. Are all word-initial consonant clusters in PIE secondary, or has Uralic secondarily simplified some of them? There's a precedent in Middle Indo-Aryan, which simplified almost all Old IA consonant clusters (e.g. pr tr kr > p t k; rp rt rk > pp tt kk; sp st sk > (p)pʰ (t)tʰ (k)kʰ).

Fair - Proto-Uralic may have simplified its consonant clusters in a similar way. Again, we need sound correspondences to find out.

I mean like, one would come up with an Indo-Uralic correspondence set, e.g. (I don't know yet whether this set is valid or not) PIE *dh : PU *ð. But what was the PIU sound like? Taking a look at languages such as Eskimo-Aleut, or Fortescue's Uralo-Siberian reconstruction attempt, one may guess that the Proto-Indo-Uralic sound may have been *ð.

Let's suppose there is such a correspondence. The first thing I'd look into is typology: is d > ð or ð > d more likely? The more common option would seem to be the first. The second is common only in three specific cases:
– in fortition environments (word-initially in West Germanic)
– when combined with θ > t (Ossetic, Yaghnobi, Continental Nordic, Aramaic, standard Hebrew)
– as a kind of hypercorrection (standard Finnish /d/ for earlier /ð/, introduced from non-native pronunciation by Swedish speakers).
*ð in both Uralic and Eskimo-Aleut (and Chukotko-Kamchatkan) might be then simply parallel developments from an earlier *d.

We had already discussed the nature of PU *ð on the late, lamented Nostratic-L mailing list. The bottom line was that we simply don't know what it was like. Maybe Proto-Uralic did not have any voiced fricatives at all! There is no *v, only *w; Fortescue's Proto-Uralo-Siberian *v/*w distinction is based on Chukotko-Kamchatkan alone, and that's the most problematic member (which F. later dropped from his hypothesis altogether)! PU *ð and *ð' could have been lateral fricatives (as Bomhard assumes) or whatever, and of *x (*ɣ in older reconstructions and still with Fortescue), we have no good idea what it sounded like at all.

This is all theoretical anyway, though, since the actual data doesn't really have clear recurring correspondences. A few good-looking pairs are:
*dʰ ~ *ð: PIE *h₂we-dʰ- ~ PU *kuða- 'to weave'; ? PIE *Hnedʰ- ~ PU *ńiðä- 'to bind' (neither very consistently attested)
*d ~ *ð: PIE *ḱad- 'to fall' ~ PU *śaða- to rain'
*nt ~ *ð: PIE *h₂ant- ~ PU *eðə- 'front'
*rd ~ *ð: PIE *ḱerd- ~ PU *śVðämə 'heart'
*rgʰ ~ *ð(k): PIE *bʰergʰ- 'high' ~ PU *piðə 'height, length', *pið-kä 'tall, long'

This is heterogeneous enough that at least some of these comparisons are most likely wrong. The first three have also been proposed to be loans instead.

Indeed, they probably aren't all correct. At least PU *śaða- looks like a loan from Indo-Iranian to me.

As an alternative to ejectives, we might consider implosives for PIU, yielding the PIE plain voiced stops, and the PU nasals.

This may be more likely, since as per Kümmel, we can find some decent data for this (e.g. PIE *deiḱ- 'to point' ~ PU *näkə- 'to see'). This has the benefit of providing a correspondence for PU *ŋ, which has no obvious counterpart in IE.

On the other hand, PU did not have *ŋ-, while PIE has plenty enough of *g- and *gʷ-. So that would need some kind of an explanation, too. There's *g ~ *k in e.g. PIE *gerh₂ōws 'crane' ~ PU *kurkə 'crane', but this could very easily be onomatopoetic (if inherited, we would probably expect a labiovelar in IE).

Assuming Proto-Mitian *m *n *ń *ŋ has similar problems. Why does Tungusic allow all four word-initially, Turkic none of them, and most others just the front nasals (*m *n and, if applicable, *ń)?

Again, typology suggests that the situation where *-ŋ- in particular was only allowed within a word (in Uralic, Mongolic, Yukaghir, CK, EA) probably develops from an earlier system where there was no **ŋ at all, and after which there was some kind of a word-internal cluster simplification such as *ng > *ŋ (this is how this situation developed in various languages like Germanic, Avestan or innovative parts of Dravidian, after all).

On the other hand, there do not seem to be any examples of either *ŋ or *ŋk in Uralic corresponding to an IE *nK cluster.

These are valid objections. The problem probably doesn't have an easy solution; otherwise, the academic scholars working on this kind of thing would have already found out!

Here's some stuff I found on Wikipedia that can serve as a starting point:
Among the sound correspondences which Čop did assert were (1972:162) (…)

Čop's stuff is a decent starting point for data, but if you've ever actually looked at his papers (you can find scans of his compilation work Indouralica online through the usual channels), they suffer from a lot of junk analyses. For one, there's gratuitous segmentation to make things "more comparable"; e.g. he compares Uralic *uwa 'flow, stream' (reliably established) with an internally reconstructed pre-PIE *aw- 'to flow', from which *w-ed- 'water' is then supposedly a derivative. On the other, to some extent there's the usual cherry-picking problem: comparisons between isolated single-branch terms like Finnic *liiva 'sand' ~ Greek λυμα 'dirt' (the former has been rather explained as a loan from Baltic *gleiwā).

Thinking about all this, I come to the insight that trying to work on these matters as an amateur is quite hubristic! After all, there are professionals who have much better training and much better knowledge - and despite working on it eight hours a day have found nothing! I really should stop here and concentrate on matters where I can actually achieve something, such as my conlangs and my music. Yet, I am very curious of these matters and find it hard to banish them from my mind...

[KathTheDragon]

PIE *h₃negʷʰ- ~ PU *künčə 'nail'; PIE *h₂angʷʰis 'snake' ~ PU *kunčV 'worm'.

The positioning of the nasal is also interesting.

[WeepingElf]

Alternations of the type *CeRC- ~ *CReC- are not rare in PIE; they are known as Schwebeablaut 'floating ablaut'. Probably all PIE initial clusters are the result of this or other ablaut phenomena, and the *CeRC- form is the more archaic (as is well known, Uralic does not allow initial clusters, and AFAIK most other Mitian languages don't, either, so Proto-Mitian probably prohibited initial clusters, too), except in cases where a morphological explanation is to be considered (as with s mobile).

[WeepingElf]

This paper by Miguel Carrasquer Vidal (a long-time member of the Nostratic-L mailing list; thank you, Šọ̈́gala, for posting it to the Great PIE Thread) brought me to think again about stop correspondences between IE and Uralic. The three grades of PIE stops all seem to correspond simply to the single grade of PU stops, not affricates, uvulars (which PU probably didn't have), voiced fricatives (if PU had them at all!), geminates or whatever. So, how did this single grade of stops split into three in PIE? Or was there a merger in Uralic?

Carrasquer suggests that the "glottalic PIE" glottalized stops (=classical PIE voiced unaspirated stops) could have emerged from tone contours:

*TéT > *TeT'
*TèT > *T'eT

If one of the "T"s is something else than a stop, nothing happens to it, regardless of tone contour. In a root with level tone, neither stop is glottalized. The non-coocurence of voiceless and voiced (non-glottalized) stops is a simple voicing assimilation rule, and maybe there was some sort of prosodic feature that caused stops in the root to voice, if two stops were there, both were voiced. These changes could have happened in either order; if voicing precedes glottalization, glottalization deletes voicing features, and if glottalization precedes voicing, voicing does not affect glottalized stops. Labial stops are never glottalized.

The questions however are, where did the tone contours come from, and which feature triggered voicing? After all, most Mitian languages do not have phonemic tone, for starters.

Surely, no solution yet, but definitely something to think about. Of course, Proto-Indo-Uralic could simply have had three grades (of whichever kind) of stops which collapsed into one in Proto-Uralic; but then, no other Mitian proto-language has those, with the exception of Starostin, Dybo and Mudrak's Proto-Altaic which is controversial (all actual Altaic languages have just two grades of stops). Or PIU had two grades which merged in PU and split in PIE (two-grade stop systems, usually voiceless/voiced, seem to be more or less the norm in Mitian).

[Zju]

According to this paper all the words in this wordlist are borrowings, but is it really likely that a lot of basic vocabulary such as 'not', 'eat', 'come', 'breast', 'go', 'die', 'say', 'skin', 'under', 'snow', 'moon', 'mouth', 'suck' etc. are all borrowings? The differentiation of vocabulary in two layers does provide support, but it still strikes me as odd the presence of so many primary words.

[Soap]

According to this paper all the words in this wordlist are borrowings, but is it really likely that a lot of basic vocabulary such as 'not', 'eat', 'come', 'breast', 'go', 'die', 'say', 'skin', 'under', 'snow', 'moon', 'mouth', 'suck' etc. are all borrowings? The differentiation of vocabulary in two layers does provide support, but it still strikes me as odd the presence of so many primary words.

U *kumi ‟loose snow‟ 26 > EU *kuma → MY *kumø > Y *kuu „snow‟

lol.

Häkkinen apparently does not believe the two language families are related, and this paper is a response to proponents of the unity theory, who presumably had presented a similar list of cognates and used that as evidence for a common proto-language. Häkkinen's response is to separate the list into two categories, showing that each category shows a slightly different type of correlation, meaning that they must have been loans. This is fair, and reminiscent of the way that, for example, Latin loans into Germanic can be separated into those that occurred before Grimm's Law (not too many) and those that occurred after (the vast majority).

But it's certainly possible that one of the layers of supposed loanwords could actually be the reflex of true cognates, showing the natural development in both branches, while the other consists of loanwords. It's a bit interesting that the words I'd expect least likely to be loaned appear in both lists, but I'm sure there's an explanation somewhere for why Proto-Uralic /u/ sometimes corresponds to Yukaghir /u/ and at other times to /o/ that doesn't rely on the /u/~/u/ set being loanwords.

[WeepingElf]

I have no solid opinion on the Uralic-Yukaghir hypothesis (i.e., that Uralic-Yukaghir forms a valid node, as opposed to the notion that they both belong to a Mitian or whatever macrofamily), but it seems doubtful to me. It has been pointed out that the Yukaghir words resembling Uralic words have an eastern, if not outright Samoyedic look to them. That would suggest that they were borrowed from an early form of Proto-Samoyedic, perhaps when an early stage of Proto-Yukaghir was spoken somewhere near Lake Baykal.

On the other hand, many of the words on that list are pretty basic vocabulary which one would not expect to be borrowed en masse from another language. Also, there are AFAIK similarities in the Samoyedic and Yukaghir nominal (but not the verbal) inflections. Can we exclude the possibility that Yukaghir is an aberrant branch of Uralic closely related to Samoyedic, transformed by the influence of some kind of (non-Mitian?) substratum? Cases like Insular Celtic or Armenian (or, to mention an example from a wholly different region, Vietnamese) show that languages can warp away from their parent within a rather short period such that the relationship is hard to discern.

[WeepingElf]

I wish to share two thoughts with you.

1. Taking another look at Elena Maslova's sketch of Tundra Yukaghir, I think that the notion that Yukaghir is "just an aberrant branch of Uralic" can be laid to rest. Sure, such a brief inspection is insufficient to answer such a question, but if Yukaghir was a branch of Uralic, this would be evident from such things as the numerals, which are totally different in Yukaghir (though it seems that Samoyedic differs considerably from the rest of Uralic there, too!), and other parts of the language which linguists surely would have noticed by now. There is perhaps about as much reason to assume that Yukaghir was a branch of Uralic as there is to assume that Etruscan was a branch of IE (which I don't!). So the Samoyedic-looking words in Yukaghir are perhaps best explained as Samoyedic (or Para-Samoyedic) loanwords, no matter how basic they are. Yet, I am not equipped to decide on this question!

2. I think the main reason for Starostin, Dybo and Mudrak to reconstruct three grades of stops for Proto-Altaic is the inclusion of Korean, which of course has a three-grade stop system. However, I seriously doubt that the inclusion of Korean (and Japanese) is warranted at all. This leaves Indo-European as the only Mitian family with three grades of stops.

[Richard W]

(seriously, Indo-Uraliac is speculative enough!)

There's damned good evidence for the reality of Indo-Uralic. I don't have the details to hand, and I may have misremembered the numbers, but the argument is:

Of the 20 most conservative Uralic words, as measured by retention within Uralic, 10 have apparently related, well-established Indo-European correspondents.

The Uralic family seems to be big enough, and to have widely split early enough, for Proto-Uralic vocabulary to be well attested, so this is unlikely to be a chance effect.

Of course, this of itself doesn't exclude the possibility that PIE originated as a pidgin para-Uralic or similar.

[Soap]

Some people believe that PIE and Uralic are related, but that they're not each other's closest cousins. For example, some people prefer to pair Uralic with Eskimo-Aleut instead, and, while still believing PIE and Uralic are related, reject any attempt to reconstruct a shared parent language for PIE and Uralic because it would need to incorporate a significant amount of vocabulary and grammar from Eskimo-Aleut as well. This would not be at odds with finding significant shared vocabulary between PIE and Uralic, and not even at odds with there being more of it than there is between Uralic and Eskimo, since the Eskimo languages are known to have an unusually low number of roots due to the power of their polysynthetic grammar.

As above, many researchers believe that the closest living relative of the Uralic languages is Yukaghir. I agree with that position and I think the evidence for it is sound. However, when people speak of "Indo-Uralic", they often incorporate Yukaghir as simply a branch of Uralic, as WeepingElf said, and therefore the Indo-Uralic parent language includes Yukaghir and does not contradict the hypothesis that Yukaghir is the closest living relative of the narrow Uralic family.

[KathTheDragon]

My point was that if we assume that Indo-Uralic is PIE's immediate ancestor node, we must have a good enough reconstruction of PIU before we can consider further relations. Doing it the other way round is just plain insanity.