The Great Proto-Indo-European Thread's Sequel

[WeepingElf]

That map actually shows the approximate former extent of the Baltic languages, which are famed for their conservatism, and you are not the first to suspect that PIE was spoken there - but the idea is pretty much out of favour now.

[abahot]

Oh no, I don't mean PIE itself, because it seems pretty settled in the Pontic-Caspian region. I mean that, if that map represents the earliest substratum of river names, it stands to reason that some ancestor of PIE may have been spoken there before the time of PIE proper.
If not there, is there a consensus on the origin of PIE? I've read north Volga, Caucasus, and even Central Asia, but I don't know if the broader IE community has actually come to an agreement.

[WeepingElf]

There is no consensus yet, but it seems as if the Yamnaya people (probably the speakers of PIE) descend from a mixture of two populations, one genetically similar to the Proto-Uralic speakers on the Middle Volga, and one which apparently came to the Pontic-Caspian steppe from the Transcaucasian/Iranian region. This mixture seems to have happened around 5000 BC. My pet theory about PIE is that it was a language closely related to Proto-Uralic which was drastically altered by the influence of the language(s) of the Transcaucasians. In such situations, languages can change rapidly, so "Proto-Indo-Uralic" may have been spoken north of the Caspian Sea as late as 6000 BC, and much more similar to Proto-Uralic than to PIE.

[abahot]

As may be apparent, my pet theory is that PIE was an offshoot of a now-extinct group spoken around the Baltic Sea, with few relations to other modern languages but with areal influence and loanwords to and from archaic Uralic and Caucasian languages. It might also explain why Baltic languages have diverged comparatively little, having migrated into an area with a population already speaking a similar language. However, para-Indo-European languages might have been spoken as late as the 2nd millennium BC, if the substrate in Finnic languages is any evidence.
What are your reasons for believing PIU? Is it primarily morphological (acc./1sg. *-m etc.) or other?

[WeepingElf]

My reason for considering PIU plausible (not for believing in it) is primarily morphological. If two languages so close to each other (geographically) share so much morphology, the idea that they are related to each other suggests itself. Yet, similarity due to contact cannot be ruled out, even if borrowing of whole inflectional paradigms and complete sets of pronouns doesn't seem very likely.

[abahot]

In my opinion, the pronouns (and related inflectional morphology) probably represent a vast Eurasian family with immense time depth, which doesn't have Indo-European and Uralic as a particularly close node. Just speculating in the moment, perhaps the more intense similarities in the verbal and pronominal paradigms could be a result of sprachbund effects between two languages with already vaguely similar forms?

[WeepingElf]

In my opinion, the pronouns (and related inflectional morphology) probably represent a vast Eurasian family with immense time depth, which doesn't have Indo-European and Uralic as a particularly close node. Just speculating in the moment, perhaps the more intense similarities in the verbal and pronominal paradigms could be a result of sprachbund effects between two languages with already vaguely similar forms?

Yes, this morphology is also found in many other languages from Turkic to Eskimo-Aleut, and it is not certain whether IE and Uralic form a valid node within this macrofamily. Alas, this whole matter is very speculative.

[Rounin Ryuuji]

Once you get to a certain time depth, things seem irreconstructable.

[WeepingElf]

Some scholars have argued from the conservatism of the Baltic languages and the apparent absence of older strata for an origin of PIE in that region - but by that logic, Proto-Germanic would have been spoken in Iceland. What we know about Icelandic, of course, is that the absence of an earlier stratum is simply due to the fact that the island was unpopulated before the Norsemen came. Now, the region east of the Baltic Sea was not unpopulated (or maybe it was, see below), but only very scarcely populated by primitive foragers, who wouldn't have made a dent on the IE language. The other extreme are Greece, the Indus valley and Anatolia, where the pre-IE population was culturally (but not militarily) more advanced than the Indo-Europeans, and thus, Greek, Sanskrit and the Anatolian languages are full of substratum loanwords. Central Europe is somewhere between these extremes: the Central European Neolithic farmers were not more advanced than the Indo-Europeans, but about on a par with them - except, again, in military terms where the Indo-Europeans had at least the advantage of having horses. (Things are quite complex in Central Europe, though, as there apparently were two strains of Yamnaya descendants involved here, the Corded Ware and the Bell Beaker cultures, and it was a back-and-forth movement: first Corded Ware, then Bell Beaker, then again descendants of Corded Ware - the Nordic Bronze Age (Germanic) and the Urnfield culture (Italo-Celtic), so we have three different Indo-European strata here atop the non-IE languages of the Neolithic farmers.)

But there is also evidence of Yersinia pestis ravaging large parts of Europe around 3000 BC, drastically reducing the pre-IE population, and maybe it was especially bad east of the Baltic Sea such that this region was virtually empty when the Indo-Europeans, and further north the Uralians, moved in, and hence Baltic and Balto-Finnic are so conservative and lack substratum influences.

And as to the external relationships of IE, the morphological resemblances are perhaps best explained by a macrofamily I call "Mitian", using a term coined by John Bengtson, which consists of IE, Uralic, Turkic, Mongolic, Tungusic, Yukaghir, Chukokto-Kamchatkan and Eskimo-Aleut. This family may have fanned out from somewhere near Lake Baykal at the end of the last ice age. As for subgrouping, it is anyone's guess. Altaic (Turkic, Mongolic, Tungusic) may be a valid node within Mitian because the "Mitian" pronoun set has been altered in similar ways in these three families; on the other hand, I don't think that Uralic and Yukaghir are particularly close - the Uralic-looking words in Yukaghir show Samoyedic traits and are therefore probably just loanwords from a Samoyedic language (see this paper by Ante Aikio). It seems plausible that IE and Uralic form a valid node, but we can't be sure about that yet, and most of the lexical resemblances appear to be loanwords from PIE into PU and thus say nothing about this question.

[hwhatting]

Well, Baltic looks conservative regarding nominal declension, but when you look at the verb system, it has undergone quite radical changes, regardless of whether you assume the classical Graeco-Aryan reconstruction or something more similar to Anatolian for PIE. So it's questionable whether too much theroretical construction should be based on the putative conservatism of Baltic.

[WeepingElf]

Yes, and one must not forget that it is the modern Baltic languages, especially Lithuanian, that stand out as conservative among the modern IE languages. Lithuanian is actually not more conservative than Latin, but Latin was spoken 2,000 years ago, and Lithuanian is spoken now. And Latin is a good deal less conservative than Greek or Sanskrit, let alone Hittite.

[Richard W]

Yes, this morphology is also found in many other languages from Turkic to Eskimo-Aleut, and it is not certain whether IE and Uralic form a valid node within this macrofamily. Alas, this whole matter is very speculative.

PIU is still quite meaningful even if it isn't a valid node. Aren't the other well-recognised families rather young?

[Travis B.]

Yes, this morphology is also found in many other languages from Turkic to Eskimo-Aleut, and it is not certain whether IE and Uralic form a valid node within this macrofamily. Alas, this whole matter is very speculative.

PIU is still quite meaningful even if it isn't a valid node. Aren't the other well-recognised families rather young?

Probably the oldest widely-recognized family is Afro-Asiatic, whose legitimacy is based essentially solely on striking morphological similarities amongst its members rather than on shared inherited lexicons (and even then there are questions, e.g. I have heard that people reject the identity of Omotic as Afro-Asiatic, and furthermore I have heard of people who limit it to just Semitic, Egyptian, and Berber and exclude not just Omotic, but also Cushitic and Chadic).

[abahot]

Good point about the fact that the Baltic region may have been virtually depopulated when early Balto-Slavic, or perhaps even undifferentiated early IE, speakers migrated in. That would explain a lot of their conservatism.

I'm still wondering about the possibility of parts of the Finnic substratum being IE (or para-IE) in origin, given that the two loanwords I mentioned (*täštä "star", and *kümmin "ten") bear obvious similarities to PIE *h₂stḗr "star" and *déḱm̥ "ten".

And as for the hypothetical "Mitian" macrofamily, I believe that the common sets of pronouns virtually guarantee some sort of relationship, but as Rounin mentioned, the immense time depth probably means that nothing else can be reconstructed whatsoever.

[WeepingElf]

In my humble opinion, Mitian is almost as good as Afroasiatic, considering such things as Seefloth's Paradigm. I think the point is that Eurasianists tend to be "splitters" who don't accept anything less than a reconstruction, while Africanists tend to be "lumpers" who are content with less. I think that those who reject Mitian should be asked whether they have a better explanation for the morphological resemblances between these eight families, especially as the archaeogenetic evidence, sparse as it is, seems to conform with it. Yet, I don't think that either the Moscow School's or Bomhard's reconstruction of Proto-Nostratic is correct; they can't both be right, so both are probably wrong and the method is flawed. Also, I don't think Afroasiatic or (Elamo-)Dravidian have anything to do with the Mitian languages, and I am also very doubtful of a connection to Kartvelian, though here I am not that sure (Kartvelian at least has 1sg. nominative *me, but this form actually is as isolated in the paradigm as is PIE *h₁eǵoh₂). The resemblances between IE and Kartvelian are merely typological, and may reflect an affinity of some kind between Kartvelian and the old language of the Transcaucasian/Iranian component in the PIE population.

[abahot]

Given that PIE was directly to the north of the Caucasus area, might the similarities be better explained as a PIE-Caucasian sprachbund? A sprachbund might also make sense for some other things mentioned here (who was it that proposed a Great Vowel Collapse?) Or if not a sprachbund, a mutual influence between currently-reconstructed languages rather than lost ones.

[WeepingElf]

Given that PIE was directly to the north of the Caucasus area, might the similarities be better explained as a PIE-Caucasian sprachbund? A sprachbund might also make sense for some other things mentioned here (who was it that proposed a Great Vowel Collapse?) Or if not a sprachbund, a mutual influence between currently-reconstructed languages rather than lost ones.

Maybe. Though I feel that some things in IE such as the ablaut system and the tripartite verb aspect system (though the latter is found only in Northern, i.e. non-Anatolian IE) are more similar to Kartvelian than to anything found in NWC or NEC. On the other hand, the three velar series are reminiscent of NWC while there is nothing like that in Kartvelian. One point where PIE differs sharply from all three Caucasian families is that it has just one sibilant, while the Caucasian languages have rich sibilant inventories (three types of affricates and two types of fricatives at two or more places of articulation). Of course, a Caucasian substratum or sprachbund is also invoked by the glottalists, but I have defected from that camp a few years ago because I think I have found a more parsimonious solution to the problems the glottalic theory tried to resolve. (This solution is that the PIE voiceless stops once were aspirated, and the system emerged from an earlier system with aspirated and voiced stops, or perhaps voiced spirants instead of the voiced stops, with the breathy-voiced stops resulting from the voicing of aspirated stops when some kind of prosodic element attached to the morpheme. The "*b-gap" would have formed before the rise of breathy-voiced stops by weakening of *b to a spirant that merged with *w or in some cases with *m, explaining the partially stop-like behaviour of these phonemes in PIE.)

And it was me who proposed the Great Vowel Collapse.

[abahot]

This solution is that the PIE voiceless stops once were aspirated

Might this system have been preserved in pre-Proto-Germanic? In that case, it seems possible to explain Grimm's Law, as the glottalic theory invokes, by simply having "all aspirated stops fricativize" as a sound change.

"b-gap"

On a related note, on the abundance of *m, I wonder if Pre-PIE had the sound change *-n# > *-m#. It would explain why *-m is so abundant in endings and why none end in *-n.

And it was me who proposed the Great Vowel Collapse.

I came up with a similar idea a while ago, except without nearly the level of detail. Mine was roughly that some ancient ancestor of PIE had *a *i *u, with all three collapsing to a single vowel and altering preceding velar consonants, much like some current NWC inventories. However, given how likely it is that palatovelars were never actually palatal in PIE (and that the "velars" were uvular), I think the pre-GVC vowel inventory might not have been a straightforward *a *i *u.

[WeepingElf]

This solution is that the PIE voiceless stops once were aspirated

Might this system have been preserved in pre-Proto-Germanic? In that case, it seems possible to explain Grimm's Law, as the glottalic theory invokes, by simply having "all aspirated stops fricativize" as a sound change.

Grimm's Law is notoriously hard to date. There are reasons to place it early, there are reasons to place it late. My attempt at solving this dilemma is to assume that the first step, preserving (or restoring) the aspiration of the voiceless stops and devoicing the voiced stops, was early, while the spirantization of the [+breath] stops was late. Verner's Law may have happened at any point between. At any rate, Germanic appears to have been isolated from the rest of Northern IE for some time, as in Central Europe, the Northern IE language of the Corded Ware people (which would have been the "missing link" between Germanic, Balto-Slavic and Italo-Celtic) was displaced by the Southern IE language of the Bell Beaker people.

As is well known, a similar change, but without the spirantization, happened in Armenian. As there is evidence in favour of the *Dh set still being breathy-voiced in Old Armenian (some Eastern Armenian dialects preserve this, and it makes sense of the Western Armenian consonant shift which merged the aspirated and voiced stops, which becomes a simple voicing neutralization under this assumption), this simply becomes a devoicing of the voiced stops. Alas, Armenian has innovated so heavily that it is not really the right place to look for archaisms, but even the most innovative languages may be conservative in some regards.

"b-gap"

On a related note, on the abundance of *m, I wonder if Pre-PIE had the sound change *-n# > *-m#. It would explain why *-m is so abundant in endings and why none end in *-n.

I'd rather guess that final *-n became *-r, which would explain the r/n-heteroclitics that are quite numerous in Anatolian. Only after *m, final *-n was preserved, as in the many nouns with the *-mn suffix.

And it was me who proposed the Great Vowel Collapse.

I came up with a similar idea a while ago, except without nearly the level of detail. Mine was roughly that some ancient ancestor of PIE had *a *i *u, with all three collapsing to a single vowel and altering preceding velar consonants, much like some current NWC inventories. However, given how likely it is that palatovelars were never actually palatal in PIE (and that the "velars" were uvular), I think the pre-GVC vowel inventory might not have been a straightforward *a *i *u.

My model of the GVC begins with a richer vowel inventory, at least *i *e *a *o *u. Of these, all but *i and *u fall together as *a. However, before that change, high vowels before sonorants are lowered and therefore fall victim to the GVC. When a vowel was neighbouring a velar, that velar preserved the features [+front] and [+round] of the vowel. (There apparently were no front rounded vowels in the pre-GVC stage, as we don't have a fourth velar series that is both palatalized and labialized, and no **ü.) Here is an example:

1. Pre-PIE *kul- 'to turn' (seems to have cognates in other Mitian languages)
2. *u is lowered before the sonorant *l: *kol-
3. The velar *k acquires the feature [+round] from the following vowel: *k^wol-
4. Great Vowel Collapse: *k^wal-
5. Ablaut: *k^wel-

[abahot]

I'd rather guess that final *-n became *-r

That makes a lot of sense, actually. I never really thought to connect the two phenomena.