The Great Proto-Indo-European Thread's Sequel

[WeepingElf]

Yes. As were Hurrian and Urartian, spoken where those geneticists place Early PIE. This brought me to an idea which may be utterly nuts: There is this paper, which is a collaboration of a patent crackpot (Arnaud Fournet) and a less crazy but still adventurous Nostraticist (Allan Bomhard), which claims that Hurrian and IE were related to each other. But maybe Hurrian-Urartian was related to the language of those Transcaucasians who contributed to the gene pool of the Yamnaya people, and we are dealing with words from their language which made their way into PIE. But maybe there is really nothing to this at all.

Bomhard has recanted of this idea.

Yes, he has. I know Bomhard quite well; I met him (only virtually, of course) on the Nostratic-L mailing list and had some private e-mail exchange with him. He is a rather nice fellow and treats other people with due respect even if their opinions are different, and is fully aware that his ideas are not "the truth" but only ideas of what might have been. This distinguishes him from the crackpots one meets so often on the 'Net, and I would say that he s not a crackpot but merely a speculator. He also said once on Nostratic-L that there were no professional Nostraticists. Bomhard is by profession an Indo-Europeanist, meanwhile a retired one. Nostratic is merely a personal interest of him. There are no chairs for "Nostratic studies" or "macrocomparison" at any university in the world.

Fournet, in comparison, whom I also have met on Nostratic-L, showed all the obnoxious behaviour of a typical crackpot. He was convinced (or at least tried to convince everybody) that his ideas were the truth, and offended people with different opinions, including the academic mainstream. For a short time, he had a blog titled "World Web Trolls, Idiots and Assholes" devoted to the sole purpose of insulting people with different opinions. That blog did not exist for long; I guess that it was shut down by the host because it violated usage terms. And while Bomhard recanted of the IE-Hurrian idea, Fournet went even further, claiming that Hurrian was closer to Non-Anatolian IE than Anatolian! (And he never even mentioned Urartian.)

Hurrian and PIE, or at least Anatolian, do seem to have grammatical similarities, but the notion of lexical matches is not plausible.

Fair (though my impression is different). And Anatolian and Hurrian were neighbours, so one should rather think of convergence here.

At any rate, it is pretty certain that the Hittites, Luwians etc. were newcomers in Bronze Age Anatolia, arriving not long before 2000 BC, and that they came from the northwest. The northwesternmost Anatolian language, Lydian, appears also to be the most divergent one, though this is hard to say as Lydian is attested only late. See also this blog post. By the way, the map shown there, which is from one of the papers discussed here, shows an IE family tree which is utter bullfrogs: Greek is clearly not closer to Anatolian than to Italic, Celtic, Germanic and Balto-Slavic! We are clearly dealing with people who don't know the relevant facts here.

And the actual family tree published in the paper, as opposed to the map, doesn't make such a claim. Rather, the relatedness of Greek-Armenian-Albanian to the rest of core-IE comes out as not well supported by the lexical evidence. The map attempts to depict them as stay-at-homes; it can't depict deep divisions.

Perhaps the person who drew the map misunderstood what the authors of the paper wrote. The family tree looks rather sane, even if the dating is probably wrong, and the distance between Anatolian and the rest is IMHO exaggerated. Also, it groups Tocharian with Anatolian, which seems erroneous to me. Just because these two are outliers doesn't mean that they form a node.

[Nortaneous]

See also this blog post.

The claim that Tocharian, which is known to have been extensively restructured by contact with Uralic

I didn’t know this. Do you have more information anywhere?

Peyrot 2019, but consider the stop system collapse and expansion of the case system (including a perlative but lacking a dative) - both of which, however, are likely post-PToch, since the cases don't quite line up and TA and TB have different reflexes of *dʲ.

[Moose-tache]

I am very sceptical of these arguments that Tocharian's features are Uralic or Turkic influence. Tocharian split from PIE at a fairly early date, very possibly around or even before 3000 BC, although that's not set in stone. The Turkic and Uralic languages do not give us nearly that level of depth. The most ambitious reconstruction of Proto-Turkic doesn't go further back than 500 BC without giving way to pure speculation. The first attested Turkic languages were not spoken near the attested Tocharian languages, so it's hard to say when this contact would have taken place, and even harder to say how many stops Turkic would have had. For Uralic, the situation is even worse. We don't know what pre-Uralic would have looked like for most of the existence of the Tocharian branch, nor if it was spoken in the areas where Tocharian was spoken before its attested forms.

That paper by Peyrot is very careful to say X "may have happened," or Y "could be understood this way." It wisely makes no claim that any of these suppositions are anything more than speculation. We certainly cannot say "it is known" that Tocharian is the product of contact change with Uralic or any other specific language.

There are innumerable have waves, like "well maybe the changes occurred very rapidly during a short and very recent period of intense contact." But all of this is speculation, and no better than the null hypothesis: Tocharian is just like that for some other reason we don't know yet. This is my biggest personal bugbear. Saying that something is probably true merely because it feels plausable, rather than because you have ruled out the null hypothesis.

[WeepingElf]

Turkic is certainly not the language family to look at here - it is too shallow (Proto-Turkic was spoken only about 2000 years ago), and it doesn't match the phonological type (it has voiced stops, and front rounded vowels with vowel harmony, all things we don't see in Tocharian). But Uralic is another matter. Proto-Uralic is only a bit younger than PIE (about 3000-2500 BC), and lacks voiced stops. It does have /y/ (but no /ø/), though, according to the accepted reconstruction. But it was spoken nowhere near either the Tarim Basin or the Minusinsk Depression (where the Afanasievo culture, widely assumed to be the "missing link" between the Yamnaya culture and the historically attested Tocharian speakers, was located). And Kamassian, the Uralic language closest to Tocharian, doesn't fit the type - it had voiced stops and front rounded vowels just like the neighbouring Turkic languages, and probably was strongly influenced by the latter. So neither a Turkic nor a Uralic influence seems very likely here.

[WeepingElf]

I have thought again about the "hybrid theory" papers we discussed a few weeks ago, and I have an idea why they have Anatolian and Tocharian diverge so impossibly early. Well, what those people did was to feed a distance matrix based on lexical cognates into an algorithm designed to compute family trees from genetic distance matrices. Now, Anatolian and Tocharian are extinct branches whose lexica are incompletely known - Tocharian because the corpus is small and doesn't cover all subject matters, and Hittite because many words are hidden behind Sumerograms. Now, unknown words count as "not cognate", and thus the distance appears exaggerated.

[Moose-tache]

Looking at Anatolian Swadesh lists would suggest that the difference in vocabulary is not simply an artifact of preservation or presentation. I see plenty of positively identified words, but relatively few with clear IE etymologies.

[WeepingElf]

Looking at Anatolian Swadesh lists would suggest that the difference in vocabulary is not simply an artifact of preservation or presentation. I see plenty of positively identified words, but relatively few with clear IE etymologies.

Are you sure you have culled out the Sumerograms and Akkadograms? Everything transcribed in capitals is not genuine Hittite, and must count as "unknown". But I concur with you that Anatolian branched off early, though it can't be as early as the proponents of the hybrid theory claim. While Hittite has a different word for 'wheel' than most other IE languages, it appears to have a cognate in Tocharian, which doesn't form a valid node with Anatolian, which indicates that the word existed in PIE before Anatolian branched off; and Hittite has a reflex of PIE *yugom 'yoke', so the speakers of the common ancestor apparently knew animal traction, if not for carts or wagons at least for sleighs and ploughs, which rules out a divergence much before 4000 BC. And at any rate, the assumption of those scientists that the rate of lexicon replacement was constant across time and languages has been falsified long ago. It very much depends on how much contact with other languages the language has, and how stable the social institutions are. Just compare English with Icelandic, and you get the picture.

[KathTheDragon]

Hittite has a reflex of PIE *yugom 'yoke'

Hittite appears to actually reflect a neuter root noun *yéwg- (cf nom-acc.sg i-ú-uk) which was only secondarily thematised in Hittite and Late PIE separately. This doesn't really undo the argument, but it's a good point to note.

[WeepingElf]

Hittite has a reflex of PIE *yugom 'yoke'

Hittite appears to actually reflect a neuter root noun *yéwg- (cf nom-acc.sg i-ú-uk) which was only secondarily thematised in Hittite and Late PIE separately. This doesn't really undo the argument, but it's a good point to note.

Thank you. I wasn't aware of this detail.

[KathTheDragon]

Another detail - I had a look at the entry again, and Kloekhorst glosses it as "yoke, pair", which (together with the fact that the base verb means "join" rather than "yoke") may indicate that it was only specialised into a technical term post-PIE. That Anatolian and Late PIE did the same thing isn't too surprising.

[hwhatting]

Here's a recent paper showing Kloekhorst's view of the position if Anatolian in IE and about the internal branching of Anatolian.
I don't agree with him on some points - I assume that Anatolian didn't lose the distinction between Aorist and Present, but that it was only formed in core IE, and that subjunctive and optative also were developments in core IE instead of being lost by Anatolian, but that doesn't influence the branching.
He comes up with a range of 4400 - 4100 BCE for the split-off of Anatolian from PIE.

[WeepingElf]

I think this controversy highlights the difficulty of assigning split dates in dialect continua. Anatolian doesn't seem to require any different phonology than the rest of the family, but its morphology is quite different, and many Late PIE lexemes are absent. We are probably dealing with two dialect groups within PIE that gradually diverged but remained in contact with each other for several centuries, such that some innovations could spread from one to the other several centuries after they had begun to develop into different directions. If you consider just how large the Yamnaya horizon is (it is about 2,000 kilometres from the Danube to the Ural), it is only plausible that the language spoken by these people showed dialect differences.

[WeepingElf]

I'm reviving this thread in order to discuss a new thought that came to my mind last night.

It has long been theorized that PIE was a language related to Proto-Uralic altered by the influence of a "Caucasian" substratum, and this is borne out by the fact that the Yamnaya people, who probably spoke PIE, are genetically a blend of an Eastern European population related to who probably were the speakers of Proto-Uralic, and another population from south of the Caucasus (this appears to have happened between 5000 and 4500 BC, if I am not mistaken). The question is which of the extant Caucasian families the language of the latter was related to. But couldn't it have been related to something else?

My idea is that this something else may have been Semitic. Indeed, in many of the points where PIE looks very different from PU, it looks much like Semitic: three phonation types of stops, a three-vowel system in the immediate pre-ablaut stage, ablaut itself, and maybe also the *h₂e-conjugation which looks similar to the Semitic suffixal conjugation - and of course a liberal amount of loanwords, such as *steuros 'bull' and some of the numerals. (The Nostraticist literature of course is full of such comparisons, and they may be up to something.) Of course, there are differences, such as the three phonation types of stops not being the same as in Semitic (unless you believe in the glottalic theory) or the lack of sibilant and lateral affricates (both are of course also an issue with any of the three Caucasian families), and the palatalized and labialized velar stop series are unknown in Semitic, but are known from NW Caucasian. Well, PIE probably was in contact with Proto-NWC, too, and may have been influenced by it, no matter whether there is a Semitic-related substratum in it or not.

I am not sure about this idea yet, hence I am putting this up for discussion here. What do you think?

[Zju]

Wasn't Proto-Semitic spoken in the south of the Arabian peninsula? It's a stretch to posit some parasemitic substrate in the EE steppe.
Although the lexical similarities are suspicious indeed.

[WeepingElf]

Wasn't Proto-Semitic spoken in the south of the Arabian peninsula? It's a stretch to posit some parasemitic substrate in the EE steppe.
Although the lexical similarities are suspicious indeed.

According to Wikipedia, it is now mostly placed in the Levant (Israel/Syria/etc.), and the south Arabian hypothesis is out of fashion. But I am not a Semiticist, so I can't really say what is most likely. But it seems as if the "Caucasus Hunter Gatherers" (CHG), the population from south of the Caucasus that went into the Yamnaya, doesn't have much to do with the people of Neolithic Levant genetically, so there is no good reason on that count to assume that they spoke a language related to Semitic. But all this "genes and languages" business must be carried out with great care. First, genes and languages do not always travel together. Second, the sample sizes of the archaeogeneticists are very small (because sufficiently well-prepared bones are rare), so the margin of error is quite large. Not to mention that genetics can be abused for the same nefarious purposes as physical anthropology: doesn't the discovery of the Yamnaya spread "prove" that there really is something like an "Aryan race" (even though the fact that this apparently is a Mischvolk descending mainly from a Siberian and a Middle Eastern population probably is something those race theorists won't like)?

Thinking about all this, I now consider it more likely that the "CHG language" was a member of a lost, extinct lineage.

[Tropylium]

I think the typological universal about the distribution of implosive POAs isn't as much of a problem as it's made out to be - there's close precedent in Maay. (Where did implosives in Maay come from? How well reconstructed is Cushitic?) And there are a few languages with /ɗ/ and no other implosives - could this pattern develop from an earlier and more normal *ɓ *ɗ by unsystematic sonorantization[?] of the labial?

Nort has probably seen me talking about some of this elsewhere after I got into Cushitic last fall, but for general benefit: Maay, which is either a close relative or divergent dialect of Somali, has /ɗ ʄ ɠ/, no /ɓ/. All of these seem to come from earlier ejectives. /ʄ/ is from palatalization of /ɠ/ before front vowels, which in turn is from Proto-Somaloid *kʼ (corresponds with standard Somali /q/); /ɗ/ (corresponds with standard Somali /ɖ/) is from a merger of various early Cushitic glottalics already in Proto-Somaloid. There's a long list proposed: all of *tʼ *ɗ *tsʼ *tɬʼ *tʃʼ *ʄ, but they tend to merge towards one or two consonants so probably several correspondence sets are due to unidentified splits or interdialectal loaning. E.g. the Rift languages have just /tsʼ tɬʼ/, and the Agaw / Central Cushitic languages have plain voiceless stops or affricates for all of these. Hard to tell if having just *ɗ was already the case in Proto-Cushitic or Proto-East Cushitic, or if there was an early areal shift from *tʼ that affected several branches. I would think the latter more likely (would reconstruct altogether *tʼ *tɬʼ *tʃʼ) but it's really not certain at all

The lack of any labial glottalic is simply inherited from Proto-Afrasian, basically all of the Cushitic languages that have /ɓ/ and/or /pʼ/ have it only as a loanword phoneme (often from Omotic where these exist "normally") or otherwise as an innovation. At least one language has regular *b+t > *bʔ > /ɓɓ/, a few have *f+t > *fʔ > /ppʼ/ (both generally only found at morpheme boundaries), a few languages seem to reflect a rare cluster *ʕb as /ɓ/ or /pʼ/, and also at least a few have unconditional *b > /ɓ/ (may or may not go with also *d *g > /ɗ ɠ/). (Ehret reconstructs also a Proto-Cushitic *pʼ but it's basically for junk etymologies that have /b/ ~ /p/ mismatches.)

I gather from these kind of cases that while gaps tend to get filled over long enough time, especially in contact situations, just the existence of a gap really doesn't tell much about a language's history and sometimes something just has been missing for the last 5000 or 10,000 years for no apparent reason.

[abahot]

In my personal opinion, the "typological universals" are distributions of various stop consonants does not really have as much bearing on PIE phonology as people tend to assume. Such "universals" are found to be false all the time (like Northwest Mekeo and the idea of all languages having coronal consonants). Specifics aside, I don't think that discovering PIE-phonology-ese in New Guinea will magically change the likelihood of PIE having the phonology it is commonly reconstructed to have.

[WeepingElf]

In my personal opinion, the "typological universals" are distributions of various stop consonants does not really have as much bearing on PIE phonology as people tend to assume. Such "universals" are found to be false all the time (like Northwest Mekeo and the idea of all languages having coronal consonants). Specifics aside, I don't think that discovering PIE-phonology-ese in New Guinea will magically change the likelihood of PIE having the phonology it is commonly reconstructed to have.

I concur with you. Features that are typologically rare are probably diachronically unstable, and feature combinations that happen not to occur are probably just non-existent by chance. The number of possible language types exceeds the number of actually existing languages by several orders of magnitude, so many perfectly possible types won't be observed anywhere. And the diversity of stop systems found in attested IE languages probably means that whatever the stop inventory of PIE was, it probably wasn't very stable, so it should be rare rather than common. The system reconstructed by the glottalists is quite common and thus probably stable (and has indeed proven to be very stable in Afrasian and in all three Caucasian families), so it is probably not what we are looking for.

[abahot]

All that being said, what are people's thoughts on the stop system of PIE? Personally, I think it's likely that the glottalic viewpoint was true at some point in pre-Proto-Indo-European history (which would account for things like the absence of *b), but then evolved into the traditionally reconstructed PIE system during the period leading up to the fragmentation of the Indo-European language varieties (which would account for things like the breathy-voiced consonants being attested in two different subfamilies.)

The true answer might also be some variation on this theme. For example, perhaps the *D and *D^h series were originally glottalized and plain voiced in pre-PIE, but the two changes (loss of glottalization in the *D series and breathy-voicedness of the *D^h series) happened as areal changes in separate regions, maybe even at different times, propagating through a late PIE dialect continuum with varied eventual results.

[Travis B.]

All that being said, what are people's thoughts on the stop system of PIE? Personally, I think it's likely that the glottalic viewpoint was true at some point in pre-Proto-Indo-European history (which would account for things like the absence of *b), but then evolved into the traditionally reconstructed PIE system during the period leading up to the fragmentation of the Indo-European language varieties (which would account for things like the breathy-voiced consonants being attested in two different subfamilies.)

The true answer might also be some variation on this theme. For example, perhaps the *D and *D^h series were originally glottalized and plain voiced in pre-PIE, but the two changes (loss of glottalization in the *D series and breathy-voicedness of the *D^h series) happened as areal changes in separate regions, maybe even at different times, propagating through a late PIE dialect continuum with varied eventual results.

The key thing is that the conventional analysis of PIE stops is very consistent with what is seen in multiple branches of IE (e.g. Indo-Aryan, Greek, Armenian, and even Germanic and Italic) while being sufficiently unstable to explain why it evolved in all kinds of different ways in PIE's daughters, while as WeepingElf says the glottalic hypothesis is too common and too stable to explain the diversity of outcomes in IE, and is not consistent with changes that explain said outcomes in different branches.