The Great Proto-Indo-European Thread's Sequel

[keenir]

Hmm...

I'm starting to think that we may need to borrow from physics on this matter -- particularly the structure and nature of atoms. There are, if I recall, three ways to depict an atom, and which one a person uses, is dependant upon what they want to demonstrate or convey.

That or we borrow from light itself: it both is and isn't a wave.

[keenir]

I think you’re misunderstanding what the ‘wave model’ actually means. It doesn’t mean that every branch is strongly related to every other: that would be obviously absurd. Instead, it means that many neighbouring branches can and do share sound changes, in ways that the tree model does not predict happening.

Let’s look at Hindi. Of course, English and Hindi have shared no sound changes in the past, say, 2000 years.

This has me wondering if English and Hindi have shared any sound changes in the time since English has begun to be spoken on the Subcontinent and Hindi in the British Isles. (or do they simply influence one another, creating accents?)

But Hindi has shared sound changes with its neighbours, the other Indic and Iranian languages. Indeed, it shares enough sound changes with the other Indic language that we can group them as a single group, and then enough with Iranian languages that we can be justified in grouping them further as ‘Indo-Iranian’. Further up the cladogram, Indo-Iranian shares some sound changes with Greek, Armenian and possibly others — but not enough to draw any exclusive subgrouping, which is the requirement of the tree model.

if its too few for a grouping or subgrouping, is it enough to call it areal or coincidental?

(or, if its not enough for an exclusive subgrouping, is it enough for suggesting that those sound changes were part - but not all - of their history together before they went their separate ways?)

[Talskubilos]

(although I still think "language change looks tree-like on large scales" is undeniably true).

On this point, you just need to look at the difficulties with subgrouping in families like PIE. Some people say, for instance, that Greek forms a distinctive subgroup with Indo-Aryan; others have argued for a relationship with Armenian; still others suggest similarities with Albanian. Which of these is correct? The simplest conclusion is that all of these overlapping groupings reflect different parts of the history, with varying degrees of ‘subgroupiness’ (as François calls it) depending on how much they share.

Not to speak of poorly or not attested IE languages, which only left traces in toponymy and loanwords.

[bradrn]

(although I still think "language change looks tree-like on large scales" is undeniably true).

On this point, you just need to look at the difficulties with subgrouping in families like PIE. Some people say, for instance, that Greek forms a distinctive subgroup with Indo-Aryan; others have argued for a relationship with Armenian; still others suggest similarities with Albanian. Which of these is correct? The simplest conclusion is that all of these overlapping groupings reflect different parts of the history, with varying degrees of ‘subgroupiness’ (as François calls it) depending on how much they share.

Not to speak of poorly or not attested IE languages, which only left traces in toponymy and loanwords.

Oh, sure, but even the directly attested languages show this clearly enough.

[abahot]

How would our knowledge of Proto-Germanic be different if we didn't have Gothic? I assume we might be able to reconstruct the original nominative singular ending through runic evidence, but what else might be missing or different?

[foxcatdog]

that throws out the vocative being reconstructed

[WeepingElf]

Early last year I posted on the CONLANG list the idea that the much-discussed "Caucasian" substratum in Proto-Indo-European was related to Semitic:

https://listserv.brown.edu/cgi-bin/wa?A ... ANG&P=5567

However, I have since grown doubt against my own idea. The Transcaucasian population that contributed to the gene pool of the Yamnaya culture was genetically not much like the Levantine farmers who probably spoke Proto-Semitic, and thus more likely spoke an unrelated language (though both languages may have shared Neolithic Wanderwörter). Also, Semitic does not really fit the PIE phonology much better than Northwest Caucasian. While Semitic has fewer sibilant and lateral phonemes than NWC languages, it still has more than PIE which had only one of each, and furthermore, Semitic does not have palatalized and labialized velars which do occur in NWC.

The Maikop culture, the southern neighbours of the Yamnaya culture in the northern foothills of the Great Caucasus, probably spoke Proto-NWC, a language whose phonology is unknown: while some Russian scholars have tried to reconstruct it, they arrived at a monster with more than 150 consonant phonemes, which is unlikely to be correct. (It can be found on Wikipedia.) It seems more likely to me that the Proto-NWC phoneme inventory was *smaller* than the inventories of the modern NWC languages which may have been grown by vowel losses and cluster simplification.

The genetic studies show that a branch of the Transcaucasians moved west into Asia Minor, which the adherents of the "Southern Arc theory" (according to which "Proto-Indo-Anatolian" was spoken by the Transcaucasians and only "PIE proper" by the Yamnaya people) hold responsible for the spread of the Anatolian branch into Asia Minor. More likely, I think, the language this migration brought into Asia Minor was Hattic, a poorly-known non-IE language sparsely attested in cuneiform tablets from Hattusa which has been conjectured to be related to NWC. The phonology of Hattic is unknown because the cuneiform script is massively underspecifying; while the presence of a minimal set of phonological distinctions can be discerned, these are so generic that they fit almost any language, and there probably were more distinctions not shown in the cuneiform script.

[WeepingElf]

In this post, I shall lay out my thinking about the origins of the
Indo-European family.

Most Indo-Europeanists adhere to some form of the steppe hypothesis
about the origin of IE; so do I. About a century ago, the Dutch linguist
C. C. Uhlenbeck conjectured that Proto-Indo-European (PIE) was a
language related to Uralic influenced by a substratum related to
Caucasian languages. Recent genetic studies have come up with data that
agree well with that. Apparently, the Yamnaya people who probably spoke
PIE emerged from a mixture of two populations, the "Eastern European
Hunter-Gatherers" (EHG) and the "Caucasus Hunter-Gatherers" (CHG)
originating from south of the Caucasus, about 5000 BC (though at that
time, the CHG probably already were farmers). The Proto-Uralic speakers
probably also were EHG, while the Maikop culture in the northern
foothills of the Caucasus, the southern neighbours of the Yamnaya,
apparently were unmixed CHG. While some CHG moved north across (or
around) the Caucasus, other CHG moved west into Anatolia.

However, some geneticists (not Indo-Europeanists!) have interpreted
these results such that "Proto-Indo-Anatolian" was spoken by the CHG
before these migrations, and Anatolian was carried into Anatolia by the
westward-moving CHG, and PIE proper emerged on the steppe from the
language of the CHG moving north. I think, though, that this does not
comply well with the linguistic facts. There are undeniable
morphological matches between IE and Uralic, which would be hard to
explain if the two language families originated in such different
regions, and these morphemes are also found in Anatolian, of course.
Also, while Anatolian appears to have branched off early (or rather,
emerged from an archaic, probably geographically marginal, dialect of
PIE), it seems unlikely that it branched off 2,000 or more years before
the dissolution of "PIE proper", as there do not seem to have been many
phonological changes between pre-Anatolian and post-Anatolian PIE.

So I would say that the CHG brought in a language related to Northwest
Caucasian (NWC); the Maikop culture may have spoken Proto-NWC. Pre-PIE
was thus a Para-Uralic language on a Para-NWC substratum, which accounts
for the typological differences between PIE and Proto-Uralic; in many of
these points, PIE is quite like NWC. The Russian linguist Viacheslav
Chirikba reconstructs Proto-NWC with 100 consonant phonemes, which are
quite a lot, but most of them are just palatalized or labialized (and
also palatalized AND labialized) variants of other phonemes. These
secondary articulations cannot be separated from the impoverished vowel
inventory, which consists of only two vowel phonemes - a close one and
an open one. What apparently happened was that vowel features such as
[+front] and [+round] were transferred from the vowels to the
consonants. If one removes these secondary articulations from the
inventory, one gets a still large but not huge inventory of 39
consonant phonemes:

Code: Select all

  *p   *t   *ts  *tɬ  *tʃ  *k   *q
  *b   *d   *dz  *dɮ  *dʒ  *g   *ɢ
  *p'  *t'  *ts' *tɬ' *tʃ' *k'  *q'
  *f        *s   *ɬ   *ʃ   *x   *χ   *ħ
            *z        *ʒ   *ɣ   *ʁ   *ʕ
  *m   *n
  *w   *r        *l   *j

his may have been the consonant inventory of the Pre-Proto-NWC language
spoken by the CHG prior to their northward migration. Similar
inventories are found in the (unrelated) Kartvelian languages, as well
as in the NEC languages which are considered related to NWC by some, but
that is very uncertain. The language carried west by CHG into Anatolian
would have been an ancestor of Hattic, a poorly attested non-IE language
of Anatolia, which some linguists assume to be related to NWC. The
Hattic phonology is not fully understood, as the cuneiform spelling does
not indicate distinction alien to Hittite, the language of the scribes.
However, there are vacillations that point at phonemes unknown to
Hittite, such as frequent vaciallations between <p> and <w> which point
at a phoneme */f/, and less commonly between <t> and <l> (as in the
royal title _Tabarna/Labarna_) which can be interpreted as a phoneme
*/tɬ/. If you remove the major distinctions that are foreign to Hittite,
such as the distinction between voiceless, voiced and ejective
consonants, between alveolar and postalveolar sibilants, or between
velars and uvulars, the Pre-Proto-NWC inventory given above shrinks to:

Code: Select all

  *p   *t   *ts  *tɬ  *k
  *f        *s   *ɬ   *x
  *m   *n
  *w   *r        *l   *j

which is more or less what the spellings of Hattic words suggest.

Now back to PIE! What happened on the way from Proto-Indo-Uralic (PIU)
to PIE? It seems likely that PIU was typologically much closer to
Proto-Uralic than to PIE, as Proto-Uralic appears to be a much more
typically "Mitian" language than PIE - to such an extent that
19th-century linguists classified Uralic together with Turkic, Mongolic
and Tungusic as "Ural-Altaic"; also, Uralic is quite similar to
Eskimo-Aleut despite the great geographical and probably also
historio-linguistic distance between the two - the resemblance is
certainly due to conservatism.

So PIU would have had a singe set of stops, unmarked for voicing,
aspiration or glottalization, accompanied by sets of voiced spirants and
nasals at the same places of articulation. The labial spirant *β merged
with *w at some point. Now, these voiced spirants were alien to the
Para-NWC language. Thus, the PIU stops were rendered as voiceless stops
which were phonetically aspirated, and the voiced spirants as voiced
stops. The Para-NWC ejective stops were not used to render anything in
Pre-PIE, so it did not have such sounds (pace the proponents of the
glottalic theory). So we'd get, using the dentals as example, *t > *tʰ
and *ð > *d. In the next step, the aspirated stops were voiced in some
morphemes (probably by some kind of prosodic feature), but as they were
aspirated, these did not merge with the old voiced stops but formed a
third type: *tʰ > *dʰ. From there, it is only one step - loss of
aspiration in *tʰ - that leads to the "Classic PIE" system. The gap at
*b is explained by the old *β > *w merger, which also explains the
somewhat stop-like behaviour (as in the initial clusters *wl- and *wr-)
of *w. The exclusion of roots with two voiced unaspirated stops in one
morpheme would have been due to some kind of dissimilation rule that was
already in operation before the voiced spirants hardened to stops; this
may have turned one of two spirants in one morpheme into an approximant
or whatever.

The velar stop series split into three at the time when the PIU vowel
system was reduced as vowel features became secondary articulations of
the velars in the same way as in NWC. The palatalized velars may also
have absorbed an Indo-Uralic palatal stop series, contributing to their
high frequency. A similar effect affected the uvular fricative *χ, but
here no palatalized variant occurred because this sound had caused
backing of adjacent front vowels (try to say [χe] without either
fronting the [χ] or backing the [e], and you get the picture) - *h1 was
no palatalized laryngeal, but simply */h/. It is misleading to think of
this laryngeal as "e-colouring"; rather, it is *non*-colouring, like all
the non-laryngeal consonants.

So, if the CHG did not speak Proto-Indo-Anatolian, and the CHG moving
into Anatolia spoke Hattic rather than Anatolian, where did Anatolian
come from? From the steppe, like all IE languages. But which way around
the Black Sea? I think the western route is more likely. Apparently, the
Hittites had a tradition of once living in a region where the Sun rises
from the sea, i.e. on the western shore of a sea; IMHO this is more
likely to have been the Black Sea than the Caspian Sea. Also, the most
divergent Anatolian language appears to have been Lydian, which was the
northwesternmost Anatolian language. I would say that Anatolian
originated in the southwestern outlier of the Yamnaya culture on the
Lower Danube, which may have spoken an archaic dialect not yet affected
by the morphosyntactic innovations (feminine gender, tripartite verb
aspect system) of the PIE heartland. A similar language may have been
spoken by the Bell Beaker people, who appear to have spread from there
into Western Europe. The Italo-Celtic languages do not descend from
this, but spread across Western Europe only later. The diversity within
both Italic and Celtic is too small for these branches to have spread
that early.

About a year ago, I suggested that the "Caucasian" substratum may
actually have been a language related to Semitic, which would explain
the Semitic-like words in IE. I no longer think so; NWC is a better
candidate. The Semitic-like words may have been contributed by the
Para-NWC language which may in turn have borrowed them from Semitic as
Neolithic Wanderwörter.

Finally, let me say a word about haplogroups. I think these are
overrated, as the percentages of these lineages can be altered
substantially by genetic drift, founder effects and similar things. This
may explain why both Western Europe (where a subclade of Y-DNA
haplogroup R1b dominates) and Eastern Europe (where a subclade of R1a is
most common) show different profiles than the Yamnaya people themselves
(where a subclade of R1b *different* from the Western European one
seemed to dominate). Also, the sample sizes of the archaeogenetic
studies are so small that minor haplogroups may be missed entirely. At
any rate, it is not advisable to connect Y-DNA haplogroups to language
families (as I have myself used to do for some time before I realized
that this was fallacious).

OK, this has become quite long, and I'll shut up now. It is now open to
discussion.

[bradrn]

OK, this has become quite long, and I'll shut up now. It is now open to discussion.

To paraphrase Pratchett, I praise it as conlanging of the highest order. Whether it corresponds to any real linguistic history is another story, and quite probably an unprovable one. Even Indo-Uralic has precious little evidence behind it.

(Also, what’s with the hard-wrapping?)

I do have one question on this:

also, Uralic is quite similar to Eskimo-Aleut despite the great geographical and probably also historio-linguistic distance between the two - the resemblance is certainly due to conservatism.

I’ve seen such statements before from Eskimo-Aleut-ists (especially Fortescue), but personally I don’t see any clear connections between the two, let alone them being ‘quite similar’. So what makes you say this?

(For context, I know a fair amount about Kalaallisut, though I don’t know much about Eskimo-Aleut or Uralic more generally.)

[Raphael]

also, Uralic is quite similar to Eskimo-Aleut despite the great geographical and probably also historio-linguistic distance between the two - the resemblance is certainly due to conservatism.

I’ve seen such statements before from Eskimo-Aleut-ists (especially Fortescue), but personally I don’t see any clear connections between the two, let alone them being ‘quite similar’. So what makes you say this?

(For context, I know a fair amount about Kalaallisut, though I don’t know much about Eskimo-Aleut or Uralic more generally.)

I'm not sure about the "great geographical distance" in the first place. The different parts of the Far North are closer to each other than you'd guess from looking at them on a standard-issue rectangular, or mostly rectangular, world map.

[WeepingElf]

OK, this has become quite long, and I'll shut up now. It is now open to discussion.

To paraphrase Pratchett, I praise it as conlanging of the highest order. Whether it corresponds to any real linguistic history is another story, and quite probably an unprovable one. Even Indo-Uralic has precious little evidence behind it.

Thank you.

(Also, what’s with the hard-wrapping?)

That's because I copied and pasted it from a CONLANG list post which had hard-wrapping, and I didn't feel like editing the line breaks out as they IMHO do not impede readability much. In fact, I find it a bit easier to read because the lines aren't too long

I do have one question on this:

also, Uralic is quite similar to Eskimo-Aleut despite the great geographical and probably also historio-linguistic distance between the two - the resemblance is certainly due to conservatism.

I’ve seen such statements before from Eskimo-Aleut-ists (especially Fortescue), but personally I don’t see any clear connections between the two, let alone them being ‘quite similar’. So what makes you say this?

(For context, I know a fair amount about Kalaallisut, though I don’t know much about Eskimo-Aleut or Uralic more generally.)

There are some resemblances in the inflectional morphology and in the consonant systems of Proto-Uralic and Proto-Eskimo-Aleut (which were already noticed by Rasmus Rask in 1814).

[Zju]

What are the more plausible internal reconstructions of *éǵh₂ 'I'? I vaguely recall reading a hypothesis that *-om of *éǵh₂om could be connected with the secondary active verbal endings and *éǵh₂ with some pronoun - maybe related to *h₁eǵʰs 'out' - thus originally meaning something along the lines of "I'm here".

[WeepingElf]

In this post, I shall lay out my thinking about the origins of the Indo-European family. [...]

Since then, I have grown some doubts about this. There is the following article by David W. Anthony, a linguistically and also genetically well-informed archaeologist specializing in Neolithic and Bronze Age steppe cultures:

<https://www.academia.edu/39985565/Archa ... on_Bomhard>

Which refers to the following paper by geneticists:

<https://www.biorxiv.org/content/biorxiv ... 7.full.pdf>

As a non-geneticist, I don't find it easy to get a handle on these matters, but apparently, the "Caucasian" component in the Yamnaya gene pool is quite different from the Northwest Caucasians. I have noticed quite some time ago that the Y-DNA haplogroups are very different (but things like gene drift or founder effects can mess up such things). Apparently, the "Caucasians" contributing to the Yamnaya people came from somewhere like the Iranian highlands, and their language is lost, being clobbered by the Iranian branch of IE in their homeland in the Bronze Age.

The NW Caucasian languages do not fit typologically well, anyway. The phonologies do not really match, NW Caucasian has way more sibilants and laterals than PIE; the only resemblance is that there are palatalized and labialized velars, but NW Caucasian applies these secondary articulations to consonants at other points of articulation as well. The two families actually seem to have nothing in common except a few loanwords (as expected from neighbouring languages that have been in contact with each other, as the similarities between the Yamnaya and Maykop cultures show). Regarding the grammatical typology, the best match among the three Caucasian families is not NW Caucasian but Kartvelian (but the tripartite verb aspect system in PIE appears to be a Late PIE innovation happening only after Anatolian had broken off).

[bradrn]

Regarding the grammatical typology, the best match among the three Caucasian families is not NW Caucasian but Kartvelian (but the tripartite verb aspect system in PIE appears to be a Late PIE innovation happening only after Anatolian had broken off).

Along similar lines, it has occurred to me that Uralic has some similarities to NE Caucasian, especially in the case system. (Mind you, Burushaski also has a three-way distinction in the case system, and that seems to be recently innovated, so it could be a coincidence.)

[WeepingElf]

Regarding the grammatical typology, the best match among the three Caucasian families is not NW Caucasian but Kartvelian (but the tripartite verb aspect system in PIE appears to be a Late PIE innovation happening only after Anatolian had broken off).

Along similar lines, it has occurred to me that Uralic has some similarities to NE Caucasian, especially in the case system. (Mind you, Burushaski also has a three-way distinction in the case system, and that seems to be recently innovated, so it could be a coincidence.)

Burushaski is typologically very close to NE Caucasian, with its four-gender system and its "case construction kit". Yet, such typological resemblances cannot be used to construct a relationship hypothesis if the morphemes themselves are dissimilar. Likewise, the typological resemblances between IE and Kartvelian are not sufficient to construct a relationship between Kartvelian and the "Caucasian" substratum in PIE. Also, the Georgians are apparently genetically not very similar to the "Caucasian" component in the Yamnaya, either.

[WeepingElf]

Anthony's main concern about the NW Caucasian substratum hypothesis, if I understand his article correctly, appears to be that the Maykop people, who probably spoke Proto-NWC, had a genetic admixture from Anatolian farmers of a type that is apparently absent in the Yamnaya. But couldn't this have come into the Maykop people after their cousins mixed with the Eastern Europeans to produce the Yamnaya population? Not being a geneticist, I cannot say whether this makes sense from the genetic standpoint, but it may explain the discrepancy Anthony is worried about.

[WeepingElf]

I have taken another look at Lazaridis et al. 2024. While I don't understand everything they write, what I understand gives a rather clear picture to me. I don't question what they are doing as long as they are doing genetics. But when it comes to linguistic conclusions, well, I see some things differently than they do. To put it short, the story the paper tells can be summed up like this: "Two populations, one moving from the Caucasus northward into the steppe, the other from the Volga-Kama confluence southward down the Volga, meet, mix and merge at the Lower Volga. Then, they domesticate the horse, and go on to conquer Europe and India." So far, so good.

But which languages did those people speak? I think that the Caucasus-Lower Volga (CLV) cline can be connected to the NW Caucasian family, as its southern end is firmly rooted in Maykop/Proto-NWC territory, while the Volga cline can be connected with a language related to Proto-Uralic. Where the two clines meet sit the Yamnaya people, who speak PIE - basically a Para-Uralic language on a Para-NWC substratum. The third cline, the Dnipro cline, represents another admixture from Ukrainian hunter-gatherers who also have influenced PIE.

That means that the language that was carried from the southern end of the CLV cline into Anatolia was not, as Lazaridis at al. conclude, Anatolian, but something related to NWC, probably Hattic. The problem with Lazaridis et al.'s assumption is that Anatolian cannot have broken off the rest of IE before the latter contacted with Para-Uralic, as it clearly shows the Para-Uralic morphological elements such as the pronouns. This was my main objection against the "Southern Arc theory", and it still holds.

This raises the question when and which way Anatolian got to Anatolia instead. I think the linguistic and historical evidence points at an origin from the northwest, and that we don't see it in the genetic record is because it went to Anatolia by means of a language shift instigated by small elites, which is invisible in the genetic record. The starting point of this language shift would have been the western outlier of Yamnaya on the lower Danube around 3000 BC, where a conservative dialect of PIE may have been spoken that had not been reached by innovations from the Yamnaya heartland, as is often the case with such outliers. From the same outlier, but even further west (it extended into southern Hungary), would have been the Bell Beaker people, whose language may thus represent an even more archaic PIE dialect, but only very little can be said about that language except sheer speculation of the sort that can be rolled up into a conlang