Zompist Bboard Again

There is probably no way deciding which of those options is the correct one without reliable outgroup comparison - but we don't know what is the outgroup. Uralic seems to be the best candidate, but there are apparently too few cognates to establish sound correspondences, and at this point we cannot rule out Kartvelian, Etruscan, the two North Caucasian families (whether they are related to each other or not), or even Semitic. From genetic evidence, it seems as if the Yamnaya people resulted from the mixture of an Eastern European population with a strong Siberian component related to the likely Proto-Uralic speakers with a population of Transcaucasian origin around 5000 BC, so it seems plausible that PIE is a language related to Proto-Uralic heavily altered by a Transcaucasian substratum (whatever that may have been like), with many cultural terms including the entire agricultural terminology coming from the unknown Transcaucasian language. But as they say, genes don't speak languages (nor do pots).

Another matter I have thought about today and wish to discuss here. After the "centum/satem" dichotomy fell out of fashion, it became fashionable to divide the "classic" branches (i.e., all except Anatolian and Tocharian) into "Northwest IE" (Italic, Celtic, Germanic, Balto-Slavic) and "Greco-Aryan" (Indo-Iranian, Armenian, Greek, maybe also Albanian). Besides various words that are only found in one of the two groups, much ado is made of the "fact" that "Greco-Aryan" (actually just Ancient Greek and the oldest stages of Indo-Iranian) "preserves" the tripartite aspect system, while "Northwest IE" has "lost" it; it has recently even become fashionable to doubt that the "Northwest IE" languages ever had it.

But I think this is just an artifact of the different ages of the relevant literatures. The oldest "Greco-Aryan" literature is considerably older than the oldest "Northwest IE" literature, and in later stages, even in the "Greco-Aryan" languages the tripartite verb aspect system broke down, in very similar ways as in "Northwest IE", and not much later. The difference is merely that in Greek and Indo-Iranian, it happened in historical time, while in "Northwest IE" it happened in prehistoric time - but not much earlier because the latter group became literate later.

After the "centum/satem" dichotomy fell out of fashion

Did it? When?

it became fashionable to divide the "classic" branches (i.e., all except Anatolian and Tocharian) into "Northwest IE" (Italic, Celtic, Germanic, Balto-Slavic) and "Greco-Aryan" (Indo-Iranian, Armenian, Greek, maybe also Albanian).

Isn't Balto-Slavic closer to II than to any of "Northwest IE"? Or at least closer to II than any other branch of "NW IE" is, as the wave model would suggest.

Zju wrote: ↑Fri Feb 17, 2023 2:51 pm

After the "centum/satem" dichotomy fell out of fashion

Did it? When?

It is still common practice to describe IE languages as "satem" or "centum", but nobody believes that these categories were branches in the IE family tree. It is just one isogloss out of many.

it became fashionable to divide the "classic" branches (i.e., all except Anatolian and Tocharian) into "Northwest IE" (Italic, Celtic, Germanic, Balto-Slavic) and "Greco-Aryan" (Indo-Iranian, Armenian, Greek, maybe also Albanian).

Isn't Balto-Slavic closer to II than to any of "Northwest IE"? Or at least closer to II than any other branch of "NW IE" is, as the wave model would suggest.

Yes, Balto-Slavic has much in common with Indo-Iranian, and this shows that the idea that Post-Tocharian split into a "Northwest" branch and a "Greco-Aryan" branch is most likely bullfrogs.

Nortaneous wrote: ↑Fri Feb 17, 2023 8:34 am (Couldn't front vowels voice surrounding consonants in Turkic or something like that?)

Yep. Stop voicing is associated with advanced tongue root, which means that front vowels can voice consonants as seen in Oghuz Turkic. It can also go the other way round, like Adjarian's law in Armenian where vowels were fronted following PIE voiced aspirates, which sets an interesting parallel to Zhivlov. I'd never heard of that theory before but it could easily explain PIE as coming from **T(h) D > *T(h)~Dh D in frontness conditions, and it's certainly a lot neater than most explanations.

I was thinking about a lot of the proposed pre-PIE phonological theories (e.g. velars were actually uvulars, the whole mess of stop series realizations, *o was a low vowel, etc.) and a lot of them seem to have one thing in common. Specifically, they require an areal sound change that affects all IE dialects, but then causes different side effects in different groups. For example, many who theorize that velars were originally uvulars posit that *k shifted from original /q/ to later /k/ in all IE branches, but in satem branches this shifted original *k^' forward to "palatovelar" while in centrum branches no such thing happened and the two series simply merged. Or, that the *D series lost most (or all?) of its "glottalic" properties as an areal change across almost all IE branches, with different effects on the other stop series in many of them.
What I'm trying to ask is, is this an attested phenomenon observed in actual languages? Where an areal change affects an entire group of dialects but causes different chain shifts / mergers in different ones? Or is this more of an ad-hoc explanation used to justify PIE phonological theories that are hard to reconcile with the evidence otherwise?

It seems as if PIE underwent a massive merger of vowels (which I call the "Great Vowel Collapse", abbreviated "GVC") before (or in the course of) the rise of the ablaut system, and it may be the case that the *T(h)/*Dh split conserves old vowel quality differences, similar to the palatalization and labialization of velars (for the latter, I guess that it was frontness and roundedness of vowels - apparently, pre-GVC PIE had no front rounded vowels). Of course, the voicing of aspirated stops would be caused by a different axis of the vowel system, perhaps openness.

WeepingElf wrote: ↑Sat Feb 18, 2023 4:48 am It seems as if PIE underwent a massive merger of vowels (which I call the "Great Vowel Collapse", abbreviated "GVC") before (or in the course of) the rise of the ablaut system, and it may be the case that the *T(h)/*Dh split conserves old vowel quality differences, similar to the palatalization and labialization of velars (for the latter, I guess that it was frontness and roundedness of vowels - apparently, pre-GVC PIE had no front rounded vowels). Of course, the voicing of aspirated stops would be caused by a different axis of the vowel system, perhaps openness.

How exactly do you determine the vowels of pre-GVC PIE? That's a genuine question, not a criticism. I'd be interested in seeing how you envision the GVC-to-ablaut transition.

Darren wrote: ↑Sat Feb 18, 2023 5:22 am How exactly do you determine the vowels of pre-GVC PIE? That's a genuine question, not a criticism. I'd be interested in seeing how you envision the GVC-to-ablaut transition.

At the present state of knowledge, there is no way telling for sure, though in some cases, guesses are possible. For instance, the antecedent of *k^wel- 'to turn', the vowel apparently was back rounded, and it seems, from things I have seen in Uralic and Altaic etymological dictionaries, as if there was a Proto-Mitian root like **kul- with the general meaning 'to turn' (if Mitian is real, of course).

WeepingElf wrote: ↑Fri Feb 17, 2023 11:55 am But I think this is just an artifact of the different ages of the relevant literatures. The oldest "Greco-Aryan" literature is considerably older than the oldest "Northwest IE" literature, and in later stages, even in the "Greco-Aryan" languages the tripartite verb aspect system broke down, in very similar ways as in "Northwest IE", and not much later. The difference is merely that in Greek and Indo-Iranian, it happened in historical time, while in "Northwest IE" it happened in prehistoric time - but not much earlier because the latter group became literate later.

In general, it looks like NW-European is, at best, an areal grouping, with some shared lexicon and tendencies.
On the tripartite system:
- Italic and Celtic show clear morphological remnants, so they had it and lost it (Latin merging Aorist and Perfect into what traditional Latin grammar calls Perfect)
- Germanic developed the perfect into the simple past and shows some debated remnants of the Aorist. My best bet is to assume, again, that it had the tripartite system and lost it
- Balto-Slavic doesn't show any evidence of ever having had the reduplicated perfect; it has a continuation of *woida "know", which, being non-reduplicated, is an outlier in PIE, and probably a remnant of an earlier category that was transformed into the reduplicating perfect in a IE core group after Anatolian split off, and it has some verbs with o-ablaut that seem to be related to the perfect system (like Slavic mog- "can"), which are present tense and non-reduplicating; so they may as well they go back to the same earlier category, instead of being de-reduplicated perfects. Also the Balto-Slavic thematic present tense endings look like a weird mixture of active and mediopassive endings, which (to me) is easier to explain if they were built based on something like the Anatolian hi-conjugation merged with the athematic endings, than as a reconstruction of the Graeco-Aryan active endings. Slavic has an Aorist, but it's cobbled together from formations that are Aorist in other IE languages and from IE imperfects; so my pet theory is that the Slavic Aorist (which has no equivalent in Baltic) was formed under Iranian influence, using some of the aspect stems that were used in other IE languages, but also the IE imperfect which may have become a simple past tense at one stage in the development. So my opinion is that Balto-Slavic was an outlier that split off after Anatolian, but before the tripartite system was formed, and one member of which (Slavic) was later brought nearer to the core IE system under the influence of Iranian. All communalities with other NW IE languages then are due to areal influence. This also means that Satemisation was an areal development and not a node (which, as Elf said, seems to become the consensus view).

hwhatting wrote: ↑Sun Feb 19, 2023 4:41 am So my opinion is that Balto-Slavic was an outlier that split off after Anatolian, but before the tripartite system was formed, and one member of which (Slavic) was later brought nearer to the core IE system under the influence of Iranian.

That's quite a bold proposal! I can't say that you are wrong; and I know that you know more about these matters than I do. But Balto-Slavic occupies a central position in the IE world (both in terms of geography and isoglosses), not that of an outlier; and I have never seen an IE family tree which has Balto-Slavic forking off early. But I concur with you that both "Northwest IE" and "Satem" are areal groupings within the family.

hwhatting wrote: ↑Sun Feb 19, 2023 4:41 am So my opinion is that Balto-Slavic was an outlier that split off after Anatolian

Before or after Tocharian?

abahot wrote: ↑Sun Feb 19, 2023 9:44 am

hwhatting wrote: ↑Sun Feb 19, 2023 4:41 am So my opinion is that Balto-Slavic was an outlier that split off after Anatolian

Before or after Tocharian?

I haven't made up my mind about that, mostly because I don't know much about Tocharian beyond what you can find in general introductions to Indo-European. I need to learn more about how its verbal morphology compares to Anatolian and Core PIE in order to see where it fits into my scenario. Some day I'll find the time to seriously engage with it...

WeepingElf wrote: ↑Sun Feb 19, 2023 7:34 am That's quite a bold proposal! I can't say that you are wrong; and I know that you know more about these matters than I do. But Balto-Slavic occupies a central position in the IE world (both in terms of geography and isoglosses), not that of an outlier; and I have never seen an IE family tree which has Balto-Slavic forking off early. But I concur with you that both "Northwest IE" and "Satem" are areal groupings within the family.

I can't say that I've worked out all the details. Isoglosses can be areal; Germanic must have originated in a more central area than where we find it in historical times (all those specific Germanic-Italic isoglosses), so Balto-Slavic may actually have been at the Northern edge of IE for some time; we know that the Iranian presence in the Pontic steppe area is a later development, so it's quite likely that there originally was a continuum of dialects more divergent from the Graeco-Aryan model, of which B-S (and maybe Tocharian) were a part, which later was replaced in the steppe area by (Indo-)Iranian.

In my opinion, Germanic is sort of an outlier, though not as crassly as Tocharian, let alone Anatolian. Germanic has many innovations it doesn't share with any other (neighbouring) branch, the most famous of which being "Grimm's Law" (recte "Rask's Law")[1]. In my model, the Northern IE dialect ancestral to Germanic was cut off from the rest of Northern IE when the Bell Beaker people spread a Southwest IE dialect (or whatever those people spoke) in Central Europe, and evolved in isolation for about 1,000 years, until the southward expansion of the Nordic Bronze Age and the Urnfield expansion which brought Proto-Celtic to Central Europe (and Proto-Italic to Italy) brought it again into contact with other Northern IE languages.

[1] There is evidence of Rask's Law happening early and other evidence of it happening late. This contradiction can IMHO be solved by assuming that it went on in two stages. In the first stage, the *T set acquired (or retained) aspiration and the *D set devoiced, as (independently) in Armenian. This would have happened during the isolation phase. In the second stage, the aspirated stops became spirants. This may have happened as late as 100 BC, explaining why the Cimbri and Teutones bore seemingly unshifted names, and how voiceless stops in Celtic matron names in the Rhineland could be affected by the shift (Celtic voiceless stops appear to have been aspirated). Verner's Law, which could work as well with aspirated stops as with spirants, and accent retraction may have happened at any time between the two stages (accent retraction was of course later than Verner's Law which relies on the old PIE accent).

WeepingElf wrote: ↑Mon Feb 20, 2023 1:26 pm There is evidence of Rask's Law happening early and other evidence of it happening late.

What evidence? I've personally been very interested in this topic, of when Grimm's Law happened.

the Cimbri and Teutones bore seemingly unshifted names

In particular, how did the Latin transliteration of said "unshifted names" deal with the Proto-Indo-European *D^h set? Were they reflected in the transliteration as plain voiced stops?

abahot wrote: ↑Mon Feb 20, 2023 10:00 pm

WeepingElf wrote: ↑Mon Feb 20, 2023 1:26 pm There is evidence of Rask's Law happening early and other evidence of it happening late.

What evidence? I've personally been very interested in this topic, of when Grimm's Law happened.

Oh, I should have said, "There apparently is evidence..."; I have seen dates from 2000 BC to 200 BC mentioned in the literature. Such sound changes are very hard to date. We definitely know that Verner's Law was before the accent shift, and the Wikipedia page on Verner's Law presents a scenario similar to mine.

the Cimbri and Teutones bore seemingly unshifted names

In particular, how did the Latin transliteration of said "unshifted names" deal with the Proto-Indo-European *D^h set? Were they reflected in the transliteration as plain voiced stops?

Probably, yes.

WeepingElf wrote: ↑Mon Feb 20, 2023 1:26 pm [1] There is evidence of Rask's Law happening early and other evidence of it happening late. This contradiction can IMHO be solved by assuming that it went on in two stages. In the first stage, the *T set acquired (or retained) aspiration and the *D set devoiced, as (independently) in Armenian. This would have happened during the isolation phase. In the second stage, the aspirated stops became spirants. This may have happened as late as 100 BC, explaining why the Cimbri and Teutones bore seemingly unshifted names, and how voiceless stops in Celtic matron names in the Rhineland could be affected by the shift (Celtic voiceless stops appear to have been aspirated). Verner's Law, which could work as well with aspirated stops as with spirants, and accent retraction may have happened at any time between the two stages (accent retraction was of course later than Verner's Law which relies on the old PIE accent).

None of that needs to have been early; devoicing of the D series must have happened after the Celtic loanwords (du:n- > tu:n-, ri:g > ri:k-) were borrowed. Germanic is an outlier phonetically, but the chain shift leading to it could have happened at any time. And keep in mind that another, somewhat similar chain shift happened in historical times in High German. So Grimm / Rask is no indication that Germanic split off early. Germanic is also peculiar in its tense system (no aorist, perfect > simple past), but there is actually not much eveidence that it started out from something different to the Graeco-Aryan system. To me, it rather looks like the peculiarities of Germanic developed when the speakers of a bog-standard Western European IE dialect migrated North into Scandinavia and came in contact with a non-IE substrate population; maybe as late as the bronze age.

hwhatting wrote: ↑Tue Feb 21, 2023 6:38 am

WeepingElf wrote: ↑Mon Feb 20, 2023 1:26 pm [1] There is evidence of Rask's Law happening early and other evidence of it happening late. This contradiction can IMHO be solved by assuming that it went on in two stages. In the first stage, the *T set acquired (or retained) aspiration and the *D set devoiced, as (independently) in Armenian. This would have happened during the isolation phase. In the second stage, the aspirated stops became spirants. This may have happened as late as 100 BC, explaining why the Cimbri and Teutones bore seemingly unshifted names, and how voiceless stops in Celtic matron names in the Rhineland could be affected by the shift (Celtic voiceless stops appear to have been aspirated). Verner's Law, which could work as well with aspirated stops as with spirants, and accent retraction may have happened at any time between the two stages (accent retraction was of course later than Verner's Law which relies on the old PIE accent).

None of that needs to have been early; devoicing of the D series must have happened after the Celtic loanwords (du:n- > tu:n-, ri:g > ri:k-) were borrowed. Germanic is an outlier phonetically, but the chain shift leading to it could have happened at any time. And keep in mind that another, somewhat similar chain shift happened in historical times in High German. So Grimm / Rask is no indication that Germanic split off early. Germanic is also peculiar in its tense system (no aorist, perfect > simple past), but there is actually not much eveidence that it started out from something different to the Graeco-Aryan system. To me, it rather looks like the peculiarities of Germanic developed when the speakers of a bog-standard Western European IE dialect migrated North into Scandinavia and came in contact with a non-IE substrate population; maybe as late as the bronze age.

Your points are good, thank you. So the devoicing of *D was a late change, I had failed to think of the Celtic loanwords (if that's what they are, at least *rîg- has a cognate in Indo-Aryan). Languages can change profoundly within a few centuries, as the Insular Celtic languages show.

EDIT: I have looked again at the list of datings for Rask's Law found in a 1973 talk by Antonio Tovar, and it is clear that the datings have become later and later as time progressed:

1913 - 2nd millennium BC (F. Kluge)
1940 - 800-650 BC (W. Steinhauser)
1948 - 1500 BC (J. Fourquet)
1962 - 2000 to 200 BC (H. Stolte)
1973 - 500 BC or later (most scholars)
2023 - 100 BC (various scholars)

Apparently, early IEists assumed that Rask's Law was pretty much the first thing that happened in Germanic after the breakup of PIE, and now it seems as if it had happened only a few generations ago when the Romans showed up in Germany, and had not yet happened when the Cimbri and Teutones attacked Rome. It appears to have happened late enough to affect Celtic (or Pre-Celtic) place names and matron names derived thereof in an area that had only just been conquered by Germanic tribes in the 1st century BC.

WeepingElf wrote: ↑Tue Feb 21, 2023 7:23 am I had failed to think of the Celtic loanwords (if that's what they are, at least *rîg- has a cognate in Indo-Aryan).

It's generally taken as a loan from Celtic, because PIE /e:/ > /i:/ is a Celtic sound change that you don't find in IA or Italic, and ri:g is attested as an independent word in Old Irish (NSg. rí) and as an element in Gaulish personal names.

WeepingElf wrote: ↑Tue Feb 21, 2023 7:23 am Languages can change profoundly within a few centuries, as the Insular Celtic languages show.

Yes, that's a very good example.
By the way, this showed up today in my Academia.edu feed: Kloekhorst's paper "Anatolian evidence suggests that the Indo-European laryngeals *h₂ and *h₃ were uvular stops [2018]". As I said, there's a lot of work being done right now that hasn't yet found its way back into the standard introductions.

hwhatting wrote: ↑Tue Feb 21, 2023 9:28 am

WeepingElf wrote: ↑Tue Feb 21, 2023 7:23 am I had failed to think of the Celtic loanwords (if that's what they are, at least *rîg- has a cognate in Indo-Aryan). Languages can change profoundly within a few centuries, as the Insular Celtic languages show.

It's generally taken as a loan from Celtic, because PIE /e:/ > /i:/ is a Celtic sound change that you don't find in IA or Italic, and ri:g is attested as an independent word in Ol Irish (NSg. rí) and as an element in Gaulish personal names.

I understand.

By the way, this showed up todayin my Academia.edu feed: Kloekhorsts paper "Anatolian evidence suggests that the Indo-European laryngeals *h₂ and *h₃ were uvular stops [2018]". We live in interesting times.

I have seen it; I am not convinced. It requires the change q > χ to happen at least three times: 1. in Non-Anatolien IE; 2. in Non-Luvic Anatolian; 3. in Non-Lycian Luvic. Also it is questionable to me to build such a theory on a language which we can read but don't really know for sure how those letters were pronounced. And even if the Lycian reflex of the laryngeals was indeed [q] and not [χ] or whatever, what speaks against a change χ > q in Lycian? Maybe there was a substratum involved which had [q] but not [χ].

Kloekhorst is a creative and adventurous scholar, but some of his ideas strike me as unjustified.