The Great Proto-Indo-European Thread

[WeepingElf]

Fair. There is indeed no evidence of *o being rounded in Anatolian, where it merged with *a. Yet, I think it is quite certain that *o was rounded in Late PIE - with the exception of Tocharian (which is weird anyway), all branches that keep it separate from *a reflect it as a rounded vowel. However, I do entertain the notion that the non-high vowels huddled quite close together in Late PIE - *e was *[æ], *o was *[ɒ], which earlier may have indeed been *[ɑ].

It may have gone like this:

1. Pre-PIE *a
2. Ablaut: *a splits into *æ and *ɑ (and zero)
3. *h2 backs *æ to (new) *a; *h3 backs *æ to (already existing) *ɑ
4a. *a and *ɑ merge in Anatolian
4b. *ɑ rounds to *ɒ in Late PIE
5ba. *a and *ɒ merge in some non-Anatolian IE languages
5bb. *æ > *e and *ɒ > *o in some other non-Anatolian languages (no new phonemes created, only the vowels become more distinct in sound)

Under such a scenario, PIE *h3 may have not been labialized, but simply posterior to *h2. So *h2 may have been a front velar and *h3 a back velar fricative, homorganic to *ḱ and *k, respectively. *h1 may just have been */h/. My old "three velar series" model had not only the problem of *h2 being the most abundant, but also the problem of *h1 being lost in Hittite. Also, of course, the problem of the velar stops not colouring vowels. Question: what happened to the labialized velar fricative? Did it perhaps merge with *w? (May explain the *wl- and *wr- onsets!)

[Polka Dot]

Where does Tocharian fit in the centum–satem division ?

[KathTheDragon]

Where does Tocharian fit in the centum–satem division ?

Centum

Fair. There is indeed no evidence of *o being rounded in Anatolian, where it merged with *a. Yet, I think it is quite certain that *o was rounded in Late PIE - with the exception of Tocharian (which is weird anyway), all branches that keep it separate from *a reflect it as a rounded vowel. However, I do entertain the notion that the non-high vowels huddled quite close together in Late PIE - *e was *[æ], *o was *[ɒ], which earlier may have indeed been *[ɑ].

It may have gone like this:

1. Pre-PIE *a
2. Ablaut: *a splits into *æ and *ɑ (and zero)
3. *h2 backs *æ to (new) *a; *h3 backs *æ to (already existing) *ɑ
4a. *a and *ɑ merge in Anatolian
4b. *ɑ rounds to *ɒ in Late PIE
5ba. *a and *ɒ merge in some non-Anatolian IE languages
5bb. *æ > *e and *ɒ > *o in some other non-Anatolian languages (no new phonemes created, only the vowels become more distinct in sound)

Under such a scenario, PIE *h3 may have not been labialized, but simply posterior to *h2. So *h2 may have been a front velar and *h3 a back velar fricative, homorganic to *ḱ and *k, respectively. *h1 may just have been */h/. My old "three velar series" model had not only the problem of *h2 being the most abundant, but also the problem of *h1 being lost in Hittite. Also, of course, the problem of the velar stops not colouring vowels. Question: what happened to the labialized velar fricative? Did it perhaps merge with *w? (May explain the *wl- and *wr- onsets!)

That scenario is basically what I believe, yes. I leave some wiggle-room for the rounding of *o to have not happened in all the dialects, but at that point it's just splitting hairs. Ditto for the observations about laryngeals - with regards to the labio-velar fricative, I find it likely that it is continued in Anatolian *ḫʷ and "inner" IE *h₂w, but *w before *l and *r is very likely (and I think someone, probably Tropylium, adduced a very interesting connection along those lines)

[Salmoneus]

Fair. There is indeed no evidence of *o being rounded in Anatolian, where it merged with *a. Yet, I think it is quite certain that *o was rounded in Late PIE - with the exception of Tocharian (which is weird anyway), all branches that keep it separate from *a reflect it as a rounded vowel. However, I do entertain the notion that the non-high vowels huddled quite close together in Late PIE - *e was *[æ], *o was *[ɒ], which earlier may have indeed been *[ɑ].

It may have gone like this:

1. Pre-PIE *a
2. Ablaut: *a splits into *æ and *ɑ (and zero)
3. *h2 backs *æ to (new) *a; *h3 backs *æ to (already existing) *ɑ
4a. *a and *ɑ merge in Anatolian
4b. *ɑ rounds to *ɒ in Late PIE
5ba. *a and *ɒ merge in some non-Anatolian IE languages
5bb. *æ > *e and *ɒ > *o in some other non-Anatolian languages (no new phonemes created, only the vowels become more distinct in sound)

Under such a scenario, PIE *h3 may have not been labialized, but simply posterior to *h2. So *h2 may have been a front velar and *h3 a back velar fricative, homorganic to *ḱ and *k, respectively. *h1 may just have been */h/. My old "three velar series" model had not only the problem of *h2 being the most abundant, but also the problem of *h1 being lost in Hittite. Also, of course, the problem of the velar stops not colouring vowels. Question: what happened to the labialized velar fricative? Did it perhaps merge with *w? (May explain the *wl- and *wr- onsets!)

While this is neat, I'm not sure how feasible it is to produce all the ablaut from just one vowel. In particular, while some stressed *o and unstressed *e can be explained through later analogy, I'm not sure how this could produce *o in initial syllables (as in acrostatic nouns and causative verbs). Nor does it explain Nartens.
It seems as though there would be several ways to explain the results if you assumed two initial vowels - perhaps *a and *o, or perhaps *a and *a:
I think they'd all have to involve multiple rounds of ablaut process and stress shift, though.

[WeepingElf]

While this is neat, I'm not sure how feasible it is to produce all the ablaut from just one vowel. In particular, while some stressed *o and unstressed *e can be explained through later analogy, I'm not sure how this could produce *o in initial syllables (as in acrostatic nouns and causative verbs). Nor does it explain Nartens.
It seems as though there would be several ways to explain the results if you assumed two initial vowels - perhaps *a and *o, or perhaps *a and *a:
I think they'd all have to involve multiple rounds of ablaut process and stress shift, though.

Certainly, there are lots of open questions regarding the origin of PIE ablaut; I don't seriously expect to be able to crack this riddle - if I did, my name would be inscribed in all future PIE handbooks. Of course, breakthroughs have been made by people who weren't eminent Indo-Europeanists, as Saussure's discovery of the laryngeals as a student and Ventris's decipherment of Linear B demonstrate; but I don't expect to be the next of those wonders. All I can do is throw out ideas for discussion; some may shoot in the right direction, others not.

[KathTheDragon]

It seems as though there would be several ways to explain the results if you assumed two initial vowels - perhaps *a and *o, or perhaps *a and *a:
I think they'd all have to involve multiple rounds of ablaut process and stress shift, though.

I operate with *a and *ā - Original *ā produces *o subject to Brugmann's law, while coloured *a produces *o not subject to Brugmann's law. I actually have a solution that requires only three processes - reduction, post-tonic lengthening, and a stress shift, with various morphological changes ordered between them. See here for a fuller discussion.

[WeepingElf]

I operate with *a and *ā - Original *ā produces *o subject to Brugmann's law, while coloured *a produces *o not subject to Brugmann's law. I actually have a solution that requires only three processes - reduction, post-tonic lengthening, and a stress shift, with various morphological changes ordered between them. See here for a fuller discussion.

Interesting! I will certainly have to re-read it several times, and check how good an account this makes of the facts.

[KathTheDragon]

I'm open to any suggestions you have for improving my theory! I'll admit, a lot of it (in particular the high vowels) came to me while I was writing part 1, so I'm not expecting it to work perfectly.

[Soap]

I'm open to any suggestions you have for improving my theory! I'll admit, a lot of it (in particular the high vowels) came to me while I was writing part 1, so I'm not expecting it to work perfectly.

I like your theory a lot. Was it influenced by Caucasian languages where the only claimed vowels are /a/ and /ə/? While I think we should be cautious not to assume nothing has changed in that area in the last 5000 years, I wouldnt be surprised to find out that in the days of PIE there existed languages in that area with the same simple vowel system of just /a ə/ with massive allophony producing every other vowel.

Some Caucasian languages have a three-way contrast of /ə a ā/ ... perhaps PIE was once like that, but deleted all schwas unconditionally, meaning that, for example, /jə/ and /əj/ both became /i/, but ablaut processes could still insert a full vowel on either side of what was once just a glide. /a ā/ would still correspond mostly to canonical PIE /e o/.

[KathTheDragon]

Actually, no, the entire thing came about through internal reconstruction. I didn't want my post to follow the same process I went along to create the theory, because that would've been unnecessarily dull and I can't remember half of it anyway, so that doesn't really come across at all. On the subject of long-distance comparison, I would be interested to see how well this tallies with Uralic in particular, where there is a much more complex vowel system. All that said, I have suspected that even the length contrasts aren't original, but may go back to some kind of Pre-Pre-PIE accent, since e.g. Latin ūrīna clearly shows a root *uh₁r-, the zero-grade of what was apparently the old nominative of "water", *wóh₁r̥, where we should not expect to find any zero-grades at all if the root had an inherent Pre-PIE **ā. On the other hand, the breakdown of the original ablaut system could easily have opened the door to analogical changes, as it must have done elsewhere. This really is testing the boundaries of the capabilities of reconstruction.

[kanejam]

Latin ūrīna clearly shows a root *uh₁r-, the zero-grade of what was apparently the old nominative of "water", *wóh₁r̥, where we should not expect to find any zero-grades at all if the root had an inherent Pre-PIE **ā.

I haven't seen the root *wóh₁r̥ before, but it looks to be an variant of *wódr̥ with the same d ~ h₁ alternation seen in *déḱm̥t and *h₁ḱm̥tóm. Kortlandt has an appealing explanation of this root on page 155 of this article, which tries to disprove the existence of ablauting acrostatic nouns.

It is at least somewhat at odds with the idea that *e/o were originally **a/ā though. I did entertain the notion that *e/o were originally **ə/a, with the changes: **ə i u [+stress] > **a əi əu > *o ei eu. This goes some way to explaining ablaut as well as the fact that *e is so ubiquitous in roots, but runs into similar problems to **a/ā.

Kath, I like your blog my only suggestion would be to have a few more examples. I look forward to seeing more.

[KathTheDragon]

Thanks! I'll try to sneak in extra examples (perhaps if I can work out a good way to do footnotes, I don't want to disrupt the flow)

I have read Kloekhorst's article, although I sourced the notion of a variant of *wed- with *h₁ from this article. It is of course possible that the o/e ablauting nouns were not stressed acrostaticly, and my theory does still work if we modify the stress shift rule to accomodate it. In that case, we would instead have to suppose that the stress was retracted to a non-low vowel, and source the vowel of e-reduplication from an original *i (as opposed to the *ī of i-reduplication). This is interesting, since under Jasanoff's h₂e-conjugation theory (which I accept), the e-reduplicating presents were mi-conjugated, while the i-reduplicating presents were h₂e-conjugated, and were also uniformly stressed on the reduplicant and showed no ablaut. So we have a correlation of a long reduplicant with fixed stress on the reduplicant (though it can't be excluded that they were ending-stressed prior to the stress shift in all or some forms) and a short reduplicant with mobile stress between root and ending. It's also worth mentioning an attempt to unify the two kinds of reduplication - though I don't recall the details, I believe they proposed a paradigm with a stressed *i and an unstressed *e in the reduplicant. It's certainly compatible with my theory to have accent affect the high vowels prior to the lowering of short high vowels.

[Zju]

Thanks! I'll try to sneak in extra examples (perhaps if I can work out a good way to do footnotes, I don't want to disrupt the flow)

I have read Kloekhorst's article, although I sourced the notion of a variant of *wed- with *h₁ from this article. It is of course possible that the o/e ablauting nouns were not stressed acrostaticly, and my theory does still work if we modify the stress shift rule to accomodate it. In that case, we would instead have to suppose that the stress was retracted to a non-low vowel, and source the vowel of e-reduplication from an original *i (as opposed to the *ī of i-reduplication). This is interesting, since under Jasanoff's h₂e-conjugation theory (which I accept), the e-reduplicating presents were mi-conjugated, while the i-reduplicating presents were h₂e-conjugated, and were also uniformly stressed on the reduplicant and showed no ablaut. So we have a correlation of a long reduplicant with fixed stress on the reduplicant (though it can't be excluded that they were ending-stressed prior to the stress shift in all or some forms) and a short reduplicant with mobile stress between root and ending. It's also worth mentioning an attempt to unify the two kinds of reduplication - though I don't recall the details, I believe they proposed a paradigm with a stressed *i and an unstressed *e in the reduplicant. It's certainly compatible with my theory to have accent affect the high vowels prior to the lowering of short high vowels.

Interesting. How much does that coincide with this proposal laid out a year ago? Admittedly, not finished yet.

[KathTheDragon]

Oh my, that's a lot of stuff to read. I'll work through it and get back to you.

Edit: I found the paper I mentioned earlier. Having reread it, I realise I misremembered its conclusion, that reduplicated presents had accented e-grade and unaccented i-grade in the reduplicant. Thus, it's essentially the same as what you linked (and I suspect the theory has in fact been shared between them). I don't accept their conclusion that the "polythematic" approach is wrong, since their rejection of it is based purely on the quality of being "polythematic".

[kanejam]

I have read Kloekhorst's article, although I sourced the notion of a variant of *wed- with *h₁ from this article. It is of course possible that the o/e ablauting nouns were not stressed acrostaticly, and my theory does still work if we modify the stress shift rule to accomodate it.

I gotta say, *wóh₁r̥ ~ *udéns is some pretty cool stem alternation! I meant more the o ~ zero ablaut, where you only predict o ~ e and e ~ zero (correct me if I'm misunderstanding).

This is interesting, since under Jasanoff's h₂e-conjugation theory (which I accept) [...] I don't accept their conclusion that the "polythematic" approach is wrong, since their rejection of it is based purely on the quality of being "polythematic".

I am also a fan of the h₂e-conjugation theory, and I agree that the authors are quick to dismiss the polythematic theory, although the conclusion *Ce-CeiC-~ *Ci-CiC- *Ce-CeuC-~ *Cu-CuC- *Ce-CeC- ~ *C?-CC- (replaced by *Ci-CC-) is pretty darn satisfying.

[Soap]

According to Wikipedia, due to Kluge's law proto-Germanic is reconstructed with similarly extreme forms of stem mutation in nouns. For example tōgô "twig", genitive takkiz, and wekô "wick", genitive wukkiz.

https://en.wikipedia.org/wiki/Kluge%27s ... stem_nouns

Im a bit skeptical that they wouldnt have been regularized by the time they got that wild, but perhaps they were following a more predictable pattern that isnt visible when there just a few examples provided.

[Pole, the]

I have read Kloekhorst's article, although I sourced the notion of a variant of *wed- with *h₁ from this article. It is of course possible that the o/e ablauting nouns were not stressed acrostaticly, and my theory does still work if we modify the stress shift rule to accomodate it.

I gotta say, *wóh₁r̥ ~ *udéns is some pretty cool stem alternation!

I don't know, but *wóh₁r̥ is interestingly identical to how the same word sounds in some present-day dialects of English: [woʔər].

[KathTheDragon]

I meant more the o ~ zero ablaut, where you only predict o ~ e and e ~ zero (correct me if I'm misunderstanding).

No, you understand correctly (though I do also predict static é and ó). Frankly, there is nothing in the way of assuming that the Hittite oblique stem witen- reflects a virtual *wedén- and Kloekhorst is very quick to assume his /ɨ/ wherever there's an unexpected <e> or <i>, even where there's not even any orthographic alternation (see the verb terepp- where he assumes that the weak stem must be /terɨp/, despite the signs RE/I and E/IP being ambiguous)

the conclusion *Ce-CeiC-~ *Ci-CiC- *Ce-CeuC-~ *Cu-CuC- *Ce-CeC- ~ *C?-CC- (replaced by *Ci-CC-) is pretty darn satisfying.

It's certainly an interesting conclusion!

[KathTheDragon]

According to Wikipedia, due to Kluge's law proto-Germanic is reconstructed with similarly extreme forms of stem mutation in nouns. For example tōgô "twig", genitive takkiz, and wekô "wick", genitive wukkiz.

https://en.wikipedia.org/wiki/Kluge%27s ... stem_nouns

Im a bit skeptical that they wouldnt have been regularized by the time they got that wild, but perhaps they were following a more predictable pattern that isnt visible when there just a few examples provided.

Kluge's law was apparently a very pervasive phenomenon, which would've regularly produced synchronic alternation in n-stem nouns and nasal iteratives (from neh₂-stem verbs). It seems likely that there was a synchronic rule generating these alternations, and at least the gemination persisted at least up to Proto-Germanic, but the consonantism could be, and probably was, levelled in at least some cases (producing the secondary geminate series *þþ and *dd)

[Tropylium]

While suggestive, such long-distance comparison is hardly conclusive, and you'll also note such exceptions as *kewe "female animal" = *h₂₃ówi- "sheep" which doesn't fit the pattern whichever laryngeal you reconstruct. Also of detriment is that at least one of his PIE roots doesn't even exist! (It's *h₃pe(w)s-, see here)

Reference to obsolete or unlikely Uralic reconstructions is likewise a problem in most comparative studies of this sort as well. There are often also multiple possible comparisons, e.g. I think the "*kewe" root (Eastern Sami 'reindeer cow', Livonian 'mare') is more likely to be connected to *gʷou-; and also more likely as separate loanwords than as ancient heritage.

*h₂ ~ *k looks to me like the most solidly established connection; *h₃ ~ *k and especially *h₁ ~ *k seem weaker. Other than the fact that all three were relatively weak consonants by late PIE, I don't see much reason to assume that the laryngeals go back to some kind of a more general fricative series. I might have noted this already, but phonotactically there seems to be a clear divide between *s and *h₂ on one hand (often occurring in two-obstruent clusters such as *ḱs and *th₂), *h₁ and *h₃ on the other (which seem to be more common in clusters of the type *HR, *RH).

[WeepingElf]

One reason to assume that *h3 was labialized is IMHO perhaps the famous "triple reflex" of the syllabic laryngeals in Greek, which happened well after the breakup of Late PIE at a time (since it happened only in Greek) when *o was most likely already rounded. What do you think about this?

[Pole, the]

One reason to assume that *h3 was labialized is IMHO perhaps the famous "triple reflex" of the syllabic laryngeals in Greek, which happened well after the breakup of Late PIE at a time (since it happened only in Greek) when *o was most likely already rounded. What do you think about this?

Well, it doesn't need to be a reflex of labialization, it can be just a reflection of the backness, i.e. see how in some dialects of German, syllabic /ʀ/ is pronounced [ɔ].

[Tropylium]

One reason to assume that *h3 was labialized is IMHO perhaps the famous "triple reflex" of the syllabic laryngeals in Greek, which happened well after the breakup of Late PIE at a time (since it happened only in Greek) when *o was most likely already rounded. What do you think about this?

Did it happen only in Greek?

As far as I can see, there are three readily available ways to route the Greek development:
1) Schwa epenthesis, followed by laryngeal coloring applying more generally in pre-Greek than in other early IE dialects;
2) Triple vocalization originally occurred everywhere, with distinct vocalic values *ə₁, *ə₂, *ə₃ (we could suggest e.g. [ɪ ɐ ʊ]); later on, these merge everywhere except in Greek, where they rather merge with a distinct full vowel each;
3) There were full vowels here to begin with, and the Greek reflexes are simply archaisms. All other IE varieties instead subjected vowels in these positions to vowel reduction.

I don't think I find any of these immediately palatable, though. 1) seems problematic due to requiring a quite late date for laryngeal coloring, which is regardless rather non-trivial and applies across the entire IE family. 2) seems to require an unlikely transitional set-up with three marginal vowel phonemes, contrastive with each other but none of them contrastive with zero. Finally, 3) has the problem that we have no clear idea how, exactly, this vowel reduction would have operated. (It also requires assuming widely parallel vowel reduction, and may even suggest that zero grades also arose in parallel across the IE varieties rather than having been a feature of PIE, but vowel reduction + loss is commonplace enough to be just about trivial — see e.g. the fate of 2nd-syllable vowels in Germanic.)

[Richard W]

Where does Tocharian fit in the centum–satem division ?

The notion that satem is a sound shift that spread over a group of adjacent dialects or languages seems to work quite well. The short answer to the question seems to be 'before it' - Tocharian seems to have split off from the other non-Anatolian dialects.

[KathTheDragon]

1) Schwa epenthesis, followed by laryngeal coloring applying more generally in pre-Greek than in other early IE dialects;
...
1) seems problematic due to requiring a quite late date for laryngeal coloring, which is regardless rather non-trivial and applies across the entire IE family.

Well, the triple reflex does directly imply that the three laryngeals remained distinct beyond the breakup of PIE, at least within pre-Proto-Greek, so it is at least thinkable that colouring persisted as well. Also, it does seem that colouring was a synchronic rule, so if the pre-Proto-Greek speakers identified *ə with *e, then colouring (directly to the synchronic phones) is only natural.

Did it happen only in Greek?

No, Latin displays a "triple reflex" in front of syllabic nasals: *(h₁)n̥ > in, *h₂n̥ > an, *h₃n̥ > un, if memory serves.