The Great Proto-Indo-European Thread's Sequel

[KathTheDragon]

There may be other reasons, though.

*cough*

[Tropylium]

Speaking of the devil, here is a new "thanks, I hate it" paper according to which *h₃…Kʷ results in Hittite š, Lydian s, Luvian-Lycian-Milyan t/d (an intermediate Proto-Anatolian *dž is proposed).
The Anatolian Dissimilation Rule Revisited

So on one hand that sure sounds like labialization is involved somehow, but on the other this also points towards a front fricative or affricate.

[WeepingElf]

There may be other reasons, though.

*cough*

Fair - in fact I was alluding to the *h₂^w thing, of which I am not yet convinced, but I at least concede that such a thing may have been there and distinct from *h₃ which would then of course have to have been something else.

Speaking of the devil, here is a new "thanks, I hate it" paper according to which *h₃…Kʷ results in Hittite š, Lydian s, Luvian-Lycian-Milyan t/d (an intermediate Proto-Anatolian *dž is proposed).
The Anatolian Dissimilation Rule Revisited

So on one hand that sure sounds like labialization is involved somehow, but on the other this also points towards a front fricative or affricate.

That's interesting for sure. However, the idea that *h₃ was something like *dž^w does not seem particularly likely to me. Yet, I do have considered *f at least once, though I never thought that really fitted.

[Howl]

Speaking of the devil, here is a new "thanks, I hate it" paper according to which *h₃…Kʷ results in Hittite š, Lydian s, Luvian-Lycian-Milyan t/d (an intermediate Proto-Anatolian *dž is proposed).
The Anatolian Dissimilation Rule Revisited

So on one hand that sure sounds like labialization is involved somehow, but on the other this also points towards a front fricative or affricate.

This is the third proposal I have seen for the PIE *h₃- Hittite s- Luvian-Lycian-Milyan t- or d- correspondence. The ones I know of are:
1. An s-mobile (but it's curious that Anatolian would have an s-mobile before PIE /#h₃/ when the rest of PIE does not have one)
2. A new phoneme in PIE (Gamkrelidze: /ŝ/, Kassian/Yabukovic: /θ/)
3. This Anatolian Dissimilation Rule.

[mae]

-

[WeepingElf]

Speaking of the devil, here is a new "thanks, I hate it" paper according to which *h₃…Kʷ results in Hittite š, Lydian s, Luvian-Lycian-Milyan t/d (an intermediate Proto-Anatolian *dž is proposed).
The Anatolian Dissimilation Rule Revisited

So on one hand that sure sounds like labialization is involved somehow, but on the other this also points towards a front fricative or affricate.

This is the third proposal I have seen for the PIE *h₃- Hittite s- Luvian-Lycian-Milyan t- or d- correspondence. The ones I know of are:
1. An s-mobile (but it's curious that Anatolian would have an s-mobile before PIE /#h₃/ when the rest of PIE does not have one)
2. A new phoneme in PIE (Gamkrelidze: /ŝ/, Kassian/Yabukovic: /θ/)
3. This Anatolian Dissimilation Rule.

There is another possibility:
4. The etymologies are bogus. I have seen etymologies rejected because of lesser problems.

Maybe I missed something in Cohen & Hyllested 2018 here, but does the argument necessitate an originally coronal value for *h3? Their statements about "faucal" sounds seem to imply they still think it's a dorsal fricative, and the fronting is part of the dissimilatory process.

This would have to be a dissimilatory process then which changes more than one feature - it would not only delete [+labialized], but also change the place of articulation. That's not how dissimilations work. Usually, at least, just one feature is changed.

[Salmoneus]

Off the top of my head: what about /ɥ/? In Anatolian, it could strengthen to /ʝʷ/ outside of a nucleus - seems trivial - and then dissimilate to plain /ʝ/ before a following labiovelar. For that sound to in turn become /ɟ/ is again trivial, and /ɟ/ could easily provide a dental stop or a dental or postalveolar or palatal fricative. Such a chain of events would be very plausible and could happen rapidly and would explain all the reflexes. An approximate would also explain a lot of the laryngeal behaviour, like dropping and lengthening preceding vowels, colouring vowels, and being able to stand as a nucleus.

In fact, wouldn't it be very helpful if the laryngeals were, say, voiceless approximants? I know the diachronics aren't ideal, but something like voiced /j/, /w/ and /ɥ/ doesn't seem impossible, since we don't really know what *e, *a and *o were at the time. [Imagine, for instance, that *a was actually /ɒ/, and at least coloured-*o (which may have been distinct from *o from other sources at first) may have been, let's say, /ø/. /ø/ could then back easily enough, particularly if there's already another 'real' /o/ for it to merge into, and /ɒ/ is already low, and could easily either merge with /o/ or deround and front to /a/.]

To the above advantages of approximants, voiceless approximants would also add the virtue of weakness, helping to explain why they vanished completely almost everywhere. Meanwhile, such sounds could easily be debuccalised to /h/ to explain their descendents in Anatolian.

[of course, it could be that the three laryngeals don't all represent the same class of thing]

[WeepingElf]

Someone must have hacked Salmoneus's account! No, just kidding, but this kind of swashbuckling speculation seems a bit out of character for this well-known skeptic. I won't say that *h₃ can't be /ɥ/, but this does not seem very likely to me. /ɥ/ is a semivowel (but a rare one, even if it occurs in a major European language; and one would only expect it from a language with front rounded vowels, which PIE does not seem to have been), and why should it go such a different way than the other two? Or did I miss something?

[Salmoneus]

Someone must have hacked Salmoneus's account! No, just kidding, but this kind of swashbuckling speculation seems a bit out of character for this well-known skeptic.

The difference is, this is a dead-end question. Something like "is PIE's closest relative Uralic?" is a foundational question - it asks a question about an area we know nothing about, and the answer to that question fundamentally shapes how we go on to answer other important historical questions, like what other languages it might be related to, and where the languages originated. So I think taking a wild leap and just guessing an answer is irresponsible: it's likely to leave us trying to move on to other questions with a flawed understanding that damages our ability to answer those questions. Better to keep an open mind than close it unnecessarily. What's more, these sort of questions can be answered with rigorous methodology and an open mind.

Whereas 'what were the exact values of the PIE laryngeals?' is a question of no significance. Nothing hinges on the answer. We already know the important facts about laryngeals and their outcomes - we're just curious, for curiosity's sake, regarding the trivial matter of their exact values. So there are no stakes to being wrong in our guess. What's more, we're just guessing, and there's no better way to do it: unless new data is discovered in a Ukrainian tomb somehow, we're never going to be able to resolve this question no matter how rigorous and skeptical our methodology. So (informed, rational) guessing is as good a way of thinking about it as any.

I won't say that *h₃ can't be /ɥ/, but this does not seem very likely to me. /ɥ/ is a semivowel (but a rare one, even if it occurs in a major European language; and one would only expect it from a language with front rounded vowels, which PIE does not seem to have been),

There's no reason you'd only expect it from a language with front rounded vowels.

and why should it go such a different way than the other two? Or did I miss something?

Well, you said yourself, it's an unusual sound, and it's a sound that very commonly strengthens or merges with stronger consonants. In Spanish, for instance, it's the same phoneme as the palatal fricative/affricate.

But yes, that's partly why I made the second suggestion: a voiceless approximant.

As to it looking odd in the phonology: not really! If we imagine, for instance, that there's originally only one velar set, as you yourself suggest, then we can imagine an ur-laryngeal H, with the value /ɰ̊/ (perhaps from a voiceless velar fricative, or from /h/; for maximum neatness, pair it with /s/ and /ɸ/ (=h1)). This sound is very unusual and is quickly (or indeed immediately) reinforced to /ʍ/, as is commonplace. Your single velar set then undergoes labialisation in some context - which obviosuly doesn't affect /ʍ/ - and palatalisation in others. Palatalised /ʍ/ is just /ɥ̊/, so /ɥ̊/ (h3) and /ʍ/ (h2) fit perfectly in your triple-velar framework (exept that effectively the plain and labialised versions have merged, but that's, as I say, to be expected).

Obviously, this approach requires a degree of speculation. But otherwise, you're left looking for something else that can act as a nucleus, colours adjacent vowels, is prone to vocalising, can debuccalise to /h/ (or the like), AND can, when delabialised, result in a range of coronal stops and fricatives. The debuccalisation suggests voicelessness, and the range of other outcomes from stops through to vowels suggests something that could easily weaken/strengthen without merging with other phonemes en route, and that suggests something vaguely palatal, both because that's the area where MOA is most often conflated or changed and because that's the only vaguely 'unoccupied' area for this phoneme not to end up hitting anything else. The coronal outcomes can also be explained by a palatal(ish) consonant - in a way that makes no sense for, say, a pharyngeal or a rounded velar fricative. And if it starts out non-fricative at the time when the phonotactics are becoming fixed, that explains why it doesn't pattern with fricative /s/. [assuming *s was a fricative? I imagine someone's already suggested than one of the laryngeals could have been /s/, and that *s was actually /ts/?]. I don't see anything else making more immediate sense than /ɥ̊/, or something similar. Although of course other things are also possible.

I absolutely don't think this should be taken as gospel. But it seems not a terrible wild stab in the dark.

[Howl]

There is another possibility:
4. The etymologies are bogus. I have seen etymologies rejected because of lesser problems..

Sure, but as far as I know there no regular Hittite s- ~ Luwian t- correspondence. So that would mean that you'd need to throw away these etymologies not only for Hittite but also for Luwian.

As to the realization, I think it's likely that the laryngeals were something like approximants in late PIE. Any other sound would not disappear so easily and so completely in all the non-Anatolian daughter languages. And approximants could also explain the strange syllabification of Beekes's law (RHC → RH̥C and not R̥HC). Then h₂ could have been /ɰ/. And for h₃ I have another alternative: /ɹʷ/ like in English red /ɹ̠ʷed/.

[WeepingElf]

Whereas 'what were the exact values of the PIE laryngeals?' is a question of no significance. Nothing hinges on the answer.

Nothing hinges on the answer as long as we only consider Indo-European as a standalone unit. As long as we do so, it suffices to say that *h1 does not affect vowel qualities, *h2 adds the feature [+back], and *h3 the features [+back] and [+round] (or whatever distinguished *o from *a at that stage) to a neighbouring vowel; and that *h1 probably was a "weaker" sound than the other two since it is lost even where Hittite preserves the other two. Everything's fine with that, and hence, most handbooks do not try to sort out the phonetic values of the laryngeals.

But then, there is of course curiosity (and some people like me are working on conlangs that are meant to be related to IE in ways where such matters matter), and when it comes to Indo-Uralic or some other kind of macro-comparison involving Indo-European, it becomes a matter of interest what the PIE laryngeals are like and what they may be cognate to. But that is beyond the scope of this thread; we have the Great Macrofamilies Thread for such topics.

[Salmoneus]

when it comes to Indo-Uralic or some other kind of macro-comparison involving Indo-European, it becomes a matter of interest what the PIE laryngeals are like and what they may be cognate to.

But this is, as we've said before, working in the wrong order. You cannot use the similarity of PIE and Uralic phonologies as evidence for Indo-Uralic if you've configured your reconstruction of PIE specifically in order to support your Indo-Uralic theory.

I agree that iff Indo-Uralic is proven, then the values of the laryngeals becomes a serious question, that would need to be taken seriously. Not only would there be some significance to the answer, but there'd also be some conceivable way of judging correctness other than taste. Or, if Indo-Uralic were almost proven and the only thing in the way of that proof were those pesky laryngeals, then too the value of the laryngeals would become a serious question.

But until then, it's purely an aesthetic question - which set of changes and phonotactics most appeals to each of us aesthetically.

[howl: ok, yes, something like a labialised rhotic could perhaps do the trick too]

[WeepingElf]

when it comes to Indo-Uralic or some other kind of macro-comparison involving Indo-European, it becomes a matter of interest what the PIE laryngeals are like and what they may be cognate to.

But this is, as we've said before, working in the wrong order. You cannot use the similarity of PIE and Uralic phonologies as evidence for Indo-Uralic if you've configured your reconstruction of PIE specifically in order to support your Indo-Uralic theory.

Yes, I am aware of the peril of circularity that lurks here. So far, we can only guess what corresponds to what, and must not take these guesses as facts. I am also aware of what I have called the "system comparison fallacy" - the mistaken assumption that when dealing with two phonological inventories, like matches like (which can be illustrated with Greek vs. Armenian: both have triplets of aspirated, neutral and voiced stops, but these triplets don't match trivially).

That said, there seem to be a few morphemes where PIE laryngeals correspond to Uralic *k - but that does not really tell much more about the nature of the laryngeals than what we know from IE studies alone (there is also a seeming PIE *s: PU *t correspondence in the plural markers, but this does not necessarily mean that the laryngeals were velar fricatives, related to *k as *s is to *t), and the correspondences themselves are not known yet not to be bogus, of course.

I agree that iff Indo-Uralic is proven, then the values of the laryngeals becomes a serious question, that would need to be taken seriously. Not only would there be some significance to the answer, but there'd also be some conceivable way of judging correctness other than taste. Or, if Indo-Uralic were almost proven and the only thing in the way of that proof were those pesky laryngeals, then too the value of the laryngeals would become a serious question.

But until then, it's purely an aesthetic question - which set of changes and phonotactics most appeals to each of us aesthetically.

Just that: an aesthetic question - which becomes relevant when one goes forth building a conlang from that (which I, as you surely know, do).

[Tropylium]

Voiceless semivowels for laryngeals sounds dubious in light of how the "vocalized" versions generally result in *a (most langs), *e *a *o (Greek). Only Indo-Iranian has *i, nothing has *u or diphthongs. Non-semivowel approximants could work… I kinda like the *ɹʷ idea on first earful.

*h₂ is, per recent results from Kümmel (though I'm still waiting to see the full paper), retained in SW Iranian as /h/~ /x/. Given also the Anatolian evidence, I think this nails this down as a voiceless back fricative pretty conclusively.

[Elizabeth K.]

*h₂ is, per recent results from Kümmel (though I'm still waiting to see the full paper), retained in SW Iranian as /h/~ /x/. Given also the Anatolian evidence, I think this nails this down as a voiceless back fricative pretty conclusively.

Let us know when the paper comes out and where to find it. That sounds positively fascinating.

[Nortaneous]

Voiceless semivowels for laryngeals sounds dubious in light of how the "vocalized" versions generally result in *a (most langs), *e *a *o (Greek). Only Indo-Iranian has *i, nothing has *u or diphthongs. Non-semivowel approximants could work… I kinda like the *ɹʷ idea on first earful.

Doesn't IIr also use *i as an epenthetic vowel? How far could you get by assuming that the syllabic resonants were (say) realized as [Hə] sequences, then *ə > e in Greek, *ə > i in IIR, *ə > a everywhere else (which could potentially also explain the -a- in Latin quattuor)?

[WeepingElf]

Voiceless semivowels for laryngeals sounds dubious in light of how the "vocalized" versions generally result in *a (most langs), *e *a *o (Greek). Only Indo-Iranian has *i, nothing has *u or diphthongs. Non-semivowel approximants could work… I kinda like the *ɹʷ idea on first earful.

Doesn't IIr also use *i as an epenthetic vowel?

Tropylium says just that (boldfaced by me).

How far could you get by assuming that the syllabic resonants were (say) realized as [Hə] sequences, then *ə > e in Greek, *ə > i in IIR, *ə > a everywhere else (which could potentially also explain the -a- in Latin quattuor)?

Fair. This is indeed often assumed, and plausible; wherein the triple reflex in Greek shows that the three laryngeals still were distinct from each other at that time. I don't know why Latin quattuor has a, though, but this word never had a laryngeal.

[KathTheDragon]

quattuor is most likely due to some sort of Italic/Latin epenthesis to break up illicit clusters (de Vaan states as much, though he's sparse with details).

[Nortaneous]

Right, I've seen things like *kʷtwor reconstructed. The gemination is a little weird, but (oddly enough) there is precedent for schwa conditioning gemination of following plosives - somewhere in Austronesian IIRC.

[Tropylium]

*h₂ is, per recent results from Kümmel (though I'm still waiting to see the full paper), retained in SW Iranian as /h/~ /x/. Given also the Anatolian evidence, I think this nails this down as a voiceless back fricative pretty conclusively.

Let us know when the paper comes out and where to find it. That sounds positively fascinating.

It is out already, found in Alexander Lubotsky's Festschrift — though the publisher homepage unhelpfully currently has a broken link for the TOC. I'm looking forwards to this arriving online at some point in the future (I think Beech Stave Press allows authors to distribute articles online about a year after publication).