The Great Proto-Indo-European Thread

[Salmoneus]

FWIW, Lithuanian has /l/ for both the levir word (though in lithuanian it's changed to the other type of brother-in-law) and for 'tongue'; armenian also has /l/ in 'tongue'.

[KathTheDragon]

armenian also has /l/ in 'tongue'.

I've seen /l/ in 'tongue' words attributed to influence from 'lick' words, since the latter very commonly start with /l/.

[hwhatting]

@Valdeut: Thanks! I'll watch the podcast as soon as I find some time.

armenian also has /l/ in 'tongue'.

I've seen /l/ in 'tongue' words attributed to influence from 'lick' words, since the latter very commonly start with /l/.

This makes sense for "tongue", especially in those languages that don't have /d/ > /l/ in other words like Latin has. But "tongue" is interesting anyway, as there are only two families that have unambiguous reflexes of PIE /d/ - Germanic and, funnily enough, Latin, where dingua is attested as variant in Old Latin. Baltic and Slavic don't have an initial consonant (in Baltic, Initial /l/ is also attested, which could go back to the influence of "lick"); let's note here that Baltic shows the same lack of initial consonant in the "tear" word, which is not attested in Slavic. Celtic (OIr.) has initial /t/ (normally explained by influence of tongaid "swears"; OIr. has /d/ in "tear"). Tocharian has metathesized A käntu / B kantwo, where /t/ can go back to /*d/, but also to any other dental stop. The Indo-Iranian reflexes are the strangest; Vedic jihva:- and Av. hizu:- only share the last part *-g'^huH-, with a first part that points to *g(')i/H- (Vedic) or *si/H- (Avestan). Either the Indo-Iranian reflexes are actually a different word (maybe a compound with a different first element?) or there has been some serious reshaping going on. So we don't know whether the reshapings replaced a PIIr. Reflex of *dn.-, *n.-, or something else. Unfortunately, there doesn't seem to be a cognate in Anatolian; Hittite la:la-, Luwian la:-la/i- are probably onomatopoetic.
So we have an interesting parallelism in "tear" and "tongue": Latin /l/ (besides /d/ in "tongue"), no initial consonant in Baltic (for "tear" also in IIr., Tocharian, perhaps Anatolian, for "tongue" also in Slavic), /d/ in some (for "tear" several) other IE languages. For "brother-in-law", this is different; while Latin has /l/, all other languages have /d/, besides Iranian, wich has an unexplained /θ/. But I assume Oettinger has some candidates for reflexes without initial consonant?

[WeepingElf]

This discussion brought up an idea in me which I don't know yet whether it has any merit. Perhaps PIE had a lateral fricative which occurs in these words?

[hwhatting]

This discussion brought up an idea in me which I don't know yet whether it has any merit. Perhaps PIE had a lateral fricative which occurs in these words?

I wouldn't exclude that possibility. OTOH, with the low number of examples, it would be a very marginal phoneme.

[WeepingElf]

This discussion brought up an idea in me which I don't know yet whether it has any merit. Perhaps PIE had a lateral fricative which occurs in these words?

I wouldn't exclude that possibility. OTOH, with the low number of examples, it would be a very marginal phoneme.

As I said, I doubt the idea really has any merit. I am, for instance, deeply sceptical about the extra phonemes Gamkrelidze and Ivanov proposed - they are simply too marginal, and are introduced to explain by phonology what probably calls for a morphological explanation. What speaks against the extra phonemes is the fact that the alternations (such as *kost- ~ *Host- 'bone', which they reconstruct as *qost-) are irregular. If some languages always had *k and others always had *H, one could posit a phoneme, perhaps *q, to account for this, but both variants occur in the same language (e.g. Latin os 'bone' vs. costa 'rib'). The lateral fricative seems to be problematic for similar reasons.

[Tropylium]

Thanks for porting that. Intriguing, also the quoted proposal by Oettinger that *sH2V- > *dV.

I'm really curious what the justification is for that, but I didn't spot it. It seems to be [sχ] > [d] or [ɗ].

Edit: There also seem to be a number of words in various languages with the initial sequence *sh₂V- which do not turn up with *d, where the *s is part of the root.

Ditto. The furthest I can imagine is tχ < stχ < sχ.

You might be able to continue with tχ (> th) > tʔ > tʼ, but that makes for some very weird relative chronology. (Laryngeal loss and simplification before a GT-stype phonation shift?)

But I'm more concerned with the counterexamples of *sh₂

Hmm. How many of those have a verifiable laryngeal, and how many have a *h₂ that is posited solely because of a-vocalism? 'Salt', for example, looks fairly well-reconstructible as simply *sal-. (As a trade item, it's also quite likely that this is not a part of the oldest IE wordstock, but a loan that never went thru the ablaut-genesis period).

What speaks against the extra phonemes is the fact that the alternations (such as *kost- ~ *Host- 'bone', which they reconstruct as *qost-) are irregular. If some languages always had *k and others always had *H, one could posit a phoneme, perhaps *q, to account for this, but both variants occur in the same language (e.g. Latin os 'bone' vs. costa 'rib'). The lateral fricative seems to be problematic for similar reasons.

Yes, that looks like a dialect loaning situation — if early Italic (similarly Slavic etc.) speakers had already turned their *h₂ into /h/ or lost it entirely, and they then came in contact with IE speakers who still maintained /χ/, in loans this would be likely to be turned into /k/.

[KathTheDragon]

Hmm. How many of those have a verifiable laryngeal, and how many have a *h₂ that is posited solely because of a-vocalism?

Anything attested in Anatolian, basically. Naming examples off the top of my head, ishai "bind" < *sh₂i-, ishamai "song" < *sh₂emi- (though I might have quoted that wrongly)

[Zju]

Why does PIE *deḱsino 'right' give Sanskrit दक्षिण ‎dákṣiṇa. I thought satemisation is unconditional in Indo-Iranian.

[KathTheDragon]

It is, but all dorsals yield k before *s in Sanskrit.

[Tropylium]

The development has been *ḱs > *ćš (RUKI + satemization) > *śṣ > *ṣṣ (general lenition/ assimilation) > *tṣ > kṣ (fortition of geminate sibilants).

The only rule bleeding satemization in Indo-Iranian that I know of is *Ḱr > *Kr, found also in the other Satem branches and possibly already in PIE ("Weise's Law").

[jmcd]

What's to say the sibilant didn't just block palatalisation, like it did with Grimm's Law in Germanic?

[KathTheDragon]

Because afaik Avestan does show the expected palatalisation, hence Indo-Iranian should have, so the Sanskrit is a secondary development of that.

[jmcd]

Following on with this: https://en.wikipedia.org/wiki/Proto-Ind ... n_language, that means on the Avestan side it makes sense more or less but on on the Sanskrit side, you've got 5 different changes to arrive more or less back at the starting point. And the result in Avestan is just a single consonant, not a cluster or geminate anyway. And *kʷs yields a similar outcome in Sanskrit but not in Avestan. How about the *ĉs suequence simplifying later? In Proto-Iranian? That would make the Avestan changes the same, just a tad postponed while the Sanskrit changes are reduced to *ḱs > *ćš (RUKI + satemization) > kṣ. There is after all no Indo-Aryan evidence for the *ĉs > *šš change AFAICT. Add to that, the fact that Sanskrit has tṣ so the two would have been confused.

[Vijay]

Following on with this: https://en.wikipedia.org/wiki/Proto-Ind ... n_language, that means on the Avestan side it makes sense more or less but on on the Sanskrit side, you've got 5 different changes to arrive more or less back at the starting point.

How is that ending up back at the starting point, though? Both of the sounds in that cluster are further back as a result than they were at first.

[KathTheDragon]

Following on with this: https://en.wikipedia.org/wiki/Proto-Ind ... n_language, that means on the Avestan side it makes sense more or less but on on the Sanskrit side, you've got 5 different changes to arrive more or less back at the starting point. And the result in Avestan is just a single consonant, not a cluster or geminate anyway. And *kʷs yields a similar outcome in Sanskrit but not in Avestan. How about the *ĉs suequence simplifying later? In Proto-Iranian? That would make the Avestan changes the same, just a tad postponed while the Sanskrit changes are reduced to *ḱs > *ćš (RUKI + satemization) > kṣ. There is after all no Indo-Aryan evidence for the *ĉs > *šš change AFAICT. Add to that, the fact that Sanskrit has tṣ so the two would have been confused.

Ok, so! The fact that Avestan has only a single consonant means nothing - shortening of geminates is a fairly natural change, is it not? Why should sequences involving labio-velars be relevant in a discussion of palatovelars? Given that *ĉt > *št (and similarly before other dental stops) it's fairly believable that an identical change would take place before *s; after all, in Germanic, all PIE stops yielded fricatives in precisely this environment - before *t and *s - so *s certainly behaved similarly to *t. The assimilation of *šs to *šš is then trivial.

[Zju]

Thanks for porting that. Intriguing, also the quoted proposal by Oettinger that *sH2V- > *dV.

It suddenly occured to me - aren't we viewing it the wrong way around? What is the evidence that it was *sH2V- > *dV and not *s- + *dV- > *sh₂V- ? If the latter is the case, we could unify it with the other proposal in a single (sound) change:

sɗ sɠʷ > sð sw > sh sw / #_V

With ð being approximant in this case, debuccalising to h, then changing to h₂ if h₂ is not already [h] in your view.

Counter examples of #sdV #sgʷV could be later formations.

[KathTheDragon]

Counter examples of #sdV #sgʷV could be later formations.

There are none, at least in roots.

[jmcd]

Following on with this: https://en.wikipedia.org/wiki/Proto-Ind ... n_language, that means on the Avestan side it makes sense more or less but on on the Sanskrit side, you've got 5 different changes to arrive more or less back at the starting point.

How is that ending up back at the starting point, though? Both of the sounds in that cluster are further back as a result than they were at first.

Sure but you're still going from first element being dorsal plosive to first element being dorsal plosive in five different steps.

Following on with this: https://en.wikipedia.org/wiki/Proto-Ind ... n_language, that means on the Avestan side it makes sense more or less but on on the Sanskrit side, you've got 5 different changes to arrive more or less back at the starting point. And the result in Avestan is just a single consonant, not a cluster or geminate anyway. And *kʷs yields a similar outcome in Sanskrit but not in Avestan. How about the *ĉs suequence simplifying later? In Proto-Iranian? That would make the Avestan changes the same, just a tad postponed while the Sanskrit changes are reduced to *ḱs > *ćš (RUKI + satemization) > kṣ. There is after all no Indo-Aryan evidence for the *ĉs > *šš change AFAICT. Add to that, the fact that Sanskrit has tṣ so the two would have been confused.

Ok, so! The fact that Avestan has only a single consonant means nothing - shortening of geminates is a fairly natural change, is it not? Why should sequences involving labio-velars be relevant in a discussion of palatovelars? Given that *ĉt > *št (and similarly before other dental stops) it's fairly believable that an identical change would take place before *s; after all, in Germanic, all PIE stops yielded fricatives in precisely this environment - before *t and *s - so *s certainly behaved similarly to *t. The assimilation of *šs to *šš is then trivial.

Reduction of geminates is indeed very natural and simple and could have happened after PII. We have both given parallels using other sound changes involving PII. I have given *kʷs changes and you have given *ĉt changes. The assimilation of *šs to *šš is indeed trivial. But how else is Sanskrit tṣ explained?

[Tropylium]

But how else is Sanskrit tṣ explained?

That's simply from t + ṣ, right (and would be pronounced [t̪ʂ])?

There's independent evidence for *ṣṣ > kṣ from that also //ṣ + s// yields kṣ, e.g. dviṣ- 'to hate' : aorist dvikṣat-.

[Vijay]

When I saw <tṣ>, I thought he meant kṣ, mostly because in (Sanskrit loanwords in) Malayalam, it's pronounced [ʈʃ] (or maybe [ʈʂʃ͡]).

[2+3 clusivity]

Damn, I just returned a book with a chapter on the sources of SKT /kʂ/. Essentially, the sources of internal and initial /kʂ/ are many including PIE thorn clusters, PIE and PIr. verlar + /S/, /S/ + some PIE and PIr. Velar, and a few other odd-balls including, IIRC, a /p/ + /S/ combo. A major exception to these shifts is /*-sḱ-/ -> /-ccʰ-/ as in the /-ccʰati/ verbs and also internally in a few nouns.

Interestingly, most of the material feeding into Sanskrit /kʂ/ were both voiced and unvoiced. Colin Masica, in the Indo-Aryan languages, while discussing sound changes leading to the Middle Indo-Aryan languages, however, notes that a few languages show evidence of retained voicing -- so possible a cluster like /??*gʐ/ existed in some such languages at one point. This divergence is notable given that Sanskrit -- especially Vedic Sanskrit -- is a cousin of whatever language(s) developed into most of the modern IA languages -- except maybe some of the extreme north western ones, even that seems a stretch.

When I saw <tṣ>, I thought he meant kṣ, mostly because in (Sanskrit loanwords in) Malayalam, it's pronounced [ʈʃ] (or maybe [ʈʂʃ͡]).

Beyond loanwords, it's a fairly unstable cluster natively which makes the massive collapse towards it pretty strange. In NIA reflexes it is typically /kʰ/, /ʈʃ/, and ?? /ʈʃʰ/. A source I have states that Ga:ndha:ri:, an early North-western Prakrit, appears to have had /ʈʂ/ for <kʂ> -- which agrees with Khotanese which borrowed the same writing system and used <kʂ> for /ʈʂʰ/ (As a side note, the Niya Prakrit was basically Ga:ndha:ri: and was used by scribes in at least a few Tarim Basin city states by various non-native speakers including probably Tocharian L1s given that a lot of the documents show vacillation in voicing across all four voicing qualities found in MIA languages).

[Vijay]

Wait a minute, so it actually is /ʈʃ/ in a bunch of NIA languages, too?? I didn't realize that. I thought it was /ʈʂ/ in those languages and only Malayalam had /ʈʃ/ or something. Good to know!

[Pogostick Man]

The traditional reconstruction of PIE phonation in stops is unvoiced, voiced, and breathy-voiced. Why couldn't the breathy voice have been slack voice instead? Are there any examples of languages with an unvoiced/voiced/slack-voiced distinction?

[Atrulfal]

What is the origin of the tocharian's secondary cases?