The Great Proto-Indo-European Thread's Sequel

[KathTheDragon]

Yes, Hittite ḫasterza proves it.

[Howl]

Three interesting indirect lines of evidence for laryngeals from Anatolian, from Anthony Yates, The Phonology of Anatolian Reduplication:
– in Proto-Anatolian, *-ns- > *-ss-, but the nasal remains in *-nh₁s- (> Luwic ns, Hittite nz)
– in Hittite, *ti *tyV > zi zV (/tsi tsV/), but *th₁i *th₁yV > ti tiyV (/ti tja/);
– in Hittite, *eya > a, but *eh₁a > eya

Look at the evidence Yates provide for the second change:

a. PIE/PA *–tyo– > Hitt. šarāzziya–, Lyc. hrzze/i– ‘upper’
b. PIE/PA *h₂et-ye/o– > Hitt. hazziya– ‘strike (an instr.)’, HLuw. hazi– ‘incise; write’
c. PIE/PA *dh₁-yé/ó– > Hitt. tiye/a– ‘bind’ (not x ziye/a–)
d. PIE/PA *d(ʰ)h₁y-énti > Hitt. tianzi ‘place’ (not x zianzi)

According to Kloekhorst it is uncertain whether PIE *dʰy → Hitt. ti or Hitt. zi. So that leaves just one example of PIE dh₁y → Hitt. ti and zero examples for PIE th₁y → Hitt. ti. But I have not checked the references to see if they have any stronger evidence than this.

On this topic, would anyone happen to know of an overview on what is the evidence base on which *CVCH roots are reconstructed? I tend to see a lot of variance in if some root is assumed to have a "tailing" laryngeal or not.

Lines of evidence I'm aware of:
...
Lines of evidence I'd expect to see but am not sure if they actually occur:
– thematic vowel coloring?

In Tocharian, roots that end in a laryngeal use different verb conjugations, which are never thematic.

I'm unconvinced that schwa epenthesis should date all the way back to early PIE, because, as I said, Anatolian offers absolutely nothing to support it, and everything is just as explainable without it.

So there is nothing that proves pre-Anatolian PIE did have schwas there. But is there anything that proves that pre-Anatolian PIE did NOT have schwas in those positions?

[mèþru]

@Howl
By that standard you can claim anything existed or didn't so long as the evidence is erased somehow in the other branches

[KathTheDragon]

I would argue that the lack of any vowel anywhere in the entire Anatolian branch that could only reflect such a schwa is as close to proof as you can reasonably get when trying to prove a vowel did not exist. But there are meta-reasons to assume no schwa epenthesis - it provides a simpler explanation of the facts, as either way, you still need to assume a post-Proto-Anatolian, but still pan-Anatolian, loss of a segment, and assuming schwa epenthesis just adds unnecessarily to the set of sound changes.

[KathTheDragon]

Also, after scouring Hittite, I have managed to find one example of non-assibilation: pattai- "run, flee" < *peth₁-. The 3pl of this verb is pattiyanzi < *pth₁yénti which shows that *h₁ did indeed block the Proto-Anatolian assibiliation *ty > *ts ( and compare halzai- "call out" < *h₂let- and zai- "cross" < *h₁et- for assibilation generalised out of the plural).

[Howl]

I would argue that the lack of any vowel anywhere in the entire Anatolian branch that could only reflect such a schwa is as close to proof as you can reasonably get when trying to prove a vowel did not exist. But there are meta-reasons to assume no schwa epenthesis - it provides a simpler explanation of the facts, as either way, you still need to assume a post-Proto-Anatolian, but still pan-Anatolian, loss of a segment, and assuming schwa epenthesis just adds unnecessarily to the set of sound changes.

Thanks! My own line of thinking is that post-syncope PIE kept phonemic schwas in the vowel slots for a long time. Maybe this lasted until the breakup, but I would have to make a very detailed analysis of the sound-laws to see if that is even possible. So for now I have to assume that PIE lost these vowel-slot schwas at some point before the split-off of Anatolian.

With regards to h1 in Anatolian, Kloekhorst even thinks that h1 was present in the Anatolian languages themselves as a glottal stop which is reflected by certain spelling conventions. While the evidence for the side effects of h1 is a bit thin, and could have been presented better in the paper Tropylium mentioned, I think it is safe to assume that h1 was present in at least proto-Anatolian.

[KathTheDragon]

Yeah, some of Kloekhorst's arguments aren't as strong as he thinks, and he ascribes a lot of things to plene spelling that probably aren't a thing. E.g. ēszi is probably just marking the stress.

[Tropylium]

I would argue that the lack of any vowel anywhere in the entire Anatolian branch that could only reflect such a schwa is as close to proof as you can reasonably get when trying to prove a vowel did not exist. But there are meta-reasons to assume no schwa epenthesis - it provides a simpler explanation of the facts, as either way, you still need to assume a post-Proto-Anatolian, but still pan-Anatolian, loss of a segment, and assuming schwa epenthesis just adds unnecessarily to the set of sound changes.

Thanks! My own line of thinking is that post-syncope PIE kept phonemic schwas in the vowel slots for a long time. Maybe this lasted until the breakup, but I would have to make a very detailed analysis of the sound-laws to see if that is even possible.

I don't mean hypothetical zero-grade schwas though (even if that is IMO a hypothesis worth exploring), I'm talking about the normal development *H̥ > *a, which is well enough attested in Anatolian too (e.g. HLuw. tuwatra-, Lyc. kbatra 'daughter'). So assuming schwa epenthesis doesn't add any complexity at all! Schwa primum never stopped being something that needs to be recognized for PIE. Every single IE language reflects this process, which is strong evidence to reconstruct "vocalic allophones" *ə₁ *ə₂ *ə₃ (or a single epenthetic *ə with allophonic coloring) already in PIE. I think in positions such as these we then have intervening *ə > ∅ instead of *ə > *a.

Cf. also e.g. the lack of *ph₂ > **pʰ in Indo-Iranian 'father', which was therefore surely not "*pHtā" but rather already vocalized *pɨtā (and then *ɨ > ∅ in Avestan, *ɨ > i in Indic). Indo-Europeanists are, I think, far too eager to treat laryngeals as phonetic deus ex machinae and end up reconstructing the at nonsensical or at best highly non-trivial sound change *[x] > *a or > *i happening dozens of times independently. There must have been intermediate stages, and the very similar development of *H̥ across Indo-European should be taken to signify that some of these intermediate stages were in place quite early already.

I think it is safe to assume that h1 was present in at least proto-Anatolian.

There may well have been a phoneme continuing *h₁ in Proto-Anatolian, but I think it incorrect to conclude that it therefore would have been retained unchanged in every position.

The clearest example against this line of thinking is probably Cheung's Etymological Dictionary of the Iranian Verb; he reconstructs even nonsensical things like *Hh- from PIE *Hs-, just because there's some evidence for retention of initial prevocalic and intervocalic laryngeals into Proto-Iranian.

[KathTheDragon]

I'm not trying to deny a schwa for late PIE - I just said you can argue for exactly that. As for the "daughter" word, sure it looks like a schwa reflex, but why then do we not find any other examples? E.g. *pl̥th₂sh₂o- > Hittite palzahha-, not *paltasha-.

[Frislander]

My current pet theory is that Pre/Early PIE had a vowel system like that of modern Circassian with *ə *a *aː, equivalent to the *∅ *e *o of traditional reconstructions. This would also account for Brugmann's Law, and explain the tendency for *o *a to be confused is most of IE if we assume relatively late raising of *aː (this parallels Circassian, in that /a/ <э> is often closer to [ɛ] phonetically, but /aː/ is pretty much always low). I know comparing PIE and NW-Caucasian is a rather fruitless endeavour but I do reckon this might be one place where parallels can be drawn.

[KathTheDragon]

Funnily enough, this is not a new theory, and I've been working with it for quite some time.

[WeepingElf]

There is IMHO nothing wrong with drawing typological parallels between IE and NWC; there may even have been an areal connection as the two proto-languages probably were neighbours; though they are morphologically so different that a "genetic" relationship is pretty much out of the question. (For similar reasons, typological parallels between IE and Uralic are IMHO legitimate; again, the two proto-languages were spoken close enough together to have influenced each other, be they related by a common ancestor or not.) Alas, as long as we don't know what kind of influence there actually has been between the languages, such arguments are weak and best used with great care: that something is found in family X shows only that it could have happened in family Y, too - we still don't know whether it did.

[Tropylium]

At some point we will have to move from just proposing hypotheses to comparing them though, and at that point typological parallels start being of great value: e.g. explaining Brugmann's Law is much easier from a starting point of *a *ā than from *e *o with *o equipped with an ad hoc feature "[+tense]".

I am not aware of much other evidence for *o being or coming from a long vowel *ā, though. Another possibility for a pre-PIE vertical vowel system would be a purely height-based *ɨ *ə *a or *ə *ɐ *a (to which you'd likely want to add at least *i *u).

[KathTheDragon]

Kümmel adduces not only Brugmann's law as evidence for *o being "stronger" than *e, but also *o becoming Proto-Tocharian *e (to merge with PIE *ē) in contrast to *e merging with *i and *u in PToch *ä, as well as certain effects in Anatolian. Specifically, in Luwian, we have *éC(C)V > aCCV, while *óC(C)V > āCV, indicating that *ó counted as a long vowel for the purposes of Čop's law, while in Hittite *ó caused lenition of at least *h₂, and maybe generally if Melchert is wrong that it didn't. Kümmel goes on to explain that this assymmetry, while not proving that it developed from an original length contrast, receives the most natural explanation if it did.

[Howl]

While there may be merit to the idea that PIE *o was a long vowel, there is one problem. If PIE *o in PIE *genos was actually **genōs, then what was the PIE *ō in dʰéǵʰōm?

More generally, if you are going to reconstruct a PIE/pre-PIE vowel system, the first question you should ask is: which vowels and vowel qualities did PIE have? And already at this point you find a lot of disagreement. Leiden says there was only *e and *o. But other schools have more.
- Narten roots and -s aorists have PIE *ē.
- Then there was also the PIE *ō in words like dʰéǵʰōm , but this might be a late PIE innovation.
- Independent PIE *a and *ā is reconstructed for words where reconstructing a laryngeal runs into problems with sound-laws. For example PIE *sāl 'salt' is often reconstructed because PIE **seh₂ḷ would have a syllabic *ḷ of which we find no reflex.
- And then there is the non-ablauting *i and *u you find in case endings and particles, etc. It makes no sense to view those as semivowels. And in words like *sūs 'pig' you find a long/short alternation that is difficult to explain with laryngeals. So there is probably also a case for PIE *ū and PIE *ī.

If you accept all these vowels, you are back to the Brugmannian vowel system but with laryngeals. The main vowels are still *e and *o, and the other vowels are more rare but do exist independently in such a system. But then the next step is questioning the vowel coloring of the laryngeals. If PIE had an independent *a, do we need to reconstruct *h2 everywhere we see an *a and we can explain it with *h2? And how far do you want to go with that?

[WeepingElf]

While there may be merit to the idea that PIE *o was a long vowel, there is one problem. If PIE *o in PIE *genos was actually **genōs, then what was the PIE *ō in dʰéǵʰōm?

You are simply confusing long vowels of different time periods. The long vowels of forms like *dʰéǵʰōm arose after the old opposition of length had shifted into an opposition of vowel quality between *e and *o. (Also, there are languages with three degrees of vowel length, but we don't even need to invoke that here.) At least, that is how I understand this.

[Howl]

While there may be merit to the idea that PIE *o was a long vowel, there is one problem. If PIE *o in PIE *genos was actually **genōs, then what was the PIE *ō in dʰéǵʰōm?

You are simply confusing long vowels of different time periods. The long vowels of forms like *dʰéǵʰōm arose after the old opposition of length had shifted into an opposition of vowel quality between *e and *o.

No, I'm not.

- Then there was also the PIE *ō in words like dʰéǵʰōm , but this might be a late PIE innovation.

Because if that is true, then at least Brugmann's law can't be an archaism. And the evidence from Tocharian and Hittite also becomes more dubious.

[WeepingElf]

To be honest, I am not really convinced by this "PIE *o once having been long" thing, either, and perhaps I have missed something. Thus, I am not in a position to explain this.

[Frislander]

Is there a particular reason why **sāl must be reconstructed as opposed to **selh₂? I assume the root itself mainly has this vowel because Anatolian either shows no laryngeal or lacks this word altogether?

[Howl]

Is there a particular reason why **sāl must be reconstructed as opposed to **selh₂? I assume the root itself mainly has this vowel because Anatolian either shows no laryngeal or lacks this word altogether?

The Sonority Sequencing Principle. A vowel is more sonorous than a liquid. And liquid is more sonorous than a laryngeal. Ideally you want syllables to go from the least sonorous phoneme to the most sonorous phoneme and then to the least sonorous phoneme. For example, English has thrills but not rthisl.

The relative sonority of the phonemes in seh2l is 1(s)-3(e)-1(h2)-2(l). The l forms a sonority peak, which means that it must be syllabic in PIE. But syllabic l's have special reflexes in some language families. For example, in Germanic a syllabic *l becomes -ul, and the Proto-Germanic word is not **saul.

Now obviously not everyone (e.g. Leiden) agrees with this theory. But it is one of the reasons why PIE *a's are still reconstructed.