The Great Proto-Indo-European Thread's Sequel

[KathTheDragon]

Fair. There is no a priori reason to assume that whenever Hittite and Late PIE disagree, Late PIE must have innovated. There probably also were innovations in Anatolian. However, many of the disagreements involve Late PIE having some category - the feminine gender, some of the case forms, the three aspectually different past tenses - which Anatolian lacks and shows no traces of. Of course, there are plenty of non-Anatolian IE languages which have lost something, too, but usually the lost category leaves traces behind, such as the Latin words duo and ambo showing an idiosyncratic declension which can be explained as a residue of the PIE dual.

It would help if we had an outgroup - a language or language family more diastantly related to Anatolian and Late PIE - to compare things against, but we don't, though Uralic is IMHO (and also in the opinion of some scholars) a likely candidate. For instance, the fact that Uralic has a dual, with a marker that could be cognate to the Late PIE one, makes me think that Early PIE also had a dual, which Anatolian lost.

This is all true, but the situation isn't usually that simple. For example, even without looking at Hittite, you can deduce that the feminine was a creation of Late PIE only on the basis of evidence internal to Late PIE. The only reason this possibility was explored after Hittite was found is that nobody had any reason to explore it. Hittite lacking a feminine gender (though not its marker, apparently - Hittite just absorbed them into its a-stems) should be considered supporting evidence, not primary evidence.

[WeepingElf]

Or the laryngeals, which the young Ferdinand de Saussure reconstructed some 30 years before Hrozný deciphered Hittite, so he simply couldn't have used Anatolian data, and didn't.

The Hittite ergative case suffix *-anza, on the other hand, probably was an Anatolian innovation, there is no trace of it in other branches of IE (though the suffix is familiar from active participles, see below). But I seem to remember reading somewhere (but haven't found the source again) that several older IE languages tended to avoid neuter transitive subjects by passivizing the clause, so Early PIE may have had for what I have found the term "privative ergativity", a system where inanimate nouns had an absolutive case, but no ergative, as inanimate transitive subjects were disallowed. When the connection between animacy and grammatical gender got blurred, Hittite solved the ensuing problem - agent-worthy neuter nouns not having a form for the transitive subject slot as needed - by innovating an ergative case (by commandeering an agentive suffix which forms active participles in Late PIE), while Late PIE expanded the absolutive case to transitive subjects, resulting in the familiar nominative/accusative syncretism in neuter nouns.

To get back to the question whether Late PIE or Hittite innovated with regard to the genitive case suffixes, I feel as if an innovation in Late PIE is likelier, though this is little more than a "gut feeling". The nominative and accusative cases show cognate plural forms in both branches, so these are Early PIE. I consider it likelier that a system in which some cases distinguished number and others didn't was "fixed" by innovating plural cases, than that an existing number distinction was preserved in some cases but turned into a specificity distinction in others. As for why some cases distinguished number in Early PIE and others not, this may be due to the loss of a phonetically weak allophone of the plural marker. Uralic can be considered a typological parallel (if not an actual cognate) where the plural marker is *-t in the nominative and accusative cases and *-j- in others (well preserved in Finnish). In such a system, the oblique plural marker may easily be lost by sound change, resulting in oblique cases not distinguishing number, a situation that Late PIE and Anatolian fixed with their own devices - while the stronger allophone in the core cases never got lost.

[KathTheDragon]

I'm still not convinced we can reconstruct a large number of number-indifferent cases on the basis of, what? A few such markers that are only living cases in Anatolian and two more markers that show a distinction not otherwise found in old IE languages? Bear in mind that every case marker in Hittite with cognate case markers in core IE (excluding the genitive) match up perfectly in terms of which number they indicate, so at least the nominative, accusative, neuter direct, locative singular and dative plural (*-os, per Jasanoff) can be reconstructed in precisely those functions for early PIE. In that context, I find it really bizarre that the genitive would just happen to hold out against being assigned to a number contrast.

[WeepingElf]

The dative plural is nowhere *-os in Late PIE, is it? Sure, this morpheme is part of the *-bhyos ending, but it is not the dative plural ending itself. Jasanoff says that it is, but I just can't follow his argumentation, there are a lot of leaps of logic in it. So what makes you so sure that *-os indeed meant dative plural? What is the *-bhy- part, then? Fact is that only the nominative and the accusative are expressed the same way in the plural in both Hittite and Late PIE.

[KathTheDragon]

I'd suggest reading the article again more carefully and slowly. I found it pretty straightforward to understand, so I don't think the leaps in logic are that complicated.

He starts by grounding the adverbial morpheme *-bʰi, and looks at another clear occurence of it, in the personal pronouns where it serves as dative singular (*mégʰi, *tébʰi, with dissimilation in the 1st person). Here, we find that the various daughters added dative singular morphology to properly characterise these forms as datives. Hence, any case ending with *bʰ can in principle have been recharacterised with the morphology of the case form it replaces.
To support this, Jasanoff looks at the instrumental plural *-bʰis, and demonstrates that the original instrumental plural was *-is by adducing several adverbs in *-is from Indo-Iranian and Greek, together with the o-stem pronoun instrumental plural *-ōys, which had so far defied analysis. Jasanoff takes this to be a reduction of *-oyis by invoking Stang's law, containing the o-stem plural stem in *-oy- like the other plural cases, and the ending *-is.
Having shown this, we can compare the variants *-bʰos and *-bʰyos, and deduce that the original ending must have been *-os. Now, note that none of this analysis actually requires Anatolian! It's supported purely by internal evidence, but the dative plural we find perfectly matches the Anatolian ending, which is a strong argument for correctness.

[WeepingElf]

OK, I understand now. But this does not really answer the question of how plural case endings emerged that can't be derived from combinations of a columnar plural marker and the relevant singular case suffix. How do you derive *-os from *-X-ei and *-is from *-X-h₁ (wherein *-X- is the oblique plural marker, the (typological, maybe actually cognate) equivalent of Uralic *-j-)? (And BTW: is *-ôis the origin of Latin thematic dat./abl. pl. *-îs?)

[Howl]

I'm suspicious about this reconstructed instrumental plural -is. Perhaps -h1s is a better reconstruction here. That would also follow the pattern case-ending + plural *s of the PIE accusative (-ms > -ns). This could also be applied to the allative plural -os (also reused for the dative plural) from the allative singular -o (attested in Anatolian).

[WeepingElf]

In agglutinating languages, the plural suffix usually precedes the case suffix, as number is a more intrinsic property of the noun than case. However, the accusative plural *-ns, if from earlier *-ms, seems to show the "wrong" order, which, as I already wrote, may be due to the suffix *-m originally have been an "animacy-neutralizing" derivative suffix in an active-stative pre-stage of PIE (maybe Proto-Indo-Uralic, if that is a thing, as the same suffix also occurs in Uralic).

[mèþru]

If you say Proto-Indo-Uralic as something providing evidence I'll just ignore that post. Indo-Uralic is a separate discussion for me.

[KathTheDragon]

Yes - in the absence of firm proof of a relationship, Uralic is an invalid comparison for commenting on the quality of reconstructions.

Howl: read through the argumentation again. Jasanoff's reasoning can only yield *-is, since *-oyh₁s wouldn't be expected to lose the laryngeal, or lengthen the *o, nor would *-bʰih₁s be expected to have a short vowel reflex anywhere. And when you consider that this ending ought to have a surface vowel (all other oblique endings did - the locative singular is an exception due to being originally endingless), the discrepancy gets even worse.

WE: I'm not denying that these endings aren't further analysable, but that doesn't automatically entail that the reconstructed semantics are wrong. Sure, it may well turn out that all these oblique case endings are ultimately secondary, and go back to adverbial affices of some sort, but the key thing is that this is not synchronically the case for PIE, and there is no basis for assuming that these are anything other than case endings.

[zelda]

Let's start this thread again, after 89 pages and 1 post on the old board.

To kick off, I read this article by Kloekhorst the other day, on the origin of nominal accent-ablaut paradigms and ultimately the case-endings. He's got some interesting ideas, but I'm not sure how plausible some of his assumptions are. What do you guys think?

I know I'm a day late and a dollar short, but I've been working on a PIE lateral derivative using a lot of off-the-wall theories, and I've used another Kloekhorst paper to help define the ablaut pattern located here: http://www.kloekhorst.nl/KloekhorstIENo ... tterns.pdf, which I found to be more convincing than what was presented. It is less elaborate than his 2015 paper, which I think is a strength in his regard.

[WeepingElf]

Point taken on the invalidity of argumentation with recourse to Uralic. Indo-Uralic is not proven, so we cannot use it in arguments about the origin of anything in IE - such attempts remain speculative. Also, the parallel I drew to Uralic was chiefly meant as a typological one: an early stage of PIE could have had a plural marker that was *-s in word-final position and something weaker that was later lost when followed by an oblique case suffix. Yet, this is just speculation, and so it isn't worth much, especially as long as we don't know just what this "something weaker" was.

[Howl]

Howl: read through the argumentation again. Jasanoff's reasoning can only yield *-is, since *-oyh₁s wouldn't be expected to lose the laryngeal, or lengthen the *o, nor would *-bʰih₁s be expected to have a short vowel reflex anywhere. And when you consider that this ending ought to have a surface vowel (all other oblique endings did - the locative singular is an exception due to being originally endingless), the discrepancy gets even worse.

OK, *-h₁s does not work. But the form you cite is *tōis, not **toiis. First, what proof is there for a sound change *oii > *ōi ? And secondly, after that sound change *-bʰis could never have formed from analogy from **toi-is because it did not exist anymore. All you had was a plural stem *toi that happened to have a long vowel ō and a plural s in the instrumental. So that leaves just the adverbial modal/instrumental -s. And this would mean that instrumental plural *-bʰis was originally indifferent to number as well.

[KathTheDragon]

I'd suggest you go read up on PIE (and read through my post again), because your objection is nonsensical. Firstly, while there are no direct parallels for *-oyi- > *-ōy-, Stang's law is incredibly similar, operating word-finally on precisely this kind of input (*-VRR̥ with both *R's identical), compared to before a word-final *s. Secondly, *-bʰis was not formed on analogy with *-ōys, and I never claimed this. If you read what I wrote more carefully, you'll see that *-bʰis was created in nouns through a secondary blend of an adverbial form *-bʰi with the older instrumental plural *-is, while *-ōys was the pronominal ending.

[Tropylium]

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

Yates uses this to date the loss of *h₁ as entirely post-PAA, although I am not that convinced: we could also assume that consonantal *h₁ had already been lost, but vocalic *h₁ was still continued as *ə — so *nh₁s > *nəs, *th₁y > *təy. The third doesn't seem to imply retention of the contrast beyond PAA.

---

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:
– vocalized laryngeals in some nominal derivatives ("seṭ" roots in Sanskrit, etc.)
– *RH > *Rː in Anatolian; in "some" cases also in Germanic
– some direct evidence in other cases in Anatolian?
– *Ph₂ (and maybe *PH in more general) > *Pʰ in Indo-Iranian
– failure of Brugmann's Law in Indo-Iranian
– "long syllabic resonants" (and *ī *ū) in zero-grade CR̥H- formations

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

[Zaarin]

– *RH > *Rː in Anatolian; in "some" cases also in Germanic

And Celtic.

[KathTheDragon]

we could also assume that consonantal *h₁ had already been lost, but vocalic *h₁ was still continued as *ə — so *nh₁s > *nəs, *th₁y > *təy. The third doesn't seem to imply retention of the contrast beyond PAA.

Can we assume laryngeal vocalisation in Anatolian when not a single language has any vowel that can be confidently said to be such a vocalised laryngeal? It's honestly simpler to just assume that *h₁ was retained.

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

This is a very interesting question. I don't know of any cases where it occurs with the inflectional ending, only when followed by derivational suffices. Similarly, colouring of the *eye-causative can tell us about *h₂₃

[Tropylium]

we could also assume that consonantal *h₁ had already been lost, but vocalic *h₁ was still continued as *ə — so *nh₁s > *nəs, *th₁y > *təy. The third doesn't seem to imply retention of the contrast beyond PAA.

Can we assume laryngeal vocalisation in Anatolian when not a single language has any vowel that can be confidently said to be such a vocalised laryngeal?

I would think syllabic laryngeal "vocalization" (better: schwa epenthesis) dates already to PIE, even if the loss of the associated laryngeal consonant perhaps doesn't.

(This approach can work on some other "late-retained laryngeal" arguments too, e.g. vocalization of *HC- in Greek.)

[KathTheDragon]

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. Whether or not we can date it to late PIE is more debatable, since there are numerous commonalities.

[Frislander]

With the Greek triple reflex thing, is there any particular reason forms like *h2ster must be reconstructed with a laryngeal there, or could it not just be a weak initial vowel that in Greek split into different vowels depending on environment and was lost in other languages entirely?