The Great Proto-Indo-European Thread's Sequel

[KathTheDragon]

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?

[WeepingElf]

That article surely is very interesting, and I did not notice any obvious problems with it (but then, I am only an amateur in this field, and I may be missing something important). It is also pretty compatible with my own thinking. I have been assuming a penultimate accent for PIE1 (the pre-ablaut stage) for years, and the dynamic paradigms can IMHO be reduced to that, if one assumes that the right-shifting case endings were bisyllabic. Note that the inanimate absolutive was endingless. The notion that inanimate nouns had a defective paradigm in PIE1 is also one I have been entertaining for years (and pursued in my PIE1-based conlang Old Albic).

I don't think, though, that the PIE sigmatic nominative ever was an ergative. Rather, I suspect a topic marker here, which is something different. The animate accusative *-m may have started as an "animacy-neutralizing" morpheme of sorts, which may be the reason why the plural marker follows it in the acc. pl. rather than preceding it as one would expect from an agglutinating paradigm. This *-m, however, is clearly of Indo-Uralic age, and thus probably older than most other case suffixes in PIE.

[KathTheDragon]

The main problems I have are the sweeping and largely unexplained moves he makes to pull all the case endings into very conveniently similar forms. For example, he magically pulls an allative ending *-é out of nowhere, relegating the ending *-o (which I think itself rests on dubious evidence) to a mere unaccented byform, despite his supporting evidence for its very existence being stressed! And then his account of the instrumental ending is worse - his entire discussion of the *-h₁ ending is relegated to a footnote whose argument is wholly unconvincing, where somehow the original *-t is supposed to have voiced to *-d and then debuccalised to *-h₁ in nouns, while it escapes unaltered in the pronominal forms - never mind that the only mildly plausible examples of *d > *h₁ are all before another consonant, and followed later in the word by a coronal!

[KathTheDragon]

Oh, and then he comes up with a scenario whereby certain cases with no inherent restriction from occurring with inanimates cannot occur with inanimates, just because his theory says they can't since they have ending-stress - there's no data to back this up whatsoever.

[Nerulent]

I don't think he says a whole lot that's particularly new, but it's nice to have stuff drawn together as more of a big picture thing. I have always quite liked his kissar declension and conflation of the mobile paradigms, as well as calling acrostatic ablaut into question. I also like the connection between the dative and locative, and the internal reconstruction of penultimate stress in mobile nouns and his final paradigm is very clean and attractive.

Some of the other points don't seem so well supported and he seems to conflate different things to get a nice clean pattern to emerge. He also presents the sound laws at the beginning as propositions, but then uses them later on with explanatory power which seems to be a bit circular.

[KathTheDragon]

Yeah, everything he's said about accent-ablaut is repeated in multiple other places, and I didn't notice anything new there at all. I don't fully agree with his proposed developments (particularly the rise of *o and especially claiming all instances of *o go back to *e) and I'd like to see an actual explanation of why the hysterokinetic paradigm split off from the amphikinetic one. Other than that, I don't have any particular issues with the accent-ablaut part.

[WeepingElf]

When I said that I don't see any problems, I was referring to the general outline of the paper; indeed, some details such as his reconstruction of the allative case are IMHO questionable. Kloekhorst, while having some good and interesting ideas, tends to overvalue the conservatism of Hittite (well, he is a Hittitologist, so he tends to see things from a Hittite perspective). Of course, when reconstructing the common ancestor of just two entities (in this case, "Classical" PIE and Anatolian), it is not easy to decide which of the two has innovated where they disagree, especially if one does not have a reliable outgroup as in this case. And certainly, this paper is not the final word on this issue; Kloekhorst can't say he had cracked the ablaut riddle, but he certainly made a worthwhile contribution to its solution, even if many of his points have been stated elsewhere by other authors - which may mean that he is on the right track.

[mèþru]

I thought Indo-Hittite is not mainstream.

[WeepingElf]

I thought Indo-Hittite is not mainstream.

Indeed, it is controversial - but the label "Indo-Hittite" more so than the idea. There are many Indo-Europeanists who consider it likely that Anatolian broke off early but do not speak of "Indo-Hittite". The point is that the Anatolian languages, despite their early attestation, are so different that it is hard to account for them in terms of the "classical" reconstruction of PIE, and it thus seems likely that the branch broke off before PIE reached the stage represented by the conventional reconstruction. Alas, this is all controversial; if you consider what happened to the Celtic languages of Britain and Ireland within a few centuries, quite much is possible.

[Howl]

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?

Another paper with the Leiden model, where every kind of mobile ablaut goes back to an idealized Leiden interpretation of the ablaut in Hitt. kessar using ad-hoc accent shifts. I'm not a fan of that model.

I could write a very long post about what I think is really going on. But let me just note the nouns like kessar 'hand' that are used as proof for this paradigm are not particularly high on the animacy scale. And the Leiden model reconstructs the same form for the absolutive of inanimates and the nominative for animates here.

[KathTheDragon]

I've wondered if we couldn't take kessar back to a paradigm *ǵʰésōr ~ *ǵʰsérm̥ ~ *ǵʰsrés. As far as I can tell, the whole basis of the nominative *ǵʰésr̥ is the idea that the geminate ss is due to contact with another consonant, and so the *s and *r must have been in contact in the nominative. While the idea itself is sound, this application neglects the possibility of levelling from the other cases, and I continue to have doubts about whether or not vocalic and consonantal sonorants really would behave the same, as the Leiden no-syllabification-diacritics approach implicitly suggests.

[Howl]

There is a reason why Kloekhorst uses kessar and not tekan, even though he argues for the same ablaut pattern in tekan. It's pretty solid in the case of kessar.

First of all, a form ǵʰésr is attested in Greek χέιρ (not χέωρ <*ǵʰésōr or ξέρ <**ǵʰsér). But it is somewhat dangerous to assume that any word derives from the same ablaut pattern in multiple IE languages. In IE words can change the ablaut pattern (vṛddhi) and frequently do so.

Secondly, the a in Hittite kessar is never spelled with plene spelling. This means that the a is at most a short vowel. Then there is the geminate s you mention. But there is also the loss of *r in word-final unaccented PIE *-or# or *-ōr#, and Kloekhorst has very good examples of that.

Speaking of which, is there any example where the Erlangen amphikinetic paradigm CéCōC, CCéC, CCCé is securely attested?

Edit:
But I also want to note that there is another ablaut pattern that looks a lot like the amphikinetic one. That is the collective plural *wedṓr, *udnés. So the amphikinetic pattern could be explained with the same collective -> half-animate shift that also produced the -h₂ female suffix. And with that, it's possible to put every mobile ablaut pattern of the Erlangen model in one big pattern without any need for awkward ad-hoc accent shifts:

*CeCC inanimate absolutive
*CCeC inanimate oblique
*CCēC animate nominative
*CCCe collective oblique -> animate oblique
*CeCōC collective absolutive -> half-animate direct (nom -ōC; acc: oCm)

[KathTheDragon]

The canonical Erlangen amphikinetic is CéC-ōC ~ CéC-oC-m̥ ~ CC-C-és ~ loc. CC-éC, which is attested more-or-less as such in Germanic (though with analogical spread of the e-grade suffix the zero-grade forms)

Also, good points about kessar, I feel rather silly for not realising.

[KathTheDragon]

After a quick check of the sound changes to Hittite, it's imaginable that *ǵʰésors is a viable preform, assuming that *rs > *rr medially does imply *rs > *r word-finally. However, the lack of Szemerényi's law is awkward, since the 3pl preterite ending -er < *-ēr < **-ers implies it took place before Hittite broke off. Unless, of course, we assume that the stressed allomorph of the ending was generalised, as under these assumptions *-érs would regularly yield -er...

[Howl]

The canonical Erlangen amphikinetic is CéC-ōC ~ CéC-oC-m̥ ~ CC-C-és ~ loc. CC-éC, which is attested more-or-less as such in Germanic (though with analogical spread of the e-grade suffix the zero-grade forms)

I don't see any vowel change in the first syllable in the attested Germanic forms. All I want is something that attests that these are not just mislabeled acrostatic/hysterodynamic nouns.

[KathTheDragon]

Well of course Germanic doesn't have root ablaut, it's largely been eliminated. Ditto for most of the family, actually - it's really hard to find well-preserved synchronic root ablaut within a single language. Anyway, the existence of any sort of suffix ablaut rules out acrostatics, and root accent rules out hysterodynamics.

[KathTheDragon]

I've spent far too long on it, but I've written a blog post going into detail arguing that the pre-Proto-Anatolian nom.sg. for amphikinetic resonant stems was of the shape *CéC-oC-s, and also that hysterokinetic resonant stems had *CC-éC-s, i.e. the Erlangen nom.sg. minus Szemerényi's law.

[mèþru]

Why is the idea that Grassmann's law was either inherited by both Greek and Indo-Aryan or an areal feature a thing? Dissimilation is pretty unlikely to be an areal feature so why can't it just be an independent development?

[Frislander]

Why is the idea that Grassmann's law was either inherited by both Greek and Indo-Aryan or an areal feature a thing? Dissimilation is pretty unlikely to be an areal feature so why can't it just be an independent development?

You'd think, but borrowing sound changes is attested. My main beef with an areal contact theory here is probably how such a sound change could be spread between two branches that aren't otherwise terribly close.

[Pabappa]

Theyre the only branches that preserve the aspirates into written history*. It could have been once more widespread, but then survived in only these two branches because in the others it got moot. é.g. Slavic collapses all but one of the aspirated stops with its plain form. in fact it may be that it was productive in PIE too, since we wouldnt know, would we? that said, there are details in the two Grassmans' Lwas that suggest perhaps it was not completely developed in proto-Graeco-Aryan, or even that the original theory that it developed twice is the correct one. after all, a similar process appears in some non-IE languages too, so it could just be that languages with aspirates and variable stress tend to have laws like this no matter what the proto-lang was like.

*arm shifted, so the asps are not the original asps.