The Great Proto-Indo-European Thread's Sequel

[Salmoneus]

In fact it may have trailed off even further west, as late IE *ō gives Proto-Celtic ā.

I think the shift for the long vowel and the back vowel are different things - the eastern feature is for the short vowel, and only a subset of languages expand that to the long vowel. [and presumably celtic would have been southeast of germanic at the time, rather than west]

It does seem, though, that *o and *a, whatever they really were, were probably quite close to one another at least in post-Tocharian IE.

[WeepingElf]

Yes, *a and *o may have been something like [ɑ] and [ɒ], respectively (while *e was [æ]) - dangerously close to each other.

[Tropylium]

It is unusual that Germanic shares this characteristic with those languages... but given that it lacks their other (more unusual) innovations (like shifting the palatovelars forward, merging labiovelars with velars (except in Albanian) and RUKI), it's more likely that this is areal influence from Baltic into neighbouring Germanic.

There's a proposed Germanic parallel to RUKI - West Germanic breaking, which is triggered by *r *l *h *w (so basically RUK). This goes leftward to a vowel and not rightward to an *s, but may mean that Germanic inherited some similar allophonic preambles.

But, on the other hand, this could be just a coincidence - velarization of coda liquids is common enough (ditto sibilant retraction; the fact that also presumably palatalized *is merges with *{RUK}s in satem proper seems more significant). Same goes for all the *a / *o mergers: if these were close enough to begin with, they could've merged independently several times (no need to even assume an areal connection in all cases).

A somewhat more interesting change could be *ee > *aa, which covers (besides trivially II) at least NWGermanic, Albanian and Phrygian. These are guaranteed to not be a subgroup (because Gothic), so they probably tell us something about the PIE phonetics: why is this attested multiple times but corresponding *e > *a nonexistent? (OK, Aukštaitian aka High Lithuanian has that too, but it's too recent to tell us much about the PIE state of affairs.) Presumably because *e and *ee had some qualitative difference that made the latter more likely to be retracted.

[KathTheDragon]

A somewhat more interesting change could be *ee > *aa, which covers (besides trivially II) at least NWGermanic, Albanian and Phrygian. These are guaranteed to not be a subgroup (because Gothic), so they probably tell us something about the PIE phonetics: why is this attested multiple times but corresponding *e > *a nonexistent? (OK, Aukštaitian aka High Lithuanian has that too, but it's too recent to tell us much about the PIE state of affairs.) Presumably because *e and *ee had some qualitative difference that made the latter more likely to be retracted.

There's evidence for *e being mid-high and *ē being mid-low. Firstly, Greek preserves exactly this. You could also add Proto-Germanic i-umlaut, which affected short *e but not long *ē, implying that the former is higher than the latter. Then there's exactly this lowering under consideration (and imo lowering is more likely to be the primary change here, rather than backing - ɛː > æː (> ɑː) feels unexceptional.

[cenysor]

There is a very interesting article on Language Log about "Tocharian C"

[jal]

There is a very interesting article on Language Log about "Tocharian C"

Quite interesting indeed! Thanks for sharing.


JAL

[Moose-tache]

The existence of Tocharian C is based on the unpublished papers of Klaus T. Schmidt, who died in 2017. The papers were published last year by his long time collaborator Zimmer (also Klaus T.) as "K. T. Schmidt: Nachgelassene Schriften." I can read a little German, and even if I couldn't I would run a schoolbus off the road to get my hands on this thing, but alas, I cannot find it anywhere. It's available as a hardback for almost 90 Euros, and it's in a university library in Munich. But I can't find any record of it online. Who prints a monograph with no trace of it in any scholarly journal? What kind of monster is Klaus T. The Younger? This is the greatest tease I've ever seen in historical linguistics in my lifetime. Does anyone have any access to this text? Carsten, Hans, are you the heroes we need?

[Pabappa]

It's Stefan Zimmer, not Klaus T Zimmer. Also, are we sure this is not an elaborate April fools joke? I dont read LL

[Moose-tache]

It's Stefan Zimmer, not Klaus T Zimmer.

It's both, actually. But you're right Stefan seems to be the first author/editor. It's not an April fools joke unless the publisher is in on it, and Amazon is willing to take almost a hundred dollars of your money as a joke.

EDIT: Apparently German Tocharianists with silly repetitive names might be a whole thing. There's also Sieg and Siegling who worked together in the early 20th century. I'd love to find out which one was the mentor and which one the protege, but I haven't found their bios yet. If I can find one more pair I'd say we have a pattern on our hands.

[jal]

It's both, actually.

It's not, actually. Some fuck-up by Amazon? And it seems purchasable for a mere € 68,00 :).


JAL

[WeepingElf]

Kloekhorst has added a new paper to his web site in which he argues in favour of *h2 having been [q] (uvular stop) in Early PIE, and *h3 the same but labialized. His argumentation is basically that in Lycian, *h2 surfaces as a stop (transcribed χ) posterior to the one transcribed as k, and that the change q > χ is common enough to have happened independently in Late PIE, Hittite and Luwian. Also, he points at some points where *h2 and *h3 behave like voiceless stops rather than continuants (e.g., by appearing as geminates in intervocalic position) in Hittite, and gives some other arguments as well.

It seems reasonably plausible to me, though I am in not really convinced by this (AFAIK, Lycian phonology is very shaky ground, so using this language as sole evidence for such a reinterpretation of a PIE phoneme is quite a bold move); but what do you think about it?

[Pabappa]

They don't pattern like stops in PIE . E.G. there are no roots in PIE consisting of just 3 stops, but there are roots with 2 stops & a laryngeal, as far as I know. I didnt read the paper though.

[KathTheDragon]

Additionally, stop-stop clusters are uncommon in PIE, and are really all best explained as being the zero-grade of some other root. The phonotactics all support fricative interpretations.

Having read his article in detail some time ago, I've had a while to think about it, and I've concluded that his sole compelling point is the gemination - it's utterly inexplicable by our current understanding. But, of course, that's not enough by itself. Conceivably, we could have PIE *ḫ > PAnat *qʰ > *qː > *ḫḫ (with ḫ standing in for the uvular fricative), but this is abviously ad-hoc, and is probably typologically implausible with the fortition

[Znex]

They don't pattern like stops in PIE . E.G. there are no roots in PIE consisting of just 3 stops, but there are roots with 2 stops & a laryngeal, as far as I know. I didnt read the paper though.

The tricky thing there is that in reconstructed PIE (=late PIE by Kloekhorst), the laryngeals are already undoubtedly fricatives/approximants/etc.. The phoneme-root distribution itself may conceivably have been different at an earlier stage of PIE, but that needs to be demonstrated.

[Pabappa]

Do we really know if it was geminate? Could the hittites have been sloppy spellers, like us, where "egg" has 2 g's,"bee" & "be" rhyme, etc? They used double letters for voiceless single stops elsewhere, right?

[KathTheDragon]

The Hittites were very consistent with spelling intervocalic consonants either singly or geminately, so yes, they were assuredly geminate.

[jal]

The Hittites were very consistent with spelling intervocalic consonants either singly or geminately, so yes, they were assuredly geminate.

Knowing nothing about Hittite and PIE, but is there an offchance that double consonants represented some other sounds than geminates (I don't know, affricates, or emphatics or ingressives or whatever)?


JAL

[KathTheDragon]

It's unlikely. Any written consonant could be written geminate, and in the case of resonants, the graphic geminates are largely the result of consonant assimilations, so phonetic geminates is the most likely for them. In the case of the stops, there is some subtle indication that they are phonetic geminates too, namely that they close a syllable when intervocalic.

[jal]

It's unlikely. Any written consonant could be written geminate, and in the case of resonants, the graphic geminates are largely the result of consonant assimilations, so phonetic geminates is the most likely for them. In the case of the stops, there is some subtle indication that they are phonetic geminates too, namely that they close a syllable when intervocalic.

Ok, thanks!


JAL

[WeepingElf]

Your objections are valid. Laryngeals clearly didn't behave like stops in PIE; for instance, they didn't take part in the root structure constraints that govern the distribution of stops. Even the gemination of laryngeals in intervocalic position does not necessarily mean that they were stops. Sure, this is a behaviour that puts them in a class with the voiceless stops, while *s does not show this - but *s is a sibilant, and sibilants sometimes show different behaviour from non-sibilant fricatives.

And as I already wrote yesterday: Lycian phonology is so poorly known that we can't really draw such far-reaching conclusions from it! Who says that a letter like χ, which appears to be the regular reflex of *h2 and *h3 in Lycian, always represents a stop, and never a fricative? And while q > χ is common as dirt and χ > q quite rare, nobody can say that the latter couldn't have happened in this language. Think about how English speakers tend to pronounce a name like Bach as [bæk]. Perhaps there was a substratum in SW Anatolia that lacked dorsal fricatives, so [χ] was changed into [q] or [k].