The Great Proto-Indo-European Thread's Sequel

[dewrad]

This is much like how kw later became p in Gallic/Welsh, but gw did not merge with the already extant b.

Contentious! The existence of *gw in Proto-Brythonic is far from assured.

Those who favour the Insular Celtic hypothesis (i.e. that the Brythonic and Goidelic languages form a branch on the Celtic family tree distinct from the Continental Celtic languages), assume that Proto-Celtic *gw remained as such in "Proto-Insular Celtic", giving reflexes of g in Goidelic and gw in Brythonic. For example, PC *gwed-yo- 'to pray' (< PIE *gʷʰedʰ- 'wish') gives OI guidid and MnW gweddïo.

However, we have a further reflex of the same etymon in Gaulish: uediíu-, which certainly indicates an initial /w/. Given that initial *w in Brythonic has the reflex gw anyway, there is a large body of opinion that would group Brythonic with Gaulish here.

Personally, I suspect that if there was a gap in Proto-Brythonic where one would expect a voiced labiovelar stop, it would be a further motivating factor "pushing" *kw towards *p.

[Znex]

This is much like how kw later became p in Gallic/Welsh, but gw did not merge with the already extant b.

Contentious! The existence of *gw in Proto-Brythonic is far from assured.

To add to this, Insular Celtic languages (and Celtic loanwords in Romance languages in part) underwent a sort of initial fortition, presumably to keep mutated forms distinct from isolated forms.

This initial fortition affected all sonorants (as attested still in Old Irish and partially in Welsh), including *w:
PC *m n l r w s > IC *M N L R W S >> Welsh /m n ɬ r̥ g(w) s~h/, OI /m N L R f s/ (so initial sonorant consonants in isolated forms merged with geminate fortis consonants)

Intervocalically in comparison:
PC *m n l r w s > IC *ˡm ˡn ˡl ˡr ˡw ˡs >> Welsh /v n l r w~∅ h~∅/, OI /ṽ n l r w~∅ h~∅/

So in Brythonic, adding intervocalic lenition,*gʷ became indistinguishable from *w. *gʷ on the other hand is still distinguishable in Old Irish because it never became identical to *w in initial position (instead *gʷ merged with *g initially).

[Moose-tache]

Just to set the record straight, I was using Gallic p as an illustration; I'm not committed to the idea that this is exactly what happened. The idea still stands that gw > b while gwh /> bh is a perfectly normal sound change that requires no special explanation.

[jal]

A pre-Celtic speaker may even have wondered if b was an allophone of gw or gw was merely an allophone of b.

Pre-Celtic linguist be like "[ b ]/[gʷ] = /b/ or /gʷ/"?


JAL

[Howl]

That said, I think that Holzer's etymologies are bogus and "Temematic" never existed.

I'm reading Matasovic paper on Balto-Slavic substratum words (the one refuting Holzer's Temematic). The usual strategy is presenting alternative etymologies using loose semantics. But that does not work when the meaning of the word is very specific:

Tm. *berg’-, *borg’-: 'furrow' [...] The Baltic and Slavic forms cannot be derived from the same prototype, so the derivation from Temematic amounts to a root etymology. An alternative etymology (Vasmer I:109) relates the Slavic words to PIE *bhers- ‘point’ (OHG burst ‘bristle’, OIr. barr ‘top’); PIE *bhors-dheh2 would regularly yield PSl. *borzda, but in this case the Baltic words must be unrelated. Smoczyński (61—2) derives them from the PIE word for ‘birch’ (*bherHg’h- > OCS brěza, Germ. Birke, etc.). It would originally have denoted a furrow delineated by birch branches, which is semantically difficult.

Tm. *trono-, *tronto- ‘drone’ [..] These words indeed appear to be of substratum origin, since the initial *t in BSl. cannot correspond regularly to Germanic *d- and Greek *th-. Note, however, that the Slavic and Baltic formations are not identical, so they may have been borrowed from different sources, or through different intermediaries.

Outside of Temematic, I also found this gem:

However, although there are many Uralic loanwords in individual BaltoSlavic languages, especially in Russian and Latvian, there do not seem to
be any Uralic loanwords that could be attributed to Proto-Slavic, or to Balto-Slavic periods (Kallio 2005).
[...]
BSl. *pel-/ *pāl- ‘burn’ > PSl. *paliti ‘burn’ ... The root does not seem to be attested outside Balto-Slavic, but we think it might be derived by metathesis from PIE *leh2p-

Big facepalm.

[WeepingElf]

I won't deny that a language with such sound changes (basically a loss of voicing distinction, followed by a voice onset shift: T D Dh > T T Th > D D T) could have existed somewhere, but Cimmeria strikes me as the wrong place for that. It could, for instance, have been a Corded Ware language in an area later clobbered by Uralic, such as Estonia or the Upper Volga.

Addition: In such a location, the loss of voicing distinction may be due to the areal influence of Uralic, or the substratum influence of a Para-Uralic language, as Proto-Uralic is known to have lacked voiced stops. The subsequent voicing onset time shift may not even be real, but have happened when the words in question were borrowed into Balto-Slavic, which had a voicing distinction but no aspiration distinction.

[WeepingElf]

Here's some speculation on the origin of the PIE stop system, which is based on this blog post by Tropylium, but modified a bit - not really that much, I have to concede.

The idea is that things started with a "typically Mitian" system with one set of stops unmarked for either voice or breath (thus the *T set), and a corresponding set of weak obstruents, possibly voiced fricatives (the *Ð set). There was a dissimilation rule against two *Ð set consonants in a morpheme. First, the *Ð set lost its labial member by merging with *w, which also explains the frequent occurence of *wl- and *wr- in PIE. (If Proto-Uralic had a *Ð set at all - it is far from certain that *ð, *ð' and *x were such consonants - it seems to also have lost the labial member, perhaps the same way? Did this already happen in Proto-Indo-Uralic?) The remaining consonants of the *Ð set hardened to voiced stops, thus becoming the *D set, while the *T set gained aspiration (thus becoming a *Th set). In the next step, some autosegmental prosodic feature connected to morphemes came into play, perhaps a residue of a pre-GVC vowel feature of some sort, though I don't have a good idea of what kind of feature does the job. This affected the *Th set such that all of them in a morpheme with the feature were voiced; thus the *Th set split into a *Th set and a new *Dh set which do not co-occur within a morpheme.

This is almost the traditional reconstruction of the PIE stop system, except that the *T set is aspirated in this model (some scholars assume that the *T set was aspirated in Proto-Celtic, but I don't know on what grounds). In Germanic and Armenian, the *D set then devoiced, while in most other branches of IE, the *Th set (and in many branches, also the *Dh set) lost its aspiration, and there you are.

[KathTheDragon]

In the next step, some autosegmental prosodic feature connected to morphemes came into play, perhaps a residue of a pre-GVC vowel feature of some sort, though I don't have a good idea of what kind of feature does the job. This affected the *Th set such that all of them in a morpheme with the feature were voiced; thus the *Th set split into a *Th set and a new *Dh set which do not co-occur within a morpheme.

This step really is under-supported and thus a severe weak point in your argument; what feature causes aspirated voiceless stops to become breathy-voiced stops? Is there such a feature?

[WeepingElf]

In the next step, some autosegmental prosodic feature connected to morphemes came into play, perhaps a residue of a pre-GVC vowel feature of some sort, though I don't have a good idea of what kind of feature does the job. This affected the *Th set such that all of them in a morpheme with the feature were voiced; thus the *Th set split into a *Th set and a new *Dh set which do not co-occur within a morpheme.

This step really is under-supported and thus a severe weak point in your argument; what feature causes aspirated voiceless stops to become breathy-voiced stops? Is there such a feature?

I admit that this is a weak point; I have no idea what kind of feature could exert such an influence. This is just a brain fart of mine which swirled round and round in my head, and all I intended with my post was to let it out and see what others think about it.

[KathTheDragon]

So far the only vaguely-plausible way I've seen to project the *T and *Dʰ series back to be originally the same is to assume tone-induced voicing, which requires *Dʰ being modally voiced immediately after the sound change.

[mae]

-

[Nerulent]

I can’t remember exactly, but I believe Kortland mentions in passing that he follows a similar theory except that the mobility of accent is the conditioning factor, due to the fact that mobile stems more often have breathy voiced consonants (apparently). Either I think there is at least some merit to the idea that there were originally only two MoA’s of stops and that the T and Dh series share a common origin.

[KathTheDragon]

Either I think there is at least some merit to the idea that there were originally only two MoA’s of stops and that the T and Dh series share a common origin.

Sure, it does work well enough, although I have yet to see any compelling reasons within PIE itself to assume this theory (or indeed, any other theory professing to eliminate a stop series)

[Tropylium]

Wrt. Tropylium's linked post, the rarity of T + Dh roots in contrast to the frequency of Dh + Dh roots doesn't demand that the breathy feature was originally vowel specific and then spread, because long distance consonant voicing assimilation is attested

Bonus content: this part of the wild speculation has been inspired by a hypothesis by Ray Cote (WeepingElf may remember him from the Nostratic-L, haven't heard much of him in recent years) that PIE *Dh would correspond to Uralic stop + back vowel, PIE *T to Uralic stop + front vowel.

[WeepingElf]

Two thoughts on the idea that we are just dealing with an assimilation rule here.

1. The domain of the assimilation rule seems to be the morpheme rather than the word, which AFAIK is a rather unusual condition with such rules. There is no such assimilation at the word level, as such well-reconstructed forms as *bhereti demonstrate.

2. Whatever mechanism led to the incompatibility of the *T and *Dh sets in roots, at that time, the *D set can hardly have been phonologically intermediate between the two but probably was marked by a feature that set it off the other two. Various features have been proposed: [+ejective] (Gamkrelidze & Ivanov; Hopper), [+implosive] (Kortlandt), [+stiff voice] (Clackson) and a few others; further possibilities are [-aspirated] (would require that the *T set was aspirated back then) and even [-stop].

Neither of the two exclude an assimilation rule within PIE, but the first is supportive of the notion of an autosegmental feature connected with morphemes.

As for Ray Cote's hypothesis, it doesn't ring a bell with me, though I think I have read that name once or twice. It may well be that the autosegmental feature in question preserves some pre-GVC distinction in the vowel system (if the GVC happened at all, of course), though front/back doesn't seem like a plausible candidate to me. Alas, without reliable correspondence sets between IE and Uralic (or whatever is most closely related to IE), this remains speculation.

[KathTheDragon]

Counterpoint: point 1 is unprobative as the sound change can easily be very old, and have been completely fossilised within synchronic PIE. This allows for, at any point in pre-PIE, productive morphemes like 3sg -ti to have been restored from environments where they would not have been affected (in particular, note that word-initial *sT- was indifferent to the stop constraints, suggesting *T(R)e(R)s- would always have *-ti, and from there *-ti could easily be restored into *T(R)e(R)H-, then *T(R)eR- (and *T(R)e(R)T- depending on how local assimilation worked at the relevant time). Roots containing no stops at all would similarly all have *-ti)

[mae]

-

[KathTheDragon]

2pl -tha has been claimed to be < *-th₁e, cf Greek -τε

[Kuchigakatai]

Does anyone know the etymology of Latin adverbial -ē?

[Kuchigakatai]

Does anyone know the etymology of Latin adverbial -ē?

I have been informed Michael Weiss (the author of Outline of the Historical and Comparative Grammar of Latin) says it is uncertain, there being two possibilities:
- 1) an ablaut variant of the masculine/neuter ablative singular -ōd
- 2) the Proto-Indo-European instrumental case ending -eh1, apparently after receiving an analogical -d as influence from the ablative (the -d is relevant because in Old Latin the ending is attested as <-ED>, e.g. <FACILVMED> for Classical facillimē)