The Great Proto-Indo-European Thread

[Zaarin]

I understand it's tradition, but it feels very inconsistent to me. The book doesn't write Latin deivos and datus as <DEIVOS> <DATVS> nor does it put Vedic in Devanagari (and those two are as important to the field as Greek, especially Sanskrit, without which the entire field of Historical Linguistics might not exist (or a least be heavily delayed in coming forth)! Nor does anyone put Hittite in Akadian Cuneiform or Old Norse in runes.

It is by no means a dealbreaker, but it does feel like a bizarre and archaic appeal to a pointless tradition. It's as if mathematicians used roman numerals for no real reason whenever talking about geometry but nowhere else.

Shhh! Don't give them ideas! As if geometry weren't hard enough to understand already!

[Dewrad]

I understand it's tradition, but it feels very inconsistent to me. The book doesn't write Latin deivos and datus as <DEIVOS> <DATVS> nor does it put Vedic in Devanagari (and those two are as important to the field as Greek, especially Sanskrit, without which the entire field of Historical Linguistics might not exist (or a least be heavily delayed in coming forth)! Nor does anyone put Hittite in Akadian Cuneiform or Old Norse in runes.

It is by no means a dealbreaker, but it does feel like a bizarre and archaic appeal to a pointless tradition. It's as if mathematicians used roman numerals for no real reason whenever talking about geometry but nowhere else.

It's not just tradition, it's the audience. Most people studying IE diachronics have Greek*. If they've learnt Sanskrit it's in transliteration (I know I couldn't deal with Sanskrit in Nagari), and everyone knows Latin in its classical form. It's not really an unreasonable assumption.

*Greek. Not Ancient Greek. Unmarked, "Greek" refers to the language of Homer. The debased horror spoken in the Dodecanese these days is always Modern Greek. This is, to my knowledge, unique.

[zompist]

Another couple of reasons: 1) the Greek alphabet is better writing system for ancient Greek than Roman, most notably because it has seven vowel symbols; 2) any transliteration ties the romanization down to a period and to a particular dialect, and makes a lot of assumptions.

[R.Rusanov]

I kinda agree. You can't just write Laocoon cause no one's gonna know what you mean, is it two omicrons, an omicron and an omega, two omegas, what about the first one, etc

[Pole, the]

I kinda agree. You can't just write Laocoon cause no one's gonna know what you mean, is it two omicrons, an omicron and an omega, two omegas, what about the first one, etc

Laokóωn. CHECKMATE

[Astraios]

*Greek. Not Ancient Greek. Unmarked, "Greek" refers to the language of Homer. The debased horror spoken in the Dodecanese these days is always Modern Greek. This is, to my knowledge, unique.

Modern Hebrew and Neo-Aramaic are somewhat the same - everyone assumes I'm some kind of religious scholar if I say "I know Hebrew and Aramaic".

[Salmoneus]

I kinda agree. You can't just write Laocoon cause no one's gonna know what you mean, is it two omicrons, an omicron and an omega, two omegas, what about the first one, etc

That's a terrible reason. It's not like Greek has some magically incommensurability to it that means it can't be transcribed just as every other language can be transcribed.
[Odd example, too - if you write Laocoön, everyone's going to know you mean Laocoön, because who else are you likely to be talking about when you say 'Laocoön'?]

FWIW, older philosophy texts used the greek alphabet, and I think you can still find it in books marketed toward the classical philosophy people, but generally newer books seem just to transliterate. You're more likely to find the greek letters in a quoted phrase than in a technical word, though - but that might just be a relic of the old "make your quotations as unintelligeable as possible" tradition.

[KathTheDragon]

To stave off boredom whilst listening to professors at the Univeristy I was visiting today, I was thinking about PIE's obstruent system, and I decided that it's quite possible that PIE exhibited a neat 4-way constrast between voiced and unvoiced stops and fricatives. The supposed development would make h₂ /χ/ and h₃ /xʷ/. h₁ is represented by various phonemes, including /x/ and /ɸ/. It would also make laryngeal colouring a very early development, well before Anatolian split from PIE. The voiced aspirate series was originally voiced fricatives, which fortited to voiced affricates in all dialects but those ancestral to Germanic and Italo-Celtic (the latter had the fortition later, after Italic and Celtic split from each other). In Graeco-Aryan, the frication developed into breathy-voice, while in the others, it was lost and the series merged with the voiced stops. Other developments went as normal.

Thoughts? This is probably a load of rubbish, given that I came up with it over the course of about an hour.

[WeepingElf]

To stave off boredom whilst listening to professors at the Univeristy I was visiting today, I was thinking about PIE's obstruent system, and I decided that it's quite possible that PIE exhibited a neat 4-way constrast between voiced and unvoiced stops and fricatives. The supposed development would make h₂ /χ/ and h₃ /xʷ/. h₁ is represented by various phonemes, including /x/ and /ɸ/. It would also make laryngeal colouring a very early development, well before Anatolian split from PIE. The voiced aspirate series was originally voiced fricatives, which fortited to voiced affricates in all dialects but those ancestral to Germanic and Italo-Celtic (the latter had the fortition later, after Italic and Celtic split from each other). In Graeco-Aryan, the frication developed into breathy-voice, while in the others, it was lost and the series merged with the voiced stops. Other developments went as normal.

Thoughts? This is probably a load of rubbish, given that I came up with it over the course of about an hour.

I don't think it is a load of rubbish. Indeed, I consider it not unlikely that PIE was just as you fancy. Though I'd expect /ɸ/ to have merged into *h3. Laryngeal colouring is indeed a matter of Pre-Anatolian PIE, as it is found in Hittite.

And what regards the transliteration sub-topic: Fortson transliterates his Greek, using macrons to disambiguate eta and omega from epsilon and omicron. On the other end, Bopp's Vergleichende Grammatik (1833-1847) gives Sanskrit, Avestan and OCS in both native script and transliteration.

[KathTheDragon]

Actually, I posit it the other way round: /xʷ/ > /ɸ/, as Anatolian seems to treat h3 like h1 after vowel colouring (afaik). /ɸ/ later debuccalises into /h/, as does /x/. /χ/ fronts to /x/, parallel to the stops (I really like the idea of uvulars as being traditional 'plain' velars, cos it makes explaining the laryngeals so much easier), and in non-Anatolian (post-Anatolian?) PIE, also debuccalises. A possible middle step in the development of /x/ is that it fronted with the velars to some palatal, then debuccalised. A pseudo-Satem Anatolian would explain Luwian.

The one thing I can't do (but neither can the traditional reconstruction) is explain how /b/ would become /m w/ in many envrionments, as evidenced by /m w/ being able to form clusters with /r l/ while /n j/ cannot, and apparently some alternations between /m/ and /w/ within morphemes.

Edit: Is Anatolian subject to delabialisation of labiovelars adjacent to /u/?

[Zju]

The one thing I can't do (but neither can the traditional reconstruction) is explain how /b/ would become /m w/ in many envrionments, as evidenced by /m w/ being able to form clusters with /r l/ while /n j/ cannot, and apparently some alternations between /m/ and /w/ within morphemes.

That's why I like the idea that voiced stops were implosives, while breathy voiced stops were infact plain voiced - implosives are known to easily change into sonorants. (nearly) all instances of ɓ would then have turned into m or w, in this way explaining the rarity of b. Afterwards implosives were lost everywhere, causing chain shifts in different directions, one of them leading to the traditional reconstruction and eventually to Indo-Aryan branch. In another, they became ejectives and yet in another - preglotalized voiced stops.
This explains nicely all developments and doesn't require that much steps and assumptions at all.

[R.Rusanov]

how would implosives have gotten there short of contact with africa

[Zju]

Why do they need to be African influence? Europe by no doubt was typologically different back then.

[Nortaneous]

re phonotactics, possibly relevant:

Consonant clusters in Kabardian are predominantly regressive, i.e. the point of articulation of the first element is closer to the lips than that of the second element.

*dʰǵʰyes-
*dʰǵʰu-
*psten-
p. NWC but this would predict there'd be pP clusters, which there aren't. (and why are there no non-breathy PP clusters? i only looked at roots; what's it like for derived words?)

also why have b > m w instead of b > w? does wn- not occur?

[8Deer]

Why do they need to be African influence? Europe by no doubt was typologically different back then.

Not to mention the fact that implosives occur in many regions other than Africa and that at least one IE language (Sindhi) has them.

[TaylorS]

Over on the blog Paleoglot I have seen proposed that the 3 sets of PIE plosives were plain voiced, creaky voiced, and unvoiced. This shifted to the standard system in the late PIE dialects that became Indo-Aryan, Hellenic, and Armenian.

[KathTheDragon]

That's just as marked as the traditional reconstruction, isn't it?

re phonotactics, possibly relevant:

Consonant clusters in Kabardian are predominantly regressive, i.e. the point of articulation of the first element is closer to the lips than that of the second element.

*dʰǵʰyes-
*dʰǵʰu-
*psten-

But zero-grade formations can form clusters from consonants of any two PoAs, cf. *gʷʰedʰ-, zero-grade *gʷʰdʰ-. Also, do Gr, Gl clusters not count? And for that matter, what does Kabardian have anything to do with?

and why are there no non-breathy PP clusters?

Apparently for much the same reason that there are (nominally) no DeG- roots.

also why have b > m w instead of b > w? does wn- not occur?

I don't get what you're trying to say here. We can find root-initial clusters of *ml, *mr, *wr and *wl, but not **nl, **nr, **yl, **yr. Given the extreme rarity of *b, it is natural to say that many instances of *b became *m and *w.

[Nortaneous]

i searched a list of roots, and the only ones with initial P(s)P clusters fit the kabardian model. (that pattern is actually common to NWC and kartvelian.) was wondering how far the caucasian parallels ran.

different phonotactics in roots vs derived forms is possible, but n = 3 so probably just chance, dhgh- could be derived forms and i've heard that some people don't reconstruct *psten- at all.

i have also heard it hypothesized that all *b became *w. why say it was a split into *m *w? ml- mr- clusters are far less strange than wl- wr-.

[KathTheDragon]

I have seen a suffix cited as *-men- ~ -wen-.

[R.Rusanov]

that could be an m or w infix followed by the extremely common suffix en

[KathTheDragon]

That may be so, but I would expect to see **nl and **nr as well, if *ml and *mr were original.

[Tropylium]

The argument is "'it's unlikely, therefore it didn't happen' is invalid", not "it's unlikely, therefore it happened".

No one is making argument #1 here. Lots of people are making the argument "hypothesis A is typologically way less plausible than hypothesis B, therefore hypothesis B is preferrable".

It's funny how the more we work on PIE, the less SAE it becomes. PIE is incredibly un-european.

The same thing happened to reconstructions of Proto-Semitic, IIRC. It seems to look less and less like Hebrew and Arabic and more and more like Amharic. It is even reconstructed to have /tɬʼ/, fer Christ's sake!

One thing that has to be understood about the "old" proto-languages (particularly PIE, but also Proto-Semitic, Proto-Uralic, and I imagine to an extent other things like Proto-Dravidian) is that they were kinda pre-scientific; worked up before anyone had an idea how sound changes usually work. Instead they were initially very heavily based on the idea that there is one language in the family that is the "key" language and should be assumed conservative in every respect until proven otherwise. (Arabic, Sanskrit and Finnish respectively, though there were some conflicting opinions too.)

And so you had people gleefully proposing that *c sometimes goes to *kʷ under unclear conditions, or that a correspondence *tt in lang 1 ~ *ht in lang 2 indicates that the proto-language had an alternation *kt ~ *ɣt under some sort of now-levelled conditions, or that *sˤ becomes /ts/ unconditionally. The most obvious problems have been at least proposed to be done away with by now, but otherwise the cleanup process is still ongoing. There's lots of room for optimization towards less bizarrity.

re phonotactics, possibly relevant:

Consonant clusters in Kabardian are predominantly regressive, i.e. the point of articulation of the first element is closer to the lips than that of the second element.

*dʰǵʰyes-
*dʰǵʰu-
*psten-
p. NWC but this would predict there'd be pP clusters, which there aren't. (and why are there no non-breathy PP clusters? i only looked at roots; what's it like for derived words?)

the 3rd well-known example is voiceless: *h₂artḱo- "bear"

[KathTheDragon]

And a counter-example (though everyone says it's clearly a borrowing), *ksweks "6".

[Nortaneous]

That may be so, but I would expect to see **nl and **nr as well, if *ml and *mr were original.

why?

the whole point of bringing up caucasian is that, if the caucasian protolangs were at all like the current caucasian languages in that regard (which i don't know) and if there was caucasian influence on PIE (which seems fairly well established), it's not implausible that some of the regression constraint would be present in PIE.

even if both m and w come from b, the regression constraint would still be there.

it's not *necessary* to posit caucasian influence to get the regression constraint -- there wasn't any caucasian influence on tibetan, where b- can occur before a preinitial. and it may be that it wasn't even present in the relevant protolangs.

so there are three issues here:
1) regression constraints appear to be an areal feature of the caucasus, and may have existed even at the time of PIE. (if it seems implausible that something could last that long, remember that the very unusual clusters wl- wr- persisted into the middle english period. wr- was only lost a few centuries ago, and for all i know may still exist somewhere in britain. no words with wl- survived into modern english.) this is the weaker case.
2) regression constraints are things that exist. constraints toward cluster POA difference exist -- in english. (pl bl *tl *dl kl gl) the constraint that appears to be present in PIE isn't that crosslinguistically weird, other than the presence of *wl- *wr- clusters, which can be explained if *w is from earlier *b. this is the stronger case.

as for *b splitting into *m *w, one affix is not enough evidence -- especially since it contains another nasal, and could be a case of feature spreading. if *b split into *m *w, there would still be a regression constraint, but either one of *m *w was already present or the split somehow managed to produce *ml *mr *wl *wr. it seems more likely to me that *b just became *w, at least initially.

[Tropylium]

I am entertaining the hypothesis that the three-vowel system */a i u/ of the pre-ablaut stage resulted from what I call the "Great Vowel Collapse" (GVC), wherein all non-high vowels of a richer inventory (at least */a e i o u/) of a yet earlier stage (perhaps Proto-Indo-Uralic?) collapsed into */a/, and that the three velar series preserved the features [+front] and [+round] of the collapsed vowels. So */ke/ > */ḱa/, */ka/ > /ka/; */ko/ > */k^wa/. But this is just personal speculation and may be entirely wrong-headed!

k wasn't palatal tho
and q wasn't either

This isn't a particularly good counterargument, though. We could well assume that pre-PIE *e caused no coloring of *k, while *a caused uvular coloring: *ke *ka *ko > *ka *qa *kʷa? A good parallel is Proto-Uralic >> Northern Mansi:
*ki > *kæ > /ka/
*ka > *qa > /χa/
*ku > *qʷa > *χʷa > /χo/
*ky > *kʷæ > *kʷa > /ko/
(mid vowels omitted; these generally become the new high vowels)

My previous idea that *kʷ = [qʷ], *ḱw = [kʷ] would work pretty well with this too; if pre-PIE had an *y, we expect it to have been rarer than *u, and indeed *ḱw is rarer than *kʷ.

OTOH what I think is a worse problem is that this seems to replicate an issue of the slanted PIE vowel inventory: since EPIE *k > trad. *ḱ is way more common than EPIE *q > trad. *k, this setup implies that pre-GVC there was also a vowel system where /e/ was way more common than /a/?

I suspect only parts of this Great Vowel Collapse scheme are on the right track. There might have indeed been a vowel collapse accompanied by a change like *ko > *kʷa (>> *kʷe ~ *kʷo) — but the uvulars ("plain velars") don't have to be of the same age. Perhaps they're older; perhaps younger.

As long as we're speculating, it may or may not be worth noting that none of the families proposed to be in some fashion related to IE have labiovelars. Plenty of them however contrast velar/uvular — Kartvelian, Chukotko-Kamchatkan, Eskimo-Aleut; Semitic has *χ *ʁ — or palatal/velar consonants —Uralic, Turkic, Mongolian, Tungusic, Dravidian, CK & EA again. Elsewhere in Afrasian there are labiovelars attested in some far-off branches but we know too little to say if they're innovations or not. (Given that at least the labiovelars in Ethiopian Semitic & some Berber dialects are innovations, I lean on a "nope". They seem to be sufficiently unstable consonants that they should not be assumed offhand to be 10000 years old. I don't think they're found as monophonemes in any IE language since Old Irish, for example.)