The Great Proto-Indo-European Thread's Sequel

[Tropylium]

For the root *kes 'to comb, to scrape' I can find reflexes in all branches except Tocharian and Armenian.

I'm using a slighly more closely division with e.g. Northwest Germanic, Gothic, Old Prussian, East Baltic and Slavic counted separately. Not to suggest that Germanic or Balto-Slavic aren't branches, but to measure how well roots are doing at surviving. Also, my chart compiles just the LIV+NIL data, which leaves out a lot of one-off nominalizations. For poorly attested roots many are probably dubious. (Sometimes even for otherwise clear ones. I would for example not accept Albanian ka 'ox' as a probative reflex of *kes-.)

I don't doubt that *k *g *gʰ existed even in early PIE, but the fact that they're rare in the best-distributed core vocabulary, and especially in Anatolian, suggests that they're recent pre-PIE introductions (similar to e.g. initial /v z/ in English). Maybe first as loanword phonemes (uvular stops?), then later on with more *k ~ *k correspondences generated thru developments like exceptions to satemization and post-PIE or inner-PIE loanwords.

[KathTheDragon]

Why is it important to look at widely-attested cases? For example, *ḱenk- "hang" is only in Germanic, Latin, Sanskrit, and Hittite, but its stem-final velar can't be secondary, nor does pre-PIE loaning seem particularly likely.

[Frislander]

I'm more inclined to follow a theory where the palatovelars represent an innovation of the satem languages, though clearly relatively early, and probably any pre-satem dialect(s) may have co-existed during the time of PIE itself, but I am not convinced that the system currently reconstructed in a single PIE dialect for long.

[mèþru]

I really doubt that, for instance, Balto-Slavic and Indo-Iranian developed satemisation at the same time and have a common ancestor.

[Howl]

Also, my chart compiles just the LIV+NIL data, which leaves out a lot of one-off nominalizations. For poorly attested roots many are probably dubious. (Sometimes even for otherwise clear ones. I would for example not accept Albanian ka 'ox' as a probative reflex of *kes-.)

But then there is still Albanian kezë 'woman’s head-dress, bonnet, hair-net' which is directly cognate to Balto-Slavic *kasā́ˀ 'braided hair', so that is not even a 'one-off nominalization'. And if you read the introduction of NIL you would know that it is far from comprehensive. In other words, there are serious gaps in the data set you are using. Which simply means that you can't draw any conclusions from the absence of anything in your data set.

[Vijay]

This talk of satemization has me wondering: Is anyone familiar with a language called Bangani, spoken in Garhwal up in the Himalayas?

[jal]

But then there is still Albanian kezë 'woman’s head-dress, bonnet, hair-net' which is directly cognate to Balto-Slavic *kasā́ˀ 'braided hair'

What's a "direct cognate"? And why would it be one? To me it seems you're drawing unjustified conclusions a bit too quick here.


JAL

[Frislander]

I really doubt that, for instance, Balto-Slavic and Indo-Iranian developed satemisation at the same time and have a common ancestor.

I don't think they necessarily had a common ancestor "proto-satem", the sound change (and also the RUKI rule) could have spread by areal diffusion, just that at the time of PIE these dialects would have co-existed with centum dialects. The current models do seem to assume some areal diffusion of these sound changes anyway, I just reckon that this kind of diffusion would only really be feasible during or even before PIE itself, when the language was still relatively small and the dialect groups that would become the primary branches were still geographically close together and major areal contact between them was still a possibility.

My main objection to the current palato-velar reconstruction is a phonetic one; simply that no branch preserves the "palato-velars" as palatalised velars or true palatals - all the branches show them minimally as sibilant affricates, and most have further lenited them to fricatives (Sanskrit doesn't count simply because the sound values are uncertain at best, and again as no living Indo-Aryan languages shows true palatals I'm also dis-inclined to reconstruct them as such). This indicates to me that the palatalisation change must have occurred early and produced these affricates very quickly, else we would expect to see at least one branch preserve them as palatals. From there the idea of a "centumisation" seems much less credible, because palatal to velar shifts may be sparsely attested but affricate to velar shifts are totally unknown.

This talk of satemization has me wondering: Is anyone familiar with a language called Bangani, spoken in Garhwal up in the Himalayas?

I am not, why do you mention it?

[mèþru]

Well pronunciation of Sanskrit is well preserved due to folks like Pāṇini and religious tradition.

[Vijay]

This talk of satemization has me wondering: Is anyone familiar with a language called Bangani, spoken in Garhwal up in the Himalayas?

I am not, why do you mention it?

It's been argued to be a centum language.

[WeepingElf]

I really doubt that, for instance, Balto-Slavic and Indo-Iranian developed satemisation at the same time and have a common ancestor.

I don't think they necessarily had a common ancestor "proto-satem", the sound change (and also the RUKI rule) could have spread by areal diffusion, just that at the time of PIE these dialects would have co-existed with centum dialects. The current models do seem to assume some areal diffusion of these sound changes anyway, I just reckon that this kind of diffusion would only really be feasible during or even before PIE itself, when the language was still relatively small and the dialect groups that would become the primary branches were still geographically close together and major areal contact between them was still a possibility.

(Non-Anatolian) IE may have been a dialect continuum until as late as 2000 BC. In this continuum, what are now the branches of IE would have been dialect groups, and innovations such as satemization with RUKI spread from innovation centers across the continuum. The result would have been the overlapping isoglosses we observe in the family today.

[Tropylium]

And if you read the introduction of NIL you would know that it is far from comprehensive.

In other words, there are serious gaps in the data set you are using. Which simply means that you can't draw any conclusions from the absence of anything in your data set.

No, statistical discrepancies can be used to draw conclusions just fine (they're not the exact same thing as "gaps"). Think of the LIV+NIL data as an available set of measurements: they're not perfectly comprehensive, but they don't need to be, they only need to be representative.

Adding in all nominalizations not in NIL probably would add e.g. a few more examples of *g in Anatolian; but, then, it would surely also add more examples of *ǵ, and we'd be still in a situation where *g is not just rarer than *ǵ, but also even more rarely distributed. The latter is the key fact: a priori we don't expect words/roots with rare phonemes (e.g. *g) to be any less likely to be reflected in daughter languages. If there are, say, 120 reconstructions with palatovelars and 40 of these have reflexes in a daughter language group, then given 30 reconstructions of plain velars, we'd still expect to see about 10 reflexes, not a mere 2-3.

Why is it important to look at widely-attested cases?

Because word survival is not completely random: core vocabulary changes slower. You can also do the Swadesh route and pre-define a list of core vocabulary, but I would lean towards the real width of attestation and use that to define what's the core vocabulary specifically in IE. (There is a cultural aspect: "cow" is core vocabulary in IE, "arrow" is core vocabulary in Uralic, the inverse does not hold.) In either case, it will be reasonable to assume that this core vocabulary is also likely to be old within the pre-PIE period. Therefore all statistical internal reconstruction is better based on such data, not on the whole comparative material, which is otherwise bound to include also recent and even post-proto-language innovations.

For example, *ḱenk- "hang" is only in Germanic, Latin, Sanskrit, and Hittite, but its stem-final velar can't be secondary, nor does pre-PIE loaning seem particularly likely.

"Can't" is a bit strong: I could e.g. suggest that actually original *Ḱ… Ḱ dissimilates in satem languages, giving *ć…k rather than *ć…ć.

Or maybe more profitably, that *-k- is a root extension: it's one of the more common extensions even in attested cases, and *-RK- roots are actually more common overall than *-RḰ- roots (*-Rk- 43, *-Rg- 29, *-Rgʰ- 18; *-Rḱ- 22, *-Rǵ- 12, *-Rǵʰ- 18, *-Rḱ-). (Though this could also mean that there has been some kind of a root-final neutralization process involved.)

[KathTheDragon]

You conveniently neglect the root constraint against identical consonants, according to which *ḱenḱ- is impossible.

[Howl]

And if you read the introduction of NIL you would know that it is far from comprehensive.
In other words, there are serious gaps in the data set you are using. Which simply means that you can't draw any conclusions from the absence of anything in your data set.

No, statistical discrepancies can be used to draw conclusions just fine (they're not the exact same thing as "gaps"). Think of the LIV+NIL data as an available set of measurements: they're not perfectly comprehensive, but they don't need to be, they only need to be representative.

So let's look at your statistics:
"The median distribution for roots beginning with clear *g- is 3 subgroups. For *ǵ it's median 4. For *gʷ, median still 3."
That does not convince me of anything.

And the rest, like the maximum number of 'sub-groups', are outliers. And data gaps do matter for those.

But then there is still Albanian kezë 'woman’s head-dress, bonnet, hair-net' which is directly cognate to Balto-Slavic *kasā́ˀ 'braided hair'

What's a "direct cognate"? And why would it be one? To me it seems you're drawing unjustified conclusions a bit too quick here.

Both words seem to come from PIE *koseh₂ 'hairdo' [A concise historical grammar of the Albanian language, page 61]. So it is not even a form that is restricted to one branch. Which is what I guess Tropylium means with 'one-off nominalizations'.

[jal]

Well pronunciation of Sanskrit is well preserved due to folks like Pāṇini and religious tradition.

I'm pretty sure that when the same argument would be used for Latin, its pronunciation would be reconstructed quite differently than we currently do.


JAL

[mèþru]

Actually, Latin pronunciation was written about, but there wasn't much of an effort to stick to the original pronunciations as Romance dialects evolved. Hindus, on the other hand, felt that it was important to preserve the original pronunciation instead of reading texts with the same pronunciation as the local languages.

[Vijay]

Hmm, IME most Hindus pronounce Sanskrit with the same pronunciation as their native languages.

[mèþru]

Maybe it's specific to those who serve as religious leaders?

[Vijay]

Maybe even more specific to people who know better. Some religious leaders actually know their shit, others just like to show off. Like in any other religion, I suppose.

My sister-in-law is Hindu, so she and my brother had a Hindu wedding. I'd say the priest who officiated their wedding pronounced Sanskrit fairly accurately (at least compared to the other priest I'm about to talk about...). Within a few years after that, her brother got married. The priest at his wedding was terrible, a really arrogant guy who said he'd try to keep the length of the wedding down to only two hours and also made the fire alarm go off. At one point, he was discussing the pronunciation of the Sanskrit word yajña, i.e. the sacrificial offering in front of a sacred fire performed at weddings (among other ceremonies). He said something like "we [North Indians] say [jəgj], but South Indians say [jəd͡ʒj]. South Indians have the best pronunciation of Sanskrit." (Yes he did actually make word-final [gj] and [d͡ʒj]). Then he said, "Hail to them!" (this part I remember verbatim).

As flattering as that was, in reality, we say neither of those things but rather [jəd̚ˈɲa]. Somehow, though, I doubt any of these was really the original pronunciation of that word.

[mae]

-