The Great Proto-Indo-European Thread's Sequel

[Travis B.]

Middle Chinese *ljuwk

On a totally unrelated note, I am always amused by how almost every word in MC contains a <j> in its onset (despite there being no evidence of a palatal glide in most cases in reconstructed PTB/PST).

As I recall (possibly wrongly), MC syllables are reconstructed in two classes, and the one class is conventionally written as palatalised. That wasn’t necessarily how they were pronounced though.

Yes, the MC 'type A' versus 'type B' distinction, where the (more common) 'type B' syllables are conventionally written as palatalized. Of course, while older OC reconstructions carry this back into OC, Baxter-Sagart found the idea that most OC syllables were palatalized to be silly, hence their introduction of reconstructed pharyngealization of 'type A' syllables to explain why the 'unpalatalized' 'type A' syllable were less common than the 'palatalized' 'type B' syllables.

(And there is good reason to believe that the idea that OC 'type B' syllables were palatalized is silly, not the least because loanwords into Sinitic from other languages typically were marked with 'type B' syllables despite not being palatalized in their source languages.)

[Richard W]

I'd contest the idea that numbers are hard to reconstruct. Bantu, Austonesian, Afro-Asiatic, Dravidian, Algonquian, Mayan all seem pretty solid.

These are generally accepted as families rather than mere phyla. Afro-Asiatic is the most uncertain case (though I think it admits 35 content words). But what are the reconstructed Afro-Asiatic numerals?

I notice you said 'Bantu', and not 'Niger-Congo'. But perhaps Niger-Congo is not as disparate as it at first looks.

From my numbers list, I'd suggest that the notable fact is not how often numbers are borrowed, but how little. The above sets are the exception, not the rule. Historically, people don't throw out their number sets when they're conquered.

And there's an organisational development - having enough of a state organisation to be conquered.

[Travis B.]

I notice you said 'Bantu', and not 'Niger-Congo'. But perhaps Niger-Congo is not as disparate as it at first looks.

There is reason to doubt 'Niger-Congo' as a phylum, particularly since there is good reason to suspect Mande might be an independent family ─ but if one restricts oneself to Atlantic-Congo then one is on much more solid ground.

[zompist]

I'd contest the idea that numbers are hard to reconstruct. Bantu, Austonesian, Afro-Asiatic, Dravidian, Algonquian, Mayan all seem pretty solid.

These are generally accepted as families rather than mere phyla. Afro-Asiatic is the most uncertain case (though I think it admits 35 content words). But what are the reconstructed Afro-Asiatic numerals?

I notice you said 'Bantu', and not 'Niger-Congo'. But perhaps Niger-Congo is not as disparate as it at first looks.

I may well be out of date-- when I was collecting numbers there were reconstruction of Bantu, but not Niger-Congo.

Some of those big groupings are based only on mass comparison and may not hold up. Lyle Campbell has some very mean things to say abut Greenberg's work in the Americas.

[WeepingElf]

Oh brother, not Octaviano again...

My name here is Talskubilos.

Yeah, Octaviano.

Let's be fair to him and (1) don't call him Octaviano if he doesn't want to be called that anymore, (2) don't nail him down on hypotheses he no longer maintains and (3) treat him as a sincere person. He doesn't offend anyone; he just maintains ideas most of us disagree with, but it is just linguistics and nothing to be upset about.

[Richard W]

Let's be fair to him and (1) don't call him Octaviano if he doesn't want to be called that anymore,

There is a legal maxim that a man's name is what he is known by.

[Talskubilos]

Sino-Tibetanists happily reconstruct the numbers. I suppose Classical Tibetan *drug , Sunwar ruku, and Written Burmese k’rok could be from Middle Chinese *ljuwk, but they're explained better by proto-Sino-Tibetan *t-r’uk.

This numeral has parallels in NEC and appears to be derived from a lexeme 'hand' found in Yenissean *ŕɔŋ and Burushaki -reŋ.

[Richard W]

Sino-Tibetanists happily reconstruct the numbers. I suppose Classical Tibetan *drug , Sunwar ruku, and Written Burmese k’rok could be from Middle Chinese *ljuwk, but they're explained better by proto-Sino-Tibetan *t-r’uk.

This numeral has parallels in NEC and appears to be derived from a lexeme 'hand' found in Yenissean *ŕɔŋ and Burushaki -reŋ.

Which looks rather like Balto-Slavonic *ránkā 'hand, arm', which has possible connections in other parts of Indo-European, but with initial cluster wr-. Alas, Bayesianly speaking, this connection has a good chance of being wrong, which itself is plausibly connected to the Balto-Slavonic word.

[Talskubilos]

Sino-Tibetanists happily reconstruct the numbers. I suppose Classical Tibetan *drug , Sunwar ruku, and Written Burmese k’rok could be from Middle Chinese *ljuwk, but they're explained better by proto-Sino-Tibetan *t-r’uk.

This numeral has parallels in NEC and appears to be derived from a lexeme 'hand' found in Yenissean *ŕɔŋ and Burushaki -reŋ.

Which looks rather like Balto-Slavonic *ránkā 'hand, arm', which has possible connections in other parts of Indo-European, but with initial cluster wr-. Alas, Bayesianly speaking, this connection has a good chance of being wrong, which itself is plausibly connected to the Balto-Slavonic word.

Actually, Starostin's reconstruction of NEC 'six' has the structure HRVNCV, where C stands for a lateral affricate, which behaves much like the classical PIE palato-velars, because it's reflected as a velar stop in some branches but a lateral in others. Starostin's PNC looks much older than commonly though and thus closer to the purported Proto-Sino-Caucasian (aka Dene-Caucasian), which I regard as a source of Wanderwörter rather than a macro-family.

As for the connection between Balto-Slavic 'hand' and English wrong, I think this is precisely what happened with the Iberian anthroponym formant iskeŕ (with a fossilized demonstrative is), which I translate as 'hand' and is the likely source of Spanish izquierda, Gallego-Portuguese esquerda, Catalan esquerra and Basque ezker 'left (hand)'.

[WeepingElf]

I'd contest the idea that numbers are hard to reconstruct. Bantu, Austonesian, Afro-Asiatic, Dravidian, Algonquian, Mayan all seem pretty solid.

These are generally accepted as families rather than mere phyla. Afro-Asiatic is the most uncertain case (though I think it admits 35 content words). But what are the reconstructed Afro-Asiatic numerals?

I notice you said 'Bantu', and not 'Niger-Congo'. But perhaps Niger-Congo is not as disparate as it at first looks.

I may well be out of date-- when I was collecting numbers there were reconstruction of Bantu, but not Niger-Congo.

Some of those big groupings are based only on mass comparison and may not hold up. Lyle Campbell has some very mean things to say abut Greenberg's work in the Americas.

Mass comparison confuses asking the question with finding the answer. What those lists say is not, "These languages are related", they say, "Are these languages related?". They stop where real comparative work begins; and many armchair linguists do the same. Of course, asking the right question is the first step towards finding the answer, and in some cases, Greenberg may have asked the right question, but he did not really give us answers, and as long as we don't have an answer, we don't know whether the question was the right one.

Amerind is widely rejected for these reasons, but also because it is useless for classification purposes: it just shoves all indigenous American languages except the two northernmost families into one big sack. Also, it presupposes the "Clovis first" theory, which since then has been falsified. Greenberg's African language classification algorithm is a bit more useful. It seems to work as follows:

1. If the language appears to share morphology with Semitic, it goes to Afroasiatic.
2. If the language appears to share morphology with Bantu, it goes to Niger-Congo.
3. If neither, and the language has clicks, it goes to Khoisan.
4. If neither, and the language doesn't have clicks, it goes to Nilo-Saharan.

I don't know whether Proto-Niger-Congo has been reconstructed by now; for Proto-Afroasiatic, I have heard of two reconstructions that differ in a number of points, so they can't both be entirely correct, which probably means that both contain errors. Which, of course, doesn't mean that the language family is unreal; Bopp's, Schleicher's and Brugmann's reconstructions of PIE also erred in some points, and the currently accepted reconstruction probably isn't the last word either, yet it is wide accepted that the family is real.

[bradrn]

for Proto-Afroasiatic, I have heard of two reconstructions that differ in a number of points, so they can't both be entirely correct, which probably means that both contain errors.

I recall a paper which convincingly argued that both of those ‘reconstructions’ are essentially worthless. (I remember it because it cited zompist’s page on chance resemblances to argue so!)

Which, of course, doesn't mean that the language family is unreal; Bopp's, Schleicher's and Brugmann's reconstructions of PIE also erred in some points, and the currently accepted reconstruction probably isn't the last word either, yet it is wide accepted that the family is real.

I’ve heard more than one person express skepticism that Afroasiatic is a real family, even excluding Omotic (which may not even be monophyletic).

[Travis B.]

Even Wikipedia does not appear to accept 'Niger-Congo' as actually a demonstrated node, if one can speak of Wikipedia accepting anything, and Glottolog (mind you they are notorious splitters) does not accept it either. Atlantic-Congo, OTOH, is widely accepted.

[Travis B.]

for Proto-Afroasiatic, I have heard of two reconstructions that differ in a number of points, so they can't both be entirely correct, which probably means that both contain errors.

I recall a paper which convincingly argued that both of those ‘reconstructions’ are essentially worthless. (I remember it because it cited zompist’s page on chance resemblances to argue so!)

Which, of course, doesn't mean that the language family is unreal; Bopp's, Schleicher's and Brugmann's reconstructions of PIE also erred in some points, and the currently accepted reconstruction probably isn't the last word either, yet it is wide accepted that the family is real.

I’ve heard more than one person express skepticism that Afroasiatic is a real family, even excluding Omotic (which may not even be monophyletic).

About Afroasiatic, the main evidence for its cromulence is its morphology, with Proto-afroasiatic being at such a time depth that reconstructing really any of its lexicon is all but impossible.

[bradrn]

Glottolog (mind you they are notorious splitters)

For this reason my usual heuristic is, ‘if Glottolog lists it, it’s definitely a valid family’. (Though not necessarily the converse.)

About Afroasiatic, the main evidence for its cromulence is its morphology, with Proto-afroasiatic being at such a time depth that reconstructing really any of its lexicon is all but impossible.

Agreed. I personally find it convincing, at least minus Omotic.

[Talskubilos]

Mass comparison confuses asking the question with finding the answer. What those lists say is not, "These languages are related", they say, "Are these languages related?". They stop where real comparative work begins; and many armchair linguists do the same. Of course, asking the right question is the first step towards finding the answer, and in some cases, Greenberg may have asked the right question, but he did not really give us answers, and as long as we don't have an answer, we don't know whether the question was the right one.

Amerind is widely rejected for these reasons, but also because it is useless for classification purposes: it just shoves all indigenous American languages except the two northernmost families into one big sack. Also, it presupposes the "Clovis first" theory, which since then has been falsified. Greenberg's African language classification algorithm is a bit more useful. It seems to work as follows:

1. If the language appears to share morphology with Semitic, it goes to Afroasiatic.
2. If the language appears to share morphology with Bantu, it goes to Niger-Congo.
3. If neither, and the language has clicks, it goes to Khoisan.
4. If neither, and the language doesn't have clicks, it goes to Nilo-Saharan.

I don't know whether Proto-Niger-Congo has been reconstructed by now; for Proto-Afroasiatic, I have heard of two reconstructions that differ in a number of points, so they can't both be entirely correct, which probably means that both contain errors. Which, of course, doesn't mean that the language family is unreal; Bopp's, Schleicher's and Brugmann's reconstructions of PIE also erred in some points, and the currently accepted reconstruction probably isn't the last word either, yet it is wide accepted that the family is real.

IMHO, proto-languages are useful tools for comparative purposes, but they can't be mistaken for real languages actually spoken by people. In that sense, they're a particular class of conlangs.

[Ketsuban]

That's a difference more in attitude than anything else. Linguists recognise that Proto-Indo-European is a scholarly reconstruction which has limits (if every surviving branch regularised an irregularity in the same way, the reconstruction cannot recover the original irregularity) but are comfortable using "Proto-Indo-European speakers" to mean "speakers of the languages ancestral to the modern Indo-European family of which Proto-Indo-European is an attempt at reconstruction".

[Talskubilos]

That's a difference more in attitude than anything else. Linguists recognise that Proto-Indo-European is a scholarly reconstruction which has limits (if every surviving branch regularised an irregularity in the same way, the reconstruction cannot recover the original irregularity) but are comfortable using "Proto-Indo-European speakers" to mean "speakers of the languages ancestral to the modern Indo-European family of which Proto-Indo-European is an attempt at reconstruction".

In my opinion, among the +2,000 lexical items reconstructed for PIE there must be hundreds of borrowings of different chronologies (substrates, cultural loanwords, Wanderwörter, etc.), while the true native lexicon remains obscured. For example, both 'wolf' (e.g. Latin lupus) and 'fox' (e.g. Latin vulpes) share a nucleus *wl̥ with 'wool', suggesting they're derivated from a common lexeme referring to animal hair.

[bradrn]

In my opinion, among the +2,000 lexical items reconstructed for PIE there must be hundreds of borrowings of different chronologies (substrates, cultural loanwords, Wanderwörter, etc.), while the true native lexicon remains obscured.

I may regret asking this, but if all these things are not the ‘true native lexicon’, then what do you think is?

[Talskubilos]

In my opinion, among the +2,000 lexical items reconstructed for PIE there must be hundreds of borrowings of different chronologies (substrates, cultural loanwords, Wanderwörter, etc.), while the true native lexicon remains obscured.

I may regret asking this, but if all these things are not the ‘true native lexicon’, then what do you think is?

If we were able to identify and classify all those loanwords (by means of external comparison, internal analysis, etc.), the remainder must be the native lexicon, although there probably would be still a number of dubious items for lack of enough information. For example, some kindship words have a *-h₂tēr suffix attached to a lexeme which can be further compared (in the case of 'father' and 'mother' they're the nursery words pa/ma.)

[Richard W]

Also, it presupposes the "Clovis first" theory, which since then has been falsified.

You've got that the wrong way round. While Amerind may follow from Clovis first, it doesn't require it. The same linguistic results would follow (to first order) if the descendants of Clovis simply replaced everything else.