Zompist Bboard Again

bradrn wrote: ↑Mon Oct 14, 2019 12:36 am Do you have any idea what the sound changes are? I’d like to see how those are related!

I have that info! See Jacques Guy's explanation (under Austronesian). (He compares it to Shark Bay rather than Tolomako.)

zompist wrote: ↑Mon Oct 14, 2019 1:13 am

bradrn wrote: ↑Mon Oct 14, 2019 12:36 am Do you have any idea what the sound changes are? I’d like to see how those are related!

I have that info! See Jacques Guy's explanation (under Austronesian). (He compares it to Shark Bay rather than Tolomako.)

Interesting! So it looks like Sakao has lenited all its consonants and completely overhauled its vowel system, which explains the cognates well. Also, that’s an excellent overview of language families; I’m not sure how I haven’t seen it before… This probably isn’t the place for it, but reading through it I do have a couple of questions/critiques/omissions:

• I can’t see any mention of the other branches of Afro-Asiatic; you only mention Semitic.
• Khoisan isn’t actually a language family, but you don’t mention that.
• I thought Niger-Congo was pretty well-established by now.
• Similarly, I thought that Altaic had been pretty thoroughly debunked.
• I thought that Pama-Nyungan was also pretty well-established, but you don’t mention it in the Australian section.

Keep in mind that I’m not an expert on this topic by any means, and in fact learnt most of the above from Wikipedia, so I’m very probably wrong in the above. This thread also probably isn’t the best place to discuss this, so I’m happy to discuss this in a new thread if you want me to.

That's not actually an intro to language families... it's a "fun facts" page to go along with my numbers list.

Niger-Congo is not (unless things have changed in the last 20 years) actually reconstructed. It's based on Greenberg's mass comparison-- which Africanists more or less accepted, but Americanists hate.

zompist wrote: ↑Mon Oct 14, 2019 3:56 am That's not actually an intro to language families... it's a "fun facts" page to go along with my numbers list.

Thanks for the clarification! But it certainly can act as an intro as well…

Niger-Congo is not (unless things have changed in the last 20 years) actually reconstructed. It's based on Greenberg's mass comparison-- which Africanists more or less accepted, but Americanists hate.

Interesting! Looking into this a bit further, it seems that Niger-Congo consists of Atlantic-Congo (which comprises the vast majority of languages in it) plus Dogon, Mande, Ijoid, Kordofanian and a few others. Wikipedia and Glottolog both imply that Atlantic-Congo is fairly widely accepted, which would mean that even if Niger-Congo is false, then most of the languages within it are still related.

Also, isn’t ‘mass comparison’ the procedure Greenberg used to ‘reconstruct’ proto-World? Or is that something different? (That’s also one of my favourite articles of yours, by the way; my favourite is the one on Mangling Foreign Dialects, which I feel obligated to mention here. If anyone here hasn’t read that article, you should do it now!)

bradrn wrote: ↑Mon Oct 14, 2019 4:45 am Interesting! Looking into this a bit further, it seems that Niger-Congo consists of Atlantic-Congo (which comprises the vast majority of languages in it) plus Dogon, Mande, Ijoid, Kordofanian and a few others. Wikipedia and Glottolog both imply that Atlantic-Congo is fairly widely accepted, which would mean that even if Niger-Congo is false, then most of the languages within it are still related.

This is also my understanding, with the addition that recent thinking seems to be that there's no good reason to class Dogon, Mande, Ijoid, and Kordofanian languages with Atlantic- (or Niger-)Congo. Güldemann's `chapter' (it's almost 400 pages long) in The Languages and Linguistics of Africa (2018), which he edited, gives arguments to this conclusion.

Although I'm not sure that's a universal thought as yet; it's also possible that some of these languages are related and others aren't (I've seen someone argue that Mande is part of the family, but that Dogon isn't).

Note, though, that 'Kordofanian' is an areal term. Specifically, the Talodi-Heiban languages are generally considered Atlantic-Congo (i.e. they seem to have inherited noun classes), while Katla and Rashad are outside Atlantic-Congo and may not be Niger-Congo at all.

That's fair.

bradrn wrote: ↑Mon Oct 14, 2019 4:45 amAlso, isn’t ‘mass comparison’ the procedure Greenberg used to ‘reconstruct’ proto-World? Or is that something different? (That’s also one of my favourite articles of yours, by the way; my favourite is the one on Mangling Foreign Dialects, which I feel obligated to mention here. If anyone here hasn’t read that article, you should do it now!)

"Mass comparison" is just the process of looking at a large number of languages and grouping them based on obvious similarities.

It's uncontroversial as the first step in analyzing families. After all, all existing families were found when people noticed large numbers of similar words. Where Greenberg and Ruhlen get controversial is when they argue that the comparative method is unnecessary, declaring that all you need is mass comparison to "validate" a family.

Greenberg applied his methods to African languages in the 50s and 60s. His classification is widely accepted, not least because it was far better than previous attempts. He used the same methods on Amerindian languages in the 70s and 80s, with far less happy results. (One reason is surely that the Amerindian languages are just far more various and just don't fit together nicely.)

In neither case did he provide reconstructions based on the comparative method, which is why I say that Niger-Congo is not proven. I'm not sure what "Atlantic-Congo" is, but I'd note that given adam's summary, Güldemann's classification is quite different from Greenberg's!

(Also, the "vast majority" of Niger-Congo languages is not too relevant— half of Niger-Congo consists of the Bantu languauges, which are clearly related, but that does not help validate the higher-order groupings.)

The "Proto-World" stuff (~ 1990) is an even more fanciful stage, and captured a lot of media attention. It's about at this stage that books on historical linguistics have to come with chapters explicitly criticizing Greenberg...

zompist wrote: ↑Mon Oct 14, 2019 6:55 pm I'm not sure what "Atlantic-Congo" is, but I'd note that given adam's summary, Güldemann's classification is quite different from Greenberg's!

Summary version: Atlantic-Congo is everything with noun classes that doesn't look like it recently borrowed its noun classes. Or in other words, it's Greenberg's Niger-Congo, but minus Mande, Ijo (with or without Defaka), Dogon, Katla, Rashad (which has noun classes, but seemingly borrowed), and Kadu. Niger-Congo today is basically Greenberg's Niger-Congo minus Kadu, and possibly a few other isolates and oddities.

So, my layman's impression is that Greenberg's Niger-Congo has held up reasonably, although my impression is that Dogon is now thought to probably not be related, and that the other non-Atlantic-Congo languages may not be either. There are also some individual languages or branches that people have questions about, which may be Niger-Congo-ized independent families or may not.

His internal structure for Niger-Congo seems to have held up a bit less well, but still not that terribly.

(Also, the "vast majority" of Niger-Congo languages is not too relevant— half of Niger-Congo consists of the Bantu languauges, which are clearly related, but that does not help validate the higher-order groupings.)

Actually, apparently some people now doubt that Bantu (in the traditional sense) is actually a grouping at all. It's doubtful whether a Proto-Bantu can be reconstructed that includes all the divergent Bantu languages without also including a bunch of Bantoid languages.


I think a huge problem for Niger-Congo is its size, and in particular the size of Bantu. A real Proto-Bantu (or Proto-Bantoid, or whatever turns out to be reconstructable) would be hugely important - with that, you could probably construct a decent Proto-Benue-Congo, at which point you'd be in a reasonable position to sort through the mess in Kwa/Volta-Niger, and to a lesser extent then in 'North Volta' (i.e. the tricky stuff), working out what's related and how much. Proto-Bantu would be like having Sanskrit. But the problem is, there are SO MANY Bantu languages, many of which are so little studied, that it's going to take a lot of work just to get the internal classifications in Bantoid sorted out well enough to enable a real Proto-Bantu reconstruction...

OK, here's a representation of Greenberg's Niger-Congo (as revised by Ruhlen). The red branches are those that are removed from Atlantic-Congo.

It's really held up even worse than the diagram indicates, because so far as I can see Africanists no longer accept West Atlantic as a grouping, and reject the top-level groupings ("Central Niger-Congo", divided into North and South branches).

Caddoan guy, stop it, we all know oligosynthesis is unnatural so stop acting like it's a sensible way to build a language family. Also did you even start from an actual proto-lang or did you just cobble together each language from similar base components of morphology and syntactic structure and make the rest up new each time?

Frislander wrote: ↑Wed Oct 16, 2019 11:14 amCaddoan guy, stop it, we all know oligosynthesis is unnatural so stop acting like it's a sensible way to build a language family. Also did you even start from an actual proto-lang or did you just cobble together each language from similar base components of morphology and syntactic structure and make the rest up new each time?

The phonotactics of Wichita in particular are also curious, with many voiceless stops (including glottal stops), compared to the more normal-souding Arikara and Caddo.

čukú načootšáka ? nuuháʔi̥ tičootšáka, nikuteekḁhčiišawataáʔḁ. [Arikara]
'Where is the church? The church is over there; it's the white building.'

šuxwaákoʔ san̥i̥štaakḁhníniʔ [Arikara]
'Say it in English.'

háhʔánássaʔ háhʔánkisaʔ [Caddo]
'He's cold while standing.'

kúsidí·, haʔímáy hússayahʔniʔaʔ ! náttiʔ dikahí·yah, kúyatáyánnáhduʔ natdahʔní·way. [Caddo]
'After a while, there was a big log lying there. There he [Turtle] got stuck, as he couldn't get by it when he climbed up the log.'

kiyari·cé·hire·weʔekih [Caddo]
'God' (literally, "the one in charge of everything")

niya·hkʷírih tìkìteʔècaki [Wichita]
'He's sitting on top of the tree.'

kà·ʔà·ʔà·kó·khá·rʔa [Wichita]
'Once upon a time, ...' (introduces a story, literally "long ago it happened that...")

híraciya·ká·hí·kʔih neʔerʔíriwah [Wichita]
'Mother Earth' (addressee of prayers before eating, literally "our mother who is lying down")

tiʔikhánthirisʔih [Wichita]
'yesterday'

ki·ckharikwitath wéraʔ hi·ckwakhari hára·sʔe·ʔiyákiciháka·rʔistiri·sa [Wichita]
'When they went to Round-Tops [the city of Wichita, Kansas], they planted their gardens there too.'


That said, I do highly approve of the names of the days of the week in Wichita.

ná·sa·khíʔinnih 'Sunday' ("when it's His day")
kíriwaré·sa·khíʔinnih 'Monday' ("when it's not His day anymore")
wicha kínné:sa·khíʔinnih 'Tuesday' ("when it's not His day again")
nackháti·kih 'Wednesday' ("the middle day")
ní·cʔarhiʔirhé:sishah 'Thursday' ("the day they go to stay overnight")
acs nasa·khaskhírih 'Friday' ("when a good day is going by")
wa·khácsa·khirʔa 'Saturday' ("cow day")

I will say that I do not believe the reports of this-and-that South American language having /t̪ʙ̥/ as a phoneme. I'm not saying that as a joke, either: like "haha that is so outrageous, anadew amirite?". I am being serious: I literally do not believe it. Napoleon Chagnon in one of his books talks about an entire village making up fake and lewd-sounding names for each member of their tribe as a prank to pull on Chagnon, and sustaining that prank for months without anyone ever breaking character. Methinks the [t̪ʙ̥] business is another example of Amazonian tribal humor.

Frislander wrote: ↑Wed Oct 16, 2019 11:14 am Caddoan guy, stop it, we all know oligosynthesis is unnatural so stop acting like it's a sensible way to build a language family. Also did you even start from an actual proto-lang or did you just cobble together each language from similar base components of morphology and syntactic structure and make the rest up new each time?

Mind linking to a lexicon?

Frislander wrote: ↑Wed Oct 16, 2019 11:14 am Caddoan guy, stop it, we all know oligosynthesis is unnatural so stop acting like it's a sensible way to build a language family. Also did you even start from an actual proto-lang or did you just cobble together each language from similar base components of morphology and syntactic structure and make the rest up new each time?

I thought there were no oligosynthetic natlangs? (The only proposed one I’m aware of is Navajo, and I thought that was debunked.)

Ser wrote: ↑Wed Oct 16, 2019 2:33 pm The phonotactics of Wichita in particular are also curious, with many voiceless stops (including glottal stops), compared to the more normal-souding Arikara and Caddo.

The samples you’ve posted don’t seem too unusual to me.

That said, I do highly approve of the names of the days of the week in Wichita.

I agree — those are excellent names! Maybe I can steal some of them if I ever get to that stage in a conlang…

Xephyr wrote: ↑Wed Oct 16, 2019 2:45 pm I will say that I do not believe the reports of this-and-that South American language having /t̪ʙ̥/ as a phoneme. I'm not saying that as a joke, either: like "haha that is so outrageous, anadew amirite?". I am being serious: I literally do not believe it.

I can believe it all too easily. I find /t̪ʙ̥/ to be actually quite easy to pronounce compared to some other natlang phonemes: the various pharyngeals and epiglottals, for instance, or /ɺ/, or tenuis clicks, or ejectives (in the sense that I managed to get /t̪ʙ̥/ right first time but it took a while to figure out how to pronounce those other consonants). Compared to those, /t̪ʙ̥/ is so easy to pronounce that Iʼm surprised there aren’t more languages which use it.

Looking further into this, Wikipedia lists four languages with /t̪ʙ̥/: Wariʼ, Oro Win, Pirahã, and Sangtam. Comparing each language:

• Wariʼ seems to mainly use it for onomatopoeic words, which is understandable.
• Wikipedia doesn’t have much information on Oro Win, but it’s closely related to Wariʼ.
• Pirahã is suspicious: According to Wikipedia, Everett states he only discovered the phoneme in 2004. However there is no citation given, so this claim does seem suspicious.
• Sangtam is spoken in India, unlike the other languages. Wikipedia doesn’t have much information on it, but states that it has /t̪ʙ̥/ in both aspirated and unaspirated versions.

So it seems that even though the claim of /t̪ʙ̥/ in Pirahã is debatable, there is enough other evidence to conclude that the phoneme is real.

Napoleon Chagnon in one of his books talks about an entire village making up fake and lewd-sounding names for each member of their tribe as a prank to pull on Chagnon, and sustaining that prank for months without anyone ever breaking character.

Now this is something I don’t believe. Are you sure he wasn’t making that up?

Zju wrote: ↑Wed Oct 16, 2019 3:18 pm

Frislander wrote: ↑Wed Oct 16, 2019 11:14 am Caddoan guy, stop it, we all know oligosynthesis is unnatural so stop acting like it's a sensible way to build a language family. Also did you even start from an actual proto-lang or did you just cobble together each language from similar base components of morphology and syntactic structure and make the rest up new each time?

Mind linking to a lexicon?

Sure, here's a site that does a dictionary of Pawnee and Arikara.

bradrn wrote: ↑Wed Oct 16, 2019 5:49 pm

Frislander wrote: ↑Wed Oct 16, 2019 11:14 am Caddoan guy, stop it, we all know oligosynthesis is unnatural so stop acting like it's a sensible way to build a language family. Also did you even start from an actual proto-lang or did you just cobble together each language from similar base components of morphology and syntactic structure and make the rest up new each time?

I thought there were no oligosynthetic natlangs? (The only proposed one I’m aware of is Navajo, and I thought that was debunked.)

Well firstly you sure you're not getting confused with Blackfoot, which is the one I've seen argued be this? (if anything Navajo is actually less synthetic than even other parts of Athabaskan). Also I will admit to exxaggerating a little, but I will also argue that if any language family comes close to oligosynthesis it's Caddoan.

Well firstly you sure you're not getting confused with Blackfoot, which is the one I've seen argued be this?

No; I was actually getting confused with Nahuatl. (See https://en.wikipedia.org/wiki/Oligosynthetic_language, which lists Nahuatl as an example.)

Also I will admit to exxaggerating a little, but I will also argue that if any language family comes close to oligosynthesis it's Caddoan.

After seeing that dictionary, I must agree. But I do still have a query about this: Oligosynthesis is usually defined as creating new words by combining pre-existing morphemes with a specific meaning. Caddoan is certainly reminiscent of this: for instance, the Sikri Pawnee dictionary you’ve linked lists aciks ‘think, feel’, aciksaar ‘to think of’, aciksasaah ‘to think about’, aciksawaaks ‘feel relieved’ etc. But the individual ‘morphemes’ aar and asaah don’t seem to have any specific meaning of their own, and in fact don’t seem to have any specific meaning when used with other morphemes as well. (awaaks seems to be unusual in that it does have a relevant, independent meaning of its own.) So Sikri Pawnee at least cannot be considered a ‘true’ oligosynthetic language, but merely one with a lot of repeated sounds in similar words. (I’m hesitant to even call it compounding, as the ‘morphemes’ taking part in the compounding don’t seem to have any particular meaning on their own.)

Frislander wrote: ↑Tue Jul 23, 2019 3:58 am Dear creator of Japanese, what the heck's going one with your vowel devoicing process? Everyone else just uses it on all vowels regardless of quality, but you've linked it to height of all things! What on earth is it about high vowels that makes them more prone to voicelessness than other vowels I ask you! I certainly can't think of anything.

Something similar must have happened in Proto-Slavic, where Balto-Slavic /i/ /u/ became reduced vowels (written ь, ъ), still attested in Old Church Slavic and Old Russian, which in the later stages of the Slavic languages were either dropped or fortified to full vowels, depending on position and environment, with different results in the individual languages.

It's not that weird for high vowels to end up semi-voiceless or fricativized--think of all the spirantization that occurs before them; it's a short leap from there.

A couple examples: the Bantu high vowels *i *u caused fricativization on a preceding consonant in many instances; in many of the most closely related languages of Cameroon their cognate segments are apical or syllabic fricatives. And the Chinese province of Sichuan, which is [sɨʈʂʰwan] (tones ignored) in Beijing newscaster standard, is [sz̩tsʰwã] to the locals.

hwhatting wrote: ↑Fri Oct 18, 2019 9:12 am

Frislander wrote: ↑Tue Jul 23, 2019 3:58 am Dear creator of Japanese, what the heck's going one with your vowel devoicing process? Everyone else just uses it on all vowels regardless of quality, but you've linked it to height of all things! What on earth is it about high vowels that makes them more prone to voicelessness than other vowels I ask you! I certainly can't think of anything.

Something similar must have happened in Proto-Slavic, where Balto-Slavic /i/ /u/ became reduced vowels (written ь, ъ), still attested in Old Church Slavic and Old Russian, which in the later stages of the Slavic languages were either dropped or fortified to full vowels, depending on position and environment, with different results in the individual languages.

In (Old) English, high were dropped entirely in prosodically weak contexts (after a heavy syllable, or after a sequence of two light syllables), while mid and low vowels were (at that point) retained. That would seem to suggest a similar process.