Ubykh: requiem and revival
« previous post | next post »
I begin with an e-mail from Martin Schwartz, sent to me on 3/14/16:
Last September in Istanbul a fair-haired academic there, a colleague of my wife, said she is of Çerkes background, and went on to say a relative of hers was the last Ubykh speaker. Dumêzil had been to her family's home, grouchy that there were apparently no Ubykh speakers to be found, when the Ubykh speaker knocked on the door….
A few words about who the Ubykh people are:
The Ubykh (Ubykh Circassian: пэху, туахы (tʷaχə), убых; Russian: убыхи; Turkish: Ubıhlar, Vubıhlar) are one of the twelve Adyghe (Circassian) tribes, representing one of the twelve stars on the green-and-gold Adyghe flag.
And about the language:
Ubykh, or Ubyx, is an extinct Northwest Caucasian language once spoken by the Ubykh people. The Ubykh language is ergative and polysynthetic, with a high degree of agglutination, with polypersonal verbal agreement and a very large number of distinct consonants but only two phonemically distinct vowels.
And here is an amazing story about the language (recounted by R. Fenwick):
Victor Mair: "When human beings hear others speaking but are unable to comprehend what is being said, to what do they compare such speech?" (first sentence here)
Along a similar vein is an old story about the languages of the Caucasus, which I understand to be probably Turkish or Persian in origin, but in the early 20th century was adapted and told to the German linguist Adolf Dirr specifically about (and in) Ubykh. I translate moderately freely here from the Ubykh text, narrated by an unknown speaker:
There once was a country in the east, and in that country there was a great ruler, a man of great intelligence and good heart, who had by his side a very well-read scribe. One day the sheikh said to his scribe: "You have learned much, even though you are still young; I will give you whatever money you need to travel around all the countries of this world, and learn the languages of all the people in them. Travel for five years, and then return to tell me about the languages you have learned."
That scribe got on his way, and went to each country, staying in those countries for five years altogether. At last he returned, adorned gorgeously in silks and velvets, carrying a sack on his back. He came up to the sheikh and bowed, and said, "I have learned all the languages of the world."
"Show me what you have learned!" replied the sheikh. And the scribe began to speak to him in Arabic, Turkish, Armenian, Greek, and all the other languages he had learned. Then the sheikh said, "That is wonderful, but what is inside the sack on your back?"
"Well," replied the scribe, "if I said I had spoken to you in all the languages there are, I would be deceiving you; there is still one more." And he took the sack from off his back, opened it, and poured out a stream of pebbles onto the floor.
"What is this?" asked the sheikh.
And the scribe replied, "This is the Ubykh language!"
From Peter B. Golden:
This is, allegedly, someone telling a story in Ubykh, and another.
Reminds me of my joys with Georgian (and Georgian has its full share of vowels, but with an ergative and all the other characteristics noted of Ubykh). Georgian belongs to the K’art’velian groupings of languages in the Caucasus along with Svan and Mingrelian (and in Turkey – the Laz). Ubykh is related to Circassian (Çerkez), the various subgroupings of which have a large number of consonants and very few vowels. Although there are periodic attempts to demonstrate that all the “Caucasian” languages (including the Nakh grouping [Chechen, Ingush, Batsbi] and the host of languages in Daghestan (Avar being the leading one, see here) are related, no one, as far as I know, has made the case. I think that they have given up on trying to prove a relationship with Basque, although more than a few attempts have been made.
From Stefan Georg (referring to the recordings provided by Peter Golden:
Yes, this is nice, but the first guy is not a native speaker, since the last one, Tevfik Esenç (who worked mostly with Dumézil, but also with Vogt and Charachidzé), passed away in 1992 – his last taped words in Ubykh were „Here ends the Ubykh language.“ The second sample, that’s him in person. Ubykh was the record-holder in the number of phonemic consonants (by some accounts 86), at least in Eurasia, if not the world. Nobody, apart from the usual trolls, defends „Iberocaucasian“ (i.e., all C. groups) as a genetic grouping today (it enjoyed some popularity in the USSR after Soviet science recovered from Marrism, but the defendants were few even then), and few (if any) outside of Moscow would defend „North-Caucasian“ today (North West + North East C.). WC is inspectionally obvious as a family, whereas East (Nakh + Daghestan) is less obvious, but demonstrably a family. Various attempts to join some languages of the Ancient Near East (usually Hurro-Urartian), though at times supported by Diakonoff, have been unsuccessful and have been (and if they haven’t been, they should be) abandoned (again, outside of Moscow, where nothing is ever abandoned). For completeness' sake, it should be added that another ANE lg., Hattic, attracted some literature which tried to show that it belongs to NWC, but the results of this were nothing to write home about. The number of clearly demonstrated and viable language families is three.
From Marcel Erdal:
Proto-Altaists and the like might spare a moment for the obviously adopted family name of Tevfik Esenç. It looks like a blend consisting of Old Turkic esen 'safe and sound' and enç 'at peace', a warning to those drawing far-reaching prehistoric conclusions based on proper names.
Proper name 'evidence' is crucial also for many of the entries in the 2011 "West Old Turkic" work.
From Stefan Georg:
His original family name was different, of course – Zays˚a.
Some advocates of Ubykh are trying to revive the language. My friend John Colarusso is one of them. I wish them well.
Readings
"GoFundMe for the Ubykh Dictionary Project" (3/5/18)
"Caucasian words for tea" (1/26/17)
"The languages of the Caucasus" (8/25/08)
"Obituary: The Ubykh Language" (1/25/12)
"Ur-etyma: how many are there?" (7/6/14)
Victor Mair said,
February 26, 2019 @ 8:07 pm
From Peter Golden:
There is a Çerkes “colony” in New Jersey, many of them immigrants from Turkey and Jordan (where the king’s bodyguard and not a few generals are Circassian, descendants of those who fled the Russian conquest in the 19th century). I have had a number of students of that background and most of them were still fluent in their respective Adyghe/Circassian languages, some 30 years ago. I don’t know what the situation is today. Some even brought me books (school books actually) to show that the language was still being actively taught within the community. This is a phenomenon that is quite common among immigrant groups. I know Poles, Estonians, Chinese and others who attended “national” schools on Saturdays or Sundays (e.g. the “Estonian School” to which my neighbors’ children dutifully went) or had classes in Polish in local parochial schools in predominantly Polish neighborhoods. The Ukrainian community stressed Українознавство in their schools. By the third generation in the US, the languages are usual vestigial at best, limited to a phrase or two, the names for certain foods and some curse words. Those of my Çerkes students from Turkey also spoke native Turkish. Of those from Jordan (and also several from Syria, especially the Golan Heights where there had been a Çerkes colony), the younger generation was comfortable in Arabic, but their parents, so they told me, less so. I have also had a number of students (from Syria, usually the city of Ma'lūlā) whose parents were still fluent in Aramaic, but that is another story.
Victor Mair said,
February 26, 2019 @ 8:22 pm
From Peter Golden:
NJ is an interesting state with pockets of everything. The son of the President of the Qarachay community (a Turkic people from the North Caucasus ) was a student of mine. I am still in contact with several of my Qarachay students, now middle aged, one a banker, the other recently retired from the UN. I also gave a special tutorial in the Georgian language to the granddaughter of Noe Zhordania (the last Menshevik President of Georgia), irony of ironies.
cameron said,
February 26, 2019 @ 8:45 pm
I heard a version of that story sbout the learned vizier traveling the whole kingdom and learning all the languages when I was a child. I heard it in Persian from my father. In the version that I heard, the end of the story has the vizier shaking some walnuts in a sack and declaring that that was the Semnani language.
I think the story is better suited to Ubykh. I don't think Semnani is particularly strange
Victor Mair said,
February 26, 2019 @ 9:00 pm
Ah, cameron, but it's wonderful that what's in the Semnani sack are walnuts, which have occupied us greatly since this post of yesterday:
"'Rondle it!'"
http://languagelog.ldc.upenn.edu/nll/?p=41940
Chris Button said,
February 26, 2019 @ 11:33 pm
It's funny how this has vexed linguists when on the surface Ubykh, like any other language, has plenty of vowels. It reminds me of how the two vowel analysis of Old Chinese is often challenged as "unnatural" when in reality it comes down to little more than properly distinguishing underlying phonology from surface phonetics.
Victor Mair said,
February 27, 2019 @ 12:02 am
I knew that Chris would be pleased by this post on Ubykh.
Philip Taylor said,
February 27, 2019 @ 12:54 am
Nothing to do with Ubykh per se but definitely to do with endangered languages. I had occasion to read an Australian Health Service article on Total Hip Replacement yesterday, and was delighted to see that it was available in a number of languages in addition to English. Delighted, that is, until I realised that not one of the languages into which it had been translated was an aboriginal Australian language …
Philip Taylor said,
February 27, 2019 @ 12:56 am
Sorry, the link appears to have become faulty in copying. It should read http://www.healthtranslations.vic.gov.au/bhcv2/bhcht.nsf/PresentDetail?Open&s=Patient_Guide_to_Total_Hip_Replacement_%28THR%29_Surgery or perhaps (with ampersand escaped) http://www.healthtranslations.vic.gov.au/bhcv2/bhcht.nsf/PresentDetail?Open&s=Patient_Guide_to_Total_Hip_Replacement_%28THR%29_Surgery
Bob Ladd said,
February 27, 2019 @ 2:06 am
@Philip Taylor: None of those links work(s).
Keith said,
February 27, 2019 @ 3:50 am
Maybe this link will work:
http://www.healthtranslations.vic.gov.au/bhcv2/bhcht.nsf/PresentDetail?Open&s=Patient_Guide_to_Total_Hip_Replacement_(THR)_Surgery
David Marjanović said,
February 27, 2019 @ 5:28 am
I would say it has three. Analyzing it as having only two, as done in the source used on Wikipedia, requires that the vowel cluster /aa/ is remarkably common, while /əə/, /əa/ and /aə/ never occur.
Historically, this third vowel /aː/ comes from the sequences */ʔa/ and */aʔ/, which are preserved as such in Abaza. Abaza (spoken on the other side of the mountains from Abkhaz, to which it is closely related) thus really has only two phonemic vowels; and so, by general consensus, did Proto-Northwest-Caucasian.
I thought this is from Evliya Çelebi?
I thought even Johanna Nichols had come around to it some 10 or 15 years ago? Or was that just about NEC (which she had doubted for a long time, too)?
KeithB said,
February 27, 2019 @ 9:49 am
"NJ is an interesting state with pockets of everything."
Along with that, there is a common expression in the press which is similar with the "100 words for snow" meme.
That is: City Y has the highest population/concentration of X outside of their native Z.
For example: Glendale California has the largest populations of Armenians outside of Armenia.
I here it all the time and wonder how they know.
Victor Mair said,
February 27, 2019 @ 10:14 am
I always thought that Watertown, MA, where I lived for a couple of years, had the largest Armenian population in the US, but I was wrong. From the Wikipedia article on Watertown:
=====
Watertown is a major center of the Armenian diaspora in the United States, with the third-largest Armenian community in the United States, estimated as numbering 7,000 to over 8,000 as of 2007. Watertown ranks only behind the California cities of Glendale and Fresno. Watertown is also the venue for the publication of long-running Armenian newspapers in English and Armenian, including:
Baikar Association Inc.'s
Armenian Mirror-Spectator
Baikar
Hairenik Association Inc.'s
Armenian Weekly
Հայրենիք (Hairenik Weekly)
Armenian Review
Hairenik Association also runs a web radio and a web TV station.
[notes and links stripped out]
=====
Patrick said,
February 27, 2019 @ 10:20 am
@Chris: "It reminds me of how the two vowel analysis of Old Chinese is often challenged as "unnatural""
And yet people accept the phonology of Proto-Indo-European just fine.
Chris Button said,
February 27, 2019 @ 10:45 am
@ David Marjanović
Doesn't the "long" vowel also come from /ə/ as well as /a/ ?
Personally I don't understand the point of an underlying phonological analysis that intentionally does not go deep enough. The line between underlying phonology and surface phonetics should be made as clear as possible.
Returning to the Old Chinese case above, it provides a classic example in that regard where people have attempted to reconstruct the underlying phonological system while simultaneously constraining their analyses on the basis of surface phonetic rhyming tendencies! I just don't understand why people can't see the inherent problem with such an approach.
@ Patrick
Yes it's absolutely ridiculous. Even more so given that people seem content to treat it as e~o rather than ə~a which actually has typological support!
Phew… rant over…
Chris Button said,
February 27, 2019 @ 3:09 pm
@ David Marjanović
Also, I think you mean the voiced pharyngeal fricative/approximant /ʕ/ rather than the glottal stop /ʔ/
David Marjanović said,
February 27, 2019 @ 6:19 pm
Oh, oops! Yes, I mean */ʕ/; a **/ʔ/ is not even reconstructed.
I may have become confused by the fact that the same change has happened in Abkhaz, where, however, */ʕʷ/ – a different phoneme – survives to this day as [ɥ].
(And that makes phonetic sense because Caucasus-area "pharyngeals" aren't generally pharyngeal strictly speaking, but epiglottal. Pharyngeals pull vowels towards [ɑ] much like uvulars do, while, epiglottals pull them towards [æ].)
Not that I know of, but I don't think I've read any primary literature on this.
In many languages, there are actually arguments for at least two underlying levels instead of just one, more or less corresponding to phonemic vs. morphophonemic transcription. I recommend this paper, which, however, does not use that transcription convention, and Ricardo Bermúdez-Otero's work which speaks of "stem-level" and "lexical-level" phonology.
In this case we have so much information from attested languages that we must reconstruct e~o at least for the last common ancestor of the IE crown-group. However, it has often been suggested in the last 20 or 30 years that that was a short-lived stage which derived from an older ə~a system. It's just hard to test as long as IE doesn't have any widely accepted close relatives, or in other words, as long as there's no Proto-Indo-Uralic reconstruction (or whatever).
Chris Button said,
February 27, 2019 @ 10:35 pm
@ David Marjanović
– It does not seem to me that a distinction between epiglottal and pharyngeal approximants/fricatives is that well-defined (Ladefoged and Maddieson suggest most are epiglottal while maintaining that a distinction can still be found), but I am not at all knowledgeable in this area.
– My understanding is that /ə/ lowered to merge with/a/ in the environment that created the "long" vowel. I'm hoping R. Fenwick will chime in soon regarding Ubykh.
– I believe in two levels: underlying phonology and surface phonetics. Anything in between is simply a means of getting from one to the other.
– W. Sidney Allen wrote about a structural relationship between Proto-Indo-European and the "ə/a" of Abaza back in 1956.
Chris Button said,
February 28, 2019 @ 2:58 pm
In response to R. Fenwick's comment here:
http://languagelog.ldc.upenn.edu/nll/?p=41832#comment-1561089
Very broadly there are three series (plain, palatal, labial) although, as you might expect, the type of phoneme and its position as onset or coda affects how it actually surfaces and what features it can bear.
For Old Chinese I tend to view the origin of the palatal and labial series as a result of the ability of the glides /j/ and /w/ to function as the syllable nucleus to give CjC (or CjəC) and CwC (or CwəC) syllables, in which the schwa is inherent, along with apophonic variants CjaC and CwaC. Crucially, Old Chinese only allows the glides (at the top of the sonority hierarchy) to function as the nucleus whereas in Proto-Indo-European other sonorants could also provide that role (although only /j/ and /w/ afford us with the transcriptional flexibility of /i/ and /u/ and hence the artificial perception of high vowels in PIE).
Incidentally, Pulleyblank later (1991 onward) added a labiopalatal series to Old Chinese, but I'm not convinced.
I couldn't agree more and made some similar comments above.
R. Fenwick said,
March 1, 2019 @ 1:14 am
@Victor Mair:
Ah, cameron, but it's wonderful that what's in the Semnani sack are walnuts, which have occupied us greatly since this post of yesterday:
You may also enjoy the fact that in Ubykh, the "stones" (laqʲá) in the scribe's sack are just a minimal pair away from "walnuts" (laqʲ’á)! :)
In response to Stefan Georg's comment in the original post:
Yes, this is nice, but the first guy is not a native speaker, since the last one, Tevfik Esenç (who worked mostly with Dumézil, but also with Vogt and Charachidzé), passed away in 1992 – his last taped words in Ubykh were „Here ends the Ubykh language.“
Both the videos shared by Peter Golden are of the same story ("Eating Fish Makes You Clever"), told by two different speakers. The speaker of the second video is indeed Tevfik Esenç – the man once allegedly referred to by the great Ossetic scholar Vasily Abaev as the "Ubykh Homer" – who was the last fully fluent heritage speaker. Two partial L1 speakers still live, Erol Esenç (Tevfik's son) and Özcan Komaç, who is the son of Hidayet Komaç, another informant of Dumézil's. (I had the pleasure of meeting and working with Özcan bey in Hacıosman during the last two summers. He was raised speaking Ubykh as his L1, but has lost much of his ability through simple lack of use; nonetheless, he can still express some ideas in the language, has helped me with providing Turkish glosses for some Ubykh terms, and he started our session last year with the following encouragement to me: wanagʲə́ aʃáw, aʃəgʲə́ wbjáw; amdwá[n]da jadagʲə́ wbjáw "all right, you will see what happens; if it [Ubykh] is not dying, you will see a lot!")
The speaker in the first video is Ali Berzeg, who is indeed not a native speaker but is ethnically Ubykh; he's of the Barzagʲ clan, one of the old princely families from before the Ubykh Exile and the clan to which the teenage informant of Baron Peter von Uslar's 19th-century sketch belonged. Ali has worked for some years, from what I understand, to reacquire some of his ancestral language, especially some of the old tales told by Tevfik Esenç.
Here's a free-ish translation of the story itself:
Once, there were two men who were travelling together as companions. Both went one day to purchase provisions for them to eat. One of them bought cheese and bread, and the other one bought bread and fish.
When they got back on their way, the one who had bought the cheese asked of his companion: "You people eat a lot of fish; why is it that you eat so much fish?"
His companion responded: "If you eat fish, you become much smarter; it's because of that that we eat a lot of fish."
"If that's true, then give me a little of your fish!" the first man said.
"Very well, I'll give you some," said his companion, and cutting a small piece from the fish's head, he gave it to the first man.
"How much do I owe you for it?" said the first man.
"One piece of gold," replied his companion.
"All right," said the first man, and he gave his companion one piece of gold for the piece of fish, and they continued on their way.
After a short while, the first man turned to his friend from whom he had taken the fish. He asked, "How much did you pay for that fish?"
"I bought it for one piece of gold," replied his companion.
The first man was incensed. "Look, you sold me just a little piece of its head for one piece of gold, and you still get to keep the rest of the fish!"
His companion replied, "You see? You just had a little taste of its head, and you're much smarter already! Who knows what you could have become if you'd eaten all of it?"
R. Fenwick said,
March 1, 2019 @ 1:34 am
@David Marjanović:
I would say it has three. Analyzing it as having only two, as done in the source used on Wikipedia, requires that the vowel cluster /aa/ is remarkably common, while /əə/, /əa/ and /aə/ never occur.
This refutation is problematic because Ubykh phonology would quite simply explain these absences in other ways. The general absence of initial *ə– and the vacillation between unstressed schwa and zero made a two-vowel analysis eminently plausible: *-əə– and *-aə– clusters would be absent purely because word shapes didn't permit initial *ə– at all (therefore it can never appear as the second element of a cluster), and *-əa– clusters would generally reduce to –a– by loss of unstressed schwa, leaving –aa– as the only functionally common vowel cluster. Nevertheless, Vogt's old claim that ạ was often the result of coalescence of *-aa– has since been shown to be wrong anyway: initial a– in Ubykh lexical roots behaves as ạ– phonologically, so the medial vowels of (for example) ʧạ́χʲa "stable" and gʷəmạ́χʲa "byre" are the same despite the fact that they arise from distinct underlying sequences: *ʧə-aχʲa (horse-pen), *gʷəma-aχʲa (cow-pen).
Historically, this third vowel /aː/ comes from the sequences */ʔa/ and */aʔ/, which are preserved as such in Abaza.
Frankly, I've never before heard that claim for Ubykh, only ever for Abkhaz (and only the southern Abkhaz dialects at that; Ashkharywa preserves surface ʕ). Not even Nikolayev and Starostin were willing to make such a claim unqualified. In any case, there's no evidence that Ubykh ever underwent a similar process, and in fact there are many solid cognates suggesting the Proto-NWC pharyngeal fricatives *ʕ ʕʲ ʕʷ ʕᶣ simply became the corresponding uvulars *ʁ ʁʲ ʁʷ ʁʲ in Ubykh, occasionally expressively pharyngealised:
PNWC *abʕa "skinny, bony": Bzyp abáː, Ubykh abˁʁˁá "id."
PNWC *gᶣəʕʷa "to believe, to suspect": Bzyp a-gʷɥá-ra, Ubykh gʲəʁʷa– "id."
PNWC *ʑaʕʷa "plague, pox": Bzyp a-ʑɥá, Ubykh ʑaʁʷá "id."
PNWC *məʕᶣa "road, way, path": Bzyp á-mɥa, Ubykh məʁʲá "id."
I believe the a/ạ distinction in Ubykh rather arose as a result of the intense influence upon it from Circassian, whose older two-vowel system *ə/a similarly increased to three *ə/e/a through a regularly-conditioned split of Proto-Circassian *a in non-final open syllables (note that Ubykh also only allows –ạ– in non-final syllables). This then allowed for further innovations within Ubykh, of course (as well as its introduction through subsequent loans), but in origin it remains the result of Circassian influence. Ubykh terms of the form CạCa, for instance, are almost always Circassian (or later Turkish) loans: a perfect example is bạʤá "fox" (compare Temirgoi Circassian baʤe "id."), which coexists with the native Ubykh cognate in bˁagʲáʃʷ "jackal" (with diminutive -ʃʷ) as well as Abaza bagá "fox", all going back cleanly to Proto-NWC *baga "a wild canid".
I thought this is from Evliya Çelebi?
Perhaps the general story does come from Evliya Çelebi, but if it is, it's not in reference to Ubykh; to my knowledge Evliya efendi only discusses Ubykh in his small travel phrase lists, and there misconstrued Ubykh as lisān-i Ṣadša-Abaza even so. In any case, though, in the meantime it's become widely told in the Near East. As I expressly said in the sentence immediately before the one you quoted, this specific version was told about, and in, Ubykh to the German linguist Adolf Dirr; if you'd like to check it out for yourself the text is on p. 46 of Dirr's Die Sprache der Ubychen (Verlag der Asia Major, Leipzig, 1928). If I can find the spoons I might even look into rerecording it.
R. Fenwick said,
March 1, 2019 @ 1:57 am
@Chris Button:
Very broadly there are three series (plain, palatal, labial) although, as you might expect, the type of phoneme and its position as onset or coda affects how it actually surfaces and what features it can bear.
I'm very intrigued by this. I should state upfront that I categorically reject all the fringe long-range proposals for Sino-Caucasian or other such things, but with that said, I find the typological parallel absolutely fascinating (and doubly so when considered alongside PIE, which almost certainly interacted with PNWC as well). Are there sources you'd be able to recommend for a linguistically-competent novice to Chinese historical linguistics?
For Old Chinese I tend to view the origin of the palatal and labial series as a result of the ability of the glides /j/ and /w/ to function as the syllable nucleus to give CjC (or CjəC) and CwC (or CwəC) syllables, in which the schwa is inherent, along with apophonic variants CjaC and CwaC.
From what little I've been able to glean in the Papua New Guinean literature, the situation in Sepik-Ramu languages seems to bear some similarity to this. The only overt discussion I've found so far is from William Foley's 1991 grammar of Yimas, but there's probably more in the literature that's relevant to the topic.
My understanding is that /ə/ lowered to merge with/a/ in the environment that created the "long" vowel. I'm hoping R. Fenwick will chime in soon regarding Ubykh.
As I note in the above comment, David Marjanović's statement regarding the Ubykh vowels is off base (probably by confusion with Abkhaz). The Abkhaz long vowel aː does come from Common Abkhaz-Abaza *ʕ adjacent to the open vowel a. Conventional analysis of Abkhaz doesn't consider an origin from the close vowel ə in the same position, but that's because this vowel merges unconditionally with a adjacent to unrounded pharyngeal fricatives anyway, whether voiced or voiceless – sequences of *ħə *əħ are globally prohibited in Abkhaz as well.
However, Ubykh appears to reliably distinguish three underlying vowel phonemes, /ə ɜ ɐ/, in a system that I think has probably developed under prolonged intensive influence from Circassian. They're symbolised in Dumézil's practical orthography as ə a ạ, since the distinction between /ɜ ɐ/ is often very subtle and is realised primarily as whether or not the vowel in question is subject to phonetic colouring from adjacent (usually preceding, but in the case of the glides /w y/, also following): s(ə)kʲ’an [sɨ̥ˈc’ɛn] "I go", but ʃ(ə)kʲ’ạn [ʃɨ̥ˈc’ɐn] "we go". (The difference between a and ạ can be quite hard to hear at times, especially when the nearby consonants bear no secondary articulation that might offer colouring to the vowel.)
That said, the three-vowel system is subject to some major caveats: /a ạ/ are not distinct word-initially or word-finally, leaving only a binary distinction in these locations; also, since the canonical Ubykh lexical morphosyllable is of the shape (C)(C)V, the three-way phonemic distinction is possible only in polymorphemic words or in non-final syllables. Moreover, word-initial /ə/ is restricted to a single allomorph of a single morpheme (the allomorph ə- of the third-person absolutive pronominal prefix), and even there only appearing as an uncommon optional variant under stress of a more usual form jə-. Further complicating things is that only rarely does unstressed schwa appear to be phonologically distinct from zero, and the question of whether purely epenthetic schwa counts as a phonological vowel or not is a fairly vexed one, particularly in NWC. (I try not to get too hung up on it, myself; far better phoneticians and phonologists than I have gotten directly involved in arguing about the NWC vowel systems, including Allen, Kuipers, Trubetzkoy, Smeets, Catford, Halle, and others.)
Chris Button said,
March 1, 2019 @ 11:11 am
@ R. Fenwick
Thanks for clarifying that. So not only is the origin different, it's not even a long vowel! Yes, I was connecting DM's comment to the comparison between Abaza and Abkhaz by people like W.S. Allen.
Yes, I agree with that (at least in terms of available evidence).
Regarding Old Chinese, the best basic introductions in English remain Karlgren's "Compendium of Phonetics in Ancient and Archaic Chinese" (1954) and Baxter's " Handbook of Old Chinese Phonology" (1992) since they are not too heavy going and provide enough background context to the non-specialist. However, neither of these works touch upon a "ə/a" system and both blur the line between underlying phonology and surface phonetics (albeit far more in the case of the former than the latter).
For "ə/a", you need to turn to Edwin Pulleyblank whose immense output is unfortunately scattered across multiple journals and often quite a challenging read. In "An Interpretation of the Vowel Systems of Old Chinese and of Written Burmese" (1963) and "The Final Consonants of Old Chinese" (1977-8), he resolves the distribution imbalances of some earlier reconstructions by proposing a "ə/a" system. While there are certain issues of detail with his proposals therein, the fundamentals are nonetheless sound. In terms of onsets, his article "The Chinese Cyclical Signs as Phonograms" (1979) proposes a binary split of Cʲ- and Cʷ- which, although highly problematic not least in terms of an association with the cyclical signs, is at least thinking along the right lines. A decade or so later, clearly influenced by his ruminations on Proto-Indo-European, he overhauls his 1978 article as "The Ganzhi as Phonograms and Their Application to the Calendar" (1991). This is again extremely problematic; most notably he has shifted to a C-, Cʲ-, Cʷ-, Cᶣ- system which now only occur after velars (in accordance with PIE k-, kʲ-, kʷ- to which he believes kᶣ- can also be added).
In this regard, I think Pulleyblank's otherwise truly brilliant work was thrown by two things: a mistaken analysis of Burmese in which he believed that the palatal codas were primary (which he then extended to his analysis of Old Chinese) rather than secondary (incidentally Old Burmese has an underlying "ə/ɐ" system when the data is not forced); he stopped focusing exclusively on what the data was actually telling him and started trying to force it into a Proto-Indo-European shell (much as how people try to force vertical vowel systems into triangular ones). The solution actually comes from Baxter's 1992 reanalysis of the traditional Old Chinese rhyme categories based on surface phonetic tendencies which is an incredibly useful analysis only marred by his attempt to apply it to underlying phonology. Strikingly, his reanalysis only disrupts rhymes before non-velar codas in Pulleyblank's 1977-8 reconstruction; even more strikingly, no-one seems to have commented on that fact! Essentially what broadly happened was the following: the velar codas -ɣ, -ŋ, -k shifted to -j, -ɲ (< -ŋʲ), -c ( CʲəC) or CwC (> CʷəC) with apophonic CʲaC or CʷaC.
Thanks that's interesting to note. I will check Foley's work out.
Chris Button said,
March 1, 2019 @ 11:20 am
For some reason, the last part of the final paragraph of my second section above was truncated (perhaps due to length restrictions?) as:
It should have said the following:
Essentially what broadly happened was the following: the velar codas -ɣ, -ŋ, -k shifted to -j, -ɲ (< -ŋʲ), -c ( CʲəC) or CwC (> CʷəC) with apophonic CʲaC or CʷaC.
Chris Button said,
March 1, 2019 @ 11:22 am
Ok it just did it again. I think it is my forward and backward arrows. Third time lucky….
Essentially what broadly happened was the following: the velar codas -ɣ, -ŋ, -k shifted to -j, -ɲ (from -ŋʲ), -c (from -kʲ) after a palatalized onset and -w, -ŋʷ -kʷ after a labialized onset simultaneously pulling the palatal and labial features away from the onsets. These distinctive palatal and labial codas largely prevented any cross-rhyming such that while CʲaC might still have occasion to rhyme with CaC, it wouldn't with CaCʲ). Hence my take that /j/ and /w/ (but, unlike in PIE, not the other sonorants lower down in the sonority hierarchy) were able to function as a syllabic nucleus that was later encoded as CjC (as CʲəC) or CwC (as CʷəC) with apophonic CʲaC or CʷaC.
Chris Button said,
March 2, 2019 @ 7:19 am
Regarding Pulleyblank, it would be remiss of me not to mention his two 1962 articles "The Consonantal System of Old Chinese" (1962) and "The Consonantal System of Old Chinese: Part II)". I did not include them in the very small selection above since he had not developed his "ə/a" system by then, but his proposals there provided the foundations for many of the things now taken as a given in Old Chinese. Needless to say, he is one of the few people for whom I would think the overused term "genius" is wholly appropriate.
David Marjanović said,
March 2, 2019 @ 2:16 pm
Ah, so I not only confused ʔ and ʕ, but I also confused Ubykh and Abkhaz at the same time! That explains a few things! Many thanks for the detailed clarifications. :-)
There does not seem to be any language with a phonological distinction between pharyngeals and epiglottals. Aghul, an East Caucasian language, has them as allophones: pharyngeals with back vowels, epiglottals with front vowels. Within Arabic, the distinction is geographic instead – western Arabic has epiglottals, and so do Berber and Korandjé.
(On top of that, epiglottal stops exist, while pharyngeal stops appear to be impossible for humans to articulate.)
No surprise there – Karlgren believed that the phoneme concept was just a fad that was going to disappear any year now. In the 1950s he wasn't alone in this, of course.
David Marjanović said,
March 2, 2019 @ 2:52 pm
If Mandarin can be analyzed as having just two underlying vowels without much effort, and if the same can be done for Old Chinese, the next logical question may be: can the same be done for Middle Chinese?
(My initial impression is "no way", but I haven't tried.)
Chris Button said,
March 3, 2019 @ 8:49 am
The problem with asking a question like that is that it makes it seem like it is just some theoretical game. The approach should rather be one of reconstructing the underlying phonological system and seeing where the data takes you. That Old Chinese, Proto-Tibeto-Burman and Proto-Indo-European all end up with such a system is not surprising given the depth of the reconstructions. Sometimes it can be happened upon at much shallower depths (e.g. Old Burmese by virtue of the inscriptional evidence from 900 or so years ago, or Northwest Caucasian languages today).
R. Fenwick said,
March 5, 2019 @ 2:46 am
Ah, so I not only confused ʔ and ʕ, but I also confused Ubykh and Abkhaz at the same time! That explains a few things! Many thanks for the detailed clarifications. :-)
Wana aš°wáma (no worries)! :) There are so many complex phonological things going on in the Caucasus, and so many of them occur across several languages, that honestly I'm pleasantly surprised such confusions don't occur more often…
There does not seem to be any language with a phonological distinction between pharyngeals and epiglottals. Aghul, an East Caucasian language, has them as allophones: pharyngeals with back vowels, epiglottals with front vowels.
The Richa dialect of Aghul does appear to raise the distinction to the phonemic level, as reported by Sandro Kodzasov in 1987; he didn't give any minimal pairs, but for near-minimal pharyngeal/epiglottal contrasts before Aghul –a-, cp. ħaw "udder" vs. ʜatʃ "apple", ʕan "belly" vs. ʢakʷ "light".
With that said, more recent literature does seem to be leaning towards treating the distinction as one not of place (as pharyngeal vs. epiglottal), but of manner (as fricative vs. trill) within a single more general pharyngeal/epiglottal place of articulation; the only location in the pharynx where trills can be produced is at the aryepiglottic folds, and any constriction in this area sufficient to produce frication tends to develop trilling anyway.
(On top of that, epiglottal stops exist, while pharyngeal stops appear to be impossible for humans to articulate.)
My understanding was that some humans are incapable of articulating true pharyngeal stops. But in any case, yes, there appears to be no language that has true phonetic pharyngeal stops, while epiglottal stops are found in a surprising range of individual cases around the world (a brief skim turns up examples in Amis, Dahalo, Haida, Nuuchahnulth, and Thompson River Salish as well as North-East Caucasian Archi, Aghul, Lezgi, probably Tsakhur, and Chechen and Ingush; oddly, epiglottal stops don't exist in Bats, the third of the Nakh languages).
R. Fenwick said,
March 5, 2019 @ 3:01 am
@Chris Button:
The problem with asking a question like that is that it makes it seem like it is just some theoretical game. The approach should rather be one of reconstructing the underlying phonological system and seeing where the data takes you.
I don't disagree at all, though at the same time I do understand what David means by the question – which is to say, in view of underlyingly bivocalic analyses of both Old Chinese and apparently now also Mandarin, might previously-completed reconstructions of Middle Chinese (with more apparently pedestrian vowel systems) have been somehow influenced by biases regarding what should be possible as a phonological system?
Tangentially, many thanks for your reading recommendations. I'm looking forward to diving into them!
Chris Button said,
March 5, 2019 @ 5:41 am
@ R. Fenwick
That's a fair point. Incidentally, Charles F. Hockett proposed a two-vowel analysis of Mandarin way back in 1947 so it's not really something new in that regard.
Chris Button said,
March 5, 2019 @ 12:10 pm
Having said that Pulleyblank himself does not reconstruct such a system for his Early and Late Middle Chinese, although it's worth noting that "Middle Chinese" is a somewhat artificial construct anyway based on written evidence in rhyme dictionaries and rhyme tables.
It's also worth noting that Pulleyblank's "vowelless" analysis of Mandarin (clearly inspired by Kuipers' analysis of Kabardian) allows /a/ to pattern as a feature of openness in the same way that /j/ and /w/ pattern as features of palatility and labiality that only become vocalized when occupying the requisite slot in the syllable. This seems to differ from Old Chinese and Proto-Indo-European in which /a/ aligns with otherwise inherent /ə/ (as its apophonic derivative) as a basic building block of the syllable, rather than with /j/ and /w/.