No, I haven't forgotten about the mystery vowel of Manchu. I'll get back to it next week.

In the meantime, I've been looking at the Cyrillic letter ӯ which is reminiscent of the ū used to transcribe the Manchu mystery vowel. The macron atop у - normally [u] in Cyrillic - implies a long vowel, but sources at hand imply otherwise.

As far as I know, ӯ is only currently used in Tajik. It too is a mystery vowel. I've found three different sound values for it:

According to Comrie (1996: 708), ӯ stands for [o], whereas о stands for [ɒ].

One Wikipedia page states that ӯ is [ø] (though Tajik is close to Persian which lacks front rounded vowels) and another states that it is a rounded schwa [ɵ] (which only occurs in 1.11% of UPSID*).

The Arabic script equivalents of ӯ (presumably initial او and و elsewhere) imply that it orignated from a long *uu and its Hebrew script equivalent (אוֹ) implies a long oo.

According to Windfuhr (1987: 543), Tajik ӯ is from early New Persian oo:

Early New Persian i ii ee a aa u uu oo
Tajik и е а о у ӯ
Dari e i ee a aa o u oo
Farsi e i æ ɒ o u

Windfuhr uses the Czech-like symbol ů for Tajik ӯ, probably because Czech ů [u] is also from long *oo.

ӯ used to represent [ʊ] in the Kazakh alphabet (cf. ū for Manchu [ʊ]) but was replaced by ұ in 1957. Wikipedia has an explanation that sounds like an urban myth: the newspaper «Социалистік Қазақстан» misprinted

Ӯлы Сталин 'Great Stalin' (cf. Turkish ulu 'great')


Улы Сталин 'Poisonous Stalin' (does улы have a Turkish cognate? - I can't find any)

ұ was created to avoid the problem of accidental diacritic omission. Is it true?

This story reminds me of the North Korean neologism

원쑤 wŏnssu 'enemy'

to avoid ambiguity with

원수 wŏnsu 'marshal'

the title of Kim Il Sung. The two are still both 원수 wŏnsu in South Korea and can be disambiguated with Chinese characters:

원수 (怨讐) wŏnsu 'enemy'

원수 (元帥) wŏnsu 'marshal'

*This figure includes Manchu, which is listed in UPSID as having seven instead of six vowels:

i u
e ø ɵ o

I have not seen UPSID's source (Austin 1962), so I can only assume ø and ɵ are pronunciations of written Manchu ū. This is a very unusual system. None of the four other languages with ɵ in UPSID have this system. ҚАЗАҚ КИРИЛ ӘЛІПБИІ

David Boxenhorn sent me a link to this Kazakh video with [i] and [u]. The Wikipedia article on the Kazakh alphabet analyzes them as underlying vowel-glide sequences:

The letter И represents the tense vowel [i] obtained from the combinations ЫЙ /əj/ and ІЙ /ɪj/. The letter У represents /w/ and the tense vowel [u] obtained from the combinations ҰУ /ʊw/, ҮУ /ʉw/, ЫУ /əw/ and ІУ /ɪw/.

Is this really necessary? Even if Kazakh lost native /i/ and /u/ (which shifted to [ɘ] and [ʊ]), why can't it have new simple vowel phonemes /i/ and /u/ in Russian loans?

The Kazakh Cyrillic alphabet (Қазақ кирил әліпбиі [qazaq kiril ælɘpbiɘ]?) consists of the Russian alphabet plus letters for Kazakh. Some of the latter are interesting:

ә is not [ə], but [æ]; the Azerbaijani Cyrillic and Latin alphabets have the same shape-sound correspondence.

ғ is [ɣ] as in Azerbaijani Cyrillic. There are no consistent Cyrillic letters for voiced back fricatives:

g ɣ ɢ ʁ ɦ
Kazakh г ғ
Azerbaijani г ғ
Uzbek г ғ
Sakha г ҕ
Kirghiz, Tatar, Turkmen г г
Abkhaz г ӷ (formerly ҕ)
Avar г гъ
Kabardian (none!) г гъ
Chukchi (none!) г
Nivkh г ғ ӷ ӻ
Tajik г ғ
Ukrainian ґ (formerly г) г
Belarusian (г)
Russian, Bulgarian, Serbian, Macedonian г (No voiced back fricatives)

(21:27: Expanded table: split Azerbaijani into its own row and added Uzbek, Sakha, Abkhaz, Nivkh, and Tajik.)

(10.17.14:48: Expanded table: added Kirghiz, Tatar, Turkmen, Avar, Kabardian, Chukchi based on Comrie in Daniels and Bright 1996.)

The phonetic value of г is predictable in Tatar and presumably also in Kirghiz and Turkmen: [g] before front vowels and [ʁ] before back vowels.

Ukrainian and Belarusian г is from *g. Ukrainian ґ [g] is only in loanwords and onomatopoeia and was once written as кг (Shevelov 1993: 950-951) and as г (as in Russian). Belarusian г [ɣ] is [g] only in loanwords and onomatopoeia.

The descender of қ [q] reminds me of the descender oaf Latin q (which represents [g] in Azerbaijani - g is Azerbaijani [gʲ]).

The descender of ң [ŋ] implies potential phonetic similarity with қ, but the two are quite different: ң is velar and қ is uvular. [ŋ] is ñ (cf. Spanish ñ [ɲ]!) in the Kazakh Latin alphabet.

ө [ɥʉ] is not unlike IPA ø.

ү [ʉ] is distinct from у [u] which has a nonvertical descender.

ұ (ү [ʉ] with a bar) looks like a yen sign ¥ but is actually [ʊ].

һ is [h] and is presumably directly borrowed from the Latin alphabet. Its capital letter is Һ since Н is [n] in Cyrillic. Cyrillic is often Greek-like but not in this case. Greek Η (classical [ɛɛ], later [i]) corresponds to Cyrillic И [i], not Cyrillic Н [n]. WHAT'S M-I-SSING FROM KAZAKH?

Thanks to David Boxenhorn for bringing up ATR (which I call 'high/low') during our discussion of Mongolian vowels. According to Wikipedia, Kazakh has a 'high/low' vowel system like earlier Korean, my interpretation of the Standard Manchu vowel system, and Khalkha Mongolian* but with a striking difference: it has no i!

nonlow achromatic glide + nonlow achromatic low achromatic glide + labial labial
High series ɘ (typo? should this be jɘ?) æ ɥʉ ʉ
Low series ə (none) ɑ ʊ

That stunned me so much I didn't even notice it was also missing u as a phoneme**. It's the first language I've ever seen with an eye-straining (and probably also ear-straining) contrast between ɘ and ə:

The modern Kazakh system orignated from a more Turkish-like system with fewer exotic vowels:


high nonlabial mid nonlabial low mid labial high labial
Palatal *i *e *y
Nonpalatal (none) *o *u

Modern Turkish

high nonlabial nonhigh nonlabial nonhigh labial high labial
Palatal i e ø y
Nonpalatal ɯ a o u

Here are the frequencies of the unusual vowels of Kazakh in UPSID:

ɘ: 4.88% (this includes Bulgarian and Thai whose central vowels are usually regarded as ə and which lack a contrast between ɘ and ə)

ʉ: 1.55% (Nishida 1964 and Li Fanwen 1986 reconstructed this vowel in Tangut, but I don't)

Diphthongs with ʉ: 0%!

I don't know how the pre-Kazakh system turned into the modern Kazakh system, so I can only guess:

1. Mid vowels broke into glide + vowel sequences:

*e > *je

> *ɥø

*o > *wo

Such breaking (or bending) is common: e.g, in Italian:

'foot': pedem > piede

'fire': focus > fuoco

Late Old Chinese mid vowels also underwent similar changes after nonemphatic initials:

支 'branch': *ke > *kie > *tɕie

駒 'colt': *ko > *kuo

2. Front vowels backed to central:

*i >

*je > (typo for jɘ?)

*ɥø > *ɥɵ

*y > ʉ

3. Most high vowels lowered:

> ɘ

> ə

*u > ʊ

ʉ could have lowered to the extremely exotic vowel ʊ̈ , found only in Somali in the UPSID. Perhaps ʉ is phonetically [ʊ̈].

4. Labial mid vowels merged with labial back vowels, reducing the number of vowels in the system:

*ɥɵ > ɥ ʉ

*wo >

The only vowels that remained unchanged were æ and ɑ.

The *e > *je > breaking in Kazakh is like the breaking I proposed for Korean which accounts for correspondences with foreign languages: e.g.,

Middle Korean 셤 sjəm < *semV 'island' : early Japanese *sema (borrowing from a Koreanic language?)

Sino-Korean 저 tɕə < 뎌 tjə < *te 'low', borrowed from Middle Chinese 低 *tej before breaking

*High-low systems are found in all 'branches' (subtypes) of Altaic other than Japonic.

Here's a high-low chart of Khalkha Mongolian vowels:

Neutral Nonback Back nonhigh Back high
High series i e o u
Low series a ɔ ʊ

And here are their earlier Mongolian sources:

Neutral Nonback Back nonhigh Back high
Palatal i e ø y
Nonpalatal a o u

This system lacks a nonpalatal ɯ ciorresponding to i, but is otherwise identical to the modern Turkish system.

**In Kazakh Cyrillic, у represents the glide w, not u. There is no need for a letter like Belarusian ў for w since Kazakh doesn't have both u and w as phonemes.

According to Wikipedia, Kazakh /ʊw/, /ʉw/, /əw/, /ɪw/ are all pronounced [u]. Couldn't one say Kazakh has a neutral /u/? Or are there phonological alternations favoring a vowel-glide analysis?

I presume Russian у [u] is borrowed as Kazakh у [u]. Would that [u] also be analyzed as one of the above four phoneme sequences? WHAT WAS THE SIXTH VOWEL OF MANCHU? (PART 4)

In my last post, I asked,

Does the Ū- of Ūlət correspond to the diphthong of Oirat?

Was Ūlət originally *Uilət?

Or *Oilət? No, no, and no.

This idea was initially tempting because Manchu ū looks like o + i in initial and medial position in Manchu script*. However, there are three problems:

1. Manchu allows oi- in initial position, so Mongolian oi- should have been borrowed as Manchu oi-. (However, one could claim that pre-Standard Manchu *oi became SM ū, and that SM oi- is from something other than *oi in pre-SM.)

2. Unlike its Altaic** neighbors Korean and Japanese, Manchu distinguishes between r and l***. So there would be no reason to borrow Mongolian Oirad 'Oirat' as Manchu *Uilət or *Oilət with -l-.

3. Mongolian Oirad 'Oirat' has an a which should have been borrowed as Manchu a. But the Manchu word has ə.

The different final consonants are not a problem since Manchu words cannot end in -d****, so a foreign -d would be changed to -t.

Gertraude Roth Li (2000: 406) equates Ūlət with Eleuth 'Oirat', identified by Wikipedia as a French missionary spelling of "something like Ölöt", another name for the Dzüüngar 'Left Hand' (eastern wing) group of Oirat tribes. (I would expect a French spelling of Ölöt to be something like Euleute.) Eleuth corresponds to modern Mongolian Өөлд Ööld.

If Ūlət is from "something like Ölöt", then was one source of Standard Manchu ū [ʊ]?

Next: How could end up as [ʊ]?

*Manchu oi is written as if it were o + i + i in initial and medial positions to avoid confusion with ū. There is no ambiguity in final position because ū has its own unique final form.

**I use 'Altaic' to refer to a type of Asian language without implying anything about genetic relationships. Altaic languages all share some characteristics. There is no doubt that this similarity is due to contact, but it is debatable whether some of it is due to common ancestry.

***Standard Korean has both r and l as allophones of a single liquid phoneme /r/:

[l] in syllable-final position and in the medial combination /rr/ [ll]

[r] elsewhere (there is no /rr/ [rr])

On the other hand, there are /l/ and /r/ minimal pairs in Manchu: e.g.,

ala- 'to inform'

ara- 'to do'

ələ 'still more'

ərə 'this'

****The only -d word I can find at anaku.cn is uzhed [udʒəd]?*****, glossed as Mandarin 老實 laoshi 'honest'. I suspect the word is a typo for something else, since there is no final form for -d in the Manchu script.

*****Anaku entries are in a mixture of romanizations. Anaku zh can correspond to j [dʒ] in other romanization systems. WHAT WAS THE SIXTH VOWEL OF MANCHU? (PART 3)

The distribution of the mystery vowel transcribed as ū may hint at its phonetic value.

In Standard Manchu, velars and uvulars are in complementary distribution except in Chinese loanwords which contain velars before the yang vowels a and o (but not ū):

Vowel class k g x q ɢ χ
Neutral i ki gi xi no neutral or yin vowels after uvulars
u ku gu xu
Yin ə
Yang a (ka) (ga) (xa) qa ɢa χa
o (ko) (go) (xo) qo ɢo χo
ū no ū after velars ɢū χū

I once thought of ū as a post-uvular allophone of /u/, but such an analysis does not lead to a much simpler system:

Vowel class /k/ /g/ /x/ /q/ /ɢ/ /χ/
Neutral /i/ ki gi xi no i after uvulars
/u/ ku gu xu ɢū χū
Yin /ə/ no yin vowels after uvulars
Yang /a/ (ka) (ga) (xa) qa ɢa χa
/o/ (ko) (go) (xo) qo ɢo χo

Is it better to view velars and uvulars as allophones of earlier velar phonemes conditioned by the following vowel?

Vowel class /k/ /g/ /x/
Neutral: velar allophones /i/ ki gi xi
/u/ ku gu xu
Yin: velar allophones /ə/
Yang: uvular allophones /a/ qa ɢa χa
/o/ qo ɢo χo
/ū/ qū ɢū χū

Such an analysis would only be valid prior to the borrowing of Chinese ka, ko, etc. which were distinct from native qa, qo, etc.

/u/ and /ū/ must be distinct phonemes since they both occur in initial position:

ulə- 'to sew a straight seam'

Ūlət 'Oirat'

(both from enenggi.com; I confirmed the second word in Gertraude Roth Li's textbook, but I can't find the first word there or at anaku.cn)

If ū really were an allophone of /u/ after uvulars, it would not occur in initial position.

Ūlət and some other ū-words contain the yin vowel ə. How is this possible? Could ū have two sources in earlier Manchu, one yin and one yang?

Alexander Vovin taught me that ū is from *ui. I don't know if he meant that all instances of ū are monophthongizations, but I wonder if some are. Does the Ū- of Ūlət correspond to the diphthong of Oirat? Was Ūlət originally *Uilət? I don't think so, and I'll explain why next time. SARA I BANJIHA INENGGI (translation*)

Today is the birthday of my friend

Sara (see the cursive Manchu here)

(typed in Manchu using Andrew West's BabelPad which has a romanization-to-Manchu converter)

Last year I wished her a happy birthday in Tangut, and this year I'll do it not just once, but twice, in the nearly extinct Manchu language (manju gisun):

Tumen jalafun jecen akū!

[tumən dʒalafun dʒətʃən aqʊ]

(see the Manchu script here; cursive here)

lit. 'ten-thousand long-life border there-is-not'

'May you have an infinitely long life!'

(a calque of Chinese 萬壽無彊 'ten-thousand long-life without boundary')

Banjiha inenggi sebjen okini!

[bandʒiχa inəŋgi səbdʒən okini]

(see the Manchu script here; cursive here)

lit. 'born day joy be-may'

'May your birthday be joyous!'

(a calque of Chinese 生日快樂 'birth day joy' = 'happy birthday')

*'Sarah's birthday' in Manchu:
















See the Manchu script here. The cursive is here.

Manchu Sara is 'sah-rah'. There is no Manchu vowel like the a of English Sarah [særa]. Manchu has no [e], so [sera] is not possible. The e in Manchu romanization is phonetically [ə], not [e].

Sara also means 'umbrella' in Manchu.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2009 Amritavision