At a glance, my high-low analysis of Manchu vowels looks doubtful. How could ū be a 'low' vowel? Even if I were to say that ū is relatively lower than high u, I would still have to deal with the following problems:

- Thanks to Andrew West, I got to see a couple of pages of 동문유해 ( 同文類解 Tongmun yuhae / Classified Analysis of the Same Text, 1748), a Chinese-Korean-Manchu lexicon. (See pp. 35-36 of this PDF.) Manchu ū is transcribed as ㅜ u in hangul.

- Jurchen is thought to be ancestral to Manchu. The Jurchen vowel corresponding to Manchu ū was transcribed in Chinese as if it were *[u] (Kane 1989: 117).

One might be tempted to regard both u and ū as high, and reinterpret 'yin' and 'yang' in terms of palatality:

Neutral Pair 1 Pair 2 Pair 3
'Yin' = 'palatal' i (neutral; has no paired vowel; was there originally a nonpalatal [ɯ]?) e u [y] (originally ø?)
'Yang' = 'nonpalatal' a ū [u] o

However, if Manchu u were [y], why wasn't it transcribed as ü by Europeans? Möllendorff (1892: 1) wrote,

[Manchu] a, i, o, u, ū [are pronounced] as in German.

German orthography does not use macrons, so I was puzzled by this statement. Ramsey (1987: 218) thought Möllendorff was equating ū with a long [uu] which does exist in German: u(h) [uu]. Ramsey also surveyed other interpretations of ū:

- Harlez (1884) thought ū was a long [oo] as in German Sohn [zoon].

(But Möllendorff wrote that ū was not [oo]. Was he thinking of Harlez?)

- Haenisch (1961) thought ū was between u and o.

- The Korean Translators' Bureau transcribed ū as o (but the pages of the KTB publication Tongmun yuhae I've seen have the transcription u; see above)

If ū were a long vowel, it should appear in environments similar to oo, which can appear after all initial classes, but as far as I know, ū can only appear after uvulars and in word-initial position in native words in Standard Manchu. (I'll discuss Old Manchu later.)

I interpret ū as [ʊ], originating from a lowered allophone of u after uvulars. Arabic /u/ also lowers to [ʊ] after uvular q, which is why Qurʔaan has been Anglicized as Koran.

But how could ū also appear in initial position if there were no uvulars to condition lowering? And if ū really is a relatively 'low' vowel, how can it coexist with the relatively 'high' vowel e in the following words*?

ūlen 'house'

Ūlet 'Oirat'

ūnglingge ejen 'almsgiver; benefactor' (ejen is 'lord')

ūren 'idol; puppet; image; tablet of a deceased person'

I'll try to answer these questions next time.

*The only other ū-initial words I know of are

Ūdui 'a name of a Muslim noble' (and hence not a native Manchu name)

ūn cecike 'bullfinch' (cecike is 'a small bird'; I presume ūn is not an error for un 'a place for pigs to sleep')

I found all but Ūdui at enenggi.com (named after the Manchu word for 'today') which is not an entirely accurate source. enenggi.com sometimes has ū instead of u, perhaps because of OCR errors: e.g., bucembi 'to die' is misspelled as būcembi (sic). Conversely, akū 'there is not' is sometimes misspelled as aku (sic), so some ū-initial words may be listed as u-initial at enenggi.com. Unfortunately, I can't confirm ū- in the above words because I left Norman's Manchu-English Lexicon at home. How many more exist?

(10.11.14:36: Expanded note on enenggi.com.) WHAT WAS THE SIXTH VOWEL OF MANCHU? (PART 1)

Earlier this week, The New York Times printed a story about attempts to revive Manchu (emphasis mine):

Indeed, with virtually no native speakers [of standard Manchu] left, it isn't always clear how to speak the words. In the Qing dynasty, a textbook had been developed for Chinese wanting to learn their rulers' languages, with Chinese characters to suggest how to pronounce Manchu letters. That helped, as did a system of transcribing Manchu script into Roman letters devised by European missionaries and academics. But even today, Manchus can't agree on how to pronounce one of the vowels, let alone how to make the language flow naturally.

The mystery vowel is often transcribed as ū. The macron is arbitrary and not meant to represent vowel length. It does not appear atop any other Manchu vowels in transcription.* It merely signifies that ū was somehow different from u.

Like Korean, Manchu has partial vowel harmony with three types of vowels: neutral, yin, and yang:

i (neutral) u (neutral)
ū (yang)
e [ə] (yin) o (yang)
a (yang)

The table above implies that

'neutral' = high

'yin' = mid central

'yang' = nonhigh but not mid central (peripheral nonhigh?)

But such a yin-yang opposition seems unlikely.

I wonder if the Manchu vowel system can be analyzed in terms of high vs. low like the Middle Korean vowel system:

Neutral Pair 1 Pair 2 Pair 3
'Yin' = 'high' i (neutral; has no paired vowel) ə u (none; merged with u?)
'Yang' = 'low' a ū o

Next: Testing the high-low hypothesis.

*One might erroneously think that ū was the only long vowel in Manchu since ā ē ī ō do not exist in transcription. Long [oo] is transcribed as oo. INTERNAL AND EXTERNAL EVIDENCE AGAINST KOREAN VOWEL ROTATION

I first encountered the term 'vowel rotation' in the mid-90s when I learned that some scholars believed it had occurred in the history of Korean:

Earlier Korean vowels (PVR1 = pre-vowel rotation 1; based on the table in Lee and Ramsey 2000: 282)

*i *y *u

Modern Korean vowels (excluding new vowels developed from vowel-glide sequences; rotated vowels in bold; the vowel ㆍ is no longer in Korean)

i ɯ u
ɤ o

Not all vowels rotated. It is a fact that the vowel ㆍ arae a disappeared from Korean. So if it was *ɔ, its loss might create a gap that was filled by ㅗ u which 'dragged' other vowels into new positions. Note that not all vowels participated in the shift (cf. Coptic).

Some linguists think Korean vowel rotation postdated 15th century Middle Korean (MK) while others think it predated the 15th century, but I have never believed in Korean vowel rotation. I reconstruct an earlier Korean vowel system which is very similar to the modern system:

*i *u
(not in modern Korean)

Although this system is unusually center-heavy, it is still neater than the pre-vowel rotation system which looks quite random and hence improbable:

- why only one front rounded vowel *y?

- or in Kim Wan-jin's reconstruction, the unusual high central rounded vowel found in only 1.6% of the languages in the UPSID

- why instead of *o to match *e and *ə?

I could also rewrite my earlier Korean system as

*i *ə *u
*a (not in modern Korean) *o

I could try to make the pre-vowel rotation system neater (cf. the table for early Korean vowels in Lee and Ramsey 2000: 278):

*i *y *u
*e *o

However, even this system (PVR2 = pre-vowel rotation 2) doesn't fit either the internal or the external evidence.

Internal evidence: harmonic vowel pairs

These 'yin-yang' pairs must have some phonetic similarity, or else they wouldn't be pairs. Notice how pairs 1 and 3 are written with hangul letters pointing in opposite directions:

Neutral Pair 1 Pair 2 Pair 3
'Yin' ㅣ(has no paired vowel)

In my reconstruction, the top row is the 'high' series and the bottom row is the 'low' series:

Pair 0? Pair 1 Pair 2 Pair 3
'Yin' = 'high' *i *u
'Yang' = 'low' (*e > ㅕ jɤ?) *a *o

In PVR2, the top row would be the 'palatal' series and the bottom row would be the 'nonpalatal series':

Neutral Pair 1 Pair 2 Pair 3
'Yin' = 'palatal' *i *e *y
'Yang' = 'nonpalatal' (*ɨ merged with *i?) *a *o *u

The top series is less clearly defined in PVR1. Pairs 1 and 3 are palatal vs. nonpalatal, but Pair 2 is unrounded vs. rounded:

Neutral Pair 1 Pair 2 Pair 3
'Yin' *i *e *y
'Yang' (*ɨ merged with *i?) *a *u

ㅡ and ㆍ are the 'minimal' vowels of Middle Korean. Their letters are two of the three building blocks of hangul vowel letters (the third being the neutral vowel ㅣ I) Which is a more likely pair of minimal vowels?

My reconstruction: ㅡ and ㆍ

PVR2: ㅡ and ㆍ *o

PVR1: ㅡ and ㆍ

My and ㆍ are achromatic (colorless; neither palatal nor labial), whereas PVR2 has two labial vowels (including a palatal and labial vowel!) and PVR1 has one achromatic vowel and labial vowel.

External evidence: Chinese loanwords

Korean has many loanwords borrowed from Middle Chinese (MC) approximately seven centuries before Middle Korean. Although the reconstruction of MC is disputed, all reconstructions and non-Korean evidence that I have ever seen agree on two points:

- the 侯 rhyme borrowed into Korean as the vowel written as ㅜ in MK was nonpalatal (unlike PVR2 and PVR1 *y but like my nonpalatal *u which matches present-day ㅜ u)

- the 痕/欣/登 rhymes borrowed into Korean with the vowel written as ㅡ in MK were nonpalatal and nonlabial (unlike PVR2 and PVR1 *y but like my nonpalatal and nonlabial which matches present-day ㅡ ɯ)

- the 支/脂/之 rhymes borrowed into Korean as the vowel written as ㆍ in MK were nonlabial (unlike PVR2 *o and PVR1 but like my nonlabial ㆍ *ʌ)

10.8.1:24: ADDENDUM: Pulleyblank's Middle Korean vowels

EG Pulleyblank (1984: 100-101) proposed that one of the MK vowel letters represented a glide-vowel sequence:

Neutral Pair 1 Pair 2 Pair 3
'Yin' *ji *e *i *u
'Yang' (*ɨ merged with *i?) *a *o

Pulleyblank believes the 'yin' vowels were once all front (presumably before the borrowing of Chinese loanwords, as he does not reconstruct ㅜ as *y).

This reconstruction has several problems:

1. It requires the frequent reconstruction of *ji, even after coronals: e.g., 디 *tji. Is there any language with numerous contrastive pairs like 드 *ti and 디 *tji, or worse yet,*pstji, even more complex than the *psti everyone else reconstructs for the MK verb stem meaning 'be steaming'.

2. No other Altaic-type languages have Cji syllables.

3. The letter ㅣ *ji is also used to write *j. Why write *uj as ㅟ u-ji instead of as u-i?

4. Why don't ㅣ *ji and ㅡ *i form a more/less palatal pair analogous to ㅓ *e : ㅏ *a (palatal : non palatal)? Why are ㅣ *ji and ㅡ *i in completely separate categories (neutral and yin)?

5. There is no consistent logic underlying the three yin-yang pairs beyond high-low (differences are in bold):

Pair 1 (*e : *a) mid front palatal vs. low (front? central?) achromatic

Pair 2 (*i : *ə) high front palatal vs. mid central achromatic

Pair 3 (*u : *o): high back labial vs. mid back labial

I suspect that the original system was

Pair 0 (*i : *e): high front palatal vs. mid front palatal

Pair 1 (*ə : *a) mid central achromatic vs. low central achromatic

Pair 2 (*ɯ : *ʌ) high back achromatic vs. mid back achromatic

Pair 3 (*u : *o) high back labial vs. mid back labial

Each pair was in the same column (front/central/back) with the same quality (palatal/achromatic/labial), differing only in relative height.

I also reconstruct a high/low series distinction in Tangut and Old Chinese, but for very different reasons.

Next: a high/low series analysis of Manchu vowels. COPTIC VOWEL ROTATION?

The Coptic alphabet indicates vowels absent from hieroglyphics:

'Egypt': Coptic κημε kēme vs. hieroglyphic k-m-t

'answer': Coptic ουωϣβ ouōshb [woʃb] vs. hieroglyphic w-sh-b

Should Coptic vowels simply be projected back into the language recorded into hieroglyphics? Did k-m-t and w-sh-b represent *kemet and *woshb? Not necessarily.

Suppose that earlier English were written in hieroglyphics whereas later English was written in an alphabet. If we projected current English vowels back in time, our reconstructions of earlier English would not reflect the Great Vowel Shift.

The Egyptian language may have undergone its own GVS. Peust (1999: 223) proposed what I call vowel rotation in Coptic:

i/e < (via y/ø?) u
v ^
a > o

Perhaps Coptic kēme and ouōshb [woʃb] were from earlier kumut and washb.

Peust (1999: 222) reconstructed a six-vowel system for Paleo-Coptic:

*ihigh *uhigh
*ilow *ahigh *ulow

I would rewrite his PC vowels as

*i *u

I could also use mid vowel symbols

*i *u
*e *o

though Peust does not give any examples of *ulow = = *o transcribed as o in late 2nd milllennium BC cuneiform.

Here is what those six vowels eventually became in Sahidic Coptic (Peust 1999: 224-225). Developments under special conditions are in parentheses (Peust 1999: 231, 238):

ι, (η after μ, ν; maybe also before ι [j]?) η (ι sometimes before λ, ρ)
α ω, (ου [u] after μ, ν, before ι [j], sometimes before λ, ρ) ε, α, ο
ο, (α before back spirants)

κημε 'Egypt' and ουωϣβ 'answer' may have once been *kumʊt and *wɐshb (or *kumot and *wəshb) in my PC notation.

Keeping *a as low α before back spirants and raising to a high ου [u] matching an ι [j] resembling a high vowel make sense to me. ι might have dissimilated to η before ι [j].

However, I don't understand why nasals either lower or raise following vowels:

Lowering: *i > η after μ, ν

Raising: > ου [u] after μ, ν

Nor do I understand why liquids raise preceding vowels:

> ου [u]

I wonder if *u shifted to a palatal vowel ι (via *y?) before λ, ρ because Egyptian liquids "might have had a somewhat palatal articulation" (Peust 1999: 243).

might have become a palatal vowel *y before palatal liquids, later shifting to ου [u]. ΟΥΩϢΒ

The ουωϣβ 'answer' to my previous question is Coptic. Let me expand on my three hints:

1. Look at the last four letters of the name of the language and its speakers.

κημε kēme is close to k-m-t, ancient Egyptian for 'Egypt'. The vowels of k-m-t are unknown, since its hieroglyphic spelling only indicates the consonants:

k-m (first glyph to the right of the dotted vertical line)

m (birdlike glyph)

t (dome-shaped glyph)

(the X in a circle glyph has no phonetic value here; it indicates that the adjacent word represents a place)

(10.6.00:15: The X in a circle glyph may be analogous to the sinographic element 阝 in characters for place names and surnames presumably baed on place names: e.g., 鄧 Deng, a place sounding like 登 deng 'climb' [and also the surname of 鄧小平 Deng Xiaoping].)

2. The font in this post is not quite correct.

The Coptic alphabet is similar in appearance to the Greek alphabet and most of it can be transliterated into Greek. However, I should ideally use a Coptic font, though I don't have one, and I doubt most of my readers have one.

3. Here are a few more words in that language from The World's Writing Systems:

εϥϣηϥ 'it being deserted'

ν̅γ̅ϩεϯε 'and you flow'

ευϫοσε 'they being high'

Those words contain five of the seven Coptic-only letters descended from hieroglyphics (scroll to the bottom of this section):

ϣ sh (10.6.0:16: which reminds me of the unrelated, similar-sounding Cyrillic letters ш sh and щ - generally shch but sht in Bulgarian)

ϥ f

ϧ x

ϩ h

ϫ j (10.6.0:18: not to be confused with gamma - Greek γ or IPA ɣ!)

ϭ q

ϯ ti

Most of these new letters represent sounds absent from the variety of Greek known to the Copts. I presume that Greek at this point

- had not shifted the aspirated stops ph kh to fricatives f x yet

- had lost initial h- (which only had a diacritic on Greek initial vowels; there was no way to write noninitial h in Greek)

Greek had no fricative sh or voiced affricate j (once a voiced palatal stop ɟ according to Peust 1999: 92).

The letter ϭ transliterated here as q may have once represented a palatal stop c from a pre-Coptic k. (The letter is derived from a hieroglyph for k.) This c, a sound absent from Arabic, then shifted to sh, presumably under Arabic influence (Peust 1999: 92).

I have no idea why ϯ ti has its own letter. Why not write ti as τι?

(10.6.1:05: ϯ ti means 'give' in Coptic. Perhaps this word was so common that it became a logogram doubling as a phonogram for the syllable ti.)

Oddly, no letter w was invented. (The Greek letter digamma Ϝ w was extinct by the time the Coptic alphabet was created.) Hence Coptic w- is written as ου ou-: e.g., in the title of this post,





descended from earlier Egyptian w-sh-b 'id.' See its hieroglyphic spelling at the bottom right of Fischer 2004: 39 (added glyph explanations 10.6.0:36):

w (birdlike glyph)

sh (horizontal box glyph)

b (foot-shaped glyph)

(the person in the box is a semantic element 'speak' with no phonetic value here)

One might expect the Chinese character 答 for 'answer' to also have a semantic element like 'speak', but instead it has 竹 'bamboo' over its phonetic 合!

10.6.1:24: I should have pointed out a fourth giveaway: the use of supralinear strokes to indicate the absence of vowels in a syllable. I don't know of any other script with a similar graphic device.

The closest analogues might be the vowel cancellation marks in Indic scripts: e.g., the viraama of Devanagari. But such markers simply convert CV symbols into consonant symbols: e.g.,

ka > क् k

whereas the supralinear strokes of Coptic are added to letters of syllables with syllabic sonorants: e.g.,




(five syllables; the first three have syllabic nasals)

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