13.5.18.23:59: MANCHU VOWEL P-Ū-ZZL-E

I initially wrote in "J_(r)_en",

In Manchu, u ... e (a yin-yin sequence) and ū ... a (a yang-yang sequence) are possible, but not ū ... e (a yang-yin sequence).

However, the yang-yin sequence ū ... e was in Old Manchu Jūsen (sic!). I quickly caught my mistake and added two words (in bold)

In standard written Manchu, u ... e (a yin-yin sequence) and ū ... a (a yang-yang sequence) are possible, but not ū ... e (a yang-yin sequence).

Did vowel harmony rules change between standard written Manchu and Old Manchu? No, but orthographic conventions changed:

Was the variety of Mongolian known to the Manchu (technically still Jurchen until 1635) at the turn of the 17th century like modern Khalkha Mongol? Old Manchu spelling conventions are more or less what I'd expect from a Khalkha speaker. (There are exceptions: e.g., [suwəni] 'you' is spelled as ᠰᠤᠸᠠᠨᠢ <suwani> as well as ᠰᠦᠸᠡᠨᠢ  <sūwani> in Old Manchu.)

The big mystery to me is why the ᡡ  grapheme was reassigned to [ʊ] in standard Manchu.

5.19.13:09: I used to think that the change in spelling reflected a shift from a Turkic-style vowel system with palatal harmony to a Khalkha-style vowel system with height harmony:

In this scenario, original [y] and [ø] merged into [ʏ] which then backed to [ʊ] and raised to [u] while original nonlow back vowels lowered: [u] > [ʊ] (only after uvulars?), [o] > [ɔ]. The use of <ü> for [ʊ] would date from stage 3 below:

The trouble is that I don't know of any text with stage 3 characteristics: i.e., <ū> after both nonuvulars and uvulars. The closest thing I can find in the Old Manchu texts I have on hand (those in Roth Li's 2010 textbook) is "Manchu-Chinese cooperative living" (1621-1622) in which <ū> is in yin-voweled words except for ᠦᠯᠭᠢᠶᠠᠨ <ūlgiyan> ~ ᠤᠯᠭᠢᠶᠠᠨ <ulgiyan> which must be a yang-voweled word because of its <a>. I am hesitant to draw any conclusion from such an isolated error.

Worse yet, those early texts sometimes have both <u> and <ū> in yin-voweled words in the same text - a practice that the above table fails to predict: e.g., "Manchu-Chinese cooperative living" has ᠵᠤᠰᠡᠨ <jusen> as well as ᠵᠦᠰᠡᠨ <jūsen> 'Jurchen'.

Let's start over again ... in 1599 when Nurhaci commissioned the adaptation of the Mongolian alphabet to Manchu, Manchu had at least two types of vowels that were written with the Mongolian letters ᠤ <u> and ᠦ <ü> (a combination of <u> and <i>).ᠦ <ü> is ū with a macron instead of a in the Möllendorff romanization of Manchu. Presumably these two vowel types (not necessarily two vowels)

- sounded like the vowels written with those letters in Mongolian

- were the ancestors of the three vowels [u ʊ ɔ] of later standard Manchu

In Mongolian, the letter ᠦ <ü> cannot follow uvulars, and its Old Manchu counterpart ᠦ <ū> also did not follow uvulars in the texts I have seen. However, in standard Manchu, ᠦ <ū> usually follows uvulars (!). How did that happen? Here's one last scenario:

- Old Manchu had four labial vowels [y u ʊ ɔ]

- [y] was losing its palatality and merging with [u]

- [y] was written as ᠦ <ū> (ᠤ <u>  after a yin vowel) and the rest were written as ᠤ <u>

- [y] was lost by the time the alphabet was finalized, leaving three labial vowels [u ʊ ɔ]

- [u ɔ] were much more common than [ʊ] which was mostly after uvulars:

Vowel	Yin/yang	Uvulars	Velars	Nonuvulars and nonvelars
[u]	neutral	no	yes	yes
[ʊ]	yang	yes	no	rarely
[ɔ]				yes

- Therefore [u ɔ] should be written with characters requiring less effort than [ʊ].

- [u] was written as yang <o> with a dot by analogy with noninitial [ə] which was written as yang <a> with a dot:

Dotless yang / lower vowels	Dotted higher vowels
[o] ᠣ  <o>	[u] ᡠ <u>
[a] ~ [ɑ] ‍ᠠ‍ <a> ‍	[ə] ‍ᡝ‍ <e>

- The existing letter ᠦ <ū> - similar to ᠣ  <o> but differentiated by more than a dot - was then recycled for the low-frequency vowel [ʊ]. This does not necessarily mean that [ʊ] came from an earlier [y] and/or [u]; graphic continuity (i.e., the reuse of a shape) does not entail phonetic continuity (i.e., a sound change bridging the two sound values).

Similarly, the fact that the Latin letter y once stood for the Azerbaijani vowel [y] but now stands for the Azerbaijani consonant [j] does not mean that Azerbaijani [y] became [j]. (Compare Azerbaijani alphabets here.)

I still don't find that last account entirely convincing. A real answer would require a detailed study of spelling in all early Manchu texts. One might hope that a reconstruction of Old Manchu phonology - if it's really that different from the phonology of standard Manchu only decades (!) later - would map nicely onto what is known about Ming Jurchen phonology, but I fear that might not be the case, since written Manchu is not a direct descendant of the dialects transcribed by the Chinese.