got me thinking about Marshallese vowels and the perpetual mystery of Tangut rhymes again for the first time since 2014. The very first time I thought of comparing Marshallese with Tangut was in 2010. And nearly a decade later, it was the sight of Kwajalein in IPA that got me on that track again:


The complex vowels of Marshallese are analyzed as just four basic vowel phonemes /a ɜ ɘ ɨ/ that 'warp' under the influence of consonants with various qualities.

Similarily, the complex vowels of Tangut could have been just six basic vowels (u i a y e o) that 'warped' under the influence of consonants with various qualities (pharyngealized, uvularized, and plain from a Xun Gong-type perspective).

The 'grades' of Tangut correspond to those qualities. I write grades as numbers after basic vowels: e.g., ka1 is grade I ka.

I could write Marshallese using a similar notation: e.g., /kʷɨwatʲlʲɜjɜnʲ/ (?) 'Kwajalein' as k1ɨw1at3l3ɜjɜn3. I can't place the 'grade' numbers after the vowels since vowels are influenced by consonants on either side, and not all consonants are followed by phonemic vowels. In my Marshallese 'grade' system, 1 is labial(ized) and 3 is palata(lized); 2 - not in 'Kwajalein' - is velar(ized).

2.4.10:30: Here's my (mis?)understanding of how /kʷɨwatʲlʲɜjɜnʲ/ (?) surfaces as [kʷuɒ͡æzʲ(æ)lʲɛːnʲ]

1. /ɨw/ becomes [u] after /kʷ/.

2. /a/ becomes [ɒ͡æ] (starting labial like /w/ and ending palatal like /tʲ/) between /w/ and /tʲ/.

3. Palatal [æ] is inserted to break up palatalized /tʲ/ and /lʲ/.

4. /tʲ/ voices to [zʲ] between vowels.

5. /ɜ/ becomes palatal [ɛ] between palatal(ized) consonants (/lʲ/ and /j/; /j/ and /nʲ/).

6. /VjV/ contracts to a long vowel [ɛː].

I have doubts about whether abstract phonemic forms like /kʷatʲlʲɜjɜnʲ/ represent what speakers are thinking. The phonemic-phonetic gap seems enormous:

ɜ j
ɒ͡æ æ ɛː
Marshallese spelling
Ø l
English spelling

I am reminded of Bernard Karlgren's (1954: 366) criticism of an

intellectual sport - to write a given language with as few simple letters as possible, preferably no other than those to be found on an American typewriter.

/ʷ/, /ʲ/, and /ɜ/ obviously aren't found on an American typewriter (or any typewriter unless it's been customized, I imagine), but the problem remains: how far should a phonemic analysis go before it no longer corresponds to reality? RYUMUNADESU

Thirty years ago tonight, リュムナデスのカー サ Ryumunadesu no Kāsa 'Limnades Caça' made his animated debut on Saint Seiya. I had first seen him in the manga some months before that. That was my first exposure to the name of a kind of naiad. I had assumed the Greek name was Lymnades since Japanese borrows Greek y as yu. But in fact the closest Greek name is Λιμνάδες Limnádes with i, not y.

Could mangaka Kurumada Masami have arbitrarily changed ムナデス Rimunadesu to リュムナデス Ryumunadesu? I have doubts because I don't remember him altering any other foreign mythological names. This page lists many of those names as spelled in his manga/the anime: e.g., スキュラ Skyura 'Σκύλλα Scylla' (with the expected yu : Greek y correspondence).

The same katakana spelling appears in 門あさ美 Kado Asami's song title リュム ナデス Ryumunadesu from 1985 - three years before the Ryumunadesu in the Saint Seiya manga. Did Kurumada get his spelling from the song, or do both attestations of Ryumunadesu independently derive from a common source?

The fact that this entry in 幻想世界神話辞典 Gensō sekai shinwa jiten 'Fantasy and World Mythology Dictionary') is titled リュムナデス Ryumunadesu and cites two sources

ギ リシア神話小事典 Girisha shinwa shōjiten (A Small Encyclopedia of Greek Mythology), a 1979 translation of Bernard Evslin's Gods, Demigods, and Demons: An Encyclopedia of Greek Mythology (1975)

世 界の妖精・妖怪事典 Sekai no yōsei·yōkai jiten (An Encyclopedia of the World's Fairies and Mythical Creatures), a 2003 translation of Carol Rose's Spirits, Fairies, Leprechauns, and Goblins: An Encyclopedia (1996)

suggests that the リュムナデス Ryumunadesu spelling has a life beyond and a history predating the Seiya character and the song title.

Might リュムナデス Ryumunadesu have originated as a error by some Meiji period translator who confused Greek i with y? I'm guessing the spelling might go as far back as Meiji since I can't imagine the Japanese only learning about the Limnades during the last century. Unfortunately Google Books Ngram Viewer doesn't do Japanese yet, so I can't see any attestations of the spelling in old books. JURCHEN 1284: MAHILA 'HAT'

If I had more time, I'd write an English dictionary of Jurchen characters, building upon the foundation that Jin Qizong laid in his 1984 女真文辞典 Nüzhenwen cidian 'Jurchen dictionary'. Ideally it'd be online so I could continually update it. But in reality ... you'll get random blog entries like this one about this character or that.

Tonight's character is numbered 1284 in N3788¹. It is only attested as the first half of mahila 'hat' in the Sino-Jurchen vocabulary of the Bureau of Translators (Kiyose #547):

1284 0176 <HAT la>

Although 1284 does not appear in which seems to be the earliest surviving list of Jurchen characters, I suspect that it was originally a standalone character for mahila 'hat' in the early 12th century, and that <la> was later added to it as a phonetic clarifier at some point prior to the compilation of the Sino-Jurchen vocabulary in the 15th century. I agree with Jin Qizong who regards it as a pictograph.

The second character 0176 is a common phonogram for la. See Kiyose (1977: 70) for a list of its other occurrences within the vocabulary and Jin Qizong (1984: 36-37) for examples in other texts. It is apparently the sole Jurchen character pronounced la.

I think of 0176 as Chinese 友 'friend' with an extra dot, but the first stroke of the part of 0176 resembling the 又 component (originally a drawing of a hand, though it does not represent a word for 'hand' in Chinese) stretches further leftward, crossing over the 丿 stroke (part of 𠂇, a drawing of another hand). How many Chinese students of Jurchen miswrote 0176 as 友 plus a dot?

Speaking of hands, the Tangut word for 'hand' is 𗁅 3485 1laq1 < *S-lak. 1laq1 and other Tibeto-Burman (i.e., non-Chinese Sino-Tibetan) words for 'hand' sound like 0176 la. Is 0176 a repurposed character originally intended to write 'hand' in  some Tibeto-Burman language²? That hypothesis makes no geographic sense, as there were no Tibeto-Burman languages spoken in Manchuria³. I regard the correspondence between  the 又 hand shape and Tibeto-Burman lak-words for 'hand' as a coincidence.

¹If I use N4631 numbers for Khitan large script, I might as well use N3788 numbers for the Jurchen (large) script.

²1.28.20:35: The Tangut script is a rich source of pareidolic stimuli. After 23 years, I suddenly 'saw' the hand-shape in the right-hand component 𘦳 of 𗁅 3485. (I still don't know why that component, often regarded as 'hand', cannot stand alone and needed a vertical stroke to be a standalone character.) If one pulls apart 又 into its component strokes フ and 乀, inserts two more 丿 between them, and adds two strokes 丷 on top, the result is  𘦳.

One could also subtract what I've called the 𘡊 'horned hat' and see the remaining 𘢌 as Chinese 手 'hand' tilted 45 degrees, but I think the resemblance between the two elements is coincidental. 𘢌 is often (but not always!) 'person', and Grinstead (1972) has derived it from a variant of Chinese 人 'person' with two extra intersecting strokes on the bottom right. (Alas, that variant is not yet in Unicode. Here is a similar variant with three nonintersecting strokes.)

³1.28.21:14 (expanded 22.32): The Jurchen script is an offshoot of the Parhae script of Manchuria. But even if the roots of that script go back westward to the lost 'Serbi script' (to use the term from Shimunek 2017: 121), that script was for Serbi (Xianbei), not a Tibeto-Burman language.

Thirty years ago, Kwanten (1989: 19) wrote,

I have recently come in possession of a number of early T'ang documents written in a script that bears very close similarity with Tangut. These documents will be the subject of a later communication, but they appear to solve the mystery [of the origin of the Tangut script] discussed above. I wish to thank Prof. Edward S.I. Wang of the Chinese Culture University in Taipei for having drawn my attention to these documents.

Unfortunately, to the best of my knowledge Kwanten never wrote about those documents or about Tangut again.

If I assume that those documents (which I have never seen) indeed contained a Tangut-like script from the early Tang, and if I take into account the fact that the Tangut ruling house claimed descent from the Tuoba of the Northern Wei (see Dunnell 1994: 157-158 for a discussion of interpretations of that claim), I can come up with this highly speculative and almost certainly wrong scenario:

- The Tuoba rulers spoke both Serbi and a Tibeto-Burman language (pre-Tangut?)

- The lost Serbi script was an offshoot of the Chinese script designed to write both languages (cf. Pahawh Hmong which was intented to write both Hmong and Khmu, though no examples of Khmu in Pahawh Hmong have survived)

- The Tangut script is a western descendant of this script, and the  Parhae script is an eastern descendant. Khitan and Jurchen large scripts both descend from the Parhae script.

One huge problem with this is that I am unaware of any evidence for any Tibeto-Burman language in the Northern Wei. The Chinese transcriptions of Middle Serbi analyzed by Shimunek (2017: 125-163) are Mongolic-like (Janhunen's Para-Mongolic, a term Shimunek rejects), not Tibeto-Burman.

Another huge problem is that there is no resemblance between the Tangut script on the one hand and the Parhae/Jurchen/Khitan (PJK?) scripts on the other beyond a shared set of Chinese stroke types. No one is going to confuse Jurchen

0176 la

with the Tangut element (not character) 𘦳 'hand', much less the actual Tangut character for 'hand', 𗁅 3485 1laq1.

Tangut Yinchuan font copyright © Prof. 景永时 Jing Yongshi
Tangut character image fonts by Mojikyo.org
Tangut radical and Khitan fonts by Andrew West
Jurchen font by Jason Glavy
All other content copyright © 2002-2019 Amritavision