Last night's post dealt with the Taiwanese surname 偕 Kai (Mandarin Jie; normally read xie) 'together' which consists of 亻 'person' (semantic) plus 皆 Tw kai, Md jie 'all'.

The two words are obviously cognate and belong to the same Old Chinese (OC) word family (Schuessler 2007: 310-311):

皆  Md jie < OC *kri 'all'

偕 Md jie < OC *kri(ʔ), Md xie < OC *N-kri 'together'

諧 Md xie < OC *N-kri 'be harmonious'

and perhaps

階 Md jie < OC *kri or *krəj 'steps, stairs' (a harmonious collection of stairs?)

which may be related to 陔 Md gai < OC * 'steps; stair' (Schuessler 2007: 248)

I don't understand the structure of the character 皆 for the root which Shuowen analyzes as

比 Md bi < OC *piʔ 'compare'

白 Md bai < OC *brak 'white'

Neither sound like 皆  Md jie < OC *kri 'all', so neither can be phonetic.

However, on closer examination, 比 can also mean 'be associating' (cf. 'together') or 'equal; similar' (cf. 'be harmonious'). It also has the OC readings

*piʔ-s 'to put together, match, assemble'

*N-piʔ-s 'to join; go together with, close together, successive [like stairs?]'

But I am still puzzled by 白 'white'. 大修館新漢和辞典 (not the best source, I know) says 皆 represents everyone (比 lined up) 白 speaking. 白 has a secondary meaning 'to declare': i.e., make clear/white. However, 皆 'all' is not a verb, much less a verbal verb, so 白 seems irrelevant.

If I had to pick a second component for 皆, it would be 立 'stand': 'standing together' = 'all'.

I'd suggest 众 (three 人 people) as a graph for 'all', but it  already stands for an unrelated word 'be numerous, all, multitude' (also written 眾) OC *tuŋs.

At least I grasp half the structure of 皆 'all', whereas the analysis of the Tangut character for zi 'all' is unknown:

That tangraph must be the basis for

to 'all' (borrowed from Middle Chinese 都 *to 'all'?)

Note the mysterious 'horned hat' in place of 二 on top.

Together they represent the redundant compound

to zi 'all' (lit. 'all-all')

The shared bottom components of to and zi look like they could comprise a third tangraph, but I can't find any such character.

Next: Taking apart to and zi. 偕萬來 TEN THOUSAND COMING TOGETHER

After writing about Korean baby names, I revisited the various East Asian name articles on Wikipedia. One passage in the Chinese name article stood out:

Among the Taiwanese Presbyterian Christians, the family name 偕 (Kai in Taiwanese Pe̍h-ōe-jī [白話字 romanization]) is of particular interest as an example of a Chinese-like surname with a non-Chinese root. According to the clan's tradition, the name was adopted to honor the Canadian missionary George Leslie Mackay, also known as Má-kai (馬偕). This family name is actually rarely seen even among Presbyterian Christians. Taiwanese Christians of other sects do not carry this tradition.

I have never heard of Chinese adopting an invented surname in modern times, so I investigated further and discovered that the surname 偕 Kai (Mandarin Jie) had been adopted by some of the indigenous 噶瑪蘭族 Kavalan of Taiwan. Until recently their existence was not officially acknowledged:

Following a traditional ritual of blessing to repel evil spirits, the Taiwan government declared that the Kavalan people--most of whom are plains aborigines living on the island's East Coast--were an equal and legitimate member of Taiwan's Austronesian aboriginal family. The tribe thereby became the 11th official aboriginal group.

"The government is returning to you the indigenous status and tribal name that rightfully belong to you," said Premier Yu Shyi-kun, while offering a government apology to the Kavalan people. "I hope hereafter that the age-old language and cultural heritage of the Kavalan community will be well preserved and passed on to future generations."

I predict the non-Chinese language of the Kavalan (distantly related to Indonesian and Hawaiian) will be preserved in publications like this dictionary instead of being passed on as a living language. There were only 24 speakers in 2000, and I presume some have passed away by now.

When most people think of the island's indigenous groups--which number 413,513 people or less than 2 percent of Taiwan's population--they think of the Atayal, Saisiyat, Bunun, Tsou, Shao, Paiwan, Rukai, Puyuma, Ami and Yami. The common misconception is that aboriginal tribes were mostly mountain-dwellers.

In fact, official recognition of aborigines extends only to mountain peoples who are thought to exhibit cultural uniqueness. Plains-dwelling tribes, on the other hand, have not yet had their ethnic identities officially recognized ...

... [T]he achievement [of Kavalan recognition] did not happen overnight. It followed a time-consuming and difficult journey against social prejudice. About 17 years ago, the then middle-aged Chieh Wan-lai walked into the Ilan County Culture Center, claiming Kavalan status.

His words led to further investigation into whether Kavalan was an existing demographic designation on the island. Chieh therefore moved to help restore the identity of his tribe as a separate ethnic identity.

"The Kavalan were once regarded as a dead ethnic group. With assistance from academic and official sectors, however, the group has been revived," enthused the 72-year-old Chieh. "Wanan," said Chieh, using his mother tongue to say "thank you." For him, having the same legal status as other aboriginal groups is only the beginning. The Kavalan people have another sacred mission: saving their language and culture from extinction.

Chieh must be at least quadrilngual. His family probably knew Taiwanese as well as Kavalan (see below) and he lived through the Japanese colonial period and the switch to Mandarin under the Kuomintang.

The World Summit of Indigenous Cultures site reveals a bit about his background:

Today, the preservation and protection of genuine Kavalan culture and customs is a matter of the utmost urgency, and no one has done more for this cause than Mr. Chieh Wan-lai of the Hsin She community. His father was a Kavalan missionary trained by Dr. Mackay — which is where the character “chieh” for his family name comes from (the [Mandarin] Chinese transcription of “Mackay” is “ma-chieh” [equivalent to Taiwanese Má-kai].

Mackay's biography doesn't mention him mastering Kavalan, so I presume that he spoke to Chieh's father in Taiwanese.

I wonder if Chieh is still alive. He wasn't in good health in 2003:

The Kavalan elder Chieh Wan-lai, who has been suffering from kidney problems for many years, is also present, physically supported by members of his tribe. Now well over 70, Chieh Wan-lai has in recent years rarely left his ancestral home on the coast of Hualien.

The title of this post refers to his name:

Gloss together ten thousand come
Taiwanese kai bān lâi
Mandarin (Wade-Giles romanization) chieh wan lai
Mandarin (Pinyin) jie wan lai

chieh/jie is more commonly read as xie. I'll discuss its two readings and etymology next time. THE SOUNDS OF FEMININE TRUTH (PART 2)

Daniel Kane (1989: 115, 251) reconstructed Ming Jurchen *dz on the basis of a single Ming Mandarin transcription of the Ming Jurchen word for 'scissors':


Ming Mandarin *xatsa

Kane's Ming Jurchen *xadza or *xadʒ(x)a

cf. Kiyose's Ming Jurchen *xadʒixa on the basis of another MM transcription (*xatsɨxa?)

cf. Manchu χasaχa

Note how MM unaspirated voiceless *ts corresponds to MJ *dz. The reconstruction of *dz fits the general pattern of MM transcriptions of MJ:

MJ *p [pʰ] *t [tʰ] *tʃ [tʃʰ] *k [kʰ] (~ [qʰ]?)
MM transcription *pʰ *tʰ *tʃʰ *kʰ
MJ *b *d *dʒ *g [g] (~ [ɢ]?)
MM transcription *p *t *tʃ *k

(I'll discuss the question of whether MJ had uvulars later. Kane's reconstruction has no uvulars.)

Like modern standard Mandarin, MM had no voiced obstruents. (I am ignoring the [ʐ]-type variant of MSM /r/.) The transcription of MJ voiced obstruents with MM voiceless unaspirated obstruents has parallels in MSM borrowings like

柏林 Bolin [pwɔlin] 'Berlin' (but note 巴黎 Bali [pali] 'Paris', 巴基斯担 Bajisidan [patɕisz̩tan] 'Pakistan' - French and Urdu p are unaspirated)

匹兹堡 Pizibao [pʰitsz̩paw] 'Pittsburgh' (not entirely consistent; I'd expect Picibao [pʰitsʰz̩paw]).

MJ clearly must have had aspirated obstruents, but romanizations of Manchu ignore the aspiration, implying that it was nonphonemic. I wonder if MJ (and some varieties of later Manchu, with or without Mandarin influence?) had an MM-like opposition based on aspiration instead of voicing. If so, then romanized Manchu has Latin voiced letters corresponding to voiceless sounds: e.g., b [p].

Kane's MJ *dz is distinct from his *z (which he described as phonetically [dz]!). I will discuss his *z in part 3. For now, I only want to propose MJ *χas(a)χa with *-s- or even *χaz(a)χa with *-z- as alternate interpretations of the transcription of 'scissors' as 哈雜. *z could be an intervocalic allophone of MJ */s/ (cf. Middle Korean z < intervocalic *s) absent in Manchu.

If the MJ word for 'scissors' were disyllabic *xadza, *xadʒ(x)a, *χasχa, or *χazχa, it cannot be the ancestor of Manchu trisyllabic χasaχa unless Manchu -a- were inserted later to break up a consonant cluster. MJ *dʒ otherwise corresponds to Manchu dʒ, not s.

Perhaps the original word was *χaCχa and different dialects inserted different vowels (Kiyose's MJ *-i-, Manchu *-a-) to break up the cluster *-Cχ-. An even more complicated possibility is that the original word was *χaCiχa (cf. Kiyose's reconstruction) which was reduced to *χaCχa and then later expanded to Manchu χasiχa with a new epenthetic vowel.

If the MJ word for 'scissors' were trisyllabic *xadʒixa, it cannot be the ancestor of Manchu χasaχa unless MJ *i became Manchu a, which normally corresponds to MJ *a. Kane's (1989: 117) list of unusual MJ : Manchu vowel correspondences does not include MJ *i : Manchu a, though his list is not comprehensive*.

I don't think the dialect of MJ reconstructed by Kane is the ancestor of Manchu because there are too many unusual correspondences that regular sound changes cannot account for**.

10.23.0:54: I can't find any Tungusic cognates for the Jurchen and Manchu words for 'scissors' in Starostin's database.

10.23.1:03: There is no connection between Japanese hasami 'scissors' and Manchu χasaχa 'scissors', since Japanese *h is from *p and Manchu (and Jurchen) χ is from Proto-Tungusic *k. (Alexander Vovin derived Manchu k from PT *Nk, but I don't understand how Manchu words like kalimu 'whale' can have initial k- unless these are archaisms or PT had initial prenasalized stops.)

*10.23.1:28: I found one case of MJ *i : Manchu a:

'boat' (Kane 1989: 109, 257)

MJ *dixa (transcribed as MM 的哈 *tixa) (< *dijaχa?)

Manchu dʒaχa

See the following footnote regarding MJ *d corresponding to Manchu dʒ.

**10.23.1:54: There is no doubt that MJ *t and *d palatalized before *i, resulting in Manchu affricates (Kane 1989: 109):

'pillow' (Kane 1989: 109, 254)

MJ *tirgu (transcribed as MM 替兒古 *tʰir̩ku)

Manchu tʃirku (not tʃirgu)

'blanket' (Kane 1989: 109, 334; not in his index!)

MJ *dibexun (transcribed as MM 的伯洪 *tipəxuŋ)

Manchu dʒibəxun

'to hear' (Kane 1989: 109, 290; also not in his index)

MJ *dondi- (transcribed as MM 斷的 *tonti)

Manchu dondʒi- AGIIRUM

looks like a Latin word, but it's a Korean phrase:

아기 agi 'child'

이름 irŭm is 'name'

agiirum.co.kr is an online dictionary of the 5,151 characters officially permitted in Korean baby names:

- the 1,800 characters taught in schools from 2000 to present

- 3,079 additional characters approved for baby names in 1991-2003

- 159 additional characters approved for baby names in 2005

- 159 additional characters approved for baby names in 2007

Here's a list of the characters pronounced 가 ka (with or without a final consonant) and 개 kae. The columns indicate when a character was approved for use.

The approved list doesn't seem to include any hanja-hangul hybrids, so some retro names with non-Sino-Korean syllables can't be recycled.

I am amazed that the approved list includes

惡 악 ak 'evil' (one of the 1,800 characters taught in schools)

魔 마 ma 'devil' (not in the 1,800 list; added in 2001 - who requested this?)

邪 사 sa 'evil' (one of the 1,800)

殺 살 sal 'kill' (one of the 1,800)

and others I would rather not type.

Maybe such graphs are supposed to be in names opposing evil like

懲惡 징악 Ching-ak 'punish evil'

降魔 항마 Hang-ma 'vanquish devils'

破邪 파사 Pha-sa 'smash evil'

莫殺 막살 Mak-sal 'don't kill'

but I don't remember seeing names like those.

No, actually, I do. Almost. There's a character in the Chinese (not Korean) comic book 中華英雄 Chinese Hero called


Situ Mowen 'Don't-Ask Situ'

(this could be Koreanized as 사도막문 Sado Mak-mun)

But 問 'ask' is still not a negative word like 殺 'kill'.

While Googling 莫殺 'don't kill', I found a Chinese saying


Qiemo sha ji qu dan.

'be-sure-to-don't kill chicken take egg'
Can you guess its English equivalent?

The first graph 切 qie 'be sure to' usually means 'cut' (and contains a 刀 knife).

Can you guess the meaning of this example sentence from 重編國語辭典修訂本?


Jiu hou qiemo kai che, yi mian fasheng weixian.

'liquor after be-sure-to-don't open car, to avoid emit-birth danger-danger'

I'm deliberately translating morpheme by morpheme to make the overly easy mildly interesting. THE SOUNDS OF FEMININE TRUTH (PART 1)

(The title refers to the Chinese transcription of 'Jurchen' as 女眞 'woman' + 'true'.)

One might think that the reconstruction of Jurchen, a language often thought to be ancestral to Manchu*, would shed light on the mystery vowel of Manchu (last seen here). Unfortunately, there is no recoverable *u : distinction in Daniel Kane's (1989) reconstruction based on the Sino-Jurchen vocabulary of the Ming Dynasty's Bureau of Interpreters. Kane reconstructs only five vowels for Ming Dynasty Jurchen (MDJ):

*i *u
*e *o

The placement of the vowel symbols in the table is mine. Kane does not state whether he believes MDJ *e was phonetically *[ə] as in Manchu. Perhaps the table could be rewritten as

*i *u

MDJ *u corresponds to both Standard Manchu u and ū.

The absence of in MDJ can be interpreted in at least two ways:

A. MDJ merged *u and *ū, whereas the ancestor of Old and Standard Manchu preserved the distinction.

B. MDJ had an *u : distinction which could not be transcribed in Ming Dynasty Mandarin, a language that lacked such a distinction. The inability to record or recover a distinction does not necessarily entail its absence.

Kane's inventory of MDJ consonants is similar to that of Standard Manchu. Kane's symbols are in parentheses when they differ from the IPA.

*p *t *ts *tʃ (č) *k *'
*b *d *dz *dʒ (ǰ) *g
*m *n *ŋ (ng)
*f *s *ʃ (š) *x (h)
*z *ʒ (ž)
*w *l, *r *j (y)

Two MDJ consonants that have no Manchu counterparts are *' (intervocalic hiatus) and *z. Kane lists hiatus as a velar, though I suspect hiatus might have phonetically been a glottal stop *ʔ. I'll discuss another possible phonetic value in a future post. I'll also explain my doubts about MDJ *z, *dz, and (but not *dʒ!) later.

*Alexander Vovin and I think that Standard Manchu is not the direct descendant of the Jurchen dialect(s) that have been been transcribed in Chinese, though those languages are obviously very closely related. I may explain my reasoning in the future. 夜音 THE SOUND OF NIGHT

㖱 is an interesting hanja-hangul hybrid because it has two readings (얌 yam, paem). The first reading is a straightforward combination of the reading of the top and bottom halves:

夜 야 ya 'night' + the hangul letterㅁ m = 㖱 얌 yam

One might guess that 夜 had a second reading 배 pae, but it doesn't. However, the native Korean word for 'night' is 밤 pam. Hence 㖱 can also be interpreted as

夜 밤 pam.'night' + the hangul letterㅁ m = 㖱 뱀 paem [pɛm] < paym in earlier Korean)

A slight mismatch between the readings of a character and its phonetic is normal in sinography.

I'm surprised that 㖱 isn't also read as 밤 pam. Maybe 'night' wasn't in any premodern Korean names.

I would have expected a combination of a 배 pae graph with ㅁ m for 뱀 paem, but pae graphs were rare until modern times*.

In 향가 (鄕歌) hyangga, the earliest Korean poetry, there are several sinographic spellings of 'night':
Source Spelling Premodern Sino-Korean readings My reconstructed Old Korean Gloss
處容歌 line 8 (ya) *pam 'night'
慕竹旨郎歌 line 2 夜音 (ya)-ɯm
薯童謠 line 4 夜矣 (ya)-ɯy *pam-ɯy 'at night'
請佛住世歌 line 5 夜未 (ya)-mɯy

Premodern Sino-Korean ya is in parentheses because it is irrelevant to the interpretation of the native word for 'night'.

Conversely, the meanings of 音 'sound', 矣 'final perfective particle', and 未 'not yet' are irrelevant to the Old Korean sounds that they represents.

Old Korean *-ɯy (or *-əy?) is a locative case postposition cognate to Modern Korean -에 -e.

It is not absolutely certain that the first spelling 夜 represented *pam because it has no second graph confirming a final *-m. And one could even propose that some or all of these spellings represented an unrelated Korean *-m word meaning 'night' that is now extinct, though it is simpler to assume that the Old Korean *-m word for 'night' is the ancestor of Middle and Modern Korean pam 'night'.

The second spelling has a 音 representing final *-m. This 音 also appears in

'heart' + *-m (隨喜功德歌 line 10)

thought to represent the Old Korean ancestor of MIddle Korean mʌzʌm 'heart'. 音 is one of several clarifiers in Old Korean added to core characters to distinguish native Korean words from Chinese loans:

core character by itself = Chinese loan (Sino-Korean)

夜 SK ya 'night'

心 SK shim 'heart'

core character + clarifier = native Old Korean word ending in sound(s) represented by the clarifier

OK *pam 'night'

OK *mʌCɯm 'heart' (middle consonant uncertain)

Next: Core-clarifier composites.

*10.20.1:22: In modern Sino-Korean, there are many graphs read as 배 pae, but nearly all of these graphs were read as ᄇᆡ pʌy in premodern SK. The only exception I can quickly find in 새字典's index of premodern readings is 牌 which is normally read 패 phae with an aspirated initial. 牌/ㅁ might have been misread as 팸 phaem. I wish I still had copies of 東國正韻 (though its readings are idealized) and 全韻玉篇 with me. Martin (1992: 129) lists Middle Korean 배 pay (> later pae) as a possible syllable with three different tones in 東國正韻 but doesn't list any graphs. HANJA-HANGUL HYBRIDS

Last week, Andrew West asked me about two classes of made-in-Korea 'Chinese' characters. I'll deal with the second next time. The first consists of hanja-hangul hybrids with the structure:

top: hanja (Chinese character) representing CV syllable

bottom: hangul letter representing final consonant

Hybrid 국자 (國字) kukcha 'national characters' (i.e., made-in-Korea characters) are 음역자 (音譯字) ŭmyŏkcha 'sound translation characters' transcribing native Korean syllables:

-ㄱ -k -ㄴ -n -ㅁ -m -ㅇ -ng
加 가 ka 㔖 각 kak 㔔 강 kang
巨 거 巪 걱 kŏk
斗 두 tu 㪲 둑 tuk 㪳 둥 tung
豆 두 tu 䜳 둔 tun
老 로 ro 㖈 놈 nom (not rom!)
於 어 ŏ 㫇 억 ŏk 㫈 엉 ŏng
夜 야 ya 㖱 얌 yam, paem
者 자 cha (<쟈 chya) 䎞 작 chak (earlier쟉 chyak?)

I compiled this table of hybrids using these two lists. I welcome any additions.

Modern Korean names no longer include hybrids and tend to be Sino-Korean. (The obvious exception is 서울 Seoul which is a native Korean word.)

I've known about hybrids since at least 1996, but I never saw them in use until tonight, when I found a few Chosŏn Dynasty names containing hybrids in this list:

㔖白 각백이 Kakpaegi

㔖孫 각손이 Kaksoni

㪲間 둑간이 Tukkani

The name suffix -이 -i is unwritten in hanja.

The first of those names is in this manuscript by 박재춘 (朴栽春) Pak Chae-chhun:

朴㔖白 박각백 Pak Kakpaek (각백이Kakpaegi?)

(I can only see the typed transcription and not the scan of the original text.)

Some of the hybrids puzzle me:

1. Hybrids were created for

- syllables that did not exist in Sino-Korean: e.g., 㪲 둑 tuk

- syllables that did exist in Sino-Korean: e.g., 㔖각 kak (the SK reading of 各, 角, 覺, etc.)

I understand why new graphs were needed to unambiguously write non-Sino-Korean syllables, but I don't understand why 각백 Kakpaek couldn't be transcribed with nonhybrids as 各白, 角白, 覺白, etc. If hybrids for syllables that existed in Sino-Korean were intended to be semantically vacuous, why not write the entire name with hybrids instead of mixing the two?

2. 놈 nom is 㖈 with 老 로 ro 'old' rather than 奴 노 no 'slave' on top. This choice reflects how r- was pronounced as n- in word-initial position. (Modern Korean allows initial r-.) It may also reflect an unwillingness to use a semantically negative phonetic.

There is also a native Korean word 놈 nom 'guy'. I wonder if this is the source of nom in names and if it was ever written as 㖈.

I also wonder if 㖈 can be read as 롬 rom.

3. 㖱 has two readings, the expected 얌 yam and 뱀 paem which has a 배 pae unlike 夜 야 ya. I'll discuss this character further tomorrow in "The Sound of Night".

10.19.0:07: You may have noticed that the table only contains hybrids for syllables ending in -k and nasals (-m, -n, -ng). Why aren't there any hybrids for syllables ending in other consonants: e.g., 돌 tol? I plan to tackle this question on Tuesday.

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