The following HTML text was created from the original print version. Although care has been taken to transcribe the original correctly, errors may remain. Please refer to the PDF as the text of record for citation. If you encounter a mistranscription, please report it to the editor (

Manchu was first written at the end of the sixteenth century. Spoken Manchu was undoubtedly dialectally differentiated at that time. An interesting question is whether at this time a kind of interdialectal koine had formed among the growing army of the nascent Manchu nation. Since such koines commonly come into being when an army is composed of speakers of different dialects and languages, in this case Manchu dialects, Mongolian, Korean and Chinese. Koine Greek, for example, is commonly thought to have developed in the armies of Alexander the Great.[1] If such a Manchu koine existed, it was most probably the basis of the new written language.

Once a language is reduced to writing, it begins to undergo a development independent of the vernacular, progressively developing its own rules and lexicon which in the course of time comes to differ more and more from vernacular forms of the language.

Let us skip ahead to the Kangxi period (1662-1723). At this time, some seventy years after the creation of the written language, it stands to reason that a certain distance already existed between the written language and the contemporary vernaculars. Let us imagine a young Manchu who spoke a variety of colloquial Manchu at home when he began to learn how to read and write the literary language. In addition to learning the essentials of the script, he would have to remember how things were spelled. In many cases his vernacular and the phonological forms of the written language would have been different. Suppose, for example, our young Manchu said something like [tuiɣə] for 'cloud'; in school he would have to learn to spell the corresponding written word {tugi} correctly.[2] Like a long-suffering English or American student, he would probably simply have to memorize the correct spelling of the word for 'cloud'. On the other hand, he may have learned to read the written word for 'cloud' as [tu'gi] in conformity with its spelling. This is precisely what I discovered when I began to study Manchu with Professor Kuang Lu at Taiwan University in 1965. He read Manchu aloud in strict conformity with the standard spelling, but when he spoke with members of his family, he always used vernacular Sibe pronunciation. When reading aloud, he would pronounce the word for 'six' [ɲiŋun] but when speaking with his family he would say [ɲyŋun]. A reading or school pronunciation existed side by side with a vernacular pronunciation. I believe this dichotomy developed early in the history of Manchu. Therefore, when we speak of Manchu pronunciation, we must distinguish between a school pronunciation based on written forms and various types of vernacular pronunciation. Fortunately several varieties of vernacular Manchu survived into the twentieth century. One of these, Sibe, is still very much alive in Xinjiang. Since most of us are primarily interested in written Manchu, it is its pronunciation that I will concentrate on in this paper.

In written Manchu there are six vowel signs: i e a u o and ū. The grapheme {i} was pronounced [i], as it is in many modern European languages. Judging from transcriptions of Manchu in Chinese, Korean and most importantly from its pronunciation in modern vernaculars, {e} was a central or back unrounded vowel pronounced [ə] or even [ɯ].[3] My teacher, Professor Kuang Lu, also pronounced {e} in this way. The letter {a} was pronounced as a low, back unrounded vowel [a], and {u} was a high back rounded vowel [u]. The fifth vowel {o} was a lower-mid, back rounded vowel [ɔ]. Finally we come to a problem that has bewildered several generations of Manchu scholars, the letter {ū} ({ô} in German transcription). This letter is identical to the Mongolian letter that represents ü/ö. We must remember, when thinking about Manchu, that in the Eastern Mongolic dialects with which the Manchus would have come in contact with pronounced these vowels not as front rounded vowels ([y] and [ø]) but as central vowels ([ʉ] and [ɵ]).[4] However, I do not think that Mongolian holds the key to this problem. Part of the difficulty is that the grapheme {ū} has more than one function. In a few cases, it seems to preserve old spellings inherited from the pre-reformed script; an example of this is the verb {tūmbi} 'to hit, to pound'; this word, as can be seen from variant spellings, was actually pronounced [dumbi].[5] Such archaic spellings tell us very little about the usual pronunciation of the letter {ū}. There are only a few such words and most of them have variant spellings. An examination of this vowel sign reveals that it most often occurs after k, g and h. When we look closer at such examples, we discover that the k, g and h written before the vowel {ū} are actually written with different Manchu letters from those found in the syllables ku, gu and hu. In the Von Mollendorf and German transcriptions two contrastive series of consonants, a velar series [k g x] and a uvular series [q G χ] were confused. In the nineteenth century very few native speakers of Manchu were to be found in Beijing, so when Europeans learned Manchu, they most likely learned their pronunciation from non-native speakers who lacked the original native distinction between velars and uvulars. At any rate, this is a difficult distinction for speakers of European languages, so even if they had heard the relevant letter pronounced correctly, they would probably have been unable to reproduce the distinction. In other words, they heard ku and kū as the same; but since words containing such syllables were still distinct graphically, the transcriptions had to remain distinct. So what was the actual difference between minimal pairs such as suku 'tumbleweed' and sukū 'skin'? The first word suku is "front vocalic" and was pronounced [suku] with a velar [k]; the second word was "back vocalic" and was pronounced [suq℧] (cf. Sibe [soqw] and Sanjiazi ['sɔ:qɔ]). Hence, we see that when {ū} follows k, g or h, it is a high back rounded vowel. When it occurs in this position, it is pronounced somewhat lower than [u] and is more pharyngealized. However, in such cases {u} and {ū} are not distinct phonemes; the difference in quality is caused by the "throaty" nature of the uvular sounds.

Now, some further comments on {o}. As an independent vowel, as pointed out above, {o} was pronounced [ɔ], but {o} also has other functions. As the second member of a diphthong it is pronounced [u]: ao [au], eo [əu], oo [ɔu]. It is probable that {oo} was pronounced [ɔ] in some varieties of Manchu; cf. boo 'house' Sibe [bɔ], Sanjiazi [bɔu] ~ [bɔ]. In Professor Kuang's reading pronunciation {oo} was always read [au]. The combination {io} is ambiguous; in back vocalic words it represents [jɔ] but in front vocalic words it was most likely read [ju]. Cf.

Manchu Sibe Sanjiazi

niohe 'wolf' [juxw] [ 'niu:ɣo]

nioron 'rainbow' [niorun] [ 'niɔ:rɔn]

The imperative form jio 'come!' was probably pronounced ju:

ManchuSibe Sanjiazi

jio 'come!'[dʐu] [dʐɔ]~ [dʐu]

Note in this regard that the imperative form for verbs ending in the suffix -njimbi is -nju. We should also take into account the imperative of gajimbi which is gaju. One wonders how the io in the ending -mbio was pronounced. Here the modern vernaculars are of no help since they do not employ the interrogative suffix -o. In the Qingwen qimeng (Wuge 1730) the Chinese transcription of -mbio was 密幽, suggesting that genembio may have been pronounced as [gənəmiju].[6]

The consonants {b}, {d} and {g} were voiceless unaspirated sounds when occurring initially. This seems to be an areal feature since we can observe the same situation in both Northern Chinese and Mongolian. Intervocalically they were undoubtedly voiced. Moreover, {b} and {g} may have been voiced fricatives in this position, {b} becoming [v] and g becoming either [ɣ] or [R] depending on the vocalism. This is what we find in modern vernaculars:

Manchu Sibe Sanjiazi

aga 'rain' [ARA] [a:Ra]

yabumbi 'go' [javəm] [javume]

Intervocalic {d} was simply voiced.

It seems likely that intervocalic {g} and {h} were pronounced the same way. In traditional texts many variant spellings suggest this: buge ~ buhe, dogo ~ doho, cige ~ cihe, sigambi ~ sihambi, etc. The modern vernaculars also bear our this supposition:

Manchu Sibe Sanjiazi

agūra ~ ahūra 'weapon' [aRur] [aRur]

boihon 'earth' [biɔRun] [biɔRɔn]

jugūn 'road' [dʐɔRun] [dʐɔRɔn]

juhe 'ice' [dʐuɣo] [dʐuɣo]

yohoron 'ditch' [jɔRurun] [jɔyɔrɔ]

cige ~ cihe 'louse' [tɕiɣə] [tɕiɣ]

A similar situation can be observed with {b} and {f}. In vernacular Manchu both intervocalic {b} and {f} are pronounced [v].

Manchu Sibe Sanjiazi

aba 'hunt' [avə] [a:v]

ubu 'part' [uv] [u:vo]

jube 'story' [dʐuvo] [dʐu:vo]

gebu 'name' [gəv] [gəvə]

ifimbi 'sew' [ivim] [ivi:me]

ufa 'flour' [uva] [uva:]

jefu 'eat!' [dʐəv] [dʐəvə]

hafan 'official' [χavən] [χavʌn]

In the case of {b} and {f}, however, we hardly find any variant spellings; this suggests that in early Manchu they were still kept separate intervocalically.

In vernacular Manchu {ngg} is pronounced either [ŋ] or [ŋŋ], that is, it lacks a velar or uvular stop [g] or [G]. In the Qingwen qimeng this also appears to be the case. For example, in this work tanggū is transcribed 湯屋, which implies [taŋu]; further examples from the Qingwen qimeng: nanggin 那英因切 [naŋin], onggolo 惡硬窩囉 [oŋolo]. The examples below illustrate the situation in Sibe and Sanjiazi.

Manchu Sibe Sanjiazi

angga 'mouth' [aŋ] [aŋŋa]

inenggi 'day' [inəŋ] [iniŋŋə]

ninggun 'six' [nyŋun] [niuŋŋun]

mangga 'hard' [maŋ] [maŋŋa]

Dagur contains many Manchu loanwords. Some of these were probably borrowed directly from the written form of Manchu but others were undoubtedly borrowed from a vernacular form of Manchu. In these loanwords, regardless of their origin, {ngg} is pronounced [ŋg]: Manchu enduringge is [əndu:riŋgə]; mangga is [maŋgə]; ninggun is [niŋgun] (the last form occurs in [niŋgunbe:] 'the sixth month'. The word monggo is clearly a loan from Mongolian mongɣol; the {ngɣ} in this word is pronounced [ŋg} in virtually all the Mongolic languages and dialects: Chahar [moŋgol, Kharchin [moŋgol], Alashan [moŋgol] (Sun 1990). It is likely that Dagur in this case preserves a general Mongolic pattern, that is, it has substituted a native phonetic habit when pronouncing words containing {ngg}. For this reason I prefer the reading [ŋ] for {ngg}, although I see no reason to reject the Dagur and general Mongolic style pronunciation [ŋg].

Finally I would like to comment on stress. In written Manchu there is no indication of where stress falls. In vernacular Manchu the situation is not entirely clear. Both Sibe and Sanjiazi agree in placing final stress on words ending in {n}. For words that in written Manchu end in vowels, the evidence is unclear and somewhat inconsistent.

In two syllable words ending in a vowel the stress sometimes falls on the first syllable and sometimes on the second or final syllable:

Manchu Sibe Sanjiazi

jaka 'thing' [dʐaq] [ 'dʐa:qa]

baita 'matter' [bait] [bait]

leke 'whetstone' [lək] [lək]

doro 'rite' [dor] [dɔ:rɔ]

sabu 'shoe' [sav] ['sa:ve]

The words above were clearly accented on the first syllable.

Manchu Sibe Sanjiazi

muke 'water' [mu 'ku] [mu 'kuo]

sunja 'five' [sun 'dʐa] [sun 'dʐa:]

jiha 'money' [dʑi 'Ra] [dʑi 'Ra:]

buda 'food' [bə 'da] [bu 'da:]

ilha 'flower' [il 'Ra] [il 'Ra:]

hu/da 'price' [χu 'da] [χu 'da:]

The words in this second group were clearly accented on the second syllable. It is notable that in all these cases the first syllable contains either {i} or {u} in the first syllable (Qingeertai 1998: 260) whereas those with the stress on the first syllable have other vowels. However there are exceptions to this apparent regularity:

Manchu Sibe Sanjiazi

onco 'wide' [on 'tȿu] [ɔn 'tȿɔ:]

amba 'large' [am 'bu] [am 'ba:]

beki 'firm' [bə 'ki] [bə 'ke]

In addition there are cases where Sibe and Sanjiazi have different stress patterns:

Manchu Sibe Sanjiazi

cihe 'louse' [tɕi ‘ɣə] [tɕi:ɣ]

lefu 'bear' [tɕi ‘ɣə] [lə’və]

jalu 'full' [dʐa 'lu] [ 'dʐɔulo]

turi 'bean' [ty 'ry] [‘tyre]

While there are some rules and general tendencies that can be observed about Manchu stress, it appears impossible to reconstruct the entire system on the basis of modern vernaculars. The main reason for this is that not all Manchu words have survived in Sibe and Sanjiazi. Moreover, the evidence, as shown above, was ambiguous. Professor Kuang Lu, when he read Manchu aloud, generally stressed words on the last syllable. Since many final vowels were lost in colloquial pronunciation, this practice has the merit of helping a student remember what the proper final vowels are in the written language.


Browning, Robert. 1969. Medieval and modern Greek. Cambridge: Cambridge University Press.
Engkebatu et al. 1984. Daɣur kelen-ü üges. Hohhot: Nei Menggu Chubanshe.
Engkebatu. 1995. Mǎnyǔ kǒuyǔ yánjiū. Hohhot: Minzu Chubanshe.
Li Shulan et al. 1984. Xībóyǔ kǒuyǔ yánjiū. Beijing: Minzu Chubanshe.
Li Shulan et al. 1986. Xībóyǔ jianzhì. Beijing: Minzu Chubanshe.
Qingeertai (Cinggeltei). 1991. Měnggǔya yǔfǎ. Hohhot: Nei Menggu Renmin Chubanshe.
Qinggeertai (Cinggeltei). 1998. Mínzú yánjiū wénjí. Beijing: Minzu Chubanshe.
Sun Zhu. 1990. Měnggǔ yǔzǔ yǔyán cídiǎn. Xining: Qinghai Renmin Chubanshe.
Wuge Shouping. 1730. Qīngwén qǐméng.

1. According to Browning (1969) an "expanded Attic" began to develop in the late fifth century B.C.; it was on the basis of this language that the Koine developed and was spread by Alexander's armies. Perhaps such a comparable koine had already begun to develop among the Tungusic tribes of the Ussuri Basin in the period preceding Nurgaci's unification efforts and was subsequently spread by his military campaigns.return to text

2. In the present day Sanjiazi dialect of Manchu, 'cloud' is pronounce [tuiyo]; the Qingwen qimeng of 1730 shows a similar pronunciation.return to text

3. I will distinguish elements of the script (graphemes) from actual pronunciations by placing the former in curly brackets and the latter in square brackets.return to text

4. For the modern Mongolian pronunciation see Qinggeertai (Cinggeltei) (1991:71)return to text

5. There are many examples of the spelling {dumbi} in the Manchu translation of the Sanguo yanyi.return to text

6. The question particle -o is probably a borrowing of Mongolian -u, having the same meaning; Sibe uses

na for this function; Sanjiazi uses nɔreturn to text