There are no dialects in strict sense,[clarification needed] but three local varieties (tuyu) can be found: Suonanba (ca. 50% of all Dongxiang speakers), Wangjiaji (ca. 30% of all Dongxiang speakers) and Sijiaji (ca. 20% of all Dongxiang speakers).
Phonology
Compared to other Mongolic languages, there is no vowel harmony, except for a handful of suffixes. However, the rules governing this phenomenon are by far not as strict as those of Mongolian.[3]
The aspirated stops /pʰ,tʰ,kʰ,qʰ/, at varying degrees, tend to be affricated before close vowels as [p͡ɸʰ,t͡θʰ,k͡xʰ,q͡χʰ] respectively:[5]
/qʰ/→[q͡χʰ]: This has the most noticeable affrication, seen in ghudogvo[q͡χʰutoˈʁo] (knife).[6]
/pʰ/→[p͡ɸʰ]: Still commonly affricated, e.g. pixie[p͡ɸʰi̥ˈɕi̯ɛ] (belt).[7]
/tʰ/→[t͡θʰ]: Affrication is salient, but seems to be more apparent with [ɯ] than [u], as in tiigha[t͡θʰɯ̥ˈqɑ] (chicken).[7]
/kʰ/→[k͡xʰ]: This consonant is the least affricated, but is still attested in kiqie[k͡xʰi̥ˈt͡ɕi̯ɛ] (to lie down).[6]
Likewise, before non-close vowels non-aspirated stops do not show affrication e.g. da[tʰɑ] (to guess), korolon[kʰoroˈlõŋ] (footprint), or khalun[qʰɑˈlũŋ] (hot).[5]
Phonemes /t͡sʰ,t͡s,ʐ/ are only found in loanwords:
/t͡sʰ/ is only found in Chinese-derived words, as in cai/t͡sʰɑj/ (vegetable; from Chinese 菜cài). It is frequently deaffricated to the native [s].[8]
/t͡s/ is found within Chinese and Persian borrowings, with Persian زمینzamin (earth) rendered as zemin/t͡səˈmi(ŋ)/.[8]
/ʐ/ in the same way is seen in words of Chinese and Arabic origin, as in renshen/ʐəŋˈʂəŋ/ (ginseng; Chinese 人參rénshēn).[9]
The nasals /n,ŋ/ assimilate to [m] before labials, as in unba/unˈpɑ/→[ũmˈpɑ] (to swim) and banban/pɑŋˈpɑŋ/→[pɑ̃mˈpɑ̃ŋ] (cudgel) respectively. /ŋ/ also has another allophone [ɴ] before uvulars, cf. zhangha/t͡ʂɑŋˈqɑ/→[t͡ʂɑ̃ɴˈqɑ] (walnut).[9]
Nasals are commonly deleted in the coda, which is an areal feature around the border between Gansu and Qinghai. Although it is not restricted to word-final situations, it is rarer to find deletion there as the following consonantal onsets somehow reinforce the nasal. It should also be noted that vowels preceding nasal consonants are also nasalized themselves.[10]
The central approximants /j,w/ are fricativized [ʝ,v] when syllable-initial, mainly before close vowels. /w/ can further be found before front vowels and even low back vowels. An example of fricatization for each is yibai[ˈʝipɛi̯] (one hundred) and weila[və̝i̯ˈlɑ] (to cry).[11] This phenomenon can also be found in neighboring languages, including Dungan and Monguor.[12]
/ʁ/, which is usually realized as a fricative, can be pronounced as an approximant [ʁ̞] in fast speech.[9]
/r/ is usually a trill, and length is dependent on the speed of speech; similarly in rapid speech it can become a tap [ɾ].[13]
Vowels
Dongxiang has 7 vowels.[4] Unlike other neighboring Mongolic languages, it has limited vowel harmony and no distinctions of vowel length.[3]
/i/, which is usually a close front vowel, has a backened allophone [ɨ] after retroflex consonants, and a slightly opened one [ɪ] after alveolar affricates and fricatives. Compare chi[t͡ʂʰɨ] (you [2nd sing.]) with misi[miˈsɪ] (to fly).[14]
Close vowels are devoiced when in an unstressed initial syllable and between voiceless consonants, for example: chighin[t͡ʂʰɨ̥ˈqə̃ŋ] (ear), sidara[sɪ̥tɑˈrɑ] (to catch fire) and khidun[q͡χʰɯ̥ˈtũŋ] (hard). This is not the case though before a voiced consonant, especially if it is of native Mongolic origin, as evidenced with chila[t͡ʂʰɨˈlɑ] (to be exhausted) and suma[suˈmɑ] (arrow).[15]
In Chinese loans, devoicing is optional, i.e. chugui (cupboard; from 櫥櫃chúguì) can be pronounced either [t͡ʂʰu̥ˈku̯ə̝i̯] or [t͡ʂʰuˈku̯ə̝i̯].[15]
The mid-central vowel /ə/ has also been recorded voiceless in the aforementioned environment: pese[pʰə̥ˈsɛ] (again).[15]
Devoicing of /i/ can also occur regardless of the following sound and in non-initial syllables, under the condition it occurs after /s/: basi[ˈpɑsɪ̥] (tiger) or kewosila[kʰəˌwɑsɪ̥ˈlɑ] (children). Interestingly, it is not often the case, as in bositu[pʷosɪˈtʰu] (pregnant).[15]
/ɚ/ is always a syllabified as its own, as in ershi[ɚ.ʂɨ] (twenty), with the exception of one Arabic loanword with glottal /h/ as the onset: ruhher[ʐu.hɚ] (spirit; from رُوحrūḥ).[16]
/o/ is often pronounced with a considerable degree of labialization [ʷo], however this can be inconsistent, even in the same utterance; cf. olien[ʷoˈli̯ɛ̃n] (cloud) vs. onghono[õɴqõˈno] (to shout).[16]
When stressed, /ə/ is typically realized as mid-front [ɛ]: enzhegve[ə̃nt͡ʂəˈʁɛ] (donkey). Before /ŋ/, [ɤ] is used instead: amen[ɑˈmɤ̃ŋ] (rice). Otherwise in unstressed syllables it is mid-central [ə].[16]
[ɑ] is typically an open back vowel, e.g. apa[ɑˈpʰɑ] (barley), but before alveolar nasals it is fronted to [æ]; an example is dan[tæ̃n] (carrying pole with loads). Before velar nasals, as in dan[tɑ̃ŋ] ([political] party), the back form is used.[17]
Glides
Although true diphthongal sequences (i.e. complex vowel combinations) do not exist in Santa, combinations with the vowels /u,ə,ɑ/ and the glides/j,w/ do.[18] The glides are phonetically realized as non-syllabic vowels [i̯,u̯] and the vowels have the following phonetic realizations below, regardless of any previous allophonic rules as mentioned above.[19]
Distinct phonetic vowel realizations (note that vowels in other environments take the cardinal form):
/ə/:
Before /j/, the vowel is raised to [ə̝], e.g. khighei[q͡χʰɯ̥ˈqə̝i̯] (pig), falling somewhere in between [ə] and [i].[21]
If also after /w/, the vowel is optionally dropped, as in ghuilu[qu̯(ə̝)i̯ˈlu] (to become).[22]
After /j/, it is pronounced like stressed [ɛ]; for instance nie[ni̯ɛ] (one) or bierei[pi̯ɛˈri] (wife).[23]
Before /w/, backness is often applied so that it varies between [ə~ɤ]: hotou[xʷoˈtəu̯~xʷoˈtɤu̯] (maggot).[21]
/ɑ/:
Before /j/, it may be raised, varying from [ɑ~ɛ], see bai[pɑi̯~pɛi̯] (to stop), kuaisun[kʰu̯ɑi̯ˈsũŋ~kʰu̯ɛi̯ˈsũŋ] (navel).[21]
Before /w/, the vowel is often raised and backed, ranging from [ɑ~o] in nao[nɑu̯~nou̯] (to hit the mark), oqiao[ʷoˈt͡ɕʰi̯ɑu̯~ʷoˈt͡ɕʰi̯ou̯] (old).[24]
Phonotactics
The general syllable template for Santa is (C)(G)V(G/N), where the brackets represent optional phonemes. C represents any consonant apart from /ŋ/ or a glide, G is one of the glides /j,w/, V is a vowel and G/N is either a glide or a nasal /n,ŋ/ (occasionally also /r/).[25] However, there are further constraints to this scheme:
The obstruents/pʰ,qʰ,f,x,h/ in words of Monogolic origin can only be found at the beginning of words.[26]
Also, /ɕ/ can occur at any position syllable-initially, but not after nasals of glides.[26]
/r,ʁ/ on the other hand can only occur after vowels. It is also possible for /ʁ/ to occur after nasals, but not /r/ due to the sonorant rule.[26]
/ʂ,ɕ,t͡ɕ,m/ can occur word-initially but between syllables only after vowels.[26]
It is impossible for non-compounded words of Mongolic origin for a sonorant (with the exception of the glide [i̯]) to follow a nasal after a syllable boundary. However, three exceptions exist: manlou[mɑ̃ŋˈləu̯] (forehead), danlei[tæ̃nˈləi̯] (palate) and chanlie[t͡ʂʰæ̃nˈli̯ə] (to listen).[27]
Words borrowed from Mandarin Chinese do not follow this rule, seen with rinmin[ʐəŋˈmin] (people; from 人民rénmín).[28]
Stress
The majority of words in Santa are stressed on the final syllable. Examples include yawu/jɑˈwu/ (to walk), funiegvan/funjəˈʁɑŋ/ (fox) and agven/ɑˈʁəŋ/ (village). Stress also shifts to the suffix when added to a word: funiegvan-ni/funjəʁɑŋˈni/ (fox, genitive case), agven-de/ɑʁəŋˈdə/ (village, locative case).[29]
However, several exceptions to this rule exist:
Certain suffix clitics prevent the stress from shifting to the final syllable, as seen in yawu=ne/jɑˈwunə/ (walk!, imperative), tomo=ghala/tʰoˈmoqɑlɑ/ (hammer, instrumental). However, not every clitic follows this rule.[29]
The suffixes +jiwo and +senu receive penultimate stress: jawu+jiwo/jɑwuˈt͡ʂiwo/ (to walk, progressive) and ire+senu/irəˈsənu/ (as soon as one comes).[30]
Similarly, basi/ˈpɑsi/ (tiger) and bosi/ˈposi/ (cloth) are stressed on the initial syllable. There is a minimal pair with bosi where the stress on the final syllable (/poˈsi/)means 'to wake up'.[30]
Loanwords may also violate the final stress rule; this is discussed below.[30]
In loanwords
In Chinese loanwords, stress is not predictable; this is likely due to the tonal influence from the original pronunciation. However, these tones are based on the dialects of the Hui people rather than the standard dialect. Examples of non-final stress include hushi/ˈxuʂi/ (nurse; from 护士hùshì), daozi/ˈdɑwt͡si/ (rice paddy; from 稻子dàozi) and dadou/ˈdɑdəw/ (soybean; from 大豆dàdòu). Minimal pairs, though uncommon, also exist. One example is shizi/ˈʂit͡si/ (persimmon; from 柿子shìzi) and shizi/ʂiˈt͡si/ (lion; from 狮子shīzi).[31]
There are a number of Arabic loanwords which also violate ultimate stress; these include aghili/ˈɑqili/ (wisdom; from عَقِيْلʿaqīl), mehheri/ˈməhəri/ (dowry; from مَهْرmahr), ghalebu/qɑˈləbu/ (body, form; from قَالَبqālab), etc.[32]
Some words with unknown origin may also stress on other syllables, including tonghori/ˈtʰoŋqori/ (wild goose), dawala/ˈtɑwɑlɑ/ (urinary bladder) and bawa/ˈpɑwɑ/ (great-grandfather).[32]
Voice category (relationship between action and its subject)
Voice
Particle
Example
Meaning
Active voice
-∅
Causative voice
-gva
bai-gva
to establish
Cooperative voice
-ndu
jie-ndu
to eat
Numeral: Singular has no suffix, plural suffix is -ndu, the same as Cooperative voice.
Numbers: Singular has no additional component, plural's additional component is -ndu, the same as the additional component for Cooperative voice.
Adverbial verb
Adverb
Particle
Example
Meaning
Coordinative adverbial verb
-zhi
shira-zhi ijie
to roast and eat
Repetitive adverbial verb
-n
wila-n wila-n yawulawe
walking while crying and crying
Purpose adverbial verb
-le
bi nie uzhe-le irewe
I came to take a look
Separative adverbial verb
-dene (-de)
chi ijie-dene bi ijieye
you eat first, then I'll eat
Conditional adverbial verb
-se
ghura bao-se bi ulie irene
if it rains, I won't come
Choice, Limit adverbial verb
-tala
khara olu-tala wilie giewe
work until it gets dark
Concessive adverbial verb
-senu
eqiemagva bosi-senu shu uzhezho
read a book after getting up in the morning
Stative verb
Tense
Particle
Example
Meaning
Past
-san (-sen)
ene baodei bolu-sanwe
this wheat is ripe
Future
-ku / -wu
echi-ku mo gholowe
the road that needs to be taken is long
Present
-chin
enende sao-chin kun wainu uwe
is there anyone living here?
Existential verbs bi and wi
bi
wi
Active verb
wai-ne
wi-we
Stative verb
Past stative verb
bi-san
wi-san
Future stative verb
bi-ku
wi-ku
Present stative verb
bi-chin
wi-chin
Adverbial verb
Coordinative adverbial verb
bi-zhi
wi-zhi
Separative adverbial verb
bi-dene
wi-dene
Conditional adverbial verb
bi-se
wi-se
Choice, Limit adverbial verb
bi-tala
wi-tala
Concessive adverbial verb
bi-senu
wi-senu
Tense
Particle
Example
Meaning
Speculative
bi-mu
hhen-de baer bi-mu
he might be rich
Recollective
bi-lai
hhe nasun giedun bi-lai bi matawe
I forgot how old he is
Syntax
In common with other Mongolic languages, Dongxiang is spoken as an SOV language. In Linxia, however, under the influence of the Mandarin Chinese dialects spoken by the neighbouring Hui people, sentences of the SVO type have also been observed.[34]
Writing system
Knowledge of Arabic is widespread among the Sarta and as a result, they often use the Arabic script to write down their language informally (cf. the Xiao'erjing system that was used by Hui people); however, this has been little investigated by scholars. As of 2003[update], the official Latin alphabet for Dongxiang, developed on the basis of the Monguor alphabet, remained in the experimental stage:[35]
^When the preceding consonant is gh, kh, h or gv, this vowel is pronounced the same as ii.
Numerals
Numeral
Dongxiang
IPA
Classical Mongolian
1
nie
/njə/
nigen
2
ghua
/qwɑ/
qoyar
3
ghuran
/quˈrɑŋ/
ghurban
4
jieron
/t͡ɕjəˈroŋ/
dörben
5
tawun
/tʰaˈwuŋ/
tabun
6
dzigvon
/t͡ʂiˈʁoŋ/
jirghughan
7
dolon
/toˈloŋ/
dologhan
8
naiman
/nɑjˈmɑŋ/
naiman
9
yesun
/jəˈsuŋ/
yisün
10
hharan
/hɑˈrɑŋ/
arban
Some numbers in Dongxiang have been heavily influenced by Mandarin. These include 30 (sanshi) and 40 (sishi). Meanwhile, both khorun/qʰoˈruŋ/ and ershi/ˈɚʂi/ are used for 20.
There are about 20,000 people in the north-eastern part Dongxiang County, who self-identify as Dongxiang or Hui people who do not speak Dongxiang, but natively speak a Dongxiang-influenced form of Mandarin Chinese. The linguist Mei W. Lee-Smith calls this the "Tangwang language" (Chinese: 唐汪话), based on the names of the two largest villages (Tangjia and Wangjia, parts of Tangwang Town) where it is spoken and argues it is a creolized language.[36] According to Lee-Smith, the Tangwang language uses mostly Mandarin words and morphemes with Dongxiang grammar. Besides Dongxiang loanwords, Tangwang also has a substantial number of Arabic and Persian loanwords.[36]
Like Standard Mandarin, Tangwang is a tonal language, but grammatical particles, which are typically borrowed from Mandarin, but are used in the way Dongxiang morphemes would be used in Dongxiang, do not carry tones.[36]
For example, while the Mandarin plural suffix -men (们) has only very restricted usage (it can be used with personal pronouns and some nouns related to people), Tangwang uses it, in the form -m, universally, the way Dongxiang would use its plural suffix -la. Mandarin pronoun ni (你) can be used in Tangwang as a possessive suffix (meaning "your").
Unlike Mandarin, but like Dongxiang, Tangwang has grammatical cases as well (however only four of them, unlike eight in Dongxiang).[36]
Bao, Saren (包萨仁) (2006), 从语言接触看东乡语和临夏话的语序变化 [Sequential Changes in Dongxiang Language and Linxia Dialects from the View of Linguistic Contact], Journal of the Second Northwest University for Nationalities (Social Science Edition) (2), ISSN1008-2883, archived from the original on 2011-07-13
Wei, Li Xue; Stuart, Kevin (1989), "Population and Culture of the Mongols, Tu, Baoan, Dongxiang, and Yugu in Gansu", Mongolian Studies, 12 (The Owen Lattimore Memorial Issue), Mongolia Society: 71–93, JSTOR43194234