Altaic Elements in the Chinese Variety of Tangwang: True and False Direct Loans

Abstract

This paper foccusses on the Tangwang language, a Chinese variety spoken in southern Gansu that has been in contact with the Dongxiang language, a Mongolic language. Tangwang is believed to be a highly altaicised variety, as it demonstrate several traits that are usually absent in this language family are are reputed ‘typical’ of the Turkic-Mongolic languages. However, most of these traits are present in the other northwestern chinese varieties and are the result of reanalysis, thus, it is difficult to trace their exact origin. This paper aims at analyzing the influence of Mongolic languages on Tangwang from the perspective of borrowings, and in particular direct loans. Taking the formally identical features that are shared in Dongxiang and Tangwang as a starting point, we will try to determine which form can be seen as a direct borrowing due to the adstratal influence of Dongxiang and which one is probably due to an earlier altaic influence. We will try to classify which form is a ‘true’ direct loan from Dongxiang and which form could be the evidence of an earlier substrate. From the results, and based on the existing models on languages contact, we will try to understand which mechanisms from relexification, grammaticalization, and language shift is the most probable in the case of Tangwang.

1. Introduction

All varieties spoken in Northern China have been influenced by Altaic languages, a process known as the ‘Altaicisation of Chinese’ (Hashimoto 1976, 1986). It has led to the division of the Chinese dialects into two main areal–typological complexes which correspond to the northern and the southern areas (Norman 1988). During the course of history, various groups of languages, including Mongolic, Turkic, and Uralic, have been in contact with the Chinese varieties spoken in Northern China and have induced changes to the phonology and morpheme structure, but also to syntax and morphosyntax. Because of the historical nature of these contacts, Chinese has received several layers of substratal and superstratal influences of the Altaic type (Janhunen 2015), and, in modern times, most direct contact between northern Chinese dialects and Altaic languages have ceased. However, some dialects also have been in prolonged contact with Mongolic languages on the local level, leading to the adstratal influence on those varieties. These include the Tangwang dialect, spoken by about 15,000 speakers in the village of Tangwang (唐汪镇 Tángwāng zhèn), situated in the north-eastern part of the Autonomous County of Dongxiang (东乡族自治县 Dōngxiāngzú zìzhìxiàn) at the border of Gansu and Qinghai provinces, in the People’s Republic of China. This region, known as the Hexi Corridor complex, is the home of speakers where languages from different families (Mongolic, Turkic, Sinitic, and Bodhic) have been in contact, leading to the formation of a linguistic area (Dwyer 1995; Janhunen 2004; Slater 2001; Szeto 2021, etc.). Tangwang, along with some other dialects spoken in this region (including Wutun 五屯话, Gangou 甘沟话, Zhoutun 周屯话, Linxia 临夏话, and Xining 西宁话), features a certain number of non-Sinitic traits including nominal case, tense, and aspectual markers, verbal dependance marker (equivalent to Mongolic converbs), dominant SOV order, multiple plural markers on unanimated objects, etc., which are believed to have been induced through language contact with Altaic languages, and in particular, with the surrounding Mongolic languages of Monguor, Bonan, and Dongxiang spoken by their neighboring communities (Peyraube 2017). The lexicon in all these dialects demonstrates only a few lexical loans, and their vocabulary remains strictly Chinese. Because of the high degree of ‘mixture’ of these dialects on the syntactic level, they have often been considered as a kind of Chinese-based “creoles”, “mixed languages”, or “Chinese hybrid dialects”, which show some degree of mutual intelligibility with other Sinitic varieties (Dwyer 1992; Wurm 1995; Lee-Smith 1996; Slater 2001; Janhunen 2004; Kawasumi 2018; Bell 2020). However, most of these features are formally very different from their Mongolic counterparts, especially from the local Mongolic languages they have been in contact with, and are either the result of a calque or reanalysis process. Another possibility would be that they were borrowed from other now extinct Altaic varieties we have not fully understood yet (e.g., para-mongolic Khitan) or an unknown Altaic language that we do not have any written record of. In any case, the exact source of the influence is often difficult to establish, especially for cases of nominal case markers (Xu 2015). Some other forms seem to be related to those found in other Chinese dialects, especially those of the Shanxi–Shaanxi area, spoken further east, but which are yet not fully reconstructible. While the hybrid dialects of the Hexi corridor all share these syntactic features, Tangwang demonstrates a few additional direct loans from Mongolic that are found only in this variety and that will be the object of this study.

This paper aims at analyzing the influence of Mongolic languages on Tangwang from the perspective of borrowings, and in particular direct loans. Taking the formally identical features that are shared in Dongxiang and Tangwang as a starting point, we will try to determine which form can be seen as a direct borrowing due to the adstratal influence of Dongxiang and which one is probably due to an earlier Altaic influence. This paper is organized as follows: in Section 2, we provide an overview of the context, including the sociolinguistic environment and the contact-induced features found in Tangwang’s syntax that are seen as Altaic, but that are commonly shared with other northwestern varieties and should be identified as a superstatal influence. In Section 3, we give an overview of the vocabulary borrowed from and through Dongxiang in Tangwang, as seen in the previous studies, and discuss in detail one particular element found over the different Chinese hybrid dialects of the Hexi corridor, i.e., the formal distinction for the word ‘small’ (ka 尕 and ɕʲɔ 小) and its particularities in Tangwang compared to Dongxiang. In Section 4, we discuss the pronoun wɤlũ 喔咙 ‘1sg.log’ and compare it with the Dongxiang reflexive pronoun orun ‘1sg.log’. In Section 5, we describe two possessive suffixes, namely the third-person possessive -ni 尼 and the reflexive possessive -ne 呢. In Section 6, we compare two case markers, the limitative thala 他啦 and the instrumental -la 啦 with their Dongxiang counterparts. In the discussion, we will try to classify which form is a ‘true’ direct loan from Dongxiang and which form could be the evidence of an earlier substrate. From the results, and based on the existing models of contact language, we will try to understand which mechanisms from relexification, grammaticalization, and language shift is the most probable in the case of Tangwang.

2. Context

Tangwang is spoken in the village of Tangwang (唐汪镇 Tángwāng zhèn) where both Han (汉族 Hànzú) and Hui Muslims (回族 Huízú) cohabitate. This village has been named after the two main families present in the village, respectively, the Tangs (唐) and the Wangs (汪). Based on the legends collected among the Tangwang people, but also on limited historical evidence as well as on genetic affinities, it has been shown that the population inhabiting this village was of different origins, including Tibetan, Mongolian, and Han Chinese (Xu 2014). Three main narratives have been developed regarding the origin of the people of the Tang family. The first one is that they were partly of Mongolian origin. Their ancestors, a Yuan general, (called Baima jiangjun 白马将军 ‘General with the White Horse’), and his three wives (including one of Tatar or Mongolian origin) would have led a group from Sichuan, situated south of Gansu, to settle in the actual location of the Tangwang village (Tang 2011). This general is still worshipped as the clan god by the local Han Chinese in the local temple and is believed to be the ancestor of the Tang family. The second narrative is that they are of Dongxiang or Mongolic origin that have been Sinicized and that their initial language was relexified with the Chinese lexicon; hence, there are remaining Altaic syntactic features in the Tangwang language. The last narrative is that a group of four Chinese thieves flew from Sichuan to Tangwang and that one of them had converted to Islam. Their descendants would be today’s Hui and Han communities (Lefort 2023b, Fieldwork). The Wang family is mostly constituted of Han Chinese populations that have migrated to their actual location during the Ming dynasty and have partially converted to Islam, becoming Huis (Xu 2014). In any case, the time of formation of this community would be between the late Yuan dynasty (1271–1368) and the early Ming dynasty (1368–1644), that is to say, at the junction of the XIIIth and the XIVth centuries. Because of their religious proximity, Hui males from Tangwang and Dongxiang women have frequently intermarried. The Tangwang Hui males usually do not speak Dongxiang, but the Dongxiang women are usually bilingual. Therefore, there exist differences between Hui and Han speakers of Tangwang, a phenomenon that has not been widely described so far. This study focuses on Hui Tangwang speech and data partially collected with native informants in Tangwang during August 2023.

Tangwang remains an under-described variety, both in China and in the West. Ibrahim (1985) was the first to partially describe the Tangwang language, which was then included in Sun’s work on Mongolic languages (Sun 1990). It was only mentioned or partially described by scholars working in this area (Wurm 1995; Lee-Smith 1996; Janhunen 2004) before Xu Dan published her full monograph (Xu 2014). In addition, a few articles on the grammatical characteristics of Tangwang and phonological changes (Mo 2015; Lei 2017; Djamouri 2013, 2015) have been published. More recent research includes Ibrahim’s work on Tangwang vocabulary (Ibrahim 2017), Djamouri’s glossed and translated material of the Tangwang language and the glossary of the Tangwang language (Djamouri 2022a, 2022b), along with his grammatical description of the Tangwang language (Djamouri, forthcoming). As the Tangwang village forms part of the Dongxiang Autonomous District where Dongxiang (or Santa Mongolian), a Mongolic language from the Shirongolic branch, is spoken, contact-induced change descriptions have mainly focused on analyzing the influence of Mongolic languages on Tangwang by taking Dongxiang as the source (Ibrahim 1985, 2017). However, the influence of Mongolic on Tangwang is more complex than has often been described, involving several layers of contact with different Altaic (possibly Mongolic) languages. Along with the other above-mentioned Chinese hybrid dialects, Tangwang is considered to be a highly ‘Altaicised’ variety due to the several non-Sinitic traits that have been identified in these languages and the resemblance to what is found in Mongolic and Turkic languages. Those include head final, nominal case markings, converbial markers, quotative, aspectual/tense markers, third-person, and distal demonstrative pronoun syncretism, multiple categories of nouns that can receive plural markers, etc. In addition, Tangwang has a reduced tone system compared to other Sinitic languages. It is unclear if Xu and Wen (2017) believe that these changes have arisen solely from contact with Dongxiang, but at least, they do not propose another scenario and state that all changes have been induced through contact with Mongolic languages. However, although these traits bare some resemblance with Mongolic (and even Turkic languages) on the functional level, they are not completely identical and, in most cases, they formally differ from Mongolic languages.

The Tangwang nominal case includes the objective case xa 哈 (regrouping the dative–locative and benefactive functions), the ablative case ɕʲɛ 些~liɕʲɛ 里些, the comitative–instrumental case la 啦, locative cases li 里, and limitative t^hala 他啦 (Djamouri 2015). The limitative case marker t^hala 他啦 is not found with this function in Mongolic languages but could have developed based on the Mongolic converb t^hala 他啦. This point will be further developed in Section 6. The origin and the exact source for the nominal case have not been demonstrated (up to date) and some functions of the objective case xa 哈 overlap with those of the Dongxiang counterparts (Xu 2015; Lefort, forthcoming). The formally identical dative–locative–benefactive marker in Mongolic have two forms in Tangwang, the genitive and accusative cases phonetic forms have merged in Dongxiang as -ni/ji, but in Tangwang, we have two clear totally makers (dʑi 底 and xa 哈), and the formally identical accusative–dative–benefactive marker xa 哈 is expressed through two different markers in Mongolic. The only marker formally identical to its Dongxiang counterpart is the instrumental -la, which could have been borrowed. This point will be further developed in Section 6. Therefore, the main strategies adopted for the transfer of those features are probably calques and reanalysis, and the source is most probably not Dongxiang itself, and maybe not even Mongolic at all1 for the objective case (see Lefort, forthcoming for details). Consider

Table 1. Tangwang, Dongxiang, and Middle Mongol nominal case markers.

Tangwang	Function	Dongxiang	Middle Mongol
dʑi	Genitive	-ni/ji	-yin, -n (-Un, -in, -Ai/-(n)U(’Ai))
xa 哈 (a 啊/ang 昂)	Accusative	-ni/ji	-yi (-ni/-i)
	Dative	-de	-du(r)/-da
	Benefactive		-du(r)/-da
li 里 **	Locative		-du(r)/-da/-a/-na ***
thala 他啦	Terminative	-	-cAA
ɕʲɛ 些	Ablative	-se	-ca (-sa/-asa)
	Comparative
la 啦	Comitative	-le/-la	-luga/-laa
	Instrumental	-ghuala/-la	-ar/-iar/-niar

* ji 底 is not a nominal case marker per se, as it can be combined with adjectives and verbs. However, its functions partially fulfill the definition of case marking, as it can be combined with nouns, which will be further explained in the section below. In Mongolic, the genitive case is also a part of the case system, as it is the only case that connects two nouns. ** The spatial locative postposition li 里, ‘inside’, (along with ʂã 上 ‘on’, tʂũ 中 ‘in’, and xa 下 ‘under’) are attested in Chinese, but the marker li 里 in Tangwang has other functions that makes it closer to a case marker. *** According to Janhunen (2003, pp. 66–68), there were two distinct forms for the locative, one dative–locative case, -du(r) / da, marking indirect objects as well as a variety of local and temporal functions, and a “pure” locative case, -a / -na, which later disappeared and was completely replaced by the dative–locative case.

below:

The aspectual markers include the prospective li 咧, perfective lʲɔ 寮, and imperfective tʂɛ 寨. Although post-verbal aspectual markers are attested in other Chinese dialects, this paradigm is closer to what is found in Altaic languages. However, as for nominal case markers, the aspectual markers demonstrate many differences. They formally do not correspond at all to their Dongxiang counterparts, not even to Middle Mongol: Tangwang uses two different markers to express the prospective and imperfective while Tangwang uses two different markers to express the prospective and imperfective while Dongxiang uses only one, and Tangwang lacks progressive and habitual tense markers, consider

Table 2. Tangwang tense–aspect markers (indicative).

Aspect	Tangwang	Dongxiang	Middle Mongol
Future	li 咧	-ne	-m~mU~-mUi
Imperfective	tʂɛ 寨
Perfective	lʲɔ 寮	-wo	-bA~-bAi~bi
Confirmative	-	-	-lU’A~.-lU’Ai~-li’i
Progressive/Durative		-zhiwo	-nAm
Habitual		-zhi	-

below:

Other Gansu–Qinghai hybrid dialects also feature tense–aspect markers, including Gangou (Yang and Zhao 2021), Wutun (Sandman 2016), and Linxia (Liu et al. 1996). Aspectual and tense markers are also found in Jin languages, e.g., northern Shanxi–Shaanxi and southern Inner Mongolia (Arcodia 2021; Xing 2006, 2020), but do not function exactly in the same manner and are thought to be developed from contact with Altaic languages.

One important gap in Tangwang’s paradigm compared with the Mongolic languages is the lack of participle markers (or ‘nominalizer’), which enable verbs to receive a further nominal morphology (for example, there are three nominalizers in Dongxiang, and thirteen participle markers in Middle Mongol).

Dependence verbal marks that could possibly be of converbial Mongolic origin include the progressive tʂɚ 着 and t^hala 他啦. The progressive marker tʂɚ 着 is found across all northern dialects and is believed to be an early loan from Altaic languages (Zu and Gao 2022). The form t^hala 他啦 is found in all Mongolic languages and the hybrid languages of the Gansu–Qinghai area, and is probably a secondary loan. It is worth noting that the Tangwang language, along with other hybrid Chinese dialects, feature only a very restricted set of converbs compared to the Mongolic languages (seven converbs in Dongxiang, for example).

Other traits include the quotative ʂuo 说 ‘to say’, which has been grammaticalized in most northern dialects, including the Beijing dialect, where it is found as a quotative marker, a complementizer, a topic marker, etc. (Fang 2006). In the Gansu–Qinghai complex, it is attested as a quotative and evidentiality marker in Xining (Zhang 2007), Wutun (Sandman 2016, p. 344), and Tangwang (Djamouri 2013). Those are thought to have been induced through contact with Mongolic languages and in particular, could be a calque Mongolic verb ke—‘say’. In Tangwang, the quotative ʂuo 说 also has epistemic and alethic functions that seem to be internal developments (Djamouri 2013, p. 17).

In addition, Tangwang’s distal demonstrative nə 呢 (and its related forms) is formally identical to the third singular personal pronoun nə, a phenomenon that is found across Mongolic and Turkic languages. This syncretism is also present in Wutun (third-person pronoun gu 箇 and distal demonstrative gu 箇), while in the other Chinese hybrid dialects of the Gansu–Qinghai area, the third-person form seems to be derived from the proximal demonstrative form (e.g., Lanzhou dialect third-person pronoun la 那 and distal demonstrative lei 那). Furthermore, in many dialects along the Yellow River in the area of Shanxi and Shaanxi, including in the Jin dialects, distal pronouns and the third-person form are often formally identical (Hou 2012, p. 314). Those forms are believed to have been induced by contact with Altaic languages (Zhang and Zhang 2007, p. 332; Zhen 2002). Moreover, formally identical demonstratives and third-person pronouns are well attested in other languages in the world and are not always the result of language contact (Bhat 2013). Therefore, although contact with Dongxiang could have reinforced the calquing process, this syncretism is probably inherited from an earlier altacised variety of Chinese.

The use of the plural marker mu 仫 with non-animated objects in Tangwang have not been investigated in detail, but is mentioned by Xu and Wen (2017) who believe it is the result of contacts with Mongolic languages. This trait is also very common to all northern dialects, even those that have not been in contact with Mongolic languages in modern times, and the particularities of the plural marker in the other Gansu–Qinghai dialects have been extensively described (Xu 2012; Mo 2004; Yang 2014, etc.). Compared to other Mongolic languages, Dongxiang has a reduced set of plural markers, which are -la and -sila and no other Altaic language features a plural marker from which Tangwang could have borrowed the marker mu 仫. Hence, the marker itself is not of Dongxiang origin, and if the use extention of this marker is the result of language contact, it has gone through a reanalysis process and has probably been inherited from an earlier altaicised variety.

At last, the head-final OV dominant word order has been proven to cohabitate with a VO underlying word order (Djamouri 2013; Xu 2014). Therefore there is no complete SOV to SVO word order shift in Tangwang.

All the above traits are reputed to have been induced through contact with Mongolic (or more generally with Altaic) languages, but most of them are also found in other northern dialects that have not been in contact with any language of this family in modern times. Therefore, we believe they form part of a substratral influence of Altaic languages shared in all northern dialects, including Tangwang. The only trait that is not found in dialects outside of the Gansu–Qinghai area is the case system. We have seen that it does not exactly match the Mongolic nominal case system, both functionally and formally; hence, it is unlikely to be solely the result of the contact with these languages. The other aspectual, converbial, quotative, and plural markers are formally completely different from any Mongolic markers, and therefore should also be identified as calques or reanalyzed forms.

Leaving aside the calques of these syntactic features (case markers, aspect, quotative, plural markers, third-person distal demonstrative syncretism) and the loans that are found cross-dialectally in the northern region (progressive converb), in the next sections, we will turn to what can be identified as the direct influence of Mongolic languages, that is to say, lexical and grammatical borrowings that are formally identical with their Dongxiang counterparts, or formally very close to them.

3. Borrowed Lexicon

3.1. Lexical Loans and Calques

One of the characteristics of the Tangwang language is the contrast between the very low number of lexical borrowings and the high number of non-Sinitic grammatical elements. The non-Sinitic vocabulary represents less than 1.5% of the Tangwang vocabulary (Xu 2017, p. 41) and is mainly borrowed from the Dongxiang language. Secondary sources of loans include Arabic, Persian, and Turkic, mainly related to religion, that have been borrowed through Dongxiang. This is not a feature unique to the Tangwang language, but a common trait in all other Chinese hybrid dialects of the Hexi corridor (Zhou 2023; Sandman 2016). Only two out of the two hundred words listed in the Swadesh list are of Turkic origin, namely ana 阿娜 ‘mother’, ‘mom’ (CT ana) and ata~ada 阿达 ‘father’, ‘dad’ (CT ada). It is shared in Dongxiang and could have been borrowed through the mothers when intermarriage took place. However, they are most probably the result of areal diffusion, as they are also found in other hybrid dialects of the region, including Wutun (Sandman 2016, p. 36), Linxia (Liu et al. 1996, p. 91), the Linxia city Bafang variety (Lefort 2023b, fieldwork), Hetan (Lefort 2023b, fieldwork), Gangou (Feng and Stuart 1992, p. 5), and partially in other Mongolic languages2. The core vocabulary remains strictly Chinese, mainly shared with the vocabulary of northern Chinese dialects, but with two exceptions3 from southern Chinese, especially Sichuan, which, according to Xu (2020), corroborates the narratives regarding the settlement of the Tang family in the region.

In his work on the Tangwang vocabulary, Ibrahim (2017) has listed about 150 loanwords in Tangwang that he believes were borrowed from Dongxiang (including words of Mongolic, Turkic, Persian, and Arabic origin). Most of them are direct loans, but some others are more probably Chinese borrowings to Dongxiang (most likely from Linxia) rather than the contrary. For example, vuəvuə 窝窝 ‘a kind of doughnut’ is identified as a Dongxiang loan but was borrowed first from Linxia wɤwɤ 窝窝 to Dongxiang and then loaned to Tangwang, if not already present in Tangwang initially. Furthermore, a significant number of the words listed are common to the hybrid Chinese dialects and the Mongolic languages of the region. For example, Tangwang aka 阿尕; Linxia aka 阿尕 ‘elder brother’; Wutun aka ‘elder brother’; Gangou akə ‘elder brother’; Dongxiang aga ‘elder brother’; Mongguor aġa ‘elder brother’; eastern Yughur ağa ‘elder brother’; Bonan a’gha dou ‘brothers’, etc. Therefore, it should be considered as a common areal lexicon, or regionalism, probably the result of different contact layers, rather than direct loans from Dongxiang.

Dongxiang has undergone important phonological changes compared with Middle Mongol, which most of its Mongolic vocabulary has been inherited from (Field 1997; Lefort 2012). Therefore, the Dongxiang vocabulary in Tangwang is easily identifiable. Most of the Mongolic vocabulary borrowed from Dongxiang is related to everyday life and includes words from kin relationships to cultural-related matters through cuisine and animal names. Most of the borrowings are nouns, followed by verbs, while only a few adjectives and adverbs have been loaned. A few examples are provided hereunder (Ibrahim 2017; Djamouri 2022a):

• TW 苏唤赤 suxʷɛ̃ntʂʰə ~ suxɛ̃tʂʰə ‘bridesmaid < DX suxʷɛ̃ntʂi
• TW 粘连 tʂɛ̃lʲɛ̃ ‘companion; friend’< DX dʐənliən
• TW 窝咧称 wɤlʲɛtʂʰə̃ ‘orphan’ < DX oliətʂin [< MMo SHM onecit]
• TW 哈眯柔 xamiʐʉ ‘a person who squints’< DX χamirəu ‘squinted-eyed’
• TW 合里哦屋 xəliɔvu ‘hawk’ < DX həliəwu [< MMo HY hele’e]
• TW 猴娃 xʉwa ‘spinning top’< DX χauwa ‘spinning top’
• TW 马且 matɕʰʲɛ ‘steamed twisted rolls’< DX matɕiə ‘steamed twisted rolls’
• TW 哈楼 xalɔ ‘ugly’ < DX χalao ‘ugly’
• TW 果果 kʷɤkʷɤ ‘breast’< DX gogo [< MMo HY köken].

One of the most often used adverbs borrowed from Mongolic is ətə (< ede) ‘now’, which is not attested in other hybrid languages. It is almost certain that it is a direct loan from Dongxiang, as neither Shirongolic language forms (Mongguor do; Bonan da ~ de; Kangjia da; Eastern Yughur ɔdɔː) nor Middle Mongol ones (MMo SHM edo’e; HY edö’e) are likely to be the source:

(1)	阿德	三十	年	成寮
	ətə	sɛ̃ʂʅ	nʲɛ̃	tʂʰə̃ -lʲɔ
	now	thirty	year	finish—PFV.
		‘It happened thirty years ago’. (Djamouri 2022b, p2)

In addition, a few Turkic, Persian, and Arabic loans have been borrowed to Tangwang, e.g.:

• 吾心 vuɕin ‘face’ < uɕin (Turkic)
• 罗波罗 lʷɤp’olʷɤ ‘a kind of meat congee’ < loporo (Persian)
• 杜瓦呐 tuvana ‘a kind ritual bread used during funerals’ < duwana (Arabic).

Those words are also present in Dongxiang and have probably been loaned to Tangwang via Dongxiang. Other non-Mongolic lexical borrowings are mostly related to religious matters (Xu 2017, pp. 127–28).

In addition to lexical loans, there are a certain number of expressions that are similar to those found in Dongxiang. Their meaning and structure are copied from the Mongolic languages by using Chinese morphemes, and therefore should be identified as calques rather than lexical loans per se. A few examples are provided below (Ibrahim 2017):

TW 压黑 jaxi (lit.‘press black’) ‘to have a nightmare’ < DX qara daru lit.: black + press ‘have a nightmare’.

TW 五形没有 wuɕʰĩ mɔjʉ lit.: face + have.not ‘ugly’ < DX uɕin wiwo lit.: face + have.not ‘to be ugly’ (not ‘loose face’ as in other Chinese dialects).

TW 往前来 wã$\tilde{t}$ɕʰʲɛ̃ lɛ lit.: from before + come ‘to recover from a serious illness’ < DX məliə irə lit.: from before + come ‘to recover from a serious illness.

Some of those expressions, such as 压黑 jaxi, ‘to have a nightmare’, are clearly calques from Mongolic as they are attested in Middle Mongol and have even been conserved in Modern Mongolian: ᠬᠠᠷ᠎ᠠ ᠳᠠᠷᠤᠬᠤ qara daruqu lit.: black + press ‘have a nightmare’. They could be loans from earlier contact with other Mongolic varieties rather than from Dongxiang itself, but this is unlikely to be the case.

To conclude, although a certain number of lexical borrowings can be found in Tangwang, the influence of Dongxiang on the vocabulary is sporadic and most of the Tangwang lexicon remains Chinese.

3.2. Formal Distinction of the Word ‘Small’

A lexical feature which seems to be shared among most of the Chinese hybrid languages of the Gansu–Qinghai area is the two distinctive forms for the word ‘small’. In Tangwang, it is found under the form ka 尕 (2) and ɕʲɔ 小 (3):

(2)	过去	驴	走着	过不哈底	中底	尕	路
	kʷɔʨʰi	lɥ	tsʉ-tʂə	kʷɔ-pu-xa-ʨi	tʂũ -ʨi	ka	lu,
	past	donkey	walk-subv	pass-neg-pot-sub	like.this	small	road
	阿德	成哈	大卡车	寮。
	ətə	tʂʰə̃-xa	ta-xatʂʰə	lʲɔ
	now	become-res	big-truck	fp

‘Roads that used to be so small that even a donkey could not go through have now become so big that trucks can drive on them’.

(Djamouri 2022b, p. 48)

(3)	这	个	人	小气底很。
	tʂɛ	kɛ	ʐə̃	ɕʲɔtɕʰi-tɕi-xə̃
	this	cl	man	mean-subs-very

‘This man is very mean.’

(Djamouri 2022a)

The adjective ɕʲɔ 小 is also found in other lexicalized items such as ɕʲɔtʰʷi 小腿 ‘calf’ and ɕʲɔɕĩ 小心 ‘careful’ (see Djamouri 2022a for details).

The form ɕʲɔ 小 is shared among all Chinese dialects, while the origin of ka 尕 is more uncertain. According to Dwyer (1992), the source language for ka 尕 is clearly not of Chinese origin and appears in loans and onomatopoeias4. This distinction is also found in Linxia, Wutun, Xining and Lintan, but also in some Chinese dialects that have been in contact with Mongolic languages in Xinjiang (see, for example Wang and Zhu (Wang and Zhu 2021) for the Jimsar dialect). The main difference between ka 尕 and ɕʲɔ 小 is that ɕʲɔ 小 appears in bound morphemes, while ka 尕 is a free one. ka 尕also bears emotional overtones of affection that ɕʲɔ 小 does not. In addition, ka 尕 is used as predicate, while ɕʲɔ 小 is used in an attributive form in Linxia. In Tangwang, the functional distribution of ɕʲɔ 小 and ka 尕 appears to be similar to what is found in Linxia. ka 尕 can be used both in predicative and attributive positions, while ɕʲɔ 小 appears only in the attributive position as a bound morpheme (all the examples below are from (Djamouri 2022a)):

(4)	尕尕底	尕驴
	kakaːtɕi	ka.lɥ
	small-small-subs	donkey

‘A small donkey; donkey foal’.

(5)	土盖啦	泥哈寨	个	尕	炉子
	tʰukɛ-la	ni-xa-tʂɛ	kɛ	ka	lutsɿ
	clod.of.earth	mud-res-ipfv	cl	small	stove

‘Make a small stove out of mud and clod of earth.’

(6)	这	衣裳	尕	些
	tʂʅ	jiʂã	ka	ɕʲɛ
	this	clothes	small	a.bit

‘These clothes are a little small’.

The Tangwang ga 尕 is also a prefix which has lost its primary meaning (such as in ka.lɥ 尕驴 ‘donkey’ in example (7)), which is also the case in Linxia.

In Dongxiang, there are also two distinctive forms for the word ‘small’, which are ga and mila.

(7)	niə	ga	maoɚ	osi-səŋ	saiʁaŋ	no	wo	dʑiuʂi
	one	small	cat	look-p.nmlz	pretty	very	cop	indeed
	daŋdzi-ni	mila	no	wo.
	courage-3sg.refl	small	very	cop

‘There was a beautiful kitty, but he was very fearful.’

(Lefort 2023a)

In Dongxiang, mila is used in idiomatic expressions, such as in mila ʂida, ‘(when I) was small’, a calque of the Chinese ka-ʨʰjɛ̃tsɿ 尕前子, ‘(when I) was small’, where it cannot be replaced by ga. It is also found in a series of lexical calques borrowed from Chinese with a predicative function, e.g., mila otɕin (ɕjɔ nɥɪ 小女儿) ‘little girl’, mila sara (ɕjɔʝɥɛ 小月) ‘a month with less than 30 days’, mila udʑə (ɕjɔkʰæ̃ 小看) ‘to scorn’, etc.

In Dongxiang, ga is mostly used in a predicative manner, to express ‘young’ rather than ‘small’, as in ‘small in height’, and bears the emotional overtone Dwyer (1992) was referring to in Linxia:

(8)	ənə	ga	wawa	saiʁaŋ	wo.
	dem	small	girl	pretty	cop

‘This little girl is cute’

(Lefort 2012)

ga is also found in a series of calques borrowed from Chinese, but the distinction with mila used in a predicative manner is very fuzzy as ga and mila are used as bound morphemes and can alternate most of the time: ga baɚ/mila baɚ ‘small change’ (<ɕʲɔ ʨʰjɛ̃ 小钱), ga bəilin/mila bəilin ‘younger member of a family’/‘premature’ (<ɕʲɔ pe 小辈), ga dʑiawu/mila dʑiawu (<? ɕʲɔɕʲatsu 小家族5) ‘families made up of closely relatives’, ga pudʐa/mila pudʐa ‘green pea’ (<ɕʲɔ tʉ 小豆); ga oliə/mila oliə ‘less than half’ (<ɕʲɔpɛ̃ 小半), etc.

It is interesting to note that the Dongxiang mila can replace ga in all situations, but not the contrary, while in Tangwang, ka 尕 can replace ɕʲɔ 小 most of the time, but not the contrary. Hence, the use of mila in Dongxiang corresponds to ka 尕 in Tangwang, while the use of ga in Dongxiang corresponds to ɕʲɔ 小 in Tangwang.

This syntactic split (predicative vs attributive) exists in other Shirongolic languages, as well as in Amdo Tibetan. The word mila is not attested in Middle Mongol sources with the meaning ‘small’, which reflects a regionalism6. Mila, ‘small’, is shared by all the other Shirongolic languages, except Bonan, while ga is found only in Dongxiang, which suggests it is a secondary borrowing due to areal diffusion. It could be an initial borrowing from the Mongolic baga, ‘small’, to Chinese (with an initial syllable drop), which has then been re-borrowed to Dongxiang as ga. Baga ‘small’ and mila ‘small’ are both found in Eastern Yughur, which could be an indication that Dongxiang initially featured the same form. This would suggest that the origin of ka 尕 differs from the same word attested in southern dialects. Furthermore, the distinction between predicative vs attributive functions of the word ‘small’ is attested in modern Mongolian as baga ‘small’ and dʑidʑig ‘small’. Therefore, although the origin of the lexical stem itself cannot be established, the functional distribution could be the result of an earlier calque from Mongolic. The fact that the Dongxiang mila and the Tangwang ka 尕 (and to some extent the Dongxiang ga and the Tangwang ɕʲɔ 小) are used in the same manner suggests that it is a language contact-induced phenomenon. However, it is difficult to state which language has influenced the other, and how ga is a direct loan in Dongxiang from the local variety of Chinese ka 尕, but with restrictive features compared to its Chinese counterpart, while mila may be a claque of the Chinese ka 尕. On the other hand, the Chinese ka 尕 could also be a calque of the Mongolic mila.

4. First-Person Logophoric Pronoun

The Tangwang first-person logophoric pronoun wɤlũ (vəlũ) 喔咙 is formally very close to its Dongxiang counterpart oruŋ~oroŋ. The Dongxiang oruŋ~oroŋ corresponds to the common reflexive pronoun *öxer-i-xe/n found in all Mongolic languages. This pronoun is not attested in other Chinese dialects, nor in other Chinese hybrid dialects such as Linxia or in Wutun. This indicates that wɤlũ (vəlũ) 喔咙 has been borrowed from Dongxiang itself rather than from an earlier common Mongolic source.

In Dongxiang, even if it was attested to be used as a reflexive pronoun in earlier sources (Buhe 1987; Kim 2003), oruŋ and its plural form oruntaŋ are nowadays used almost exclusively for reported speech7, while the function of the reflexive pronoun is carried out by the Linxia loanword guadʑia (gojə) (<gedʑia 各家):

(9)

hə

madə

kiəliə-dʐiwo:

‘oruŋ

maʁaʂi

Bəijin-də

3sg

1 sg.ben

say-prog

1sg.log

tomorrow

Beijing-loc

əchi-nə’

giə-dʐi

go-ipfv

tell-cont

‘He told me that he will go to Peking tomorrow’ (Lefort 2023b, fieldwork).

(10)

əminə

kiəliə

bi

jiən-nə

gojə

waʁa-ne

tʂi

goye-ne

Amin

say

be

clothes-refl

self

wash-ipfv

2sg

self-refl

nadu-le

etʂi.

play-pur

go

‘Amin said, “I’ll wash my clothes myself, you go and play!”.

(Lefort 2023a)

In Tangwang, wɤlũ 喔咙 is also attested as a logophoric pronoun (Ibrahim 1985, 2017; Djamouri 2022a):

(11)	呢个人	我哈	说着咧,	“喔咙	明个	北京
	nekɛʐə̃	wɤ=xa	ʂʷɔ-tʂəlje:	wɤlũ	mĩkɛ	pɛitɕɪŋ
	dem-cl-man	1sg-acc	say- impf	1log	tomorrow	Beijing
	去哩	说。
	tɕʰi=li	ʂʷɔ
	go-fut	quot

‘The man said to me: ‘tomorrow, I will go to Beijing’

(Fieldwork, Lefort 2023b)

wɤlũ 喔咙 can be used as the first person when stating a long-term situation, such as identity, and be used in place of the regular first person pronoun wɤ 我. Consider the examples below (Lefort 2024, fieldwork):

(12)	喔咙	是	老师	是
	wɤlũ	ʂʅ	lɔʂʅ	ʂʅ
	1log	COP	teacher	COP

‘I am a teacher’

(13)	我	看	书
	wɤ	kʰ ɛ̃	ʂu
	1	read	book

‘I am reading a book’

It is also used as an inclusive first-person plural pronoun corresponding to standard Chinese zánmen 咱们, ‘ourselves; we’ (inclusive) (also attested in Ibrahim 1985) which is the counterpart of wɤ-mu 我仫 ‘we’ (non-inclusive)8:

(14)	喔咙	庄稼人	当寨
	wɤlũ	tʂʷãtɕʲaʐə̃	tã-tʂɛ
	1sg.log	peasants	be-ipfv
	‘We are (all) peasants’			(Lefort 2023b, fieldwork).

This semantic distribution of pronouns (logophoric and reflexive) shows that wɤlũ 喔咙 was probably borrowed at an early stage and further developed to an inclusive pronoun based on the original Dongxiang reflexive pronoun.

In addition, both the Dongxiang oruŋ and the Tangwang wɤlũ (vəlũ) 喔咙 are used in an idiomatic expression in which it refers to the person’s overconfidence: Dongxiang oruŋ gie (lit.: ‘to do self’) ‘to be arrogant’, and Tangwang wɤlũ ‘to be arrogant’:

(15)	你	喔咙	什么咧?
	ni	wɤlũ	ʂʅma-li
	2sg	arrogant	what-pf
‘Why are you being so arrogant (toward me)? (Lefort 2023b, fieldwork).

The use of the reflexive pronoun *öxer in the meaning of ‘arrogant’ is not attested in Middle Mongol, but the semantic derivation between ‘self’, ‘selfish’, and ‘arrogant’ is quite easy to make and this lexical development in Tangwang is most likely based on the reflexive pronoun itself: it was probably first borrowed in Tangwang as a reflexive pronoun that later extended its semantic meaning in Tangwang and was borrowed back to Dongxiang9. The use of gie, ‘do’, after oruŋ in Dongxiang could indicate that it is borrowed from Chinese, as gie is also a borrowed verb auxiliary10. Furthermore, wɤlũ 喔咙 ‘to be arrogant’ is also attested in Hetan, but not as a logophoric pronoun. However, as far as we are aware, it is not attested in other Chinese hybrid dialects. Hetan is structurally closer to Linxia than to Tangwang, but has also been extensively in contact with Dongxiang, which strongly suggests that the loan of this lexical item is the result of language contact.

5. Possessive Markers

There are two direct loans from Mongolic languages, and most probably Dongxiang, which are found in Tangwang, namely the third-person possessive marker -ni 尼 and the possessive marker -ne 呢. These two suffixes are solely found in Tangwang and no other hybrid Chinese dialect of the region. As this subject has been discussed by other scholars (Xu 2014, pp. 93–99), we will only make a condensed description here.

Reflexive and possessive suffixes are found in all Mongolic languages and share more or less the same functions. There are specific first, second, and third-person forms that can be added to nouns. In addition, there are two other reflexive–possessive markers whose forms vary from one language to another and have been developed on the possession marker *-ni and possessive reflexive marker *-xAn.

In Dongxiang, the connective (genitive-accusative) marker -ni is formally identical to the third-person marker, as in other Mongolic languages, with reference to the possessor. It is usually attached to the possessed noun phrase, and can also be used when the subject is omitted:

(16)	kiəli-ni	fugiə	fugiə	tɕingən	olu	da- ne
	belly-3	big	big	at.last	give.birth	cannot-ipfv
	‘Her belly was extremely big but she could not give birth.’ (Lefort 2012)

Tangwang possessive suffix -ni functions in a similar manner, and is also attached to the possessed noun phrase:

(17)	我	呢	尕桶尼	bjɛ̃ãː底	气着	撂着	过给寮
	wɤ	nə	katʰũ -ni	bjɛ̃ãː-ʨi	ʨʰi -tʂə	lʲɔ-tʂə	kʷɔ-ki-lʲɔ
	1sg	3sg	bucket-poss	onom-sub	angry-pfv	throw-subv	pass-caus-pfv
‘I was so angry at him I took his bucket and throw it on the floor with a crash’ (Djamouri 2022b, p. 27)

Both in Tangwang and Dongxiang, the possessive marker -ni can be preceded by other case markers.

In addition, Dongxiang features a reflexive possession marker, -ne, which derives from the common Mongolic suffix *-xAn. As the marker -ni, it is affixed to the possessed noun phrase (not the possessor), but unlike -ni, it can be used with first, second, or third-person forms. It indicates that one owns X, where X is the possessed noun phrase (Field 1997), and can be translated as ‘own’ most of the time:

(18)	Bi	aɢɯli-nə	a-lə	ətʂi-yə
	1sg	wisdom-refl	fetch-pur	go-des
	‘I will go (to) fetch my own wisdom’ (Lefort 2023b, fieldwork).

It is also found used in the same manner in Tangwang and, as in Mongolic languages, it can be used in combination with other cases:

(19)	腰里呢	还	背寨些,	背寨		一	圪塔
	jɔ-li-nə	xɛ̃	pi-tʂɛ ɕʲɛ	pitʂɛ		ʝi	kɛta
	hips-loc-poss	also	carry-ipfv	few	carry–ipfv		one	a.few
	‘(We were) also carrying it on our hips and a little bit on our back.’ (Djamouri 2022b, p. 22).

In this case, it is certain that Tangwang has borrowed these two reflexive suffixes from Dongxiang because they are identical on the formal and the functional levels while the set of suffixes differ in all other Mongolic languages of the Hexi corridor, i.e., Bonan 3sgr -ne and poss -ne; Shira Yughur 3sgr –(i)ni and poss -y-aan; Mongghul 3sgr -ni and poss -naa; Mangguer 3sgr -ni and poss -nang.

6. Case Markers

There are two case markers in Tangwang that are formally identical with two Mongolic markers: the limitative case -t^hala and the instrumental -la. The following section examines the possibility of a borrowing from Dongxiang in both cases.

6.1. Limitative Case/Terminative Converb t^hala

The limitative (also found as terminative in the scientific literature) -tala is a Mongolic aspectual converb already attested in Middle Mongol that can be used after verbs to express ‘until X is complete’, where ‘X’ is an action:

(20)	Ger-ün	belge	sün	tusür-ëd	esüg-i-yen	söni-de	üdür
	yurt-gen	sign	milk	pour.out-after	kumiss-refl	night-loc	day
	cayi-tala	büle-ku		bü-lee.
	dawn-lim	churn-nmlz		be-pfv

‘The sign of the yurt (was that) after pouring the milk there would be churning of the kumiss at night until day dawned’.

(Street 1957, p. 38)

In Dongxiang (21) and Tangwang (22), the limitative converb is used in a similar manner:

(21)	qara	olu-tala	wiliə	giə-wo
	black	arrive- lim	work	do- prv
	‘He worked until dark’ (Kim 2003, p. 360).

(22)	学校	里	去-他啦	你	这里	来
	ɕᶣɛɕʲɔ	li	tʂʰi-thala	ni	tʂʅli	lɛ
	school	inside	go-lim	2sg	here	come
	‘Before going to school, come here’ (lit. until you go to school) (Ibrahim 1985, p. 44)

In addition, -tala and -t^hala are found both Dongxiang (23) and Tangwang (24) in comparative constructions:

(23)	坐-他啦	睡觉 咧
	tsu-thala	ʂʷitɕʲɔ li
	sit-lim	sleep-pf
	‘Sleeping is better than sitting’ (Ibrahim 1985, p. 44)

(24)	Landʐou	bi-tala	badza	sao-dʐi	wo
	Lanzhou	be-lim	Linxia	live-prog	cop
	‘Better live in Linxia than in Lanzhou’ (Kim 2003, p. 360)

In Middle Mongol sources, in addition to its terminative functions, -tala is sometimes used with the meaning of ‘although’, but not the alternative action meaning ‘rather than’ or ‘better’, which suggests that the comparative function of -tala is probably a secondary development.

In addition, Tangwang has developed the word nũntʰala 弄他啦 (<nũ 弄 ‘this.way (v.) + t^hala 他啦 LIM (conv.) ‘before doing X’, ‘rather’, which has become an indecomposable adverb:

(25)	你	东乡里些	买	山野	去着	弄他啦	这它些
	ni	tũɕʲãɕʲɛ̃-li-ɕʲɛ	mɛ	ʂɛ̃jɛ	tɕʰi-tʂə	nũntʰala	tʂətʰa-ɕʲɛ
	2sg	dongxiang-loc-abl	buy	potato	go-prog	better	here-abl
	买	吧啦!
	mɛ	pala
	buy	pf
‘You are going to Dongxiang to buy potatoes, (you’d) better buy them here’
(Djamouri 2022a)

In Dongxiang, the equivalent expression ingie-tala is not attested, which suggest that -tala in Tangwang also functions as a productive suffix that developed independently and is not due to language contact.

It is most probable that the converb -t^hala is a direct loan from Mongolic languages, but it is not certain that it was loaned from Dongxiang. Indeed, it is found in all Mongolic languages (except Moghol) as a terminative converb and is attested in the Shirongolic languages as a comparative converb in Bonan and Mangghuer (but not in Shira Yughur nor in Mongghul). Moreover, both terminative and comparative functions are found in Dagur. In the Chinese hybrid dialects of the Hexi corridor, -t^hala is found as a terminative converb in all dialects but is not attested as a comparative converb in Xining, Wutun, and Linxia.

In addition to the terminative converb 他啦, Peyraube (2018) has identified a terminative nominal case in Tangwang, t^hala 他啦, which sets the boundaries of a moment in a given set of actions:

(26)	你	早上-些	这	会-他啦	做	嘛寨
	ni	sɔʂɛ̃-xie	tʂɛ	xʷi–thala	tsu	ma-tʂɛ
	2SG	morning-abl	dem	time-lim	do	what-ipfv
	‘What have you been doing from this morning until now?’ (Xu 2017, p. 97)

A marginal nominal terminative marker *-cAA is attested in Middle Mongol, Buryat, Mongol proper, Ordos, and Oirat (with Kalmuck) (Schönig 2003, p. 416), which is also used in equative constructions. It is rarely used, and the dative–locative case is preferred. The terminative functions of the Mongolic marker *-cAA are similar to Tangwang’s t^hala 他啦 but is formally very different and unlikely to be the source of a borrowing.

The most common strategy in Mongolic languages to express ‘until’ is usually verbal, through the use of converbs (*tala), verb-based postpositions developed by the aggregation of the verb and the terminative converb -tal (e.g., Modern Mongolian hurtel ᠬᠦᠷᠲᠡᠯ᠎ᠡ ‘until’< hur- ‘arrive’ + -tel term; boltol ᠪᠣᠯᠲᠠᠯ᠎ᠠ ‘until’< bol- ‘arrive’ + -tol term), or verbs in combination with the ablative case on the head-closed noun (which sets the beginning of the action). Examples from Dongxiang are provided below:

(27)	bi	eqiemagva-se	yizhi	gongzuo	gie-zhuo	xieni	kuru-wo
	2sg	early.morning-abl	all.the.way	work	do-prog	night	arrive-pfv
	‘From this morning, I have been working until dark’    (Buhe 1987, p. 31)

(28)	Jiu	ede	liaozhe,	hhe	goyene	xiaojiao-ne	usu-de	nie	chujie-gve-se
	so	now	until	3sg	own	shadow-refl	water-loc	one	see-cs-seq

daishun	qiorun-ne	taila-ne.
right.away	head-refl	raise-ipfv

‘So, until now, each time he sees his reflection in the water, he raises his head up right away’.

(Buhe 1987, p. 71)

It is most probable that the terminative case t^hala was developed from the limitative converb borrowed from Mongolic languages’ t^hala and is the result of an internal development rather than a reanalysis of the Middle Mongol terminative case.

6.2. Comitative–Instrumental Case -la

Tangwang features a comitative case (29) and an instrumental case (29), which are formally identical:

(29)	我		呢啦			气	打哈寨		(Xu 2017, p. 91)
	wɤ		nə-la			tɕʰi	ta-xa-tʂɛ
	1sg		3sg-com			angry	fight-res- ipfv
	‘I have quarrelled with him’.
(30)		这		个	老奶奶			过来过去		簸箕啦	护寨 (Djamouri 2022a, p. 39)
		tʂə		kə	lɔnɛnɛ			kʷɔlɛ-kʷɔʨʰi		pɔtʰɕi-la	xu-tʂɛ
		dem		cl	old.woman			come-go		sieve-inst	cover-ipfv
		‘This old woman was coming and going covering (her body) with a sieve’.

In Standard Chinese, comitative and instrumental functions are borne by the prepositions gen 跟 ‘with’ and yong 用 ‘with’11. Therefore, the Tangwang marker la 啦, being formally very close to Mongolic languages, is most probably a direct loan. However, this marker does not seem to be directly related to Dongxiang (see

Table 3. Instrumental and comitative case in the Mongolic languages.

	Middle Mongolian	Mangghuer	Mongghul	Dongxiang	Bonan	Shira Yughur
Instrumental	-ar~-iar~-niar	-la	-la	-ghuala~-ghala	-ghuala~ghala	-aar
Comitative	-lU’A > -laa	-la -tai (poss)	-la -dii (-poss)	-le ~ -la	-ghuala~-ghale	-la

below). The Middle Mongol instrumental case*-xAr > *-AAr is not attested in any of the Mongolic languages of the Hexi corridor, except Shira Yughur. As stated by Rybatzki (2003, p. 337), this could mean that the instrumental case was never present in these languages, but it is more likely that it was secondarily lost and that it has been replaced by the comitative case. The use of the comitative case for the instrumental case is also quite common among other languages in the world (Heine and Kuteva 2005, p. 139). In Mongghul and Mangghuer, the function of the instrumental case is filled by the comitative case in *-lUxA > *-lAA > *-la and has developed an alternative form, -tai and -dii, based on the possessive case, like in modern Mongolian. In Dongxiang, the comitative and the instrumental cases feature two distinct forms: -le and–ghuala~-ghala. The Dongxiang (and maybe Bonan) instrumental case -ghuala ~ ghala (ghua-la lit.: two-pl ‘the two’) is probably a secondary development and a calque from the Linxia instrumental case marker liangge 两个 (lit.: two-cl ‘the two’), which has then been replaced by the marker formally identical to the comitative12 case.

The different forms of the instrumental and comitative cases in the Mongolic languages of the Gansu–Qinghai area are summarized in Table 3 below:

Tangwang instrumental and comitative could have been borrowed either from other Mongolic languages in the region, or even from other Mongolic sources. Given the geographic distance between Mangghuer and Mongghul, it is quite unlikely that Tangwang borrowed the instrumental and comitative case ‘as is’ from those languages through relatively recent contacts. It is more probable that, as it is the case in other Mongolic languages of the Hexi corridor, the instrumental case has been first replaced by the comitative case in Dongxiang, from which Tangwang borrowed both the instrumental and the comitative cases. Later, the instrumental case has been replaced by a calque of Linxia liangge 两个 (lit.: two-cl ‘the two’) in Dongxiang, developing the secondary form ghuala~ghala (ghua-la lit.: two-pl ‘the two’), while only the comitative form la 啦 remained in Tangwang for both the comitative and instrumental cases. Another hypothesis is that Tangwang borrowed (or inherited) the comitative and the instrumental cases at its early stage of formation from Middle Mongol and, as other Mongolic languages of the region, has replaced the instrumental case by the comitative case form. This would be an internal development.

7. Discussion

Although Tangwang features many characteristics that resemble Altaic languages, only a few can be attributed to direct contact with Dongxiang. Dongxiang and Tangwang partially share formally identical functional words and suffixes, but the direct borrowing from Dongxiang is certain only for the possessive markers (third-person possessive -ni 尼 and reflective marker -ne 呢) and the logophoric pronoun (wɤlũ 喔咙). Other syntactic features (instrumental -la and limitative t^hala) are probably of Mongolic origin, but the certainty of Dongxiang being the source cannot be established. From the above, we can classify Tangwang’s ‘Altaicised’ features resulting from in contacts in at least three types:

(1)

Regionally shared features: Western northern Chinese common Altaic features (plural markers on unanimated objects, tense, and aspectual markers, distal demonstrative and third person syncretism, progressive converb, quotative markers);

(2)

Areally shared features: features shared in the Gansu–Qinghai area (partial SOV word order, cases, terminative converbs, formal distinction for the world ‘little’ with attributive and predicative functions);

(3)

Tangwang-specific features: Dongxiang loans (few direct content words, first-person logophoric pronoun, reflexive suffix, possessive suffix).

The first type of contact has probably preceded the second and third type, but the development between the second and the third type is uncertain. Certain evidence, such as the mismatch between the comitative–instrumental case form with the Dongxiang form, or the fact that the distinction for the world ‘little’ is shared in all Gansu–Qinghai hybrid Chinese languages, tend to point at an earlier influence from Mongolic languages on the areal level and a formation context common to other Chinese hybrid dialects.

The mechanisms involved in the contact-induced process based on the existing models have been partially discussed by Xu and Wen (2017), who believe that the language shift with imperfect learning from Dongxiang to Sinitic is not possible given the results of their genetic analysis13. Rather, and according to them, the situation of Tangwang (partially) corresponds to Thomasson and Kaufman’s borrowing scale’s “very strong cultural pressure: heavy structural borrowing structure” (which includes word order, inflectional morphology, and other major syntactic changes). However, the problem is that the ‘strong cultural pressure’ can be only partially proved14 and that those kinds of heavy pressure situations are ‘preceded by more casual contact situations, characterised by borrowings of content and function words’ (Thomason and Kaufman 1988). In the same manner, Heine and Kuteva’s (2005, p. 14) concept of grammaticalization is defined as a process leading from lexical to grammatical forms and from grammatical to more grammatical forms. This does not correspond to the situation of Tangwang, as we have seen, in which borrowings are scarce. For the most part, the Tangwang vocabulary remains strictly Chinese, and only little evidence of borrowings from the Mongolic lexicon can be seen. A good part of the vocabulary that is listed as a Dongxiang loan in previous studies is, in fact, common Mongolic areal vocabulary that can be found across most of the other Chinese hybrid languages spoken in the region.

Another scenario would be relexification, which would explain the scarcity of non-Chinese vocabulary and would involve the hypothesis that the population initially spoke a Mongolic language. According to Xu and Wen’s genetic evidence, this could be possible (Xu and Wen 2017). Relexification can occur in two kind of situations: bilingualism (Muysken 1981) and second language acquisition with imperfect learning in a dominance situation, such as in the case of the emergence of creole languages (Lefebvre 1998). In the case of bilingualism, recent evidence from bilingual communities which speak both Mongolic languages and Chinese has shown that the syntactic elements remain, for the most part, Mongolic, while the content items are replaced by their Chinese counterpart—see, for example, Dörbed Mongolian (Bao 2011) and Shineken Buriat (Yamakoshi, forthcoming). In that case, how could the formal discrepancies with the Mongolic case system be explained? If further grammatical reanalysis, or ‘relabeling’ in Lefevre’s meaning (Lefevre 1998), with Chinese material had occurred in Tangwang, the functions of the Mongolic nominal cases would have remained, as this process involves the use of equivalent stems from the receiving language. Moreover, more diverse converbial and participle markers would have been found, mirroring the complexity of the Mongolic verbal system. In the case of second language acquisition with imperfect learning, the process of relabeling involves a “very limited access to the superstrate data” (Lefevre 1998). This means that Chinese would have been the dominant language, but not accessible to the Tangwang speakers. However, we lack historical evidence to confirm this fact. In any case, the main point of opposition for the relexification hypothesis is that it contradicts the scenario which involves the migration of the Tangwang people from Sichuan and fails to explain why traces of the Sichuan dialect are still found in today’s lexicon.

Other models involving restructuring, such as metatypy, which Ross (Ross 2006) defines as a diachronic process whereby the morphosyntactic constructions of one of the languages of a bilingual speech community are restructured on the model of the constructions of the speakers’ other language, which is preceded by lexical loans and calques, seem to match better, as they can involve ‘imitation’ rather than direct copying. However, this kind of syntactic change emerges in situations where the features are replicated from the model language (in this case Mongolic), a dominant language, and very often, in bilingual communities. In the modern contexts, Dongxiang is not a dominant language and bilingualism is only partially shared among the Tangwang community. The partial bilingualism with Dongxiang can only explain the direct loans. If bilingualism had happened, it would have been during the Mongolian period, before the supposed settlement of the Wang family in the region.

Therefore, it is possible that the ancestors of the Tangwang people already spoke a Chinese variety before their settlement to their actual location. Prolonged contact with Dongxiang would have helped to maintain the Altaic changes that were already present in their speech. The superstratal influence visible in all northern Chinese dialects has evolved differently in dialects spoken further east, because the contact with Altaic languages had ceased. Due to the local contact with Dongxiang, Tangwang has further developed. However, the presence of nominal case and even direct loans cannot be solely explained by the contact with Dongxiang.

8. Conclusions

This paper analyzed the formally identical function words found in Tangwang and Dongxiang and it has been shown that, like the direct loans of content words, Tangwang’s direct loans of function words from Dongxiang are rare and can be certain in only three cases. Generally speaking, Tangwang is very hermetic to direct loans and adopts a calque strategy, which makes it difficult to fit in the existing models of language contact. However, this trait is not exclusive to Tangwang and is shared by all Chinese dialects.

Although the Tangwang people are surrounded by Dongxiang speakers and partial bilingualism is attested in today’s Tangwang community, it is difficult to prove that Dongxiang was the sole source language of influence that can explain both syntactic changes and the presence of function words resembling their Mongolic counterparts. On the contrary, the Dongxiang influence seems to be sporadic and several layers of Altaic influence can be identified. However, we still lack historical evidence to fully understand the process, and further research should be made in that direction.