Kazakhstani Gansu Dungan as a Contact Language: An Analysis of Russian Influence

Abstract

This paper discusses extensive language contact and its results in Kazakhstani Gansu Dungan, a divergent variety of Mandarin Chinese. Based primarily on recorded conversational source materials, this study offers a contact linguistic overview of the language, introducing both phonological and morphosyntactic contact phenomena. It is shown that Kazakhstani Gansu Dungan is currently under extensive Russian influence. The influence permeates all layers of the language and exceeds lexical borrowing mentioned in earlier Dungan studies. For instance, clause combining and complex clauses in Dungan have shifted to the direction of a Russian model, which makes the language stand out among other Sinitic varieties. This study demonstrates that, in addition to introducing new structures, extensive Russian influence on Dungan also reinforces earlier development that has led the language further away from the Sinitic prototype. In all, Kazakhstani Gansu Dungan forms its own kind of divergent Russianized Sinitic variety and thus offers a contribution to both researching language contact in the Russophone world and to understanding the typological diversity of Sinitic languages.

1. Introduction

The present paper analyzes Kazakhstani Gansu Dungan from the viewpoint of contact linguistics and focuses on Russian influence on the language, termed ‘Russianization’1 in the analysis and discussion that follow. The introduction consists of four parts: goals of this paper (Section 1.1), sources and methodology (Section 1.2), conventions (Section 1.3), and structure of this paper (Section 1.4).

1.1. Goals of This Paper

In recent years, ‘divergent’ Sinitic varieties that typically manifest either Altaic or Tibetan contact history have received increasing attention in Chinese linguistics. The trend is exemplified, among others, by the descriptive grammar of Wutun (Sandman 2016), a contact language of the Amdo Sprachbund, and Zhoutun (Zhou 2022), a Tibetanized Northwestern Mandarin variety in the same area. At the same time, Dungan varieties offer yet another Sinitic contact variety type largely neglected in earlier research, especially in Western Sinology.

Dungan research has a long history that has resulted in a sizeable body of publications in various languages, including several descriptive grammars. While earlier research, such as the grammar of Lin (2012) and Wang et al. (2015), mentions and acknowledges Russian contact and its linguistic consequences, it primarily analyzes Dungan as yet another Mandarin Chinese ‘dialect’. The sentiment is concisely summarized by Comrie (1981, p. 273): “In terms of its structure and basic vocabulary, Dungan is a form of Mandarin (Northern) Chinese, not standing out particularly from other Northwestern dialects of Mandarin”. Such an approach focuses on Russian influence primarily in the lexical domain, where it surfaces saliently due to numerous borrowings. On occasion, morphosyntactic influence on Written Dungan also receives some mention. The present paper, however, demonstrates that Russian influence in spoken Kazakhstani Gansu Dungan permeates the whole language and is far more extensive than previously thought.

Based on recordings of spoken Gansu Dungan, including natural spontaneous language use, the present study argues that the language is taking steps that lead it away from features that characterize the well-known prototype of a Sinitic language. Chao (1968, p. 13) famously claimed that “There is practically one universal Chinese grammar”. While the Sinitic languages have many core features in common, research in “outlier” cases, such as Dungan, is gradually showing that many deviations from the prototype that matches most closely with the features of Standard Mandarin exist.

Russian influence in Dungan extends to morphosyntax and has caused the remodeling of Kazakhstani Gansu Dungan grammar to follow Russian patterns more closely. Unlike earlier influence from Turkic languages also identifiable in Dungan, most conspicuous forms of Russian influence are likely relatively recent and thus not yet well reflected in older Dungan research. In addition to introducing new structures based on a Russian model, extensive Russian influence also reinforces earlier development of Dungan that has led the language to deviate from the Sinitic prototype.

The present study is not an exhaustive analysis of all contact-induced changes in contemporary Kazakhstani Gansu Dungan. On the contrary, by selecting many of the most prominent contact phenomena neglected in earlier work for analysis, the present study lays a foundation on which further contact linguistic research of the language can be built. Also, contact-induced change in Written Dungan is largely omitted since the study focuses on conversational language use.

From the viewpoint of contact linguistics, this study offers two contributions. First, it connects to the ongoing study of ‘Russianization’ in languages of the Russophone world. Many contact features that Dungan manifests are not restricted to the language but surface in other languages that have been in intense contact with Russian. Second, Dungan additionally offers a contribution to investigating the scope and possibilities of typological variation across Sinitic languages.

1.2. Sources and Methodology

This study is based on Dungan source materials collected through linguistic fieldwork in Kazakhstan in 2022–2023. The materials were collected from Astana, the capital of Kazakhstan, and Dunganovka, a Dungan village outside Taraz in the southern part of the country. Most of the source materials that this study relies on originate from proficient native speakers who continue using Dungan in their daily lives, rather than from speakers who have learned the language only imperfectly. The speakers are bilinguals with a fully proficient command of Russian, since finding monolingual Dungan speakers is becoming increasingly difficult, especially among the younger generations. At present, approximately six hours of recordings exist, of which we have fully annotated and translated approximately an hour with multi-layer annotations and metadata. A considerable portion of the source materials consists of recorded conversations, and most examples selected for the paper represent this speech genre.

1.3. Conventions

The present study follows a five-line glossing system instead of the usual three-line approach common in linguistics. The first line offers an etymological spelling where Sinitic elements are written with Chinese characters. In turn, Russian and Turkic are expressed with the Cyrillic script used for Turkic languages in Kazakhstan, such as Kazakh and Uyghur. Dungan has a written language with some degree of use and standardization, particularly in Kyrgyzstan.2 Consequently, the second line presents the Dungan Cyrillic spelling of the example. To some extent, Kazakhstani Gansu Dungan differs from Written Dungan. To address this discrepancy between the varieties and to make this study more approachable to readers with no knowledge of the Cyrillic script, the example is also written with the International Phonetic Alphabet with tone marking. The following three signs are used: $\bar{\mathrm{V}}$ ‘去声 realized as a high tone (45, 44)’, $\acute{\mathrm{V}}$ ‘平声realized as a rising tone (13)’, and $\grave{\mathrm{V}}$ ‘上声 realized as a falling tone (41)’ (tonal values from Wang et al. 2015, p. 40). The historical 入声 category is distributed across the three tonal categories above. The absence of tonal marking indicates the presence of a ‘neutral tone’, namely the neutralization of tonal distinctions. This typically occurs with grammatical enclitics, some of which frequently undergo phonetical erosion in their articulation, as demonstrated in the present article. Finally, while the tone symbols appear visually identical to those used in Hanyu Pinyin, the Dungan tonal system should not be interpreted through them.

Reliable and extensive Dungan lexicographical resources are rare, a situation shared with many other Sinitic varieties. As stated by Janhunen (2020), Salmi’s (2018) dictionary of Dungan is the first linguistic dictionary of the language. Consequently, in addition to the fieldwork investigation, the present study uses the dictionary to ascertain the tonal values of the Dungan lexemes.

1.4. Structure of This Paper

This paper is structured as follows: After the introduction, chapter two offers an overview of Dungan and the speakers of the language, particularly in the context of Kazakhstan. Following, chapter three demonstrates how intensive language contact with Russian is now solidifying phonological development that likely started in an earlier era of Turkic contact. In turn, chapter four presents a selection of morphosyntactic phenomena that result from intensive language contact with Russian. The study focuses on the verbal system, clause combining, and complex clauses that earlier research on contact languages of the Russophone world identified as common domains for structural and formal borrowing. Finally, chapter five, concluding this study, contextualizes Dungan as a Northwestern Sinitic contact variety, examines the future trajectory of the language, and reiterates the presented key arguments.

2. Introduction to Dungan

The present chapter offers a brief introduction to Dungan. It starts with an overview (Section 2.1) with a focus on the term ‘Dungan’ and the sociolinguistic situation of the language. Following, the Dungan varieties and their intelligibility with Standard Chinese (Section 2.2) and the historical origins and migrations of the Dungan speakers (Section 2.3) are discussed. (Section 2.4) addresses why earlier research has tended to neglect contact phenomena in Dungan. The chapter also presents a typological morphosyntactic sketch of Kazakhstani Gansu Dungan (Section 2.5). It concludes with an illustration of Dungan diglossia and how the formality of the speech registers correlates inversely with the presence of Russian influence in the language (Section 2.6).

2.1. Overview: The Term ‘Dungan’ and the Sociolinguistic Situation of the Language

The Dungan language is known in Written Dungan as Хуэйзў йүян /hwízú jỳján/ 回族语言 which means ‘the language of the Hui nationality’. The Hui (Ch. huízú 回族) are Sinophone Muslims, the main population concentration of which lies in the post-Soviet states of Central Asia and Western China. The etymology of the ethnonym ‘Dungan’ remains unclear. Focusing on the source language, Fougner (2012, pp. 23–27) divides the existing proposals into three groups: Chinese, Russian, and Turkic-based. Many oft-repeated and well-known etymological proposals are Sinitic, such as the toponymically motivated dōnggān 东甘 ‘Eastern Gansu’. The term ‘Dungan’ is also attested in the regional Turkic languages, e.g., Chagatay tungan تونکان (see Schluessel 2018). In Russian, it is attested as дунгане.

The Dungan language is becoming increasingly endangered, yet its speakers retain a strong ethnic consciousness. Religion plays an important role in the identity formation and maintenance process of the Dungans, who identify themselves strongly as Muslims and follow the Hanafi School of Sunni Islam (Akiner 1986, p. 355). The influence of Islam also presents itself through Perso-Arabic loanwords.

Ethnologue classifies Dungan, with approximately 110,000 speakers, as endangered (Eberhard et al. 2022). Language loss is attested among some Dungans (Hai 2004). Speakers of the language are aware of an intergenerational gap that is emerging. The young have difficulties understanding the elderly and frequently communicate in Russian, at least with their siblings and peers. Spoken Dungan with Russian admixture is deemed less proper and correct than the idea of a ‘pure’ Dungan where everything can be expressed with native resources without resorting to borrowing or code-switching (see also Section 1.2). The following excerpt from an interview summarizes the current sociolinguistic state of the Dungan language in Kazakhstan and young speakers’ common self-perception as speakers of the language. Since the speaker narrated everything in fluent Dungan, the total lack of linguistic competence is an exaggeration that likely results from the inability to express everything with Dungan lexical resources. At the same time, the concern expressed about the intergenerational gap is real.

That (i.e., Dungan) is what my grandma and grandpas speak. Myself, I don’t know anything. I don’t understand half of their talk. (I speak Dungan) only at home, with my mum and dad and with the elders. Among the siblings, we speak Russian. All young children speak Russian now.

(interview recording of a young male in Taraz 2023)

2.2. Dungan Varieties and Mutual Intelligibility with Chinese

The Dungan language has two main varieties: Gansu (甘肃) and Shaanxi (陕西) Dungan. Kazakhstani Gansu Dungans call their language җунянхуа /pf$\acute{\mathrm{ə}}$nj$\acute{\mathrm{ə}}$xwā/ (中原语) ‘Central Plains language’. In this paper, ‘Dungan’ is used as a shorthand for Kazakhstani Gansu Dungan, and the findings should not be considered generalizations valid for all Dungan varieties.

Phonology offers a quick way to identify a Dungan variety. Therefore, in addition to the speakers’ self-assertion as Gansu Dungan, I have verified the studied variety from the known phonological differences. For instance, Shaanxi Dungan possesses the initial /ŋ/ that is lacking in Gansu Dungan, as in /ŋɛ/ vs. /nɛ/ 爱 ‘love’ (Zavyalova 2017a). Also, the Shaanxi Dungans are considered culturally more conservative (Rimsky-Korsakoff Dyer et al. 1992, p. 245). The present study does not investigate the Shaanxi variety. It remains to be seen whether the ‘cultural conservatism’ has linguistic consequences, namely whether the findings apply to Kazakhstani Shaanxi Dungan as well, and if so, to what extent.

The two Dungan varieties differ noticeably, e.g., in the number of their phonemic tones. Because of considerable differences, it is a matter of analysis whether the two should be called Dungan dialects or separate Dungan(ic) languages. The present article leans towards the latter option. To illustrate, when communicating with each other, speakers of Gansu and Shaanxi Dungan in Kazakhstan resort to Russian, the lingua franca of the country. To avoid over-emphasizing the differences, however, it should be mentioned that high competence in Russian may direct towards this communicational choice. Many Dungans already speak Russian in their daily lives. The use of Russian consequently obviates the possibility of linguistic discomfort resulting from efforts to make sense of a variety with noticeable differences from one’s own.

The “Chineseness” of the Dungan language is often emphasized. To illustrate, the claim that a speaker of Dungan can converse with that of Standard Chinese or the Beijing dialect is often repeated upon mentioning the language in brief discussions of a general nature (see, e.g., Dong 2020, p. 176). Yet no systematic study of mutual comprehension exists, and the degree of such claims should be rigorously measured. At a higher level, this reflects the “myth of mutual intelligibility” among spoken varieties of Mandarin. As Szeto et al. (2018, pp. 21–242) show, homogeneity and mutual intelligibility among Mandarin varieties are often exaggerated, whereas in reality, speakers of different Mandarin dialects may struggle to understand each other even in the same province.

2.3. Historical Origins and Migrations of the Dungans

The Dungans are an ethnic minority of Kazakhstan, a multinational state. Despite the group’s marginal status, the Dungans are nevertheless rarely discussed as a ‘minority’ in the domestic context. This reflects what Dave (2007, p. 131) has identified as “the absence of the term‘minority’ in either official or public discourse on ethnic relations”. The Dungan settlements of Kazakhstan are concentrated in the south along the Kyrgyzstani border. Among other things, economic opportunities have drawn a sizeable Dungan population into the north, epitomized by the capital Astana, where several hundred households now reside.

The Dungan originate from a relatively compact area that covers southern Gansu and western Guanzhong in Shaanxi (Zavyalova 2017a). The Dungan presence in Central Asia was caused by the power politics of the 19th century. The British defeated the Qing dynasty in the first Opium War (1839–1842), which triggered local revolts around the empire and weakened it further (Khalid 2021, p. 79). As a result of a failed revolt against the Qing government, the Dungans had to cross into the Russian Empire to seek refuge. Several migration waves occurred, and they also involved migrating Uyghurs, with whom the Dungans often settled into a new territory in the Russian Empire (see Smagulova 2016, p. 77). According to Rimsky-Korsakoff Dyer (1990–1991, p. 301), two major migrations took place between 1864 and 1887, the first one from the Ili valley and the second one consisting of supporters of the autonomous state of Kashgaria. Rather than being one-off events, Dungan migrations continued on a smaller scale even during the early 20th century until the Sino-Soviet border was closed in the 1960s (Allès 2005, p. 122).

The migrations have a linguistic dimension. First, migrating and settling together with the Uyghurs meant that Dungans remained in contact with the Uyghur language even after the relocation. Jochelson (1928, p. 105) reference to “a corrupt Turkic dialect of Uigur origin” among the Dungan illustrates that Uyghur was widely spoken among the Dungans in the past. Even today, some Kazakhstani Dungans can speak Uyghur (Smagulova 2016, p. 77).

Second, as the links with the Sinophone heartland weakened, Dungan lost many of the elements belonging to the register of Written Chinese (书面语). Since Dungan was typically a spoken language even prior to the migrations, it is hard to ascertain to what extent such features have ever existed in the language. In any case, as Harbsmeier (2015) points out, the independence of Dungan from the Chinese writing system has shaped the development of the language. For instance, severed links with Written Chinese resulted in the loss of literary expressions and sayings, such as chéngyǔ (成语).

Standard Chinese possesses many words that are polysyllabic but monomorphemic, such as the famous example of húdié 蝴蝶 ‘butterfly’. Dungan, however, has additionally monomorphemized several Sinitic words that are etymologically polymorphemic. In other words, the compositional etymological structure of the word has been lost, and this may be due, at least to some extent, to losing contact with the Chinese writing system acting as a reminder for the morphological structures of the words.3 As a representative case, бу щин /p$\acute{\mathrm{ə}}$ɕín/ 不行, ‘to be sick’ that etymologically means ‘not to go well’, is now perceived as a single morpheme rather than two. The monomorphemization originates from the negation of the Sinitic verb 行 xíng, ‘to go, be alright’. If the monomorphemized verb is to be negated in contemporary spoken Dungan, a sequence of two etymological negation markers results (1). The conclusion of this paper will return to the topic of Sinitic etymological awareness by contemplating the possibility of re-establishing more intensive linguistic contact with Chinese and its potential linguistic effects.

(1)	都不不行了。
	Ду бу бущинли.
	tú	p$\acute{\mathit{ə}}$	p$\acute{\mathit{ə}}$ɕín=le.
	all	neg	be.sick=pfv
	‘... (and let) no-one be sick.’ (New Year’s wish)

2.4. Neglect of Contact Phenomena in Earlier Research

There are three main reasons why the extent of Dungan’s morphosyntactic contact with Russian has received less attention than it deserves. First, some of the earlier research materials that have played a pioneering role in establishing Dunganological studies are already relatively old, e.g., Rimsky-Korsakoff (1967), a study built on publications in Written Dungan. Turkic language contact has a long history with Dungan and likely predates the Dungan westward migrations into the Russian Empire. On the other hand, the intensification of language contact with Russian in its full magnitude is a more recent phenomenon. During the Soviet era, the Dungans were predominantly agriculturalists inhabiting rural areas where they participated in collective farming, often in ‘Dungan kolkhozes’ (Rimsky-Korsakoff Dyer 1990–1991, pp. 302–3). At present, however, the Dungan communities are witnessing a “collapse of the traditional trade and way of living” (Smagulova 2016, p. 83). Many migrate to cities where interaction with other ethnic groups in Kazakhstan becomes a daily matter, often in Russian. In sum, it is possible that not all the contact phenomena described in the present paper existed, at least in their full form, when some of the earlier studies were conducted.

Second, Kazakhstan is known for strong prescriptive views and attitudes regarding language and its perceived ‘purity’. At the national level, the aim of ‘purity’ revolves around the Kazakh language. Ethnic Kazakhs are expected to fulfill the performative rule of being ‘pure’ and ‘authentic’ (Fleming 2019, p. 72). Consequently, code-switching to Russian or speaking a Russianized variety of Kazakh can be seen as an obstacle to the imagined ‘purity’ of Kazakh. Kazakhstani Dungans share this purist language attitude. Therefore, an investigator must be careful when researching the language to be able to capture the real spoken form of Dungan, rather than an idealization of the language that is easily offered to an outsider.

Finally, the principle of “you get what you ask for” that applies to linguistic fieldwork in general holds relevance for Dungan. The reason Dungan research has not documented many of the key contact features presented here lies in the neglect of conversational source materials, which remains a persistent issue in linguistic fieldwork more broadly. Elicitation, especially when conducted through translation from a lingua franca, the collection of folk stories, and analysis of Written Dungan will not show the extent of language contact, although, as Hai (2006, p. 148) demonstrates, even the formation of the literary standard for Dungan manifests a noticeable Russian impact that separates Written Dungan further from Standard Mandarin. To conclude, Dungan is a contact language, the full extent of which is hidden in plain sight. Only a focus on natural conversation reveals the depth and intensity of the language contact.

2.5. A Brief Typological Morphosyntactic Sketch

This section provides a brief morphosyntactic-typological sketch of Dungan. Dungan is an outlier Sinitic language. While the core of the language remains recognizably Sinitic, extensive language contact throughout history has shaped Dungan even before the current wave of Russianization. To keep this section manageable, some of the offered brief examples originate from non-conversational contexts to illustrate patterns confirmed to exist in spoken Dungan.

The pronominal system of Dungan corresponds to three-term systems that are relatively common in Sinitic languages (Chen 2015, p. 94). Dungan adnominal demonstrative pronouns make a three-way distinction (Julie Lefort p.c., 4 August 2023). In other words, a different pronoun is used for referents that are close to the referee, some distance away, or further away (

Table 1. Dungan demonstrative pronouns.

Distinction	Pronoun	Gloss
Proximal	/ts$\overline{\mathrm{ə}}$(kə)/ 这（个）	this
Medial	/n$\overline{\mathrm{ə}}$(kə)/ 那（个）	that
Distal	/nē(kə)/ 那（个）	that (yonder)

). The same seems to apply to locative demonstratives that equally manifest three distance-based distinctions, a topic for further research.

Dungan has a typical Northern Sinitic person pronoun system where the plural personal pronouns are formed with the addition of the pluralizing enclitic му /=mə ~ =m/ 们 (

Table 2. Dungan personal pronouns.

Person	Number
	Singular		Plural
1	/v$\grave{\mathrm{ə}}$/ 我	I	/ò=mə/ 我们	we
2	/nì/ 你	you (SG)	/nì=mə/ 你们	you (PL)
3	/tʰà/ 他她	s/he	/tʰà=mə/ 他们	they

). Only in the case of the first person does asymmetry arise in spoken Dungan regarding plural formation. Instead of the expected */v$\grave{\mathrm{ə}}$=mə/, the first-person plural pronoun is вәму /ò=mə/ 我们. Also, the second-person plural pronoun has evolved a secondary function of politeness when used in the singular, a topic discussed later in this paper (see Section 4.1). Finally, the language has a reflexive pronoun in the shape of гәҗя /kétɕá/ 个家 ‘self’, which is widely shared in the Sinitic languages of the region.

Dungan is less topic-oriented than Standard Mandarin (Salmi 2023, p. 16). In other words, the language has taken steps from topic to subject prominence. Among other things, this manifests through the extensive use of clitics in noun phrases to mark case relations. In terms of the history of their forms, Dungan case enclitics are mostly Sinitic, as the ablative чoчў /ʈʂʰōpfə= ~ ʈʂʰōpf=/ 朝着 coding source (2).

(2)	朝着рoзетка上取掉。
	Чoчў рoзеткахoн чүдё.
	tʂʰōpf=razjetka=xōŋ	tɕʰỳ=tō4
	abl=socket=loc	take=comp
	‘Take it (i.e., the device) from the power socket’. (conversation)

Several Northwestern Sinitic contact varieties, such as Linxia (临夏), Tangwang (唐汪), and Gangou (甘沟), have evolved a case system in which Altaic languages of the Turkic and Mongolic types serve as the sources of some of the case markers (Peyraube 2017). Dungan shows less Altaic influence in this respect. However, following Sandman’s (2016) terminology from Wutun, a terminative case тала /=tʰala/ ‘until’ can be identified in the language (3). It must have been borrowed into Dungan early on, namely prior to the westward migrations (see Section 2.3). The ultimate origin of this regionally attested morpheme likely lies in the Mongolic languages (Peyraube 2017, p. 125).

(3)	我到后半儿тала闲的呢。
	Вә дo хубартала щяндини.
	v$\grave{ə}$	tō	xūb$\acute{ə}$=tʰala	ɕján=tini.
	1sg	reach	evening=term	be.free=ipfv
	‘I am free until the evening’. (constructed)

Clitics are not only a property of the nominal system but also surface in verb phrase syntax, where they, among other things, mark TAM distinctions and verbal complementation.5 They can undergo chaining in what takes the maximal form of three enclitics carried by the predicate (4) in the source materials.

(4)	把人家也喊上了吗？
	Ба жынҗя е ханхoнлима?
	pá=r$\acute{ə}$ntɕà= jè	xàn=xōŋ=le=ma.
	acc=3=also	call=comp:up=pfv=q
	‘Did you invite them also?’ (conversation)

The close bond of the enclitics with their hosts is substantiated by their phonological erosion in the spoken language, e.g., the imperfective aspect marked by the enclitic tini > tni (5). This phenomenon is affecting Dungan phonotaxis discussed further in Section 3.

(5)	Зульфия, 你咋么个？好的呢吗？
	Зульфия, ни замугә? Хoдинима?
	zuljfija	nì	tsàmúkə.	xò=tni=ma?
	Zulfiya	2sg	how	be.good=ipfv=q
	‘How are you, Zulfiya? Are you doing well?’ (WhatsApp message)

A prominent feature of Dungan is the frequent use of the versatile dative-locative enclitic хoн /=xōŋ/ 上, one of the most commonly occurring lexemes of the language. The main functions of the enclitic include encoding spatial location (6), goal (7), and possessor (8). Also, as example (8) demonstrates, the existential-possessive verb ю /jù/ 有can be omitted in Dungan. Cognates for /=xōŋ/ exist in other Sinitic languages of the adjacent regions, such as Zhoutun (see Zhou 2022, pp. 26–28).

(6)	我把你们见了Инстаграм上。
	Вә ба ниму җянли Инстаграмхoн.
	v$\grave{ə}$	pá=nì=mə	tɕjān=le	instagram=xōŋ.
	1sg	acc=2=pl	see=pfv	Instagram=loc
	‘I saw you on Instagram’. (WhatsApp message)

(7)	Базар上去，打了些肉，买了些子青货。
	Базархoн чи, дали ще жу, мэли щезы чинхуə.
	pazə=xōŋ	tɕʰī,	tà=l	ɕé	rū,	mè=l	ɕé-zə	tɕʰíŋxw$\overline{ə}$
	market=loc	go	hit=pfv	some	meat	buy=pfv	some-dim	fruit
	‘I went to the market. I bought some meat and some fruits’. (daily activities)

(8)	我们上三个娃娃：两个儿子，一个丫头。
	Омухoн сангә вава: лёнгә эрзы, йигә яту.
	ò=m=xōŋ	sán=kə	vávà	ljàŋ=kə	$\acute{ə}$zə,	jí=kə	játʰu.
	1=pl=loc	three=clf	baby	two=clf	son	one=clf	daughter
	‘We have three children: two sons and one daughter’. (personal history)

Another oft-noted feature that distinguishes Dungan lies in its high preference for marking the P argument with ба /pá=/ 把 (9a), even in cases where Standard Mandarin cannot do so, as in (9b). The use of /pá=/ in Dungan qualifies as differential object marking (see Chappell 2015, pp. 19–24 for a typological survey of DOM constructions in Sinitic). Many scholars, such as Hai (2011, p. 143) and Lin (2003, p. 84), attribute it to Altaicization, since in the ‘Altaic languages,’ P arguments generally precede the predicate.6 In Dungan, P arguments high in their referential properties of animacy and definiteness, such as pronouns and personal names, must be marked with /pá=/. This occurs even with emotive verbs of low transitivity and controllability where the marking is not permitted in Standard Mandarin (see also Hai and Wang 2002, pp. 49–50), e.g., нэ /nē/ 爱 ‘to love’, щён /ɕòŋ/ 想‘to miss’, and җыдo /ʈʂ$\acute{\mathrm{ə}}$tó/ 知道 ‘to know’. In short, Dungan has progressed further in its development towards morphologized coding of noun phrases' semantic roles with a case marker. The question of frequent differential object marking with /pá=/ also directly relates to the canonical word order discussed at the end of this section.

(9a)	我们把Тараз想呢。
	Вәму ба Тараз щённи.
	ò=mə	pá=taraz	ɕòŋ=ni.
	1=pl	acc=Taraz	miss=stat7
	‘We miss Taraz.’ (personal history)

(9b)	我们很想塔拉兹。
	Wǒ=men	hěn	xiǎng	tǎlāzī.
	1=pl	very	miss	Taraz
	‘We miss Taraz (a lot).’ (constructed)

The nominal numeral classifier system has collapsed in Kazakhstani Gansu Dungan, so that гә /=kə/ 个 remains the only remnant of a more complex erstwhile system. As Hashimoto (1978, p. 251) points out, “the unification of noun classifiers” is likely yet another contact-induced change that has occurred in Dungan. While the present study glosses /=kə/ as a classifier, the ‘classifier’ plays a dummy role in phrase-level syntax. Some previously published Dungan studies report complex classifier systems (see, e.g., Wang et al. 2015, pp. 525–27; Lin 2012, pp. 265–66). The present study could not corroborate this in Kazakhstani Gansu Dungan. While this erosion of nominal numeral classifiers is not solely a feature of Dungan, Sinitic languages that have lost their nominal numeral classifier systems are typically contact varieties. Example (10) demonstrates how Dungan applies the same classifier to all countable nominal referents.

(10)	a.	一个娃娃		b.	两个狗		c.	那个书本子
		йигә вава			лёнгә гу			нэгә фубынзы
		jí=kə	vávà		ljàŋ=kə	kù		n$\overline{ə}$=kə	f$\acute{ə}$p$\grave{ə}$nzə
		one=clf	child		two=clf	dog		that=clf	book
		‘a child’			‘two dogs’			‘that book’

The Dungan system contrasts with that of Standard Mandarin, which retains a complex and contrastive set of nominal numeral classifiers. Standard Mandarin counts with different classifiers for all three referents that Dungan treats equally with /=kə/ (11).

(11)	a.	一个娃娃		b.	两只狗		c.	那本书
		yī=ge	wáwa		liǎng=zhī	gǒu		nà=běn	shū
		one=clf	child		two=clf	dog		that=clf	book
		‘a child’			‘two dogs’			‘that book’

Comparative formation forms a core typological variable among the Sinitic languages. Two main patterns exist: 1. Compare and 2. Surpass (Chappell and Peyraube 2015, p. 134) Dungan follows the Compare pattern of Standard Mandarin, where би /pì=/ 比 ‘than’ is sandwiched between the comparee and standard of comparison (12).

(12)	我们房子里有三个娃娃呢。第二个是我的妹妹。她比我两岁尕。
	Ому фoнзыни ю сангә вавани. Ди эргәсы вәди мими. Та би вә лён суй га.
	ò=m	fóŋzə=ni	jù	sán=kə	vávà=ni.	tī-$\overline{ə}$=kə=s$\overline{ə}$
	1=pl	house=loc	exv	three=clf	child=stat	ord-two=clf=cop

	v$\grave{ə}$=t	mīmi.	tʰà	pì=v$\grave{ə}$	ljàŋ=swī	ká.
	1sg=mod	little.sister	3sg	cmpr=1sg	two=year	be.small
	‘In our household, there are three children. The secod is my little sister. She is two years younger than me.’ (personal history)

In turn, superlative formulations are structurally identical to those of Standard Mandarin, but the superlative marker zuì 最 ‘most’ has been replaced by its Turkic semantic equivalent йин /in/ ‘most’ through borrowing, as in (13). Written Dungan uses дин /tìŋ ~ tɕìŋ/ 顶 ‘most’ for the same purpose.

(13)	我是房子的йин大的儿子。
	Вәсы фoнзыди йин дади эрзы.
	v$\grave{ə}$=s$\overline{ə}$	fóŋzə=ti	in	tā=ti	$\acute{ə}$zə.
	1sg=cop	home=mod	supe	be.big=mod	son
	‘I’m the oldest son in my home.’ (personal history)

Dungan word order is SV in intransitive clauses (14) while transitive clauses show variation. In cases without /pá=/, the canonical Sinitic word order AVP dominates (15). When the object is formally marked, the word order appears as APV (16). Lin (2012, p. 291) attributes such preposing of the object with the resulting APV word order to Altaicization in Northwestern Sinitic languages.8 In any case, establishing a basic word order is more challenging for Dungan than for Standard Mandarin. Since /pá=/ has not reached the status of a universal object marker, the structure can be seen as a marked adjustment of an underlying AVP word order in line with most Sinitic languages. This distinguishes Dungan from many contact languages in Northwestern China with strict APV word orders, such as Zhoutun (see Zhou 2022, p. 18).

(14)	А你夜里哪去了？
	А ни ели на чили?
	a	nì	jēl	nà	tɕʰī=le.
	and	2sg	yesterday	where	go=pfv
	‘And where did you go yesterday?’ (conversation)

(15)	我今儿买鱼了。
	Вә җер(гә) мэ йүрли.
	v$\grave{ə}$	tɕ$\overline{ə}$ː	mè	jɥ$\acute{ə}$=le.
	1sg	today	buy	fish=pfv
	‘I bought fish today.’ (conversation)

(16)	一千九百七十二年我妈Тараз上把我养下了。
	Йичян җюбый чишы эр нян вә ма Таразхoн ба вә ёнхали.
	jí-tɕʰjàn	tɕù-píj	tɕʰíʂ$\acute{ə}$-$\overline{ə}$=nján	v$\grave{ə}$	má	tazaz=xōŋ
	one-housand	nine-hundred	seventy-two=year	1sg	mother	Taraz=loc

	pá=v$\grave{ə}$	jòŋ=xā=le.
	acc=1sg	give.birth=comp:down=pfv
	‘My mum gave birth to me in 1972 in Taraz.’ (personal history)

The behavior of ditransitive clauses constitutes another typological division line in the Sinitic languages. With the verb ‘to give,’ two dominant patterns exist for marking the recipient (R) and theme (T): 1. V R T and 2. V T R, the former dominating in Northern Sinitic (17) and the latter in Southern Sinitic. (Szeto 2019, pp. 70–71).

(17)	我给他钱。
	Wǒ	gěi	tā	qián
	1sg	give	3sg	money
	‘I give him money.’ (Standard Mandarin; Szeto 2019, pp. 70–71)

In Dungan, the lexical verb ги /kì/ 给 ‘to give’ forms a ditransitive construction where a preverbal /kə= ~ k=/ 给corresponds to a dative proclitic marking the recipient and the predicate ги /kì/ 给indicates the action of transferral (18). This is relatively common in Northwestern dialects of Mandarin; see, e.g., Wang and Wang (2003) on the Lanzhou dialect.

(18)	就那一天给我给了медаль了。
	Җю нэ йитян ги вә гили медальли.
	tɕū	nē	jí=tʰjàn	kə=v$\grave{ə}$	kì=le	mjidalj=le.
	dis	dem.dist	one=day	dat =1sg	give=pfv	medal=sfp
	‘On that day, (they) gave me a medal.’ (Dungan history)

In ditransitive constructions involving other verbs, three patterns are attested. They correspond to the typological tendency of the ‘indirect object’ or R to precede the ‘direct object’ or T in Northern Sinitic languages, with Southern Sinitic showing the opposite tendency (Chappell 2015, p. 17). In all the patterns, the R argument is compulsorily marked with the enclitic /kə= ~ k=/. First, when the /= pá / construction is absent from the ditransitive clause, the canonical order surfaces as A /kə/ R V T, as in (19). Following Peyraube’s (2015, pp. 67–68) six typological classes of ditransitive clauses in Standard Mandarin, this is the only pattern shared both by Standard Mandarin and Dungan.

(19)	雀给个家盖窝窝子的呢。
	Чёр ги гәҗя гэ вәвәзыдини.
	tɕʰ$\overline{ə}$	kə=kétɕá	kē	v$\acute{ə}$vəzə=tini.
	bird	dat=refl	build	small.nest=ipfv
	‘A bird is building a small nest for itself.’ (constructed)

In the source materials, however, the /pá=/ construction appears frequently in ditransitive clauses, with the resulting order of A / kə=/ R /pá=/ T V (20).

(20)	妈，我给那们把билеты拿上了。
	Ма, вә ги нэму ба билеты нахoнли.
	má	v$\grave{ə}$	kə=nē=mə	pá=biljet-ɨ	ná=xōŋ=le.
	mother	1sg	dat =dem.dist=pl	acc=ticket-pl	take=comp:up=pfv
	‘Mom, I bought the (flight) tickets for them.’ (WhatsApp message)

Finally, Dungan possesses a double ги /kì/ 给 benefactive-causative construction where the affected is marked by the first /kə= ~ k=/ playing the role of a dative proclitic, and the predicate is followed by a second enclitic /kì/ that occupies the same slot as other verbal complements (21). This construction occurs only with a limited list of predicated, most noticeably фәги /f$\acute{\mathrm{ə}}$=kì/ 说给 ‘to tell’, мэги /mè=kì/ 买给 ‘to buy’, and the currently only attested causative instance in the source materials, namely канги /kʰān=kì/ 看给 ‘to show, lit. to give to look’. The phenomenon exists in other Northern Sinitic languages, as documented by Chappell under the term ‘applicative and causative syncretism’ (Chappell 2023, forthcoming).

(21)	你给他заранее说给。
	Ни ги та заранее фәги.
	nì	kə=tʰà	zaranjije	f$\acute{ə}$=k.
	2sg	dat =3sg	in.advance	say=comp:ben
	‘Tell him in advance.’ (conversation)

2.6. Different Registers of Dungan and Language Contact

Dungan qualifies as a diglossic language. The high (H) variety, namely written Dungan, has a codified written standard and a body of publications, although Dungan publishing has become less common since the collapse of the Soviet Union. Written Dungan is held in high prestige in comparison to the less ‘pure’ spoken register of the language under heavy Russian influence. No speakers of Kazakhstani Gansu Dungan consulted in the present study speak the H variety in their everyday lives, although those who have mastered the H variety may adjust their speech to approximate it more closely in formal settings, e.g., in giving a religious speech. In contrast, the L variety lacks prestige and is often perceived as less proper than H due to Russian influence (see also Section 1.2 on language attitudes and ‘linguistic purity’).

The intensity of Russian presence is intrinsically linked to Dungan diglossia and correlates inversely with the formality of the speech register, as summarized in

Table 3. Different registers and Russian presence.

Speech Register	Russian Presence	Formality
Written Dungan (H)	low	high
Planned speech (M)	moderate	moderate
Conversational speech (L)	high	low

. A similar correlation between formality and speech register does not exist with Turkic presence. In other words, Turkic borrowing generally appears in all three registers with no regard to the formality of the speech event when this is pragmatically feasible.9 Since many Turkic borrowings appear to be more established in Dungan than the newest influx of Russian borrowings, this is to be expected.

Written Dungan shows the least amount of Russian lexical and morphosyntactic influence. This is most noticeable in the lexicon in the relative scarcity of Russian loanwords. To illustrate, (22) is a small description of winter weather first written down in Dungan and subsequently read aloud and recorded. The example contains only lexemes of Sinitic origin.

(22)	今儿个外头冷得很。
	Җергә вэту лындихын.
	tɕ$\acute{ə}$ːkə	vētʰu	l$\grave{ə}$ŋ=ti	x$\grave{ə}$ŋ.
	today	outside	be.cold=comp	int
	‘Today, it is very cold outside.’ (prepared text read out loud)

More casual speech sees the emergence of a wider range of Russian loanwords (23). The Russianization of Dungan syntax begins particularly at this level, a topic discovered with examples in Section 4.

(23)	我们的房子，我们的oтель就мoре跟前呢，一百метр.
	Вәмуди фoнзы, вәмуди oтель җю мoре гынчянни, йибый метр.
	ò=mə=ti	fóŋzə,	ò=mə	atelʲ	tɕū	morʲe	k$\acute{ə}$ntɕʰjàn=ne,
	1=pl=mod	house	1=pl=mod	hotel	dis	sea	next.to=stat

	jí-píj	mjetr.
	one-hundred	meter
	‘Our house (self-correction), our hotel is next to the sea, one hundred meters.’ (travel narrative)

Finally, casual registers of Dungan speech usually contain a noticeable portion of elements of Russian origin. In (24), only the subject pronoun and the interrogative enclitic ма /=ma/ 吗 remain Sinitic, while the rest of the utterance is constructed with words of Russian provenance. In the chapters that follow, the present study focuses primarily on this register of Dungan neglected in earlier research.

(24)	你мoжешь吗oтпрoситься？
	Ни мoжешьма oтпрoситься?
	nì	moʐ-eʂ=ma	atprasjitjsja?
	2sg	can.ipfv.pres-2sg=q	be.excused.inf
	‘Can you get leave from work?’ (conversation)

Many earlier descriptions of Dungan mention contact influences from Persian and Arabic. While this is correct and such influence certainly exists, it seems to be of a lexical nature only, rather than penetrating into morphosyntax as well.10 Moreover, Perso-Arabic loanwords have often entered the Dungan lexicon through an intermediary Turkic language rather than directly from Persian and Arabic (Wexler 1980, p. 297). To illustrate, the Persian حاضِر /hɒːzer/ ‘present,’ itself an Arabic loan, was adopted to Turkic languages of Central Asia, e.g., Kazakh қазір /qazɯr/ ‘now’ and Uyghurھازىر /hɑzɪr/ ‘now.’ The word was subsequently borrowed into Dungan, likely via Uyghur: Written Dungan has хазыр /xazə/ ‘now’ and Spoken Kazakhstani Gansu Dungan ха /xa/ ‘now.’

3. Dungan Phonology in Contact

The present chapter offers a brief overview of the phonological results of Russian language contact. After the introduction (Section 3.1), it is shown that Russian contact reinforces Dungan short forms that emerged before the present intensive contact with Russian (Section 3.2). Finally, consonant palatalization constitutes one of the most conspicuous results of Russian influence in spoken Dungan.

3.1. Introduction

Dungan has been in linguistic contact long before the intensification of the current Russian contact, the contact having been particularly intensive with the Turkic languages. Older loanwords in the language have undergone phonological adaptation. To illustrate, two Turkic loanwords for Kazakhstan exist: 哈萨国 Хасагуй /xásákɥí/ and Қазақстан /qazaqstan/. The former has been subject to phonological adaptation, such as tonalization and fronting of the velar plosive absent from native Dungan words. In contrast, the latter is pronounced identically to the Kazakh pronunciation of the word, with no perceptible phonological adaptation.

Tonalization, phonotactic adjustment, and other types of adaptation are sometimes mentioned to characterize loanwords in Dungan (see Lin 2012, pp. 95–96). In spoken Kazakhstani, Gansu Dungan, this nevertheless only applies to old, established loans. For instance, unlike Standard Mandarin, no tonalization of new borrowings takes place. The new tendency toward direct, non-adaptive integration of loanwords into Dungan makes differentiating such borrowing from other multilingual discourse phenomena, such as code-switching, a challenging task.

3.2. Sinitic Syllable Structure and Dungan Short Forms

All modern Sinitic languages are known for their simple syllable structure, represented as CGVX, where C stands for consonant, G for ‘glide’, V for vowel, and X for consonant or the second part of a vowel or a diphthong (Duanmu 2011). The resulting phonotactically maximal syllables are illustrated in (25) from four Sinitic languages.

(25)	Standard Mandarin:	nián 年	‘year’
	Gan:	sioŋ42 想	‘to think’ (Li 2018, p. 17)
	Shaowu:	kuɔŋ21 光	‘light’ (Ngai 2021, p. 39)
	Cantonese:	gwok3 國	‘country’

Many Dungan words and clitics appear in two forms that the present paper calls full and short. The phenomenon has been identified in pioneering Dungan research, e.g., Dragunow and Dragunowa (1936, p. 38) and Hashimoto (1974, p. 24). The mentions over 80 years ago offer solid evidence of the fact that the phenomenon does not originate from language contact with Russian but possibly from language-internal development. In what follows, however, the present paper argues that the phenomenon has been reinforced by Russian contact.

The short forms result from the deletion of the vowels /i/ and /ə/ only, underlining the high likelihood of their deletion as a result of language-internal development. To illustrate, the word for ‘stone’ corresponding to Standard Mandarin shítou 石头alternates between the full form /ʂ$\acute{\mathrm{ə}}$tʰù/ and its short realization [ʂtʰù]. The short forms do not follow the expected canonical syllabic pattern CGVX. On the contrary, ‘stone’ is CCV in its short form, forming a consonant cluster. From a comparative viewpoint, the phenomenon of short and full forms resembles Matisoff’s (1973, p. 86) sesquisyllabic structure, also known as minor syllable, a phenomenon with “a final heavy syllable preceded by a light and phonologically reduced first syllable” (Butler 2015, p. 443). Sesquisyllabic word structures are widely attested in languages of Southeast Asia.

In rarer cases, the short forms have become the only pronunciations of a word. To illustrate, ‘yesterday’, spelled conservatively as елигә, is now pronounced disyllabically more innovatively as /jēl.kə/. This is the only acceptable pronunciation for the word, and trisyllabic pronunciations, such as /jē.li.kə/, are deemed incorrect by native speakers. The etymological origin of the word is clearly trisyllabic: 夜里 ‘yesterday’ (Zavyalova 2017a) with the generic classifier ge 个. Leaving this and a few other potential exceptions aside, the colloquial short forms are not yet phonemic but in free variation with the long form considered ‘standard’ by the speakers. The phonologized short forms result in new syllable coda consonants and thus create new syllable structures in the syllabic inventory of Dungan

In addition, final vowel deletion occurs frequently in spoken Dungan with /i/ and /ə/. The phenomenon is attested especially frequently with grammatical enclitics in the nominal and verbal domains, as demonstrated in

Table 4. Main weak forms of Dungan grammatical enclitics.

Dungan	Written Chinese	Function
=m ~ =mə	=men 们	pluralizer
=k ~ =kə	=ge 个	generic classifier
=t ~ =ti	=de 的	modification marker
=l ~ =le	=le 了	aspectual marker
=tni ~ =nni ~ =tini	=dene 的呢	aspectual marker
=n ~ =ni	=ne 呢	aspectual marker
=ɕ ~ =tɕʰī	=qù 去	complement verb

. It should be noted that the short forms result in the realization of final consonants not typically present in Northern Mandarin (e.g., -m) or absent from Sinitic languages altogether (e.g., -ɕ).

The reduced forms surface with far more frequency in everyday conversation than in carefully planned speech. To illustrate, see (26). In (27), only the first instance of the perfective aspect enclitic undergoes reduction, highlighting the variability and non-phonemic status of the phenomenon.

(26)	他们把我写的呢
	Таму ба вә щедини.
	tʰà=m	pá=v$\grave{ə}$	ɕjè=tni.
	3=pl	acc=1sg	write=ipfv
	‘They are recording me.’ (conversation)

(27)	阿够了够了все。
	А, гули гули, всё.
	a	kū=l	kū=le	fsʲo.
	interj	be.enough=pfv	be.enough=pfv	all
	‘Yeah, that is enough.’ (conversation)

The short forms emerged either because of Altaicization or through internal development, even before the westward migrations of the Dungans. Heavy Russian-language contact in the present era has nevertheless reinforced the phenomenon. Russian is known for its complex phonotactics and consonant clusters. In short, Dungan short forms have further evolved towards phonotactic adjustment resulting from Russian phonotactics and syllable structure encroaching into the language. To illustrate, the Dungan short form of шыту [ʂtʰu] ‘stone’ discussed above matches perfectly with the Russian syllable CCV, as in чтo /ʂto/ ‘what’. Similarly, the Dungan short form of дини /=tini/ [=tnʲi] ‘imperfective aspect’ is close to identical with the Russian дни /dnʲi/ [dnʲi], save the voicing in the initial consonant. Finally, by altering the syllable structure, frequent use of the short forms in conversational Dungan affects the rhythm of the language and brings it closer to that of Russian.

3.3. Palatalization

Russian manifests phonological palatalization in its consonant system, often illustrated with the minimal pair мат /mat/ ‘checkmate’ vs. мать /mat^j/ ‘mother’. Similarly, Dungan consonants often undergo palatalization before a front vowel. To illustrate, the locative enclitic ни /=ni/ 里 surfaces as [=n^ji]. No minimal pairs where palatalization is contrastive can be identified. Consequently, in contemporary spoken Dungan, palatalization remains a phonetic rather than a phonological phenomenon. It is nevertheless clearly caused by carrying over articulatory habits from Russian that many of the younger generations feel more comfortable with.

4. Morphosyntactic Remodeling

The present chapter offers an overview of morphosyntactic remodeling in Dungan. After the introduction (Section 4.1), the following key topics are covered in the chapter: nominal system (Section 4.2), verbal system (Section 4.3), copular omission (Section 4.4), clause combining (Section 4.5), complex clauses with a focus on relative clauses and conditionals (Section 4.6), and discourse marking (Section 4.7).

4.1. Introduction

In comparison to lexical borrowing, which has been extensively studied, the morphosyntactic results of language contact in Dungan have received less attention. Sometimes scholars consider them minimal. As stated by Hashimoto (1978, p. 257):

While influences of the surrounding languages are often talked about with respect to Zhunyanese [Dungan], in the area of syntax they are at best marginal.

Some earlier research nevertheless addresses morphosyntactic remodeling briefly. This is because Written Dungan where such remodeling is studied has been generally accessible due to a sizable body of literature and other publications.11 It has been highlighted that the constituent-internal order of Written Dungan now approximates that of Russian more closely (Kalimov 1957, p. 165; Lin 2005, p. 45; Hai 2007, p. 40). Lin (2003, p. 84) argues that the contact features in Written Dungan result from early Dungan scholars having been educated in Russian. Consequently, when they contributed to the early development of Written Dungan, they were influenced by their own linguistic background. As evident from (28a–c), Written Dungan replicates the Russian phrase-internal order in coding the semantic roles of time and goal, which makes it the opposite of Standard Mandarin in this respect. To illustrate with time, Dungan places the month before the year, which is identical with Russian. On the other hand, Standard Chinese follows the order year-by-month.

(28a)	打伊尤勒1943 年我到哩果拉德车尔尼果夫哩。
	Да июль 1943 нян вә дoли гoрoд Чернигoвли.
	tà=ijunj	1943=nján	v$\grave{ə}$	tō=le	gorat	tɕirnjigaf=le.
	abl=July	1943=year	1sg	arrive=pfv	city	Chernigov=sfp
	‘In July 1943 I arrived in the city of Chernigov.‘(Kalimov 1957, p. 165 discussed further in Hai 2007, p. 40, original representation in the Chinese characters maintained, IPA glossing added, and spelling adjusted)

(28b)	В июле 1943 гoда я прибыл в гoрoд Чернигoв.
	v	ijulj-e	1943	god-a	ja	pribɨ-l	v	gorat
	loc	July-inst	1943	year-gen.sgM	1sg	arrive.pfv-pst.sg.M	loc	city

	tɕirnjigaf
	Chernigov
	‘In July 1943 I arrived in the city of Chernigov.’ (Kalimov 1957, p. 165 discussed further in Hai 2007, p. 40, IPA and glossing added)

(28c)	1943年七月我就到了车尔尼果夫市了。
	1943=nián	qīyuè	wǒ	jiù	dào=le	chē’ěrníguǒfū	shì=le.
	1943=year	July	1sg	dis	reach=pfv	Chernigov	city=sfp
	‘In July 1943 I arrived in the city of Chernigov.’ (Kalimov 1957, p. 165 discussed further in Hai 2007, p. 40, Pinyin and glossing added, translation modified12)

Russian influence on Written Dungan is not discussed further in the limits of the present paper, which aims to focus on more neglected phenomena of the spoken language. What follows is not an exhaustive analysis and aims merely to give an overview of several noticeable changes in spoken Gansu Dungan, each of which must be investigated in detail in future research. This study uses the concepts of matter (MAT) and pattern (PAT) borrowing, which constitute the two basic ways of linguistic borrowing (Sakel 2007). To illustrate, borrowing a Russian conjunction corresponds to MAT, while restructuring relative clauses with native elements to correspond more closely to a Russian model corresponds to PAT.

Strikingly similar contact phenomena have been identified in all corners of the Russophone world, regardless of genealogical factors. For instance, Mithun (2021, pp. 521–22) reports the use of Russian modal verbs conjunctions and complementizers in Mednyj Aleut, an extinct mixed language of Russian and Aleut ancestry. Stolz and Levkovych’s (2022) extensive study of loan conjunctions in the Russophone world provides a solid starting point, but more comparative studies of Russian-influenced contact phenomena are needed to discover the full range of outcomes and limits to variation in language contact situations with Russian.

4.2. Nominal System

The nominal system of spoken Dungan contains both MAT and PAT borrowings from Russian. Three major contact-induced changes in the nominal system of spoken Dungan are the emergence of a polite second-person pronoun, mixed indefinite pronouns, and a caritative preposition.

Starting with the pronominal system, the second-person plural pronoun ниму /nì=mə/ 你们 is used also for a single person as a means of showing respect, like the French plural pronoun vous, a type of PAT borrowing (29) Among others, Hai (2002, p. 74) and Wang et al. (2015, p. 264) have identified this phenomenon and claim that it is likely caused by linguistic contact with Russian, where the second-person plural pronoun вы /vɨ/ also functions as the polite pronoun in the singular.13 In (29), two people are conversing on the phone. They address each other as вы /vɨ/, as evidenced by the first part of the example where code-switching into Russian occurs. When speaking in Dungan, the use of the second-person plural is retained in the conversation.

(29)	Вы не пoверите, 我给你们说一个прикoл吗?
	Вы не пoверите, вә ги ниму фә йигә прикoлма?
	vɨ	nje	pavjer-itje,	v$\grave{ə}$	kə=nì=mə	f$\acute{ə}$	jí=kə	prʲikol=ma?
	2.pl	neg	believe-2pl	1sg	dat= 2=pl	tell	one=clf	joke=q
	‘You won’t believe (it); shall I tell you a joke?’ (conversation)

Dungan has evolved mixed indefinite pronouns with both Russian and Dungan components. They are formed with a Sinitic interrogative pro-form carrying the Russian suffix -тo /-ta/ and attempt to replicate a Russian model, e.g., чтo /ʂto/ ‘what’ > чтo-тo /ʂto-ta/ ‘something’. For this reason, they show characteristics of both MAT and PAT borrowing. While genuinely mixed contact structures remain rare in Dungan, the case of indefinite pronouns nevertheless proves that they exist in the language. In (30), the Sinitic base form is са /sā/ 啥 ‘what’ while in (31), натар /nàtʰé/ 哪搭 ‘where’ serves as the base.

(30)	中原们的туәй，那是啥та。
	Җунянмуди туәй, нэсы са-та.
	pf$\acute{ə}$nj$\acute{ə}$=m=ti	tʰoj	n$\overline{ə}$=s$\overline{ə}$	sā-ta.
	Dungan=pl=mod	wedding	dem.med=cop	what-indef
	‘Dungan weddings are (quite) something.’ (explanation of Wedding practices)

(31)	哪搭та山上 长好少树。
	Натар-та санхoн җoн хoшo фу.
	nàtʰé-ta	sán=xōŋ	ʈʂòŋ	xóʂò	f$\overline{ə}$.
	where-indef	mountain=loc	grow	many	tree
	‘There are a lot of trees growing somewhere on the mountain.’ (constructed)

Finally, spoken Dungan uses a caritative preposition без /b^jez/ ‘without’ (32), a MAT borrowing from the Russian без /b^jez/ ‘without’. This is likely because no direct Sinitic equivalent exists for the pronoun. At the same time, other Russian prepositions are not attested in the source materials, possibly because Dungan already possesses native resources for encoding the corresponding semantic relations. It remains to be analyzed how widely spread this behavior is among Dungan speakers.

(32)	都安稳 без打仗 без骂仗。
	Ду нанвын, без даҗoн, без маҗoн.
	tú	nánw$\grave{ə}$n,	bjes	tàtʂōŋ,	bjes	mātʂōŋ.
	all	safe	without	war	without	fight
	‘Everything is safe and secure, without war, without fighting.’ (New Year plans)

4.3. Verbal System

In terms of individual items in the lexicon, Dungan manifests both MAT and PAT borrowing. As an example of the latter, ще /ɕjè/ 写 ‘to write’ has evolved into a novel meaning ‘to record’ because in Russian, the two verbs (на)писать /(na)p^jisat^j/ ‘to write’ and записать /zap^jisat^j/ ‘to record’ originate from the same root. The present section, however, focuses on the morphosyntactic aspects of borrowing in the verbal system of Dungan rather than on individual verbs. Dungan adopts Russian verbs with direct insertion. No light verb strategy is attested in the language, even though it is common in other contact scenarios with Russian, particularly in Turkic languages (see Forker and Grenoble 2021). Verbs of Russian origin show two patterns: infinitival and fully conjugated. The former host Dungan TAM enclitics (33) and the latter are incompatible with them, retaining Russian TAM and argument indexation (34).

(33)	你咋了？知不道吗谁скидывать的呢？几时скидывать的呢？
	Ни зали? Җыбудoма? Сый скидывать дини? Җисы скидывать дини?
	nì	tsà=le?	ʈʂ$\acute{ə}$	p$\acute{ə}$	tó=ma?	síj	skʲidɨvatj=tini	tɕìs$\acute{ə}$
	2sg	how=pfv	know	neg	reach=q	who	send.ipfv.inf=ipfv	when

	skʲidɨvatj=tini?
	send.ipfv.inf=ipfv
	‘What happened (with you)? Don’t you know? Who sent (the picture)? When was it sent?’

(34)	Хазыр我给你скину, 把那个。
	Хазыр вә ги ни скину, ба нэгә.
	xa	v$\grave{ə}$	k=nì	skʲinu	pá=nē=kə.
	now	1sg	dat=2sg	send.pfv.1sg	acc=dem.dist=clf
	‘I will send you that (i.e., the information) now.’ (conversation)

The example pair shows the behavior of the Russian verb pairs скидывать /skʲidɨvat^j/ (imperfective aspect) and скинуть /sk^jinut^j/ (perfective aspect), ‘to send.’ If a distinction between borrowing and code-switching is made in Dungan, in the absence of available phonological clues, morphosyntactic integration (see Matras and Adamou 2021, p. 240) draws the best demarcation line. In other words, the infinitival pattern that Dunganizes a Russian verb is closer to borrowing, whereas the inclusion of conjugated Russian verbs into the discourse is code-switching. This distinction is necessary, since if it is not made, it follows that Dungan has paradigmatically borrowed the complex system of Russian verbal morphology as well. At the same time, Russian verbal morphology never appears with etymologically Sinitic verb roots.

In the verbal system, a particularly salient phenomenon is the high frequency of modal verbs of Russian origin. This constitutes a phenomenon of alternation rather than complete replacement. Dungan modal verbs of Sinitic origin, such as ё /jō/ 要 ‘need, to have to’ and хуэй /xwī/ 会 ‘can, to be able,’ still occur in the language (35).

(35)	Ну, 把两个娃娃要拿上去呢дан。
	Ну, ба лёнгә вава ё нахoнчини дан.
	nu	pá=ljàŋ=kə	vávà	jō	ná=xōŋ	tɕʰi=n=dan.
	well	acc=two=clf	child	must	take=comp:up	go=stat=sfp
	‘Well, we need to take the two children there.’ (conversation)

Even so, the spoken language is replete with modal verbs of Russian origin. To illustrate, надo /nada/ ‘must’ (36) and its past counterpart надo былo /nada bɨla/ ‘should have’ surface in an unconjugated form, as in Russian. The same applies to нельзя /n^jil^jz^ja/ ‘must not’ (37).

(36)	我们где-тo表四个上началo. Надo四个上到去的。
	Вәму где-тo бё сыгәхoн начала. Надo сыгәхoн дoчиди.
	ò=m	g(dj) eta14	pjò	s$\overline{ə}$=k=xōŋ	natɕala.	nada	s$\overline{ə}$=k=xōŋ
	1=pl	about	o’clock	four=clf=loc	began.N.sg	must	four=clf=loc

	tō=ɕ=ti.
	arrive=comp:go=nom
	‘The beginning is at about four o’clock. We need to arrive at four.’ (conversation)

(37)	连他两个，нo把人家нельзя那个的дан。
	Лян та лёнгә, нo ба жынҗя нельзя нэгәдидан.
	lján=tʰà	ljàŋ=kə,	no	pá=r$\acute{ə}$ntɕà	njilzja	nē=kə=ti=dan.
	com=3sg	two=clf	but	acc=3	must.not	dem.dist=clf=mod=sfp
	‘With those two (people), you must not do that (i.e., let them down).’

4.4. Copular Omission

Dungan retains the inherited Sinitic copula сы /=s$\overline{\mathrm{ə}}$/ 是. Its formal nature and relationship with the copular subject require more investigation. For instance, Dragunow and Dragunowa (1936, p. 42) argue that the copula has already evolved into a suffix. In the present paper, it is interpreted as an enclitic phonologically dependent on its host and occasionally undergoing erosion into /=s/, vowel deletion being identified earlier as one of the defining characteristics of Dungan grammatical enclitics. Especially in more conscious and planned speech, the copula is frequently retained (38).

(38)	我是回族。
	Вәсы хуэйзў.
	v$\grave{ə}$=s$\overline{\mathit{ə}}$	xwítsú.
	1sg=cop	Dungan
	‘I am Dungan.’ (self-introduction)

Utterances with a zero copula, however, also appear in the language (39). Dragunow and Dragunowa (1936, pp. 42–47) are possibly the first researchers to mention limited copular erosion and even full omission in Dungan. Consequently, the phenomenon might originate from language-internal development, a view Hashimoto elaborates on as an alternative to a contact explanation (Hashimoto 1978, p. 257).

(39)	你мoжешь把个家рассказать。你谁？几岁？
	Ни мoжешь ба гәҗя рассказать. Ни сый? Җи суй?
	nì	moʑeʂ	pá=kétɕá	raskazatj.	nì	ø	síj.	tɕì=swī.
	2sg	can	acc=refl	tell.pfv.inf	2sg	Ø	who	how.may=old
	‘You can tell about yourself. Who are you? How old are you?’

Regardless of whether copular deletion arose because of language-internal development or through language contact, it has been reinforced through contact with Russian. Russian is known for its zero copula in the present tense (40). Copular omission has not yet become universal. Even in (39), when analyzing the recording for the first time, the main consultant reported the presence of the copula, nì s$\overline{ə}$ síj. On the second replay, she noticed that the copula was missing. In sum, more normative language maintains the copula, but it is sometimes omitted in the L register of spoken Dungan.

(40)	Ктo вы? Скoлькo вам лет?
	kto	ø	vɨ	skoljka	vam	ljet
	who	Ø	2pl	how.many	2pl.dat	years
	‘Who are you? How old are you?’ (constructed; cf. 39)

4.5. Clause Combining

Dungan clause combining constitutes one of the syntactic domains most heavily restructured by Russian language contact.

Table 5. Summary of Dungan conjunctions for clause combining.

Dungan	Russian	Function
и /i/	и /i/	coordinative: ‘and’
или /ilʲi/	или /ilʲi/	disjunctive: ‘or’
a /a/	a /a/	adversative: ‘and, but’
нo /no/	нo /no/	adversative: ‘but’

lists the most frequently encountered devices from the source materials. In this, Dungan follows the common trend of adopting Russian conjunctions and conjunctional adverbs (Forker and Grenoble 2021, p. 276), a type of MAT borrowing.

Stolz and Levkovych (2022) identified Russian loan conjunctions in 71% of the surveyed languages of the Russophone world (sample size 137), which underlines the ubiquitousness of the phenomenon. While the authors classified Dungan as a language in which Russian loan conjunctions could be verified, the present study shows that Dungan, too, follows the mainstream pattern of Russianized contact languages.

Unlike Standard Mandarin, which has a frequent zero strategy, Dungan marks clause coordination with high frequency. This divergence from the Sinitic pattern results from language contact with Russian. To begin with, (41) contains two clauses with separate predicates connected with the coordinative conjunction и /i/.

(41)	把我Дoм Дружбы上拉上，и пoтoм把我们拉上去。
	Ба вә Дoм Дружбыхoн лахoн, и пoтoм ба oму лахoнчи.
	pá=v$\grave{ə}$	dom	druʐb-ɨ=xōŋ	lá=xōŋ	i	patom
	acc=1sg	house	friendship-gen.sg=loc	take=comp:up	and	then

	pá=ò=m	lá=xōŋ	tɕʰī.
	acc=1=pl	take=comp:up	go
	‘Pick me up from the House of Friendship and then, take us (to the circus).’ (conversation)

Similarly, example (42) illustrates the disjunctive conjunction или /ilʲi/ ‘or’. A mother and a son are cleaning an office before going back home. The mother needs to unpluck an electronic device, and she uses the Kazakh expression ‘to take from electricity’. Somewhat bemused, the son reacts to this by asking whether she wishes to unpluck the device from electricity or from the power socket. Subsequently, the mother corrects herself and uses the more precise Russian word рoзетка /raz^jetka/ ‘socket’.

(42)	Хаз(ыр)等一会儿，把这个朝着тoк上要取掉呢。
	Хаз(ыр) дын йихур, ба җыгә җoчў тoкхoн ё чүдёни.
	xaz	t$\grave{ə}$ŋ	jíx$\overline{ə}$,	pá=ts$\overline{ə}$=kə	ʈʂʰōpfə=tok=xōŋ	jō	tɕʰỳ=tō=ni.
	now	wait	bit	acc=dem.prox=clf	abl=electricity=loc	must	take=comp=stat
	Speaker A: ‘Now wait for a minute, (we) need to unplug this (lit. take this from electricity).’

	Тoк上吗 или рoзетка上？
	Тoкхoнма или рoзеткахoн?
	tok=xōŋ=ma	ili	razjetka=xōŋ?
	electricity=dat=q	or	power.socket=loc
	Speaker B: ‘From electricity or from a power socket?’ (conversation)

Following a Russian model, spoken Dungan contains two adversative conjunctions, a /a/ and нo /no/. The language does not use the Sinicic conjunction dànshì 但是 ‘but’. Kalimov (1968, p. 485), however, lists it as кəсы ‘but’ in the repertoire of conjunctions illustrated with examples, corresponding to the Standard Chinese kěshì 可是 ‘but’. Consequently, it is possible that half a century ago, the conjunction was still in use. Now, the Russian-based conjunctions a /a/ ‘and, but’ and нo /no/ ‘but’ appear in the source materials with a very high frequency (43, 44).

(43)	蒸饭。А菜哪？
	Җынфан. А цэ на?
	ʈʂ$\acute{ə}$ŋ-fān.	a	tsʰē	nà.
	steam-rice	and	side.dish	where
	‘(There is) steamed rice. And/but where are the side dishes? ’ (conversation)

(44)	Билет чут贵下了。Нo我拿上了。
	Билет чут гуйхали. Нo вә нахoнли.
	biljet	tɕut	kɥī=xā=le.	no	v$\grave{ə}$	ná=xōŋ=le.
	ticket	bit	be.expensive=comp:up=pfv	but	1sg	take=comp:up=pfv
	‘The tickets have gotten a bit more expensive, but I bougtht them.’ (WhatsApp message)

Forker and Grenoble (2021, p. 278) argue that many languages that have borrowed the conjunctions /i/ and /a/ from Russian have also adopted their discourse function of marking the continuation of a topic with /i/ and conversely, topic switch or contrast with /a/. This applies to Dungan as well. Longer examples with turn taking are omitted in the limited confines of the present paper.

4.6. Complex Clauses

The complex clauses of Dungan have undergone structural remodeling that brings them closer to Russian.

Table 6. Borrowed devices for forming complex clauses.

Dungan	Borrowing Type	Function
tɕìs$\acute{ə}$ 几时	PAT < Ru. кoгда /kogda/ [kagda]	adverbial (temporal): ‘when’
jeslji	MAT < Ru. если /jeslji/	adverbial (conditional): ‘if’
xatja	MAT < Ru. хoтя /xotja/ [xatja]	adverbial (concessive): ‘if’
ʂto	MAT < Ru. чтo /ʂto/	complement: ‘that’
ʂtobɨ, ~ ʂtob	MAT < Ru. чтoбы /ʂtobɨ/	complement: ‘(so) that’
katorɨj	MAT < Ru. кoтoрый /kotorɨj/ [katorɨj]	relative: ‘who’
nàtʰé 哪搭	PAT < Ru. где /gdje/	relative: ‘where’

summarizes the principal devices encountered. The patterns include both MAT borrowing, e.g., the relativizer кoтoрый /katorɨj/ ‘that, which’, and PAT borrowing, e.g., Dungan натар /nàtʰé/ 哪搭 ‘where’ the original function of which is an interrogative pro-form. In Russian, где /gd^je/ ‘where’ can be used to form locational relative clauses, and this function has been borrowed for the Dungan /nàtʰé/.

In (45), the speaker, who has been trying to record new Dungan materials at home, expresses his frustration with not remembering to record something when a good opportunity arises. Similarly, in (46), the children of the speaker promise to come for a visit when they have a holiday. The temporal relation is marked with the interrogative pro-form җисы /tɕìs$\acute{\mathrm{ə}}$/ 几时 ‘when’ that replicates the use of Russian кoгда /kogda/ ‘when’.

(45)	А几时喧谎开了,忘掉的呢。
	А җисы щүанхуoнкэли, вoндёдини.
	a	tɕìs$\acute{\mathit{ə}}$	ɕɥánxòŋ=kʰéle	wōŋ=tō=tni.
	but	when	talk=	forget=comp=ipfv
	‘But when I start talking, I forget (to record).’ (conversation)

(46)	Ой, 几时我们缓开了，我们去呢说的。
	“Ой, җисы вәму хуанкэли, вәму чини,” фәди.
	“oj	tɕìs	ò=mə	xwàn=kʰél	ò=mə	tɕʰī=ni”	f$\acute{ə}$ =ti
	interj	when	1=pl	have.holiday=inc	1=pl	go=prosp	say=mod
	‘They said: “We come when we have a holiday”.’ (conversation)

Spoken Dungan now applies double marking in conditionals with high frequency. In addition to the native enclitic кели /=kʰéle/ 开了serving the function of an adverbializer in the protasis clause (and also encoding temporal and inceptive meanings, as in (45, 46), conditionality is frequently reinforced with the conjunction если /jesl^ji/ ‘if’ borrowed from Russian (47). The pattern resembles that of many Turkic languages of Central Asia that, in addition to a native suffix, have also borrowed the Persian-based اگر /agar/ ‘if’ as a device for forming conditional clauses.

(47)	А если我一天не пoлучится开了做啥呢？
	А если вә йитян не пoлучитсякели зў сани?
	a	jeslji	v$\grave{ə}$	jí=tʰjàn	nje	palutɕitsja=kʰéle	tsū	sā=ne.
	and	if	1sg	one=day	neg	be.available.inf=cond	do	what=prosp
	‘And what to do if I am not available on a certain day?’ (conversation)

The use of the independent conditional marker если /jesl^ji/ ‘if’ is not compulsory, although it surfaces with high frequency (48). Also, as the example demonstrates, the conditional enclitic may be further reduced into a monosyllabic form =kʰe in rapid speech.

(48)	哪搭想去开多贵。
	Натар щён чикэ дуә гуй.
	nàtʰé	ɕòŋ	tɕʰī=kʰé	tó	kwī.
	where	want	go=cond	very	be.expensive
	‘If you want to go somewhere, it is very expensive.’ (description of life in Almaty)

The Russian complementizer чтo /ʂto/ ‘that’ used among others with verba dicendi has been borrowed into Dungan (49).

(49)	说的呢，чтo她丫头没喧的дан说的。
	Фәдини, чтo та яту мә щүандидан, фәди.
	fə=tini	ʂto	tʰa	jatʰu	mə	ɕyan=ti=dan	fə=ti.
	say=ipfv	that	3	daughter	neg	want=mod=sfp	say=mod
	‘(My neighbor) says that her daughter said that does not (want to) talk (with that person). (WhatsApp message)

Purposive clauses in spoken Dungan follow a Russian pattern with the borrowed conjunction чтoбы /ʂtobɨ/ ‘in order to’ (50). The borrowing is not limited to the conjunction. Rather, the whole construction follows the Russian model, in which the verb that follows the conjunction must be in the past tense regardless of the pragmatic time reference. The Dungan copy of the structure tries to replicate the construction. The language, however, lacks grammaticalized tense. Consequently, past tense is replicated with the closest available counterpart, namely the perfective aspect marked by the enclitic ли /=le/.

(50)	我一会儿шы给папа说给呢чтoбы他把娃们拉上了。
	Вә йихуршы ги папа фәгини чтoбы та ба ваму лахoнли.
	v$\grave{ə}$	jíx$\grave{ə}$=ʂə	kə=papa	f$\acute{ə}$=k=ni	ʂtobɨ	tʰà	pá=vá=mə
	1sg	monent=foc	dat=father	say=ben=prosp	in.order.to	3sg	acc=children=pl

	lá=xōŋ=le.
	take=comp:up=pfv
	‘After a moment, I will tell (your) father to take the children (from the kindergarten).’ (conversation)

Moving to relative clauses, example (51) with натар /nàtʰé/ 哪搭 ‘where’ illustrates PAT borrowing from Russian. It can be contrasted with its Russian equivalent below, which shows an identical structure (52). In contrast, Mandarin Chinese disallows such post-head relative structures (53), like many other languages in East Asia.

(51)	Пoтoм 把我们拿上去рестoран 上哪搭光卖鱼都的呢。
	Пoтoм ба вәму нахoнчи рестoранхoн натар гуoн мэ йүр дудини.
	patom	pá= ò=mə	ná=xōŋ	tɕʰī	restoran=xōŋ	nàtʰé	kóŋ	mē
	then	acc=1=pl	take=comp:up	go	restaurant=loc	where	only	sell

	jy$\acute{ə}$	tú=tini.
	fish	all=cont
	‘Then, (the guide) took us to a restaurant where they only sell fish (dishes).’ (travel narrative)

(52)	Пoтoм нас oтвели в рестoран, где гoтoвят тoлькo рыбу.
	patom	nas	atvji-l-i	v	rjistaran	gdje	gatovj-at
	then	1pl.acc	take.pfv-pst-pl	loc	restaurant	where	prepare.ipfv-3pl

	toljka	rɨb-u
	only	fish-acc
	‘Then, we were taken to a restaurant where they only sell fish (dishes).’ (Russian translation of example 43 by Zhalilya (Mia) Shegetayeva)

(53)	然后他把我们带到了一个只卖鱼的餐厅。
	ránhòu	tā	bǎ	wǒ=men	dài=dào=le	yī=ge	zhǐ	mài
	then	3sg	DOM	1=pl	take=reach=pfv	one=clf	only	sell

	yú=de	cāntīng
	fish=mod	restaurant
	‘Then, he took us to a restaurant where they only sell fish (dishes).’ (Chinese translation of example 43 by Ular Nuerlan)

Some evidence exists to point out that the restructuring of relative clauses along a Russian model is a Central Asian phenomenon of spoken language that is more widely spread than previously documented. To illustrate, in addition to Turkic-type relative clauses, Spoken Kazakh now frequently builds relative clauses with the Russian relativizer кoтoрый /katorɨj/ (Akanov 2022), as in (54).

(54)	Үлкен апай-лар бар=ғoй кoтoр-ые қoрық-па-й-ды ештеңе-ден.
	ʏlkjen	apaj-lar	bar=ʁoj	katorɨje	qorɯq-pa-j-dɯ	jeɕteŋje-djen
	big	woman-pl	exv=cgm	rel-pl	fear-neg-prs-3	nothing-abl
	‘You know there are old women who do not fear anything.’ (Akanov 2022, p. 33)

4.7. Discourse Marking

Spoken Dungan now incorporates many discourse markers of Russian origin (

Table 7. Examples of borrowed discourse markers in Dungan.

Dungan	Borrowing Type	Function
да /da/	MAT < Ru. да	agreement
нет /njet/	MAT < Ru. нет	disagreement, negation
ну /nu/	MAT < Ru. ну	alerting of disjunction (Bolden 2016)
вoт /vot/	MAT < Ru. вoт	deictic: directing attention
җыма /ts$\overline{\mathrm{ə}}$ma/ 这马	PAT? < Ru. вoт	deictic: directing attention
всё /vsjo/	MAT < Ru. всё	leading to conclusion
кoрoче /karotɕe/	MAT < Ru. кoрoче	summarizing, changing direction
значит /znatɕit/	MAT < Ru. значит	summarizing, interpersonal management

). Hai (2000, p. 60) briefly mentions the existence of the phenomenon without any full examples, considering it a manifestation of the young with high bilingual capacities. Twenty years later, the discourse markers now appear in the speech of more elderly Dungans in the source materials of the present study. An intergenerational study comparing not only the frequency of Russian-based discourse marking but also Russian borrowing as a whole among different age cohorts of Dungan speakers is clearly needed but not attempted in the confines of the present paper.

Discourse marking is commonly placed either at the beginning or at the end of a phrase. To start with, ‘yes’ and ‘no’ have been borrowed from Russian ‘yes’ (55) and нет /n^jet/ ‘no’, respectively.

(55)	Да, 今儿个表四个上。
	Да, җергә бё сыгәхoн..
	da	tɕ$\acute{ə}$ːkə	pjò	s$\overline{ə}$=k=xōŋ.
	yes	today	o’clock	four=clf=loc
	‘Yes, it is today at four o’clock.’ (conversation)

The Russian attention-directing discourse marker вoт /vot/ has been copied into Dungan (56). A Dungan equivalent also exists җыма /ts$\overline{\mathrm{ə}}$ma/ 这马, literally ‘(Is it) this?’ It may result from PAT borrowing, whereby a functional equivalent to вoт /vot/ is constructed with native resources. Both discourse markers surface occasionally together, as in the example.

(56)	Вoт这嘛说的我们朝着садик上回去了，回去的呢。
	Вoт, җыма фәди: “Вәму чoчў садикхoн хуэйчили, хуэйчидини”.
	vot	ts$\overline{\mathit{ə}}$ma	f$\acute{ə}$=ti:	“ò=m	ʈʂʰōpf=sadik=xōŋ	xwí=ɕ=le,
	dis	dis	say=mod	1=pl	abl=kindergarten=loc	return=comp:go=pfv

	xwí=tɕʰī=tini”.
	return= compl:go=ipfv
	‘This is what they (i.e., the grandchildren) say: “We went back from the kindergarten; we are going back (self-correction)”.’ (conversation)

Similarly, the Russian-based discourse marker всё /vs^jo/ ‘well, alright’ permeates casual Dungan conversation. In (57), it tells the interlocutor that the topic has now been discussed sufficiently and the conversation is consequently reaching its conclusion.

(57)	Всё, 你把那两个娃娃拉上啥。
	Всё, ни ба нэ лёнгә вава лахoнса.
	vsjo	nì	pá=nē =ljàŋ=kə	vávà	lá=xōŋ=sa.
	alright	2sg	acc=dem.dist=two=clf	child	take=comp:up=imp
	‘Alright, take those two children.’ (conversation)

The discourse marker кoрoче /karotɕe/ ‘in short’ is very frequent in Dungan. It appears when the speaker wishes to shorten a longer argument, but it also carries other pragmatic overtones that have not been researched sufficiently despite the advanced stage of research on Russian discourse markers (58).

(58)	我气上来了，кoрoче.
	Вә чи шoнлэли, кoрoче.
	v$\grave{ə}$	tɕʰī	ɕōŋ=léj=le	karotɕe
	1sg	anger	go.up=comp:come=pfv	in.short
	‘In short, I got angry.’ (conversation)

The discourse marker значит /znatɕit/ has the literal meaning ‘it means’. It is used in pragmatic interpersonal management, summarizing the ongoing topic of discussion between the speaker and the addressee in an effort to establish a common understanding. In (59), the speaker, who is the addressee’s mother, lets him know that he will come to the office at one o’clock the following day.

(59)	朝着明个значит你表一个上来。
	Чoчў мергә значит ни бё йигәхoн лэ.
	ʈʂʰōpfə=mékə	znatɕit	nì	pjò	jí=k=xōŋ	léj.
	abl=tomorrow	means	2sg	o’clock	one=clf=loc	come
	‘From tomorrow, you will come at one o’clock.’ (conversation)

The phenomenon of discourse-marking borrowing is attested elsewhere in the Russian-speaking region, where ‘inegalitarian bilingualism’ exists. Inegalitarian bilingualism refers to a situation where the languages have asymmetric power so that one “exerts formidable pressure upon the other” (Hagège 2009, p. 79).

5. Implications and Conclusions

This chapter concludes this study with the following goals: First, it briefly compares Dungan with other ‘divergent’ Sinitic varieties of the adjacent regions and addresses the question of whether Dungan has evolved into a ‘mixed language’ (Section 5.1). Following, the chapter examines an ongoing trend, namely deepening connections between China and the Kazakhstani Dungans, from the viewpoint of Kazakhstani Gansu Dungan’s future development as a contact language (Section 5.2). The end of the chapter reiterates the key arguments of this study and underlines the importance of researching Dungan further with interactional source materials (Section 5.3).

5.1. Dungan as a Northwestern Sinitic Variety

While Dungan has undergone Russianization, the language manifests many similarities with other regional Sinitic varieties that have undergone various kinds of linguistic interaction in their history. As Salmi (2007) states, Dungan is better analyzed as the westernmost extension of a Sinitic dialect chain rather than a linguistic island separated from its origin. Focusing on Wutun and Zhoutun, two relatively known divergent varieties of Mandarin,

Table 8. Comparison of Sinitic contact varieties.15

Feature	Kaz. Gansu Dungan	Wutun	Zhoutun
Speaker population	ca. 110,000	ca. 4000	ca. 800–900
Contact languages	Russian, Turkic	Tibetan, Mongolic	Tibetan
Tonal contrasts	3	none	2
Polite 2SG pronoun	present	absent	absent
dem distinctions	3	2	2 (basic) + 2
Reflexive	个家kétɕá	个家gejhai	个家kuɤtɕia16
Inanimate plural	limited	present	present
Coding of sem. roles	enclitics, proclitics	suffixes	enclitics
Aspect marking	enclitics	suffixes	enclitics
Zero copula	possible	no	no
REL clause position	pre- and post-head	pre-head	pre-head
Conditional marking	single or double	single	single
Basic word order	SVO, (SOV)	rigid SOV	rigid SOV
Nominal CLFs	only generic	only generic	many present
Verbal CLFs	many present	many present	many present
Evidentiality	not present	present	only reportative

compares the three languages along with sixteen parameters that are relevant in analyzing Sinitic languages in the region.

In terms of its phonology, morphosyntax, and lexicon, the core of Dungan remains Sinitic. Unlike Wutun and Zhoutun, the language retains a more complex tonal system and SVO word order in several (but not all) contexts. At the same time, Dungan manifests many properties identical to those of other Northwestern Sinitic contact varieties, such as the collapse of the nominal numeral classifier system, also visible in Wutun.

It is important to consider whether Dungan qualifies for a ‘mixed language’ in the common understanding of the term, namely a language with mixed ancestry (Velupillai 2015, p. 69). At the current stage, the answer is no. Since the term ‘mixed language’ is used in several ways in the literature, this nevertheless also depends on how we define the concept. Summarizing the research consensus, O’Shannessy (2021, p. 325) argues that languages classifiable in a genetic tree of a single-parent language fail to qualify as mixed languages. Despite heavy Russian influence and numerous borrowings that have resulted in noticeable morphosyntactic restructuring, the core of Dungan remains clearly recognizable as Sinitic.

Challenges in differentiating borrowing and code-switching complicate the research on spoken Dungan. Unlike older layers of loanwords, new Russian loanwords do not undergo any phonological adjustment anymore, which results in difficulties in placing them exactly along what (Matras 2020) calls the codeswitching-borrowing continuum. Moreover, as earlier Dunganological research, such as Lin (2012, pp. 70–71), has pointed out, frequent code-switching has emerged as a discourse strategy. This is particularly observable in coding the predicates. Also, Dungan has witnessed the emergence of some genuinely mixed structures where bound Russian elements have been borrowed into Dungan (60).

(60)	我ха(зыр)找一个谁呢，连谁нибудь去呢。
	Вә ха(зыр) зo йигә сыйни. Лян сыйнибудь чини.
	v$\grave{ə}$	xa	tsò	jí=k	síj=ne,	lián=síj-nibudj	tɕʰī=ni.
	1sg	now	search	one=clf	who=sfp	com=who-indf	go=prosp
	‘Now I will search for someone. I will go with someone.’ (conversation)

Nevertheless, such mixed structures remain limited, reflecting Weinreich’s (2011, p. 50) remark on the relatively low transferability of bound morphemes in borrowing. The language contact with Russian, however, shows no signs of abating. If the current mode of contact with the resulting linguistic developments continues unabated, a likely scenario, it cannot be ruled out that Dungan eventually becomes a more ‘mixed language’ with a very high degree of relexicalization and morphosyntactic restructuring, with the caveat that not all theories concerning the emergence of mixed languages recognize a possible birth process merely as a result of heavy code-switching.

5.2. The Future Development of Dungan and the Role of Chinese

This study has demonstrated that, currently, heavy borrowing from Russian is reshaping both the Dungan lexicon and morphosyntax. It affects the phonological system as well by reinforcing earlier development. In short, the dominant source language for borrowing has shifted from earlier Turkic to Russian. This is supported by the Dungans’ linguistic self-evaluations in Russian and Kazakh. In 1999, 95.1% of the surveyed Dungans evaluated themselves as proficient in Russian, yet in Kazakh, the corresponding figure was only 35.9% (Suleimenova et al. 2007, cited in Smagulova 2016, p. 67). The figures are somewhat dated and do not fully reflect ‘Derussification’ and ‘Kazakhization’, two sides of the same coin, now operating in independent Kazakhstan. Notwithstanding, they clearly highlight a situation that continues to this day among the speakers of Dungan in Kazakhstan: Russian, rather than Kazakh, has become the socially dominant language for many. It is also the language most of them feel the most proficient in.

When addressing the future of Dungan as a contact language, the role of China and Chinese cannot be ignored. As Ding (2007, p. 46) demonstrates, the collapse of the Soviet Union and the resulting cessation of support that the Dungans had earlier received, e.g., for the development of the literary language and mother tongue education, created major challenges that the Dungan language is currently facing. At the same time, it can be argued that this has created an opportunity for re-establishing a more robust connection with China and the Chinese language.

Most Dungans are unable to read Chinese characters (Lin and Ding 2002, p. 70, confirmed by personal fieldwork), save in cases where an individual has been living in China, usually because of studies or work. China has been projecting its influence on Central Asia and beyond, a part of which goes under the Belt and Road Initiative (一带一路). The long-term results of relative isolation of the Dungan from other Chinese-speaking groups may consequently move towards integration into the Sinosphere, which may also include efforts of an orthographic reform by introducing Chinese characters (Janhunen 2020, p. 358; see also Zavyalova 2017a) and increased teaching of Standard Chinese with or without the Chinese characters (Allès 2005, pp. 133–34). At present, however, little concrete action has materialized, save attempts to build Dungan neologisms from a Chinese model. Also, China offers sponsored tours for Dungans to visit the country and scholarships to pursue higher education at a Chinese university (Smagulova 2016, p. 79). Taken as a whole, these activities manifest what Kokaisl and Hejzlarová (2023) perceive as the contemporary Chinese ideology emphasizing the cultural similarities of the Chinese and the Dungan.

If the cultural and linguistic rapprochement between the Dungans and the Chinese intensifies in the future, it will certainly affect the current patterns of linguistic contact as well. It is too early to speculate whether an eventual ‘de-Russianization’ and shift back towards a more Sinitic model is possible for Dungan. Although we enter a more speculative ground, a possible outcome of this could, among other things, be a trend reversal in‘monomorphemization’ discussed in Section 2.3 and an increased etymological awareness of the Sinitic lexicon. In sum, Central Asian Dungan as we know it emerged because of the power politics of the time. Once again, such power politics, fully external to language, have the chance of influencing the direction of the cultural and linguistic development of the Dungans.

5.3. Conclusion: Summary of Contact-Induced Russian Features in Dungan and the Role of the Language in Sinitic Typology and Contact Linguistics

The present study demonstrated that phonological and morphosyntactic contact phenomena are plentiful in spoken Kazakhstani Gansu Dungan and manifest as both MAT and PAT borrowing. In phonology, contact with Russian reinforces the short forms, a development that predates intensive contact with Russian. Also, consonantal palatalization likely results from transferring this phonetic feature from Russian, which has become the dominant language for many younger speakers of Dungan. In terms of the nominal and verbal systems, major contact-induced developments include the use of the plural form of the second person pronoun ниму /nì=mə/ 你们to show respect, the emergence of a caritative preposition без /b^jez/ ‘without’, borrowing of modal verbs of Russian origin, and frequent copular omission, a development not likely caused by Russian contact but reinforced by it. Regarding clause combining and complex clauses, Dungan has borrowed a wide range of conjunctions and devices for forming complex clauses from Russian. Additionally, the patterns of forming complex clauses often approximate those found in Russian. Similarly, many devices of discourse marking identifiable in spoken Kazakhstani Gansu Dungan have their origins in Russian.

Rimsky-Korsakoff Dyer, a pioneering Dungan researcher whose contributions helped to establish the field in the West, stated as follows regarding the nature of Russian influence on Dungan:

Dungans do not use Russian verbs, adverbs, pronouns, adjectives, conjunctions, prepositions, particles, etc. The only Russian words that are adopted are nouns or adjectives...

(Dyer 1965, p. 55, cited in Rimsky-Korsakoff Dyer 1977, p. 354)

The statement was possibly correct at the time. More than half a century later, the present study nevertheless demonstrated that the list must be reconsidered. Earlier research on the Sinitic languages emphasized unity and similarity across the language group. On the other hand, the ongoing documentation of Sinitic varieties and their typological comparison that goes hand in hand with the former has made the research community increasingly aware of wide internal variation. The Sintic languages certainly share many core features together, to the extent that we can talk of a Sinitic prototype. At the same time, there is no universal Sinitic grammar (see Szeto and Yurayong 2021) shared in full by all Sinitic languages. By being a highly ‘divergent variety of Sinitic,’ Dungan plays a key role in mapping the possibilities and extent of Sinitic variation. In conclusion, the present paper aims to offer a springboard to discover the language further and to question some of the fixed ideas about what a Sinitic language should look like. At the same time, Dungan offers new perspectives for the study of contact languages in the Russophone world.