Atlas of multilingualism in Daghestan: A case study in diachronic sociolinguistics

This paper introduces the future Atlas of Multilingualism in Daghestan, a project based on extensive field study of the language repertoires of the residents of rural highland Daghestan. The Atlas will provide quantitative data on multilingualism across a relatively compact linguistic area, which is, culturally and socially, both homogeneous and diverse. It will represent a wide range of ethnic contact situations in a qualitatively and quantitatively comparable way. The data are collected by the method of retrospective family interviews, which is designed to obtain data about bilingualism in the past. The paper gives a brief sociolinguistic overview of Daghestan, describes the method and its restrictions, explains the design of the future Atlas, and provides two sample chapters. One of the chapters describes three villages in northeast Daghestan, and the other describes two villages in southern Daghestan.


Introduction
This paper introduces the future Atlas of Multilingualism in Daghestan, a project based on an extensive field study of the language repertoires of the residents of rural highland Daghestan 1 .
The Atlas will make a contribution to studies of the phenomenon which is referred to as small-scale multilingualism (Lüpke 2016) (it has also been termed reciprocal by Jourdan 2007, balanced by Aikhenvald 2007, traditional by Brandl & Walsch 1982, Di Carlo 2016, and egalitarian by François 2012. Small-scale multilingualism is typical for small socio-political groups, which have no overarching hierarchical political structure joining them together (Singer & Harris 2016). This type of societal multilingualism is characterized by the absence of power or prestige relations between languages.
Investigating situations of small-scale multilingualism is important for reconstructing social conditions that favored linguistic diversity in the precolonial world (e.g. Evans 2010: 10, Evans 2017, Lüpke 2016. Another emergent reason is the 1 The article was prepared within the framework of the HSE University Basic Research Program and funded by the Russian Academic Excellence Project '5-100'. We express our deep gratitude to the hundreds of our consultants who hosted us, fed us, and shared their life stories. We also thank the reviewers who helped us to improve the manuscript of this paper.
Daghestanian multilingualism is discussed in some detail in several surveys (Wixman 1980, Chirikba 2008, Magomedxanov 2008, but has never been studied systematically, i.e. by comparing various locations, according to a similar set of parameters, and on the basis of quantitative data. The project described in this paper aims at collecting quantitative data about the multilingualism of Daghestanians in a representative set of locations across the region and presenting it in the form of maps and descriptions. The project was launched in 2009 and is run by a large team of researchers collecting and processing data 2 . The aim of this paper is to explain the methodology of the project and to present the structure and design of the Atlas. We will give a very brief overview of Daghestan as an area of high linguistic and ethnic density (Section 2), describe the method used to collect data on multilingualism (Section 3), explain the design of the Atlas (Section 4), and provide two sample chapters, one devoted to two villages in southern Daghestan (Section 5), the other to three villages in northeast Daghestan (Section 6). Section 7, the conclusion, provides an outlook on what kind of theoretical problems can be addressed by using the data of the Atlas.

Daghestan and its languages
This section provides a brief overview of the linguistic situation and language density in the region.
Daghestan is the area of the highest language density in the whole Caucasus. Most Daghestanian languages belong to the East Caucasian (Nakh-Daghestanian) family. There are also speakers of three Turkic (Kumyk, Nogai, Azerbaijani) and two Indo-European languages (Tat and Russian) (for overviews of languages see van den Berg 2005, Koryakov 2002, Tuite 1999, Hewitt 1981, Geiger et al. 1959, Wixman 1980. East Caucasian languages form a deep-level linguistic family, comparable to, or deeper than, Indo-European. Even languages located closely on the family tree, within one branch, such as Lezgian and Aghul (both Lezgic), are linguistically distant and mutually unintelligible. At least two major languages, Avar and Dargwa, show a high degree of dialect divergence, and in the case of Dargwa some dialects are increasingly considered separate languages.
In terms of its landscape, Daghestan is mostly a mountainous terrain. Until relatively recent times, the lowlands near the Caspian Sea were not densely populated. The mountains provided protection from invasions and had a climate more adapted for traditional husbandry based on shepherding and crops, while the lowlands were hot and damp, with malaria as a constant threat. As of today, highlanders live in villages with a population ranging from one hundred to several thousand residents. Most villages reportedly date back to the Middle Ages or deeper in time, though exact dates for specific villages are usually not known.
With so many languages packed into a relatively small area, two neighbouring villages in a walking distance one from another often speak different L1s. It is on such language neighbourhoods that the project is focused. It is assumed that mountain ridges, rivers and other natural barriers contribute to maintaining language density. But in addition to the landscape, language boundaries were also, if not more, maintained by traditional village-level endogamy, typical of most parts of Daghestan. In Daghestanian villages, marriage partners were taken almost exclusively from the same village and often from the same patrilineal clan (Comrie 2008, Karpov 2010. Linguistically mixed marriages were and still are very uncommon (Karafet et al. 2016). Endogamy in highland Daghestan is presumably not a recent innovation. There is historical evidence that the tradition of endogamic marriages goes far back in time (more than a thousand years ago, according to Lavrov 1978, see overview in Bulaeva et al. 2006).
In our observations, villages with 300-400 households can have as few as one or two mixed families. In the infrequent cases when a man married a woman from a different language community, after moving to her husband's village the woman was expected to learn the local L1 and to talk to her children in the language of their father. Today, villagers often say that marriage rules are not as strict as they used to be, but mixed couples prefer to move to the towns or lowland villages where strict highland endogamy is not observed. At present, mountain villages continue to be almost strictly mono-ethnic. Exceptions are villages with important markets (cf. e.g. Kumukh and Tsudakhar in Materialy 1927), and bigger towns like Kizljar or Khunzakh. This fosters the vitality of the languages in the villages, where children still usually communicate in the village's native language.
Unlike some other areas of high language density (cf. Franҫois 2012 on Vanuatu and Campbell & Grondona 2010 on Misión La Paz), interethnic communication was not on a daily basis, within speech communities, but was outbound and required special circumstances, occurring weekly rather than daily; though, of course, that depended on individual occupations (Nichols 2013). In the Soviet period, some of the collective farms were meeting points of workers who spoke different languages, and probably became arenas for interethnic communication on a daily basis. For some locations, we have reports that this communication was in one of the local languages. In the 1980s and 1990s, as new roads were constructed, and individual transport was becoming a norm, connections between neighbouring villages loosened. Today, people prefer to go for shopping to towns lower in the plains. They are less attracted to local markets and locally available goods. Intervillage communication has become less important.
In any case, inter-ethnic communication required a shared language. The lowlands were dominated by speakers of Turkic languages, which led to widespread use of their languages in communication between different ethnic groups, Kumyk in the north and Azerbaijani in the south (both often referred to as Tatar in early reports) (Wixman 1980: 108-19). Avar has been another lingua franca, used in at least some parts of the northern highlands (Dobrushina, Zakirova 2019). Arabic, used in local administration in the pre-Soviet period, was known by few highlanders (and read phonetically by more). On the other hand, our research shows that, in some areas, the more typical multilingualism pattern is what we will call vicinal, which we propose as a technical term for multilingualism whereby people from neighboring villages communicate among themselves in the language of one of the villages rather than in a third language such as a regional lingua franca.
People from neighbouring villages communicated between themselves in the language of one of the villages rather than in a third language. Multilingualism was widespread. Villagers spoke the language of their village, the language of their neighbouring village(s), and additionally a more widespread language of the area (when different from these two) . Bilingualism in distant major languages, within Daghestan and across its borders (Chechen, Azerbaijani, Georgian, Russian), was more, and sometimes much more, restricted.
After the establishment of the Soviet administration in the 1920s, and especially starting from the mid 1930s, Russification started a major sociolinguistic shift (Dobrushina 2016). Russian was the first language to be established as one single lingua franca and the language of written administration common for the whole area (Chirikba 2008: 30;Šaxbanova 2011, Magomedov 2010, Daniel et al. 2010. As of today, vitality of local languages has not yet been impaired in the villages, but vicinal multilingualism is in obvious decline. Most people under 20, or, in some villages, under 40, use Russian when communicating with their neighbours from other villages. In order to capture the traditional patterns of vicinal multilingualism in Daghestan -data of paramount importance to the social and linguistic history of the area -we need to act quickly.

Method
Research on traditional bilingualism in Daghestan requires data on local multilingual patterns from the period before the spread of Russian. That period, which we call late traditional bilingualism, provides a window into the sociolinguistic past of the region. We do have some documentary sources, such as works and notes by Uslar, Dirr and other Russian researchers and civil servants who worked in the Caucasus in the 19th century (see e.g. Kavkazskie gorcy 1992). There are also several reports on bilingualism by Soviet anthropologists (Genko 2003, Lavrov 1953, 1978, Volkova 1967). All such documents, however, provide qualitative and fragmentary information. The information is usually given in the form of general observations and assessments (cf. "During the historically accessible period, the Tsakhurs are in a state of a permanent Tsakhur-Azerbajani bilingualism", Magomedxanov 2008:47) without providing specific quantitative details, such as counts of people who had a command of specific languages as L2. Such details may be necessary for studies not only of the social background of language contact but also of its structural corollaries; cf. for example (Seifart 2015) who treats the problem of direct vs. indirect affix borrowing in relation to the rate of bilingualism.
Information from such sources is also often biased. Most reports from the 19th century were compiled by the Russian administration whose primary concern was the local command of Russian and Arabic. They were much more interested in their own communication with the highlanders than in the communication between local ethnic groups. In addition, the data usually concerned major ethnic groups at politically important locations. The situation of minority languages, less known but not less crucial for our understanding of social and linguistic aspects of language contact, remains hardly visible in these reports.
The data of the Atlas are meant to be both quantitative and diachronic. They are focused on specific locations, often remote from administrative centers, and target areas of contact not only between major languages but also between a major language and a minority language or between several minority languages (or, sometimes, dialects). They are collected by the method of retrospective family interviews, which was specifically designed to obtain quantitative data about multilingualism in the past (Dobrushina 2013). The respondents are interviewed about language inventories, both their own and those of their older -often deceased -relatives. Only those relatives whom they claim they remember clearly are added to the database.
Questions about grandparents, including what were their language repertoires, are considered natural in a Daghestanian village. Daghestanians have large extended families where children live together with their paternal grandparents or often stay at their place. It was usual for the youngest son's family to stay with his parental home. Relatives who did not live in the same house nevertheless had intensive contacts with their parents, both on father's and mother's side. Cousins and second cousins formed strong social networks of support and communication. Traditional highland villages were poor in space. Houses were located on a very small territory, most often one over the other terrace-like on a hillside. One's children were often looked after by grandparents and cousins. Members of one extended family spent a great deal of time together.
Interviews about relatives allow us to reach back into the 19th century, starting with people born around 1850, with more dense data from the 1880s on. This time span covers the situation typical of the village before the drastic social changes of the 20th century.
The interviews were all held in Russian. In very few cases, when the respondent had poor command of Russian, we communicated via an interpreter, most often a younger relative, who translated our questions and answers between Russian and the local language. The answers were put together in a table (spreadsheet) and aggregated (see Sections 5 and 6 for examples). Table 1 shows an example of a completed questionnaire.

Interviewed in Chabanmakhi
The interviewer was talking to Umaidat Family relation to the respondent father of Umaidat The method is vulnerable to both individual mistakes and systematic biases. First of all, the multilingual situation is often stereotyped and generalized and extended to the self's relatives. ("Our elders all spoke this language, so my parents did, too.") 3 In some areas, this may be supported by an ethnic self-ascription which is a relatively recent development. For example, the speakers of Archi (Lezgic) and many speakers of various Andic or Tsezic languages often identify themselves as Avars (at least as their second identity). This may, at least partly, result from the demographic policy of Daghestanian officials in the second half of the 20th century. Archi people were assigned Avar ethnicity in their passports, and have been taught Avar in L1 classes (Russian уроки родного языка, literally lessons of mother tongue) at schools. In censuses and scientific reports from the late 19th and early 20th century, however, they were identified as ethnic minorities rather than Avars. An emergence of the secondary Avar identity may have led highlanders to generalize the current multilingual patterns to the past.
The difference between data obtained directly and indirectly is statistically assessed in (Daniel et al., under revision); our conclusion is that there is no visible difference in terms of command of local second languages, while knowledge of Russian is underestimated in indirect data as compared to direct, but only very slightly.
Another issue is the evaluation of the command of a language. There are various and sometimes sophisticated approaches to evaluating language proficiency developed in the studies of language shifts (Vakhtin 2001). Such studies are very time consuming, requiring the researcher's command of the language for which proficiency is evaluated. For a large-scale cross-Daghestanian study this was not a feasible approach. In any case, language proficiency cannot be directly evaluated for the dead. As a result, we relied almost entirely on the respondent's self-assessment and his or her assessment of the recollected multilingualism of the older relatives. The trustworthiness of individual evaluations is questionable, as some people overestimate or underestimate their own command of a language, as well as that of the others. Claiming that you or your grandmother spoke Kumyk may refer to fluent use or to the ability to produce several phrases. Sometimes the respondents could be more specific, but in the majority of cases the answers were simply yes or no.
Finally, for the eldest relatives (e.g. born in 1880), only multilingualism in later age could be reported. Although people acquire their basic language repertoire in their youth, their linguistic inventory may gradually change over the course of their lives, especially if the community undergoes considerable social changes (Chambers 2002: 358). As in any kind of apparent time study -which this project in a certain sense is -there is an effect of life-span change distorting the true dynamics of trends of change (on similar issues in studying structural language change, see Sankoff &Blondeau 2007 andBowie &Yaeger-Dror 2014).
All this results in distortions of the targeted sociolinguistic reality in our analysis. However, the only bias that we expect to be truly systematic in our data is bias towards the present community's stereotypical image of past multilingualism. Such stereotypes may be established in the course of the interviews and then taken into account, at least in the qualitative interpretation of the quantitative data. All other vulnerabilities, although decreasing the reliability of individual records, probably go both ways and should not result in systematic mistakes. In some cases, we have evidence on language proficiency of an older person obtained from different living relatives (though no systematic effort at such cross-validation has yet been done). We thus believe that the collected data, biased as they are, provide empirical evidence beyond social stereotypes. Here, an important even if not rigorously academic point to make is, this is all we can get.
Prior to dozens of short interviews based on the questionnaire, the project leader (who is also the first author of this paper) conducted several extensive interviews with knowledgeable members of the community to 'set the sociolinguistic scene' of the location. The targeted information included the typical language repertoire of the past (with an eye to possible stereotypes); dynamics of village population (supported by Svod 1893 and Materialy 1927 and later censuses); ethnic identity; typical marriage patterns (especially local endogamy); the geography of communication and important local connections; locally important markets; the economic situation of the village, also as compared to the neighbours, including traditional and current occupations, objects of manufacture and trade, seasonal jobs, etc. These data serve as the basis for the questionnaire used in the interviews. In the chapters of the Atlas, such qualitative data are summarized in the sections that describe specific villages (cf. Section 5 and 6 here).

Data
The data were collected by Nina Dobrushina and Michael Daniel in 2009, Dmitry Ganenkov in 2013, and by Nina Dobrushina leading a field team of students in 2013 -2019 (see footnote 1). As of this writing, 62 villages have been surveyed.
The villages form eighteen geographic clusters of two to four villages with two to three native languages per cluster (see Figure 1).

# Villages
Year ( The cluster did not necessarily include villages whose socioeconomic relations were especially tight. A cluster is a unit of analysis more than a real social unit. The idea behind studying clusters rather than individual villages is that bilingualism is (at least) a binary relation between two (or more) ethnic groups. Counts of bilingualism within one of the groups only cannot provide for a robust sociolinguistic interpretation of the patterns of interethnic communication unless complemented by similar counts from the other group(s). All neighbouring villages in Daghestan are closely related in terms of socioeconomic interaction. Each village in our clusters is adjacent to at least one of the others in the cluster. The villages are usually within 20 to 90 minutes of walking distance. Local communication in a multilingual setting is essentially a pattern of shared practices of language choice. The surveys were initially conducted by two people only and are now performed by a group of six to eight people. Obviously, the amount of the data collected in different clusters varies significantly. In our database, the smallest amount is 30 entries, obtained in Lezgic village Arkhit in southern Daghestan, and the largest is 323, obtained in Chuni, an Avar enclave in a Dargwa-speaking area. The amount of collected data also correlates with the size of the village. For example, the Lak villages of Uri and Mukar are almost deserted. Even working in a large team, we were only able to collect some 80 entries in total.
We will come back to the analytic issues that may be addressed based on these data in the conclusion.

The Atlas of multilingualism of Daghestan: aims and design
The Atlas of multilingualism of Daghestan is intended to appear as a paper edition in addition to the already existing electronic open-access resource 4 . Linguistic atlases are usually dedicated to mapping first languages and their features (Moseley & Asher 1994;Wurm et al. 1987;Koryakov 2006). One of the few atlases dedicated to non-first languages across a large area is the Atlas of Languages of Intercultural Communication in the Pacific, Asia, and the Americas (Wurm et al. 1996). It maps contact languages and lingua francas of the area across different historical periods (see also Sichra ed. 2009). So far, quantitative data have only been collected for individual cases (locations), most often in urban communities. The Atlas of multilingualism of Daghestan is unique in that it will contain quantitative data on multilingualism across a contiguous linguistic area which is, culturally and socially, both homogeneous and diverse. It will represent a wide range of ethnic contact situations, which will allow a cross-case quantitative comparison.
Each chapter of the future book represents one geographic cluster containing from two to four neighbouring villages. It contains a text which describes the cluster in economic and social aspects, and reports the results of our sociolinguistic survey. Chapters have an introductory section about the cluster as a whole, followed by several more sections, one for each village in the cluster. These are further divided into subsections providing general background data on the recent past of village (as reconstructed from the interviews with the villagers) and its present. In the last section of each chapter, the multilingual patterns of the cluster are summarized. Apart from the description, each chapter contains several maps.
Separate chapters are planned that discuss general topics such as the level of Arabic literacy in different parts of Daghestan as it is observed in our field research, the distribution of major languages as lingua franca (Avar, Kumyk, Azerbaijani), and gender patterns of multilingualism.
Quantitative data on multilingualism is available at multidagestan.com, and the database is updated after each field season. The webpage provides flexible tools for filtering the data by multiple parameters and visualization of descriptive statistics. In 2019, the data from the earliest demographic surveys of Daghestan (Svod 1893, Materialy 1927) were added to the website of the project (see Census data bookmark).
Below we provide two sample chapters. We chose two clusters which are located in different parts of Daghestan: the south (Kina and Gelmets) and the northwest (Rikvani, Zilo and Kizhani). These two clusters show very different patterns of language contact between neighbours. The former is an area where a lingua franca is in use, with Azerbaijani spoken in interethnic communication. The latter is a pure case of asymmetrical vicinal bilingualism. Kina also shows (or, rather, used to show in the late traditional period) seasonal nomadism, of a type unusual for Daghestan -the whole village population moved to Azerbaijan for the winter period.

The cluster
This cluster was examined in 2016-2018. It includes two villages, Kina and Gelmets. Kina speaks Rutul and Gelmets speaks Tsakhur. Both languages belong to the Lezgic branch of East Caucasian. They are close relatives, but there is no mutual intelligibility between them. Both languages show some dialect variation between villages.
Rutul and Tsakhur had no literacy before the end of the 20th century. Azerbaijani was used for writing documents and religious texts. Residents of both villages could not recall any texts written in Tsakhur or Rutul before the 1990s. After Perestroika, Rutul and Tsakhur officially adopted writing systems, in 1989 and 1990 respectively. Since that time, teaching Rutul and Tsakhur is included in the school curriculum, and several textbooks have been published. There is a newspaper Rutul'skie novosti (Rutul news, http://rutnov.ru/), with most texts in Russian and some in Rutul; and a newspaper Nur, similarly combining Russian and Tsakhur.
Both villages are very close to the border with Azerbaijan. It took between 6 and 7 hours by foot to reach a big Azerbaijani village, Qax. Contacts with Azerbaijani villages there were an important part of the villagers' life, because Azerbaijan was richer than Daghestan and had more fertile soil. Many Daghestanians used to go there for work. Contacts with Azerbaijan have been in decline from the second half of the 20th century. The complete separation came after the collapse of the Soviet Union when the border with Azerbaijan became a national border, and especially from the 2000s, when the border started to be closely inspected. People cannot follow the usual paths across the border anymore, and some extended families are separated.
In the 19th century, the villages were a part of the Samur district, together with other Rutul, Tsakhur and some Lezgian villages. In 1928, the territory of the Samur district was divided into two districts with administrative centers in the villages of Rutul and Akhty. Kina and Gelmets belong to the district with the administration in Rutul. Informally, the district is also considered to be divided into two areas, Lower and Upper Magal (a traditional division in some Islamic countries). The border between the two falls exactly between Kina and Gelmets, probably based on the language boundary.
Kina and Gelmets are located high in the mountains. The highest -and the oldest -parts of both villages lie at about 1,800 meters above the sea level. Newer houses are at 1,500-1,600 meters. Generally, Tsakhurs live further up the valleys and thus higher than Rutuls.
Kina is the last settlement in the chain of Rutul villages going up the Samur River. Gelmets is the first Tsakhur village further upriver (see Figure 2). The walking distance between Kina and Gelmets is considerable. It takes one to one and a half hours to walk from one village to the other.
According to the villagers' own accounts, both communities were very poor. The gorge is narrow and rocky, and arable land was scarce. Available pastures were not enough to support many cattle. At present, people have some fruit trees in their gardens, but they only started to plant them well into the Soviet period. Principal crops were rye and barley; potatoes were planted as well. Both villages did not have enough corn and had to buy it from their neighbours, though Gelmets was slightly better-off than Kina. The village could not house many more people because of the subsistence conditions. There was an acute shortage of arable land and, even more so, of land for cattle herding. Kina people did not have a lot of cows or sheep because there was no place to herd them. They had to buy corn, meat and wool from Azerbaijan and the neighbouring villages.
Difficult conditions and the proximity of Azerbaijan resulted in active seasonal migrations. Whole families including small children would leave Kina to go to Azerbaijan in late September (before the first snow) to come back in March, after the snow was gone. Very few people stayed in the village in the winter. The residents of Kina describe their ancestors as nomads. This was the second reason why Kina people did not have many cattle, in addition to land shortage: they could not take their cattle with them to Azerbaijan. A man born in 1950 recalls that he would spend winters in Azerbaijan with his father, brothers and sisters, while his mother stayed in Kina with the cattle. Only recently have Kina villagers started to have more cattle. There is a common expression that the locals use to describe the situation of their village in the past: "Zdes' nechego bylo lovit'" (lit. "Here there was nothing to catch", colloquial for "here, one had no chances whatsoever").
In Azerbaijan, Kina villagers worked for farmers as hired hands, mainly shepherds. They also had the reputation of being good builders and carpenters. As in most of southern Daghestan, Kina women wove carpets.
The road passing in the mountains over Kina was used by those who went to Azerbaijan from Rutul and Lezgian villages located downstream. The road went past Kina and Gelmets and further to Azerbaijan.
The main advantage of Kina was a forest near the village. Kina residents sold logs and charcoal they made. These goods were especially popular in the Rutul village of Shinaz, where people forged daggers and sabres.
Kina had one mosque, and no madrasah. The secular school was opened in the 1930s, and the education was first in Azerbaijani. Later, in the 1950s, Russian became the language of instruction.
Kina villagers do not recollect strict prohibitions on marrying out or taking women from outside, but most often wives were taken from within Kina. There are about ten women from the neighbouring Tsakhur villages Gelmets and Kurdul, and one elderly Azerbaijani woman. Comparing to other highland villages (including Gelmets), this is quite a few incomers. Most wives from Gelmets are rather young (below 40 years old). It seems that the practice of taking wives frоm Gelmets is new, since only three women who were born outside Kina are older than 50. At present, there are several women from other Rutul villages (Luchek and Shinaz), and about ten women from Gelmets and Kurdul, the closest Tsakhur neighbours of Kina.

Present
According to the latest censuses, Kina had 976 residents in 2002 (Tablica № 2с 2004 5 ) and 653 in 2010 (Mikrodannye 2010), being one of the biggest villages in the district. The population increased very sharply in the 1980s, and many new houses were built where no houses stood before. Later, the population started to decrease. Many people moved from Kina to their kutan 6 in Babayurtovsky district (a lowland district in the north of Dagestan, with an ethnically mixed population).
Before the 1980s, life was hard in Kina. Villagers strongly depended on economic relations with their neighbours. A woman born in 1958 recalls that she used to go to Gelmets to exchange carrots and beans for sheep wool when she was young.
Electricity was brought in the 1970s, and television reached the village in the end of the 1980s.
Kina people speak a dialect of Rutul. The villagers know that the variety differs from idioms of other Rutul villages, but in their opinion, there are Rutul villages where dialectal differences from the accepted standard (the Rutul of Rutul) are stronger (Khnov and Ikhrek).
Kina has a full school (11 grades, from 6-7 to 17 years old). As of 2016, there were 110 schoolchildren. School education is in Russian. There are two classes in Rutul, weekly; but, in the opinion of the teachers, Rutul lessons are not popular among the pupils. The main purpose of the course is to teach children to read in Rutul, but the written language is almost useless in everyday life. There is only one Rutul newspaper, published once a week. Ideally, the newspaper is meant to be half Rutul, half Russian, but in practice issues have only one or two articles in Rutul. Another problem for Kina villagers is that the written standard is based on the dialect of the village of Rutul, deemed by them to be different from the dialect of Kina. As a result, most residents of Kina can only read in Russian.
In the 1970s to 1980s, the village had a small carpet factory. At present, there are no factories in Kina or anywhere in the vicinity. Some people earn money by selling cow and (much more expensive) sheep cheese. The production of sheep cheese was still alive in Kina in 2019, while in most other places in Daghestan it has been abandoned, because milking sheep demands great efforts (only done by men) and communal cooperation (taking turns to go to relatively distant pastures).

Past
Gelmets (local name [gɨmlecʼ / gɨlʲmec']) was slightly bigger than Kina. In 1886, there were 358 residents (Svod 1893). According to the 1926 census, the population was 403 people (Materialy 1927). According to the villagers' own estimates, by the middle of the 20th century the population of Gelmets continued to be bigger than that of Kina.
Compared to Kina, Gelmets had more land, especially more land suitable for pasturing. There were more cattle, and the village was better off. Nevertheless, Gelmets was also poor. There was not enough corn, and in winter people used to go away for seasonal jobs. In autumn, the male population went to Azerbaijan to work as shepherds and builders. There were several tinsmiths who spent winters in Georgia. The important difference between Gelmets and Kina was that, in Gelmets, the families stayed in the village with the cattle, and only men worked outside the village. Relocation from Gelmets to Azerbaijan, however, was popular. For example, several families originally from Gelmets live in the Azerbaijani village of Qaxbaş. In general, the Tsakhurs had closer links with Azerbaijan than the Rutuls, probably because the historical Ilisu sultanate, important in late middle ages, was split between the modern territories of Azerbaijan and Daghestan. Yet, at least in the late traditional setting that our method can reach, Kina seems to have stronger ties with Azerbaijan than Gelmets.
The advantage of Gelmets was its closeness to Azerbaijan. Villagers earned a living by working in Azerbaijan. In their opinion, they lived better than people from other Rutul and Tsakhur settlements located farther away from Azerbaijan.
Our consultants also emphasized the fact that their relations with Azerbaijan were much closer and more important for them than their relations with the lands of the Rutul. According to their claims, they had almost no economic interests linking them to Kina, their closest Rutul neighbour, let alone any other Rutul villages.

Present
According to the latest censuses, Gelmets had 571 residents in 2002 (Tablica № 2с 2004) and 611 in 2010 (Mikrodannye 2010), and is now smaller than Kina. Gelmets is located on a mountain ridge, which makes it very beautiful, but there is almost no space for new houses near the village, and no new houses are built in the valley. Gelmets has preserved its old picturesque layout with very dense disposition of houses.
In Gelmets, we found only two women from Kina and several from the Tsakhur village Kurdul.
There is a common belief widespread in Gelmets that their village is the oldest in the area, and their dialect of Tsakhur is the most beautiful and 'genuine' (Russ. pravdivyj). Our consultants express annoyance at the fact that the written standard of the Tsakhur language was based on the dialect of the village of Tsakhur instead of Gelmets.

Patterns of multilingualism
Four languages are currently spoken in the area, including Rutul, Tsakhur, Azerbaijani and Russian. The command of Lezgian and Georgian is reported in some individual cases. Rutul is the native language of Kina. Tsakhur is the native language of Gelmets. Native languages are not exposed to loss, while the command of the languages of the respective neighbours has significantly decreased over the observable time.

Villagers born before 1919 7
Unlike some other locations, especially in central Daghestan, the inhabitants of Kina and Gelmets did not show a good level of command of their neighbours' language. Only 25 percent of Kina villagers could speak Tsakhur. Only 7 percent of Gelmets residents could speak Rutul (see Table 3). Both in Kina and Gelmets, more than 90 percent of people spoke Azerbaijani. The communication between Kina and Gelmets people was thus most likely in Azerbaijani, which is also confirmed by our consultants. It helps to understand why people in Kina and Gelmets could hardly speak Lezgian, although it is present in the vicinity, and, at some periods, was even the language of administration. Most likely, Azerbaijani was also used for communication with Lezgians.
The following table, like similar tables below, reads as follows. Columns are grouped according to the villages, in this case Kina and Gelmets. Rows show linguistic repertoires of the villagers. For each village, the first two columns show the number of people who were reported to speak a language, men and women. The last column shows the percentage of people who were reported to speak the language, in both sexes, out of all people we have data for 8 . Command of Russian was low, as everywhere in Daghestan (28 percent in Kina, 12 percent in Gelmets).

Kina
Command of the Arabic script is as average across Daghestan. 19 percent of the population could read Arabic letters in Kina, and 18 percent in Gelmets; 13 percent could translate (understand) Classical Arabic in Kina, and 21 percent in Gelmets.
Gendered patterns of multilingualism are observed only in the generations born before 1919. While the command of Azerbaijani was typical for all population, less common languages (Tsakhur in Kina, Rutul in Gelmets, and Russian in both villages) were spoken almost exclusively by men. Command of the Arabic script was also restricted to men.

Villagers born after 1919
In the generations born after 1919, the command of Russian has risen up to 84 percent in Kina and 78 percent in Gelmets (see Table 3). The youngest person reported not to speak Russian was born in 1979 in Kina, and in 1972 in Gelmets. Meanwhile, the command of other second languages rapidly went down. People stopped acquiring Azerbaijani a long time before the border between Russia and Azerbaijan was closed. The youngest person who speaks Azerbaijani was born in 1977 in Kina and in 1984 in Gelmets.
The command of Arabic script is also almost entirely lost. Unlike some other parts of Daghestan (see Section 6), after Perestroika, learning Arabic did not become popular in Kina or Gelmets.

The cluster
This cluster includes three villages: Rikvani, Zilo and Kizhani. The villages lie in the Unsatlen valley where most settlements speak Andi. Andi belongs to the Avar-Andic branch and is now spoken by circa 35-40,000 speakers (see Figure 1). There are 14 Andi villages in the mountains with a total population of 16,900. According to Aglarov 2002, there are about 20,000 Andi people living outside the Andi valley. The dialects of Andi show a strong variation.
Andi has no literacy tradition. Andi-speaking people read and write in Avar, which is only distantly related to Andi. There is no mutual intelligibility between Andi and Avar. The Andis learn Avar as a foreign language. Until recently, Andi people were registered as Avars in their passports.
In Rikvani and Zilo two different varieties of Andi are spoken. The villagers of Kizhani speak Avar. Kizhani is located at the border between the Avar and Andi speaking areas. The territories of all three villages share borders. The distance from Rikvani to Zilo and from Rikvani to Kizhani is between 6 and 7 km (about 60-80 minutes of walking). The distance from Zilo to Kizhani is a slightly less (5-6 km, or 40-60 minutes of walking).
Administratively, the villages belong to the Botlikhsky district. The most important settlements in the vicinity are Botlikh and Khasavyurt. Botlikh speaks a language of its own, spoken in only one other village, Miarso. According to the locals, communication with the residents of Botlikh was in Avar in the past. At present, it is either in Avar or in Russian. The markets were located in the Andi villages Andi and Gagatli and in Botlikh.
In the 19th century, the cluster was part of the Andi naibstvo (a traditional administrative division in northern Daghestan) of the Andi district.
Rikvani is located at 1,800 m above sea level and is the highest of the three villages. Zilo and Kizhani are lower, about 1,500 m.
The cluster borders with Chechnya. Compared to Daghestan, Chechnya was considered richer and more fertile territory. The Andis had intensive economic and social relations with Chechnya, disturbed, however, by territorial conflicts (Kapustina 2015). Villagers report that, in the past, children from Andi villages were sent to Chechnya for several months to learn Chechen.
In 1944, the Chechens were deported to Kazakhstan, and some Daghestanian people were forced to leave their home villages to resettle in Chechnya. In the area, the Avar village Kizhani was subject to resettlement but Andi-speaking villages were not forced to move. In Rikvani, there were several families who moved of their own will, seeking better life conditions. In 1957, the Chechen people were allowed to come back, and most Daghestanian people had to return to Daghestan, either to their original settlements or to the lowlands.
A Soviet school opened in 1932 or 1933. In the 30s and the 40s, the school had Russian and Avar teachers and headmasters. Rikvani people who obtained higher education and became certified teachers were born in the 1940s and came to work at the school in the 1960s.
Although Rikvani is located very high (1,800 m), there were enough crops to survive. Rikvani has good natural irrigation and fertile soil, so there was no need for irrigation. The main crops included wheat, barley, rye and oats; potatoes were also planted. Rikvani had several water mills. Because of the cold climate, maize was not cultivated. The villagers had sheep and cattle and made cheese. Seasonal jobs outside the village were not practiced.
Rikvani women manufactured felt cloaks (called burkas). According to Dirr, manufacturing felt cloaks was the main trade and signature product of the whole Andi area (Dirr 1906); Andi burka is an expression known in other areas of Daghestan.
Meat, cheese and felt cloaks were sold and exchanged for maize and nuts in Chechnya. The closest Chechen villages are Kharachoy (three to four hours of walking) and Dargo.
Rikvani people emphasized their strong connections with Chechens until the recent past. Most families had friends in Chechnya with whom they could stay when they came to Chechnya for commercial or other purposes. We were once told that a boy from Rikvani was brought up by a Chechen family because his father died and it was difficult for his mother alone to provide for the large family.
There were several markets in the vicinity. Travel to the markets in Andi and Gagatli took about one day, and there were also markets in Botlikh and Akhvakh, which were farther away. When people went to Botlikh or Akhvakh, they usually stayed there overnight.
Spouses were most often preferably chosen among second cousins and from within the same patrilineal clan. Nowadays, there are several wives from other Andi villages, and also a few Kumyk, Avar and Russian women.

Present
According to the latest censuses, Rikvani had 645 residents in 2002 (Tablitsa 2с) and 769 in 2010 (Mikrodannye 2010). According to the authors' personal estimate, in 2015 Rikvani had about 350 residents. Many families (probably about 100) moved to Kalmykia.
Rikvani has a full secondary school (1 to 11 grades, from 6/7 to 17 years old). In 2015, there were about 80 schoolchildren. In 2014, only eight children started school. The teacher of the elementary school reports that children start education without knowing Russian or Avar. At school, children are taught Avar (so called L1 classes) and Russian. All textbooks are in Russian. Andi is not included in the school curriculum. Children have to learn two non-native languages from their first year at school.
Adherence to religion is very strong. There are three mosques, and one more is under construction (compare to the period before the Soviet times when the village was twice as big but had only one mosque). Drinking spirits and smoking is strictly forbidden, and alcoholic drinks and cigarettes are not sold in the shops. There are two madrassas (one for men and one for women). During the day they are attended by children, and by older people in the evening.
Zilo is located lower than Rikvani, at about 1,500 m. This results in a significant difference between their economies. The main products were maize and potatoes. Unlike in Rikvani, the fields were artificially irrigated, and cattle were herded in distant pastures during the whole year. Some women made felt cloaks, but this skill was not as widespread in Zilo as in some other Andi villages.
As other villages in the area, the residents of Zilo had intensive contacts with Chechens. They used to go to Chechnya (through the village of Andi) to exchange meat and felt cloaks for wheat and maize. Some men went to Chechnya for seasonal jobs in husbandry, such as weeding and herding. In general, however, occupations outside the village were not common.
There is no forest near Zilo. There is a common opinion among the villagers that the forest was burned down in the 19th century during the forest burning campaign of the Russian army in the Caucasian war.
The school opened in the 1930s. The villagers of Zilo usually married within their village. Several people reported that they or their relatives were sent to another Andi village, Muni, to learn to read the Quran (i.e. to know the phonetic values of letters and to read aloud).

Present
According to the latest censuses, Zilo had 1,107 residents in 2002 (Tablitsa 2с) and 1,220 in 2010 (Mikrodannye 2010). In reality, however, the population of Zilo is comparable to that of Rikvani, numbering some 300 people. The residents of Zilo report that in the Chechen village of Urus-Martan there are people who speak Chechen as L1 but claim to have originated in Zilo.
Because of a milder climate, some people have started growing apricots, although they are not as good as the apricots coming from lower villages of the district and the lowlands.
Zilo has a full secondary school (grades 1111 grades, from 6 or 7 to 17 years old). There are three mosques but no madrasah. Some people teach their fellow villagers to read the Quran at home.
Kizhani, the only Avar village in the neighbourhood, is closer to Zilo than to Rikvani. Zilo is also closer to the large Andi village of Muni. Muni is usually considered an Andi-speaking village, but people in Rivani and Zilo say that the language of Muni is difficult for them to understand.
Zilo people do not call their language Andi. They usually refer to it as 'the language of Zilo' (ziludirab micʼi). There are some dialectal differences between Zilo and Rikvani, concerning lexicon, phonetics and morphology, but they are not considered as important as those between either of them and Muni.
Both Rikvani and Zilo are provided with natural gas, but only very recently in the case of the much higher Rikvani.

Past
Kizhani (local name [k'ižani]) is located at the same altitude as Zilo (about 1,500 m), but the life conditions differ, because Kizhani has fewer water sources. There is also a shortage of land appropriate for pastures and for agriculture, and no forest. The closest forest belongs to the Avar village Ichichali. In 1886, Kizhani was half the size of Rikvani and Zilo: 309 residents according to (Svod 1893). According to the 1926 census, the population of Kizhani was the same, 309 (Materialy 1927). The territory of Kizhani borders with Andi villages (the closest is Zilo) and with Avar villages (the closest is Ichichali).
Residents of Kizhani say that, in the 20th century, their village underwent four resettlements. In 1944, when Chechens were deported to Kazakhstan, the Kizhani people were forced to move to Chechnya, to the village of Tsa-Vedeno. In 1957, the Chechens returned, and most people from Kizhani came back to their home village. Some of them did want to return to their homeland, but most came back simply because they had to give the houses back to the Chechens. Later, in 1970, Kizhani suffered an earthquake. Some houses were destroyed, and some cracked. The Kizhanis started moving to Tukhchar, a settlement in Novolaksky district (155 km from Kizhani) where a viticultural kolkhoz needed hands. Those who stayed in Kizhani relocated within the village territory, from uphill where the old houses were built to lower, more convenient locations that were traditionally kept for agricultural purposes.
There was an elementary school in Kizhani before the resettlement to Chechnya. In the 1960s it reopened.
Kizhani people cultivated potatoes, wheat, and maize. Handicrafts were not widespread, but several people had forges and made pitchers. According to the villagers, the skill of making copperware was learned from the nearby Avar village of Ichichali, where most people were involved in blacksmithing.
Kizhani was not a rich village, as it was short of both land and water. There was only one water spring, located in the upper part of village, and women had to queue to fill their pitchers. There was not enough grain, so that it was brought from Chechnya in exchange for meat, butter and maize. The closest Chechen village is Kharachoy, reached via the village of Andi.
There is a common opinion that, in the past, there were mountain Jews (an ethnic group practising Judaism and speaking Tat, a southwestern Iranian language) living in the neighbourhood, and that the ruins of their villages still remain.
Many people from Kizhani live in the village of Tukhchar, located in the Novolaksk district at a distance of about 150 km.
There are 70 children at the local school. The school has 9 grades (from 6/7 to 14 years old). After that, children continue their education in the towns or in Tukhchar. They do not go to the schools in the neighbouring Andi villages.
At present, people earn for their living by growing potato and maize, breeding cattle and leaving seasonally for construction work. Most houses have had gas since 2013-2014.

Patterns of multilingualism
Command of four languages is common in the area, including Andi, Avar, Chechen, and Russian. A few people could also speak or understand Botlikh, Kumyk and Chamalal (another Andic language, designated by the villagers as 'the language of Agvali', which is the Chamalal speaking administrative center of the Tsumada district). Andi is L1 in Rikvani and Zilo. Avar is L1 in Kizhani. As in most highland villages of Daghestan, command of the native languages remains unwavering.

Villagers born before the year 1919
Most inhabitants of the two Andi villages, Rikvani and Zilo, could speak Avar. In Rikvani, command of Avar was reported for all people born before 1919, and in Zilo it was reported for 94 percent of our entries (see Table 5). Dirr (1906) wrote that Avar was spoken by all male population in all Andi villages. Our data did not show any gender biases.
For the Kizhani villagers, Avar is the L1. The language of their closest neighbours, Andi, was not spoken by the majority of the villagers of Kizhani. Only two out of 18 residents born before 1919 were reported to speak Andi. One of them, according to his relatives, worked as a shepherd together with L1 speakers of Andi.
The asymmetry in language relations found between the Andis and the Avars is a common pattern across Daghestan. The language that is spread over a larger territory (in this case, Avar) usually dominates a minority language.
Contacts with Chechens were reported for all three villages. Chechnya was more fertile and rich, and people used to go there for money. For example, an orphan from Zilo, born in 1886, was sent to Chechnya by his relatives so that he could work there in exchange for food. Three residents of Rikvani, born in 1900Rikvani, born in , 1904Rikvani, born in and 1910 were, according to their relatives, sent to Chechnya for one or two years for the purpose of learning Chechen. That the command of Chechen was present among the Andi people is confirmed by Dirr's observations at the turn of the century (Dirr 1906).
Exactly how widespread Chechen was differed among the three villages. Rikvani was the village closest to Chechnya, and Zilo was closer to Chechnya than Kizhani. Many villagers of Rikvani have lived in Chechnya for several years and /or had regular trade interactions with the Chechens.
It follows from our survey that 36 percent of the residents of Rikvani, 33 percent of the residents of Zilo, but only 13 percent of the residents of Kizhani could speak Chechen. The difference between the Andi villages and Kizhani is remarkable, because almost all Kizhani villagers spent part of their life in Chechnya (see 5.2.4.1). For example, a man born in 1894 spent in Chechnya fourteen years, and yet did not speak Chechen (although he could understand it). This situation is unclear; we could not establish which language was the main medium of communication. One of the possible explanations could be that the Kizhani Avars stayed in Avar environment while in Chechnya. We did not however collect any such evidence from the villagers.
The level of Russian among people born before 1919 was not very high. About 37 percent of residents in this generation spoke Russian in Rikvani, 23 percent in Zilo, and 29 percent in Kizhani.
Knowledge of Kumyk is reported only in Rikvani. It was not high (18 percent), and was usually explained to us as due to winter shepherding in the Babayurt district where Kumyk was the dominant language. Dirr (1906: II) indicated that command of Kumyk (referred to as Tatar by Dirr) is not common among the Andis, in contrast with some other parts of Daghestan («According to my observations, the command of Tatar, which is used as a lingua franca in eastern Daghestan, is less widespread in the valley of the Andi Koisu River»).
No one in this age group is reported to have been able to speak or understand the Botlikh language, although there were some contacts with Botlikh.
A striking feature in all villages is a very high level of Arabic literacy. Ability to read the Quran was reported for 50 percent in Rikvani, 38 percent in Zilo, and 72 percent in Kizhani. Few people, however, could understand the meaning of the text (translate from Arabic). Among the villagers born before 1919, the command of some languages was gendered. In all villages, women had less proficiency in Chechen, Kumyk and Russian -languages which were not acquired in contacts with nearby villages. Women were not involved in the activities associated with travelling to distant villages. Interestingly, the command of Andi in Kizhani was also typical only of men, although the language was spoken nearby, while vicinal bilingualism is not usually gendered in our data. Also unexpected is that the ability to read the Quran was almost the same among men and women in Rikvani and Kizhani. Zilo shows the pattern which is more typical for Daghestan: men were able to read the Quran more often than women. The command of Avar in Rikvani and Andi was equally high among men and women.

Villagers born after 1919
As elsewhere in Daghestan, people born after 1919 show a better command of Russian and start losing local languages as L2. This tendency is less prominent with minority groups, such as the Andis, whose L2 (Avar for the Andis) has been supported by its use in school. At present, however, there are many young people (born in the late 1990s) who do not speak Avar. Although for Kizhani the level of knowledge of Andi seems to have been preserved and even, percentagewise, increased (from 11% to 16%), this is most probably due to the sparse data from the earlier period.
The command of Chechen decreased significantly in all villages of the cluster. Even more important, people no longer acquire Chechen. The youngest person who speaks Chechen was born in 1974 in Rikvani, in 1955 in Zilo, andin 1940 in Kizhani. Knowledge of Kumyk is not reported after the 1925 year of birth.
Meanwhile, Russian has spread in all villages. Among people born after 1919, proficiency in Russian is observed in 80 percent of people reported in Rikvani, 80 percent in Zilo, and 82 percent in Kizhani. The oldest adult person who does not speak Russian was born in 1973in Rikvani, in 1984in Zilo, and in 1977 The only language skill that has not been lost and has even increased is the ability to read the Quran. Under the Soviet authorities, all Daghestan was secularized. There was a strict prohibition on religious practices, and these villages were not an exception. The post-Soviet revitalization of Islam was however so intensive in Andi villages and Kizhani that a large part of population now studies the Quran and can recite texts in Classical Arabic. Note that the knowledge of Arabic was also high in pre-Soviet times (see above in 6.5.1).
The gender patterns of multilingualism were lost in the generations born after 1919. Russian is spoken by almost everyone, as is Avar in the two Andi villages. Except for Arabic, knowledge of other languages has been lost almost completely by men (and was never widespread among women).
The language repertoire of Rikvani, Zilo and Kizhani villagers born after before 1919 is shown in Table 6. high language density,the highest in Europe (see Section 2). These traditional patterns, which were vigorous in the highlands prior to the intensive Russification of the 20th century, are now virtually lost and can only be accessed by interviewing villagers about their older relatives, in a way essentially similar to the methods used in oral history. The method is described in Section 3.1. It provides a window into the sociolinguistic past of the region, a window which is about to close. The need for documentation of the traditional multilingualism, now lost, is in a sense similar to the need to document endangered languages. The former is as important to sociolinguistics as the latter is important to linguistic typology.
The data collected in the course of the project allows us to raise novel questions and issues in sociolinguistic typology that could not be addressed before such a large amount of quantitative data on traditional multilingualism came into existence. In the last few years, we have started to address some of them.  show that the dynamics of acquisition of Russian, the emergent lingua franca, was clearly gendered. Dobrushina & Moroz (submitted) show that speakers of smaller languages were more multilingual than speakers of larger languages to a statistically significant extent. Other questions are being considered.
How universal was it, for traditional Daghestan, that one of the neighbouring languages was dominant in the sense of being used in interethnic communication (as was the case in Rikvani, Zilo, and Kizhani)? So far, our data shows that asymmetrical vicinal bilingualism was more common than the use of a lingua franca, while symmetrical vicinal bilingualism of the egalitarian type as described for Vanuatu (François 2012) was probably absent altogether. Were there any substantial differences between knowing distant languages and thoses of the closest neighbours? Can we see systematic differences between the data obtained directly and indirectly, a result that would not only test the degree of reliability of our data but may also shed light on sociolinguistic stereotypes connected to historical multilingual practices? Also, from sheer bilingualism rates, we are moving on to research topics related to language structure. Is there a correlation between the level of bilingualism and the intensity of feature sharing between contacting languages, including lexical borrowing (Daniel et al., submitted; Chechuro et al., submitted)? To what extent was the direction of borrowing dependent on asymmetry of bilingual patterns? The Atlas is designed to accumulate empirical data for research related to these or other issues in sociolinguistic history and typology.