# A grammar of Sanzhi Dargwa

Diana Forker

Languages of the Caucasus 2

### Languages of the Caucasus

Editors: Diana Forker (Universität Jena), Nina Dobrushina (National Research University Higher School of Economics, Moscow), Timur Maisak (Institute of Linguistics at the Russian Academy of Sciences, Moscow), Oleg Belyaev (Lomonosov Moscow State University).

In this series:


# A grammar of Sanzhi Dargwa

Diana Forker

Forker, Diana. 2020. *A grammar of Sanzhi Dargwa* (Languages of the Caucasus 2). Berlin: Language Science Press.

This title can be downloaded at: http://langsci-press.org/catalog/book/250 © 2020, Diana Forker Published under the Creative Commons Attribution 4.0 Licence (CC BY 4.0): http://creativecommons.org/licenses/by/4.0/ ISBN: 978-3-96110-196-2 (Digital) 978-3-96110-197-9 (Hardcover)

DOI:10.5281/zenodo.3339225 Source code available from www.github.com/langsci/250 Collaborative reading: paperhive.org/documents/remote?type=langsci&id=250

Cover and concept of design: Ulrike Harbort Typesetting: Diana Forker, Felix Anker, Felix Kopecky Proofreading: Ahmet Bilal Özdemir, Andrew Spencer, Aniefon Daniel, Daryl MacDonald, Felix Kopecky, Ivica Jeđud , Jeroen van de Weijer, Jezia Talavera, Laura Arnold, Laurentia Schreiber, Mykel Brinkerhoff, Jean Nitzke, Sebastian Nordhoff, Sune Gregersen, Tom Bossuyt, Alena Witzlack, Yvonne Treis Fonts: Libertinus, Arimo, DejaVu Sans Mono Typesetting software: XƎLATEX

Language Science Press Unter den Linden 6 10099 Berlin, Germany langsci-press.org

Storage and cataloguing done by FU Berlin


### **I Phonology 17**






### **III Verbal morphology 203**




### **IV Syntax 339**






## **Acknowledgments**

This grammar could not have been written without the support of many people and institutions.

First of all I am indebted to the Sanzhi community, in particular to my main assistant and consultant Gadzhimurad Gadzhimuradov, his wife Fatimat ('Batichaj') and his entire family. His passion for Sanzhi culture, history and language, his extraordinary patience in dealing with my endless questions and his linguistic intuitions made the work on this grammar possible and in fact a wonderful experience. He and his family hosted me during the entire project and allowed me to fully concentrate on my work while being their guest in Druzhba.

The grammar is one of the results of the language documentation project *Documenting Dargi languages in Dagestan – Shiri and Sanzhi* (2012–2019). The project was supported by many other members of the Sanzhi community in addition to Gadzhimurad, in particular by Isakadi Bakhmudov and Asabali Gadzhimuradov. A big *barkalla* goes to all of you!

The project has received financial support through a grant from the Volkswagen Foundation (DobeS program) to Diana Forker (Grant Number 86 357). This generous support is gratefully acknowledged. I am particularly thankful for the help of Vera Szöllösi (Volkswagen Foundation) and for the administrative support of Geoffrey Haig (University of Bamberg).

The project has been conducted by an international team of researchers, first of all Rasul Mutalov, whose idea it was to document endangered Dargwa languages and to apply for a grant from the Volkswagen Foundation. I also wish to thank my other two fellow project members, Oleg Belyaev and Iwona Kaliszewska, as well as the student assistants André Müller, Teresa Klemm, Felix Anker, and Nils Schiborr for scientific, technical and practical support at all stages of the project.

The grammar has profited from many discussions with researchers, colleagues and friends, especially the very active "Caucasian community" scattered around the globe, but united by the interest and the fascination for the languages of the Caucasus. In particular I want to thank the reviewers who kindly provided many comments, corrections and useful suggestions: Gilles Authier, Natasha Bogomolova, Ilya Chechuro, Denis Creissels, Misha Daniel, Nina Dobrushina, Dima Ganenkov, Linda Konnerth, Yura Lander, Timur Maisak, Zarina Molochieva, George Moroz, Jérémy Pasquereau, Stefan Schnell, Nina Sumbatova, and Jussi Ylikoski. I am also grateful to Sebastian Nordhoff, Felix Kopecky and the proof readers from Language Science Press for making it possible to publish high quality books that are accessible to everybody. Finally I am lucky to have the best family support I can think of.

Tanzt eure Revolution!

## **Spelling conventions**

The writing system used in this grammar largely follows previous works on other Dargwa varieties (Sumbatova & Mutalov 2003; Sumbatova & Lander 2014). Given below as well is the Cyrillic orthography, which I use with the Sanzhi community and which has been established in Forker & Gadzhimuradov (2017). It is almost identical to the established orthography of Standard Dargwa (plus sounds that Standard Dargwa lacks, minus sounds that do not exist in Sanzhi Dargwa). The letters given in brackets represent phonemes that occur only in loan words.


## **Glosses and other abbreviations**

### **Glosses**



### **Other abbreviations**



### **1.1 The Sanzhi community and the Sanzhi language**

Sanzhi Dargwa is an East Caucasian (i.e. Nakh-Dagestanian) language from the Dargwa (or Dargi) subbranch and belongs to the South Dargwa varieties (Glottocode: sanz1248). In the literature, there is no unique terminology referring to Dargwa languages, dialects or peoples, but several terms exist: Dargwa, Dargva, Dargi, or Darginskiy. For reasons of uniformity and unambiguousness I restrict myself to the label and the graphic representation *Dargwa* and will not use the other terms. Sanzhi Dargwa is spoken by approximately 250 speakers and is critically endangered. The self-designation of the Sanzhi people is *sunglan-te* (Sanzhi.person-pl) and the language is called *sunglan ʁaj* (lit. Sanzhi.person language).

More than 40 years ago, all Sanzhi speakers left the village of Sanzhi, their village of origin, in the Caucasian Mountains. Sanzhi is located in the Dakhadayevskiy rayon in central Dagestan (today part of the Russian Federation), which is predominantly inhabited by speakers of Dargwa languages.

The village of Sanzhi is located on the sunny side of the Ulluchay river valley, at an altitude of about 1,500 meters (Figure 1.5). The closest neighboring villages are Itsari, Shari, Khuduts, Ashty, and Amukh. The distance from Makhachkala is around 200 kilometers, from the regional center of the Dakhadayevskiy rayon, Urkarakh, it is 66 kilometers, and from Derbent around 150 kilometers. There is no direct road to Sanzhi. In order to reach the village, people go to Itsari by car or minibus and then walk around six kilometers until they reach Sanzhi. Currently, the Sanzhi territory is part of the nature park Itsari.

The village consists of approximately 30 houses, which are in very poor condition and not inhabited anymore (Figures1.1–1.3). The only house with a roof that is relatively well kept is the former school building. Sanzhi people regularly go to Sanzhi in the summer to spend a few days fishing, berry picking, and doing other activities in their former village. The village is surrounded by terrace fields that have been used for centuries to grow crops such as rye, wheat, barley, oats, and in the recent past also carrots, radishes, potatoes, and others. The traditional occupations of the Sanzhi people were farming and breeding, in particular sheep breeding. Not far from the village, ancient rock paintings can be found that, according to the Sanzhi people, have been the subject of investigation by several researchers from Russia. Unfortunately, I was not able to find literature on the paintings or the research expeditions.

From 1968 onwards, within a relatively short time span, all Sanzhi people moved to the lowlands to ethnically and linguistically mixed settlements. The major reason for the resettlement was the difficult life in the mountains. There was and still is no road leading

Figure 1.1: The village of Sanzhi in 2011 (courtesy of Gadzhimurad Gadzhimuradov)

Figure 1.2: The village of Sanzhi in 2013 (courtesy of Iwona Kaliszewska)

Figure 1.3: An old picture of Sanzhi, around 1957 (courtesy of the Sanzhi community)

Figure 1.4: The village of Druzhba in the winter of 2014 (picture by Diana Forker)

to Sanzhi, and also no electricity. From grade five on, children had to walk by foot to the school in Itsari every day and in all weathers.

Today, the majority of Sanzhi speakers live in the village of Druzhba in the Dagestanian lowlands (Kayakentskiy Rayon) (Figure 1.4) and to a lesser extent in other settlements in Dagestan and other parts of Russia. Druzhba is an ethnically and linguistically heterogeneous settlement with speakers of other South Dargwa varieties, other East Caucasian languages such as Tabasaran, Agul, Lezgian, and Lak, and also a very few Kumyk (Turkic) and Russian speakers. In Druzhba, people make a living by working in the local vineyards that used to be part of a *sovkhoz* (Soviet state farm). Many inhabitants, especially men, commute to other parts of Russia to work there and support their families back home. A map of Dagestan with Sanzhi and Druzhba is given in Figure 1.5.

### **1.2 The sociolinguistic situation of Sanzhi**

All languages of the Republic of Dagestan are official languages, but only 14 of them have the status of being officially written languages. Sanzhi Dargwa, like many other comparatively small languages and varieties spoken on the territory of Dagestan, does not belong to the written languages.

Before the arrival of Russian in the remote parts of the central Dagestanian mountains, where the original village of Sanzhi is located, Kumyk served as the language of interethnic communication in the wider area. The main traces of contact with Kumyk are the numerous Turkic loan words (e.g. the first part in *ač barq'ij* 'open' originates from the Kumyk verb *ač-maq*, *baχča* 'garden' (identical in Kumyk), *qːʷaz* 'goose' from Kumyk *qaz*, and many more). Nevertheless, among the Sanzhi speakers with whom I worked, nobody claimed to have a significant command of Kumyk. All villages, except for one<sup>1</sup> in the immediate neighborhood of Sanzhi, are Dargwa villages with Dargwa varieties closely related to Sanzhi, so that communication was and still is easily possible just by sticking to one's own variety.

Today, all Sanzhi speakers are bilingual or multilingual to various extents because they know at least some Russian. Russian serves as the main language of interethnic communication and is the only language used in education and administration, and more generally in the public sphere in Dagestan. The degree of bilingualism varies from speaker to speaker, but simplifying somewhat, it is possible to say that women of the oldest generation (60 years and older) are the only group for whom Sanzhi is the dominant language. Men of the oldest generation as well as many members of the middle generation (age 30 to 60) are more or less balanced bilinguals, and use the two languages in accordance with the different functional domains (public/official vs. private/speech community). All members of the youngest generation are dominant in Russian, but everybody has at least a passive command of Sanzhi and is able to use a simplified form of the language in communication with members of the oldest generation, e.g. in interaction between grandchildren and grandparents.

<sup>1</sup>The exception is the village of Shara that was originally inhabitated by speakers of Agul, but today it is also a Dargwa village according to my Sanzhi assistant.

Figure 1.5: Map of Dagestan

Figure 1.6: Sanzhi men at the Uraza Bayram, the holiday at the end of Ramadan in 2013 (Gadzhimurad Gadzhimuradov, who is dressed in dark clothes, is standing on the left side) (picture by Diana Forker)

Thus, the contact situation is largely language maintenance for the oldest and middle generation. Among the youngest generation language shift is observable, and it is reasonable to assume that members of the youngest generation in particular who are still children today will not pass on Sanzhi to their children. Some children and young people in Druzhba still learn Sanzhi as their first language (this depends on the family situation), but they come in contact with Russian right from the first day of their life. Russian becomes the dominant language at the latest when children start attending kindergarten. Therefore, they generally have a limited and mostly passive command of Sanzhi and prefer to speak only Russian. Sanzhi people of the young generation, including small children, speak predominantly Russian with each other. More and more Sanzhi people speak Russian not only to their neighbors in Druzhba, many of which are from other ethnic groups, but even at home. Although the people have a positive language attitude and are proud of speaking their own language, Russian is considered to be not only more prestigious, but extremely necessary for the future of their children (see Forker 2018c for more information).

Another factor influencing the linguistic situation is marriage between women and men from different ethnic groups, which usually does not lead to bilingual children acquiring both the language of the mother and of the father, but to children speaking only Russian at home, as the parents use Russian to communicate with each other. I estimate

### 1.3 Genealogical affiliation

that there are only a few families left in which both husband and wife are competent Sanzhi speakers that have grown up in the village of Sanzhi. We can assume that in the past the situation must have been different and the vast majority of wives were either from Sanzhi or from the surrounding villages (Itsari, Chakhri, Kunki, Duakar, Dzilebki are the main villages of origins of mothers and wives of the Sanzhi speakers with whom I worked).

Since Sanzhi Dargwa is not employed in the public domain (e.g. administration, education, media, court) the language is unwritten and used only for oral communication within the Sanzhi community. The only printed material so far is Forker & Gadzhimuradov (2017), a collection of traditional stories and other texts. In school, Sanzhi children have around two hours of mother tongue education per week, during which they learn Standard Dargwa. Sanzhi speakers do not understand literary Standard Dargwa, because Akusha Dargwa, the base for the standard language, is a Northern Dargwa variety and quite different from Sanzhi. Therefore, in spite of the school classes, Sanzhi children usually do not learn Standard Dargwa well and are not able to speak, write, or read in Standard Dargwa, or make use of the few newspapers and TV programs that exist.

### **1.3 Genealogical affiliation**

Sanzhi (Glottocode: sanz1248) belongs to the Dargwa (Dargi) languages, which form a subgroup of the East Caucasian (Nakh-Dagestanian) language family. The exact number of languages belonging to this family is unknown, but it can be estimated to be around 40. The internal classification of the family has not yet been unanimously resolved. Figure 1.7 shows one of the possible classifications (namely the classification according to Kibrik 1996: xi). The internal division of the Dargwa branch into subvarieties is largely taken from Koryakov (2006). Dargwa languages are commonly divided into a Northern Dargwa group and a Southern Dargwa group, whereby Sanzhi belongs to the latter. The spelling of the names for languages and varieties in Figure 1.7 follows the conventions established in the literature and in the recent handbooks on East Caucasian languages (Polinsky 2020; Koryakov et al. Submitted). Unfortunately, in a few cases this leads to differences between the spelling of a village name and the spelling of the language spoken in it (e.g. the village of Itsari vs. Icari Dargwa).

### **1.4 Dargwa languages and the problem of the "Dargwa ethnicity"**

Today, all languages spoken in the the Republic of Dagestan have the status of official languages (see the article 11 of the constitution of Dagestan, 2003). This includes Standard Dargwa and Russian, among others. There is a distinction between the so-called "unwritten" and the "written languages" of Dagestan. The latter are (in addition to Russian), Avar, Agul, Azerbaijani, Kumyk, Lak, Lezgian, Noghay, Rutul, Tabasaran, Tat, Tsakhur, and Chechen. Written languages of Dagestan are, in principle, taught in school and used

Nakh branch *Chechen, Ingush, Tsova-Tush (Batsbi)* Avar-Andic-Tsezic subbranch Avar-Andic *Avar* Andic *Andi, Botlikh, Godoberi, Karata, Akhvakh, Bagvalal, Tindi, Chamalal* Tsezic subbranch *Tsez, Hinuq, Khwarshi, Bezhta, Hunzib* Dargwa subbranch *Akusha/Standard Dargwa, Urakhi, Mugi, Tsudakhar, Gapshima-Butri, Mjurego-Gubden, Kadar, Muiri, Mehweb, Sirkhi, Amukh-Xuduc, Shiri, Qunqi, Icari, Sanzhi, Chirag, Kajtag, Kubachi-Ashti Lak Khinalug* Lezgic subbranch *Udi, Archi, Lezgian, Agul, Tabasaran, Tsakhur, Rutul, Kryz, Budugh*

Figure 1.7: A family tree of East Caucasian

to some extent in the media (e.g. newspapers, journals). Until 1928, speakers of Dargwa varieties used the Arabic script, but there was no standard orthography. From 1925 onwards, the first newspaper in a Dargwa language was published (Abdullaev 1954: 15). This newspaper, as well as most books and other materials, was published in Akusha Dargwa, the language which was later chosen as the basis for the literary standard Dargwa language. There are several reasons for this choice: Akusha was and still is the Dargwa variety with the most speakers, and the village of Akusha together with the surrounding villages formed an autonomous center (*vol'noe obščestvo*) for a long time. In 1930 at the first Dagestanian conference on orthography, Akusha was appointed to be the basis for the literary standard Dargwa language. In 1928, a Latin alphabet was developed for a number of Dagestanian languages including Dargwa, Avar, Lak, Lezgian, and Tabasaran. In 1938 the policy changed completely, and for all Dagestanian literary languages Cyrillic alphabets were introduced (Grenoble 2003: 48–51). In the following years the Dargwa alphabet underwent several changes.

Dargwa people are officially considered to be one group that shares a common ethnicity, and to speak various dialects of one and the same Dargwa language (see below for the viewpoint of linguistics on this). According to the data of the Russian census from 2010, for instance, about 510 000 people consider themselves to be ethnic Dargwa, and

### 1.4 Dargwa languages and the problem of the "Dargwa ethnicity"

thus represent the second biggest ethnic group in Dagestan (after the Avars). The vast majority of them claim to speak Dargwa.

Dargwa languages are spoken in the central part of Dagestan (traditionally in the districts Akushinskiy, Levashinskiy, Dakhadayevskiy, Sergokalinskiy, Kaytagskiy, and also partially in the districts of Gunibskiy, Buynakskiy, Karabudakhkentskiy, and Agulskiy), in a territory with a length of about 100 km and a breadth of about 70 km (Figure 1.8). In the west, this area borders on Lak and Avar territory. In the north and east, the Dargwa area borders on Kumyk lands, and in the south on Tabasaran lands.

The term *Dargwa* with its current reference was only introduced during Soviet times. There was a policy at the time to create names for peoples and languages that often lacked significance for the people themselves, and to introduce ethnic boundaries all over the Northern Caucasus (Grenoble 2003: 114). The use of these names is nowadays fully established and is largely maintained for political reasons (Šaxbanov 2009).

Historically, the term *Dargwa* (or *Dargi*) does not refer to an ethnic group (Abdullaev 1954: 13). There were seven unions of settlements in central Dagestan that referred to themselves with a proper name and the term *Dargwa*: Akusha Dargwa, Bukun Dargwa, Gutsi Dargwa, Kaba Dargwa, Utsmi (or Kaytag) Dargwa, Khamur Dargwa, and Sirkha Dargwa (Magomedov 1999: 13). That is, *Dargwa* referred to settlement centers that consisted of a number of small villages forming a unit, which were able to defend themselves and their own interests against enemies (*vol'noe obščestvo*). Other urban centers in the north, like Kadar and Gubden, whose inhabitants are also considered to be Dargwa people today (and to speak Dargwa varieties), did not belong to those units to which the term *Dargwa* was applied. They formed one administrative unit with Kumyk villages (Abdullaev 1954: 12), and used Kumyk as their lingua franca (Dobrushina et al. 2020; Wixman 1980: 58–59).

Similarly, there was not one single language with the name *Dargwa*, but a group of related languages, in reference to which the names of the urban centers were used (Uslar 1892: 1). But since Soviet times, the classification of the Dargwa varieties as dialects of one and the same Dargwa language has persisted in many publications and in all official documents (e.g. Abdullaev 1954; Gasanova 1971; Musaev 2002; WALS<sup>2</sup> ; Ethnologue<sup>3</sup> ).

Following the most recent publications on the internal classification of the East Caucasian language family (Koryakov 2006; Koryakov & Sumbatova 2007), the Dargwa branch consists of 19 languages and about 40 dialects (see Figure 1.7 above). The biggest are Akusha Dargwa (about 42 000 speakers), Mjurego-Gubden Dargwa (ca. 39 000), Urakhi Dargwa (ca. 35 000), followed by Kajtag Dargwa (ca. 21 000), and Tsudakhar Dargwa (ca. 19 000). Speakers of many Dargwa languages do not understand speakers of other Dargwa varieties, and the variation between them is much bigger than between the Andic languages, another subbranch of the East Caucasian family. The break-up of the Proto-Dargwa language can be estimated to have occurred about two millennia ago (Sumbatova, p.c.). However, the exact number of Dargwa languages is still subject to debate, because descriptions are lacking for many of the individual languages and dialects. Thus, Figure 1.7 will likely need to be corrected in the future.

<sup>2</sup>http://wals.info/

<sup>3</sup>http://www.ethnologue.com/

Figure 1.8: The East Caucasian (i.e. Nakh-Dagestanian) language family (map courtesy of Yura Koryakov)

### 1.5 Typological overview

The place of the Dargwa languages inside the East Caucasian family is also debated. Some authors consider them to form a separate branch of the East Caucasian language family (Gigineishvili 1977; Kibrik 1996: 142), others group them together with Lak (Haspelmath 1993; Koryakov 2006; van den Berg 2005).

### **1.5 Typological overview**

Sanzhi Dargwa is typologically similar to other East Caucasian languages. It has a relatively large consonant inventory including pharyngeal and ejective consonants, and a medium number of vowels. With respect to its morphosyntactic structure, Sanzhi is predominantly dependent-marking with a rich case inventory. The grammatical cases are ergative, absolutive, dative, and genitive. In addition, there is a plethora of spatial cases. The morphology is concatenative and predominantly suffixing. Sanzhi has an elaborate system of TAM forms. Verbal stems come in pairs that express imperfective and perfective aspect, and many can take spatial preverbs. Salient traits of the grammar are two largely independently operating agreement systems: gender/number agreement and person agreement. Gender/number agreement operates at the phrasal and at the clause level. Within the clause, it is mainly controlled by arguments in the absolutive case and shows up on verbs, adverbs, and on nouns in some of the spatial cases. Person agreement operates at the clausal level only, and functions according to a person hierarchy. Sanzhi has ergative alignment at the level of morphology. SOV is the most frequent constituent order.

Features of Dargwa languages that have attracted the attention of typologists and linguists working within various theoretical frameworks include gender and person agreement (Sumbatova 2011; 2013; Belyaev 2013; 2017a,b; Ganenkov 2018; Forker 2016b), complement constructions including reported speech (Ganenkov 2012; Forker 2019c), experiencer constructions (Comrie & van den Berg 2006; Ganenkov 2006; 2013), local and long-distance reflexivization (Forker 2014), backward control and long-distance agreement (Serdobolskaya 2009; 2010; Belyaev 2016), the expression of space (Ganenkov 2010; Forker 2019a), information structure (Sumbatova 2009; Forker & Belyaev 2016; Forker 2016b), and the problem of finiteness (Kalinina & Sumbatova 2007).

### **1.6 Literature on Dargwa languages, Dargwa people, and previous works on Sanzhi**

In comparison to some other Dagestanian languages, the description of Dargwa languages has a relatively long tradition. However, despite the impressive number of monographs and articles that have been dedicated to various Dargwa languages, the scope and the quality of many of these works cannot satisfy modern scientific standards. Thus, in the following I will mention only those works that are still in use and represent valuable documentations and analyses of Dargwa. For a more detailed overview on the history of the study of Dargwa languages, see Magometov (1983) and also the references in Temirbulatova (2005).

The first scientific treatment of a Dargwa language (Urakhi) comes from Uslar (1892), who visited the Caucasus in the second half of the 19th century. The next key scholar is Said Abdullaev, who published a Russian-Dargwa (i.e. Akusha) dictionary and a grammar of Akusha (Abdullaev 1950; 1954). Since the 1950s, Saida Gasanova has written many articles and books about various Dargwa languages and dialects, concentrating mainly on Muiri, Mjuregi, Urakhi, and Tsudakhar (e.g. Gasanova 1961; 1971). Other important scholars are Zapir Abdullaev, who worked on Standard Dargwa and occasionally on Urakhi and Kajtag (e.g. Abdullaev 1961; 1969; 1971; 1986; 1993; Abdullaev et al. 2014), and Magomed-Said Musaev, who investigated various Dargwa varieties, including Chirag and Akusha (e.g. Musaev 1975; 1978; 1983; 1980; 1984). There are also works on Sikhi (Kadibagomedov 1998), on Kajtag (Temirbulatova 2005) and most notably on Kubachi (Magometov 1963). Recently, two new dictionaries have been published (Jusupov 2005; 2009). Rasul Mutalov, one of the key participants in the language documentation project resulting in this grammar, has written a number of papers and books on Icari Dargwa and Standard Dargwa (Mutalov 1992; 2002; 2018).

In 1999, the first book in English on a Dargwa language (Akusha), written by van den Berg was published, followed by a descriptive grammar of Icari Dargwa, which was co-authored by Nina Sumbatova and Rasul Mutalov (Sumbatova & Mutalov 2003). Icari Dargwa is closely related to Sanzhi Dargwa; the two varieties are mutually intelligible and the Icari grammar was a fruitful source of inspiration for this grammar of Sanzhi.

In Moscow, a group of linguists works on a number of Dargwa languages, of which the major results are comprehensive studies of Tanti (Sumbatova & Lander 2014), Shiri (Belyaev In Preparation), Mehweb (Daniel et al. 2019), Ashti (Belyaev 2012) and Chirag (Ganenkov Submitted). Other important works from the same group are Kalinina & Sumbatova (2007), Sumbatova (2009; 2010; 2011; 2013), Lander (2008; 2010), and Serdobolskaya (2009; 2010). Sumbatova (Submitted) provides a recent overview on Dargwa varieties. Sketch grammars in preparation include Ganenkov (Submitted) and Forker (Submitted).

Topics in the morphosyntax of Sanzhi and other aspects of Sanzhi have been treated in Forker (2016b; 2014; 2019b; 2018c; 2019c; Accepted). A collection of texts with Russian translations and a Sanzhi-Russian and Russian-Sanzhi dictionary is Forker & Gadzhimuradov (2017).

There is not much to say with respect to the ethnographic literature on Dargwa people. There are only two older monographs (Schilling 1949; Gadžieva et al. 1967).

### **1.7 Documenting and describing Sanzhi Dargwa**

This grammar is the result of a language documentation project, *Documenting Dargi languages in Dagestan – Shiri and Sanzhi*, funded by the DoBeS program of the Volkswagen Foundation. The project officially started in 2012 and ran until 2019. Within this project, three linguists (Diana Forker, Rasul Mutalov, Oleg Belyaev), one anthropologist (Iwona Kaliszewska), and student assistants from the Universities of Bamberg and Leipzig (André Müller, Teresa Klemm, and Felix Anker) documented, described, and analyzed the two endangered East Caucasian languages Shiri Dargwa and Sanzhi Dargwa.

### 1.7 Documenting and describing Sanzhi Dargwa

Detailed information about the project, the languages and many texts, recordings and pictures can be found on the project website.<sup>4</sup> All materials gathered in the project are accessible upon request via the Language Archive hosted by the MPI Nijmegen.<sup>5</sup> The major results of the project are, in addition to the grammar of Sanzhi, a book with narratives, legends and other texts for the Sanzhi community (Forker & Gadzhimuradov 2017), the electronic corpus of Sanzhi texts with audio recordings for every text and many video recordings (around 24 hours of natural speech), and an electronic dictionary. Around 15 hours of speech have been transcribed in ELAN, translated into Russian, and are deposited in the Language Archive.<sup>6</sup> A subcorpus of around 10 hours, which amounts to more than 46 000 word tokens, has been fully glossed with FLEx<sup>7</sup> and translated into Russian and English. The texts have almost exclusively been recorded by myself in the village of Druzhba. During the recordings I was accompanied by Rasul Mutalov, my fellow project member, linguist and native speaker of the neighboring Icari dialect, or by Gadzhimuard Gadzhimuradov, my main language assistant, who led the conversation and explained the aims of the project to the Sanzhi speakers. After recording the text were transcribed in ELAN by using a Cyrillic orthography (page xvii) and by making use of the help of native speakers. They also provided a Russian translation. In the ELAN file I added a Latin transliteration following the orthography, which is also employed in this grammar (page xvii). From the transcribed texts I chose a subcorpus, transferred the Latin transcription into FLEx, glossed it and partially added English translations to the Russian translations.

The glossed corpus has been put on the internet and is freely is accessible.<sup>8</sup> This corpus consists of 75 texts from 24 speakers of Sanzhi who were between 21 and 80 years old when the texts were recorded (mostly between 2012 and 2015). Only three of the speakers were 35 years or younger, whereas most were older than 50. Slightly more than half of the speakers were female, but the majority of texts originate from male speakers.

The corpus contains the following types of texts:


<sup>4</sup>http://www.kaukaz.net/dargwa/sanzhi/lexicon/index.htm

<sup>5</sup>http://dobes.mpi.nl/projects/shiri\_sanzhi/

<sup>6</sup>https://archive.mpi.nl/

<sup>7</sup>https://software.sil.org/fieldworks/

<sup>8</sup>http://web-corpora.net/SanzhiDargwaCorpus/search/index.php?interface\_language=en


The natural data has been complemented by many hours of elicitation. All natural examples originating from the corpus are not further marked in this grammar. All examples which have been elicited are marked by (E).

The electronic dictionary of Sanzhi was built up with Lexique Pro<sup>10</sup> and has been published with *Dictionaria*. <sup>11</sup> The dictionary contains around than 5 500 entries written with Cyrillic and Latin script, Russian and English translations, grammatical information, and example sentences as well as audio recordings for (almost) every entry. The dictionary is also accessible via the project homepage.<sup>12</sup>

In August 2017, my main assistant Gadzhimurad Gadzhimuradov and I were able to print a book with community materials and present it to the Sanzhi community in Druzhba (Figure 1.9). The book contains 42 texts of various genres taken from the corpus (fairy tales, legends, anecdotes, descriptions of games and recipes, oral history, and a poem) written in the Cyrillic Sanzhi script with a sentence-by-sentence translation in Russian, as well as a Sanzhi-Russian and a simplified Russian-Sanzhi dictionary, which is also available on the project website.

Within the project I have undertaken more than ten field trips to Druzhba (including two short trips to Sanzhi in 2013 and 2016) in order to gather materials on the language. My major language assistant and consultant during all these years was and is Gadzhimurad Gadzhimuradov (Figure 1.6), a videographer and cameraman from Druzhba, who was born in Sanzhi. After spending his first five years there, his family moved to Druzhba, but he has ever since kept close relationships with the village and is a strong patriot in the best sense. Without the support and friendship of him and his family, in particular his wife Batichay, neither the grammar nor the entire project could have been realized. Gadzhimurad Gadzhimuradov not only helped me to gather, transcribe, and translate materials, he also made many recordings by himself, translated texts into Sanzhi and raised the interest of the Sanzhi community in the project. Patiently he sat down endless hours with me to go through morphological and syntactic paradigms. This grammar could not have been written without his assistance.

<sup>9</sup>http://catalog.paradisec.org.au/collections/SocCog

<sup>10</sup>http://www.lexiquepro.com/

<sup>11</sup>https://dictionaria.clld.org/contributions/sanzhi

<sup>12</sup>http://www.kaukaz.net/dargwa/sanzhi/lexicon/index.htm

Figure 1.9: Gadzhimurad Gadzhimuradov presenting the first book in Sanzhi (courtesy of Gadzhimurad Gadzhimuradov, 2017)

# **Part I Phonology**

## **2 Phonology**

Sanzhi phonology is typical for East Caucasian languages with its relative large consonant inventory (§2.1) and medium vowel inventory (§2.2). Other topics covered in this chapter are the syllable structure (§2.3), pharyngealization (§2.4), stress (§2.5), and phonological and morphophonological alternations (§2.6).

### **2.1 Consonant inventory**

Table 2.1 displays the consonant inventory for Sanzhi. The table gives the phonemic value of the consonants and displays the orthographic representation used in this grammar in italics (see also page xvii for the Cyrillic orthography). The three series of stops are, in the order given in the table: voiceless non-ejective, voiced, and voiceless ejective. The two series of fricatives are voiceless and voiced. All velars and uvulars also occur in labialized form. All voiceless non-ejective stops and fricatives (except for the pharyngeal/epiglottal and the glottal sounds) also occur as geminates (i.e. tense).

The uvular stops /q/ and /qʷ/ have strong friction that makes them sound almost like affricates /q͡χ/ and /q͡χʷ/. The friction is absent from the ejective /q'/ and the geminates /qː/ and /qːʷ/.

The phonemic glottal stop is found in the noun *beʔe* 'blood' and at the end of some words, for instance in the root-final position of two verbs *ha-ʔ-* (pfv)/*h-erʔ-* (ipfv) 'say' and *b-erʔ-* (pfv)/*b-uʔ-* (ipfv) 'rot' and the numeral *kːaʔ-al* 'eight'. Except for *beʔe* 'blood', only loan words and names contain the glottal stop in root-medial position (e.g. *daʔim* 'continuation', in the male name *žaˁbraˁʔil*).

A non-phonemic glottal stop, which is not written, occurs before word-initial nonpharyngealized vowels, e.g. *aba* [ʔaba] 'mother', including vowel-initial words in compounds, for example *ca-ibil* [t͡saʔibıl] 'first' (one-ord), or occasionally at other morpheme boundaries of inflected words, for example, *a-uk-un* 'not eating' (neg-eat.ipfv-icvb) can be pronounced [aʔukʊn] or [aʊ̯kʊn].

The semivowel /w/ is realized as a voiced labiodental fricative [v] or as a labial-velar approximant [w].

In addition to the segments listed in Table 2.1, the voiceless labiodental fricative /f/ is attested in the ideophone *uf b-ik'ʷ-ij* 'blow' (whew hpl-say.ipfv-inf) and in loan words, mostly from Russian, e.g. *forel* 'trout'. In older loans it had been replaced with /p/, e.g. *purma* 'uniform' (< *forma*).

All plain consonants occur in word-initial, word-medial, and word-final position. Geminates are never found in syllable-final position. Three labialized consonants (/q'ʷ/, /χʷ/, /ʁʷ/) are also not attested in syllable-final position. Table 2.2 shows the distribution of

### 2 Phonology


Table 2.1: The consonant inventory of Sanzhi Dargwa

consonants by means of example words. The table contains a number of morphologically complex words for which the relevant sound happens to occur at the end of the root, but within the stem because the root is followed by suffixes (the root is given in boldface).

Final voiced stops do not undergo devoicing. Final voiceless non-ejective stops (/p/, /t/, /k/, /q/) are post-aspirated. Stops in final position are released. They are also released when a homorganic consonant follows, e.g. *urek-c'al* 'sixty', *ħaˁžat-ce* 'necessary' (needdd.sg), *c'elt-ne* 'gravestone-pl', *le-d=nu* (exist-npl=prt). If the voiceless stops /t/, /k/, or the voiceless affricate /t͡s/ occur at morpheme boundaries and are followed by homorganic consonants, all consonants are fully pronounced and released (1). Neither /t/ nor /k/ nor /t͡s/ become geminates under the described conditions, although gemination is otherwise a frequent process that applies across morpheme boundaries (§2.6.11). However, the ejective stop /k'/ can turn into a plain stop as shown in the examples in (1c).

### 2.1 Consonant inventory


### Table 2.2: Distribution of consonants

*<sup>a</sup>*The relevant roots of morphologically complex words are given in boldface. In these words, the respective sound occurs at the end of the root, but within the stem because the root is followed by suffixes.

### 2 Phonology


(1) a. *b-uˁc-ce*/*b-uˁc-te* 'thick' (n-thick-dd.sg/n-thick-dd.pl)


All velar and uvular consonants occur in plain and labialized forms. The labialized velars and uvulars can be followed by all vowels except /u/. Labialization is mostly found with syllable-initial consonants, but as Table 2.2 shows, there are also words with labialized consonants in final position. In most words, labialization is restricted to one consonant per root, but there are a number of words with two labialized consonants, e.g. *gʷagʷa* 'flower', *gʷargʷal* 'onion', and *xʷixʷit'* 'pipe'. In addition to labialization in roots, deletion of the vowel /u/ triggers labialization of the preceding consonant or following consonant (§2.6.10). Labialized consonants are mostly found in nouns, numerals, adjectives, adverbs, and verbs and also attested in a few particles, but not in pronouns or suffixes. Labialization is absent from Standard Dargwa and therefore speakers who have been trained in the Standard Dargwa orthography do not write them in Sanzhi, although they pronounce them. Younger speakers often replace labialized consonants by plain consonants and change a preceding or following *a* to *o* (in speech and writing). Minimal pairs for some labialized consonants are given in (2).


Geminates are always voiceless, non-ejective, and unaspirated. All voiceless non-ejective obstruents, except for pharyngeal/epiglottal and glottal segments, occur as geminates, and even a number of labialized consonants are geminates. The phonemic status of geminates is proven by the minimal pairs and minimal oppositions in (3).


2.2 Vowel inventory


Geminate fricatives are not always easy to identify because fricatives can be tense in emphatic pronunciation. But geminate stops and affricates are clearly audible as such, because there is a significant difference in the closure duration between singletons and geminates. Gemination can probably be analyzed as a difference between lax and tense consonants, but the exact phonetic properties of geminates still need to be clarified by future research.

In addition to their occurrence in stems, geminates occur at morpheme boundaries (see §2.6.11 below). A few sonorants can also occur as tense consonants within roots (/n/, /m/, /l/, /r/, and /w/) and/or at morpheme boundaries, but their phonemic status needs further clarification. Only geminates of /n/, /r/, and /l/ are found in native items (4); the other sonorants are only found in loan words (5).


### **2.2 Vowel inventory**

Sanzhi has four plain vowels and three pharyngealized vowels, of which one (*iˁ*) is very rare and whose phonemic status needs further clarification. Pharyngealized vowels and pharyngealization are treated in §2.4. The vowels *i*, *e* and *u* have lax and tense variants, whose distribution is not entirely clear. Table 2.3 shows the vowel inventory with the orthographic symbols used in this grammar.

There is a long vowel [aː], which is not phonemic, but occurs relatively frequently. It shows up only as sequences of homorganic vowels at morpheme boundaries (6) (most often in negated verb forms), and occasionally as a stressed variant of short vowels. Long vowels mostly occur in open syllables, but can occasionally also be found in closed syllables. The negative present-tense copula-auxiliary normally has a short vowel, but when it is used as existential or locational copula, the first vowel becomes long (6d).

	- b. *aːčːib* < *a-ačː-ib* 'did not get' (neg-get.pfv-pret)
	- c. *čiaːžib* < *či-a-w-až-ib* 'did not see' (spr-neg-m-see.pfv-pret)
	- d. *b-aːkːu* 'does not exist' vs. *akːu* 'is not' (cop.neg.prs)

The long high front vowel [iː] is rarely found when spatial preverbs are prefixed to some verbs having [i] as stem vowel (see §2.6.5 below for examples).

### 2 Phonology


Table 2.3: The vowel inventory of Sanzhi Dargwa

Sanzhi also has four diphthongs [ʊɪ], [aɪ ̯ ], [εɪ ̯ ], and [aʊ ̯ ̯] that can be analyzed as consisting of two phonemes, a vowel, and a semivowel. Examples are given in (7).

(7) *čuj* 'for themselves' (refl.pl.dat) *nejg* 'milk' *ʁaj* 'word, talk' *alaw* 'around' *caw* 'is' (cop.prs.m); 'himself' (refl.sg.m)

### **2.3 Syllable and word structure**

The minimal syllable consists of a single vowel. Initial vowels are always preceded by a non-phonemic glottal stop not indicated in the orthography. The syllables in monomorphemic native words are V, VC, VCC, CV, CVC and CVCC. In other words, syllables never have complex onsets, but can have complex codas. The general syllable structure is shown in (8).

(8) (C)V(C)(C)

In the onset, every consonantal phoneme can occur (see Table 2.2 above for examples), whereas in the coda not all consonants are allowed. Note, however, that simple underived verbs have stronger restrictions because they can basically only have /l/ and /r/ in the onset as well as the pharyngeal stop /ʡ/ (in addition to gender exponents and consonants used in the deixis/elevation preverbs, see §11.2 for more details on the structure of verbs). Labialized consonants in syllable-final position are rarer than in syllableinitial position, but they are attested. Ejective consonants are also found (Table 2.2). By contrast, geminate (i.e. tense) consonants are prohibited in the coda of syllables. Thus, geminate consonants in roots that happen to occur at the end of syllables in morphologically complex words, for instance after suffixation, are regularly shortened (see §2.6.11 for examples). The nucleus consists of one vowel, which under certain circumstances can be long (§2.2). The minimal syllable (and word) consists of the nucleus only (9). Some

2.3 Syllable and word structure

words can be seen as containing diphthongs, but diphthongs are analyzed as a sequence of a vowel and a semivowel. The most frequent syllable type is CV (10), but VC (11) and CVC (12) are also relatively common. By contrast, syllables of the type V are relatively rare, and *u* 'you' is the only native word that consists just of a minimal syllable.

(9) V

*u* 'you' 2sg *a.law* 'around'

(10) CV

*šːi* 'village' *qu* 'field' *χːʷe* 'dog' *ʁuˁ.ra* 'hare' *a.ba* 'mother' *du.ra.zi* 'threshing floor'

(11) VC

*at* 'to / for you' (2sg.dat) *eb.la* 'in spring'

(12) CVC


As mentioned in §2.2, there are no phonemic long vowels. Long vowels occasionally show up at morpheme boundaries or when the vowels are stressed or emphasized.

The only types of superheavy syllables are VCC (13) and CVCC (14), with only sonorants (/r/, /l/, /n/, /m/, /j/) and /b/ permitted in the position of the first consonant in the coda. Up to now I found only one exceptional noun that has a fricative before the second consonant, this being *q'ast* 'aim, intention, plan'. This noun is a loan ultimately from Arabic (*qaṣ̊d*); in Standard Dargwa its form is *q'as*. The syllable-final consonants of superheavy syllables can only be plain stops, fricatives, or affricates including ejectives, geminates, and labialized stops (i.e. obstruents). Although they are mostly voiceless, there are also a few examples of voiced fricatives in the final position of (C)VCC syllables (13), (14).

(13) VCC


(14) CVCC *laˁbz* 'mortar' *daˁrqʷ* 'barn' *nejg* 'milk' *laˁmc'* 'lightning' *kabc* 'skin, fell' *c'ult.mi* 'plum' *jebš* 'base' *t'ult'* 'bread' *b-ark* 'inside' *b-arx* 'direct, straight' *ku.bart* 'pressed dung' *qːuˁš.tːunk'* 'rolling pin' *q'ast* 'target, intention, idea'

### 2 Phonology

There are no native words with syllable-initial consonant clusters. Consonant clusters in (older) loans are broken up by insertion of epenthetic vowels either between initial consonant clusters or before them. In the first case, the vowels vary and are often identical to the following vowel, as the first three words in (15) show. In the second case, the vowel is /i/, as in the last two words:


Another possibility is to apply metathesis, though this process is very rare, for example Russian *brigadir* > Sanzhi *birgadir* 'brigadier'.

The minimal word (i.e. free root) has the shape V (see the example in (9) above). Minimal bound roots seem to consist of a single consonant and are only found among verbs. Examples are *ha-ʔ-* (pfv) 'say' and *ka-xʷ-*/*ha-xʷ-* (pfv) 'pour, add'. These verbs obligatorily contain preverbs and the vowel can be analyzed as either belonging to the preverb (which results then in the monosegmental verb stems) or to the verbs, or two both (*ha-* + *aʔ-* > *haːʔ-* > *haʔ-*).

### **2.4 Pharyngealization**

The most frequent pharyngealized vowel is /aˁ/, but /uˁ/ is also relatively common, whereas [iˁ] is restricted to very few words. The vowel /aˁ/ has phonemic status in Sanzhi as the following minimal pairs and minimal oppositions show (16).


The vowel /uˁ/ is far less frequent than /aˁ/, and thus I so far have found only one minimal pair and only a few examples of minimal oppositions in which the pharyngealized vowels only occur after uvular and pharyngeal sounds (17).


2.4 Pharyngealization

There are a few words that seem to have a pharyngealized high front vowel, e.g. *biħ-iˁb* 'they fought' (hpl-fight.pfv-pret), *w-irʡ-iˁb* '(they) betrayed him' (m-betray.pfvpret), *b-iˁʡ-iˁj* 'steal' (n-steal.pfv-inf), *čːiˁħri* (village name). However, speakers are uncertain about the presence of [iˁ] in Sanzhi words. Furthermore, I do not have any (near-) minimal pairs with pharyngealized and non-pharyngealized high front vowels, and thus further research is needed.

The vast majority of pharyngealized vowels occur in the adjacency of the uvular or pharyngeal consonants (see Table 2.2). When pharyngealized vowels occur in roots that contain those consonants, the vowels most frequently follow the consonants, but can also precede them (18). The respective consonants are /q/, /q'/, /qː/, /χ/, /ʁ/, /χː/, /ʡ/, /ħ/ for /aˁ/ and /uˁ/, and for /aˁ/ also the labialized consonants /qʷ/, /q'ʷ/, /χʷ/, and /ʁʷ/. The remaining uvular and pharyngeal consonants (/qːʷ/, /χːʷ/) are in general rare and I have not found any words that contain both the consonants and pharyngealized vowels.


The pharyngeal stop /ʡ/ cannot be followed by non-pharyngealized /a/ or /u/, but only by non-pharyngealized /e/ or /i/, that is \*/ʡa/ and \*/ʡu/ (18). And the pharyngealized vowels /aˁ/ and /uˁ/ are never followed by the glottal fricative /h/, but only by the pharyngeal fricative /ħ/, that is \*/aˁh/, \*/uˁh/.

Nevertheless, pharyngeal /aˁ/ and to a lesser extent /uˁ/ can also be found in stems that do not contain uvular or pharyngeal phonemes (19) (see also the first minimal pair in (16) above).


There are a number of words that contain two pharyngealized vowels. The vowels can be either identical (i.e. both vowels are /aˁ/ or both vowels are /uˁ/ or they are /aˁ/ and /uˁ/ in either order (20).


There is also some variation with those words that contain two pharyngealized vowels, in the sense that some speakers pharyngealize only one vowel whereas others pharyngealize both. The vowel that is optionally pharyngealized can be the first (21) or the second vowel (22).

### 2 Phonology

	- a. *puˁšːuˁk'*/*puˁšːuk'* 'blister'
	- b. *maˁlʡuˁn*/*maˁlʡun* 'snake'
	- c. *ʡuˁruˁs*/*ʡuˁrus* 'Russian'
	- a. *daˁʡaˁna*/*daʡaˁna* 'secret, secretly'
	- b. *durħuˁ*/*duˁrħuˁ* 'boy, son'

I found very few words that contain only one vowel that can optionally be pharyngealized (23).

	- b. *zaˁnʁ*/*zanʁ* 'ring' (ideophone)
	- c. *čaˁʁir*/*čaʁir* 'wine'

There are also two derivational suffixes containing pharyngealized vowels. The suffixes are *-q'aˁ* and *-uˁq'*. They are not productive and derive agent nouns from other nouns, infinitives, and parts of compound verbs form (§3.5.2). These suffixes do not have allomorphs with plain vowels.

Pharyngealization is a suprasegmental feature that spreads to inflectional prefixes and suffixes, even in those words that do not contain pharyngealized vowels in the root, but uvular/pharyngeal consonants. Only those prefixes and suffixes are affected that start with the vowels /a/ and /u/ such that in affixes only the pharyngealized vowels /aˁ/ and /uˁ/ occur, but no other pharyngealized vowels. Other affixes that contain the same vowels but start with a consonant do not have pharyngealized variants, for instance the vowel in the local participle suffix *-na* cannot be pharyngealized (24).

(24) *guči d-urq-aˁdi* 'I gathered' vs. *guči d-urq-na* 'the place of gathering'

In the case of the negation prefix *a-* this leads to a long pharyngealized vowel (see §2.6.5 for long vowels resulting from sequences of identical vowels):

(25) a. *či-aˁ-aˁħ-un* > *či-aˁːħ-un* 'did not fly on (something)' (spr-neg-fly.pfv-pret) b. *aˁ-w-aˁq-ib* > *aˁ-aˁq-ib* > *aˁːq-ib* 'did not hit' (msc.) (neg-hit.pfv-pret)

Furthermore, only affixes in immediately preceding or following syllables are affected. Pharyngealization does not spread over the entire word. For nouns the suffixes containing pharyngealized vowels are the plural and oblique plural suffixes, as well as one suffix deriving actions nouns (*-a*; §3.5.5). For verbs, the suffixes can be derivational (the causative suffix, the spatial preverbs) or inflectional (negation prefixes *a-* and *ma-*, various TAM suffixes). Examples are provided in (26).

2.4 Pharyngealization

	- b. *baliqː-aˁ-lla* 'of the fish' (fish-obl.pl-gen)
	- c. *ruˁrq-uˁl* 'boiling' (boil-icvb)
	- d. *b-iħ-aˁq-ib* 'made fight' (hpl-fight-pfv-caus-pret)
	- e. *b-iʡ-uˁn* 'stole' (n-steal.pfv-pret)
	- f. *b-aˁq-aˁjaˁ* 'hit it!' (n-hit.pfv-imp.pl)

The pharyngealized articulation associated with the vowel is maintained when the vowel changes, that is when there is vowel mutation *a* > *u*, as, for instance, with plural forms of some nouns (27).

(27) a. *ʡaˁrʡaˁ* 'chicken' > *ʡuˁrʡ-e* 'chickens' (chicken-pl) b. *q'ʷaˁl* 'cow' > *q'uˁl-e* 'cows' (cow-pl)

There is one verb 'go' that occurs without a root vowel when prefixes are attached and with a root vowel that can be pharyngealized or plain otherwise. The suffixes used with this verb are obligatorily pharyngealized (28a), (28b), whereas for prefixes pharyngealization is optional (28c).

(28) a. *maˁ-q'-aˁtːa* 'do not go!' (proh-go-proh.sg) b. *b-uq'-aˁq-ij* 'to make it go' (n-go-caus-inf) c. *saˁ-q'-aˁn* vs. *sa-q'-aˁn* 'going' (hither-go-ptcp)

At least with some affixes, pharyngealization is optional, and one can find one and the same inflected word form with and without affixes that contain pharyngealized vowels (29).


Pharyngealization includes loan words, even recent borrowings from Russian (30), which are not pharyngealized in the donor language. It is even noticeable when (older) Sanzhi people speak Russian.


### 2 Phonology

### **2.5 Word stress**

Stress is not a very prominent category in Sanzhi Dargwa. The stress is quite weak and the stress properties of words are very hard to determine, but they seem to be largely lexicalized. There are a few examples of minimal pairs or near minimal pairs that differ only in stress (31).


Some affixes attract stress, so that the position of stress in roots and in inflected word forms of one and the same lexeme may differ. For instance, plural suffixes of nouns normally attract stress (32).

(32) a. *qːap* 'sack' > *qːup-né* 'sack-pl'

b. *kur* 'pit' > *kur-mé* 'pit-pl'

The factors influencing placement of stress require further research.

### **2.6 Phonological and morphophonological alternations**

Sanzhi Dargwa has a variety of phonological and morphophonological alternations that affect vowels and consonants. Some of the processes that target vowels result from the fact that hiatus is not allowed, and therefore the underlying word forms have to be changed. A number of processes such as vowel deletion, alternation in the form of enclitics/affixes, degemination are syllable repair mechanisms, but others do not serve this function.

Processes affecting vowels are vowel deletion (§2.6.1), alternations in the form of enclitics/suffixes (§2.6.2), glide insertion (§2.6.3), glottal stop insertion (§2.6.4), long vowels resulting from sequences of identical vowels (§2.6.5), pharyngealization and formation of diphthongs (§2.6.6), and vowel mutation (§2.6.7).

Processes affecting mainly consonants are assimilation (§2.6.8), palatalization (§2.6.9), labialization/delabialization (§2.6.10), gemination (in combination with devoicing) and degemination (§2.6.11) (although labialization and delabialization also have an effect on vowels).

### **2.6.1 Vowel deletion (vowel syncope)**

Vowel deletion (vowel syncope) is one means of avoiding two subsequent vowels at a morpheme juncture. It is mainly found with encliticized negative auxiliaries. There are three types of vowel deletion within the domain of verbal morphology. First, sequences

### 2.6 Phonological and morphophonological alternations

of identical vowels might lead to the deletion of one vowel or to vowel deletion in combination with mutation (§2.6.7). Second, the initial vowel of the negative auxiliary is deleted when the auxiliary is used as an enclitic (33).

	- b. *qːuʁace + akːu* > *qːuʁa-ce=kːu* (beautiful-dd.sg=cop.neg)
	- c. *kabišːible + akːʷadi* > *ka-b-išː-ib-le=kːʷadi* (down-n-put.pfv-pret-cvb=cop.pst.neg.1)
	- d. *χʷalle + akːʷi* > *χʷal-le=kːʷi* (big-advz=cop.pst.neg)

Third, there is one verb 'go, direct oneself, move' of which the root vowel *u* is deleted when the gender agreement is masculine singular and the verbal root is preceded by the deixis/elevation preverbs or the negation prefixes (34). These preverbs and prefixes end in *a*. The process is accompanied by the labialization of the root consonant, that is, the labial feature turns from a vowel feature into a consonantal feature. See §2.6.10 below for more examples with the same verb stem.

(34) *ha-(w)-ulq-an* > *halqʷan* 'the one that goes upwards' (up-m-direct.ipfv-ptcp) (compare with *ha-b-ulq-an* up-n-direct.ipfv-ptcp)

As for nominal morphology, vowel deletion that is not caused by sequences of vowels is a regular component of plural formation. The deletion of *a*, *i*, or *u* in the final syllable of mostly disyllabic nouns is found with the suffixes *-be*, *-me*, *-re*, *-e*, *-ne*, *-upːe*, *-urbe*, and *-ube* (§3.2). If the last (usually stem-final) obstruent is a geminate it undergoes degemination. Examples are:

(35) a. *šuša* > *šuš-ne* 'bottle'


### **2.6.2 Alternations in the form of enclitics/suffixes**

There are suffixes and enclitics that have allomorphs whose use depends on the syllable structure of the item to which the suffixes or enclitics are added. The general rule is that suffixes/enclitics consisting of a single consonant are attached to vowel-final words and suffixes/enclitics of the form CV or VC to consonant-final words. Relevant suffixes and enclitics are:


### 2 Phonology


The last enclitic has another allomorph *=ew* that only occurs after the imperfective converb suffix *-ul*, perhaps to avoid having two identical vowels in two adjacent syllables (although this is generally allowed).

One can argue that the underlying form of the suffixes is the one with the vowel (*-li*, *-le*) and that the vowel is deleted in the appropriate contexts. However, there is no phonotactic need for such a deletion and it even goes against the general preference of open syllables in final position (§2.3).

### **2.6.3 Glide insertion**

Glide insertion represents a regular form of allomorphy. It is only found with the palatal glide *j* and only before certain suffixes or enclitics that start with the vowel *a*. The respective suffixes and enclitics are:


### **2.6.4 Glottal stop insertion**

Another means of avoiding two adjacent vowels is the insertion of a glottal stop. This occurs when spatial preverbs and negation prefixes are added to vowel-initial verbs. In the following examples the glottal stops are given (although they are normally not written in this position).

### 2.6 Phonological and morphophonological alternations


As can be seen in the above examples, in some cases alternative processes can be applied, namely the formation of a long vowel in the case of a sequence of two identical vowels (36e), (36g), and the change from /i/ to /j/ if the first vowel is /a/ and the second is /i/ (36b). The same can be observed for combinations of three vowels or compound verbs. For instance, if the verb form *sa-ʔargul* given in (36e) is negated, we get *sa-a-ʔargul*. In principle, it is possible to pronounce all three vowels separately, although this is not the preferred variant in practice. Similarly, in compound verbs the first part can be pronounced as a phonologically independent word or the two vowels can fuse across word boundaries, for example *dak'u uqandel* vs. *dak'uːqandel* 'if he appeared'. The precise conditions of glottal stop insertion still require clarification.

### **2.6.5 Sequences of identical vowels**

Long vowels can be the result of a sequence of two identical vowels or of the vowel /i/ plus the semivowel /j/. The latter happens when the dative suffix is added to nominals. The only long vowels are /aː/, /aˁː/, /iː/, and in one case after vowel mutation /eː/. The emergence of long vowels from two identical vowels is in many cases optional, with the insertion of a glottal stop being the usual alternative. Two identical vowels at morpheme boundaries occur only with verbs, either when the negation prefixes *a-* and *ma-* are used or with spatial preverbs. Note that in particular with the negation prefix it is the long vowel that carries the meaning of negation. If the two identical vowels would be shortened, the resulting verb form would be identical to the affirmative verb form and the negative meaning would be lost. Examples of the sequence are given in (37) and for the long pharyngealized low central vowel in (25).

	- b. *sa-arg-ul* > *saːrgul* 'coming here' (hither-go.ipfv-icvb) (alternative: *saʔargul*, (36e))
	- c. *qum a-art-u* > *qum aːrtu* 'does not forget' (forget neg-forget.ipfv-prs.3)
	- d. *sa-a-ka-b-išː-ib* > *saːkabišːib* 'did not put down' (in.front-neg-down-n-put.pfv-pret)

### 2 Phonology


Furthermore, the masculine singular agreement prefix *w-* can optionally be deleted when it occurs between two identical vowels (and, more generally, before /i/). This process also leads to long vowels [aː] and [iː] (38).

	- b. *ma-w-ax-utːa* > *maːxutːa* 'do not go!' (proh-m-go-proh.sg) (alternative: *ma-w-ax-utːa*)
	- c. *w-i-a-w-ax-an=de* > *wiʔaːxande* 'you will not go inside' (m-in-neg-m-go-ptcp=2sg) (alternative: *w-i-a-w-ax-an=de*)
	- d. *či-w-ig-ul=de* > *čiːgulde* 'you see him' (spr-m-see.ipfv-icvb=2sg) (alternative: *čiwigulde*)

### **2.6.6 Other general processes affecting vowels: Pharyngealization and formation of diphthongs**

Pharyngealization is a frequent process that is attested with verbal and nominal affixes containing *u* or *a*. The pharyngealization feature of verbal and nominal stems spreads to the closest prefixes or suffixes, but not to the entire word. The nominal affixes that have pharyngealized allomorphs are the plural suffix *-upːe*, oblique plural suffix *-a* and the suffix *-a* deriving actions nouns from verbs. The verbal suffixes are a variety of derivational and inflectional suffixes. See §2.4 above for more details and relevant examples.

There are two diphthongs that arise when vowels are followed by the semivowels /w/ and /j/. The diphthongs [aɪ] and [aʊ ̯ ̯], written *aj* and *aw*, are found in a few roots, e.g. *ʁaj* 'word, talk', *alaw* 'around'; more are given in (7). They also arise during certain inflectional or derivational processes. The first diphthong is attested with verbs having *i* as the root vowel and consisting only of one consonant, that is, verbs of the structure *(b-)iC(ː)*. They may have or may not have gender prefixes. When a spatial preverb *ka-*, *ha-*, *sa-*, or the negation prefixes (*a-*, *ma-*) are added, the result is *a* + *i* > *aj*. For verbs with gender prefixes the process only applies when the gender prefix *w-* for masculine singular is omitted, which is always possible for verbs that have the root vowel *i*. The forms with the gender prefixes *b-* or *r-* that do not contain the diphthong are given in brackets.

(39) *a* + *i* > *aj* (iC(ː) > jC(ː))


### 2.6 Phonological and morphophonological alternations

This process is optional to some degree. This means that under certain circumstances that need further investigation, the two adjacent vowels *a* and *i* can be pronounced separately, not forming a diphthong. For instance, *majsːit* can alternatively be pronounced *maʔisːit* (36b).

The diphthong [aʊ̯] arises when spatial preverbs or negation prefixes with the final vowel *a* are added to verbs with the root vowel *u*. This can be verbs with a gender prefix (*b-uC(ː)-*) that are inflected for masculine singular gender agreement (§20.2). The masculine singular prefix *w-* is regularly dropped before verbs with the root vowel *u* (e.g. *uc-ib* 'caught him' vs. *r-uc-ib* 'caught her'), and then the combination of the two subsequent vowels turns into a diphthong that will be written *aw* (40). In the following examples forms with overt agreement prefixes *b-* or *d-* are given in brackets at the end of the example lines.

	- a. *tːura-(w)-uq-un* > *tːurawqun* 'he went outside' (outside-m-go.pfv-pret) (vs. *tːura b-uq-un*)
	- b. *sa-(w)-uq-un* > *sawqun* 'he came' (hither-m-go.pfv-pret) (vs. *sa-b-uq-un*)
	- c. *gu-sa-(w)-uc-ib* > *gusawcib* 'kept him' (down-hither-m-keep.pfv-pret) (vs. *gu-sa-b-uc-ib*)
	- d. *a-(w)-uq'-idel* > *awq'idel* 'should I not go' (neg-m-go-modq) (vs. *a-d-uq'-idel*)

The same happens to verbs that do not have a gender prefix (*uC(ː)-*) when the root is preceded by prefixes ending in *a* (41). And again there are exceptions to the rule, e.g. in *sauq'ij* 'go towards, go to meet' the two vowels do not form a diphthong, but are separated by a glottal stop.

	- a. *ka-utː-ij* > *kawtːij* 'tear off, rip off' (down-tear.ipfv-inf)
	- b. *ha-utː-ij* > *hawtːij* 'pull out, disassemble, take apart' (up-tear.ipfv-inf)
	- c. *ha-uχːaq-ij* > *hawχːaqij* 'ignite, set fire' (up-sparkle-caus-inf)

### **2.6.7 Vowel mutation (apophony)**

Vowel mutation is found with inflected nouns and verbs. In the case of nouns, it is triggered by suffixation, in the case of verbs by prefixation.

The vowel *a* in the final syllable of nouns ending in a consonant is raised and backed when one of the plural suffixes *-e*, *-te*, *-be*, and *-re* containing close-mid vowels is added, that is, there is vowel mutation *a* > *u* (including *aˁ* > *uˁ*). The process can be accompanied by delabialization (§2.6.10 shows examples). There are also one instance each of *e* > *u* and *e* > *i* under the same conditions. Relevant examples are given in (42). See §3.2 for more nouns.

### 2 Phonology

(42) *a* > *u*; *e* > *u*; *e* > *i*

a. *qːap* > *qːup-re* 'sacks'


Vowel mutation with verbs occurs when the spatial preverbs or negation prefixes with the final vowel *a* are prefixed. The first type of verbal vowel mutation happens with verbs containing the stem vowel *i* that are inflected for masculine singular or lack gender agreement prefixes. The gender prefix is dropped and the two vowels merge. Verb forms with overt gender prefixes are given in brackets at the end of the example lines for comparison.

(43) *a* + *i* > *e*


This process is optional, but again the circumstances under which alternatives are allowed need to be clarified (44).

	- b. *a-irʁ-ib=da* > *erʁibda*/*aʔirʁibda* 'I did not understand' (neg-understand.pfv-pret=1)

The second type of verbal vowel mutation happens with verbs that have the stem vowel *e* and lack gender agreement prefixes (45). Note that in the first verb given below the vowel mutation results in a long vowel because the negation prefix *a-* assimilated to the stem vowel and this, in turn, leads to a sequence of two identical vowels, which then becomes a long vowel. This process commonly occurs when the negation prefix is added to verbs beginning with the vowel *a* because if the sequence would be shortened, the negated form would be identical to the affirmative form and negation could not be expressed (37). The same logic applies to *eːrčur* (45a).

(45) *a* + *e* > *e*


Finally, the combination of spatial preverbs ending with *i* and a verb without a gender prefix and *a* as stem vowel or a following preverb *ha-* 'upwards' also leads to vowel mutation. In the second case, when two preverbs combine, then the vowel mutation is 2.6 Phonological and morphophonological alternations

initiated by the disappearance of the glottal fricative. The affected preverbs are *či-* 'on' + *ha-* > *če-, kʷi-* 'in the hands' + *ha* > *kʷe-, hitːi-* 'behind, after' + *ha* > *hitːe-, b-i* 'in, inside' + *ha-* > *be-*. We can analyze this process as lowering of the vowel of the second preverb (46). Again the process is optional and does not occur in slow, careful speech.

	- a. *či-ag-ur* > *čegur* 's/he went' (spr-go.pfv-pret)
	- b. *či-ha-b-išː-ib* > *čebišːib* 's/he put it up' (spr-up-n-put.pfv-pret)
	- c. *kʷi-ha-b-uc-ib* > *kʷebucib* 's/he kept it in the hands' (in.hands-up-n-keep.pfv-pret)

### **2.6.8 Assimilation**

Progressive assimilation occurs with all verbal and nominal suffixes that have initial *l*. The liquid assimilates to a preceding sonorant *n* or *r* (47). The following suffixes are affected:


(47) *n* + *l* > *nn*; *r* + *l* > *rr*


With many words the process is optional, and in careful speech no assimilation takes place.

### **2.6.9 Palatalization**

Palatalization of velar consonants occurs with verbs when the causative suffix *-aq* or suffixes starting with the front vowels *i* (48a), (49), (50a), (50b) or *e* (48b), (50c) are added, or occasionally when the masdar suffix *-ni* is following (50d).

(48) *x* > *š*, *xː* > *šː*


2 Phonology

```
(49) g > ž
      b-ug-ul 'remaining' (n-stay-icvb)
        vs. b-už-ib 'remained' (n-stay-pret)
```
	- a. *b-uk-ul* 'gathering' (n-gather-icvb) vs. *b-uč-ib* 'gathered it' (n-gather-pret)
	- b. *b-ikː-a* 'give it!' (n-give.pfv-imp.sg) vs. *b-ičː-ib* 'gave it' (n-give.pfv-pret)
	- c. *er w-erk'-araj* 'in order to look at him' (look m-look.pfv-subj) vs. *er w-erč'-e* 'Look!' (look m-look.pfv-imp.sg)
	- d. *b-ebk'-a* 'death' (n-die.pfv-nmlz) vs. *b-ebč'-ni* 'death' (n-die.pfv-msd)

When the masdar suffix is added the process is optional, at least with some verbs (51) (although it occurs when other suffixes are added). However, with a few verbs such as *er b-ik'ʷ-ni* 'looking' (look -n-say.ipfv-msd) it is ungrammatical.

(51) a. *ubč'-ni*/*ubk'-ni* (die.m.ipfv-msd) < *b-ubk'-* (n-die.ipfv-) b. *b-arč-ni*/*b-ark-ni* (n-find.pfv-msd) < *b-arkː-* (n-find.pfv-)

### **2.6.10 Labialization and delabialization**

There are two instances of labialization of stops triggered by the round vowel *u*. In the first instance, a preceding vowel is lost and the loss is compensated for by labializing the following stop (another example with the same verb has been provided in §2.6.1 above):

(52) *w-i-ha-(w)-ulq-an* > *wihalqʷan* 'the one that goes inside' (m-in-up-(m)-go.ipfv-ptcp) (compare with *b-i-ha-b-ulq-an*)

The second instance represents the combination of the two spatial preverbs *gu-* 'under' and *ha-* 'upwards'. The glottal fricative between the two vowels is lost and the round vowel disappears, leaving the initial stop labialized, that is *gu-ha-* > *gʷa-* (53).

(53) c'a fire gʷa-b-iq'-un from.under.up-n-set.fire.pfv-pret ca-b cop-n '(She) set up a fire.'

Delabialization is a more widespread and predictable process. It occurs when verbs that contain labialized stem consonants take suffixes beginning with the round vowel *u* (i.e. one of the preterite allomorphs *-ub*, *-ur,* or *-un*):

	- b. *kaxʷ-ij* 'kill' > *kax-ub* 'killed' (kill.pfv-pret)
	- c. *ergʷ-ij* 'sieve' > *erg-ur* 'sieved' (sieve.pfv-pret)

### 2.6 Phonological and morphophonological alternations

With nouns delabialization occurs in the formation of the plural. When the plural suffix or the oblique plural suffix is added to nouns that have a vowel *a*/*aˁ* in the root that undergoes vowel mutation *a*/*aˁ* > *u*/*uˁ*, then the mutation is accompanied by delabialization of a stop that precedes or follows the mutated vowel. Furthermore, plural suffixes containing *u* also trigger delabialization of preceding consonants when they are added (55).


Other plural suffixes do not lead to vowel mutation, and thus labialized consonants are preserved, for example:


### **2.6.11 Gemination and degemination**

Gemination is not a common process, whereas degemination is frequent. There is optional gemination in combination with devoicing, which always involves at least one gender affix. This process occurs only with the gender affixes *b* (neuter singular/human plural) and *d* (neuter plural/first and second person plural). The two lax voiced consonants become tense and devoiced when they are preceded or followed by an identical consonant. This can either be the same gender affix or the past tense enclitic *=de*, the attributive plural suffix *-te* or occasionally when a preverb ending in *p* is used in a complex verb (there are no preverbs ending in *d*). Examples are given in (57) and (58). In careful speech the two consonants are pronounced individually, and no gemination and devoicing take place.

	- a. *či-d-d-iχ-un* > *čitːiχun* '(they) tied them' (spr-npl-npl-tie.pfv-pret)
	- b. *le-d=de* > *letːe* 'we were there' (exist-npl=pst) (*le-b=de*)
	- c. *xari-d-te* > *xaritːe* 'the ones down' (down-npl-dd.pl) (*xari-b-te*)
	- a. *gu-b-b-iči-b* > *gupːičib* 'it lost' (sub-n-n-occur.pfv-pret)
	- b. *χːap b-arq'-ib* > *χːapːarq'ib* 'grabbed it' (grab n-do.pfv-pret)

Gemination does not occur when two voiceless consonants follow each other, for example *ħaˁžat-te* (need-dd.pl).

Furthermore, a number of verbal suffixes such as the habitual present suffixes contain geminates, e.g. *-tːe* (-2sg.prs) (§13.1). These suffixes are probably diachronically complex in their morphology, but since they synchronically function as entire morphemes that are not further split up, they are not treated here.

### 2 Phonology

Geminates are regularly degeminated when they end up in syllable-final position, because geminates in syllable-final position are prohibited (see §2.1, §2.3). Therefore, when suffixation leads to resyllabification, then degemination takes place, that is, tense consonants become lax. Voicing is not affected. Within the nominal morphology we find degemination of stops, fricatives, and affricates when the plural suffixes *-be*, *-ne,* and *-me* are added (59).


Similarly, a number of nouns have underlying geminates (stops and fricatives) in the word-final position that are only pronounced as geminates when suffixes that begin with a vowel (e.g. the plural suffixes *-e* and *-upːe*) are attached (60). In those plural nouns the geminates occur in syllable-initial position. By contrast, when the nouns are used in the singular or when suffixes that start with consonants are added (e.g. the ergative suffix *-li*), then the stops and fricatives are degeminated. More examples can be found in §3.2. In the examples in (60) first the plural forms are given and then the singular forms.


Within the verbal system, degemination can only occur when consonant-initial suffixes are added to verbal roots that have geminated consonants. The only relevant suffixes are the masdar suffix *-ni* (or *-ri*) and the locative participle *-na*.

	- b. *ka-b-ičː-ij* (down-n-cut.up.pfv-inf) > *kabič-ni* (masdar)
	- c. *b-arcː-ij* (n-get.tired.pfv-inf) > *barc-ni* (masdar)
	- d. *akːʷ-* (cop.neg) > *akʷ-ni*/*akʷ-ri* (masdar)
	- e. *b-učː-ij* (n-drink.pfv-inf) > *buč-na* (locative participle)

## **Part II**

## **Nominal categegories**

## **3 Nouns**

The grammatical categories of nouns and other nominals in Sanzhi are gender, number and case. There are three genders: masculine, feminine, and neuter. With respect to number, nouns distinguish singular and plural. In addition, there is an associative plural. Sanzhi Dargwa has four grammatical cases, namely absolutive, ergative, dative, and genitive, and many more semantic cases. Most of the latter are spatial cases.

This chapter describes gender (§3.1), number (§3.2), and case (§3.4) as well as the derivation of nouns (§3.5) and the formation of nouns by means of compounding and reduplication (§3.6).

### **3.1 Gender**

Sanzhi has the typical Dargwa gender system of three genders that have a transparent semantic basis: masculine, feminine, and neuter. To the feminine and masculine gender belong only those nouns that denote humans or are perceived as humanoids or similar to humans. This means that gender for humans follows natural gender, and all other nouns are neuter. Gender agreement is a major grammatical trait of East Caucasian languages, including Sanzhi. The combined gender–number agreement affixes are given in Table 3.1. All forms except the zero marking for masculine singular agreement can occur as prefixes, suffixes, and infixes (only with two words). For more information on gender agreement see §20.2.

Table 3.1: Agreement affixes in Sanzhi


Gender is normally not marked on nouns, but there are a few nouns that do carry overt gender markers in word-initial position that seem to go back to gender prefixes. They can be divided into two groups. The first group is kinship terms and the noun 'owner' that differ in their form depending on the gender of the referent (1).

	- b. *ucːiq'ar*, *rucːiq'ar* 'cousin' (male, female)

### 3 Nouns


The second group is nouns that express the gender of the (implicit or explicit) possessor (2). Most of the words of the second group denote body parts. The first noun controls neuter plural agreement since it is morphosyntactically a compound noun (§3.6.2). The second noun controls neuter singular agreement, independently of the agreement prefix used. The third noun controls agreement according to the referent and therefore in accordance with the prefix it has. For instance, *babq'i (χalq')* 'half (of the people)' controls human plural agreement. It thus behaves similar to the nouns in (1).

	- b. *wag*, *rag*,*bag*, *dag* 'middle, waist'
	- c. *wabq'i*, *rabq'i*, *babq'i*, *dabq'i* 'half'

There is another word *daˁʡ* 'face' that clearly contains a frozen gender prefix and also occurs as part of compound verbs, e.g., in *b-aˁʡ-či-aʁ-ib* 'direct', or as the root of spatial adverbs such as *b-aˁʡ-gubal* 'upside-down'. In principle, it is also possible to form the masculine singular, feminine singular, and neuter singular variants *waˁʡ*, *raˁʡ* and *baˁʡ* that seem to have the meaning 'face, muzzle' (of a man, of a woman) and 'muzzle of an animal, wall, facade', but they are not used in natural speech and speakers have trouble to find a context in which they could occur.<sup>1</sup>

Nichols (2007) calls the overt marking of gender on nouns in (1) and (2) "head gender". Nichols (2007) argues that synchronically, the initial segments of these words cannot simply be considered to be gender agreement prefixes because for most nouns head gender does not change in the plural, in contrast to agreement affixes on verbs and other parts of speech. For instance, the plural of *ucːi* 'brother' and *rucːi* 'sister' is *ucbe* and *rucbe* respectively, not *bucbe*. For the nouns of the first group in (1), the head gender is determined by the natural gender of the referent and not controlled by another nominal. Nichols also writes that only few such nouns are likely of verbal origin. As for the Sanzhi words given in (1) and (2), I am unable to say anything about their origin.

### **3.2 Number**

Most nouns in Sanzhi can be marked for plural by means of a suffix. The singular has no special marking. Plural suffixes can be divided into three groups according to their frequency and productivity:

<sup>1</sup>There is another noun *baˁʡ* 'leaf, sheet of paper, page' that is arguable a cognate of *baˁʡ* 'muzzle, wall, facade'. The two nouns can be distinguished through their plural marking: *baˁʡ* > *buˁʡre* 'leaves, pages' vs. *baˁʡ* > *baˁʡuˁrme*, *baˁʡme* 'muzzles, walls, facades'.

3.2 Number


The first group is the only one that can be used with recent loan words from Russian (Russian loans are indicated in the lists in (4) to (14)). The last group is restricted to one or two lexical items. Many of the nouns undergo morphophonological processes before the plural suffix is added. Plural suffixes containing the vowel /u/ have allophones with the pharyngealized vowel *uˁ*, e.g. *-uˁpːe*.

### **3.2.1 Frequent and productive plural suffixes**

In this and the following section, I provide examples for every suffix illustrating at the same time the morphophonological processes. The morphophonological processes that are applied when forming the plural of nouns are summarized in §3.2.3.

**3.2.1.1** *-e*

	- a. *t'ult'* > *t'ult'e* 'bread'
	- b. *sːurrat* > *sːurrate* 'picture'
	- c. *unc* > *unce* 'ox'
	- d. *qːačuʁ* > *qːačuʁe* 'bandit'
	- e. *χurejg* > *χurejge* 'food'
	- f. *q'ampit'* > *q'ampit'e* 'chocolate' (Russian loan)
	- g. *student* > *studente* 'student' (Russian loan)
	- h. *praznik'* > *praznik'e* 'holiday' (Russian loan)
	- a. *juldaš* > *juldašːe* 'friend'
	- b. *baliq* > *baliqːe* 'fish'
	- c. *ʡuˁrus* > *ʡuˁrusːe* 'Russian'
	- d. *ħaˁšuk* > *ħaˁšukːe* 'pot'
	- e. *miriqʷ* > *miriqːʷe* 'worm'
	- f. *t'up* > *t'upːe* 'finger'
	- g. *tusnaq* > *tusnaqːe* 'prison'

### 3 Nouns

	- a. *halmaʁ* > *halmuʁe* 'fiancée'
	- b. *kːazat* > *kːazute* 'newspaper' (Russian loan)
	- c. *qːabaʁ* > *qːabuʁe* 'pumpkin'
	- d. *q'ʷaˁl* > *q'uˁle* 'cow'
	- a. *murgul* > *murgle* 'man'
	- b. *k'apːur* > *k'apre* 'leaf'
	- c. *amχa* > *umχe* 'donkey'

### **3.2.1.2** *-te*

Most of the nouns taking this suffix are disyllabic loan words ending in a resonant. This suffix is also used for the plural form of long adjectives and, more generally, of predicates (§9.6.1).

	- a. *šːišːim* > *šːišːimte* 'worry'
	- b. *tuχtur* > *tuχturte* 'doctor'
	- c. *qːanaw* > *qːanawte* 'ditch'
	- d. *ħaˁkim* > *ħaˁkimte* 'ruler'
	- e. *pawur* > *pawurte* 'cook' (Russian loan)
	- f. *mašin* > *mašinte* 'car' (Russian loan)
	- g. *bazar* > *bazarte* 'market'
	- h. *salam* > *salamte* 'greeting'
	- i. *kːapan* > *kːapante* 'shroud'
	- j. *kep* > *kepte* 'drinking'
	- k. *ukul* > *ukulte* 'injection'
	- l. *ħaˁjwan* > *ħaˁjwante* 'animal, livestock'
	- m. *dušman* > *dušmante* 'enemy'
	- n. *mahar* > *maharte* 'marriage'
	- a. *χabar* > *χaburte* 'story, news'
	- b. *šajt'an* > *šajt'unte* 'devil'
	- c. *darman* > *darmunte* 'medicine'
	- d. *bajram* > *bajrumte* 'holiday'
	- e. *qːajtar* > *qːajturte* 'club'
	- f. *q'uˁšːem* > *q'uˁšːumte* 'handcuff'

(10) other:

*admi* > *adimte* 'person, human being, man'

### **3.2.1.3** *-be*

	- a. *ul* > *ulbe* 'eye'
	- b. *milic'a* > *milic'abe* 'police(man)' (Russian loan)
	- c. *pːalaženija* > *pːalaženijabe* 'situation' (Russian loan)
	- d. *xːun* > *xːunbe* 'road, way'
	- e. *ʡuˁnru* > *ʡuˁnrube* 'life'

### (12) vowel mutation:

	- a. *cula* > *culbe* 'tooth'
	- b. *rursːi* > *rursbe* 'girl, daughter'
	- c. *hinci* > *hincbe* 'apple'
	- d. *rucːi* > *rucbe* 'sister'
	- e. *cːacːi* > *cːacbe* 'thorn'

### **3.2.1.4** *-me*

	- a. *pikru* > *pikrume* 'thought'
	- b. *buh* > *buhme* 'bundle'
	- c. *dus* > *dusme* 'year'
	- d. *t'uˁ* > *t'uˁme* 'leg'
	- e. *irk* > *irkme* 'threshing board'
	- f. *peč* > *pečme* 'oven' (Russian loan)
	- g. *šalakbluk* > *šalakblukme* 'building block made from concrete' (Russian loan)
	- a. *kːurtːi* > *kːurtme* 'dress, shirt'
	- b. *kːalkːi* > *kːalkme* 'tree'
	- c. *beretːa* > *beretme* 'ax'

3 Nouns


### **3.2.2 Modestly frequent plural suffixes**

The following plural suffixes are not used with recent borrowings.

### **3.2.2.1** *-re*

Many monosyllabic nouns ending in a consonant take the plural suffix *-re* after vowel mutation. Disyllabic undergo vowel deletion before the suffix is attached.

	- a. *qːap* > *qːupre* 'sack'
	- b. *maχ* > *muχre* 'wheelbarrow' (used to carry hay)
	- c. *t'at'* > *t'ut're* 'thread'
	- d. *k'at'* > *k'ut're* 'drop'
	- e. *lac* > *lucre* 'fence, wall'
	- f. *baˁʡ* > *buˁʡre* 'leaf, page'
	- a. *migʷi* > *migʷre* 'large basket'
	- b. *ʡaˁmi* > *ʡaˁmre* 'window'
	- c. *ʁuma* > *ʁumre* 'block'

*xːunul* > *xːunre* 'woman'

### **3.2.2.2** *-ne*

The plural suffix *-ne* is mostly found with disyllabic nouns ending in *-a* or very occasionally in *-u/-uˁ*) after deleting the final vowel (and degemination of the previous consonant if it is a tense consonant). There are also a few monosyllabic nouns that make use of the suffix.

	- a. *ʁʷab* > *ʁʷabne* 'ploughshare'
	- b. *bar* > *barne* 'day'

<sup>2</sup>This example slightly differs from all the others because the vowel, which is deleted, occurs in the final syllable, but it is followed by consonant and thus does not represent the word-final segment.

	- a. *sːika* > *sːikne* 'bear'
	- b. *ajarq'a* > *ajarq'ne* 'hunter'
	- c. *žaq'a* > *žaq'ne* 'boar, pig'
	- d. *šuša* > *šušne* 'bottle'
	- e. *likːa* > *likne* 'bone'
	- f. *čatːa* > *čatne* 'cover'
	- g. *qːarqːa* > *qːarqne* 'stone'
	- h. *c'eltːa* > *c'eltne* 'gravestone'
	- i. *ečːa* > *ečne* 'she-goat'
	- j. *durħuˁ* > *durħne* 'boy, son'

### **3.2.2.3** *-upːe*

This suffix occurs with a handful of nouns and by means of simple suffixation, final vowel deletion, or surface appearance of underlying geminate consonant (in the last noun).

	- b. *daˁqaˁ* > *daˁquˁpːe* 'wound'
	- c. *muza* > *muzupːe* 'corner'
	- d. *ʡaˁq'lu* > *ʡaˁq'lupːe* 'mind'
	- e. *qːuˁnq* > *qːuˁnqːuˁpːe* 'nose'

### **3.2.2.4** *-urbe*

The suffix *-urbe* is mostly attested with disyllabic nouns ending in *-a* or *-i*. It can be simply added to nouns ending in consonants; otherwise final vowel deletion applies.

	- b. *dahag* > *dahagurbe* 'slope'
	- c. *uncːa* > *uncːurbe* 'door'
	- d. *duʁa* > *duʁurbe* 'hayloft'
	- e. *mašːi* > *mašːurbe* 'farmstead'
	- f. *ʡaˁči* > *ʡaˁčurbe* 'work'

### **3.2.2.5 Other suffixes**

There are a few rare suffixes that only occur with a very small number of nouns. Some of these nouns also make use of alternative, more common plural markers.

3 Nouns

```
(23) -urme:
```

```
(24) -ude:
```
*χːʷe* > *χːude* 'dog'

(25) *-une*:

*ʡaˁdat* > *ʡaˁdatune* 'habit, tradition'

(26) *-(u)bne*:

*malla* > *mallubne* 'mullah'

### **3.2.3 Morphophonological rules and other restrictions**

The distribution of the plural markers is basically lexical. For certain derived nouns plural marking is predictable (e.g. agent nouns ending in *-či* take the plural suffix *-be*, abstract nouns with the suffix *-dex* take the plural suffix *-e* and undergo gemination of the suffixfinal consonant). There are many nouns that can attach more than one plural suffix, and both in texts and in elicitations one frequently comes across variation between speakers as well as within the speech of individuals. Examples are provided in (27).

	- b. *ʡaˁdat* > *ʡaˁdaturme/ʡaˁdatme/ʡaˁdate/ʡaˁdatune* 'habit, tradition'
	- c. *dard* > *dardane/dardme* 'sorrow'

Plural suffixation is accompanied by a number of morphophonological processes:


With a few nouns, the last vowel shifts to *u* or *uˁ*, and, as a consequence, the preceding labialized consonant is automatically delabialized as in *mikʷa* > *mikupːe* 'fingernail', *χːʷe* > *χːude* 'dog', and *q'ʷaˁl* > *q'uˁle* 'cow'. However, in most cases, no vowel shift takes place and thus labialized consonants are not delabialized, and only the above-mentioned 3.3 Gender–number mismatches and exceptions

morphophonological processes take place (vowel deletion, degemination), for example *mikʷ* > *mikʷbe* 'oak', *gʷagʷa* > *gʷagʷne* 'flower', *kaˁχːʷi* > *kaˁχʷne* 'dustpan', and *žilixʷa* > *žilixʷme* 'saddle'. See §2.6 for more information on the morphophonological processes.

When case suffixes are added to nouns overtly marked for plural, then the final vowel of the plural suffix changes from *-e* to *-a* (see §3.4 for examples).

### **3.2.4 The associative plural**

In addition to the normal plural, Sanzhi has an associative plural formed with the suffix *-qal* that probably originates from the noun *qal* 'house'. The associative plural is only used with nominals that have specific reference:


Some kinship terms (e.g. 'uncle', 'sister') and personal names can also form the plural by means of regular plural suffixes, but there is a clear difference in meaning:

	- b. vs. *Muʔminat-qal* 'Muʔminat and the people associated with her'
	- c. *acːi-be* 'uncles'
	- d. vs. *acːi-qal* 'a specific uncle and his relatives and associates'

Cases are directly suffixed to the associative plural marker.

### **3.3 Gender–number mismatches and exceptions**

There a couple of nouns whose behavior deviates from the majority as described in the previous sections. This section provides examples of the different groups of divergent nouns. There are three important parameters along which the divergent nouns can be grouped:


### 3 Nouns

The first clearly identifiable group consists of nouns that denote liquids and other substances composed of small or minimal parts such as grains and dust-like materials. These nouns normally control *d*-agreement, but *b*-agreement is possible if the noun is interpreted as denoting a specific quantity (e.g. a bottle or a glass in case of liquids; one grain or one ear in case of sand or cereals). The specific quantity reading occurs frequently with some nouns (e.g. vodka, flour) and is therefore easier to obtain in elicitation. The nouns do not have a morphological plural. Examples are:

(29) liquids

*čaˁʁir* 'wine', *nejg* 'milk', *kːamput'* 'homemade juice', *čaˁj* 'tea', *beʔe* 'blood', *nerʁ* 'soup', *ʡaˁraq'i* 'vodka', *hin* 'water'

(30) cereals, etc.

*ač'i* 'wheat', *sːusːul* 'rye', *ʡaˁjlač'i* 'corn', *birinž* 'rice' (with *b*-agreement: a sack or single grain), *t'ut'i* 'grapes' (with *b*-agreement: a single grape)

(31) other substances

*qːum* 'sand', *cːe* 'salt' (with *b*-agreement: a specific quantity or one grain), *pisuk'* 'caster sugar' (with *b*-agreement: one bowl or sack), *bet'u* 'flour' (with *b*-agreement: one sack)

The second group contains mass nouns that control only *b*-agreement and lack a morphological plural, for instance *bergʷa* 'smoke', *erza* 'dew', *duˁħi* 'snow', and *baˁqaˁla* 'butter'. They can be reasonably treated as controlling neuter singular agreement.

The third group consists only of one noun *χalq'* 'people(s)', which controls *b*-agreement and lacks a morphological plural. Because of its semantics it is classified as human plural.

The fourth group is composed of mass nouns that control only *d*-agreement and also lack a morphological plural: *mura* 'hay', *pːala* 'wool', *nekʷ* 'straw', *qʷesːa* 'ashes', *dalga* 'tool, product, detail', *wajaˁħ* 'thing', and *šuˁt'a* 'saliva, spittle'.

The fifth group consists of nouns that lack a singular form and only occur with what seems to be a frozen plural suffix. These nouns control plural agreement (*d*-agreement), for example *mecːe* 'stinging nettle', *t'alaħne* 'dishes', *cːurbe* 'heaven', and *susme* 'throat'.

The last four groups are given in (32–35). It is not always possible to clearly identify the mass noun reading. For all words in (32) and (33) that have the label 'many' in parenthesis after the English translation, the label 'many' refers to the normal (collective or distributive) plural reading, e.g. many individual apricots.

(32) *b-* and *d-*agreement with plural meaning and/or mass noun reading; no morphological plural:


### 3.3 Gender–number mismatches and exceptions


<sup>3</sup>This use not very common since there is a noun *q'ačme* with the mass noun reading of 'hair'.

### 3 Nouns

(34) *b-* and *d-*agreement with no clearly distinct meanings (normally including mass noun interpretation); no morphological plural:


(35) *b-* and *d-*agreement with no clearly distinct meanings; with morphological plural:


### **3.4 Case**

Sanzhi Dargwa has four grammatical cases and 19 core semantic cases as well as one minor directional suffix. The grammatical cases and the comitative are given in Table 3.2. The 18 core spatial cases are provided in Table 3.3. The essive is shown in the neuter singular/human plural form with the gender–number suffix *-b*. Illustrative partial paradigms of a few nouns can be found in Tables 3.4–3.5.

Table 3.2: Grammatical cases


Case suffixation is (almost) completely regular and predictable. Like in many other East Caucasian languages, including other Dargwa varieties, case suffixes in Sanzhi for the most part do not directly attach to the nominal root, but are preceded by a so-called oblique marker. For nouns in the singular, the oblique marker is identical to the ergative suffix *-li* and will be glossed with obl. Demonstrative pronouns in the singular have *-i* as the oblique marker; all nouns and demonstrative pronouns in the plural have *-a*.


Table 3.3: Spatial cases

Table 3.4: Partial paradigms of two nouns (vowel-final stems)


Table 3.5: Partial paradigms of two nouns (consonant-final stems)


### 3 Nouns

There are a few differences between nouns ending in a vowel and nouns ending in a consonant with respect to the distribution and the usage frequency of oblique markers in the singular. With nouns ending in a vowel all case suffixes are mostly directly added to the nominal stem (Table 3.4), but occasionally the oblique marker *-l(i)* precedes suffixes of semantic cases, as in *bušːukala-l-cːella* broom-obl-comit 'with the broom', *q'aca-l(i) šːu* 'to the goat' (he.goat-obl-ad). With nouns ending in a consonant (Table 3.5), the oblique marker obligatorily precedes the dative and the comitative, and is normally also used before all spatial cases except for the loc-series with *-le*. If nouns are marked for the plural, then overt case suffixes are always added to the plural oblique marker *-a*, never directly to the plural stem.

The suffixes for the ergative and for the genitive can and most frequently do assimilate after *n* and *r* (and *s* in the noun *dus* 'year') to *-ri*/*-ni*/*-si* (ergative) and *-ra*/*-na* (genitive). The genitive of nominals in the plural is frequently realized as *-lla* instead of *-la*, but this phenomenon seems to be at least partially subject to variation. It is hard to notice in audio recordings of natural texts and speakers are not always aware of it. Therefore, it will mostly not be acknowledged in the examples.

### **3.4.1 Functions of grammatical cases**

### **3.4.1.1 Absolutive**

The absolutive case is zero-marked and not indicated in the glosses. It occurs in the following contexts:

	- (36) di-la 1sg-gen nuˁq-be arm-pl ʡaˁbħ-ib get.tired.pfv-pret ca<d>i cop<npl> 'My arms got tired.'
	- (37) du-l 1sg-erg ka-d-iqː-an=da down-npl-carry.ipfv-ptcp=1 qix-be nut-pl 'I will bring nuts.'
	- (38) itːa-j those.obl-dat ʡuˁrus Russian ʁaj language d-alχ-ul npl-know.ipfv-icvb akːʷ-i=q'al cop.neg-hab.pst=mod 'They did not know the Russian language.'

3.4 Case

	- (39) du 1sg baliqː-a-l fish-obl.pl-erg ∅-uk-un=da m-eat.ipfv-icvb=1 'I (masc.) eat fish.'
	- (40) χalq' people kːuš-le=de hungry-advz=pst 'The people were hungry.'
	- (41) het that durħuˁ boy aždaha monster ca-w cop-m ' That boy is a monster.' (E)
	- (42) d-alt-u npl-let.ipfv-prs.3 ʁajal twenty bari day '(They) leave (them) for 20 days.'
	- (43) ʡaˁbal three dus year kelg-un=da remain.pfv-pret=1 'I stayed there for three years.'
	- (44) ce what r-ik'-utːe, f-say.ipfv-2sg.prs aba? mother 'What do you say, mother?'

### **3.4.1.2 Ergative**

The ergative suffix is *-li* (allomorphs *-ni*, *-ri* after *n* and *r* respectively, and *-l*, which can only be added to vowels). The ergative occurs in the following contexts:

	- (45) žaq'-ne boar-pl a-d-uk-i neg-npl-eat.ipfv-hab.pst nušːa-l, 1pl-erg akːʷ-i=w? cop.neg-hab.pst=q 'We did not eat boars, right?'
	- (46) hin-ni water-erg heχ dem.down lus around b-ik'-aq-u n-move.ipfv-caus-prs 'The water turns this around.'

### 3 Nouns

	- (47) kːiši-l hunger-erg imc'a-l additional-advz χalq' people b-ebč'-ib hpl-die.pfv-pret 'Many people died of hunger.'
	- (48) du-l 1sg-erg ka-b-irčː-ul=da down-n-cut.ipfv-icvb=1 t'ult' bread dis-li knife-erg / / dis-li-cːella knife-obl-comit 'I cut the bread with a knife.' (E)
	- (49) a. il-i-la that-obl-gen azbar yard b-ic'-ib n-fill.pfv-pret ca-b cop-n ʡuˁrʡ-aˁ-l, chicken-obl.pl-erg qːuz-r-a-l, goose-pl-obl-erg k'urk'ur-t-a-l turkey.cock-pl-obl-erg 'His yard was filled with chickens, geese, and turkeys.' b. b-ic'-ib n-fill.pfv-pret hel that buq'a=ra hut=add sːusːul-li rye-erg

'And (they) also filled the hut with rye.'

	- (50) il that ħaˁjwan animal tuχtur-ri doctor-erg kelg-un remain.pfv-pret 'He was a/the veterinarian.'

### **3.4.1.3 Genitive**

The genitive suffix is *-la* (allomorphs *-na*, *-ra* after *n* and *r*, and allomorph *-lla* with many nouns and pronouns marked for plural, and in some other contexts). It is used in the following contexts:

### 3.4 Case

	- (51) nišːa-la 1pl-gen sungli-la Sanzhi.person.obl-gen šːi-l-cːe-b village-obl-in-n 'in our village of Sanzhi' (lit. 'in our village of the Sanzhi people')
	- (52) hel=ʁuna that=eq admi person ʡaˁʁuni-l needed-advz ca-w cop-m nišːi-j 1pl-dat ca one χːula-ce big-dd.sg hunar-ra strength-gen w-ah m-owner

' We need such a man, one with huge strength.'

	- (57) qal-la house-gen sala in.front 'in front of the house'
	- (58) qːarqːa-la stone-gen hila-b behind-n 'behind the stone'

### 3 Nouns

	- (59) hin-na water-gen b-erčː-a! n-drink.pfv-imp 'Drink (some) water!' (E)
	- (60) b-ic'-ib-le, n-fill.pfv-pret-cvb hin-na water-gen b-erčː-ib n-drink.pfv-pret ca-b cop-n 'Having filled (the glass) with water (he) drank it.'
	- (61) *qalla + b-arq'-ij* 'marry off'; *qalla + ka-b-at-ij* 'marry off' (house.gen + downhpl-let.pfv-inf)

a. ca one qal-la house-gen r-arq'-ib f-do.pfv-pret cin-na refl.sg-gen rursːi girl '(She) already married off one daughter.'

<sup>4</sup> In the constructions in (61a), (61b), (62) and (63) the direct object (e.g. *rursːi*) can perhaps be interpreted as the possessum and the genitive noun (*qal-la*) as possessor such that we would deal with a genitive phrase. The phrase would, however, have the reverse case distribution of normal genitive phrases. The possessor is normally a human referent and the possessum can be inanimate, but in the four examples it is the other way around. Furthermore, in genitive phrases possessor and possessed usually occur next to each other in the order genitive + noun, which is also not the case in these examples, but occasionally other orders are possible (121) (§21.1.2). In sum, an analysis in which the direct objects and the nouns in the genitive syntactically form genitive phrases needs to be rejected.

3.4 Case

	- (64) ala 2sg.gen r-uˁq'-aˁn! f-go-imp '(You) yourself (fem.) go away!' (E)

### **3.4.1.4 Dative**

The dative suffix is *-j*. The dative occurs in the following contexts:

	- (65) ce what ħaˁžat-le, need-advz ∅-ik'ʷ-ar, m-say.ipfv-prs at 2sg.dat betsat here.there w-ič-itːaj m-lead.ipfv-subj.2 durħuˁ boy 'What need is there, he says, for you to bring the boy here and there.'
	- (66) dam 1sg.dat simi anger d-ulq-u npl-direct.ipfv-prs 'I am angry.' (lit. 'anger directs to me') (E)
	- (67) χatːaj grandfather at 2sg.dat b-uχːar-re=w? n-be.cold-cvb=q 'Grandfather, are you cold?'

### 3 Nouns

	- (68) dam 1sg.dat / / di-cːe 1sg.obl-in b-urs-ib, n-say.pfv-pret … '(They) said to me, …'
	- (69) b-irq'-an=da n-do.ipfv-ptcp=1 rursːi-j girl-dat ʡaˁħ good meq wedding '(I) will make a big wedding for my daughter.'
	- (70) gu-lik'-an down-listen.ipfv-ptcp ʡaˁħ-le good-advz ʡaˁt'-n-a-la frog-pl-obl-gen t'ama-j sound-dat 'listening carefully to the sounds of the frogs'
	- (71) Murad-li-j Murad-obl-dat χʷal-le big-advz urk'ec'i pity či-d-ulq-u spr-npl-direct.ipfv-prs Madina-j Madina-dat 'Murad feels very sorry for Madina.' (E)
	- (72) na now il-i-j that-obl-dat w-iχči m-believe ag-ur-re go.pfv-pret-cvb durħuˁ boy sa-r-∅-uq-un in.front-abl-m-go.pfv-pret ca-w cop-m 'Now the boy believed it (= the mouse) and ran away.'
	- (73) žergʷ-ne wasp-pl hitːi after d-uq-un-ne npl-go.pfv-pret-cvb ca-d cop-npl durħ-a-j=ra boy-obl.pl-dat=add kac'i-j=ra puppy-dat=add 'The wasps flew after the boy and the puppy'
	- (74) iž=ra this=add xunul-li-j woman-obl-dat miši-l similar-advz ca-r cop-f hel that 'This one also (= the person on a picture) is similar to a woman.'
	- (75) hel that qːaq Kak ħuˁsen-na Hussein-gen ʡaˁħ-dexː-li-j good-nmlz-obl-dat hel=ra that=add kelg-un-il=de remain.pfv-pret-ref=pst 'Because of the benevolence of Kak Hussein that (man) also remained (alive).'

3.4 Case

	- (77) palata-li-j ward-obl-dat ca-b cop-n har every bari-j day-dat k'ʷel two azir thousand '(The price) for a ward is 2000 (rubles) per day.'
	- (78) har every dusːi-j year.obl-dat k'ʷel two q'ap'a hat isː-i buy.ipfv-hab.pst 'Every year he (usually) bought two hats.'
	- (79) du-l 1sg-erg b-irq'-id n-do.ipfv-1.prs dars homework ca one saˁʡaˁt-li-j hour-obl-dat 'I will do the homework in one hour.' (E)
	- (80) ixtːu-b there-n čːaˁʡaˁl-li-j morning-obl-dat sːaˁʡaˁt hour kːaʔal-li-j eight-obl-dat w-iχʷ-ij m-be.pfv-inf ʡaˁʁuni-l needed-advz ca-w cop-m hel that admi person 'That person needs to be there in the morning at 8 o'clock.'
	- (81) qːuruš-li-j ruble-obl-dat b-ic-ib n-sell.pfv-pret ca-b cop-n hel that amχa donkey '(He) sold the donkey for three rubles.'
	- (82) čaˁkʷa bird kːalkːi-le tree-loc / / kːalkːi-j tree-dat či-ka-b-iž-ib spr-down-n-be.pfv-pret ca-b cop-n 'The bird sat down on the tree.'

### **3.4.2 Functions of semantic cases**

Apart from the comitative all semantic cases have a basic spatial meaning. Table 3.3 provides the core spatial cases. As in most other Caucasian languages the spatial cases are formally and functionally rather transparent and organized along two dimensions: location and direction (movement). There are six suffixes that express different ways of locating an item with respect to a reference point:

### 3 Nouns


There is a semantic distinction between animate reference points (normally used together with the ad-series) and inanimate reference points (usually marked with the locseries). Furthermore, not all conceivable spatial constellations are covered by the location suffixes. For instance, meanings such as 'near' and 'above' can only expressed by means of postpositions (Chapter 8).

Furthermore, there is a three-way distinction in terms of direction (movement):


The direction markers can be directly suffixed to spatial postpositions/adverbs and some other nominals that have inherent locational meaning (e.g. place names). With all other items, direction markers only occur in combination with the location markers. In addition to the core spatial cases given in Table 3.3 there is one minor spatial case whose use is somewhat restricted, the directional *-gm-a* (§3.4.2.8).

The lative is formally unmarked and expresses direction and movement to a reference point. The essive is formally marked through gender agreement suffixes that agree with the item whose location is expressed. This is normally the absolutive argument, i.e. agreement of the essive adjunct confirms to the general rules of gender agreement. The absolutive argument does not need to be overtly expressed in order to control the agreement, but can be left implicit (83).

(83) qili home sa-∅-jʁ-ib=qːella hither-m-come.pfv-pret=when xːunul-la woman-gen bek'-le-w head-loc-m či-w on-m w-alkː-un-ne m-importune-pret-cvb k-erg-ul=de down-sit.ipfv-icvb=pst

'(After drinking alcohol the husband,) when he came come, he pestered his wife.' (lit. 'sat on the wife's head')

3.4 Case

However, as example (84) shows, it is also possible for other prominent arguments to control gender agreement on an essive adjunct. In this example, it is the implicit agent, the first person pronoun in the ergative case referring to the masculine speaker that controls agreement on the noun in the in-essive (see §20.2.4 for more details).

(84) hež this sawχuz-li-cːe-w sovkhoz-obl-in-m aʁʷ-c'al four-ten dus year ʡaˁči work b-arq'-ib=da n-do.pfv-pret=1 'In the sovkhoz I (masc.) worked for 40 years.'

The ablative has two meanings, 'from' and 'through'/'along'. It is most frequently expressed by the suffix -*r*, but -*rka* is also possible with apparently no difference in meaning. The latter suffix is morphologically complex consisting of -*r* and -*ka* and less frequently used than simply -*r*. Diachronically, -*ka* might go back to an elevation marker *ka* 'down' (and thus be related to the elevation preverb *ka*- 'down', see §11.6.2).<sup>5</sup>

The spatial cases are functionally and partially also formally close to spatial adverbs (§7.1.2) and postpositions (§8.1) and can be used alone or together with them. Furthermore, there are semantic and formal resemblances with spatial preverbs (§11.6). Spatial cases are also used for non-spatial purposes, e.g. as part of valency frames, in certain constructions such as comparison or to express non-canonical agent constructions. In the following, spatial and non-spatial functions will be described in more detail. Microtoponyms, some other place names (§10) and spatial adverbs (§7.1) diverge from ordinary common nouns when inflected for spatial cases. In a nutshell, they have an inherent locational meaning and are only inflected for direction (lative, essive, and ablative).

### **3.4.2.1 Comitative**

The suffix of the comitative is -*cːella*. Diachronically it is probably complex consisting of the in-lative -*cːe* and the genitive -*(l)la*. It is used with nominals having animate referents in the comitative function (85) as well as with inanimate nouns in the instrumental function (86), (87) and to express manner (88) or experiencers (89) as well as in other contexts roughly corresponding to the use of English *with* (90).


<sup>5</sup>Tanti Dargwa, a relatively closely related variety, has four orientation markers, among them -*ka*, that are only suffixed to nominals inflected for the lative or the ablative. (see Sumbatova & Lander 2014: 69–70 and Forker 2019a).

### 3 Nouns


'Inside the pot there are 2, 3, 4 things like this with holes on the lower side.'

### **3.4.2.2 loc-lative -***le***/-***ja***/-***a***, loc-essive -***le***-***b***/-***ja***-***b***/-***a***-***b* **and loc-ablative -***le***-***r***/-***ja***-***r***/-***a***-r**

The loc-series is together with the in-series (§3.4.2.4) by far the most frequently used series among the spatial cases. It can be broadly described as a general location marker that expresses the most common location of a figure with respect to the ground. Both its formal make-up and its functions are rather complex and deserve a future study. In this section, I can only provide a sketch of its formation and its meaning. Formally, the loc-series is very heterogeneous (in contrast to all other spatial case suffixes). It is either expressed by dedicated suffixes *-a* and *-ja* or by a change of the final vowel *i > e*. The vowel change almost exclusively affects nouns that take *-li* as their ergative suffix and oblique stem marker, and thus we get *-li* > *-le*. However, not all nouns that have the ergative/oblique suffix *-li* undergo the vowel change, but some of those nouns take a suffix as loc-series marker. The occurrences of the allomorphs (suffixes or vowel change) can only partially be predicted. I will first give the usage constraints for each marker, describe its functions by means of examples and in the end compare it to similar markers from other Dargwa varieties. In addition to the relatively regular ways of forming the loc-series by means of the just listed allomorphs, there is a special class of nouns that has inherent locational meaning and can be said to semantically express the loc-series as well although synchronically no case suffix can be identified. This class consists of native place names (mostly names of villages and microtoponyms, Chapter 10).

The default way of forming the loc-series for nouns in the singular is the vowel change of the ergative/oblique suffix *-li* > *-le*. The combinations *\*-li-le* (-obl-loc) or *\*-l-le* (with a deletion of the vowel of the ergative/oblique suffix) are ungrammatical. The suffix *-le* is used after consonants and vowels. There is one noun, *neqːi* 'cave', which has *neqːe* as the loc-lative (in addition to the regular form *neqːi-le*), i.e. we have again the vowel change *i > e*.

3.4 Case

The suffix -*ja* is only used after the vowels *a*, *i* and *u*. It occurs with most nouns in the plural, personal and demonstrative pronouns, plural reflexive pronouns as well as with very few other nouns in the singular, e.g. *qu-ja* (field-loc), *aba-ja* (mother-loc).

The suffix *-a* is used with a number of nouns of which at least some make use of *-li* as ergative/oblique suffix. Examples of such nouns are *ħaˁšak* 'pot' (90) (comitative *ħaˁšakli-cːella*), *mistːik'* 'mosque', *qːatːa* 'canyon', *musːa* 'place', *daˁrqʷ* 'barn, cattle-shed', *šːi* 'village'. With a few of these nouns the suffix is simply added to the noun, e.g. *mistːik' a* 'to the mosque' and *ħaˁšuk-a* 'into the pot'. With those nouns that have stem-final *a*, the locative case differs from the base stem in the pitch accent that switches to the final vowel, e.g. *musːá* 'place/to the place' (92).<sup>6</sup> Thus, we can assume that *a + a > aː > á*. The noun 'village' has the special locative form *šːa* '(in)to the village'. This form is not the oblique stem because it does not serve as the base form for the formation of other cases. Furthermore, the more common way of saying 'in the village' is to use the in-essive (51). The suffix *-a* is also attested for some plural nouns that make use of -*ne* as the plural suffix, e.g. *mus-n-a* (place-pl-obl.loc) 'to the places', *kis-n-a-b* (pocketpl-obl.loc-n) 'in the pockets', *buruš-n-a-r* (mattress-pl-obl.loc-f) 'on the mattresses'. These examples can perhaps be analyzed as undergoing a vowel change *e > a* for the formation of the loc-series.

The meaning of the loc-series is rather broad. It has a basic general spatial and directional meaning indicating movement to a goal, static location at a reference point and movement away from a reference point. Usually the location is the most typical location. The reference points can be places (92), place names, villages, cities, buildings, institutions (91), body parts, vehicles and other means of transport, containers (94), and so on. The loc-series translates into English as 'to, in, on'. Its meaning includes vertical location, e.g. on a wall (93), and also location inside a reference point (94), (95). Note that instead of the loc-series it is possible to use the in-series in examples (93–95) with no difference in meaning (§3.4.2.4). However, it seems that with certain locations there are conventionalized uses of the one or the other suffix. For instance, with names of settlements the loc-series occurs (96), whereas with the noun *šahar* 'town' the in-series form *šahar-ri-cːe* is used. With *ħaˁšak* 'pot' the loc-series is clearly preferred (95), but the loc-series form *ħaˁšak-li-cːe* is also attested (116). Further research is needed in order to arrive at a more detailed picture about the semantic similarities and differences between these two spatial case series.


<sup>6</sup>This refers only to the nominals that take the suffix *-a*. It is not the case that every nominal ending in *a* takes the suffix *-a*, e.g. *ʡaˁrmija-le* 'in the army'.

### 3 Nouns


When used together with the postposition *či-b*, the loc-series can also express the meaning 'above' (see §8.1.7 on postpositions for an example).

When the locative suffix is followed by the ablative case, the meaning is 'from, through' (96), (97).


In the basic meaning, the loc-series is only used with inanimate nouns and can be opposed to the ad-series (§3.4.2.3), which is used with animate nouns to express the same general meaning. Thus, compare (98) and (106):

(98) du 1sg wabše at.all uškul-le school-loc ag-ur-il go.pfv-pret-ref akːʷa-di cop.neg-1 'I did not go to school at all.'

In (99), the participle bearing the loc-ablative refers to a picture showing people who drink and the speaker is asked to begin his story with this picture.

(99) hel-tːi that-pl b-učː-an-t-a-ja-rka n-drink.ipfv-ptcp-pl-obl-loc-abl w-aʔ-∅-ač'-e m-begin-m-come.pfv-imp caj-na! one-time 'First begin with these who drink!' (said to a man)

The loc-essive can also be used for the expression of metaphorical location and direction, e.g. *di-la ʡuˁnru-le-b* (1sg-gen life-loc-n) 'in my life', and for a number of constructions denoting feelings and emotions that are located in body parts or in persons (100), (101).

3.4 Case


The loc-series can also be used with animate nouns. In this case the spatial meaning is 'on, onto' and thus more specific than when used with inanimate nouns:


Occasionally, one can find expressions for points in time marked by the loc-essive, e.g. *sːaˁʡaˁt kːaʔal-le-b* (hour eight-loc-n) 'at eight o'clock'. However, other cases such as the dative are more common in this function. The loc-ablative is regularly used in phrases with the meaning 'after (time)', e.g. *k'ʷel ʡaˁbal minut'-le-r* (two three minute-loc-abl) 'after two, three minutes'.

Finally, the loc-ablative occurs in comparative constructions (104) marking the standard of comparison (§30.1):

(104) ala 2sg.gen qːuʁa-l beautiful-advz arg-ul go.ipfv-icvb ca-b cop-hpl di-la-ja-rka 1sg-gen-loc-abl 'Yours (i.e. your story) is better than mine.'

Synchronically, all markers are allomorphs of the same case. For instance, in *tusnaq-leb* (prison-loc-hpl) 'in the prison' vs. *tusnaqːa-ja-b* (prison.obl.pl-loc-hpl) 'in the prisons' the only difference is the number of the noun to which the case suffix is attached. Furthermore, it is ungrammatical to have both the vowel change and a suffix *-a* or *-ja* with one and the same nominal, e.g. \**ša-ja* (village.loc-loc). Normally each nominal can apply only one operation to form the loc-series, but there are a few examples that prove that there is some variation, e.g. *neqːe* vs. *neqːi-le* 'into the cave' (cave.loc vs. cave-loc). As the above description has shown, the distribution of the allomorphs used is at least in part lexically determined and needs further study.

Diachronically, all markers go back to formally and functionally unrelated markers. This becomes clear when we compare Sanzhi to other Dargwa varieties. In her comparative paper on spatial cases in Dargwa, van den Berg (2003b) provides spatial case paradigms of ten Dargwa varieties from north to south. For the analysis of the Sanzhi locative marker three groups of suffixes are relevant:

### 3 Nouns


The Sanzhi locative case marker seems to be a mixture of all three groups. Formally its exponents correspond to suffixes from all three groups, and functionally the marker unifies the three different meanings. Other Dargwa varieties show a similar picture. For instance, Tanti Dargwa, another south Dargwa variety, has a super-series expressed with the suffix *-ja* that is part of the regular paradigm of spatial cases. In addition, it has a category 'location' (*lokalizatsija*) that is only formed from the direct stem of nouns in the singular (Sumbatova & Lander 2014: 66–68). This special form is used when expressing the most natural location of a figure with respect to the ground. As the Sanzhi locative, its formation is very heterogeneous by means of unproductive suffixes (*-na*, *-ni*), vowel change (*i > e*) or a switch of the pitch accent to the word final vowel *a*. The Tanti examples parallel the examples of the Sanzhi locative given above.

Mekegi, a northern Dargwa variety, has a general locative suffix *-le* that is directly added to the nominal stem. This suffix is mentioned in van den Berg 2003b, but unfortunately she does not provide examples or a description of its meaning. In the same paper, van den Berg suggests that this marker has cognates in Akusha Dargwa (*-la*) and Urakhi Dargwa (*-la*) and notes that its precise meaning requires further investigation. The Akusha Dargwa grammar by the same author provides a few examples of the suffix *-la*, which is only added to inanimate nouns (van den Berg 2001: 24). On the same page, the grammar also mentions a couple of nouns with irregular locative forms that have shapes analogous to some of the Sanzhi words discussed in this section.

### **3.4.2.3 ad-lative -***šːu***, ad-essive** *-šːu-b***, and ad-ablative -***šːu-r*

The series of spatial cases formed with the suffix -*šːu* denotes movement to a goal (the moving item is not further specified for precise location with respect to the goal), general location that can be broadly translated with 'at, by, with', and movement away from a source. The goal, location or source, i.e. the noun bearing the spatial case suffix, mostly has an animate referent. Thus, the loc-series and the ad-series are in a kind of animacy opposition. However, as (107) and (108) show, inanimate reference points are also allowed.

(105) na now istikan-na glass-gen juldašː-a-šːu friend-obl.pl-ad ʁudur mix ∅-ič-ib m-occur.pfv-pret ca-w cop-m hel that 'He mingled with his drinking friends.'

<sup>7</sup>Van den Berg (2003b) further hypothesizes that there is a connection between the markers for 'in a hollow space' and the general locative markers.

3.4 Case


### **3.4.2.4 in-lative -***cːe***, in-essive -***cːe-b***, and in-ablative -***cːe-r*

The suffix of the in-series is -*cːe*. The locational meaning can be roughly translated as 'in' (111), or 'on, at' (112), and its directional meaning is 'to' (113).


Note that in contexts such as 'in a settlement', 'in a container-like object' or 'on a vertical surface', it is possible to use the loc-series instead of the in-series with no semantic differences between the two variants. Thus, compare (111) with (96), (114) with (93), and (116) with (95).

(114) sːurrat picture baˁʡ-li-cːe-b wall-obl-in-n sa-r-h-aq-ib ante-abl-up-hang.pfv-pret ca-b cop-n ' The picture is hanging on the wall.' (E)

### 3 Nouns


The in-ablative does not only translate as 'from' (117) and more specifically as 'from within, out of' (118), but is also used to denote 'among, along, through' (119). Thus, we find it in superlative constructions (120) (§30.1):


The in-lative has also more metaphorical uses when marking the goal-like argument of the verbs *aq*- 'go through' (121) or *b-arχː*- 'be engaged in' (122) and other predicates (123).


3.4 Case

Young speakers use the in-ablative alone (124) or in combination with the postposition *b-alli* 'together' to express the comitative. Older speakers reject such a usage by pointing out that the comitative case -*cːella* that can be optionally combined with the same postposition (§3.4.2.1) is the only grammatical variant.

(124) hel-ka that-down sa-∅-jʁ-ib hither-m-come.pfv-pret nik'a small durħuˁ boy welisipjed-li-cːe-r bike-obl-in-abl 'From over there a little boy with a bike came.'

The in-essive is used in the temporal expression 'in the year X' (125).

(125) xːunul woman ka-r-iž-ib=da down-f-be.pfv-pret=1 urek-c'a-ra-ib six-ten-num-ord dusːi-cːe-b year.obl-in-n 'I married in 1960.'

There are a number of non-spatial functions that the in-series fulfills. The in-essive expresses temporarily limited possessors (126).

(126) hel that ručka=ra pen=add le-b exist-n iž-i-cːe-b this-obl-in-n 'He also has a pen (in his hands).'

The in-lative denotes temporarily limited recipients (127), addressees (128) and causees (129) (see §24.3 for reported speech constructions and §19.2.2 for causativization).


The in-ablative marks causers and causes (130), involuntary agents (131) and other non-canonical agents (132). In the involuntary agent construction, the verb cannot be transitive, i.e., it cannot have a genuine agent argument, but must be intransitive or labile. The added involuntary agent is thus rather an adjunct than an argument.

### 3 Nouns


### **3.4.2.5 sub-lative -***gu***, sub-essive -***gu-b***, and sub-ablative -***gu-r*

The suffix of the sub-series is *-gu*. The spatial meaning of the sub-series is 'under'. It is added to the oblique form of the noun, but for many nouns the oblique form can be identical to the citation form (133–135).


The sub-series has some more lexicalized (135) and metaphorical uses (136), (137).


'Here also (i.e. in this case) I am guilty in front of my daughter-in-law.' (lit. 'I occurred under their sins, of that son's wife')

There is a spatial preverb that has the same form and the same meaning as the case marker (§11.6.1) and is often used in clauses that contain nouns bearing the spatial case suffix (138). There is also a formally and semantically identical spatial postposition/ adverbial that recurrently appears after the spatial case marker (139), (140).

(138) qajqaj-li-gu jaw-obl-sub nuˁq-be=ra hand-pl=add gu-ha-d-uc-ib-le sub-up-npl-catch.pfv-pret-cvb ka-∅-jž-ib down-m-remain-pret ca-w cop-m

'He is sitting with his hands holding his head (lit. 'catching the hands under the jaw').'

(139) ču-la refl.pl-gen d-urkː-ar npl-find.ipfv-prs iχ-tːi dem.down-pl ganza-l-gu-d ground-obl-sub-npl gu-d under-npl daˁʡle like ca-d cop.npl

'Their, these (vegetables) are probably like (growing) under the ground.'

(140) Ulučaj Uluchaj b-ik'-ul hpl-say.ipfv-icvb erk'ʷ river ca-b cop-n ka-b-ax-an down-n-go.ipfv-ptcp nišːa-la 1pl-gen šːi-l-gu-r village-obl-sub-abl gu-b-a down-n-dir

'The river with the name Uluchaj is passing by down from our village.'

### **3.4.2.6 ante-lative -***sa***, ante-essive -***sa-b***, and ante-ablative -***sa-r*

The broad meaning of the ante-series is location in front of a reference point or on flat surfaces. It can be translated into English with 'in front, by, at', but also with 'on'. In the latter meaning it is functionally equivalent to the loc-series and the in-series. Thus, instead of *baˁʡ-li-sa-b* in example (142) it is possible to use *baˁʡ-le-b* (93) or *baˁʡ-li-cːe-b* (114). Similarly, a more common alternative to *xːun-ni-sa-b* (143) is *xːun-ne-b* (way-locn), but *xːun-cːe-r* (way-in-abl) is also attested, and instead of *burma-sa* (141) also *burmacːe-b* (threshold-in-b) can be found. I leave the more precise analysis of the semantic similarities and differences between the three cases for future research.


'There is also a rifle (hanging) on the wall.'

### 3.4 Case

### 3 Nouns


The ante-ablative is also used with a number of experiential predicates such as 'be afraid, fear', 'long for, miss', 'be embarrassed', and 'be ashamed', with which it denotes the source-like stimulus of the experience (145), (146). When an animate noun bearing the ante-essive or the ante-lative is used together with a verb of movement or a locative predicate the meaning is 'herd, pasture; look after, care for' (147).


### **3.4.2.7 post-lative -***hara***, post-essive -***hara-b***, and post-ablative -***hara-r*

The meaning of the post-series is location behind (148), (149) and next to a reference point (150), (151), although this is not always reflected in the English translation. Sometimes it occurs together with the postposition *hitːi* that roughly has the same meaning (149), (151), see §8.1.4.


3.5 Derivation of nouns

(151) rursːi-la girl-gen kulpat family lak' throw či-r-ka-b-arq'-ib-le spr-abl-down-hpl-do.pfv-pret-cvb heχ-i-la dem.down-obl-gen sub-li husband-erg r-alt-u=w f-let.ipfv-prs=q di-hara 1sg-post hitːi-r? behind-f 'Will her husband allow her (i.e. my daughter) to throw away the family and (sit) with me?'

### **3.4.2.8 Directional** *-b-a*

There is a further directional suffix containing a gender/number agreement marker, namely *-gm-a*. This suffix is almost exclusively used with spatial adverbs (85), (140) (see §7.1 for more examples). But it can occasionally be used with nouns (152), (153).

(152) il=ra that=add χːula big b-irʁ-an n-come.ipfv-ptcp ca-b cop-n hel that qal-la house-gen baˁʡ-li-cːe-r-b-a façade-obl-in-abl-n-dir 'That is also growing along the façade of the house.'

(153) paˁχ.paˁχ-la pakh.pakh-gen heχtːu-d-a there.down-npl-dir ag-ur-re go.pfv-pret-cvb hej this χːuˁrba-la graveyard-gen šːal-le-r side-loc-abl žaˁr-ra lime-gen qːatːa-d-a canyon.loc-npl-dir 'They went in the direction of pakh-pakh (microtoponym), across the side of the graveyard, through the lime canyon.'

### **3.5 Derivation of nouns**

### **3.5.1 Agent nouns with** *-či*

This suffix is found in a great number of East Caucasian languages and originates from Turkic. It productively derives agent nouns from nouns denoting objects or places. The base nouns are partially loans from Turkic or Arabic, but also include more recent loans from Russian and native words. The derived nouns can refer to men as well as to women and form the plural with *-be*.


### 3 Nouns

### **3.5.2 Agent nouns with and** *-kar***,** *-q'aˁ***, and** *-uˁq'*

These suffixes, which are not productive, derive agent nouns from other nouns, infinitives, short adjectives, parts of compound verbs and one postposition. The suffix *-kar*, which also exists in Standard Dargwa, is predominantly attested with borrowed nouns (155). Nouns with *-q'aˁ* mostly form their plural with *-ne* (after deletion of the stem-final vowel) (156); nouns with *-uˁq'* use *-e* as plural marker (157) and nouns with *-kar* employ *-te* or *-ne*.


### **3.5.3 Abstract nouns with** *-dex*

This very productive suffix derives abstract nouns from open class words (verbs, adjectives, nouns, adverbs, bound stems). The nouns derived by means of the suffix *-dex* control either neuter singular (160) or occasionally neuter plural agreement. Some of the derived nouns can form the plural, usually by means of gemination of the last segment plus suffixation of *-e*, e.g. *ʁʷabzadexːe* 'heroism, bravery, courage'. Thus, it is also possible to analyze the underlying form of this suffix as containing a geminate as final consonant that is degeminated in the singular because of the impossibility of having geminate consonants in syllable-final position. A similar analysis regarding geminate stem-final consonants in singular vs. plural forms of other nouns that take the plural suffix *-e* has been suggested in §3.2.1.

Base adjectives are mostly underived short adjectives, i.e. bare roots (158). The numeral *ca* 'one', but no other numeral, can also serve as base:

3.5 Derivation of nouns


Base nouns either denote people or abstract items. The derived nouns refer to abstract items or properties associated with the thing or the person that the base noun denotes.

	- < *ʁʷabza* 'dzhigit, hero'
	- b. *ʡaˁχːuˁl-dex* 'hospitality'
		- < *ʡaˁχːuˁl* 'kunak, guest, host'
	- c. *dajark'a-dex* 'duties of a milkmaid' < *dajark'a* 'milkmaid'
	- d. *uruχ-dex* 'fearfulness, anxiousness'
		- < *uruχ b-iχʷ-ij* 'get afraid'
	- e. *saniʡaˁt-dex* 'skillfulness'
		- < *saniʡaˁt* 'skill'

With verbs only the preterite (predominantly from perfective stems) and the *-an* participle (usually from imperfective stems) can function as base for the derivation of *-dex*nouns (160). The only regular exceptions are forms of the copula verbs, which have defective paradigms (161).


The following example shows a compound noun derived with *-dex*. There is no independent base noun *urk'i hitːi* and the complex can also not be regarded as a postpositional phrase because *hitːi* requires the dependent noun to be marked with the genitive. When *-dex* is added the complex functions as a nominal that controls agreement on the clause-final verb, and the reciprocal pronoun preceding it functions as a modifier of it (or of the noun *urk'i*).

(162) ca-lla one-gen ca-lla one-gen urk'i heart hitːi-dex behind-nmlz b-akː-u n-cop.neg-prs 'There is no pity between them (one for the other).'

### 3 Nouns

The nominalized verbs retain their arguments, but since they are nominalized they occur in argument position and can be modified, e.g. by personal pronouns (163). Thus, the subject-like arguments can either occur as preserved arguments of the nominalized verb (164), (165) or they can occur as possessors (163).


Many of the derived abstract nouns, especially those derived from verbs, have the semantic role of cause or reason and therefore bear the dative suffix. Thus, deverbal nominals occur as nominalized adverbial clauses with the meaning 'because of X'. When inflected for the dative more words are admissible as base for the derivation than would be possible without the case suffix. For instance, there is no noun *itːu-r-dex*, but if this word is inflected for the dative, it can occur as an adverbial denoting the cause (165).

(165) u 2sg itːu-r-dex-li-j there-f-nmlz-obl-dat dam 1sg.dat wahi-l bad-advz ca-d cop-npl 'Because you (fem.) are there I feel bad.' (E)

### **3.5.4 Action nouns/event nouns and tools with** *-ala*

The suffix *-ala* is used for the formation of nouns that denote tools and of action or event-denoting nouns with the meaning 'way of V-ing' from verbal stems. It is morphologically productive and rather resembles inflectional suffixes such as the masdar, but it is not commonly used in texts. The resulting nouns keep their arguments. They can be marked for plural (suffixes *-e* or *-me*) and control neuter singular agreement. Those nouns that have been derived from transitive verbs with a gender agreement prefix have a frozen gender prefix *b-* (e.g. *biχala* 'lace, shoelace', *bušːukala* 'broom'). However, the verb given in (166i) originates from an extended intransitive verb that agrees with the human agent in gender. As (167) shows, this noun has variable gender depending on the gender of the referent and thus behaves like a small number of underived nouns with gender exponents (1).


<sup>8</sup> Synchronically, this verb is no longer used in Sanzhi Dargwa and has been replaced by a compound verb *qʷaˁrš b-arq'-ij* 'wipe, sweep, stroke'. However, it is attested, e.g., in the South Dargwa variety Tanti. Thus, I suggest that Sanzhi lost the verb, but kept the noun. An alternative explanation suggested by my main language assistant is that Sanzhi borrowed the noun from Standard Dargwa.

3.5 Derivation of nouns

d. *berkʷ-ala* 'dish, food' < *b-erkʷ-ij* 'eat' (pfv) e. *irʁ-ala* 'understanding' < *irʁ-ij* 'understand' (ipfv) f. *kaxʷ-ala* 'way of killing' < *kaxʷ-ij* 'kill' (pfv) g. *haʔ-ala* 'proverb, saying' < *haʔ-ij* 'say' (pfv) h. *bet'-ala* 'boredom' < *b-et'-ij* 'be bored, long for' (pfv) i. *dalaj r-ikʷ-ala* 'way of singing songs' < *dalaj r-ikʷ-ij* 'sing songs'

(167) iχ dem.down rursːi-la girl-gen qːuʁa-l beautiful-advz kʷi in.the.hands sa-b-uc-ala=le hither-n-keep.pfv-nmlz=emph nik'a-ce small-dd.sg '(Look at) the beautiful way of the girl's holding the child in her hands!'<sup>9</sup>

### **3.5.5 Action and event nouns with** *-utːi* **and** *-a*

The suffixes *-utːi* and *-a* (allomorph *-aˁ*) derive action nouns from verbs, whereby they are almost exclusively added to the perfective stem (168), (170). They are not productive. Most of the nouns can be marked for plural, normally with the suffix *-ne* (e.g. *bebk'-ne* 'funerals'). They control either neuter singular (169) or occasionally neuter plural (170c), (170d) or masculine singular (170e) agreement. The gender prefixes of the deverbal nouns are petrified.


<sup>9</sup>The agreement prefix in this example is neuter singular, expressing the gender of the nominalized verb form itself. The noun *rursːi-la* is a genitive modifier of the nominalized verb.

3 Nouns

### **3.5.6 Other derived nouns**

There are four nouns that are formed by means of reduplication of the first syllable and the suffix *-aj*. For most of them the base is bound and unclear, i.e. it does not exist as an independent word. The last three of the four words are toys and/or names of games.


There are a fair number of Sanzhi words that end in *-aj* (172), (173) and it seems reasonable to suggest that the language once had a similar derivational suffix. This has been suggested for Standard Dargwa (Abdullaev et al. 2014: 90). Around half of the words are clearly morphologically complex. Two of them are only used in child-directed speech, and five of the words refer to human beings, more particularly, kinship relations or social roles (173).

	- b. *aqanaj* 'top, summit' < *aq* 'tall, high, loud'
	- c. *zanʁaraj* 'beanbag, rattle' < *zanʁ* 'ringing'
	- d. *pːapːaj* 'bread' (child-directed speech)
	- e. *maˁmmaj* 'scary' (child-directed speech)
	- f. *tːapːaraj* 'zany, moony, erratic'
	- g. *sːurkːupːaj* 'swing' (i.e. hanging seat)
	- h. *dalaj* 'song'

(173) a. *wanaħaˁj* 'mama's boy, lazy person'

b. *k'uk'aj* 'gossiper'


There are three kinship terms with the suffix *-q'ar* (174).


	- c. *k'uni-q'ar* 'nephew' < ?

3.6 Reduplication and compounding

A few words are derived by means of the spatial case *-gu* plus a further suffix *-(l)la*, which is formally identical to the genitive (175), and with the postposition *sala* 'in front' or the spatial case suffix *-sa* (ante) (176).


Place names are derived from verbs by means of the locative participle *-an* (see §18.1.2.4). Names for ethnic groups, inhabitants, etc. are derived by adding the suffix *-(a)n* to a root that might be the place name or some other root related to it (see §10).

### **3.6 Reduplication and compounding**

Neither compounding nor reduplication are productive or frequent ways of forming new nominals in Sanzhi. In this section, I list the majority of reduplicated and compound nouns that I have been able to identify so far.

### **3.6.1 Reduplication**

A number of nouns have the structure CV.CV(V) or CVC.CVC(V) and are composed of two (almost) identical segments following each other. Some of the nouns are clearly onomatopoetic (177), others are not (178).


### 3 Nouns

### **3.6.2 N + N compounds**

Sanzhi has a few noun plus noun compounds. Occasionally, the origin of one of the compound members is unclear. Because Sanzhi has also nominal apposition it is not always easy to differentiate between juxtaposed nouns in an apposition construction that syntactically form a phrase (§21.1) and compounds that function as one word. There are several criteria that need to be applied in order to identify compounds. First, a few words show a greater phonological cohesion (e.g. *ababa* 'grandmother') or make use of word forms that differ from the base stems or inflected forms (e.g. *kːalkːa* 'tree'in *kːalkːa zize* 'strawberry').

(179) a. *ababa*<sup>10</sup> 'grandmother' (on mother's side)

< *aba-la aba* 'mother-gen mother'

b. *atːaba* 'grandmother' (on father's side)

< *atːa-la aba* 'father-gen mother'


< *waqːa* 'skull' + *ʡaˁt'a* 'frog'

Second, a few compounds are of the dvandva type. This means that they are coordinative compounds, but in contrast to noun phrase coordination (§26.1) dvandva compounds do not make use of the additive enclitic and function morphosyntactically as one word, i.e., they take only one inflectional suffix.


Two more nouns are classified as noun + noun compounds because they are semantically not transparent. They are neither coordinative compounds nor do they have an identifiable head noun such that they cannot be analyzed as appositions (i.e. noun phrases). The second noun in (181b) seems to have been formed by reduplication.

$$\begin{array}{ccccc}\text{(181)} & \text{a. } & du\acute{a} \text{:} i \, la^{\uparrow}mc^{\uparrow} \text{ 'frefly'} & \text{<} & du\acute{a} \text{:} i \, \text{"} \text{right+} & \text{+} la^{\uparrow}mc^{\uparrow} \text{'} \text{[light} \text{ing'}\\ & \text{b. } & t^{\uparrow}ama \text{ } \text{hama } \text{'gossip}, \text{hub-hub'}\\ & & \text{.} & \text{.} & \text{.} & \text{.} & \text{.} & \text{.} \end{array}$$

< *t'ama* 'sound, noice, voice' + ?

<sup>10</sup>My main language assistant G. Gadzhimuradov said that this and the following word in (179b) are used by the younger generations, but not by elderly speakers, and may have been borrowed from other Dargwa varieties.

3.6 Reduplication and compounding

### **3.6.3 Other compounds**

There are a few more X + noun combinations that have been lexicalized as compound nouns and show phonological and morphological cohesion. The first part X can be an adjective (including numerals) or a verb (182). The order of both parts corresponds to the usual order of phrases (modifier + noun) or clauses (object + verb).


The compound nouns that contain numerals as their first part can be divided into two groups. The first group in (183) has nouns as the second part and additionally what looks like unproductive derivational suffixes (*-lan, -ar, -an*).<sup>11</sup> The base nouns are usually in the plural, which indicates that these are compound nouns and not phrases. In noun phrases with numerals, nouns normally occur in the singular. Furthermore, the numerals appear in their basic stem form that cannot be used independently, but only as the basis for other word formation processes.

(183) a. *azirt'uˁlan/azirt'uˁmar* 'centipede'

< *azir* 'thousand' + *t'uˁ-m-ar* (leg-pl-nmlz)

	- < *azir* 'thousand' + *kam-r-an* (layer-pl-nmlz)/*k'ap-r-an* (leaf-pl-nmlz)
	- < *ʡaˁb-k'apːr-ar* (three-leaf-pl-nmlz) + *q'ar* 'herbs'
	- < *ʡaˁb* 'three' + ?
	- < *ʡaˁb* 'three' + *t'uˁ-m-ar* (leg-pl-nmlz)

The second group of compound nouns with numerals has only two members (184) that denote offspring born in a multiple birth. The first part is again the basic stem form of the numeral whereas the second part looks like a noun derived from a verb by means of an otherwise unattested suffix *-i*.


<sup>11</sup>The suffixes -*ar* and -*an* both contain the same vowel *a*, which is identical to the vowel in oblique plural forms of nouns (e.g. *t'uˁ-m-a-lla* leg-pl-obl-gen) such that one could perhaps suggest that the vowel is actually not part of the suffix. However, oblique stem forms of nouns are normally only used for case formation and we would need an additional motivation for using the oblique form as the basis of derivational processes. Both suffixes are also used for the derivation of adjectives, see §5.3.

### 3 Nouns

### **3.7 Phrasal compounds**

Sanzhi has a very productive way of forming short noun phrases that have the structure of short participial clauses of the form (noun) + verb and semantically strongly resemble derived action nouns (§3.5.4, §3.5.5). In the minimal case a verb alone to which the modal/future participle *-an* is suffixed is enough. If the verb is transitive, it is often accompanied by a patientive argument. Morphosyntactically, these combinations are headless relative clauses with a transparent semantics (§23.4), but the phrases are lexicalized and perceived as complex nouns by the speakers. They denote agents, professions or tools.

	- b. *sarruˁrqaˁn* 'kidnapper'
		- < *sa-r-r-uˁrq-aˁn* ante-abl-f-drag.ipfv-ptcp
	- c. *paltar durχan* 'tailor'
		- < *paltar* 'clothes' + *d-urχ-an* npl-sew.ipfv-ptcp
	- d. *q'ʷaˁl icːan* 'milkmaid'
		- < *q'ʷaˁl* 'cow' + *icː-an* milk.ipfv-ptcp
	- e. *ʡuˁnze sarirt'an* 'handkerchief'
		- < *ʡuˁnze* 'snot, slime' + *sa-r-irt'-an* ante-abl-take.away.ipfv-ptcp
	- f. *sːurrat helt'an* 'photographer, camera' < *sːurrat* 'picture' + *ha-elt'-an* up-take.out.ipfv-ptcp
	- g. *ʁaj hadulq'an* 'intriguer'
		- < *ʁaj* 'word, language' + *ha-d-ulq'-an* up-npl-lock.ipfv-ptcp
	- h. *kːʷiš iʁan* 'dough scraper'
		- < ?*kːʷiš* 'dough'<sup>12</sup> + *iʁ-an* chase.ipfv-ptcp
	- i. *pal kerxʷan* 'fortune-teller'
		- < *pal* 'prediction'<sup>13</sup> + *ka-erxʷ-an* down-pour.ipfv-ptcp

There is a range modifier + noun combinations that have idiomatic meanings that are not transparently predictable from the meanings of the parts, but syntactically are rather phrases and not compound nouns. The modifiers used are nouns in the genitive case (186) or adjectives (187). Sometimes the origin of nouns used as modifiers in these combinations is unclear. Most examples denote animals or plants. In fact, sometimes names for herbs, healing plants or other edible plants seem to be made up on the spot and are rather descriptive.


<sup>12</sup>The word *kːʷiš* is not the regular word for 'dough' in Sanzhi, but it exists in the derived noun *kːʷiš-a* 'wooden board for making dough'.

<sup>13</sup>The word *pal* does not seem to be used in Sanzhi, but it is part of the derived noun *pal-či* 'fortune-teller'


## **4 Pronouns**

Sanzhi Dargwa has the following types of pronouns:


This chapter also includes a subsection on quantifiers such as 'some', 'every' and 'all' (§4.7).

Pronouns express the typical features of nominals, namely case, number, and to a very limited extent gender (only reflexive pronouns in the absolutive case, one type of reciprocal pronoun, essive-case forms of pronouns, e.g. of the pronoun 'where'). Case marking of pronouns is almost fully regular and identical to the case marking of nouns (and nominalized adjectives, verbs, etc.). As for number marking, only the demonstrative pronouns and the interrogative 'who' form the plural by means of special suffixes; personal and reflexive pronouns use suppletive stems and indefinite pronouns mostly do not have plural forms. The gender exponents are the usual markers that are used across all parts of speech that express gender.

### **4.1 Personal pronouns**

Sanzhi has personal pronouns for the first and for the second person. For the third person demonstrative pronouns are used (§4.2). Table 4.1 displays a partial paradigm of the personal pronouns.

It is possible to make a few generalizations about the morphophonological structure of the personal pronouns. The absolutive and the ergative make use of the same root. Most other cases are formed from a distinct oblique root that is formed via ablaut from the absolutive root (the first root vowel changes *u* > *i* for the first person, *u* > *a* for the second person; the second root vowel of the plural pronouns changes *a* > *i*). The only exceptional forms are the dative forms of the singular pronouns that do not contain segments that could be identified as dative case exponents, and the genitive forms of the

### 4 Pronouns

plural pronouns that are a mixture of absolutive and oblique stem. Note, furthermore, that the plural pronouns have *-lla* as the genitive suffix. The same allomorph of the genitive case suffixe is optionally used for plural nouns §3.4).


Table 4.1: Personal pronouns

### **4.2 Demonstrative pronouns and adverbials derived from them**

Sanzhi Dargwa has a rich system of demonstratives whose stems express number and case, but not gender. These demonstratives fulfill a variety of deictic and non-deictic functions. Their deictic uses can be exophoric (e.g. gestural) or discourse deictic when demonstratives refer to a chunk of discourse (*She said this*). Non-deictic uses of Sanzhi demonstratives can be anaphoric or cataphoric. The demonstratives are organized along several formal and semantic dimensions:


Table 4.2 displays the demonstratives that serve pronominal and adnominal functions. In the table, they are divided into three series in the columns in both the singular (*iC, heC, hiC*) and the plural (*i(C)tːi*, he*(C)tːi*, *hi(C)tːi*). The series in the columns are distinguished by the root-initial segments (§4.2.1). There are six series of pronouns in the lines of the

### 4.2 Demonstrative pronouns and adverbials derived from them

table that differentiated by their last root consonant (before the plural suffix in case of the plural pronouns), i.e. *ž (š)* vs. *j* vs. *l* vs. *t* vs. *k' (x)* vs.*χ* (§4.2.2). The series with *j* as the last root consonant is defective because it exists only for singular absolutive pronouns; singular oblique forms as well as any plural forms are unattested.

Table 4.2: Basic (i.e. absolutive case) forms of nominal demonstratives


The plural pronouns are mostly based on the singular pronouns by adding the plural suffix *-tːi* to the singular stem and some minor phonological adjustments. The oblique stems of the singular pronouns are formed by adding the suffix *-i* to the stem (Table 4.3) to which in turn case suffixes are attached. These two suffixes are not used for the inflection of nouns, but only with demonstrative pronouns. Partial paradigms of inflected pronouns are provided in Table 4.4. For the oblique stem of the plural pronouns the stem-final vowel *i* is replaced by *-a*, a suffix generally used for the formation of oblique plural stems of nouns (§3.4).

Table 4.3: Oblique stem formation of demonstrative pronouns


### 4 Pronouns

'that' 'that' 'this' 'that' 'those' 'those' abs. *il it iž hel iltːi heltːi* erg. *il-i-l it-i-l iž-i-l hel-i-l iltː-a- heltː-a-l* gen. *il-i-la it-i-la iž-i-la hel-i-la iltː-a-lla heltː-a-lla* dat. *il-i-j it-i-j iž-i-j hel-i-j iltː-a-j heltː-a-j* comit. *il-i-cːella it-i-cːella iž-i-cːella hel-i-cːella iltː-a-cːella heltː-a-cːella* ad-lat. *il-i-šːu it-i-šːu iž-i-šːu hel-i-šːu iltː-a-šːu heltː-a-šːu* in-lat. *il-i-cːe it-i-cːe iž-i-cːe hel-i-cːe iltː-a-cːe heltː-a-cːe* loc-lat. *ile/il-i-ja it-i-ja iž-i-ja hele/ hel-i-ja iltː-a-ja heltː-a-ja*

Table 4.4: Partial paradigms of some demonstrative pronouns

The deictic meaning of the demonstratives is participant-oriented. Three semantic dimensions along the scale 'proximity/distance to speech act participants' are distinguished: (i) near hearer (root-final consonants *ž/š* and *j*), (ii) near addressee (*l*), and (iii) undifferentiated or not close to speaker or addressee (*t*).

Another aspect of the deictic semantics is elevation (or height), namely higher (up) or lower (down) location than the deictic center which is most commonly the speaker. Elevation distinctions in demonstratives are widespread in Dagestanian languages (Schulze 2003; Forker 2019a), and the Sanzhi Dargwa system represent a typical instance.

### **4.2.1 The demonstrative series in the columns:** *iC* **vs.** *heC* **vs.** *hiC* **and** *i(C)tːi* **vs.** *he(C)tːi* **vs.** *hi(C)tːi*

There is a pronounced difference in frequency between the three series. The *he-* series is by far the most commonly used and the *hi-* series is only very rarely used. Speakers are aware of the three different series but do not seem to notice a difference in semantics. The phonetic difference between the *hi*-series and the *i*-series is rather small and hard to hear. Thus, one of the reasons why the latter is so rare in the corpus might be that some of the tokens might incorrectly have been transcribed as *i-*. In the following, I will only discuss the *heC* and the *iC* series.

When looking into natural texts it is not difficult to find tendencies hinting at the functional difference between the *heC*-pronouns and the *iC*-pronouns. The *heC*-pronouns preferably refer to items or persons that are or have been:


### 4.2 Demonstrative pronouns and adverbials derived from them

First of all, *heC*- pronouns are used for denoting visible referents, for instance in pointing events. For example, after (1) has been uttered the speaker stresses the fact that the person in the narrative was only pointing at the man, but not saying anything:


'He gave me such a bottle and sent me (fem.) away.'

(3) [referring to a boy that turned up unexpectedly during the conversation] aman! oh het that ceqːel when ha-∅-jʁ-ib=e? up-m-come.pfv-pret=q 'Oh! When did he come?'

The *heC*-pronouns are used for referents within the personal social sphere of the speaker such as her/his close relatives and other people well-known to the speaker (4), body parts of the speaker (5), etc. These items or persons can be assumed to be implicitly present in the discourse and can be identified via their close relationship to the speaker.


'Now the (i.e. my) lips are swollen, it is difficult to talk.'

Second, the *heC*-pronouns refer to referents that have been introduced in the preceding discourse, either in the immediately preceding sentence such that they establish a kind of topic continuity or when referring back to something said some time ago. Thus, sentence (6a) brings up a new topic, the mill forest. The speaker is then constantly referring back to the forest with the pronouns *hek'*, *het* and *hej* (6b), (6c). The first clause of (7) introduces a new referent, the sticks, and the following clause refers to them by means of a *heC*-pronoun.

### 4 Pronouns

	- b. hej=ʁuna this=eq wac'a forest k'e-b, exist.up-n urχːab-la mill-gen wac'a forest b-ik'ʷ-ar hpl-say.ipfv-prs hek'-i-j dem.up-obl-dat

'A forest like this exists there, the mill's forest it is called'

c. hextːu there hek' dem.up wac'a-cːe-b forest-in-n cik'al nothing ka-b-irčː-an down-n-cut.ipfv-ptcp akːu cop.neg 'There in the forest nothing should be cut'

(7) [ When we were little we had a game.]

dirx-me stick-pl d-umkːa npl-sharp d-arq'-ib-le, npl-do.pfv-pret-cvb ganza-l-cːe ground-obl-in d-urq-aˁ-di npl-hit.ipfv-hab-1 hel-tːi that-pl

'We made sticks and hit them into the ground.'

While speakers performed the *Family Problems Picture Task* (San Roque et al. 2012), they constantly needed to refer to the people and objects depicted on the pictures. Frequently they first used an *iC*-pronoun to establish a new referent, and then, in a kind of afterthought following the clause, repeated the reference again by employing a *heC*pronoun (8), (9). The first demonstrative pronouns in such clauses can be interpreted as deictic, whereas the second demonstratives in the same examples represent the anaphoric use. In the following two examples the relevant demonstratives are given in boldface.


This function is reflected in the meaning of the adverb *hel-i-j* (that-obl-dat) 'therefore' and the phrases *hel bahandan* 'for this reason' and *hel zamana* 'that time', which link causally or temporally connected passages in a stretch of discourse.

Third, the *heC*-pronouns denote objects and persons that the speaker assumes to be familiar for the hearer, i.e. that are common knowledge such as certain places, famous people, etc.:

4.2 Demonstrative pronouns and adverbials derived from them

(10) hej this Mažalis-la Mazhalis-gen pawarut'e-le crossover-loc w-iteʁ-ib=qːella, m-reach.pfv-pret=when … 'when (I) reached the crossover of Mazhalis, …'

The use of the *iC*-pronouns diverges from the use of the *heC*-pronouns. The *iC*-pronouns preferably occur when new topics and referents are introduced into the discourse (11) or when topics switch (12), (13).

(11) iž this Arsen Arsen b-ik'-ul hpl-say.ipfv-icvb iž=ra this=add le-w=de exist-m=pst iž this Usːan Usisha.person xːunul-li-cːe-r woman-obl-in-abl

'And this man called Arsen, he was also there, with his wife from Usisha.'

(12) [switching back the topic of the conversation to a person known to both speaker and hearer]

```
c'il
then
     it
     that
          ʡaˁbdulχaliq'
          Abdulkhalik
                        acːi-la
                        uncle-gen
                                    atːa-l
                                    father-erg
                                                ce=jal
                                                what=indef
∅-ik'ʷ-a-tːe=q'al
m-say.ipfv-hab.pst-2sg=mod
                                u
                                2sg
```
'What did you say about uncle Abdulkhalik's father?'<sup>1</sup>

(13) [Then when they had gathered all these pears, they put them again on the bike.]

il that durħuˁ boy arg-ul=de go.ipfv-icvb=pst bahla-l, slow-advz itːi dem.pl ʡaˁbal three durħ-ne boy-pl arg-ul=de go.ipfv-icvb=pst het that sa-∅-jʁ-ib hither-m-come.pfv-pret musːa-r place.loc-abl het that xːun-ni-cːe-r road-obl-in-abl

'And the boy went slowly away, and the three boys went there to that place, on that road.'

The *iC*-pronouns are also used when the referent or the topic of the conversation has been introduced into the discourse, but the speaker considers them to be out of his/her personal sphere. For instance, in (14) and (15) the speaker continues to talk about acquaintances of hers who are not close friends or relatives of herself:

(14) [ In Urkarakh there is this son of my friend.]

ik'-i-l dem.up-obl-erg har every cik'al something di-la=ra 1sg-gen=add d-irq'-u npl-do.ipfv-prs 'He does all my things. (i.e. does everything)'

<sup>1</sup>The noun 'father' bears the ergative case because the speaker intended to ask for something that Abdulkhalik's father had done, without explicitly saying so in his utterance.

### 4 Pronouns

(15) it-i-l that-obl-erg b-alχ-an n-know.ipfv-ptcp b-akːu=q'al n-cop.neg=mod r-ik'ʷ-ar f-say.ipfv-prs 'She does not know anything, (my daughter) says.'

In example (16) the speaker is talking about a person who is present but does not belong to the Sanzhi community, and who does not understand Sanzhi (later the speaker switches to *hel* when referring to the same person):

(16) iž-i-l this-obl-erg d-alc'-un npl-gather.pfv-pret q'ar plant le-d=de exist-pl=pst 'There were plants that she gathered.'

In (17) the speaker is contradicting and correcting the addressee (who is his wife) and perhaps distancing himself a bit form the referent (his sister-in-law):

(17) [Prepare (the groceries), when your little sister comes, for her to take them.]

iž-i-l this-obl-erg d-uqː-ij npl-carry.pfv-inf a-r-irχ-u neg-f-be.able.ipfv-prs itːi dem.pl 'She cannot carry them.'

However, these are only tendencies, not strict rules. Speakers play around with the pronouns, use different pronouns for one and the same referent or correct themselves. Thus, in (18) and (19) the same objects (the pills) and person (the friend) are first referred to by means of a *heC*-pronoun and then immediately later by an *iC*-pronoun. This is the opposite order of what I found in the data from the *Family Problems Picture Task* presented above in (8), (9).


c'il then ca-w refl-m ʡaˁħ good juldaš friend ca<w>i cop<m> nišːa-la 1pl-gen nu well **heχ** dem.down q'ʷila a.little q'ʷila a.little **iχ-i-cːe-b** dem.down-obl-in-n χasijat habit χe-b-il exist.down-n-ref ca-b cop-n 'He is our good friend, but he has a bit of this habit.'

In the following sections, I will discuss the differences between the horizontal series (i.e. the pronouns in the six different lines of Table 4.2) and largely ignore the differences between the columns.

4.2 Demonstrative pronouns and adverbials derived from them

### **4.2.2 Proximity, distance, and elevation**

### **4.2.2.1** *ž***-pronouns:** *iž***,** *hež***,** *hiž***,** *ištːi***,** *heštːi***,** *hištːi***; and** *j***-pronouns:** *ij***,** *hej***,** *hij*

These pronouns express proximity and normally denote referents close to the speaker. The *ž*-pronouns are preferably used as independent pronouns (16), (17), (23), (24), whereas the *j*-pronouns predominantly occur as deictic modifiers of nouns and definite markers similar to articles (20), (21), (22), but again these are tendencies, not strict rules. The *j*-pronouns have only singular absolutive forms, lacking entirely singular oblique and all plural forms.

All following examples are from the *Family Problems Picture Task* when speakers where referring to pictures and people on the pictures that were lying close to them on the table.


'This is when he came back; this (*iž*) then needs to be the (*hij*) very last.'


### **4.2.2.2** *l***-pronouns:** *il***,** *hel***,** *hil***,** *iltːi***,** *heltːi***,** *hiltːi*

These pronouns denote objects or persons that are not in the proximity of the speaker, but close to the addressee (25).

(25) [The addressee starts talking to the picture in front of her. The other speaker says to her:]

u 2sg il-tː-a-cːe that-pl-obl-in ʁaj word ma-r-ik'-ut! proh-f-say.ipfv-proh.sg 'Do not talk to them!' (i.e. to the pictures)

### 4 Pronouns

They are also employed when talking about absent referents or items located further away, not necessarily in the proximity of the hearer (26), (27). Finally, they are the default pronouns in fiction such as traditional narratives, legends, etc. (28).

(26) [talking about the colors used for the rock paintings; the conversation takes place far away from the painting]

il that kraska color atletit fly.away b-iχ-ub-le n-be.pfv-pret-cvb akːu, cop.neg il-i-j that-obl-dat cik'al nothing ag-ur-re=kːu go.pfv-pret-cvb=cop.neg

'The color does not fly off, nothing happened to it.'

(27) [I went to my house.]

hel-tːi that-pl kelg-un remain.pfv-pret heltːu-b there-hpl di-la 1sg-gen k'ʷel=ra two=add durħuˁ=ra, boy=add hel-tːi that-pl k'ʷel=ra two=add zunra neighbor admi=ra person=add

'They remained there, my two sons and those two neighbors.'

(28) aždaha monster ag-ur go.pfv-pret ca-r cop-f hel-i-cːe-r that-obl-in-abl 'She turned into a monster.'

### **4.2.2.3** *t***-pronouns:** *it, het, hit, itːi, hetːi, hitːi*

These pronouns refer to persons or objects whose location is undifferentiated, irrelevant, or impossible to determine or that are not close to the speaker or the hearer. They are used, for instance, when talking about people that are not present, or about unknown referents, of which it is not important where they are located (29), (30). They are also very frequently used in elicitation.


it that r-ax f-let r-at-ur, f-let.pfv-pret het-i-šːu that-obl-ad w-ax-ul, m-go.ipfv-icvb it that r-ax f-let r-at-ur f-let.pfv-pret hetilil-li-šːu other-obl-ad w-ax-ul m-go.ipfv-icvb 'He left her and went to that one; he left her and went to the next one.'

4.2 Demonstrative pronouns and adverbials derived from them

(31) [talking about a stone fence that the speaker is building; both speaker and hearer are located somewhere away from the fence]

c'il=ra then=add het that šːal-le-b side-loc-n lac fence či-b-irq'-an=uw? spr-n-do.ipfv-ptcp=q 'Then you also have to build the fence from that side?'

With (32) the speaker refers back to former times and (33) is the typical final statement of a traditional story that ends with the wedding of the protagonist:


### **4.2.2.4** *k'-/x***-pronouns:** *ik'***,** *hek'***,** *hik'***,** *ixtːi***,** *hextːi***,** *hixtːi*

These pronouns are used when referring to items or people located above the level of the deictic center (which is most commonly the speaker), e.g. in the mountains as in (34), (35) or higher than some other point of reference (36). For instance, in examples (34), (37), the deictic center is the speaker, but it can also be another location. See Forker (2019a) for more information on the deictic category of elevation in Sanzhi Dargwa.

(34) [talking about rock paintings located in the mountains, higher up than Sanzhi]

intersna=de interesting=pst ix-tːi dem.up-pl 'They were interesting.'

(35) [referring to the inhabitants of a legendary village that is supposed to have existed on the mountains above Sanzhi]

islam Islam prinimat accept b-irq'-an hpl-do.ipfv-ptcp zamana time hex-tː-a-l dem.up-pl-obl-erg prinimat accept b-arq'-ib-le n-do.pfv-pret-cvb a-b-určː-i neg-hpl-be.ipfv-hab.pst

'At the time when we became Muslims, they did not become Muslims.'

(36) [In one place there are trees.]

warilla.wari no.way u 2sg ix-tː-a-j dem.up-pl-obl-dat er look či-ma-hark'-utːa! spr-proh-look.ipfv-proh.sg 'Whatever may happen, do not look at them (=trees)!'

### 4 Pronouns

The factual elevation with respect to the deictic center can be minimal as long as speakers perceive a difference in height. For instance, the village of Druzhba where most Sanzhi speakers live is located on the flat land around 5 km from the Caspian sea coast. It stretches from the highway that runs parallel to the coast and to a point approximately one kilometer before the slops of some hills. If there is any difference in elevation between the two ends of the village it is minimal and not visible, but the part of the village closer to the sea is conceptualized as 'lower' whereas the part closer to the hills is regarded as 'higher'. Example (37) originates from a conversation about a woman who lives in the 'higher' part of the village and the speaker uses *hek'* with reference to that woman. Thus, it is not necessarily the location at the time of speaking that is relevant, but the usual location of the referent in relation to the deictic center can be decisive for the use of demonstratives.

(37) [talking about a woman who lives in the 'upper part' of the village]

hana now hek' dem.up hek-ka dem.up-down ka-r-eʁ-ij=al down-f-go.pfv-inf=indq gargar trembling gargar trembling r-ik'-ud f-say.ipfv-1.prs du 1sg 'If now she comes from over there I am trembling.'

Sentence (38) has been uttered during a *Family Problems Picture Task* discussion. The picture shows the arrest of the protagonist by the police. His wife is sitting on the ground and he is carried away by two policemen. In the picture, he is depicted higher than the woman. Example (39) is also part of a *Family Problems Picture Task* discussion. The two demonstrative pronouns refer to the main protagonists who are depicted in little bubbles above the main scene of the picture.

	- first dem.up-pl beautiful-advz life hpl-be.ipfv-icvb hpl-stay-pret cop-hpl hex-tːi dem.up-pl er look r-erč'-e f-look.pfv-imp k'ʷel=ra two=add canille! together

'It turns out that they lived well in the beginning. Look at them both together!'

ca-b;

### **4.2.2.5** *χ***-pronouns:** *iχ, heχ, hiχ, iχtːi, heχtːi, hiχtːi*

These pronouns denote referents located below the deictic center. For instance, example (40) originates from a conversation about a woman who lives in the part of the village closer to the sea and the speaker continuously uses *iχ* with reference to that woman. Examples (41) and (42) refer to people and items on pictures. The people are sitting down and the pumpkins (referred to as watermelons) on the picture are lying on the ground.

4.2 Demonstrative pronouns and adverbials derived from them


In the discourse deictic function, mostly the *χ*-pronouns occur (43), but the *k'-/x*pronouns can also occasionally be found.

(43) taman end ca-b cop-n heχ dem.down 'This is the end (of the story).'

Yet elevation cannot be the only criterion that governs the use of the *χ*-pronouns vs. the *k'-/x*-pronouns. For instance, in (44) the first demonstrative pronoun denotes people who the speaker has seen on TV. They are described as being located lower than some unknown point of reference. At the same time the sentence is a good example for the contrast between *iC*-pronouns and *heC*-pronouns as discussed in §4.2.1.

(44) du 1sg ∅-ik'-ul=da, m-say.ipfv-icvb=1 iχ-tː-a-la dem.down-pl-obl-gen ij=ʁuna this=eq but'a piece ca-b, cop-n di-la 1sg-gen hej=ʁuna=qːel this=eq=when 'I (masc.) say, they have a (little) piece (of a diamond) like that, while I have one like that.'

For a more thorough discussion of the elevational meaning in Sanzhi demonstratives and examples see the detailed account in Forker (2019a).

### 4 Pronouns

### **4.2.3 Pronouns and adverbs derived from demonstrative pronouns**

There are a few pro-forms derived from the demonstrative pronouns such as *hetilil* (30), *itilil* (45) 'other, next', *itil-ižili* 'one thing and another', and *it-heχ* 'this and that, various'.

(45) [He pointed with his finger on his forehead and said to the other man, Well here is the horsefly.]

itilil-li other-erg ix-ub-le throw.pfv-pret-cvb tupang weapon antːa-l-cːe forehead-obl-in zija=ra horsefly=add kax-ub kill.pfv-pret ca-b cop-n il that admi=ra person=add

'The other shot at the forehead and killed the horsefly and the man.'

The temporal/clausal adverb *helij* 'therefore' is diachronically the dative case form of the pronoun *hel* 'that'.

There is a series of manner adverbs with the meaning 'like this/like that' that is formed by adding the suffix *-itːe* to the singular demonstrative pronouns (except for the *j-* pronouns, since they are mostly used in the noun-modifying function, see Table 4.5). Their meaning is based on the meanings of the demonstrative pronouns as described in the preceding sections. Some examples can be found in (46), (47) (see also §7.3 for more examples).


(46) iž this hež-itːe this-advz b-irq'-an n-do.ipfv-ptcp ca-b cop-n 'It must be done like this.'

(47) hana=q'ar now=mod il-itːe that-advz a-haʔ-ib=de neg-say.pfv-pret=2sg 'But now you did not say this.' (i.e. you did not tell the stories that you told the other time)

There is another group of four manner adverbs with a similar meaning as the adverbs ending in *-itːe*, namely *itwaj*, *hetwaj*, *hitwaj*, and *ižwaj* 'like that, and so'. Their usage is illustrated in (48–50).

4.2 Demonstrative pronouns and adverbials derived from them


Spatial adverbs with the basic meaning 'here, there' are derived by adding the suffix *-tːu* to the pronominal stems (Table 4.6). The meaning of the spatial adverbs is transparently derived from the meaning of the demonstratives. As can be seen in the table, there are two series of spatial adverbs with the meaning 'there above' because both the singular as well as the plural pronominal stem can serve as the base for the derivation, but the adverbs with *x* are far more common than the adverbs with *k'*. The adverbs with *k'* are very rarely used and might even be switches to another dialect of Dargwa.



Since the adverbs have inherent spatial semantics, locational cases cannot be added, but only directional suffixes just as with other spatial adverbials or nominals. The lative is zero-marked, the essive is expressed through gender/number agreement, the ablative by means of the suffix *-r(ka)* and the directive through the suffix *-gm-a* including a gender/number agreement marker, e.g. lative *heš-tːu*, essive *heš-tːu-b*, ablative *heš-tːu-r(ka)*, directive *heš-tːu-b-a*. Examples can be found in ex:We crossed the border between Shurli and our (Sanzhi area), and up there we found a stoneex:Do they really allow people from here (to enter) the hospital. More examples are given in §7.1.1.

(51) šuˁrʡli-la=ra Shurli-gen=add nišːa-la=ra 1pl-gen=add dazu-la border-gen hetːu-r there-abl tːura outside d-ituq-un-ne 1/2pl-cross.pfv-pret-cvb hek'tːu-b there.up-n b-arčː-ib-il=de n-find.pfv-pret-ref=pst ca one qːarqːa stone 'We crossed the border between Shurli and our (Sanzhi area), and up there we

### 4 Pronouns


Another series of spatial adverbs denoting the source can be derived by means of the suffix *-ka* (which is probably a cognate of the second part of the complex ablative suffix *-r-ka*), e.g. *hež-ka* 'from here', *hel-ka* 'from there', etc. (§3.4.2). These adverbs can also have a temporal interpretation ('from time X on'). Moreover, there is a series of spatial adverbs with the meaning 'from X to X' containing the suffix *-k-itːu-b-a*, e.g. *hež-kitːub-a* 'from here to there' (57). This suffix is a combination of the ablative *-ka* (shortened to *-k*), the locational suffix *-tːu* and the directive *-gm-a*. Both series are available from all three stem types of demonstratives (*heC, iC,* and *hiC*), but only the adverbs based on *heC* are commonly used in my corpus. See §7.1.1 for Tables displaying all adverbs and more examples.


The equative enclitic *=ʁuna* 'like, similar' and the temporal enclitic *=qːel* 'when' can also be attached to the demonstrative pronouns leading to pro-forms used when comparing referents (2), (6b) and temporal adverbs with the meaning 'then, at this/that time' (32).

### **4.3 Reflexive pronouns**

Sanzhi Dargwa has simple reflexive pronouns (Table 4.7) and two types of complex reflexive pronouns (Table 4.8). In reflexive constructions, the reflexive pronouns refer only to third persons. For first and second person reflexivization personal pronouns are used. Reflexive pronouns are marked for gender (in the absolutive only), for number and for

### 4.3 Reflexive pronouns

case. The absolutive case of the reflexive pronoun is identical to the copula and might be diachronically related to it. For all other cases the pronoun has two stems (singular and plural).



Table 4.8: Complex reflexive pronouns


*<sup>a</sup>*with ergative controller

The simple reflexive pronouns occur in local and non-local reflexivization (including logophoric contexts across clausal boundaries, whereas the complex reflexive pronouns can only be bound within the clause. Both types of complex reflexive pronouns consist of a reduplicated form of the simple reflexive (Table 4.7). For the first variant of the complex reflexive pronouns, one part of the reflexive undergoes case-copying from the controller (in Table 4.8 exemplified with an ergative controller), and the second part takes the appropriate case-marking. In the second variant, the first part is invariably genitive. The second variant, the complex genitive reflexive, lacks a form for the genitive case, so it can never occur as possessor. Other functions in addition to local and non-local reflexivization are: emphatic reflexivization, comitative constructions and pause fillers.

All types of reflexive constructions are analyzed in more detail in §29.1 and in Forker (2014). The genitive singular and plural reflexive pronouns *cinna* and *čula* are used as pause fillers (§9.5). The absolutive reflexive pronouns occur in comitative constructions that have the formal structure of coordinated noun phrases (§30.3).

### 4 Pronouns

None of these additional functions are available for complex reflexive pronouns, which occur only in local reflexivization, emphatic reflexivization and reciprocal constructions (only plural reflexive pronouns).

### **4.4 Reciprocal pronouns**

Reciprocal pronouns are very similar to complex reflexive pronouns in form as well as in morphosyntactic behavior. They consist of a reduplicated form of the numeral *ca* 'one'. Sanzhi Dargwa has three types of reciprocal pronouns. Two of these pronouns always consist of the reduplicated numeral *ca* 'one'. Except for the genitive they fully inflect for case, but do not distinguish gender. One type of reciprocal pronouns is the equivalent of the genitive reflexive because its first part is always in the genitive. The second reciprocal has always one part in the absolutive. The third variant, *ca-b-a*, is also based on *ca* 'one', to which a plural suffix that exhibits gender/number agreement is added. It can also be reduplicated (this is not shown in the Table) and inflects for all cases. All reciprocals are shown in the partial paradigm in Table 4.9. In addition, the language also makes use of plural reflexive pronouns (Table 4.8) for the expression of reciprocity.

Syntactically, reciprocal pronouns behave similarly to complex reflexives because they are always locally bound. More information on reciprocalization can be found in §29.2 and in Forker (2014).


Table 4.9: Reciprocal pronouns

### **4.5 Interrogative pronouns**

The interrogative pronouns of Sanzhi are given in Table 4.10. Some of the pronouns are morphologically complex, consisting of the root *ce* 'what' to which other morphemes are added:

4.5 Interrogative pronouns


Table 4.10: Interrogative pronouns


The pronouns*čujna* and *čum* are also complex. They seem to contain the same root*ču-*. In order to arrive at *ču-jna* the derivational suffix *-na* (allomorph *-jna* after vowels) has been added. This suffix is also used to form multiplicative numerals (§6.5). The pronoun *kutːi* seems to be composed of a root *ku-* and an ending *-tːi*, the latter also found with plural demonstrative pronouns (§4.2).

In the following, all pronouns are described and illustrated with examples. More information on interrogative clauses can be found in Chapter 28. Embedded interrogatives are treated in §28.4.

### **4.5.1** *ča* **'who' and** *ce* **'what'**

Partial inflectional paradigms of the pronouns *ča* 'who' and *ce* 'what' are shown in Table 4.11. The pronoun *ča* has a suppletive stem *hi-* for all cases except for the absolutive (60). The pronoun *ča* can be used as a modifier to a nominal with human reference and translates then as 'which, what kind of' (58). It can be marked for plural by means of the associative plural suffix *-qal* (59) (§3.2.4).


'"Nice, who were they?" he asks the Russian woman.'

### 4 Pronouns


Table 4.11: Interrogative pronouns *ča* 'who' and *ce* 'what'

(60) il that hi-la who-gen q'am-la=de? kin-gen=pst 'From whose clan was he?'

The pronoun *ce* 'what' (61) can also be used with the meanings 'how' (62), 'where' (63) and, when functioning as a nominal modifier, 'which, what kind of'. The dative case of this pronoun *celij* translates as 'why' (§4.5.2.4).


### **4.5.2 Other interrogative words**

### **4.5.2.1** *čina* **'where'**

This pronoun has an inherent spatial meaning and can be further inflected for the directional cases just like other nominals or adverbials with spatial meaning, see §3.4.2. Thus, we obtain:


4.5 Interrogative pronouns


It can also take the genitive suffix, then denoting origin in the sense of ethnic descent (67):

(67) čina-la where-gen admi=ja person=q iž? this gurži-le-r Georgia.obl-loc-abl / / gurži-la Georgia.obl-gen ca-w cop-m 'Where is this person from? (He) is from Georgia (i.e. he is Georgian).' (E)

### **4.5.2.2** *cet'le* **'how'**

The pronoun *cet'le* refers to the manner of action.

(68) marka rain cet'le how či-b-irʁ-ul=e spr-n-come.ipfv-icvb=q ixtːu? there.up neqːe-d cave.loc-npl d-i-d=q'al npl-in-npl=prt itːi those

'How can rain fall there? They are inside a cave.'

(69) cet'le how il this r-uč'-unne? f-learn.ipfv-icvb iž this ʡaˁħ-le good-advz r-uč'-un f-learn.ipfv-icvb ca-r cop-f 'How is she studying? She is studying well.' (E)

### **4.5.2.3** *kutːi* **and** *ceʁuna* **'which'**

The pronoun *kutːi* asks for the indication of a specific item among a group of items. For instance, the first speaker in (70) wants to indicate to his interlocutor a picture that does not fit into a picture story. The second speaker does not understand to which of the pictures the first speaker is referring, and asks for clarification. It can be used as an indefinite pronoun and then be inflected for various cases (72).

(70) hež this sːurrat picture heštːu here b-al n-fit b-ič-ib-le n-occur.pfv-pret-cvb akːu. .cop.neg kutːi? which 'This picture does not fit here. Which?'

### 4 Pronouns


The pronoun *ceʁuna* literally means 'like what, similar to what' and requests the hearer to provide more information about the manner or the type as in (73). In example (74), the indefinite pronoun modifies the following noun.


### **4.5.2.4** *cel* **and** *celij* **'why'**

These pronouns are case-inflected forms of *ce* 'what', more specifically ergative *cel* and dative *celij*, and the semantics of the case suffixes together with the base pronoun transparently explains the meaning 'why' (< 'what for'). The ergative is used to express agents and instruments, and the dative for the expression of causes.

(75) χalq'-li-j people-obl-dat il that cel why a-∅-jčː-aq-ul=de? neg-m-want.ipfv-caus-icvb=pst cellij why kax-ub=e kill.pfv-pret=q il that xːun-r-a-l? woman-pl-obl-erg 'Why did the people not love him? Why did the women kill him?'

### **4.5.2.5** *ceqːel* **'when'**

This interrogative adverb is used when asking for time points. It can occur in the genitive case without a change in meaning (77).

4.5 Interrogative pronouns


### **4.5.2.6** *čujna* **'how many times'**

This interrogative adverb refers to the frequency with which a situation occurs. To the same adverb the suffixes can be added that can also be added to multiplicative numerals when they are used for the formation of expressions referring to time points (79) (§6.5).


### **4.5.2.7** *čum* **'how many'**

The pronoun *čum* 'how many' is only used as a modifier to count nouns. It can be inflected with the dative yielding *čum-li-j* if no head noun is following (81). This form is used when asking for prices. Instead of directly adding the dative to the interrogative pronoun it can also be added to the head noun, e.g. *čum q'uruš-li-j?* (how many rubleobl-dat) 'for how many rubles?'.


The forms *čum-ib* and *čum-ibil* ask for ordinal numbers (§6.2):

(82) daže even hel-tː-a-j that-pl-obl-dat d-alχ-ul npl-know.ipfv-icvb akːu cop.neg hel-tːi that-pl čum-ib how.many-ord dusːi-cːe-d year.obl-in-npl d-elk'-un-ne=l npl-write.pfv-pret-cvb=prt hel-tːi that-pl 'Even they do not remember in which year (the pictures) were drawn.'

### 4 Pronouns

**4.5.2.8** *kusa* **'how much, how many'**

The pronoun *kusa* can be used together with count nouns or mass nouns (83) and without any head nouns (84–86). It also has the more specific temporal meaning '(for) how long' (85).


'No one is able to know for how long they were away.'

(86) kusa how.much ixʷle early ag-ur-re? go.pfv-pret-cvb 'How (much) early did he go?'

### **4.5.3 Interrogative pronouns used as indefinites**

Occasionally plain interrogative pronouns are used as indefinite pronouns as in the following example (87).

(87) nuˁq-be arm-pl ʡaˁbħ-ib get.tired.pfv-pret ca<d>i cop<npl> ∅-ik'-ul m-say.ipfv-icvb ca-w cop-m ij, this "ce what d-irq'-ul=da" npl-do.ipfv-icvb=1 'He says, "My hands got tired, I do something."'

### **4.6 Indefinite pronouns**

Sanzhi Dargwa has a rather wide range of indefinite pronouns that are regularly formed on the basis of the interrogative pronouns. Most of these pronouns make use of enclitics that are also otherwise used in the grammar as complementizers (*=jal*/*=el*, *=del*), emphatic particle (*=k'u*) or additive enclitic (*=ra*). The pronominal stems are normally inflected just like the interrogative pronouns, and then the derivational markers are attached.

4.6 Indefinite pronouns


For the formation of universal indefinites the quantifier *har* 'every' or more rarely *li<b>il* 'all' is used (§4.7).

### **4.6.1 Specific indefinite pronouns**

Specific indefinite pronouns (Table 4.12) are formed by means of the complementizer *=jal* (after vowels)/*=el* (after consonants), which is otherwise used in embedded questions (see §28.4) and certain epistemic modal constructions that have developed out of embedded questions and can be labeled "insubordination" (§28.4).

Table 4.12: Specific indefinite pronouns


Exemplary case forms of *ča=jal* and *ce=jal* are:


### 4 Pronouns

(90) qːarka Karka hetːu-b there-n b-už-ib hpl-be-pret čum=el how.many=indef bari day 'In Karka the bandits stayed for some days.'

There is a second series of specific indefinite pronouns with the emphatic enclitic *=k'u* (§9.4.5) that is used when the speaker does not remember a name of a person or thing and instead uses the indefinite as a kind of filler word. Of these pronouns *ce=k'u* (what-indef) is especially frequent and can be translated as 'whatchamacallit'.


### **4.6.2 Non-specific indefinite pronouns**

Non-specific indefinite pronouns are formed by adding *=del* to the interrogative base. This suffix is morphologically complex consisting of *=de* and *=(e)l*. The first part might originate from the past enclitic *=de*. The second part represents the enclitic used for embedded questions (§28.4) and also for the formation of specific indefinite pronouns (§4.6.1). The following examples illustrate reference to non-specific indefinite persons (95), (96) and places (97), (98).


4.6 Indefinite pronouns

hi-j=del who.obl-dat=indef b-ičː-ij n-give.pfv-inf 'My cousin, Old Kurban, may his sins be relieved, brought them for me to give them to someone.'


### **4.6.3 Free-choice indefinite pronouns**

Free-choice indefinite pronouns of the 'any' or 'WH-ever' type are formed by means of the suffix *-k'a* that does not serve any other function. In the majority of the cases the pronoun is followed by the verb form *b-iχʷ-ar=ra* (n-be.pfv-cond.3=add) that has a concessive meaning that can approximately be translated with 'even if it is' (§16.3). The verb form *b-iχʷ-ar=ra* mostly has the neuter singular prefix *b-* (suspended agreement), but it can also agree with the absolutive argument or even some other salient argument (see §20.2.4 for examples). In natural speech the suffix *-k'a* is also added to the Russian free-choice indefinite pronoun *lubuj-cːella-k'a-li-j* (any-comit-indef-obl-dat) 'for anything'. The following example (99) contains not only a Sanzhi free-choice indefinite, but also the Russian free-choice indefinite *kagda-nibud* 'whenever'.

	- '15,000 needs to be given to the doctors and every day these (i.e. this amount of money), however many days you stay.'

4 Pronouns

(102) du-l 1sg-erg at 2sg.dat ce-k'a what-indef b-ikː-ul n-want.ipfv-icvb haq-itːe=ra be.enough.pfv-2sg=add lukː-an=da=n give.ipfv-ptcp=1=prt 'I will give you whatever you may want.'

### **4.6.4 Negative indefinite pronouns**

In general, the negative indefinite function of indefinite pronouns is only available in clauses with negation. In affirmative clauses none of the pronouns described in this section has a negative indefinite reading, but readings such as free-choice indefinite or universal indefinite.

The suffix *-k'al* is used for the formation of indefinite pronouns that have the negative indefinite reading if they occur in a clause with negative polarity (103), (104). This suffix can be analyzed as consisting of *-k'a*, which forms free-choice indefinite pronouns (§4.6.3) and the enclitic used for embedded questions (§28.4) and also for the formation of specific and non-specific indefinite pronouns (Sections §4.6.1, §4.6.2).


Other meanings of pronouns with *-k'al* are free-choice indefiniteness if they are used in a conditional clause (105), (106) or non-specific indefinite if simply used in an affirmative clause (107–110).


'And like this also (Isakadi's) issues, things do not finish, forever, no matter where he went.'

4.6 Indefinite pronouns


'And together with the son they are going to sit outside anywhere.'

(110) har every ce-lla-k'al-li-j what-gen-indef-obl-dat čaˁʡir=ra wine=add d-al npl-match d-irč-iri npl-occur.ipfv-hab.pst 'Wine fitted well with any of these (types of food).'

Note that the word *cik'al* (from *ce* 'what' plus*-k'al*) has been lexicalized as a noun with the meaning 'thing'. At the same time it is still used as an indefinite pronoun with the meanings 'nothing' (in negative clauses) and 'something, anything' in positive clauses (105). It can also precede nouns as negative quantifier with the meaning 'no'.

Furthermore, the additive enclitic *=ra* (§9.4.1) is used for the formation of indefinite pronouns. If these pronouns occur in clauses with positive polarity the reading is universal indefinite (111), if they occur in clauses with negative polarity the reading is universal negative (112), and if they occur in concessive clauses the reading is free choice indefinite (113).


However, in practice such indefinite pronouns are (almost) never attested in natural texts. Instead, the enclitic *=ra* is usually preceded by *-k'al* (114) or occasionally *-k'a* (115) for the negative indefinite meaning.

### 4 Pronouns


Similarly, the emphatic enclitic *=č'u* can form negative indefinite pronouns when it is attached to the usual base (interrogative pronoun or numeral 'one') and used in clauses with negative polarity. As with the additive enclitic, in almost all examples that contain the enclitic it follows *-k'al* (and frequently *=č'u* is followed by the additive enclitic) (116), (117).


akːu, cop.neg ca=č'u=ra one=emph=add akːu, cop.neg ca=č'u one=emph akːu cop.neg Ašura Ashura ka-r-iž-ib down-f-sit.pfv-pret

'No, there was no one, except for Ashura he did not marry anyone.'

There are only two examples of pronouns with *=č'u* in clauses that do not have negative polarity such that the pronouns display their free-choice indefinite meaning. The first example in (118) illustrates the use of the pronoun in combination with the concessive auxiliary *b-iχʷ-ar=ra* as it has already been described for other free-choice indefinite pronouns (§4.6.3). The second example in (119) contains two pronouns with *=č'u*, of which the first has the free-choice indefinite reading whereas the second is a negative indefinite pronoun because of the negated verb.

(118) ca=ra one=add ci-k'al=č'u what-indef=prt b-iχʷ-ar=ra n-be.pfv-cond.3=add b-uk-an-ne, hpl-eat.ipfv-ptcp-fut.3 b-ik'-ul hpl-say.ipfv-icvb … 'saying that they will also eat no matter what' (lit. 'whatever it might be') (119) heštːi these deč-li drinking-erg b-učː-an-t-a-l hpl-drink.ipfv-ptcp-pl-obl-erg ci-k'al=č'u what-indef=prt ʡaˁħ-dex, good-nmlz iš-tː-a-l this-pl-obl-erg ce what b-irq'-u=ja? n-do.ipfv-prs=q ci-k'al=č'u what-indef=prt a-b-irq'-u

neg-n-do.ipfv-prs

'The ones who are drinking anything good, what do they do? They do not do anything (good).'

4.7 Universal indefinites and other quantifiers

Other negative indefinite pronouns are *caʔarra* 'no one' and *cajnara* 'never, not once' (*ca-jna=ra* one-time=add). The first pronoun consists of (*ca-ʔar=ra* one-?=add) and seems to be related the focus-sensitive particle *arrah* 'at least' (§9.4.5)

(120) caʔarra no.one kːuš hungry ∅-iχ-ub-il m-be.pfv-pret-ref akː-i cop.neg-hab.pst 'No one was hungry.'

### **4.7 Universal indefinites and other quantifiers**

Universal indefinites are normally not formed from interrogative pronouns, but by means of the quantifier *har* 'every' (or *li<b>il* 'all') plus a following noun:


Other quantifiers are *sukːil*, *li<b>il* 'all, whole, complete', *har*, *haril*, *harki*, *harkil* 'every', *b-aqil*,*ʡaˁbra*,*ʡaˁbra-b-al* 'much, many', and *kam* 'little, few'. The quantifiers treated in this section have most morphosyntactic properties that adjectives have and, as adjectives, normally occur before teh noun when they function as nominal modifiers. But just like adjectives and some other nominal modifiers they can also follow the noun under certain circumstances. See §21.1.3 for quantifier floating.

The quantifiers *sukːil* and *li<b>il* can both be used as attributes and they can be nominalized. When they are used as attributes of nouns in the plural they mean 'all'; with singular nouns they translate as 'whole, complete'. The quantifier *li<b>il* has a gender/number agreement slot and follows the agreement rules for adjectives and other nominal modifiers, i.e. agreement with the head noun.


### 4 Pronouns

From the quantifiers listed above, *har* can only be used attributively. All other quantifiers can also be nominalized. The head noun is in the singular, but mass nouns that trigger plural agreement are also possible if an interpretation referring to a specific quantity is available.


The quantifiers *b-aqil*, *ʡaˁbra*, *ʡaˁbra-b-al* 'much, many' also show gender/number agreement with the head noun in case there is any. Otherwise they express the gender and number of the item they are referring to.


The quantifier *kam* 'little, few, less' can modify nouns, it can be nominalized (by adding the cross-categorical suffix *-ce*; plural *-te* ) and it occurs in compound verbs with the meaning 'decrease, diminish, become less' (129).

(129) qːalmaqːar-te scandal-pl kam little d-irχ-ul npl-be.ipfv-icvb akːu cop.neg 'The scandals (i.e. fights) did not diminish.'

## **5 Adjectives**

### **5.1 Introduction**

Adjectives in Sanzhi can clearly be distinguished from nouns or verbs since they are not lexically specified for gender, and they cannot take tense suffixes or other inflectional morphology reserved for verbs. They are formally rather heterogeneous (§5.2). Sanzhi adjectives cover the typical semantic domains of this word class (1–12).


### 5 Adjectives


A few underived adjectives have agreement markers as can be see from the examples above. In addition, all derived adjectives containing the essive case plus *-il, -či-b* and *-bazi-b* and all constructions with *b-ah* (§5.3) also agree. Adjectives agree with the head noun in gender and number (13), (27a), and (35). More information on gender/number agreement rules is provided in §20.2.

(13) či-r-ix-ub spr-abl-take.off.pfv-pret ca-b cop-n halak-le fast-advz ca one b-uqen n-long q'aˁli branch 'He immediately broke one long branch (off a tree).'

Adjectives can be modified by adverbs, most commonly by degree adverbs that precede the adjectives, for example *c'aq'le* (14), *ħaˁq'le* 'very', *arindan* 'too, too much' (15), *b-aq* 'much', *bah* 'most' (39), *χːʷalle* 'largely', *q'ʷila, bara, kamle* 'little, few, a bit' (16).

(14) c'aq'-le very-advz χːʷal-le big-advz ʡaˁħ good ∅-iχ-ub m-be.pfv-pret ca-w cop-m 'He was very, very good.'

5.2 Adjectives and the cross-categorical suffixes *ce* and *il*


There is no derivational means of forming negative adjectives. Only participles used like adjectives can have a negative variant if the verbal negation prefix *a-* is added, e.g. *a-b-ucː-an* (neg-n-work-ptcp) 'inoperative, spoiled, not working'. Otherwise negation is expressed on the verb that heads the clause containing the adjective (see, e.g. §22.2 on copula clauses).

Adjectives usually precede the head noun, but the reverse order is also possible. Modifying adverbs, in turn, precede the adjective. §21.1.3 provides information about constituent order in the noun phrase.

### **5.2 Adjectives and the cross-categorical suffixes -***ce* **and -***il*

As is characteristic for Dargwa varieties, adjectives occur in the form of bare roots when they are used as attributes to nominals (13), (17). Many but not all of the adjectives in (1–12) belong to the class of adjectival roots. Some of these adjectives are also used in compounding, especially for the formation of compound verbs (§12.2.3), e.g. *aq b-ik'ʷ-ij* 'increase, enlarge, elevate, rise' (high n-aux.ipfv-inf).

(17) wahi evil admi person už-ib be-pret ca-w cop-m a-b-iχʷ-ar neg-n-be.pfv-prs 'Probably he was a bad person.'

The adjectival roots cannot be used substantively or predicatively. They must take the suffix *-ce* and can then fulfill all three functions: attribution (18), predication (15), (19) and reference (20). In the plural, -*ce* is replaced by *-te* (20).


### 5 Adjectives

When occurring in the canonical position before the head noun, adjectival roots and adjectives with the suffix *-ce* do not differ in their morphosyntactic or semantic properties. For example, both types of adjectives can modify coordinated noun phrases (21). This behavior differentiates Sanzhi Dargwa from other Dargwa varieties such as Tanti Dargwa or Standard (Akusha) Dargwa, for which syntactic differences between adjectival roots and the so-called "long" adjectives have been attested (van den Berg 2001: 26, Abdullaev et al. 2014: 207–208, Lander 2014).

(21) Sanijat-la Sanijat-gen kːalas-le-b class-loc-hpl ʡaˁħ(-te) good-dd.pl [durħ-ne=ra boy-pl=add rurs-be=ra] girl-pl=add χe-b exist.down-hpl 'In Sanijat's class there are good [boys and girls].' (E)

When nominalized, case suffixes are directly added to *-ce* if the nominalized adjective occurs in the singular (22). In the plural, the suffix *-t-a* (instead of *-t-e*) is used when case suffixes follow.

(22) nik'a-ce-li-j small-dd.sg-obl-dat darman medicine lukː-unne=w? give.ipfv-icvb=q 'Does he give medicine to the little one?'

The suffix -*ce* attaches not only to adjectival roots, but also to other parts of speech such as inflected nouns or verbs. Thus, its use is not restricted to adjectives, but it applies across a range of lexical categories. Generally speaking, it forms definite descriptions that function as referential attributes, and syntactically behave like nominals. A detailed description of the functions of -*ce* is given in §9.6.1.

Apart from the suffix *-ce* Sanzhi has another suffix -*il* for the formation of referential attributes that have similar morphosyntactic properties like items with -*ce*, but its application is far more restricted. Only two quantitative adjectives need the suffix *-il* in order to be used not only attributively, but also substantively or predicatively: *har-il* 'every' and *b-aq-il* 'much, many'. Furthermore, it is arguably a part of the quantifier *li<b>il* 'all', and when added to the preterite participle of the verb *ʔ-* 'say', the resulting verb form is used as a marker for ordinal numerals (§6.2), which are also adjectival in nature. More information on -*il* can be found in §9.6.2.

### **5.3 Formation of adjectival attributes**

Sanzhi does not have very productive means of forming new adjectives, but there are a few suffixes that take nouns as base and derive adjectives. Other ways of extending the lexicon is by means of genitive attributes and a special construction with the noun 'owner' (see below). Furthermore, participles are used and nowadays Russian adjectives also occur occasionally.

Sanzhi has a number of adjectives that are derived from nouns denoting body parts and personal qualities. These adjectives express the possession of this body part. The base noun is marked for plural and then the suffix *-ar* is added whereby the final vowel of the

### 5.3 Formation of adjectival attributes

plural suffix undergoes deletion (23). This suffix might be a cognate of the participle suffix of the copula *-ar* (§16.1). The adjectives form the plural by mean of the most common plural suffix *-te*. Two examples are provided in (24a) and (24b).

	- b. *qi-m-ar* 'horned' < *qime* 'horns'
		-
	- d. *laˁpː-ar* 'big-eared, having ears' < *laˁpːe* 'ears' e. *cul-b-ar* 'having (big) teeth' < *culbe* 'teeth'
		-
	- f. *ul-b-ar* 'having big eyes' < *ulbe* 'eyes'
	- g. *k'ult'-n-ar* 'pregnant' < *k'ult'ne* 'bellies'
	- h. *piš-n-ar* 'naughty boy, scamp' < *pišne* 'habits, tricks'
	- i. *ʡaˁmul-t-ar* 'talented' < *ʡaˁmulte* 'skills, talents'
	- b. ʁiz-b-ar-te hair-pl-adjvz-pl kːaˁt-ne cat-pl 'hairy cats' (E)

There are a few adjectives involving compounding with numerals and mostly plural nouns and the suffix *-(a)n*. As with the adjectives given in (25), the nouns occur in the plural. It might be the case that this suffix is a cognate of the modal/future participle *-an* (§18.1.2.2), the locative participle *-an* (§18.1.2.4) and/or the suffix *-an* that is used for the derivation of terms denoting inhabitants of particular villages and other places (§10).


Another type of derived adjectival attributes can be formed from adjectives denoting relational qualities. To the base adjectives the suffix *-gm-azi-gm* is added and the resulting adjectives denote an extreme quality. As can be seen in (26), the base can already be a derived adjective. The resulting adjectives occur in attributive, predicative and substantive function (27a–27d). In the predicative function the suffix *-ce* (-*te*) is required (27b).

	- < *b-aq* 'much, many' b. *kam-b-azi-b* 'very few, very little' < *kam* 'few, little'

### 5 Adjectives


< *ʡaˁbra* 'much, many'

	- b. hetːi those χalq' people ʡaˁħ-b-azi-b-te good-hpl-adjvz-hpl-dd.pl ca-b cop-hpl 'Those people are good.' (E)
	- c. ʡaˁbra-d-azi-d much-npl-adjvz-npl han remember d-irk-ur npl-occur.ipfv-pret 'I remember very well.'
	- d. cara-te other-dd.pl inžener-te=ra engineer-pl=add meχanekːa-be=ra mechanic-pl=add cara-te other-dd.pl šːan-te fellow.villager-pl k'e-b exist.up-hpl ʡaˁbra-b-azi-b much-hpl-adjvz-hpl hextːu-b=ra there.up-hpl=add 'There are other (experts like) engineers, mechanics, among the villagers, there are plenty.'

I found three adjectives with the suffix *-či* followed by a gender/number agreement marker (28). The base nouns are all loans. When the adjectives occur in predicative or nominal function they need one of the cross-categorical suffixes -*ce* or *-il* (29), (30).

(28) a. *dawla-či-b* 'rich' < *dawla* 'wealth'


'She is rich.' (E)

(30) žaniwar-t-a-lla animal-pl-obl-gen χʷal-le big-advz jaˁħ=ra, conscience=add namus=ra conscience=add b-už-ib n-be-pret ca-b, cop-n nišːa-lla 1pl-gen dawla-či-b-t-a-lla-ja-r wealth-adjvz-hpl-pl-obl-gen-loc-abl

'The animals had apparently more consciences than our rich (people).'

For Standard Dargwa,Abdullaev et al. 2014: 212 have claimed that -*či* is a cognate of the spatial postposition *či* 'on, above' . However, for Sanzhi this cannot be true because the 5.4 Comparative constructions with adjectives

postposition *či* can be used with both native and loan words and it governs the loc-series or, alternatively, the genitive case (§8.1.7). The adjectivizer -*či* can also not be equated with the Turkic loan suffix -*či* (§3.5.1), which derives agent nouns because nouns do not inflect for gender and the agent nouns do not need any further suffixes in order to be used in argument position or as predicates. Furthermore, in Standard Dargwa the form of the adjectivizer is -*če*, but the form of the borrowed nominalizer is -*či* throughout all Dagestanian languages.

Nouns denoting materials and other properties or adverbs and nouns with temporal semantics can be inflected for the genitive and then yield the meaning of relational adjectives (31) (see also §3.4.1.3 for more examples).


It is possible to express attribution with a possessive construction consisting of a noun in the genitive denoting the possessed and the noun *b-ah* 'owner'.<sup>1</sup> This construction represents a standard genitive phrase. The noun agrees with the head noun in gender and number. These constructions can occur as predicates (33) and as attributes (34).

	- b. *č'imi-la b-ah* 'having a tail'
	- c. *abrazovanie-la w-ah* 'educated' (from Russian *obrazovanie* 'education')

'They are searching for who among them is guilty.'

(34) er look w-ik'-ul m-look.at.ipfv-icvb ca-w. cop-m ca one ul-la eye-gen b-ah n-owner šajt'an devil ka-b-isː-un-ne down-n-sleep.pfv-pret-cvb

'He is looking around. The devil with one eye is asleep.'

### **5.4 Comparative constructions with adjectives**

Comparative constructions can express (i) equality or similarity, (ii) comparative, and (iii) superlative. Similarity or equality can be expressed by means of the adverbs *daˁʡle* 'like, as' (35), *mišil* 'similar' or the enclitic *=ʁuna* (36) as well as through manner adverbs with the meaning 'like this, like that' (37).

<sup>1</sup>This is one of the very few nouns that has a gender marker. See §3.1 for more information.

### 5 Adjectives


Adjectives do not have a special comparative form. Instead, the standard of comparison takes the loc-ablative suffix (38).

(38) atːa-ja-r father-loc-abl χːula-te=ra big-dd.pl=add b-irχ-i hpl-be.ipfv-hab.pst 'There were (brothers) older than grandfather.' (lit. father)

The superlative is formed by means of the degree adverb *bah* 'most' (emphatic variant *bahlalla*) that occurs before the adjective (39) (for other degree adverbs see §5.1 and §7.4).

(39) bah most χːula-ce big-dd.sg w-irχ-i=w m-be.ipfv-hab.pst=q χatːaj grandfather ču-la refl.pl-gen ucː-b-a-cːe-r brother-pl-obl-in-abl 'Was (our) grandfather the oldest among his brothers?'

More details and additional examples of comparative constructions can be found in §30.1.

## **6 Numerals**

Sanzhi has (i) cardinal numerals (§6.1), (ii) ordinal numerals (§6.2), (iii) distributive numerals (§6.3), (iv) group numerals (§6.4), (v) multiplicative numerals (§6.5), and (vi) collective numerals (§6.6).

Most of the numerals have the morphosyntactic properties of adjectives or occasionally adverbs. Generally, numerals can be used as nominal modifiers with a following noun in the singular. For verbal agreement the noun phrase is nevertheless treated as plural §21.1.3. In this chapter, I also treat some other numeral expressions and basic ways of counting (§6.7). Quantifiers such as 'all' are treated in §4.7 together with indefinite pronouns.

### **6.1 Cardinal numerals**

The cardinal numerals 1–101 are given in Table 6.1. All numerals except for *ca* 'one' are morphologically complex, containing a root and a derivational suffix. The numerals 2 to 10, 20, as well as 100 are formed by means of the suffix *-al* (allomorph *-jal* after vowels). The decimal numerals 10 and 30–90 are built by adding the suffix *-c'al* to the roots. When decimals and the numerals 1–9 are combined, both the decimals and the numerals 1–9 take suffixes. On the decimals *-al* is replaced by *-nu*/*-anu*, e.g. *wec'-al* '10' and *wec'-nu*, *wer-c'-al* '70' and *wer-c'-anu*. To the numerals 1–9 the suffix *-ra* is added.

The cardinal numerals for hundreds and thousands are provided in Table 6.2. Complex numerals containing hundreds need the derivational suffix *-lim* added to *darš* '1000'. For the higher cardinal numerals (millions, billions, etc.) the Russian terms are used.

Cardinal numerals are used in counting and as modifiers of nouns in noun phrases. In the latter function the noun appears in the singular form, but it controls plural agreement on the verb (3), (4). Examples of cardinal numerals in use are (1–4).


'I would have given him money, not just 8,000, but 80,000.'



### Table 6.2: Cardinal numerals 100–20,000


6.2 Ordinal numerals


Cardinal numerals can be nominalized. Case endings are directly added to numerals ending with a consonant. With numerals ending in a vowel an oblique marker *-l* sometimes precedes the case suffixes; see Table 6.3. Examples are given in (5–6).


Table 6.3: Inflectional paradigms of selected cardinal numerals


### **6.2 Ordinal numerals**

Ordinal numerals are formed by adding the suffix *-ʔib-il* (allomorph *-ʔubil* with the stem of the numeral 'four', which contains a labialized consonant) or its short variant *-ʔib*. The first part of this suffix originates from the root of the verb 'say', which is *-ʔ-* plus the preterite suffix *-ib*. The second part *-il* in the long variant is the cross-categorical

### 6 Numerals

suffix *-il*, §9.6.2). Similar ways of forming ordinal numerals have been reported for other Dagestanian languages (e.g.Lezgian, see Haspelmath 1993: 233; Akusha Dargwa, see van den Berg 2001: 30 fn.10; Hinuq, see Forker 2013a: 401–403).



Ordinal numerals are inflected just like any other nominal, e.g. *caʔibil* 'first', ergative *caʔibil-li*, genitive *caʔibil-la*, dative *caʔibil-li-j*, in-lative *caʔibil-li-cːe*, and so on.



Ordinal numerals can also form the plural. In this case, the final *-il* part is omitted because this suffix is not compatible with plural referents (§9.6.2). Example of plural ordinal numerals are *ca-ʔib-te* 'the first ones', *k'ʷi-ʔib-te* 'the second ones', *ʡaˤb-ʔib* 'the third ones', etc. The oblique plural is formed according to the regular pattern of plural nominals with the suffix -*te*, i.e., by using -*ta*, e.g. the ergative form of 'the first ones' is *ca-ʔib-t-a-l*.

6.3 Distributive numerals

### **6.3 Distributive numerals**

Distributive numerals are formed by reduplicating the root. Optionally the suffix *-l(e)* follows the reduplicated numeral; see Table 6.5. The suffix *-l(e)* seems to be the adverbializer (cf.§9.6.3). Note that with distributive numerals the modified noun bears overt plural marking (11).

Table 6.5: Distributive numerals



### **6.4 Group numerals**

Group numerals are formed by adding the suffix *-gm-a* to the root; see Table 6.6. The gender marker shows only plural agreement (*-b* or *-d*). The suffix can also be added to the quantifier *b-aq* 'many, much'.

Group numerals denote groups or pairs of items. In my corpus, only the group numeral of *ca* 'one' is used with human plural agreement (*ca-b-a*) and its meaning is very similar to the indefinite pronoun 'some(one), somebody'. The human plural form *ca-b-a* is also used as reciprocal pronoun (see §4.4 for the case paradigm and §29.2 for one example). All other group numerals carry the neuter plural suffix *-d* (13–15). Group numerals can be attributes of nouns, which normally occur in the plural (13). Group numerals can be inflected by adding case suffixes to the suffix *-gm-a* (16). They can also take the attributive plural suffix *-te* (oblique form *-t-a-*) as in the same example.

(13) wec'-d-a ten-npl-group ʡaˤčkːa-be=ra glass-pl=add sa-d-uc-ib ante-npl-keep.pfv-pret '(The doctor) tried (with me) ten pairs of glasses.'

### 6 Numerals

Table 6.6: Some group numerals


(14) u 2sg či-r-až-ib-la, spr-f-see.pfv-pret-post wer-d-a seven-npl-group hak'-ub-le appear.pfv-pret-cvb pikru-me thought-pl le-d

exist-npl

'After I saw you, many (lit. seven groups of) thoughts arose.'


### **6.5 Multiplicative numerals**

Multiplicative numerals are formed by means of the suffix *-na* (*-jna* after vowels) that is added to the root; see Table 6.7 and (17). This suffix can also be added to the interrogative pronoun *čum* 'how much, how many' plus the suffix *-ra*, and then leads to the indefinite pronoun 'how often ever, many times' (§4.5.2.6).

(17) ʡaˤj-na three-time ag-ur-ce=de, go.pfv-pret-dd.sg=pst ij this dam 1sg.dat a-ag-ur neg-go.pfv-pret caj-na one-time arrah at.least

'He went three times (to the Hajj), to me this happened not even once.'

### 6.6 Collective numerals


Table 6.7: Some multiplicative numerals

From the multiplicative numerals expressions referring to time points can be formed by means of various derivational and inflectional suffixes. The words in (18) all mean '(at) the second time'.<sup>1</sup> Examples from texts are presented in (19–20).


(19) k'ʷi-jna two-time tːura outside w-erč-ib m-lead.pfv-pret il that nišːa-la 1pl-gen qili-rka home-abl ʡaˤj-na-la-li-j three-time-?-obl-dat tːura outside w-erč-ib m-lead.pfv-pret 'Two times (he) send him away, our (guy), out of the room, the third time (he) sent him away.'

(20) k'ʷi-jna-le two-time-loc ħaˤžitːaj=de, Hazhittaj=pst "w-alli m-together ka-Ø-jž-e," down-m-remain-imp Ø-ik'ʷ-ar, m-say.ipfv-prs "hež-i-cːella this-obl-comit w-alli!" m-together

'The second time there was Hazhittaj saying, "Sit together, with him together (in the back of the car)!"'

### **6.6 Collective numerals**

Collective numerals are formed by adding the additive enclitic *=ra* to the cardinal numerals; see (21). They can function as attributes of nouns (22) or they can be nominalized and then occur on their own (23) or be modified by demonstrative pronouns (24).

<sup>1</sup> In (18), although the suffix *-lla* in two of the given words strongly resembles the genitive, it is, at least synchronically, distinct from the case marker, since it is possible to add the genitive to an adverb with *-lla*, e.g. *k'ʷi-jna-lla-la* 'of the second time'.

### 6 Numerals


## **6.7 Other numeral expressions and compounds involving numerals**

Expressions for fractions are given in (25) and (26–27); *but'a* translates as 'piece, part' into English. The word *b-abq'i* 'half' agrees in gender and number with its head noun if it occurs in attributive function (26), (27).


'Then they went and went, and when they went half of the way, …'

(27) ʡaˤbal=ra three=add b-abq'i n-half dus year ʡaˤrmija-cːe-w army-in-m kelg-un remain.pfv-pret 'He spent three and a half years in the army.'

The Sanzhi terms for the school grades are formed by adding *-la* (*-lla* after vowels) to the root of the numerals 1–5 (with 5 being the best grade and 1 the worst): *calla* 'one', *k'ʷilla* 'two', *ʡaˤbla* 'three', *aʁʷla* 'four', *xulla* 'five' (28).

(28) ce what b-ič-ib=e n-get.pfv-pret=q at? 2sg.dat dam 1sg.dat b-ič-ib n-get.pfv-pret xulla five 'What did you get (i.e.which grade)? I got a five.' (E)

### 6.7 Other numeral expressions and compounds involving numerals

Other words that are derived from numerals are the terms *k'ʷidarq'i* 'twins' and *ʡaˤbdarq'i* 'triplets' (from the numerals 'two' and 'three' and the verb *b-arq'-ij* 'do, make'). Then there are terms for traditional events and rituals that occur after the death of a person, namely,


'After three days (people) give (alms), after 40 days, after one year, four dhikrs.'

Compound nouns and adjectives can contain numerals, e.g., *k'ʷi-dusː-an k'ašːa* 'twoyear old bull', *ʡaˤb-daˤrχ-la qul-be* (three-floor-gen house-pl) 'three-floor houses' (see §3.6.3 and §5.3 for more examples).

Counting is exemplified in (31) and (32).

(31) k'ʷel-le two-loc ʡaˤbal three či-ka-b-ix-ar spr-down-n-throw.pfv-cond b-irχʷ-u n-become.ipfv-prs arg-u go.ipfv-prs xujal

'Two plus three equals five.' (lit. if you throw three onto two five happens)

(32) weral-li-cːe-r seven-obl-in-abl gu-r-h-asː-ar sub-abl-up-take.pfv-prs či-r-h-asː-ar spr-abl-up-take.pfv-prs ca one arg-u go.ipfv-prs urekːal six

'Seven minus one equals six.' (if you take away one from seven it goes six)

five

## **7 Adverbs**

In this chapter, spatial §7.1, temporal §7.2, manner §7.3, and degree adverbs §7.4 are described as well as the productive formation of mostly manner adverbials by means of the suffix *-le* §7.5. Adverbs form a rather heterogeneous group in Sanzhi and only certain subclasses of spatial adverbs and manner adverbs have been derived by specialized adverbializing suffixes.

### **7.1 Spatial adverbs**

### **7.1.1 Spatial adverbs derived from demonstrative pronouns**

Several series of spatial adverbs can productively be derived from demonstrative pronouns. The major derivation pattern is the suffixation of *-tːu* to the stem of the pronouns (Table 7.1). The full list of the respective base pronouns is given in §4.2. The meanings of the spatial adverbs are plainly based on the meaning of the base pronouns, showing that their semantics is organized along the meaning components of demonstrative pronouns (§4.2.1, §4.2.2):


As for proximity, there is a three way distinction (see the adverbs in the first three lines of Table 7.1). Elevation distinguishes three meanings, of which 'above' and 'below' are expressed by dedicated stems (see the last three lines in Table 7.1) whereas all remaining adverbs are used when the meaning 'level' is intended. The third meaning component is expressed via the distinction of the word-initial syllable (the three columns*i(C)-* vs. *he(C)* vs. *hi(C)-* in Table 7.1). The adverbs of the type *he(C)tːu* given in the second column of the table are predominantly used when referring to the immediate geographical surroundings of the speaker (and addressee), when the conversation is about spatial reference points that have been mentioned before, are assumed to be known by the participants or are part of the personal sphere of the speaker (1), (2). In contrast, the *i(C)tːu* adverbs in the first column are commonly used when new spatial reference points are introduced or when talking about reference points whose location is unknown or irrelevant (3), (4). The adverbs of the *hi(C)tːu* type given in the third column occur only seldom in my corpus so that I am not able to make any generalizations about their meaning. Note also that

### 7 Adverbs

there are two series of adverbs with the identical meaning, being formally differentiated only by the stem consonant (*x vs. k'*). The adverbs containing *x* are far more frequently used than the adverbs with *k'*, which might even represent code switching to another Dargwa dialect.

Table 7.1: Spatial adverbs derived from demonstrative pronouns


(1) [talking about a gasoline station]

ag-ur-re go.pfv-pret-cvb heltːu-rka, there-abl badra bucket salaˁrk'a-la fuel-gen k-aqː-ib=da down-carry-pret=1 'I went away from there carrying a bucket of fuel.'


'There is nobody there without a drink in the hands.'

A second series of spatial adverbs denoting the source is derived by means of the suffix *-ka* (5) (Table 7.2). This suffix is probably a cognate of the second part of the complex ablative suffix *-r-ka* (§3.4). These adverbs can also have a temporal interpretation 'from time X on, after time X' in addition to the spatial meaning (6). As can be seen in the table, the adverbs in the first two lines have the same meaning because the base pronouns are synonyms.

7.1 Spatial adverbs


Table 7.2: Spatial adverbs denoting the source

(5) kat=q'ar down=mod ka-r-ilsː-a-di down-f-lay.ipfv-hab.pst-1 ij-ka=ra this-abl=add kːancːupːe ladder hej-ka=ra this-abl=add kːancːupːe ladder iχ-ka=ra dem.down-abl=add

'As for lying, I lay, but (there are) stairs from here, stairs from here, and also from there.' (The speaker complains that staying in the hospital is difficult for her because in order to go to the toilet she has to take the stairs)

(6) hej-ka this-abl ʡaˁħaˁd=ra Ahad=add suk meet ∅-ič-ib m-occur.pfv-pret ca-w cop-m qːačaʁ-la bandit-gen qal-sa-w house-ante-m di-la 1sg-gen durħuˁ=ra boy=add

'After this Ahad also met my son in front of the house of Kachar (lit. 'bandit')).' (Kachar is the nickname of a man)

Both series of adverbs can be inflected for the directional cases in the same way as nominals are inflected, but since the adverbs denoting source already express movement, they cannot take the essive case (Table 7.3). The ablative of the pronouns in this table can also express temporal meaning, for instance *heltːu-rka* (there-abl) 'then'.

Table 7.3: Inflectional paradigms of two spatial adverbs


A third series of spatial adverbs has the meaning 'from X to X'. It is formed by means of the complex suffix *-k-itːu-b-a* (7). The suffix is a combination of the ablative *-ka* (shortened to *-k*), the locational suffix *-tːu* and the directional marker *-gm-a* (§3.4.2.8). The last

### 7 Adverbs

suffix is, in principle, optional, although there are no examples without it in my corpus. According to Sanzhi speakers, the resulting complex adverbs are actually a short variant of combining the adverbs in Table 7.2 with the adverbs in Table 7.1, for example *hetka* + *hetːuba* > *hetkitːuba*. However, the suffix as a whole can also be added to other nominal bases such as personal pronouns, common nouns or personal names if they are inflected for the loc-ablative case first, such as *nušːa-le-r-kitːu-b-a* (1pl-loc-abl-advz-ndir) 'from us further away', *uškul-le-r-kitːu-b-a* (school-loc-abl-advz-n-dir) 'from the school further away'. The series is also available from the other two pronominal stems*iC* and *hiC*, but in my corpus there are only examples of the adverbs from the *heC*-pronouns given in (7–8), (9), (14).

	- a. *hež-kitːu-b-a* 'from here (= place of speaker) to there'
	- b. *hej-kitːu-b-a* 'from here (= place of speaker) to there'
	- c. *hel-kitːu-b-a* 'from there (= place of the addressee) to there'
	- d. *het-kitːu-b-a* 'from there (= unspecific place) to there'
	- e. *hek-kitːu-b-a* 'from above to there'
	- f. *heχ-kitːu-b-a* 'from down to there'
	- a. *hež-kitːu-rka* 'from here to there, past, by'
	- b. *hej-kitːu-rka* 'from here to there, past, by'
	- c. *hel-kitːu-rka* 'from there to there, past, by'
	- d. *het-kitːu-rka* 'from there to there, past, by'
	- e. *hek-kitːu-rka* 'from above to there, past, by'
	- f. *heχ-kitːu-rka* 'from down to there, past, by'

'if our sons go from here to there to Nizhnekamensk, …'

Finally, there is a spatial adverb *itille* 'further, to the side, sideways' that seems to be the pronoun *it* inflected for the locational suffix *-le* (10).

(10) nišːa-la 1pl-gen šːi-la village-gen itille-b further-n musːa place te-b exist-n 'At that side of the village there is a place.'

7.1 Spatial adverbs

### **7.1.2 Spatial adverbs related to postpositions**

All spatial postpositions discussed in §8.1 can also be used adverbially without a dependent noun phrase (11). Some of them have not only spatial, but also temporal semantics. They can inflect for all spatial cases expressing direction/movement (essive, lative, ablative, directional). A few examples are provided in (12), (13).


'"Now," the fox says to the wolf, "you sit down in front, and I behind!"'

(13) ank'luʁi-la Anklukh-gen šːi village ʡaˁħ-le good-advz qari-b=q'al up-n=mod 'The village of Anklukh is pretty high up.'

There are four spatial adverbs that have been derived from spatial postpositions by means of suffixing *-tːi* to the root: *gu-tːi* 'along downside, at the lower side' (< *gu* 'down, under'), *či-tːi* 'along upside, at the upper side' (<*či* 'on'), *sa-tːi* 'at/along the front, as soon as' (< *sa* 'in front, ago'), and *b-i-tːi* 'inside, through' (< *b-i* 'in, inside') (14).

(14) di-la 1sg-gen qu-la garden-gen hetkitːu-w-a, from.there.to.there-m-dir čitːi along.up gutːi along.downside ∅-iχ-ub-le m-be.pfv-pret-cvb w-erč-ib m-lead.pfv-pret ca-w cop-m Izbir-re Izberbash-loc 'Through my garden, up, down, they brought him to Izberbash.'

There are few more adverbs based on the adverbs/postpositions, namely *hitːille* 'on the back, later' (< *hitːi*), *b-atːura* 'from inside' (< *tːura*), and *qaršːa* 'upper side (of the village)' (< *qar* 'at/on the top' plus the loc-form of the noun *šːi* 'village', which is *šːa*).

### **7.1.3 Other spatial adverbs**

Sanzhi has some more spatial adverbs of which the most important ones are given in (15). A few of them are formed by means of the adverbializing suffix *-le* (§9.6.3). For spatial adverbs that have the meaning of indefinite pro-forms see §4.6.

### 7 Adverbs

(15) *kat'* 'down' *bet* 'there' *šːulum* 'by, past' (< *šːal* 'side') *sat* 'here' *alaw* 'around, in a circle' *bet-sat* 'here and there' *qili* 'at home' (< *qal* 'house') *guq-le* 'low' *haraq-le* 'far' *hek-le* 'close, near' *b-arx-le* 'directly, straight' *kʷi* 'in the hands'

### **7.2 Temporal adverbs**

Many of the spatial adverbs/postpositions listed in (11) also express temporal meaning. The adverb/postposition *gu* has the somewhat unexpected meaning 'before, in earlier times' when suffixed with the frozen neuter plural agreement suffix *-d* (16).


Adverbs for times of the day are given in (17). Deictic temporal adverbs expressing relative time in days and years can be found in (18), (19), and seasonal adverbs in (20). Some of the adverbs in (17) and (20) are formed by adding the genitive case suffix to a base noun.

	- b. *arilla* 'at midday, at lunch time' (< *ari* 'daytime')
	- c. *nisnalla* 'after lunch, afternoon'
	- d. *ʁerilla* 'early evening, at sunset' (< *ʁeri* 'sunlight')
	- e. *daˁrχːaˁlla* 'in the evening' (< *daˁrχːaˁ* 'evening')
	- f. *dučːilla* 'at night' (< *dučːi* 'night')
	- b. *hati sar bar* 'three days ago' (< *hati* 'more' + *sa-r* + *bar*)
		- c. *sar bar* 'two days ago' (< *sa-r* + *bar*)
		- d. *sːa* 'yesterday'
		- e. *ižal* 'today' (< *iž* 'this')
		- f. *čːaˁʡaˁl* 'tomorrow, morning'
		- g. *carabal* 'day after tomorrow' (< *ca-ra* 'one=add, other + ?)
		- h. *xujal bar hitːille* 'in five days' (< *xujal* + *bar* + *hitːi-lle* after-advz)
	- b. *hati sar dus* 'two years ago' (? < *hati* + *sa-r* + *dus*)
	- c. *sar dus, irig, gur dus, hit dus* 'last year' (< *gur* 'away', *hit* 'that')

7.3 Manner adverbs

	- b. *hanišalla* 'in summer' (< *haniša* 'summer')
	- c. *ibxnella* 'in autumn' (< *ebx* 'autumn')
	- d. *ganilla* 'in winter' (< *ga* 'winter')

Some more temporal adverbs are provided in (21). For temporal adverbs that have the meaning of indefinite pro-forms see §4.6. For the expression of dates and the time see the descriptions of the various spatial cases that fulfill these functions in §3.4.

	- b. *barežij* 'the whole day' (< *bar* 'day' + ?)
	- c. *ixʷle* 'early, fast'
	- d. *ixʷbel* 'long ago' (< *ixʷ* 'early' + ?*b-el* n-remain)
	- e. *q'anne* 'late' (< *q'an-ne* late-advz)
	- f. *ha* 'now, already'
	- g. *na* 'now, already'
	- h. *hana* 'now, then' (< *ha* + *na*)
	- i. *c'il(i)* 'then'
	- j. *heba* 'then, later' (?< *he-b-a* that-n-dir)
	- k. *cacajnaqːel* 'sometimes' (< *ca-ca-jna=qːel* one-one-time=when)
	- l. *cacaqːella* 'sometimes' (< *ca-ca=qːella* one-one=when)
	- m. *urkːa-urkːab* 'sometimes' (< urkːa-urkːa-b middle-middle-n)
	- n. *mah-mahle, raχ-raχle* 'sometimes, rarely' (< *raχle* 'if')

### **7.3 Manner adverbs**

Sanzhi has a productive way of deriving adverbs of manner from demonstrative pronouns. These adverbs have the meaning 'like this/like that'. They are formed by adding the suffix *-itːe* to the demonstrative pronouns in the singular (Table 7.4). Their meaning is again transparently built on the semantics of the demonstrative pronouns as described in detail in §4.2 and summarized in §7.1.1 above.

Since these adverbs are used when an action or event is compared to another event, the adverbs based on the *heC*-series in the second column are far more common than those from the other two series given in the first and in the third column. The most frequent forms in my corpus are *hel-itːe* 'like that, away from the speaker and/or close to the hearer' and to a lesser extent *hež-itːe* 'like this, like something close to the speaker' (22), but other forms such as *het-itːe* 'like that, like something further away, unspecific distance', *hek'-itːe* 'like this/that above' and *heχ-itːe* 'like this/that below' as well as very few occurrence of *ižitːe* (23), *hilitːe* (24), and *ilitːe* are also attested.

### 7 Adverbs

Table 7.4: Manner adverbs derived from demonstrative pronouns



There is another rather small group of four manner adverbs with a similar meaning that are also derived from demonstrative pronouns: *itwaj*, *hetwaj*, *hitwaj*, and *ižwaj* 'like that, and so'. Their usage is illustrated in (25), (26).


Other manner adverbs are usually formed by suffixing *-le* to a root, for example *bahlal* 'slowly', *halak-le* 'fast', *χʷal-le* 'greatly, much, a lot', *imanne* 'patiently', *ʡaˁħ-le* 'well', and so on. This is described in the next section (see also §9.6.3).

### **7.4 Degree adverbs**

Adverbs of degree express the degree of a quality and modify adjectives or other adverbs. They precede the modified item (28). From the formal perspective, degree adverbs are a

7.5 Formation of adverbials with the suffix*-le*

heterogeneous group of items. Some are simple stems (*arindan*, *bara*), but most of them contain the adverbializing suffix -*le* also used to derive manner adverbs (27) (§7.5); for comparative constructions involving degree adverbs see §30.1.


### **7.5 Formation of adverbials with the suffix** *-le*

Manner adverbs and some other adverbs are easily derived by means of the suffix *-le* (allomorphs *-l* after vowels, *-re* after *r*, *-ne* after *n*, and occasionally -*lle*). This suffix is attached to the underived short adjectives (13), (22), (29) (§5.2) and to nouns in the absolutive or in spatial cases (30) as well as to spatial adverbs bearing the essive case (§9.6.3). The same suffix is added to verbs (usually bearing the preterite or the imperfective converb suffix) in order to form simple converbs (§18.1.1). More examples of its use can be found in §9.6.3.

(29) deadjectival adverbs


(31) other adverbs

*xurc-le* 'barefoot' (\**xurc*; *xurχ b-iχʷ-ij* 'become barefoot')

## **8 Postpositions**

Sanzhi has spatial and non-spatial postpositions. Some of the spatial postpositions also have temporal readings. The majority of the spatial postpositions are widely used as adverbs and then occur without a dependent noun phrase (§7.1.2). Thus, the distinction between postpositions and adverbs is rather blurred. The distinction between postpositions and spatial cases is, by contrast, relatively clear-cut with respect to the morphosyntax, although there are no clear intonational and often also no clear semantic differences. Most postpositions govern the genitive case; otherwise two spatial cases or the absolutive case are used (Table 8.1). This is in contrast to spatial cases, which are suffixed directly to the nominal stem or to the oblique/ergative suffix. Furthermore, only the postposition *sa* has a clear cognate form used as spatial case (§3.4.2.6). The postposition *sa* is shown in (1a); examples with the cognate spatial case are given in (1b) and (1c).

	- ∅-ik'ʷ-ar m-say.ipfv-prs

'In front of his house there is also a good area, he says.'

c. it that ca-w=ra refl-m=add hel-i-sa that-obl-ante sa-ka-jsː-un-ne ante-down-lay.m.pfv-pret-cvb 'he himself also slept in front of it (a horse, in order to watch over it)'

ʡaˁħ-ce good-dd.sg ca-b cop-n

There is another class of morphemes with which postpositions formally and semantically overlap, namely spatial preverbs. The postpositions *sa*, *hitːi*, *či*, *b*-*i* and *tːura* also occur as location preverbs (§11.6.1) that can be combined with the postposition or case marker or occur on their own. In example (1c) both the spatial case *-sa* and the preverb *sa-* are used.

This chapter explores spatial postpositions (including those with temporal meanings) (§8.1) and non-spatial postpositions (§8.2).

### **8.1 Spatial postpositions**

Table 8.1 displays the spatial postpositions and the cases they govern (in the last column). Most postpositions govern the genitive case, which is typical for Dargwa varieties. Postpositions can be inflected for directional cases. The inflected postpositions in brackets can be elicited, but are not commonly used.

### 8 Postpositions

Table 8.1: Spatial postpositions


### **8.1.1** *sala* **'in front of'**

The postposition *sala* has only spatial meaning, but the cognate adverb has spatial and temporal readings (e.g. *salar(ka)* 'formerly, in former times'). It governs the genitive.


b. χːuˁrba-la tomb.obl.pl-gen sala-b in.front-n pirma farm te-b=uw? exist-n=q 'Is there a farm in front of the graveyard?'

### **8.1.2** *sa* **'in front, ago'**

The postposition *sa* is a cognate of *sala*. When having a spatial reading it governs the genitive (3a). With the temporal meaning 'ago it governs' the absolutive (3b). The ablative *sar(ka)* is more commonly used as temporal adverb with the meaning 'before, earlier, until' (§7.2).

	- b. χatːaj-la grandfather-gen xːunul woman r-ebč'-ib-il=de f-die.pfv-pret-ref=pst ʡaˁb-c'al three-ten dus year sa-r ago-abl 'Grandfather's wife died 30 years ago.'

8.1 Spatial postpositions

### **8.1.3** *hila* **'behind, after'**

The postposition *hila*, which governs the genitive, has spatial and occasionally temporal uses. It is sometimes followed by *hitːi* and it is widely used as a spatial and temporal adverb (§7.1.2).

	- b. tum-la hill-gen hila-b behind-n srazu immediately majdan field ʁuna eq k'e-b exist.up-n 'There behind the hill there is immediately something like a field.'
	- c. marka-la rain-gen hila-b behind-n cuχaˁb rainbow 'a rainbow after the rain' (E)

### **8.1.4** *hitːi* **'after, behind'**

This postposition has spatial and temporal semantics. There are examples that allow for both readings, e.g. (5b) is a sentence from the *Family Problems Picture Task* (San Roque et al. 2012) and it can refer to the spatial ordering of the pictures on the table or to the temporal ordering of the events that the pictures are illustrating. There are two examples with spatial meaning: in both examples *hitːi* is preceded by *hila* and thus it might be *hila* that, in fact, provides for the spatial interpretation (5c). It governs the genitive and mostly occurs with a preceding demonstrative pronoun and the meaning 'after this/that', e.g. *hežila hitːi*. There are also lexicalized variants of such phrases, e.g. *helila hitːi* > *helilitːi*.

(5) a. caj-na one-time marka-la rain-gen hitːi after če-r-uq-un spr.up-f-go.pfv-pret ca-r cop-f qar up qal-sa house-ante qʷaˁrš sweep b-arq'-ij n-do.pfv-inf

'Once after the rain (she) went up to sweep in front of the house.'

b. het-i-la that-obl-gen hitːi after ka-d-irxː-an-te down-npl-put.ipfv-ptcp-dd.pl ca-d cop.npl heštːi these d-ilʡ-aˁn-te npl-steal.ipfv-ptcp-dd.pl

'After this (one) must put these, these where they steal.'

c. urči-la horse-gen žilixʷa-la saddle-gen hila behind hitːi behind b-iχ-un n-tie.pfv-pret '(He) tied it behind the saddle.'

### 8 Postpositions

In addition, *hitːi* is widely used as a temporal adverb (§7.2), including temporal adverbial clauses (§18.2.5), and the short encliticized version *=itːi* occurs within compound verbs (§12.2.2, example 20).

### **8.1.5** *xːar(i)* **'down, at the bottom, under'**

This postposition has exclusively spatial meaning 'to the bottom, down, under' and governs the genitive (6a–6c). It is semantically close to the spatial case *-gu* (§3.4.2.5) and the spatial case marker can be suffixed to the postposition in which case the meaning is solely 'under'.

	- b. w-ag-la m-waist-gen xːar down w-aˁħ-un-ni=de m-get.wet.pfv-pret-msd=pst 'From the waist down I (masc.) was wet.'
	- c. heχ dem.down urx-m-a-la sea-pl-obl-gen xːari down balnicːa-b-a-j hospital-pl-obl-dat r-ax-ul f-go.ipfv-icvb 'I was going to the hospital down at the sea ...'

### **8.1.6** *qari* **'at the top, above, on, about'**

This postposition, which governs the genitive, means 'at/on the top, above' and is the counterpart to *xːar(i)* (7a–7c).

	- 'The lamp hangs above the table.' (E)

In combination with the postposition *či* (§8.1.7) it is also used to express the topic of a conversation or the contents of thoughts (8a), (8b).

(8) a. ʡaˁrz complain r-ik'-ul f-say.ipfv-icvb ca-r cop-f iχ-i-la dem.down-obl-gen qari=či-r at.top=on-f 'She is complaining about him.'

8.1 Spatial postpositions

b. cinna pause.filler d-iχ-ub-t-a-la npl-be.pfv-pret-pl-obl-gen qari=či-d at.top=on-npl b-urs-ij n-tell.pfv-inf 'to talk about what had happened'

### **8.1.7** *či* **'on, above'**

This postposition, which is often pronounced together with the complement nominal as an enclitic, normally governs a spatial case, the loc-series (9a), (9b) (§3.4.2.2), but it can, in principle, also be used with the dependent noun bearing the genitive (9c). When the loc-series is used, then the direction markers of the case and of the postposition need to coincide, i.e. both are marked for the essive (9a), lative, or ablative (9b). The postposition can be encliticized to *qari* when referring to the content of conversations or thoughts (8a), (8b).

	- b. urči-le-r horse-loc-abl či-r on-abl ka-jč-ib-le, down-occur.m.pfv-pret-cvb w-i-ka-ag-ur m-in-down-go.pfv-pret ca-w cop-m kur deep kur-ri-cːe pit-obl-in 'He fell down from the horse into a deep pit.'
	- c. dubur-ra mountain-gen či-b on-n dirixʷ fog k'e-b exist.up-n 'There is fog on/above the mountain.' (E)

Since Sanzhi also has a preverb *či-* with a very similar if not identical meaning (§11.6.1) it is sometimes not easy to decide whether an occurrence of *či* functions as postposition/ adverbial or as preverb. Thus, instead of (9b) with *či-r* as postposition, we can also write it together with the verb and interpret it as preverb (10). But we can also manipulate the constituent order in (9b) and place the verb before the postpositional phrase (11) or have both the postposition and the preverb (12). In (11) and (12), *či-r* is unambiguously a postposition.


### 8 Postpositions

### **8.1.8** *urkːa* **'between, among, within, in the middle'**

The postposition *urkːa* has spatial and temporal meanings. For the spatial reading only genitive marking on the dependent noun is admissible. When used with nouns and noun phrases denoting a plurality it means 'between, among' (13a–14a).

	- b. k'ʷel=ra two=add qič'-m-a-la rock-pl-obl-gen urkːa between hek' dem.up qːarqːa stone b-uc-ib-il n-keep.pfv-pret-ref ʁuna eq musːa place k'e-b=q'al exist.up-n=mod

'A stone kept between two rocks or the like is up there.'

c. na now xːun-re woman-pl ču-la refl.pl-gen urkːa-b between-hpl qit.qit whisper b-ik'-ul hpl-say.ipfv-icvb ca-b cop-hpl

'Now the women are whispering among themselves.'

However, it can also occur with singular nouns and the meaning 'in'. The following minimal pair illustrates the difference:

	- b. qal-la house-gen urkːa-w=da between-m=1 'I am in the house.' (E)

With the temporal reading the postposition governs the genitive (15a) or the absolutive (15b). When reduplicated the postposition can be used as an adverb with the meaning *urkːa urkːa-b* 'from time to time, sometimes'.


'After two months they were doing (the medical treatment) ...'

8.1 Spatial postpositions

### **8.1.9** *b-i* **'in, inside'**

The postposition *b-i*, which only has spatial meanings, contains a gender/number prefix agreeing with the absolutive argument of the clause to which the postpositional phrase belongs. In all examples from natural texts the postposition governs the in-series or the loc-series (depending on the noun employed, see §3.4.2.4 and §3.4.2.2) (16a), (16b). However, in elicitation the genitive is also available (16c).

	- b. du 1sg ala 2sg.gen sunduq'-le-w box-loc-m w-i-w m-in-m le-w=da exist-m=1 'I (masc.) am inside your box.'
	- c. qal-la house-gen r-i-r=da f-in-f=1 'I (fem.) am inside the house.' (E)

### **8.1.10** *šːule* **'at side, next to, near'**

This postposition exclusively expresses spatial meanings. It requires the dependent noun to appear in the genitive case. Examples (17a–17c) show the postposition inflected for the essive case. In (17d), the postposition bears the ablative case suffix.

(17) a. šːi-la village-gen šːule-d at.side-npl qič'a-la rock-gen baˁʡ-li-gu-d wall-obl-sub-npl k'e-d exist.up-npl nišːa-la 1pl-gen χːuˁrbe tomb.pl

'Next/near to the village down at a stone wall, there is our graveyard.'


This postposition probably originates from the noun *šːal* 'side'. Although it looks like it could be the loc-case of this noun, this is synchronically not the case, since the loclative of the noun is *šːal*-*le* and not *šːu(l)le* (18). Nevertheless the origin from a spatial noun explains why the postposition governs only the genitive.

### 8 Postpositions

(18) a but du 1sg rjadom next.to Isaq'adi-la Isakadi-gen šːal-li-cːe-w side-obl-in-m hej this b-aʔ-le-w=da n-edge-loc-m=1 'And I am at the side of Isakadi, at this end.'

### **8.1.11** *tːura* **'out, outside'**

This postposition governs the genitive (19a–19c). However, it more frequently occurs as an adverb and as a spatial preverb with preceding nouns in the in-ablative or locablative (§7.1.2, §11.6.1).

	- b. qːapu-la gate-gen tːura-r outside-abl wahi evil ʁaˁʁ-la scream-gen t'ama sound ha-d-eʁ-ib up-npl-do-pret ca-d cop-npl 'From outside the gates an evil scream was made.'
	- c. heχtːu there.down šːi-la village-gen tːura-b-te outside-hpl-dd.pl χalq' people li<b>il=ra all<hpl>=add 'also all people outside of the village'

The postposition *tːura* also expresses the non-spatial meaning 'apart from, except for'. In this case the governed nominal can be not only in the genitive (20a), but also in the locablative (20b), (20c). For instance, if a demonstrative pronoun precedes the postposition the whole phrase reads as 'besides, and what is more, moreover'.

	- that.loc-abl outside word spr-m-say.ipfv-icvb n-want.ipfv-icvb cop-n il that qaˁb-la neck-gen w-aˁq-ij m-hit.pfv-inf 'Moreover, he defamed (him) and wanted (him) to be beheaded.'
		- c. Rasul-la Rasul-gen / / Rasul-le-rka Rasul-loc-abl tːura outside cara=ra other=add sa-č-ib=da hither-lead.pfv-pret=1 'Apart from Rasul I also brought another one.' (E)

8.2 Non-spatial postpositions

### **8.2 Non-spatial postpositions**

### **8.2.1** *b-alli* **'together, with'**

This postposition, which seems to be a cognate of the preverb *b-al* 'matching, together, in unison', governs the comitative (21a), and with young speakers the in-ablative case (21b). It has a gender/number prefix and agreement is controlled by the absolutive argument of the clause to which the postposition belongs. Both nominals, i.e. the governed one marked for comitative or in-ablative and the noun in the absolutive, can be absent.


If the governed noun phrase is overt it mostly precedes the postposition (21a), (21b), although it can also follow it (22a) or occur in another non-adjacent position (22b). If the governed noun is absent, the noun in the absolutive frequently takes its position right before the postposition (22c).

	- b. w-alli, m-together di-la 1sg-gen hej this juldaš friend w-alli m-together le-w=de exist-m=pst di-cːella 1sg-comit 'Together, my friend was together with me.'
	- c. ulbasne glasses d-alli npl-together ha-d-iqː-a-di=q'al up-npl-carry.ipfv-hab-1=mod '[If I had known that I will look at pictures], I would have brought my glasses.'

Because of the general closeness of adverbs and postpositions, examples such as (22a), (22b), in which *b-alli* and the case-marked noun occur in the reverse order and/or not immediately following each other can be treated as adverbial uses. Similarly, in (22c) a full postpositional phrase would be 'glasses with me', but the governed nominal is absent from the clause and thus the example rather represents the adverbial use.

### 8 Postpositions

### **8.2.2** *canille* **'together, with'**

This postposition, which probably originates from the numeral *ca* 'one', can also govern the comitative case (23). However, more frequently it is used as an adverb with the meaning 'together' (24).


### **8.2.3** *bahanne/bahandan* **'because of'**

This postposition originates from the noun *bahana* 'reason'. It governs the absolutive.

(25) a. qili-b home-hpl ruc-be sister-pl b-iħ-ib-le hpl-wrestle.ipfv-pret-cvb b-už-ib hpl-stay-pret ca-b cop-hpl hel that qix nut bahanne because.of 'At home the sisters were apparently arguing because of the nut.' b. ušːa 2pl bahanne because.of caj-na one-time ka-∅-jž-ib=da down-m-remain-pret=1 'Because of you I sat in prison once.' c. "Allah bahandan w-at-abaj!" ∅-ik'-ul ca-w

Allah because.of m-let.pfv-opt.3 m-say.ipfv-icvb cop-m '"For God's sake, let me!" he says.'

### **8.2.4** *akːʷar* **'without, except, apart'**

The participle form of the negative copula with the meaning 'not being' is used in constructions that have a meaning similar to adpositions like 'except, without'. The crosscategorical suffixes -*ce* or -*il* can be added to it without changing the meaning (26b–26d). The governed nominal is in the absolutive because of the verbal origin of *akːʷar* as a copula that governs the absolutive case.

(26) a. e, yes hel that ʡaˁlibatir Alibatir akːʷ-ar, cop.neg-ptcp di-la 1sg-gen pikri thought ħisab-le, account-advz han remember w-akːu m-cop.neg dam 1sg.dat 'Yes, except Alibatir, in my mind, I do not remember.'

8.2 Non-spatial postpositions


### **8.2.5** *q'atːin(na)* **'for the sake of, because of'**

This postposition governs the absolutive. There are no examples of this postposition in my corpus, but (27a–27c) show three elicited sentences.

	- b. du 1sg q'atːin for.sb's.sake ma-w-ax-utːa! proh-m-go.ipfv-proh.sg 'For my sake do not go.' (E)
	- c. ħaˁžimurad Hazhimurad q'atːin for.sb's.sake Muslimat Muslimat heštːu here sa-r-eʁ-ib hither-f-go.pfv-pret 'For the sake of Hazhimurad, Muslimat came here.' (E)

### **8.2.6** *ħaˁsible* **'according to'**

This postposition is almost exclusively used in the phrase *dila pikri ħaˁsible* 'in my mind' (26a), (28a), but it can also be used with other nouns that always occur in the absolutive (28b). It was originally borrowed from Arabic *ħaːsib* 'counting', which has a similar meaning.

(28) a. di-la 1sg-gen pikri thought ħaˁsible following hel-tː-a-la that-pl-obl-gen ʡaˁbal three litru-la liter-gen balun can čaˁʁir-la wine-gen le-b=de exist-n=pst

' In my mind there was their 3-liter can with wine.'

### 8 Postpositions

b. a but iš-tːi this-pl juldašː-e, friend-pl ce what b-ik'-ul=el, n-say.ipfv-icvb=indq tolko only hel that sːurrat picture ħaˁsible following b-aχ-ij n-know.pfv-inf a-w-irχʷ-ar, neg-m-be.able.ipfv-prs w-irχʷ-an-ne=w? m-be.able.ipfv-ptcp-fut.3=q

'And these friends, what they are saying, only by means of the picture, (one) cannot know, can one?'

## **9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes**

This chapter discusses the morphosyntactic properties as well as the semantic and pragmatic functions of predicative particles, conjunctions, temporal enclitics, pragmatic particles, and cross-categorical suffixes. They do not form a part of speech or a homogeneous category, although they can be subgrouped into relatively coherent classes:


They are mainly treated together in one chapter because they either do not fit into any of the previous chapters or because they have a special relevance for the grammar of Sanzhi such that a separate treatment is legitimate.

### **9.1 Predicative particles**

In recent studies of Dargwa varieties researchers have introduced the term "predicative particles" to refer to a closed class of grammatical elements that fulfill the functions of copula-like auxiliaries (e.g. Sumbatova & Mutalov 2003; Kalinina & Sumbatova 2007; Sumbatova & Lander 2014). This means that they function as heads of nominal predicate clauses and similar clauses that do not contain other verbs, and that they are used in analytic verb forms together with non-finite verb forms in order to form full main clauses. In other words, they are responsible for the finiteness of certain clauses, and their use depends on the clause type and the TAM form. In the following, I discuss these particles

### 9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes

for Sanzhi. I employ the label "predicative particles", but my analysis diverges from the analysis put forward by Sumbatova and colleagues.

Table 9.1 presents the predicative particles of Sanzhi. They are enclitics because they cannot form their own phonological word. They always need a host to which they attach, but unlike suffixes they can be added to various parts of speech or phrase types, that is, to verbs, but also to nominals (noun phrases), adjectives, or adverbs.



Due to this freedom in host selection they can be used in term focus constructions (§27.3.2). However, most commonly they occur in the position in which auxiliary verbs (e.g. auxiliaries expressing aspect or modality) occur, namely following the lexical verb. In copula clauses they are normally attached to the head of the predicate (§22.2). They partially express verbal categories such as person or tense, but they are not verbs themselves.

Sumbatova & Mutalov (2003: 138–140) and Sumbatova & Lander (2014: 153–163) include in their list of predicative particles three more items: the standard copula, the negative copula, and locational/existential copulas. For Sanzhi these are the copula *ca-b* (§16.1), the locational copulas *le-b*, *te-b*, *k'e-b,* and *χe-b* (§16.2), and the negative copula *(b-)akːʷ-* in its present tense and past tense forms. However, I consider these copulas to be verbs with defective paradigms that overlap in their functions with the predicative particles because they also occur in copula clauses and analytic verb forms, but they diverge from the enclitics in Table 9.1 in a number of ways.

First, they are not genuine enclitics; they can occur on their own without a host and can form their own clause, though some of them may also be used in the form of enclitics. Second, they express far more verbal categories than the predicative particles. The negative copula shares a great number of inflectional forms with standard verbs (e.g. it can inflect for habitual present and habitual past, masdar, etc.). The copula and the locational copulas have the same gender/number agreement affix as other verbs (even though all other verbs have gender/number prefixes and not suffixes). They convey present time reference, third person agreement, and are specified for affirmative polarity. Third, the predicative particles can be attached to the copula and to the locational copulas, including those particles that express verbal categories (i.e. the person enclitics and the past

### 9.1 Predicative particles

tense enclitic), so that all copulas can express first and second person agreement or past tense (5), but the person enclitics and the past tense enclitic strictly exclude each other.

The predicative particles can be divided into two groups. The first consists of the enclitics that express categories, which are most commonly marked on the verb (person enclitics *=da* and *=de* and the past tense enclitic *=de*), and the second group are the pragmatic markers (modal particle, interrogative particles). The two groups differ in their properties:

### **Verby predicative particles (person enclitics, past enclitic)**


### **Pragmatic predicative particles (modal particle, interrogative particles)**


Predicative particles of the two groups can co-occur with each other (1), (2); the verby particles always precede the pragmatic particles, and (most) other discourse particles. This means that the interrogative markers and the modal enclitic normally occur together with a person enclitic, the past enclitic or some kind of copula, and their function is primarily pragmatic (e.g. to convey a certain modal meaning or interrogative illocutionary force) and syntactic (for the interrogative markers).

(1) itːu there a-r-ax-an=da=q'al, neg-f-go-ptcp=1=mod sːa yesterday ag-ur=da go.pfv-pret=1 ʁubza emph 'I will not go there, I went yesterday.'

9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes

(2) hel-i-la that-obl-gen ʁaj-li-j word-obl-dat qili home arg-ul=de=w go.ipfv-icvb=2sg=q u? 2sg 'Are you going home because of what she said?' (lit. 'because of her word')

There are two types of clauses that may require the use of a predicative particle instead of a copula or another type of auxiliary verb. The first type is copula clauses (§22.2) and the second type is main clauses with analytic tense forms (Chapter 14). Thus, person enclitics and the past enclitic in the sentences in (3–5) cannot be replaced by copulas or other auxiliary verbs without changing the semantics of the clause or verb form or even making the sentence ungrammatical. The copula can be added to the clauses (3a), (3b) without noticeably altering the semantics or pragmatics of the sentences, but not to (4a), (4b). An example is provided in (5). This means that in the analytic verb forms the copula can never co-occur with the person markers or with the past tense enclitic.

### **First and second person or past time reference**

	- a. du 1sg ustːa=da master=1
		- 'I am a master.' (E)
	- b. du 1sg ustːa=de master=pst 'I was a master.' (E)
	- c. u 2sg ustːa=de=w? master=2sg=q 'Are you a master?' (E)
	- a. du 1sg ħaˁħaˁ laugh r-ik'-ul=da f-say.ipfv-icvb=1 'I am laughing.' (E)
	- b. du 1sg ħaˁħaˁ laugh r-ik'-ul=de f-say.ipfv-icvb=1 'I was laughing.' (E)
	- c. u 2sg ħaˁħaˁ laughter r-ik'-ul=de=w? f-say.ipfv-icvb=2sg=q 'Are you laughing?' (E)

### 9.1 Predicative particles

In clauses with third person agreement controllers the copula is normally used (6), (9). However, it can be omitted when the pragmatic predicative particles are used if the concomitant pragmatic meaning needs to be conveyed (7a), (7b) or if the speaker wants to utter a question (7c), (7d).

### **Third person non-past time reference**

	- a. ij this badra-cːe-d bucket-in-npl d-i-d npl-in-npl hin=q'al water=mod 'In this bucket there is water.'
	- b. di-la 1sg-gen arc=el money=indq hel-tːi that-pl 'That is probably my money.' (E)
	- c. it that ustːa=w? master=q 'Is s/he a master?' (E)
	- d. ij, this ča=ja who=q iž? this 'This, who is it?'
	- a. ij this ħaˁħaˁ laugh r-ik'-ul=q'al f-say.ipfv-icvb=mod 'She is laughing.'
	- b. it that ce what r-ik'-ul=e? f-say.ipfv-icvb=q 'What is she saying?'
	- c. heš-tː-a-l this-pl-obl-erg sud trial b-irq'-ul=el n-do.ipfv-icvb=indq 'They are probably making a trial.'

It is always possible to add the copula. Thus, the following two examples show copula clauses and analytic verb forms with copulas and additional predicative particles. In (10) the negative copula together with the modal particle and the affirmative copula with the indirecet question marker encliticized to it are used.

9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes


'The girl also does not know if that is a mill or not.'

Without the predicative particles (or a copula or another type of suitable auxiliary) the copula clauses would be ungrammatical:


Clauses with analytic verb forms are not ungrammatical, but they can only be used as subordinate clauses because of the non-finite verb forms (14).

(14) u 2sg ħaˁħaˁ laugh r-ik'-ul, f-say.ipfv-icvb ka-r-iž-ib-le=de down-f-sit.pfv-pret-cvb=2sg 'You (fem.) are sitting and laughing.' (E)

### **9.2 Conjunctions**

Sanzhi does not have native conjunctions, and this is typical for East Caucasian languages. The main way of conjoining phrases is the use of the additive enclitic (§9.4.1), and at the clause level converbs are employed (§25.1). However, there are a number of borrowed conjunctions whose use varies.

The monosyndetic conjunction *wa* 'and' occurs only in translated texts. The disjunctive particle *ja … ja* 'or', 'and', 'either … or', 'neither … nor' mostly occurs in the disjunction of clauses (15) or more rarely of phrases (16). Usually both disjunctions are introduced by *ja*. However, sometimes there is only one clearly identifiable disjunction member in which *ja* occurs, and in such examples *ja* can also function as a conjunction (17). The complex form *ja=ra* (or=add) is used as well (16). See §26.1 and §26.2.4 for more information on the disjunction of phrases and clauses and their syntactic properties.

9.2 Conjunctions


' These are probably their pumpkins or watermelons.'

(17) ha-b-eʁ-ib-le, up-hpl-go.pfv-pret-cvb ja and il that kaxʷ-ij kill.pfv-inf a-b-iχ-ub neg-hpl-be.able.pfv-pret 'They went and they could not kill him.'

The conjunction *amma* 'but' introduces adversative clauses. Usually these clauses refer to situations that are contrasted with earlier mentioned events and the conjunction occurs in clause-initial position rather than between two clauses (20), but it can also be used like a normal clause conjunction between two main clauses (18) or very rarely at the end of the clause (19).


Moreover, it is employed to mark a switch of the topic of a conversation (20), just as Russian *a* is used (see examples (23), (24) below).

(20) [topic switch to back to the previous topic, namely the price of flour] amma but urek six darš-li-j, hundred-obl-dat urek six darš-li-j hundred-obl-dat wahi-l bad-advz akːu cop.neg garam=ra gram=add ' But for 600, for 600, that is not bad at all.'

The subordinating conjunction *raχle* 'if' is a native item with the morphological structure of an adverbial derived by means of the adverbializing suffix *-le* (compare *raχ-raχle* 'sometimes'). It introduces conditional clauses (21). Because Sanzhi has specialized conditionals for this function, *raχle* always co-occurs with one of the conditional forms

### 9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes

(Chapter 18.3). The use of *raχle* is optional, whereas the conditional forms are mandatory. There is another borrowed conjunction with a similar meaning, *egena* (< Persian *eger* 'if'), which occurs only in translated or elicited clauses.

(21) raχle if tusnaq-le-r prison-loc-abl tːura outside uq-ulle, go.m.pfv-cond.1 cik'al something a-b-iχʷ-ar, neg-n-be.pfv-cond.3 ʡaˁħ-le good-advz saʁ-le healthy-advz ... ' If I leave prison well, healthy, if nothing happens to me, [ I will probably become a dentist].'

Sanzhi also has a couple of conjunctions borrowed from Russian: *i* 'and', *a* 'and, but', *no* 'but', and *ili* 'or' (see Forker (2018c) for code switching between Sanzhi and Russian). Among them, *i* and *a* are very frequently used by speakers of all ages in various types of texts, most often to conjoin stretches of discourse (not necessarily sentences) in the case of *i*. The conjunction *a* is used to mark a switch of the discourse topic (23), (24). In addition, they coordinate clauses, but do not conjoin phrases, since in this function *=ra* is used (§26.1).

	- 'and/but you also already told the story of Alibatir.'

The Russian disjunction *ili* 'or' conjoins disjunctive clauses (see §26.2.4 for examples). Furthermore, it is employed in clause-initial or clause-final position when expressing uncertainty together with the indirect question marker (25) or an interrogative particle (26).


9.3 Temporal enclitics

### **9.3 Temporal enclitics**

Temporal enclitics, i.e. enclitics used in specialized converbal clauses and for the expression of other adverbial phrases, are a group of two particles that are encliticized to verbs and nominals. Their meanings are rather adverbial (27) but because they are phonologically dependent on a host and can be hosted by a variety of parts of speech (verbs, pronouns, nouns, adjectives) I do not categorize them as genuine adverbials but treat them separately. They most commonly occur with non-finite verb forms (participles and infinitive/subjunctive) in adverbial clauses, a usage which corresponds to temporal and non-temporal specialized converbs. This function is only briefly illustrated in the current section, and more information and examples can be found in §18.2.1 for *=qːel(la)* and §18.2.3 for *=sat/=satːin/=satːinna*.

(27) a. *=qːel(la)* 'when, while, because' (simultaneity, anteriority, causality) b. *=sat/=satːin/=satːinna* 'until, before, as much as, as long as'

(posteriority, manner)

The enclitics are not subordinating conjunctions even if their meaning corresponds to subordinating conjunctions in other languages, because they do not fulfill the function of syntactic subordination as genuine subordinating conjunctions or complementizers would. From a morphosyntactic point of view, they can occur in subordinate clauses because they are added to non-finite verb forms that are used to function as heads of subordinate clauses due to their non-finiteness. The enclitics themselves only contribute to the semantics of those clauses, not to their syntactic properties.

One might argue that the enclitics resemble case markers or postpositions, but in contrast to the former they are not added to oblique stem forms, and in contrast to the latter they do not govern any cases. They have phrases in their scope and they are normally encliticized to the head of the phrase that they scope over, e.g., to the noun in a noun phrase (35), (36), (97). They share this property with the focus-sensitive particles such as the additive and the modal particles (§9.4). In the following, I will describe the functions of the two enclitics in more detail, concentrating on the uses with non-verbal hosts.

The particle *=qːel(la)*, of which the short form is used more often than the long form, is encliticized to the preterite and modal participle and to the negative copula (usually in its participial form), and expresses temporal simultaneity (28) and occasionally anteriority or causality.

(28) w-ebč'-ib m-die.pfv-pret χatːaj grandfather wer-c'a seven-ten nu well wer-ra seven-num dus year ∅-iχ-ub=qːel m-be.pfv-pret=when

' Grandfather died when he was 77 years old.'

Temporal simultaneity is also expressed when it is hosted by nominals such as demonstrative pronouns, nouns, numerals or adjectives and by adverbs. In the first place, the

### 9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes

enclitic is attached to demonstrative pronouns yielding the deictic meaning 'at that time, then', which transparently derives from the meaning of the demonstrative and the meaning of the enclitic (29). As can be seen in example (30), the enclitic can be preceded by the additive, which indicates that it is not a derivational suffix that forms temporal adverbials, but rather a syntactically independent item that scopes over the entire combination of demonstrative and additive.

(29) socijalizma=de socialism=pst het=qːella, that=when het=qːella that=when het=qːella that=when het=ʁuna that=eq parjadok order le-b=de exist-n=pst hetːu-b there-n

' Socialism was at that time, order (tidiness) like this was at that time there.'

(30) it=ra=qːel that=add=when ca one ʁuna=w? eq=q

'(Is it) that time also one and the same (shirt)?'

More rarely the enclitic appears on nominals with and without additional case markers (31), and also yields the meaning 'when'. For instance, a noun denoting a profession to which *=qːella* is added is interpreted as 'when performing the relevant profession'; a noun denoting a location plus *=qːella* leads to the meaning 'when being in that location' (31). Furthermore, the interrogative adverb *ceqːel* 'when' can diachronically be analyzed as *ce* 'what' and *=qːel*, and the indefinite pronouns *ca=qːel* and *ca-ca=qːel* 'sometimes, from time to time' as *ca* 'one' plus *=qːel*.


The enclitic *=sat/=satːin/=satːinna* occurs in three different variants that are functionally equivalent, but differ in their frequency of use. It originates from the postposition *sa* 'in front, ago'. When it is used with the infinitive/subjunctive the meaning is 'before,

until' (33), i.e. temporal posteriority, which corresponds to the meaning of the postposition from which it is derived. When the enclitic occurs with the modal participle the meaning is 'as much as, as long as' (34). More examples can be found in §18.2.3.


9.4 Discourse and modal enclitics

The latter meaning is also attested when the enclitic follows nouns (35), (36). As both examples prove, the enclitic is directly attached to the stem (after plural suffixes) without additional case marking and therefore does not qualify as a spatial case. Furthermore, it has the entire noun phrase in its scope.


Finally, the enclitic can be added to demonstrative pronouns and forms manner demonstrative pronouns that are used in comparison 'like this, like that, such':

(37) hel=sat that=as.much χːula big r-eʁ-ib-le f-go.pfv-pret-cvb r-už-ib-le f-be-pret-cvb hel that rucːi sister ' ( From her small finger he pulled out his parents), so big was his sister.'

### **9.4 Discourse and modal enclitics**

### **9.4.1 The additive enclitic**

The additive enclitic *=ra* covers all of the functions typical for additives in East Caucasian languages and other language families:

	- (38) Q'ampaj=ra Kampaj=add du=ra 1sg=add ag-ur=da go.pfv-pret=1 Sanži Sanzhi 'Kampaj and I went to Sanzhi.'

### 9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes

	- (41) [The fox said to Patimat: Do such and such!] "ʡaˁħ-le," good-advz r-ik'-ul f-say.ipfv-icvb ca-r. cop-f Pat'ima=ra Patima=add razi agree r-iχ-ub f-be.pfv-pret ca-r cop-f 'And Patima agreed, saying, "Good."'
	- (42) itːi those itwaj=ra like.this=add ʡuˁrusː-e Russian-pl ʁunab-te like-dd.pl ca-b cop-hpl hana=ra now=add 'And now also they are like Russians.'
	- (43) c'il=ra then=add hel-tːi that-pl bahla.bahlal slowly ʁaj=ra language=add d-aχ-ur-re, npl-know-pret-cvb bahla.bahlal slowly islužba=ra service=add b-iqː-ul, n-carry.ipfv-icvb … 'And then slowly I got to know this language, and I did my (military) service ...'

Furthermore, the additive is used in the formation of concessive clauses by adding it to conditional markers (§18.3.6). It also has derivational uses, namely the derivation of indefinite pronouns (§4.6), collective numerals (§6.6), and direct cardinal numerals from eleven up (§6.1). For an account of additive pronouns by means of the semantic map method see Forker (2016c).

9.4 Discourse and modal enclitics

### **9.4.2 The enclitic** *=q'al*

The enclitic *=q'al* is a frequently occurring focus-sensitive modal particle that in its frequency of occurrence is only outstripped by the additive (273 occurrences of the modal particle vs. around 1,700 occurrences of the additive in a corpus with 46,000 tokens). It belongs to the class of predicative particles, that is, when it is used together with certain non-finite verb forms such as the imperfective converb it turns the clause into an independent main clause (44) (§9.1). The particle cannot occur in utterances that are of a sentence type other than assertions and exclamations in the optative mood. Commands or questions are thus excluded. It occurs in main clauses (45) and in subordinate clauses (47) and is usually hosted by the verb, but in verbless predications by the predicate (48). It co-occurs with other predicative particles and follows them (e.g. person markers, past tense marker, embedded question marker), but not together with the interrogative enclitics (including the marker of embedded questions). It can be used in term focus constructions when it is encliticized to the item that is in focus.

The function of *=q'al* is to mark an utterance as presupposed, and thus as, in principle, known to the hearer (and the speaker), but potentially in need of being activated and brought to the conscious attention of the addressee, similar to English 'you know' or Russian *že* (which is used for the translation of *=q'al* into Russian) (44–47).


Example (47) was uttered when the speaker was talking about an event in which Hamid was involved and supposes that the addressee knows Hamid, which is the case.

(47) ħaˁmid Hamid w-alχ-atːe=q'al m-know.ipfv-cond.2=mod at, 2sg.dat ... 'Hamid, if you know him ...'

The addressee is sometimes implicitly or explicitly asked to agree with the speaker (48). In (49) the speaker criticizes the use of the Russian verb *sabrat* 'gather, collect' instead of a native term and invites the hearer to agree with her and to remember the Sanzhi word.

9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes


It is also used in questions with a strong presupposition that the addressee knows the answer (50).

(50) hej this q'ar herbs ce=jal what=indef b-ik'-u=q'al hpl-say.ipfv-prs.3=mod nišːa-la? 1pl-gen ' These herbs here, how are they called in our (language)?'

Another common usage is existential clauses with the locational copula *le-b*, confirming the existence of a referent that is going to be the topic of the following discourse (51). They seem to correspond to the Russian phrase *X est' že*, a typical Dagestanian expression that is almost never used by speakers of Standard Russian.

(51) a-rurg-an neg-burn-ptcp le-b=q'al exist-n=mod it, that birikːalla.ʁut' cow.parsnip le-b=q'al exist-n=mod het that 'There is the one that does not burn, the cow-parsnip.'

If the information is new, it is still treated as presupposition that requires immediate accommodation. For instance, in (52) the speaker is talking about how his grandfather for the first time brought iron parts for ploughs to Sanzhi which were unknown in Sanzhi and the addressee does not necessarily know this fact, but it is marked as presupposed by means of *=q'al*.

(52) Sanži-d Sanzhi-npl d-a-d-už-ib-le=q'al npl-neg-npl-be-pret-cvb=mod hel-tːi that-pl.abs 'In Sanzhi, they did not have them (the iron item that you use on wooden ploughs).'

The enclitic *=q'al* widely occurs in other Dargwa varieties. In Standard (Akusha) Dargwa, there is *q'alli*, which van den Berg (2001: 748–75) analyzes as a focus particle. Tatevosov (2001) analyzes Icari Dargwa *=q'al* as a mirative marker. Sumbatova (2009) in her account of questions in Icari calls it a focus-marking clitic translated with 'but' in the glosses and described as marking the proposition "as known to both communicants". In the Icari grammar as well as in Kalinina & Sumbatova (2007) the same enclitic is also glossed with 'but' and described as "actualization particle". Sumbatova & Lander (2014: 338–339) treat Tanti Dargwa *=q'ale* as an actualizing marker with a functional range very similar to the Sanzhi particle. For a detailed analysis of the morphosyntactic and semanto-pragmatic properties of *=q'al* in Sanzhi see Forker (Accepted).

9.4 Discourse and modal enclitics

### **9.4.3 The enclitic** *=q'ar*

The enclitic *=q'ar* is a modal particle that partially overlaps in its uses with *=q'al* (§9.4.2) and *=n(u)* (§9.4.4). Like *=q'al* it cannot be used in questions together with the interrogative particles. However, it does not belong to the class of predicative particles. It bears some resemblances to German *doch* and Russian *že*.

The enclitic is used when correcting utterances (53) or contradicting expectations. Thus, the speaker of (54) expected her children to inform her about the death of her son, but in order to preserve the mother from the very devastating news they did not tell her everything, but discussed the issue only among themselves.

(53) kartuška-la potato-gen qːup-re=q'ar sack-pl=mod akːu, cop.neg ∅-ik'-ul=da, m-say.ipfv-icvb=1 χalq' people ca-b, cop-hpl ∅-ik'-ul=da, m-say.ipfv-icvb=1 heχ-tːi dem.down-pl

'These are not potato sacks, I say, they are people.'

(54) hel-tː-a-l, that-pl-obl-erg b-aχ-ur-re n-know.pfv-pret-cvb b-už-ib n-stay-pret ca-b cop-n hel-tː-a-j, that-pl-obl-dat di-cːe=q'ar 1sg-in=mod a-b-urs-ib neg-n-tell.pfv-pret

'Apparently they knew it and they did not tell it to me.'

More generally, *=q'ar* signals contrast between the utterance in which it occurs and some other utterance or previously discussed issues, i.e., it marks contrastive topicalization ('and as for X, P'). For example, in (55) the speaker contrasts the behavior of a person after he had been in prison with his behavior before he went to prison, when he never refused a drink with his friends. Similarly, (56) exemplifies a parallel structure of two clauses that immediately follow each other and contain contrasting propositions.


'The (fox) says, "For me they made a bed from herbs and grass, and for themselves they took probably cotton wool."'

Similarly to *=q'al* as described above, the enclitic *=q'ar* is also used as an actualizing modal particle that relates the utterance to the argumentative background and in this

### 9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes

way indicates what is assumed to be common ground. In other words, it signals what the speaker assumes to be known by the hearer. For instance, (57) is from a narrative about some people who stole money during the absence of the main character and his wife. The speaker stresses the fact that it is clear to everyone that the people knew the times of the day when nobody was at home and when the wife was supposed to come home again. And (58) describes the place close to Sanzhi where there are old paintings on rocks that are still visible, although they are assumed to have been made thousands of years ago, and what the speaker says is a fact known to every Sanzhi person.

(57) ʡaˁči-le-r work-loc-abl sa-q'-aˁn hither-go-ptcp zamana time sːaˁʡaˁt-e=q'ar hour-pl=mod d-alχ-ul npl-know.ipfv-icvb ca-d cop-npl hex-tː-a-j dem.up-pl-obl-dat

' They know the time when (the wife) comes back from work.'

(58) ixtːu=q'ar there.up=mod marka rain či-ikː-ul on-get.ipfv-icvb akːu cop.neg 'The rain does not reach there.'

### **9.4.4 The enclitic** *=n(u)*

The enclitic *=nu* (allomorph *=n* after vowels) is used when the speaker wants to attract the attention of the addressee. It is mostly encliticized to verbs. Its meaning can be paraphrased as 'watch out, pay attention, something is happening or is going to happen in the near future that is of relevance and important for you'. There are several contexts in which it usually occurs. For example, *=n(u)* often occurs in clauses with first person subject-like arguments when the speaker wants to stress the fact that s/he is already performing an action or is in a certain state or is about to perform an action in the near future (59).

(59) hana now t'aš stop r-icː-e! f-stand.pfv-imp r-ax-ul=da=n f-go-icvb=1=prt ' Now wait, I (fem.) am coming.'

Such clauses can also have second or third person subjects, but again they warn that soon something will happen that is of importance for the addressee (60).

(60) u=ra ukː-an=de=n, maˁ-q'-aˁtːa!

2sg=add eat.m.ipfv-ptcp=2sg=prt proh-go-proh.sg

' [Your sister turned into a monster, she ate people], and she will eat you (masc.), don't go!'

The enclitic is part of the phrase *celij akːu=n* (why cop.neg=prt) with the meaning 'because'. It introduces clauses that deliver an important explanation that the speaker wants the addressee to pay attention to (61).

9.4 Discourse and modal enclitics

(61) cellij why akːu=n cop.neg=prt tusnaq-le-w=qːella prison-loc-m=when qihin-ne difficult-advz ca-b cop-n 'Because when you are in prison it is difficult.'

The second context is the use with imperatives and optatives, because they also occur in utterances that are of special importance and relevance for the addressee who, for instance, has been ordered to do something (62).

(62) nu well b-ikː-aχː-at, n-want.ipfv-cond-cond.2 nišːa-la 1pl-gen heχtːu there.down d-uk-a=n, npl-gather.ipfv-imp=prt d-ac' npl-empty d-arq'-a=nu, npl-do.pfv-imp=prt uq'-aˁ=nu! go.m-imp=prt ' If you want, take the (stuff) and empty it there in our place, go!'

The third context is contrastive focus constructions that are used to correct wrong assumptions, assertions or beliefs. The enclitic occurs in the clause that rejects the assertion and is followed by the correction:

(63) qili home akːu=nu, cop.neg=prt witrezwitel-le sobering.up-loc uqː-ul carry.m.pfv-icvb ca-w cop-m

' No, not home, they take him to the sobering-up station.'

(64) či-haˁ-q'-uˁn-ne=kːu=n spr-up-go-pret-cvb=cop.neg=prt ka-jž-ib down-remain.m.pfv-pret ca-w cop-m hel-tː-a-cːella that-pl-obl-comit w-alli m-together

' No, he is not going towards them, but sitting together with them.'

It is not necessary that the rejecting clause contain a negation; it can also be an affirmative clause that functions as a correction. For instance, people tried to destroy a mill by hitting the turning mill stone. They did not immediately succeed although they tried hard and thus (65) contradicts the expectation that they had already finished their destruction.

(65) urk'i heart b-el=nu, n-remain.pfv=prt b-aˁq-aˁjaˁ! n-hit.pfv-imp.pl urk'i heart b-el=nu, n-remain.pfv=prt b-aˁq-aˁjaˁ! n-hit.pfv-imp.pl

'The heart remained, hit it! The heart remained, hit it!'

### **9.4.5 Other enclitics that manipulate information structure**

Sanzhi has further focus-sensitive enclitics with a more specific semantics: *=cun* 'only' (66), *=gina* 'alone, only' (67), *malle* 'even' (68), and *arrah* 'at least' (69). Note that in (67) the enclitic *=gina* is followed by a person enclitic; the reverse order would be ungrammatical. See also §27.5 for more information on the position and use of focus-sensitive particles.

9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes


' Even his sister's daughter did he throw out, after some years.'

The particle *arrah* 'at least' is used in commands (69), irrealis conditional clauses, and negative clauses together with the quantifier *ca* 'on' with a scalar additive meaning (70). It mostly occurs following nominals and then has scope over the nominals, but it can also scope over verbal predicates. In the latter case, it is possible to insert the particle between the locational and the deixis/gravitation preverbs. For instance, in (71) the verb is prefixed with the locational preverb *či*- and the particle follows it. This preverb is a lexicalized part of the verb 'see' because the root almost never occurs without the preverb, and thus the particle is inserted into a verbal stem.


'"Of our (people) not even one man fell down," he says.'

(71) w-aš-e m-go-imp a-ag-ur-il neg-go.pfv-pret-ref ∅-iχ-utːe, m-be.pfv-cond.2sg či=arrah-b-až-ij spr=at.least-n-see.pfv-inf 'Let's go if you did not go there, at least to see.'

There is an emphatic enclitic *=le*, which, however, occurs only twice in the corpus, and speakers do not have clear intuitions about its meaning, making it difficult to analyze in detail. These are the two examples:

(72) c'elt-m-a-cːe-w gravestone-pl-obl-in-m q'ar-ri-cːe-w herbs-obl-in-m hi-l who.obl-erg urkː-u=le find.m.ipfv-prs=emph dučːi-la night-gen itːu there 'Among the graves, in the grass, who finds (him), at night.'

9.5 Pause fillers, address particles, exclamatives, and interjections

(73) iχ dem.down rursːi-la girl-gen qːuʁa-l beautiful-advz kʷi-sa-b-uc-ala=le in.hands-hither-n-keep.pfv-nmlz=emph nik'a-ce small-dd.sg

' (Look at) the way the girl is holding the child in her hands.'

And there is another enclitic *=k'u* that is also roughly described as emphatic or modal. Like the two modal enclitics *=q'al* and *=q'ar* it is usually translated by *že* or *ved'* into Russian. The enclitic is also used for the formation of specific indefinite pronouns (§4.6.1). In the corpus, there are three occurrences of the emphatic/modal use, of which two are given here:

(74) "u=k'u 2sg=emph ik'ʷ-a-tːe," say.ipfv.m-hab.pst-2sg haʔ-ib=da, say.pfv-pret=1 "w-arčː-ib-le=kːu" m-find.pfv-pret-cvb=cop.neg

'I said, "You said that they did not find him."'

(75) ca one dus=k'u, year=emph ašːi-j 2pl-dat b-aχ-ij n-know.pfv-inf d-urkː-a-tːa npl-find.ipfv-hab.pst-2pl čina where musːa-t=te=l, place-pl=pst=indq Maˁħaˁmmad Mahammad

'One year, you should know it, where the places were, Mahammad.'

Interrogative markers for polar questions (§28.1), content questions (§28.2), and embedded questions (§28.4) also play a role in the information structure of utterances and are analyzed in separate sections.

### **9.5 Pause fillers, address particles, exclamatives, and interjections**

Sanzhi has two politeness particles that are used in imperatives and prohibitives in order to soften the command, *=kːʷa* and the rarely used *=ri*. The first enclitic is also used in polite questions (79). In my corpus there are 31 occurrences of *=kːʷa* (and just one of *=ri*), and two thirds of them were uttered by female speakers. Thus, it might be the case that the use of *=kːʷa* is more common among female speakers.


9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes


The genitive reflexive pronouns *cinna* (singular) and *čula* function as pause fillers. The same has been reported for the neighboring Dargwa variety Icari (Sumbatova & Mutalov 2003: 187, fn. 107). It seems that the singular pronoun occurs when the subjectlike argument is singular (80), (81), and the plural pronoun when it is plural (82). The full paradigms of the reflexive pronouns are listed in §4.3 and their use in reflexive constructions is analyzed in §29.1.


this

' He is really thinking (or worrying), and sitting, because it is like he is in prison.'

	- as cop-npl

'Well, probably they are like under the ground (growing).'

It is not always easy to identify the pause fillers because often the reflexive pronouns can be interpreted as possessive pronouns with an omitted head noun. For instance, example (82) refers to a picture showing plants or roots that grow in the earth and the reflexive *čula* could serve as a pronoun in a phrase like 'their (plants)'.

Common address particles are (83). Some examples are given in (84–86).

(83) a. *ja, wa* 'ey, hey'


9.5 Pause fillers, address particles, exclamatives, and interjections

(84) ha-jʁ-ib, up-come.m.pfv-pret r-ik'ʷ-ar, f-say.ipfv-prs "wa hey Iljas." Ilyas "haj" hi ' He came, and she said (to him), "Hey Ilyas." (He said), "Hi."'

The particle *ulkːa(s)* is used as an address term when trying to prompt the reaction of the addressee, for instance when asking him to answer a question (85), but it also seems to be a pause filler.

(85) ulkːas, hey ħaˁsanʡaˁli, Hassanali b-arx-le n-direct-advz b-urs-a=kːʷa! n-tell.pfv-imp=prt 'Hey, Hasanali, tell the truth!'

The particle *žan* 'beloved' is a loan from Persian with the meaning 'life, soul, spirit' that is used as an address particle before names or kinship terms when expressing endearment and affection towards the addressed person. It is also used as a noun with the meaning 'body, vital essence'.

(86) ellelej, prt žan beloved durħuˁ, boy haʔ-ib=da say.pfv-pret=1 ča=de=l who=2sg=q ' Ooh, my dear boy, I thought, who are you?'

There are two particles *ma* 'take' and *hara* 'come, go, look, here is, here you are' that are used in commands when requesting the addressee to take something or to come to the speaker. These particles thus function like verbs inflected for the imperative. Therefore, they can also attach the suffix *-(j)a* (87), which is used in commands and other kinds of non-indicative utterances when the addressee is plural and most often co-occurs with the imperative, the prohibitive, and the optative.

	- cop-hpl

' They say to the guys passing by, "Take (a drink)!"'

(88) "hara, come ala 2sg.gen durħuˁ," boy ∅-ik'-ul m-say.ipfv-icvb ca-w, cop-m "haˁ-q'-uˁn-ne" up-go-pret-cvb

'" Come, your son has come," he says.'

Interjections expressing astonishment or excitement are *huja, waħ* 'wow', *ellelej(-q'u)* 'oh, oh, oh' (astonishment, slightly negative evaluation) (86), and *ʁubza* 'oh man' (95). The latter originates from the noun *ʁʷabza* 'dzhigit, true man'.

Sanzhi has no real words for 'yes' and 'no', instead the copula *ca-b* is used or the respective verb forms is repeated when affirming what has been said or agreeing with the addressee. For rejection or disaffirmation the negated verb is used. However, the exclamations *e* 'yes, agreed' and *aʔa* 'no' can also be employed in these functions. For more examples of question-answer pairs, see §28.1.

9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes


Other particles and exclamations are presented in (91).

	- b. *jaʁari(b)* 'listen, my dear'
	- c. *hu* 'well, now, right, come on'
	- d. *wari* 'no, no way' (emphatic warning)
	- e. *ixʷixʷle* 'of course' (to express irony and when the speaker does not believe the addressee)

'" Oh, it turned out that the Sanzhi person is such a strong man," he says.'

(93) hu=kːʷa, well=prt sa-d-irʁ-an-ne hither-npl-come.ipfv-ptcp-fut.3 'Well, they will come (another day).'

There are a couple of exclamative phrases and words from Arabic that are common in the Muslim world and are also used by Sanzhi speakers (94), (95).

	- b. *ʡaˁlħaˁmdullilah* 'Praise be to God!'
	- c. *inša-Allah* 'if Allah wills'
	- d. *aman* 'alas, mercy, pity, oh, ah!' (lit. security, safety, peacefulness)
	- e. *wallah, wallahi tallahi, billah, wallah tallah* '(I promise, I swear) by God'
	- f. *mašaʔallah* 'God has willed it' (appreciation, joy, praise or thankfulness for an event or person that was just mentioned)
	- ' I swear by God it happened.'

9.6 Cross-categorical suffixes

Nowadays, speakers also employ Russian words or phrases as interjections or pause fillers (96).

	- b. *karoče* 'in short' (pause filler)
	- c. *značit* 'thus, this means' (pause filler)
	- d. *wat'* 'well, here is'
	- e. *tak* 'like this, so, well'
	- f. *že* (modal particle)
	- g. *dawaj* 'let's go, come' (invitations and requests)

The greeting phrase used among men is the traditional Arabic phrase *as-salam ʡaˁlaykum*. Other greetings are given in (97), (98). The first is used for greeting women, for example when they are sitting in front of their house because with women the Arabic phrase or its shorter form *salam* is not used. The phrase in (98) is uttered at night when leaving or going to bed, but not when greeting people at night.

```
(97) ka-d-iž-ib-le=da=w
      down-1/2pl-sit.pfv-pret=cvb=1=q
      'Hello!' (lit. 'Are you (pl.) sitting?')
```

```
(98) dučːi
      night
            ʡaˁħ
            good
                  d-iχʷ-ab!
                  npl-be.pfv-opt.3
      ' Good night!'
```
With outsiders, especially when they are female, Russian salutations are used (e.g. *zdrastvujte* 'hello', *dobryj den'* 'good day').

### **9.6 Cross-categorical suffixes**

These suffixes can be viewed as cross-categorical derivational suffixes that attach to a number of parts of speech (adjectives, verbs, adverbs, postpositions, nominals) and form referential attributes/definite descriptions with nominal properties (suffixes -*ce* and -*il*) or adverbials (adverbializing suffix).

### **9.6.1 The suffix** *-ce*

### **9.6.1.1 Function and distribution of the suffix** *-ce*

The semantic, syntactic and distributional properties of the suffix -*ce* are quite complex. Its syntactic impact overlaps with that of the suffix -*il* described below, but the distributions of both suffixes are rather complementary (see the end of §9.6.2 for a comparison). The suffix -*te*, which is, in fact, one of the most productive nominal plural suffixes (§3.2.1) is used as the plural form of -*ce* and for the sake of simplicity will be treated as such in

### 9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes

this section. However, there are small functional differences between both suffixes -*ce* and -*te* that will be pointed out whenever relevant.

The suffix -*ce* is added to:


The core function of the suffix can be described as forming definite descriptions that describe the referent via its location, its qualities, or its possessor:


When the referent is in the singular, -*ce* is used; when it is plural, -*te* is used. The descriptions can be used as referring expressions that function as phrasal or clausal arguments, predicates or detached topicalized items, etc. Based on the core function, the use of the suffix has further extended such that it is also optionally found on attributes such as adjectives and relative clauses that modify nominals. In the following, I will explain my approach by going systematically through the parts of speech listed above and the contexts of use.

First and foremost the suffix -*ce* is found on adjectives. In my corpus, this usage exceeds all other uses. The suffix can optionally be added to adjectives in attributive function without leading to any noticeable semantic difference (99). As the same example shows, it can be added to adjectives with gender agreement prefixes and those lacking gender agreement prefixes. If the head noun is preceded by more than one adjective, all adjectives preceding it can but not need bear the attributive suffix.


9.6 Cross-categorical suffixes

	- b. *žahil qːuʁa-ce rursːi*
	- c. *žahil-ce qːuʁa-ce rursːi*

In order for adjectives to be used as predicates (102) or nominals (104) the suffix is obligatorily added, and this rule includes Russian loan words as well (103). Adjectives that bear the suffix -*ce* are referential nominals and thus can occur in a position detached from the noun even if they semantically rather seem to function as nominal modifiers (103). Examples such as (103) do not represent discontinuous noun phrases. The adjective is rather an independent referential constituent that occurs to the right of the clause as an afterthought. This will be analyzed in more detail in §9.6.1.2 below.


Adjectives (and other items) bearing the suffix can take case suffixes after the oblique stem suffix -*li* has been added (104). In the plural, -*te* is replaced by -*ta* when cases are added (in the same way as for nouns that make use of the plural suffix -*te*).

(104) χːula-ce-li-j big-dd.sg-obl-dat ʡaˁħ-le good-advz ʡaˁq'lu mind b-alχ-u=w? n-know.ipfv-prs=q 'Does an older (person) know it better?'

Second, the suffix appears on participles (modal and preterite participle) that form relative clauses. Its use is optional and relatively rare for relative clauses in the canonical prenominal position and seems to be preferred for head nouns in the plural and mass nouns that control plural agreement (in which case -*te* instead of -*ce* is used) (106), (107). For head nouns in the singular, the use of the suffix -*il* is more common than -*ce* (§9.6.2). Example (105) is part of a translation of the famous fable 'The North Wind and the Sun'. Example (106) comes from the translation of a Standard Dargwa folktale.

(105) "či-b-ig-ul=de=w?" spr-n-see.ipfv-icvb=2sg=q b-ik'ʷ-ar n-say.ipfv-prs bari sun [q'uc' offence b-iχ-ub-ce] n-be.pfv-pret-dd.sg č'an-ni-cːe wind-obl-in

'"Do you see?" said the sun to the offended wind.'

(106) [juldašː-a-l friend-obl.pl-erg cin-i-j refl.sg-obl-dat sa-qː-ib-te] hither-carry-pret-dd.pl xunul-be gift-pl 'the gifts that the friends had brought to him'

9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes

(107) [ʁanu twenty xu-ra five-num dus year či-a-d-až-ib-te] spr-neg-npl-see.pfv-pret-dd.pl dalga=ra detail=add či-d-až-ib=da spr-npl-see.pfv-pret=1 'I also saw the details that I did not see for 25 years.'

The use of -*ce* becomes obligatory when relative clauses with the preterite participle occur in a position after or detached from the noun that they semantically belong to (108) or when they are used without a head (109), (110). In other words, relative clauses that do not function as attributes but as nominals are marked by -*ce*.


'Like the ones made by little children, from the old times, they were interesting.'

In the function of marking relative clauses the suffix in principle competes with -*il* (§9.6.2 below), but we find a clear distribution. The suffix -*ce* can only be used with singular referents (105), (109), but its use is relatively rare and -*il* is normally used instead. By contrast, in the plural -*il* cannot be used and only -*te* is available (106), (108), and (110).

Relative clauses are not the only types of clauses that can be turned into referential definite descriptions by means of -*ce*. Factual complement clauses with matrix verbs denoting emotions, cognition as well as evaluative predicates can also be marked by the preterite participle and -*ce* (as an alternative to, e.g., the masdar suffix) (§24.2.3). This use is straightforward: a fact is expressed as a proposition by means of -*ce*, i.e., as a definite description, and can then be used in argument position. In this function, the use of -*te* is not allowed.


9.6 Cross-categorical suffixes

Similarly, -*ce* (but not -*te*) can be added to the infinitive and used as the complement of the copula in existential clauses (113). The infinitive + -*ce* combinations of the verbs 'eat' and 'drink' have been lexicalized as nouns, e.g. *b-erkʷ-ij-ce* 'food' (n-eat.pfv-inf-dd.sg).

(113) cara other cik'al something b-urs-ij-ce n-tell-inf-dd.sg b-akːu n-cop.neg-prs ' There is nothing more to tell.'

There is one more context in which the suffix is used on verbs, namely for the formation of the analytic verb forms called "experiential" in this grammar. These verb forms consist of the preterite participle plus *-ce* (or *-il*) and a copula, and have perfect-like semantics. They are predominantly used when speakers talk about their own experiences and about situations they were personally involved in (§14.2.6 and §14.2.7). The semantic contribution of the suffix *-ce* to these verb forms is unclear to me, but their syntactic impact is obvious. The experiential tenses are close to forming a clause union or biclausal structure, i.e., the participle with *-ce* functions like a headless relative clause.

Third, the suffix can be added to nominals that are marked for the genitive case (114) or for the essive case (115) and also to spatial adverbs and postpositions that are inflected for the essive case (116). Thus, in (116) the suffix has the entire postpositional phrase in its scope. As with the adjectives and the relative clauses, the so-formed constituents are definite descriptions that function as attributes of nouns or are referentially independent.


' First he wanted to take one pear, when he saw the man who was in the tree.'

The difference between modifiers or adjuncts bearing *-ce* and those not bearing *-ce* can be illustrated by the following minimal pair. The first sentence has two interpretations, one in which the noun with the spatial case suffix modifies the whole clause, and another one in which it modifies only the following noun phrase. By contrast, if the suffix *-ce* is added to the noun with the spatial case, only the second interpretation is available.

9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes


### **9.6.1.2 Analyzing the suffix** *-ce* **and its cognates in other Dargwa languages**

Cognates of Sanzhi -*ce* are found in most if not all Dargwa languages (e.g. -*ci* in Standard Dargwa and Icari Dargwa, -*se* in Tanti Dargwa, -*ze* in Chirag Dargwa). In the literature, they have mostly been analyzed with respect to their occurrence on adjectives. Thus, adjectives have been divided into 'short adjectives' without the suffix and 'long adjectives' that bear the suffix.

In grammars of Standard Dargwa, the short adjectives are said to be more archaic and basically only used in poetry and other types of fictional literature as expressive means to describe emotions and feelings (van den Berg 2001: 26); (Abdullaev et al. 2014: 207–208). According to the latter grammar, adjectives with gender prefixes do not have a short form. This is in plain contrast to Sanzhi Dargwa, where they have a short form, e.g. *ca b-uqen q'aˁli* (one n-long branch) 'one long branch'. Furthermore, in Sanzhi short adjectives are at least as common as adjectives with the attributive suffix, if not more common.

Lander (2014) (see also Sumbatova & Lander 2014) describes short adjectives in Tanti Dargwa as formally and functionally marked and opposed to the unmarked long adjectives bearing the suffix *-se* (the cognate of Sanzhi *-ce*) because the former are rarely used and are restricted in their distribution. By contrast, the long adjectives allow for a large range of constructions. Lander (2014) analyzes them as basically equivalent to relative clauses. He rejects an analysis of *-se* as a nominalizer because adjectives to which *-se* is suffixed differ in some properties from standard nouns. First, they cannot be modified by short adjectives. Second, they can modify personal pronouns, indefinite pronouns, and reflexive pronouns. Third, when case-marked, long adjectives cannot follow the noun as would be expected for a noun in an appositive construction.

For Sanzhi Dargwa the question of markedness is not fully clear, but if we can apply this label at all, it is modifiers having the attributive suffix (e.g. 'long adjectives') that are marked, rather than the other way around. First, they are clearly formally marked by the suffix. Second, they seem to be slightly less common than short adjectives, can occur in positions that most nominal modifiers cannot occur in, and occasionally have marked, contrastive semantics that is absent from unmarked modifiers (see the discussion below). Furthermore, when occurring outside of their canonical position, they are syntactically not part of the noun phrase to which they semantically belong. This becomes apparent when the head noun of the noun phrase appears in a case other than the unmarked

### 9.6 Cross-categorical suffixes

absolutive. In such a case, the full adjective can only follow a noun when it is also casemarked and interpreted as forming its own phrase. In other words, it is nominalized and takes an argument or adjunct position in the clause (119), (120). A similar behavior is observed with floating quantifiers, which are also syntactically not part of the noun phrase (see §21.1.3).


Furthermore, modifiers with attributive suffixes can also be modified by modifiers without attributive suffixes, even in those cases where the former are used as nominals (121), though it would preferable to use attributive suffixes on both adjectives in this example (i.e. *b-iq'-ur-ce it'in-ce*).

(121) asː-a buy.pfv-imp b-iq'-ur n-ripen-pret it'in-ce! red-dd.sg 'Buy a ripe red one!' (E)

This behavior points again towards an analysis of the attributive suffix as a nominalization marker. If nouns bearing attributive suffixes are nominalized, we can opt for an analysis in terms of appositional constructions. In appositional constructions, the head noun is modified by one (or occasionally more than one) noun preceding it. Case marking occurs only once, namely on the head noun (122). It cannot occur on the modifier, be it a full adjective or an appositive noun.


### 9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes

A similar analysis has been proposed in the Icari Dargwa grammar: adjectives and other words bearing *-ci/-ti* are analyzed as free attributes alongside cardinal numerals, other derived adjectives and some other words. Sumbatova & Mutalov (2003: 48, 129) claim that "free attributes and nouns could probably be considered to form a single syntactic class (nouns). The main difference is that free attributes are much more common in the attributive position than nouns." Furthermore, free attributes "usually stress the restrictive character of the attribute or even imply contrastive emphasis on the attribute." This characterization fits well the Sanzhi data. Modifiers bearing the attributive suffixes can have a contrastive reading, but this reading is normally due to their position (e.g. after the noun) and is not part of the meaning of the suffixes. The suffix just makes it morphosyntactically possible for the modifier to follow the head. For instance, the following elicited example refers to a situation in which large and small plates are contrasted, but the translation of the sentence contains only one occurrence of the attributive suffix on the second adjective, because it occurs without a head noun. This means that the use of the attributive suffix has a purely morphosyntactic explanation.

(124) h-asː-a up-take.pfv-imp χːula big waq, plate kʷi-r in.the.hands-abl ka-b-ix-a down-n-throw.pfv-imp nik'a-ce! small-dd.sg 'Take the large plate, put away the small one!' (E)

Modifiers with *-ce* can precede pronouns and occur on non-restrictive relative clauses (125), which also demonstrates that they do not convey contrastive or restrictive semantics. For example, (126) does not imply that the speaker has another mother who is not old.

(125) [uniwersitet university ha-b-erχː-aq-ur-ce] up-n-fulfill.pfv-caus-pret-dd.sg at 2sg.dat ʡaˁħ good ʡaˁči work b-irk-u n-occur.ipfv-prs 'You who has finished the university will get a good job.' (E)

(126) di-la 1sg-gen r-uqna-ce f-old-dd.sg aba mother na already ixʷbel=ra long.ago=add qili-r home-abl tːura outside a-r-ax-u neg-f-go-prs 'My old mother already since long ago does not leave the house.' (E)

However, if they modify personal names the interpretation is normally contrastive. For instance, the use of a noun phrase such as (127) implies that there is another person called Murad who is not good.

(127) ʡaˁħ-ce good-dd.sg Murad Murad 'the good Murad' OR 'the Murad who is good' (E) 9.6 Cross-categorical suffixes

I finish this section with a final comment. During a guest lecture at the University of Potsdam the audience suggested that -*ce* bears some similarity to quantifiers. It might serve to express number similar to what we observe in English *the red one*, and resembles indefinite pronouns such as *some*. In fact, -*ce* is homophonous with the interrogative pronoun *ce* 'what', which can also be used as an indefinite pronoun meaning 'something'. The similarity is also attested in other Dargwa languages, e.g. Tanti (-*se* and *se* 'what'). As already mentioned, the plural marker -*te* is identical to one of the normal plural suffixes for nouns, and becomes -*ta* when further case suffixes are added. This suggests that, in contrast to -*ce*, the suffix -*te* is morphologically complex, and -*ce* and -*te* are not diachronically related, but go back to different sources. From this it naturally follows that -*ce* and -*te* do not have to have identical distributions. Following this suggestion, items bearing -*ce* could be analyzed as quantificational expressions rather than as referring expressions. However, further research is needed in order to test this and other proposals and to reach a full account of -*ce*, -*te* (and -*il*).

### **9.6.2 The suffix** *-il*

The cross-categorical suffix *-il* is functionally very close to the suffix *-ce* (§9.6.1), but shows a different morphosyntactic distribution. It is added to


As illustrated in examples (128), (129) below, the second usage is roughly identical to the employment of *-ce*.

The suffix *-il* is used for the formation of referential attributes, i.e., lexemes with attributival meaning that are used as referring expressions and can make up their own phrase, but can also occur in apposition to a noun that they modify. In the latter case they occur in the position before the noun just like other nominal modifiers (adjectives, genitives, relative clauses). With non-verbal base words (i.e. expressions marked with the essive case) the suffix is required in order to turn the spatial expression into an attribute of the noun. Without the suffix the spatial expression would function as a modifier at the event level (the same was shown for -*ce* in §9.6.1.1 above). For instance, if we omit the suffix *-il* in (128), the meaning of the sentence would change to 'The calf fell down before him.' because now the spatial expression would function as adverb and modify the action expressed by the verb.

(128) a and [cin-na refl-gen sala-b-il] front-n-ref qːačːa calf k-ag-ur down-go.pfv-pret ' And the calf before him fell down.'

9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes

(129) q'ʷani-l-cːe-w-il box-obl-in-m-ref durħuˁ boy ʁaˁʁ scream ∅-ik'ʷ-ij m-say.ipfv-inf w-aʔ-išː-ib m-begin-put.pfv-pret ca-w cop-m 'The boy in the box started to scream.'

When added to verbs the resulting construction is a relative clause that can be restrictive or non-restrictive (see Chapter 23 for examples of both types). The use of *-il* in relative clauses is not obligatory when the relative clause occurs in its canonical position before the noun and there are only very few examples in my corpus (130). But in elicitation of relative clauses the use of *-il* is common.

(130) [du 1sg hak'-ub-il] appear.pfv-pret-ref di-la 1sg-gen šːi village 'the village where I was born'

When relative clauses occur in a position detached from the head noun, e.g. following it, the use of *-il* becomes obligatory. This happens because noun phrases are head-final and modifiers can never follow the noun they modify (e.g. demonstrative pronouns). However, a relative clause with the suffix *-il* forms its own phrase and can thus directly follow the noun as in (131) or even occur after the finite verb as in (132), a position that is commonly used to express afterthoughts (see §21.1.3 for the constituent order of the noun phrase and §27.1.3 for a discussion of extraposed modifiers).

(131) dam 1sg.dat b-ičː-ib n-give.pfv-pret iž this maˁlʡuˁn-ni snake-erg [ca one kur-re pit-loc ka-b-iž-ib-il down-n-be.pfv-pret-ref dawla.či-w rich-m Ismaˁʔil-li-cːella] Ismail-obl-comit

'The snake who sat in a pit together with the rich Ismail gave it to me.'

(132) iž=ra this=add het=ra, that=add het that ʡaˁχːuˁl guest ∅-iχʷ-ij m-be.pfv-inf [xːunul-la woman-gen qajqaj-li-cːe jaw-obl-in b-aˁq-ib-il] n-hit.pfv-pret-ref

' This also and this also is probably the man who hit the woman on the jaw.'

Constituents bearing *-il* are referential and can therefore occur without a head noun. This includes headless relative clauses (for headless relative clauses formed with the preterite participle the use of *-il* or -*ce* is obligatory), but also all other constituents. For instance, without the suffix the word in (133) would not be referential.

(133) paket-le-b-il?

parcel-loc-n-ref

'The (thing) in the parcel?' NOT 'Is it in the parcel?'

Therefore, the suffix is often found in topicalization constructions in which the topicalized constituent occurs to the left of the clause in (134) or in right-dislocated afterthoughts that provide more information on the referent such that its identification

9.6 Cross-categorical suffixes

is facilitated for the hearer (135). The referential attributes are often co-referenced by nominals in the clause, as in the following two examples:


In the next example (136) the referential attributes form topicalized noun phrases that are preceding the clause and are not co-referenced in the clause.

(136) c'il then di-la 1sg-gen qu-ja-b-il, garden-loc-n-ref di-la 1sg-gen b-ax-un-il, n-sow.pfv-pret-ref di-la 1sg-gen c'idex, fruit itil.ižili, one.thing.and.another agarud-le garden-loc ag-ur-re, go.pfv-pret-cvb kumek=ra help=add b-arčː-ib-le, n-find.pfv-pret-cvb di-la 1sg-gen r-iχ-ub-il=ra f-be.able.pfv-pret-ref=add b-arq'-ib-le, n-do.pfv-pret-cvb du=ra 1sg=add ka-r-iž-ib-le=da down-f-be.pfv-pret-cvb=1 ' Then, the (things) in my garden, my sown (fields), my fruits, all the stuff, I went to the garden, I found help, my things that I was able to do I did, and then I was

sitting (relaxing).'

The verb forms to which *-il* is added are able to take case markers (preceded by the oblique suffix *-li*) and then they function as referring expressions like nominals, i.e., as headless relative clauses (137–139) (see also §18.1.2.3 and §23.4). The other items that take *-il*, i.e. the spatial expressions in the essive case, are not further inflected. For example, the form *χe-w-il-la* in (137) functions as possessor marked by the genitive, and the possessum is the clause-initial noun *kep-dex*.

(137) kep-dex drinking-nmlz či-r-ag-ur spr-abl-go.pfv-pret ca-d, cop-npl hej this admi-la, person-gen heštːu-w here-m χe-w-il-la exist.down-m-ref-gen

'T he drunkenness passed, of this person who is here down (in the picture).'

(138) [b-ikː-an-il-li] n-want.ipfv-ptcp-ref-erg aq high či-ha-d-arq'-ib-le, spr-up-npl-do.pfv-pret-cvb d-učː-i npl-drink.ipfv-hab.pst heχ-tːi dem.down-pl

'The one who wanted (milk) lifted (the cans) up and drank.'

9 Predicative particles and other particles, conjunctions, and cross-categorical suffixes


There are three more uses/meanings of this suffix that have not been discussed so far. First of all, *-il* is used for the formation of experiential forms in the same way as it was mentioned for the suffix -*ce* in §9.6.1.1. Second, the suffix *-il* can be added to the locational/existential copulas when they are followed by the standard copula, such that the result looks like an analytic verb form. In this case, the use of the suffix restricts the meaning of the locational/existential copula to the existential meaning, excluding the locational meaning. Thus, the sentence in (141) cannot be translated by 'The son of Khalirbihin is (located) down there.'.

(141) heχ dem.down χalirbihin-na Khalirbihin-gen durħuˁ boy χe-w-il exist.down-m-ref ca-w cop-m 'The son of Khalirbihin exists (i.e. is still alive).' (E)

Third, when added to the preterite participle of the verb *ʔ-* 'say', the resulting verb form is used as a marker for ordinal numerals (§6.2). It is also part of the quantifiers *li<b>il* 'all', *har-il* 'every', and *b-aq-il* 'much, many'. With the first quantifier the use of *-il* is obligatory, i.e. \**lib*. To the other two quantifiers *-il* is only attached when they are used referentially (i.e. as predicates or arguments).

The constituents marked with *-il* occasionally have a flavor of contrastiveness, but this is a pragmatic implicature from the context, not part of the meaning of *-il*. Furthermore, many example sentences with *-il* do not have a contrastive meaning. For instance, in (129) the boy in the box is not contrasted with any other boy. Similarly, the fruits of the gardens and fields in (136) are not contrasted with other items.

Finally I will briefly compare *-ce*, *-te* and *-il*. The suffix *-ce* has a larger range of applications because it is added to a greater variety of base words. Almost all morphosyntactic contexts that allow for *-il* also allow for *-ce*, but not vice versa (which might be partially explained by the fact that *-ce* starts with a consonant, and can therefore follow consonants and vowels, but *-il* can only be added after consonants):


9.6 Cross-categorical suffixes

• The suffix *-il* shares with *-ce* the restriction to singular referents and thus partially competes with it. It is primarily used in relative clauses where its use is preferred over *-ce*, but also with nominals bearing the essive case. It cannot be used with adjectives (except for three quantifiers and the formation of ordinal numerals), nor can it occur in complement clauses.

All three suffixes are used in experiential verb forms.

### **9.6.3 The adverbializer** *-le*

The adverbializer *-le* (which has the variant *-lle* and the predictable allomorphs -*ne* and -*re*) forms (manner) adverbs from short adjectives and nouns (§7.5). It is also used for the formation of the simple converbs, i.e., the imperfective and the perfective converb, which are also widely used in analytic tenses (§18.1.1), and it can follow items bearing spatial cases such as adverbs (142) or nouns (143) that then also function like manner adverbials. As example (144) illustrates, it can also be used with nouns in the absolutive case. This sentence shows a copula construction in which the copula complement of the first clause has been turned into an adverbial by means of *-le*.


## **10 Place names and microtoponyms**

Tables 10.1 and 10.2 show names for the villages, towns, and districts that are relevant to the Sanzhi people. The tables first provide the citation form of the place name followed by the essive case, i.e. the word form that needs to be used when answering the question *Where are you?* The last two place names, Druzhba and Makahchkala, morphosyntactically differ from all the others because they represent recent borrowings. In order to form the essive case they need to employ the locational case suffix -*le* (§3.4.2.2). The other place names do not need such an additional spatial case because the place names have by themselves spatial meaning just like spatial adverbials because this is their default use. With these older place names it might diachronically be possible to identify a root morpheme that represents the place name followed by a spatial case suffix, but synchronically Sanzhi has no spatial cases that consist of a vowel *i* (the most frequent word-final segment of the place names in Table 10.1). Other Dargwa varieties such as Chirag (Ganenkov Submitted), however, have a spatial case expressed by a suffix *-i* that functionally resemble the Sanzhi locational case.

The third column contains referential-attributive terms that are semantically related to the respective places. These terms are formed by adding *-(a)n* to a root that can be the place name or some other root related to it. This suffix might be a cognate of the locative participle suffix *-an* (§18.1.2.4) and/or the interrogative clauses -*an* §18.1.2.2. Another possible cognate is the adjectivizer *-(a)n*, which is used for the formation of a few adjectives involving compounding with numerals and mostly plural nouns (§5.3). The same suffix seems to occur in the derivation of the adjective *b-urkːa-l-an* 'middle' from the postposition *b-urkːa* 'between'. In the default case, these terms refer to the inhabitants of the respective places as the term *the English* can refer to English people. They are also used as attributes of head nouns that do not refer to human beings but to their language, customs, clothes, etc., just like the use of *English* in the phrase *the English language*.

Syntactically, the referential attributes in the third column function like other referential attributes formed by means of the two cross-categorical suffixes -*ce* and -*il* (§9.6.1 and §9.6.2). This means that they largely possess the syntactic properties of nouns. They are used in argument position (6) or as predicates (9). They can also modify nouns as the last column 'language' shows. The constructions in the last column, which resemble compound nouns a bit (§3.6.3), can probably be analyzed as nominal appositions similar to the combination of proper names and kinship terms (§21.1.2). Just like with referential attributes that are marked with -*ce*, plural formation occurs by means of the most common plural suffix *-te* (6).

The fourth column contains terms referring to the ethnic group. These terms are a kind of mass nouns that trigger human plural agreement like the word *χalq'* 'people'. The last column contains the terms for the language. Language names contain the word *ʁaj* 'language', which is preceded by either (i) the singular term for the inhabitants, (ii) the genitive of the term for the ethnic group, or (iii) the genitive of the place name.

### 10 Place names and microtoponyms

Table 10.1: Names for villages, towns, and districts, and their inhabitants


As can be seen in Table 10.2, the noun *qːatːa* forms the essive case by changing the pitch accent to the final vowel (this is an irregular way to form the locational case; it is also found with a few other nouns). The noun *šːi* 'village' also has an irregular locational case form, whereas *dubur* is regularly inflected for either the loc-series (suffix -*le*, assimilated to -*re*) or the in-series (suffix -*cːe*). If not specified otherwise, *šːi* 'village' refers to the village of Sanzhi.

The place names only inflect for directional cases (essive, lative, ablative). As can be seen when comparing the two columns in Table 10.1, the place names mostly have directional meaning, i.e. the lative is identical to the place names themselves. Examples are



given in (1–3). In the speech of a few younger speakers (age 30 or younger) I noticed the use of the loc-series marker with the word *sːanži*, i.e., they used the explicit marking *sːanži-le* instead of *sːanži* when talking about going to the village (12). This might be due to Russian influence because Russian place names do not have inherent locative meaning, but require explicit case marking (in Sanzhi and Russian) as the last two lines in Tables 10.1 show.

(1) du 1sg priziw-li call-erg ka-∅-ač'-ib=da down-m-come.pfv-pret=1 urkaraqari Urkarakh 'I (masc.) came to Urkarakh by call.' (i.e. 'I was called to Urkarakh.')

(2) aba mother čːiħri-r=de Chakhri-f=pst cin-na refl.sg-gen ucːi-li-šːu-r brother-obl-ad-f

'My mother was in Chakhri, at her brother's place.'

(3) [There were four ways leading to our village,] ca one ce what či-b-il on-n-adjvz bek'-le-rka, head-loc-abl ca one uc'ari-rka, Itsari-abl ca one χudec'a-rka, Khuduc-abl ca one šaˁrʡaˁ-rka Shari-abl

' one through the peak on which there is something, one from Itsari, one from Khuduc, and one from Shari.'

The referential attributive terms and the terms for the ethnic groups (fourth and fifth column) inflect like standard nouns, for example *sungul* 'Sanzhi people', ergative *sungulli*, genitive *sungul-la*/*sungli-la*, dative *sungul-li-j*, and *sunglante* 'Sanzhi villagers', ergative *sunglan-t-a-l*, genitive *sunglan-t-a-la*, and so on.


10 Place names and microtoponyms

(6) ʡaˁrz complain w-arq'-ib-le, m-do.pfv-pret-cvb tusnaq prison w-arq'-ib m-do.pfv-pret urkuqan-t-a-l Urkarakh.person-pl-obl-erg 'The Urkarakh people complained and put him into prison.'

Table 10.3 displays terms for referential attributes that mostly denotate ethnic groups of the Caucasus and the names of the respective languages. Many of the referential attributes are also formed by means of the suffix *-an*. Some examples illustrating the usage are given in (7–9). As example (7) shows, the terms that contain genitive suffixes can also be used without head nouns (e.g. *ʁaj* 'language' in this examples) if the reference is clear from context.


Table 10.3: Ethnic groups


Some microtoponyms can be found in Table 10.4. The first column provides the citation form of the name and the second column the essive case form (all other spatial cases are formed accordingly). The second column shows that the essive forms are sometimes transparently built from the loc-series (-*le*) and in one case from the ad-series (-*šːu*) (10). All terms for microtoponyms do not contain morphemes that synchronically can be identified as spatial case suffixes (11). The third column provides explanations for those place names for which I was able to find one. Unfortunately not all place names are still remembered after more than 50 years since Sanzhi people resettled from their original village to the lowlands.

Table 10.4: Microtoponyms


(10) šaˁrʡaˁ Shari d-at 1/2pl-free aˁʁ-ib-le do.pfv-pret-cvb ca one kilametru kilometer k'e-b exist.up-n b-urkː-ar n-find.ipfv-prs hextːu-b there.up-n ʁʷaž-le-r Ghwazh-loc-abl či-d-a, on-1/2pl-dir muʁar-la Mughar-gen bek'-le-r head-loc-abl 'They sent us to Shari, it is probably one kilometer, through the hill Ghwazh, through the top Mughar.'

(11) han remember b-irk-u hpl-occur.ipfv-prs ix-tːi dem.up-pl šːik'e-b Shike-hpl 'As I remember, they were in Shike.'

# **Part III Verbal morphology**

The morphosyntactic categories of verbs in Sanzhi are person, gender, number, polarity, tense, mood, aspect, evidentiality, and voice. This chapter provides an overview of the formal make-up of simple verb stems (§11.2) and the general morpheme template of verbs in Sanzhi (§11.5), the formal means of expressing gender/number and person agreement (§11.3, §11.4), spatial preverbs and their meanings (§11.6), and polarity (§11.7) since these categories are largely independent of the TAME forms and voice. It concludes with an overview of the morphophonological processes that affect the formation and inflection of verbs (§11.8).

### **11.1 Overview of the general morphological structure of verbs**

Based on their morphological make-up, verbs can be divided into the following morphological classes:


There are comparably few simple verbal stems that can be used and are actually used without having undergone additional derivational or compositional operations. Most of the verbs are morphologically complex, either making use of one or more derivational affixes, and/or being compounds.

Examples of simple underived stems (including gender prefixes) are:

	- b. *b-uq'-* (ipfv)/*b-elq'-* (pfv) 'grind, mill'
	- c. *b-isː-* 'cry'
	- d. *b-ilʡ-* (ipfv)/*b-iʡ-* (pfv) 'steal'
	- e. *b-alχ-* (ipfv)/*b-aχ-* (pfv) 'know'

The derived verbs contain spatial preverbs (§11.6) and/or the causative suffix (§12.1). The compound verbs are of various types:


This chapter includes only information on spatial preverbs §11.6, because they form a closed class and mostly are in a particularly tight connection with the verbal root, which clearly differentiates them from non-spatial preverbs and other items used in verbal compounding. Causativization and compounding are treated in a separate chapter on verb formation (Chapter 12).

### **11.2 The structure of underived verbal stems**

Simple underived verbs have the structure *(C*<sup>1</sup> *)V(C*<sup>2</sup> *)C*<sup>3</sup> *(ː)*. The only consonants and semivowels that can occur in the *C*<sup>1</sup> slot are the two resonants *r*, *l*, the glottal stop, which obligatorily occurs before vowel-initial roots and which is not indicated in the spelling, and the pharyngeal stop, which usually occurs before pharyngealized vowels and is indicated in the spelling because it cannot be predicated. If we include also the verbs with gender agreement slots before the root and spatial preverbs, which are obligatorily used with some verbal roots, we have to add the exponents of gender agreement (*b*-, *r*-, *d*-, *w*-) and the consonants of deixis/elevation preverbs (*h*, *k*, *s*) as possible in the position of *C*<sup>1</sup> . No other consonants are allowed. This clearly differentiates verbs from nouns or adjectives, which do not have similar restrictions (see §2.3 for the general syllable and word structure of Sanzhi). In the position of *C*<sup>2</sup> only sonorants (*r, l, m*) and *b* are permitted, which conforms to a general requirement of the Sanzhi syllable structure (in nouns also *n* and *j* are allowed). If we include complex stems with deixis/elevation preverbs and with the stem vowel *i*, then we have to add *j* to the list. The slot *C*<sup>1</sup> allows for a far greater variety of consonants than the other two consonantal slots because only a few consonants are excluded (*p, p', b, l, m, n, r*, and the semivowels *w* and *j*). All vowels that Sanzhi has can occur as root vowel of verbs (including all pharyngealized vowels).

In Sanzhi, just like in all other Dargwa varieties, simple underived stems come in pairs that express the aspectual opposition between perfective and imperfective. This opposition is found in most TAM forms and is also preserved in non-finite verb forms such as participles and converbs. A very small number of finite and non-finite verb forms are available for perfective as well as for imperfective verb stems; most TAM forms can be built either only from imperfective or only from perfective stems. Here I will only describe the formal side of the aspectual opposition. Its meaning is treated in the sections on the respective inflectional verb forms.

The formal expression of the aspectual pairs is largely lexicalized and cannot be predicted. However, verbs can be divided into groups that follow the same patterns. The

11.2 The structure of underived verbal stems

two different aspectual stems are cognates that seem to be derived one from the other, but there is no unique direction of derivation. They can be distinguished on the basis of stem vowels, infix-like segments from a closed class of phonemes (only *r* and *l*), and the presence or absence of a gender agreement prefix.

In the following, I will briefly describe all patterns that can be identified and provide examples for them. Since there are verbs that are only used together with preverbs or other bound morphemes, the verbs given as examples will be morphologically simple and complex. The structure of complex verbs is indicated by dots and - for morpheme boundaries, and the verbs are given with the gender agreement prefix *b-* (except for the verbs that have a fixed agreement prefix).

### **11.2.1 Differences in gender agreement**

The structure of the verbs in Table 11.1 is completely identical. The only difference is the potential for gender agreement in their prefixal form only present in perfective stems.


Table 11.1: Differences in the gender agreement

### **11.2.2 Differences in the stem vowel**

The structure of the verbs in Table 11.2 is *V(C*<sup>1</sup> *)C*<sup>2</sup> *(ː)* with *C*<sup>1</sup> being *r, b,* or *m*. The vowel distinctions attested are *i* vs. *a, u* vs. *a,* and *u* vs. *e*.

### **11.2.3 Insertion of** *r* **in the imperfective stem**

The pattern in Table 11.3 occurs without and in combination with ablaut of the stem vowel. The structure of the verbal root is always *VrC(ː)* for the imperfective and *VC(ː)* for the perfective aspect. Many of these verbs have the same root vowel in the imperfective as well as the perfective stem, with the majority of verbs having *i*. Then there are a number of verbs that have diverging root vowels. Among them there are a few that occur only with a spatial preverb (*ka-* or *ha-*). Since there are regular morphophonological process of *a* + *i* > *e* and *a* + *a* > *a(ː)* we can assume that the root vowel of these verbs is *i* for the imperfective stem and *a* for the perfective stem.

Table 11.2: Differences in the stem vowel


*a (qum.a.art-* when negated)


Table 11.3: Insertion of *r* in the imperfective stem

### **11.2.4 Insertion of** *l* **in the imperfective stem**

Apart from two exceptions, the verbs in Table 11.4 are all of the structure *VlC(ː)* for the imperfective and *VC(ː)* for the perfective aspect, with identical vowels for both verbs.


Table 11.4: Insertion of *l* in the imperfective stem

### **11.2.5 Insertion of** *r* **in the perfective stem**

The verbs in Table 11.5 have the root structure *VC(ː)* for the imperfective and *VrC(ː)* for the perfective aspect. Vowels can either be identical or diverge. There are a number of verbs in this group that lack gender agreement prefixes for imperfective stems.

### **11.2.6 Insertion of** *l* **in the perfective stem (and usually** *l***-initial imperfective stem)**

The last group of verbs has *l* in the perfective stem, see Table 11.6. Most of these verbs have divergent stem vowels. The morphological make-up of the perfective verbs belonging to this group is always *VlC(ː)*. The structure of the imperfective verbs is either *VC(ː)* (only with very few verbs) or *lVC(ː)* (majority of verbs). In the latter case the verbs do not have a slot for a gender agreement within the root.


Table 11.5: Insertion of *r* in the perfective stem

Table 11.6: Insertion of *l* in the perfective stem


### **11.2.7 Verbs with only one aspectual stem and other morphologically exceptional verbs**

There are a number of defective verbs that lack the second member of the aspectual pair and only have one stem (Table 11.7). This single stem inflects for the verb forms that are normally only or at least predominantly formed from the imperfective stem (e.g. imperfective converb, modal participle, prohibitive) as well as for verb forms that are normally only or at least predominantly formed from the perfective stem (e.g. preterite, perfective converb, imperative). In the following, I will simply call these verb forms perfective and imperfective TAM forms.

There are a few exceptional verbs that have restricted possibilities for inflection. These verbs are:



Table 11.7: Stems inflecting for all TAM forms (imperfective and perfective)

### **11.3 Gender agreement in verb stems**

Gender agreement is an important grammatical category of East Caucasian languages, and also present in Sanzhi. Most of the vowel-initial verbal stems and the two preverbs *b-i-* 'in(side)' and *b-it-* 'thither' have gender agreement prefixes. Furthermore, the locational/existential copulas (§16.2) and the copula-auxiliary *ca-b* have a slot for gender agreement suffixes (or infix in the variant *ca<b>i,* see §16.1). The agreement affixes are displayed in Table 11.8.

Table 11.8: Gender agreement affixes in Sanzhi


The agreement affix for masculine singular is always used when it occurs as a suffix. It is regularly omitted when it occurs as a prefix to a verbal root beginning with *u*, for example *ukː-unne=da* (masc.) vs. *r-ukː-unne=da* (fem.) (eat.ipfv-icvb=1) 'I will eat'. It is optionally omitted when the root starts with *i*, for example *(w-)ik'-ul* (masc.) vs. *r-ik'-ul* (f-say.ipfv-icvb) 'saying'.

The agreement prefixes disappear when the preverb *b-it-* is attached, which contains its own gender prefix (see §11.6.2 for examples). However, if the preverb is preceded by a negation prefix, then the gender agreement can be completely omitted, but such an omission is optional. Thus, the verb *b-it-eʁ-ij* (n-thither-go.pfv-inf) 'go there' has the neutral negative form *a-jt-eʁ-*, which is not specified for gender, alongside with the forms preserving the gender prefixes *a-b-it-eʁ-, a-w-it-eʁ-, a-r-it-eʁ-,* and *a-d-it-eʁ-*.

Verbal gender agreement has the clause as its domain, and in the majority of cases it is controlled by the absolutive argument of the agreeing verb. The syntax of gender agreement is treated in detail in §20.2.

### **11.4 Person agreement and stem augment vowels**

Person agreement is rather reduced, with a clear opposition of speech act participants (first and second person) vs. third person. Formally it shows up as suffixes and as enclitics. The form of the agreement exponent varies depending on the TAM form, and not all TAM forms have person agreement markers. The following verb forms distinguish person agreement:


11.5 The morpheme template of Sanzhi verbs

In the habitual present, the realis conditional and the past conditional; the person suffix for first and second persons is preceded by a stem augment vowel (*i, u,* or occasionally *a*) that indicates the valency of the verb (monovalent vs. bivalent or trivalent). Throughout this grammar, the stem augment vowel is not glossed separately, but together with the following TAM suffix. For full lists of the agreement exponents and the distribution of stem augment vowels see §20.3.

Person agreement has the clause as its domain, and the rules are rather complex and subject to variation. With monovalent predicates, it is the single argument that functions as agreement controller. With predicates that require more than one argument, only subject-like arguments (agents or experiencers) or object-like arguments (patients or stimuli) control person agreement. Person agreement follows the person hierarchy 1, 2 > 3. In the case of two speech act participants, it is often the second person that triggers the agreement, but first person subject-like arguments are also able to control agreement. The syntax of gender agreement is treated in detail in §20.3.

### **11.5 The morpheme template of Sanzhi verbs and the structure of morphologically complex verb forms**

The morphological structure of verbal predicates in Sanzhi is fairly complex. There are up to five morphemes that can precede the root and up to five that can follow it. These morphemes can be prefixes, and suffixes, but also enclitics and lexical stems functioning as first parts of compound verbs. Before the root, there are only prefixes in the form of spatial preverbs, gender/number prefixes and negation prefixes and lexical stems used in compounds. After the root, suffixes and enclitics follow. There are restrictions on the combinability of markers in the various slots, for instance TAM forms requiring person suffixes exclude the use of enclitic person or tense markers.

Table 11.9 provides a template for Sanzhi verbs. The slots are, from left to right:

**5-** first part of a compound verb (there are a few preverbs and stems used in compounding that have gender prefixes as one example in the table shows) (see §12.2 on compounding);

**4-** location preverb, optionally followed by a direction suffix that can only occur together with a preverb; the preverb b-i- 'in, inside' has an additional gender marker (§11.6.1)


**0** root


### **-3** second TAM slot

### **-4** person and tense enclitics (§11.4)

There are two slots ([5-] and [4-]) that contain items that can have gender prefixes such that the structure can even be a bit more complicated. Since only very few items in both slots are marked for gender, I did not add two additional slots for gender to the template. The slots and respective morphemes are treated in various sections of this grammar.


Table 11.9: Verb affixation order template

In principle, only the verbal root is obligatory because there is a variant of the optative that does not make use of any suffixes. There are a number of verbal roots that are bound and can only be used in combination with spatial preverbs, for example *kerxʷ-* (ipfv)/ *kaxʷ-* (pfv) 'kill', and *kert'-* (ipfv)/*kat'-* (pfv) 'pour'.

11.6 Spatial preverbs

### **11.6 Spatial preverbs**

Sanzhi Dargwa has the typical Dargwa system of preverbs, which in their original spatial meaning express location, direction, and deixis/elevation (see van den Berg 2003c for a useful overview of preverbs in Akusha Dargwa). Preverbs are generally optional, because verbs can occur without preverbs, but there are bound verbal roots for which the prefixed preverbs are obligatory.

There is a tight connection between spatial preverbs and the verbal stem, and normally they form one phonological word. The order of the preverbs is given in (2). Between the complex location/direction and the deictic preverbs, only negation prefixes and some enclitics (e.g. the additive *=ra* and *arrah* 'at least') can intervene.

(2) [(location)-(direction)]-(deixis/elevation)-root

Preverbs do not express aspectual differences, but occur with imperfective and perfective stems. The Sanzhi Dargwa system of preverbs can be characterized as being somewhere in-between regular, productive, and semantically transparent systems, like the ones found in Agul, Tabasaran, and Rutul, and non-regular systems as, for instance in Budukh, Kryz, Tsakhur, and Lezgian (Tatevosov 2000; Nichols 2003; Ganenkov 2007). It is at least partially formally compositional, in the sense that all theoretically possible combinations of location/direction and deictic preverbs are attested (§11.6.3). However, not every verbal stem takes all available preverbs or logically possible combination of preverbs. With verbs of movement and posture, the semantic contribution of the preverbs is relatively straightforward and compositional (see Table 11.11 for an example), but with most other verbs there is no real semantic transparency and the spatial meaning of the preverb is lost.

Preverbs have probably developed from spatial postpositions/adverbs, but not all spatial postpositions/adverbs are used as preverbs. For instance, *ilda* 'on the side, sideways' and spatial adverbs derived from demonstrative pronouns do not occur as preverbs. The directional markers are identical to the directional markers used for the formation of spatial cases.

### **11.6.1 Location preverbs and spatial cases expressing direction**

Location preverbs, just like the spatial cases, express location and direction. All preverbs in Table 11.10 except for the last one are identical to spatial postpositions (§8.1), though there are more spatial postpositions that are not used as preverbs. They express the location of an item with respect to a reference point. The directional affixes can be added to the preverbs (*-r* for the ablative, the gender marker for the essive, no affix for the lative), which are the same used with nominals or spatial adverbs (§3.4). The directional affixes are suffixes to the location preverbs, not prefixes to the verbal stem, because they cannot occur without location preverbs and semantically modify the meaning of the location preverbs. Thus, location preverbs and directional suffixes form a tight union.


*<sup>a</sup>*This preverb can also be used with respect to locations that do not have hands (e.g. animals, etc.). Thus, the meaning is not literally 'into the hands' anymore, and speakers do not translate it with 'into the hands'.

Examples of the preverbs with and without markers for directed motion are provided in (3–7).


For a number of verbs, the spatial semantics has been lost and has developed into a more metaphorical meaning (8). Furthermore, with verbs that do not denote the position or the movement of an item, the semantic contribution of the preverbs is synchronically opaque (9–10).

(8) dam 1sg.dat či-d-d-ač'-ib-te spr-npl-npl-come.pfv-pret-dd.pl '(It is enough what) I experienced.'

11.6 Spatial preverbs

(9) iχ dem.down gu-lik'-un sub-listen-pret ca-w cop-m ħaˁq'-le very-advz qːuʁa-l beautiful-advz 'He is listening carefully to him.'

(10) dam 1sg.dat il that či-a-b-až-ib=da spr-neg-n-see.pfv-pret=1 'I did not see it.' (E)

Because preverbs are identical to postpositions and adverbials, it is not always possible to determine whether a specific item functions as the one or the other. For instance, *či-r* in the following example (11) is interpreted as preverb by my main language assistant Gadzhimurad Gadzhimuradov, although the combination *urči-le-r=či-r* also exists as a postpositional phrase 'from on the horse'. In the example (11) the constituent order can be changed to *či-r-ka-jč-ib urči-le-r*, which excludes an interpretation of *či-r* as postposition and supports the preverb analysis. It is also possible to use both the postposition/adverbial and the preverb (12). See also §8.1.7 for some more examples in which *či-r* rather functions as postposition/adverbial and not as preverb.


For the most part preverbs can not be separated from the verb or follow it (13), but in certain contexts (that still await clarification) a separation is possible, just like it has been observed for Tanti Dargwa (Sumbatova & Lander 2014: 107) (14).

(13) \* admi person ka-jč-ib down-occur.pfv.m-pret či-r on-abl (Intended meaning: 'The man fell down.') (E) (14) a. či-r-ixʷ-a spr-abl-remove.pfv-imp qːatːi! hat 'Take off the hat!' (E) b. *ixʷ-a či-r qːatːi!* 'Take off the hat!' (E) c. *ixʷ-a qːatːi či-r* 'Take off the hat!' (E)

### **11.6.2 Deixis and elevation preverbs**

The participant-oriented and elevation preverbs, which are all deictic, are:


These preverbs immediately precede the verbal root, and the only items that can intervene are gender agreement prefixes. However, if the preverb gm-*it-* is added to verbal stems possessing a gender prefix, this prefix is omitted, e.g. *či-b-uq-ij* 'attack, hit on, fall upon' vs. *či-b-it-uq-ij* 'go on (something)', and *gu-b-aˁq-ij* 'beat from down' vs. *gub-it-aˁq-ij* 'beat from down'. The deictic/elevation preverbs cannot take the directional suffixes since they already convey motion.

The preverbs express upwards or downwards motion (elevation) with respect to a deictic center and motion to the speaker and away from the deictic center, which is usually the speaker (participant-oriented deixis). Relevant examples are (15–18).


To younger speakers of Sanzhi, the specific meanings of the preverbs *ha-* and *ka-* are not fully clear anymore, and they usually employ only *sa-* as the default form. Older speakers differentiate between:


11.6 Spatial preverbs

In contrast, the younger speakers use *sa-b-eʁ-ij* for all contexts.

Among the participant-oriented deixis and elevation preverbs, the first three preverbs *ha-, ka-,* and *sa-* are far more commonly used than the fourth preverb gm-*it-*. For instance, the verb 'carry' takes the first three preverbs, but not the fourth, that is *haqː-, kaqː-, saqː-, \*b-it-aqː-*. It still needs to be clarified whether this is due to formal reasons (presence vs. absence of gender prefixes or morphophonological restrictions) or can be explained functionally.

Tatevosov (2000) observes that the meanings of the deictic and elevation preverbs do not exclude each other, that is a movement can be upwards away from the speaker, but only one of these meanings can be realized through the use of the relevant preverb. Which preverb is actually employed seems to be an idiosyncratic lexical property of the verbal predicate.

The origin of the participant-oriented deixis and elevation preverbs is often opaque. The preverb *ka-* is possibly related to the second part of the morphologically complex ablative suffix *-r-ka,* but the synchronic semantic contribution of *-ka* in the ablative suffix is hard to determine. The deictic preverb *sa-* 'to the speaker, hither' is formally identical to the locational preverb *sa-* 'in front', the spatial case *-sa* 'in front', and the postpositions/adverbs *sala* 'in front of' and *sa-b* 'in front, ago'. If this is due to chance or due to cognacy requires further investigation.

A detailed account of the elevation preverbs in Sanzhi within a general discussion about the semantic category of 'elevation' in Sanzhi can be found in Forker (2019a).

### **11.6.3 Combinations of preverbs**

The two groups of preverbs are independent of each other, in the sense that location can be expressed without participant-oriented deixis/elevation and vice versa, but they can also be combined. The order is fixed, with participant-oriented deixis/elevation preverbs occurring closer to the stem (19), (20). Reverse combinations are ungrammatical.

(19) či-ka-b-ixː-a! spr-down-n-put.pfv-imp 'Put (it) down!'

(20) hel=ra that=add b-i-k-ert'-id n-in-down-pour.ipfv-1.prs 'I pour that in as well.'

The location preverbs have optional allomorphic variants when followed by the participant-oriented deixis/elevation preverb *ha-* because the initial fricative of the second preverb is omitted. The preverbs ending in *-i* change that vowel to *-e* under omission of *h*, for example *či-* + *ha-* > *če-, kʷi-* + *ha* > *kʷe-*, *hitːi-* + *ha* > *hitːe-*. The stop in the preverb *gu-* becomes labialized, that is *gu-* + *ha-* > *gʷa-* (21).

(21) c'a fire gʷa-b-iq'-un from.down.up-n-set.fire.pfv-pret ca-b cop-n '(Somebody) set up a fire.'

All location preverbs can be combined with all participant-oriented deixis/elevation preverbs, and almost all of the logically possible combinations are attested in natural texts. However, many if not almost all verbal roots allow only for certain preverbs and combinations of preverbs to be attached. Verbs of motion and position have, of course, the greatest freedom, since the preverbs have spatial and directional meanings. Combinations with the location preverbs *či-, hitːi-*, and *tːura-* and the participant-oriented deictic preverbs *ka-* and *ha-* are significantly more frequent than combinations with the other preverbs. Combinations with gm-*it-* are generally very rare in natural texts. Table 11.11 illustrates all combinations with the verb *b-eʁ-ij* (pfv) 'go, come'.

The negation prefixes *a-* and *ma-* follow the location preverbs and precede the participant–oriented deixis/elevation preverbs if there are any (22), (23). It is also possible to insert three particles that normally occur as enclitics into the same slot following the location preverbs. These clitics are the additive *=ra* (23), the emphatic particle *=q'ar* (24), and *=arrah* (see §9.4.5 for an example). If the enclitics occur in that position, they take scope over the verb.

(22) tːura-ma-ka-lq-ut! out-proh-down-direct.m-ipfv-proh.sg

'Do not go out (of the car)! (said to a man)'

(23) ca-w refl-m w-erc-aq-ur-il-li-j m-save.pfv-caus-pret-ref-obl-dat er look či=ra-a-w-erč'-ib spr=add-neg-m-look.pfv-pret

'He did not even look at his savior.'

(24) gu=q'ar-lik'-unne sub=mod-listen-icvb ca-b cop-n 'As for listening, s/he is listening.'

### **11.7 Negation**

Negation can be expressed through prefixes or through the negative copula, depending on the inflected verb forms. In contrast to some other South Dargwa varieties (e.g. Icari, Shiri), Sanzhi Dargwa does not express negation through reduplication of the verbal stem. There are two negative prefixes *a-* and *ma-* that occur right before participantoriented deixis/elevation preverbs and root-initial gender markers if there are any. The prefix *a-* is occasionally preceded by an additional gender agreement prefix.

The functional distribution of the negation prefixes is as follows: the prefix *a-* is used in the imperfect/preterite, resultative, pluperfect, experiential past, and sometimes also with the perfect and with non-finite verb forms. The prefix *ma-* is only used in the prohibitive and the negative optative. For all other verb forms the negative copula is employed. The negative copula has the root *akːʷ-*<sup>1</sup> (allomorphs *akʷ-, akː-)*, of which the

<sup>1</sup>The negative copla has prefixal gender agreement when it is used with locational or existential meaning, but this is impossible when it is used for the formation of analytic verb forms. See §16.1 for more information.


Table 11.11: Location preverbs with *b-eʁ-ij* (pfv) 'go, come'

initial vowel is dropped when it is encliticized to a preceding predicate, so that we get *=kːu* and *=kːʷi*. The negative copula occurs in four forms: present, past, participial, and masdar. See the sections on the TAM forms and §16.1 on the copula for examples of negated predicates.

### **11.8 Morphophonological processes affecting the formation and inflection of verbs**

There are a number of regular morphophonological processes that occur when verbs are inflected and that lead to the formation of stem allomorphs. These processes are in part optional, but occur frequently. See §2.6 for more information about the processes, their application and alternative variants.


## **12 Verb formation**

There are three types of operations that allow for the formation of complex verbal lexemes from base verbs:


In this section, causativization and compounding are discussed.

### **12.1 Formation of causative verbs**

Causativization is a productive means of deriving causative verbs from base verbs. It can be applied to most if not all verbs, including intransitive, transitive and affective verbs of imperfective and perfective aspect. The causative suffix -*aq* is added directly to the stem prior to TAM suffixes and it does not have any impact on the aspectual value of the verb or on the choice of certain inflectional suffixes (e.g. which suffix is used for the preterite). It has a pharyngealized allomorph *-aˁq*. Furthermore, suffixation of the causative marker triggers palatalization of velar consonants in the verbal root. Examples of causativized verbs and their meanings are given in (1).

	- b. *b-ikː-ul* 'wanting, liking, loving' (aff.) (n-want.ipfv-icvb) > *b-ičː-aq-ul* 'make wanting, liking, loving' (tr.)

In the majority of cases, causativization adds one argument to the valency frame of the base verb, i.e. intransitive verbs become transitive and transitive verbs become ditransitive. Causativization normally applies only once to the verbal stem, but in elicitation the causative suffix can also be added twice to a small number of verbs. However, due to the scarcity of examples the syntax and semantic properties of verbs that underwent double causativization could not be clarified. With the verb exemplified in (2) the meaning seems to be more emphatic, and the valency frame is transitive (as after single causativization).

	- > *b-iħ-aˁq-aˁq-ib* 'made fight' (tr.)

### 12 Verb formation

In addition to morphological causativization, there are other formal means for making causative constructions such as light verb change and suppletion. This operation is applied to compound verbs. Intransitive compound verbs make use of the light verbs *b-irχʷ-* (ipfv)/*b-iχʷ-* (pfv) 'be, become, can' (3). For causativization these light verbs are replaced by *b-irq'-* (ipfv)/*b-arq'-* (pfv) 'do, make' (4). A full list of available light verbs is given in §12.2.1.


See §19.2.2 for more information on the syntactic properties of causativization and more examples of causativized verbs.

### **12.2 Compound verbs**

Verbal compounds consist of two parts, the first of which can be a noun, short adjective, ideophone, bound lexical stem, or, very rarely, another verbal stem. It can be a native lexical item or a loan word. Thus, compounding is a convenient way of extending the verbal lexicon. The second part is a light verb from a closed class of verbs. It is only the light verb that is inflected and that determines all morphosyntactic properties of the compound. All light verbs used in compounding are given at the beginning of this Section in §12.2.1 with both the imperfective and the perfective stems.

### **12.2.1 Light verbs used in compounding and general remarks on compounds**

There are a fair number of light verbs occurring in compounding. The most frequent ones are *b-iχʷ-ij* 'be, become, can', *b-ik'ʷ-ij* 'say' and *b-arq'-ij* 'do'. The verb *b-ik'ʷ-ij* is widely used in compounds that denote verbs of speech and the production of other sounds, but also in many verbs of movement. Since it means 'say' when used on its own, I stick to this as an overall gloss. The two tables below display intransitive (Table 12.1) and transitive (Table 12.2) light verbs.

The verb-forming processes listed in the following sections of this Chapter are relatively freely combinable. If a specific combination is available depends largely on the semantics of the resulting verb. Thus, compound verbs can contain spatial preverbs (5–7) and causativized verb stems.

12.2 Compound verbs


Table 12.1: Intransitive light verbs


Compounding is possible with loans. Some of the nouns listed in §12.2.2 and adjectives listed in §12.2.3 have been borrowed from other languages such as Arabic, Persian or Turkic. In the last 50 years mainly Russian borrowings entered the language. Two examples

### 12 Verb formation


Table 12.2: Transitive light verbs

with Russian loans are given in (8). The first compound verb contains the infinitive of a Russian verb and the second an adverb.

(8) a. *kupatsa b-ik'ʷ-* 'bathe, take a bath' b. *žalka ag-* 'feel sorry for'

### **12.2.2 Compounds with nouns**

Many compound verbs contain a noun. These nouns are often loan words. The nouns in the compounds are non-specific indefinite and can normally not be modified or referred to by anaphoric pronouns. They occur in the absolutive case, or occasionally in the genitive or are marked by spatial postpositions.

The compound verbs can be intransitive or transitive. For intransitive verbs, the noun that is part of the compound verb cannot control the agreement (Table 12.3) (9). By contrast, the noun that serves as the subject-like argument and occurs in the absolutive controls the gender agreement on the verb. Some of the compound verbs can take clausal complements (10).


Table 12.3: Examples of intransitive compound verbs

Table 12.4: Examples of transitive compound verbs


### 12 Verb formation


'Once the sun and the evil wind argued about who is stronger.'

For transitive verbs the subject-like argument is in the ergative and the gender agreement is almost always controlled by the noun that is part of the compound. This means that the gender agreement is fixed, mostly for neuter singular. Additional arguments fulfill the semantic functions of addressees, recipients or beneficiaries and occur in the cases that are used to express these semantic roles, e.g. dative or in-lative. Examples are provided in Table 12.4 and (11).

(11) cara other adim-t-a-cːe person-pl-obl-in […] heχ dem.down tiladi request b-arq'-ib-le n-do.pfv-pret-cvb … 'when (we) asked other people, …'

However, there is at least one transitive compound verb containing a noun in the absolutive case for which not the subject-like argument, but the noun that serves as the direct object triggers the gender agreement, namely *taman* 'end' + *b-arq'-* (hpl-do.pfv-) 'finish (off), terminate' (12). And in the example in (13) the agreement prefix on the light verb does not agree with any overt noun. The first part of the compound, the noun *tilipun* 'telephone' belongs to the neuter gender. If it functioned as the object of the light verb it would trigger the prefix *b*-.


There are some nouns that are particularly productive for the formation of compounds verbs and can combine with a variety of light verbs. One is the noun *ʁaj* 'word, talk, language', that occurs in the following compounds:

	- a. *ʁaj (ka-)b-ik'ʷ-ij* (down-hpl-say.ipfv-inf) 'say, tell'
	- b. *ʁaj (ka-)b-uq-ij* (down-hpl-go.pfv-inf) 'chat, talk, communicate, converse'
	- c. *ʁaj ha-b-iž-ij* (up-hpl-be.pfv-inf) 'chat'

12.2 Compound verbs

	- a. *ʁaj d-arq'-ij* (npl-do.pfv-inf) 'say, tell'
	- b. *ʁaj d-urs-ij* (npl-tell.pfv-inf) 'say, tell'
	- c. *ʁaj b-ičː-ij* (n-give.pfv-inf) 'promise'

In addition to the noun+verb compounds there are constructions that resemble those compounds but contain nouns in the genitive. The verbs used are *b-arq'-* 'do, make' and *b-iχʷ-* 'be, become' and a few other intransitive and transitive verbs (16), (17). The nouns in the genitive case do not serve any argument functions in the clause, but form compounds together with the verb and thus contribute to the semantics of the predicate. In the predicates in (16h) and (16i) the genitive-marked nouns resemble instruments, but this cannot be said about the other predicates. Note that the last two examples in (16) differ from the others because in both cases the genitive can be explained by the morphosyntactic properties of the construction. The postposition *hitːi* in (16j) generally requires the genitive case and thus *ʁaj-la hitːi d-urs-ij* consists of a postpositional phrase followed by a verb. In (16k) the genitive functions as the modifier of the following noun such that we have a genitive phrase together with a verb. However, semantically both constructions function as compound predicates analogously to the other constructions with genitive-marked nouns. All compound predicates derive their transitivity from the transitivity of the base verb. If the base verb is intransitive the compound verb is also intransitive (16a); if the base verb is transitive, then the compound is also transitive (17). Some more examples sentences can be found in §3.4.1.3 and in §19.1.2.

	- a. *qal-la b-iχʷ-ij/qal-la ka-b-iž-ij* (house-gen hpl-be.pfv-inf/house-gen downhpl-be.pfv-inf) 'get married'
	- b. *waˁʡda-la b-iχʷ-ij* (contract-gen hpl-be.pfv-inf) 'negotiate, conspire'
	- c. *abdal-la b-arq'-ij* (fool-gen hpl-do.pfv-inf) 'take for a fool'
	- d. *qaˁb-la b-arq'-ij/qaˁb-la + b-aˁq-ij* (neck-gen hpl-do.pfv-inf/neck-gen hplstrike.pfv-inf) 'behead'
	- e. *qal-la r-arq'-ij/qal-la ka-r-at-ij* (house-gen f-do.pfv-inf/house-gen down-flet.pfv-inf) 'marry off'
	- f. *χːaˁb-la b-arq'-ij* (grave-gen hpl-do.pfv-inf) 'bury'
	- g. *dawla-lla b-arq'-ij* (wealth-gen hpl-do.pfv-inf) 'congratulate, bless'
	- h. *itul-la b-arq'-ij* (iron-gen n-do.pfv-inf) 'iron with an iron'
	- i. *qːupi-lla b-arq'-ij* (hoe-gen n-do.pfv-inf) 'weed'
	- j. *ʁaj-la hitːi d-urs-ij* 'gossip' (word-gen after nhpl-tell.pfv-inf)
	- k. *ʡaˁjib-la (w-ah) w-arq'-ij* (blame-gen hpl-owner hpl-do.pfv-inf) 'consider to be guilty'<sup>1</sup>

<sup>1</sup>The noun *w-ah* 'owner' can be omitted in this construction.

### 12 Verb formation

(17) *abdalla + b-arq'-ij* 'take for a fool' abdal-la fool-gen w-arq'-ib=q'al m-do.pfv-pret=mod itːa-l those.obl-erg it that 'They took him for a fool.' (E)

There is also one compound verb, which contains a noun marked with a spatial case (18).

(18) loc-lative case *(cin-na) ʁaj-le či-ka-b-icː-ij* 'to be true to one's word' (refl.sg-gen word-loc spr-down-hpl-stand.pfv-inf)

Finally, there are compound verbs that contain nouns with the encliticized postpositions *=či* 'on' and *=(i)tːi* 'after'. These postpositions govern the genitive or spatial cases (§8.1.4, §8.1.7), but when they are used in verbal compounding, they are directly added to the nouns without case marking:

```
(19) spatial postposition/adverb =či
```
	- a. *er=itːi sa-b-erč'-ij* (look=after ante-hpl-look.pfv-inf) 'look around, check, inspect'
	- b. *er=itːi b-ik'ʷ-ij* (look=after hpl-say.ipfv-inf) 'look (at)'
	- c. *qus=itːi b-aˁq-ij* (slip=after n-drag.pfv-inf) 'pull, drag along, after oneself'
	- d. *dukal=tːi ka-b-iħ-ij* (smile=after down-hpl-aux.pfv-inf) 'smile about somebody'
	- e. *ħaˁħaˁ=tːi b-ik'ʷ-ij* (laughter=after hpl-say.ipfv-inf) 'laugh at/about someone'

### **12.2.3 Compounds with short adjectives**

The short adjectival stems (§5.2) can easily occur in compound verbs together with the light verbs *b-iχʷ-* (pfv) 'be, become, can' (21), *b-ik-* (pfv) 'occur' (22), and *b-arq'-* (pfv) 'do, make' (23). These verbs occur in pairs of intransitive verbs that normally have an inchoative meaning and transitive verbs (24–25).

(21) intransitive compounds with the light verb *b-iχʷ-* (n-become.pfv-)


12.2 Compound verbs



Occasionally, other light verbs are used, which leads to more idiosyncratic meanings (26):

(26) qːuʁa-ce, beautiful-dd.sg ʡaˁħ good ka-b-icː-ur down-n-stand.pfv-pret musːa place het that ca-b cop-n 'This is a beautiful, pleasant place.'

### **12.2.4 Compounds with ideophones**

Sanzhi has a fair number of ideophones that combine not only with verbs of speech, but also with other light verbs and auxiliaries (Table 12.5). The resulting compound verbs denote the production of various sounds as well as verbs of movement and other activities that are accompanied by typical sounds (27–29).



Table 12.5: Examples of compound verbs with ideophones

12.2 Compound verbs

### **12.2.5 Compounds with bound lexical stems**

There is a closed class of bound lexemes that occur only in compound verbs and of which thus the meaning out of the context of a compound verb is impossible to determine. These items do not belong to any of the lexical categories that Sanzhi has. Some of the bound stems are flexible with respect to the light verbs with which they combine leading to a variety of different compound verbs containing the same bound stem (Table 12.6).

For instance, the bound stem *taˁħ* occurs together with verbs of movement or posture to yield the meaning 'jump', but it also combines with other verbs. The resulting compounds always denote movement away from a source (30–32).


Other bound stems combine only with one or two light verbs (Table 12.7, Table 12.8). Among them the verbs *b-ik'ʷ-* (n-say.ipfv-), *b-arq'-* (n-do.pfv-) and *b-uq-* (hpl-go.pfv-) are particularly frequent.

As with the compound verbs containing short adjectives (§12.2.3), there are often pairs of intransitive and transitive verbs. They can be divided into groups depending on the intransitive verbs that they make use of. Firstly, there are bound stems that are combined with *b-iχʷ-* (pfv) 'be, become, can' and *b-arq'-* (pfv) 'do, make' to form intransitive and transitive verbs, see, for instance, the examples in (33). Other light verbs cannot be used together with these stems.

(33) a. *haj b-iχʷ-*/*haj b-arq'-* 'move, drive' b. *b-ars b-iχʷ-*/*b-ars b-arq'-* 'change'

And secondly, there are bound stems that are combined with *b-ik-* (pfv) 'occur' and *b-arq'-* (pfv) 'do, make' to form intransitive and transitive verbs:

(34) a. *can b-ik-*/*can b-arq'-* 'mix, unite, meet'

b. *suk b-ik-*/*suk b-arq'-* 'meet, gather'

c. *šak b-ik-*/*šak b-arq'-* 'guess, suspect, feel'


Table 12.6: Compound verbs with bound lexical stems (Part 1)


Table 12.7: Compound verbs with bound lexical stems (Part 2)

### 12 Verb formation


Table 12.8: Compound verbs with bound lexical stems (Part 3)

Occasionally, stems can be combined with more than one intransitive auxiliary, e.g. (35).

(35) a. *ʁudur* 'mix' + *b-iχʷ-* (pfv) 'be, become, can' and *b-ik-* (pfv) 'occur' b. *qus* 'slip' + *b-ik'ʷ-* (n-say.ipfv-) and *b-ig-* (n-be.pfv-)

There are a couple of compound verbs in which the first part synchronically seems to be a verb or diachronically to originate from a verb (36). However, the compounds express verbal aspect only via the stem alternation of the second verb; the first part is invariable and not inflected except for the gender/number prefixes, which agree in exactly the same way as the prefixes, which belong to the inflecting verb (37–40).

	- b. *icːaχː-* (pfv)/*icːalχː* (ipfv) 'start to hurt' < *icː-* (ipfv) 'hurt, ache' + ?
	- c. *b-it'-b-ak'-* (pfv)/*b-it'-b-ik'-* 'pull, draw, move' < *b-it'-* (pfv) 'lure out of, from' + *b-ak'-* 'grow'?
	- d. *us.kelg-* (pfv)/*us.kalg-* (ipfv) 'go to sleep, fall asleep' < *usː-* 'lie' (pfv) + *kelg-* (pfv) 'remain, stay'
	- e. *b-iχ-(b)-it-ag-* (pfv)/*b-iχ-(b)-it-arg-* (ipfv) 'believe' < *b-iχː-* 'believe' + preverb *b-it-* 'thither' + *ag-* (pfv) 'go'
	- f. *b-iχ-čeg-* (pfv)/*b-iχ-čerg-* (ipfv) 'believe' < *b-iχː-* 'believe' + *či-ag-* (pfv) (spr-go)

### 12.2 Compound verbs


## **13 Indicative synthetic verb forms**

Sanzhi Dargwa has only two indicative synthetic verb forms that head independent clauses, the habitual present (§13.1) and the habitual past (§13.2). They are formed by adding stem augmentation vowels and person agreement markers to verbal stems that have imperfective aspect. The stem augmentation vowels occur only with first and second person forms and are also used in conditional clauses with synthetic verb forms (Chapter 18.3). They are *u* for intransitive verbs and *i* for transitive verbs in the habitual present, and *a* for all verbs in the habitual past (with the exception of the verb *b-aχ-* (pfv)/*b-alχ-* (ipfv) 'know', which also has *a* as the stem augmentation in the habitual present). The stem augmentation vowels are not separately glossed in the examples, but given together with the person/tense suffixes.

### **13.1 Habitual present**

The habitual present is formed by adding person suffixes to the augmented stem of imperfective verbs (Table 13.1). The third person has always the suffix *-u*, which can thus be interpreted as a person marker, although it most probably originates from the stem augmentation vowel for intransitive verbs. Alternatively there is the suffix *-ar* for the third person (see below for a discussion). Table 13.2 shows paradigms of three verbs. The intransitive verb 'say' is given in the female form for singular persons.

Table 13.1: Person suffixes for the habitual present (without stem augmentation vowels)


Table 13.2: Some illustrative paradigms of the habitual present


### 13 Indicative synthetic verb forms

### **Semantic domains**

	- (1) duq-n-a-lla egg-pl-obl-gen χːink'-e khinkal-pl d-irq'-id, npl-do.ipfv-1.prs wec'al ten duqu egg k-ert'-id, down-pour.ipfv-1.prs c'il then nejg milk k-ert'-id down-pour.ipfv-1.prs 'We make egg khinkal. We pour ten eggs; then we pour milk.'
	- (2) dam 1sg.dat qum.a.art-id forget.ipfv.neg-1.prs cik'al anything 'I don't forget anything.'
	- (3) u-l 2sg-erg b-arq'-ij n-do.pfv-inf w-irχ-utːe=w? m-be.able.ipfv-2sg.prs=q 'Can you do this?'
	- (4) [Tell your head of administration to wait one more day.] c'il then du 1sg hextːu-w there.up-m w-irχʷ-ud m-be.ipfv-1.prs hek'-i-la dem.up-obl-gen kabinet-le-w office-loc-m 'Then I will/could be there in his office.'
	- (5) di-la 1sg-gen šišːim-la suffering-gen dalaj song b-elč'-id=aw n-read.ipfv-1.prs=q a-b-elč'-id=aw? neg-n-read.ipfv-1.prs=q 'Should/will I sing my sad song or not?' (a more literal translation is: 'To sing or not to sing the song about my sufferings?')

The habitual/future polysemy is common for Dagestanian languages (Tatevosov 2005) and also cross-linguistically well-attested (Haspelmath 1998). The future reading has developed from the habitual reading, but it is only available for predicates that express transitory and accidental properties (4), (6). Predicates that denote temporally stable and essential properties that characterize their referents only express the habitual meaning (7), (8).

(6) hel that prosto, simply "dam 1sg.dat a-b-ikː-ar", neg-n-want.ipfv-3.prs ∅-ik'-ul m-say.ipfv-icvb ca-w, cop-m "ašːi-cːella 2pl-comit ka-jž-ij down-remain.pfv-inf w-elqː-un-ne=da" m-sate.pfv-pret-cvb=1 ∅-ik'-ul m-say.ipfv-icvb ca-w cop-m 'He simply says, "I do not want to sit with you, I had enough of you," he says.'

13.1 Habitual present

(7) ca one cik'al something a-b-alχ-ad neg-n-know.ipfv-1.prs 'I don't know anything.' (NOT: 'I will/should not know anything.')

(8) Madina-j Madina-dat du 1sg w-ičː-aq-id m-want.ipfv-caus-1.prs 'Madina loves me (masc.).' (NOT: 'Madina will/should love me.') (E)

As can be seen in Table 13.1, the third person has two suffixes, *-u* and *-ar*. The latter suffix is less frequently attested in the corpus. It is homophonous with the third person realis conditional suffix -*ar* (§18.3.1), and therefore not always easy to identify in texts. It seems that there is a slight semantic difference such that *-u* can refer to single events whereas *-ar* refers to habitually occurring events, but this difference is hard to detect and not always clear. For instance, (9) means 'remember from time to time, think of', whereas *han b-irk-u* would just mean 'remember (once)'. Similarly, *bek' icː-u* means 'the head aches (now)', whereas *bek' icː-ar* means that the head aches again and again, like when people have migraine. By contrast, example (10) shows an utterance, in which *w-irχʷ-ar* could be replaced by *w-irχ-u* without any change in meaning.


The verb *b-ik'ʷ-ij* 'say' is the most frequently used verb with *-ar*, and it is frequently but not always translated as past tense (i.e. 'said') without any habitual seamntics when it bears this suffix. I do not have an explanation for why the verb form (*r/b-/d-*)-*ik'ʷar* conveys non-habitual past time semantics. Example (11) illustrates the use of both suffixes -*u* and -*ar* with this verb in one sentence.

(11) "žaˁbraˁʔil-qal Zhabrail-assoc r-aš f-go a-b-ik'-u=w?" neg-n-say.ipfv-3.prs=q haʔ-ib=da. say.pfv-pret=1 "a-b-ik'-u", neg-n-say.ipfv-3.prs r-ik'ʷ-ar f-say.ipfv-3.prs 'I said, "Zhabrail and his family do not invite you?" She said, "They don't invite me."' (lit. 'Don't they say "Come!"')

Negation is expressed through the prefix *a-* (9). Some affective verbs allow for the ergative construction (in addition to the dative construction) with the habitual present (this phenomenon requires future research, but see §19.1.8 for some more examples).

13 Indicative synthetic verb forms

### **13.2 Habitual past**

The habitual past is the past-tense counterpart of the habitual present. It is only formed from the imperfective stem by means of a suffix *-a* that is followed by person markers (first and second person). The person markers for second person are identical to the person markers used for the habitual present such that *b-aχ-* (pfv)/*b-alχ-* (ipfv) 'know' has identical forms for the habitual present and past in the second person (compare Table 13.2 and Table 13.4). In the third person, *-a* is absent. Instead, the suffix *-i* or alternatively the longer variants *-iri* or, rarely, *-ini* are used (Table 13.3). As an alternative to *-a* plus person suffix, *-i(ri)* can also be used with first and second person without any difference in meaning.

Table 13.3: Person suffixes for the habitual past


Table 13.4: Some illustrative paradigms of the habitual past


The semantic domain is habitual situations with past time reference. The verb form is used to express habitually occurring actions in the past (12), employed in characterizing persons (13), when referring to occupations, and so on. The functional range of the habitual past also includes the expression of future-in-the-past in the protasis of past conditionals and irrealis conditionals (14). As with the habitual present, negation is expressed through the prefix *a-* (14) and some affective verbs additionally allow for the ergative construction with the habitual past (for more information see §19.1.8).


13.2 Habitual past

(14) di-la 1sg-gen aba-la mother-gen ucːi-l brother-erg du 1sg čaˁčaˁn-t-a-cːe Chechen-pl-obl-in a-r-uk-utːel neg-f-lead.ipfv-cond.pst r-ubk'-a-di f-die.ipfv-hab.pst-1 'If my mother's brother would not have brought me to Chechnya, I would have died.'

The verb *b-ik'ʷ-ij* 'say', which was mentioned in the previous section as expressing past time reference by means of the third person habitual present suffix for reasons that still await clarification is regularly inflected for the habitual past. However, the meaning is not always clearly habitual but seems also to be just a perfective past (15).

(15) a but presedatel head ča who ca-w=de cop-m=pst ∅-ik'ʷ-a-tːe? m-say.ipfv-hab.pst-2sg 'Who (masc.) did you (masc.) say was the head (of the kolkhoz)?'

The verb *b-ikː-* (ipfv) 'want, like, love', which lacks a perfective stem, shows exceptional behavior with the habitual forms. The only available forms of the habitual present are *dam b-ikː-i* 'I want' and *nišːij b-ikː-i* 'we want' and for questions *at b-ikː-i=w?* 'Do you (sg) want?' and *ašːij b-ikː-i=w?* 'Do you (pl) want?'. There are no forms for third person and the second person forms cannot be used in assertions. Furthermore, the habitual past expresses irealis modality with the first person, that is, *dam*/*nišːij b-ikː-a-di* translates as 'I/we would like, I/we would want'. It is not used with other persons apart from the first person.

## **14 Analytic verb forms**

All verb forms consisting of a lexical verb bearing a participial or converbal suffix (and possible other suffixes) followed by a person enclitic, the past enclitic, the copula *ca-b*, or the suffix -*ne* are called "analytic verb forms" and described in this chapter. When the standard copula is replaced by locational copulas or other auxiliaries, the resulting verb forms will be called "periphrastic", and they are separately treated in Chapter 15. The division between analytic and periphrastic verb forms is mainly based on differences in morphology, semantics, and frequency of use. Among the morphologically complex verb forms, analytic verb forms are the core verb forms because they are basic in terms of the semantics and pragmatics of the inflectional element that accompanies the lexical verb. This element (person enclitic, past enclitic, standard copula, suffix -*ne*) expresses basic verbal categories such as tense, person, number, and gender.<sup>1</sup> The lexical verb conveys aspectual and modal meaning. By contrast, in periphrastic verb forms the accompanying auxiliary has additional modal, locational, evidential or aspectual meanings that contribute to the meaning of the complex predicate, which is therefore more specific. Furthermore, the accompanying auxiliary verbs of periphrastic verb forms are also used as full lexical verbs, but not as semantically empty copulas in copula clauses. The latter use is only attested for person enclitics, the past enclitic and the standard copula. Because of their more general meaning most analytic verb forms occur far more frequently in texts than the periphrastic verb forms with their more specific meaning.

The analytic verb forms can be divided into two main groups: forms based on the imperfective stem (§14.1) and forms based on the preterite (§14.2). The former convey mainly present time or future time reference (and an imperfective past), whereas the latter almost exclusively convey past time reference.

### **14.1 Forms based on the imperfective stem**

The TAM forms that can be obtained from the imperfective stem can be divided into two groups, depending on whether the lexical verb bears the imperfective converb suffix or the modal participle *-an* (Table 14.1). The second group has a modal meaning due to the semantics of the participle. All forms make use of person enclitics/copula *ca-b* for present or future time reference and the past enclitic *=de* for past time reference. The following subsections treat all analytic verb forms based on the imperfective stem according to the order in the table.

<sup>1</sup>Agreement rules and agreement exponents, i.e., gender affixes, person suffixes and person enclitics, are separately treated in Chapter 20 and therefore not discussed in this chapter.

### 14 Analytic verb forms

Table 14.1: Analytic verb forms based on the imperfective stem


### **14.1.1 Compound present**

The compound present is obtained by adding the imperfective converb *-ul*/*-un(ne)* to the verbal stem,<sup>2</sup> which is in turn followed by the person enclitics (first and second person) or by the copula *ca-b* (third person).

Table 14.2: Some exemplary paradigms of the compound present


The compound present is the default tense for conveying present time reference. It covers various imperfective meanings such as progressive, habitual, or continuative.

	- (1) hana now du-l 1sg-erg b-urs-ul=da n-tell-icvb=1 χabar story 'Now I am telling a story.'

<sup>2</sup>The imperfective converb is, at least diachronically, related to the cross-categorical adverbializer -*le* (§9.6.3), and thus also to the perfective converb. However, in order to facilitate understanding I treat the converbs and the adverbializer as separate items.

14.1 Forms based on the imperfective stem

	- (3) cara-te other-dd.pl daˁʡaˁn secret b-irxː-ul hpl-put.ipfv-icvb ca-b cop-hpl 'The others hide.' (in a game of hide-and-seek)
	- (4) ʡuˁrus Russian ʁaj-la language-gen ce=jal what=indq b-ik'-ul hpl-say.ipfv-icvb ca-b be-hpl it-i-j, that-obl-dat dam 1sg.dat qum.urt-ul forget.ipfv-icvb ca-b cop-n 'What do they call it in Russian, I forget it.'
	- (5) šːi-l-cːe-w village-obl-in-m er look ∅-ik'-ul m-look.at.ipfv-icvb ca-w, cop-m ča-k'al who-indef caʔarrah not.one admi person w-akːu m-cop.neg 'He is looking around in the village, nobody is there.'

Negation can be expressed either through the negative prefix *a-* or by means of the negative auxiliary. In the former case, which represents the rarer variant, the negation suffix is simply added to the lexical verb (6). In the latter case, which is far more common, the negative copula *akːʷa-* 'be not' (cop.neg) is used (7); it is inflected for person, but not for gender (see §16.1 for the paradigm of the negative copula).


In questions with a third-person agreement controller, the copula is replaced by the respective interrogative enclitic (8). In such contexts, the interrogative enclitic acts as a predicative particle and takes over the role of the copula (§9.1).

14 Analytic verb forms

(8) hež-i-l this-obl-erg ce what b-irq'-ul=e? n-do.ipfv-icvb=q 'What is he doing?'

The compound present can also be formed by means of existential copulas instead of the normal copula, which leads to a slight change in the meaning (§16.2).

### **14.1.2 Compound past**

The compound past is formed by encliticizing the past marker *=de* to the imperfective stem that bears the imperfective converb suffix. Its semantics corresponds to the semantics of the compound present, but now we have past time reference.

	- (9) dubur-t-a-cːe-b mountain-pl-obl-in-n ca one admi-l person-erg quˁr-be pear-pl luc'-unne=de gather.ipfv-icvb=pst 'In the mountains there was a man gathering pears.'
	- (10) χalq'-li-j people-obl-dat kumek help b-irq'-ul=de n-do.ipfv-icvb=pst 'He helped the people.' (character trait of the grandfather)
	- (11) [description of a game] urcul-la wood-gen hež-itːe this-advz ka-b-irxː-ul=de down-n-put.ipfv-icvb=pst ca one krug circle b-irq'-ul=de n-do.ipfv-icvb=pst '(We) put wooden (sticks?) like this, we made a circle.'
	- (12) b-ucː-ul n-work-icvb b-el=de, n-remain.pfv=pst luk'-unne=de write.ipfv-icvb=pst '(In contrast to now, at that time my hand) worked, I wrote.'

For the negation there are again two options: prefixation of *a-* (13) and use of the negative copula inflected for the past tense, with the latter option being more frequent (14).

(13) c'il then qːačuʁ-e bandit-pl χːuˁrba-cːe graveyard-in b-uˁq'-ij hpl-go-inf a-b-irχ-ul=de=w? neg-hpl-be.able.ipfv-icvb=pst=q ' Couldn't the bandits go to the graveyard?'

14.1 Forms based on the imperfective stem

(14) hak' shake b-ulq-unne n-direct.ipfv-icvb akːʷ-i cop.neg-hab.pst mašina car 'The car did not shake.'

### **14.1.3 Future**

The future is formed by adding the person enclitics to the lexical verb that bears the participle *-an*. In the third person, the suffix *-ne* is used (Table 14.3).



Its semantic range is:

	- (15) aʁʷc'alla forty.days d-arq'-ij 1/2.pl-do.pfv-inf d-irχ-an=da 1/2.pl-be.able.ipfv-oblg=2pl ušːa-l 2pl-erg 'You (pl.) will be able to spend the 40 days (without me).'
	- (16) hana now u=ra 2sg=add ∅-ukː-an=de m-eat.ipfv-ptcp=2sg 'Now (she) will eat you (masc.), too.'
	- (17) hel-tːi that-pl.abs ħaˁjwan-qːačːa-la animal-calf-gen akːʷ-ar, cop.neg-prs cara other ce what ʡaˁči work b-irχʷ-an-ne n-be.ipfv-ptcp-fut.3 hextːu-b there.up-n b-i-b? n-in-n 'If there was not the work with the animals, what work would/should there be? '
	- (18) cara other ce what ∅-ik'ʷ-an-ne? m-say.ipfv-ptcp-fut.3 'What else should be said?'

14 Analytic verb forms

> (19) heχ dem.down u-l 2sg-erg dirxːa stick gu-r-b-uqː-ij=sat, sub-abl-n-take.out.ipfv-inf=as.much k-ercː-an=de down-stand.ipfv-ptcp=2sg heštːu here 'Until you take the stick out, you have to stand here.'

Negation is expressed by means of the prefix *a-*:

(20) a-d-irc-an=da neg-npl-sell.ipfv-ptcp=1 haʔ-ib=da say.pfv-pret=1 'I will not sell them, I said.'

### **14.1.4 Future in the past**

The future in the past is formed by adding the past enclitic *=de* to the participle *-an*. It expresses irrealis modality, referring to situations and actions that should have taken place or performed in the past (21–23). It is also used in the counterfactual apodosis of irrealis conditional clauses (24) (§25.2). The negative prefix is used for negation (23).


(23) žan-ni-cːe-r body-obl-in-abl a-r-ulq-an=de, neg-f-direct.ipfv-ptcp=pst ka-d-icː-ar=ra down-npl-stand.pfv-cond.3=add q'ijama end.of.world

'You (fem.) should not have left your body, even if the end of the world comes.'

(24) du-l 1sg-erg kiniga book b-uč'-an=de, n-read.ipfv-ptcp=pst raχle if či-d-ig-ul spr-npl-see.ipfv-icvb ∅-iχ-utːel m-be.pfv-cond.pst

'I would have read the book if I (masc.) had seen it.' (E)

### **14.1.5 Obligative**

The obligative is formally and functionally closely related to the future, but it makes use of the copula for all third persons instead of person enclitics. The meaning is usually modal referring to needs and obligations, close to deontic necessity. For negation the copula *ca-b* is replaced by the negative copula *akːu* (27), (28).

14.1 Forms based on the imperfective stem


'The elder uncle has to go; the younger aunt should not go.'

### **14.1.6 Obligative present**

The obligative present strongly resembles the future and the obligative. With both forms it shares the meaning, this means that, the obligative present expresses future and/or obligation. The only formal difference is the additional use of the cross-categorical suffixe -*ce* (plural -*te*), which is added to the participle before the person marker is encliticized.


Note that in (30), although the singular form *-an-ce* is available in elicitation, it is not attested in the corpus and it seems that for the singular the future (§14.1.3) (or the obligative, §14.1.5) is preferred. All corpus examples contain the plural suffix *-te*. The reason for this might be the general distribution of the cross-categorical suffixes -*ce* / *te* and -*il*. For plural referents only -*te* is available (29). In the singular, in principle -*ce* and -*il* compete, but in natural texts the use of -*il* is clearly preferred and only a few examples of -*ce* can be found in the corpus (see §9.6.1 and §9.6.2 for detailed analyses of the suffixes). The obligative present (and the obligative past) cannot be formed by means of the suffix -*il* for reasons that are not clear to me. Therefore, the more natural way of forming a singular from a (29) is to use the future form *du w-ax-an=da* instead of (30).

In the third person the copula is used (31). However, the copula can also be employed with first and second person, in which case the meaning of obligation is dominant (32), (33). Negation is expressed by means of the negative copula (34).

14 Analytic verb forms


### **14.1.7 Obligative past**

The obligative past is formed by replacing the person enclitic or copula of the obligative present with the past enclitic. It refers to obligations that obtained in the past and that were or were not fulfilled (35–37).


The obligative present and past forms can also have a non-modal and non-future reading when they are instead interpreted like headless relative clauses and the person enclitic, copula or past marker makes up its own copula clause (38). Thus, in the first part of this sentence the participle has been nominalized by means of the cross-categorical suffix -*te*, which corresponds to a headless relative clause ('the drinking ones'). This nominalized clause functions as subject in an an existential copula in which the encliticized past marker =*de* serves as an existential copula. The nominalized clause does not have modal or future semantics. The second part has a similar meaning, but the copula is missing such that we have only the nominalized clause, which is more complex. It also

14.2 Forms based on the preterite

contains a demonstrative pronoun and an adjunct in the ergative that serves as direct object because the nominalized clause is an antipassive construction (§19.2.1).

(38) itːu-b there-hpl b-učː-an-te=de; hpl-drink.ipfv-ptcp-dd.pl=pst iš-tːi this-pl ʡaˁči-l work-erg b-irq'-an-te hpl-do.ipfv-ptcp-dd.pl ... 'There were the drinking ones; these working ones ...'

### **14.2 Forms based on the preterite**

The preterite is the most important verbal suffix in Sanzhi not just because it is extremely common in terms of token frequency and used as the base for a wide range of TAM forms (Table 14.4), but also because it is the major indicator for verbal inflection classes.

Sanzhi has the standard Dargwa inventory of preterite suffixes: *-ib, -ub, -un,* and *-ur*. The suffix *-ib* is the most frequently used preterite suffix and thus sometimes treated as the default variant (e.g. Daniel 2015), and *-ub* is analyzed as a phonologically predictable allomorph that is in complementary distribution with *-ib* and occurs only with labialized stems (Belyaev In Preparation). The latter is, in fact, the case in Sanzhi, but the complementary distribution still needs further investigation (see §11.2 for lists of verbs and their preterite suffixes). The suffix *-un* is the second most frequent suffix after *-ib*. It occurs with verbs that have *l* in the perfective stem and/or an imperfective stem with initial *l*, labialized root consonants, or *l* in the imperfective stem. The suffix *-ur* is the least frequent one (though probably more common than *-ub*), and its occurrence cannot be predicted.

In principle, many verbs can inflect the imperfective as well as the perfective stem for the preterite, but not all verbs have this possibility (e.g. *b-alχ-* 'n-know.ipfv' cannot take any of the preterite suffixes). There are only very few corpus examples of imperfective verb stems bearing the preterite suffix, all occurring in the preterite. All other forms can be elicited, but speakers do not seem to have clear intuitions about the meanings and context of use of these forms and translations suggest that the forms are not truly part of verbal paradigms. For instance, experiential forms are normally translated with cleft constructions, suggesting a biclausal structure. Therefore, almost all verb forms based on the imperfective stem have been given in parenthesis in Table 14.4 and Table 14.5. They will not be further discussed here.

Table 14.4 displays the verb forms that can be obtained from the preterite and Table 14.5 provides one exemplary verb. The paradigm shows a clear symmetrical structure. All verb forms make use of one of the three available means for obtaining regular analytic verb forms: person enclitics, the copula *ca-b* or the past enclitic. The preterite is the base form and the preterite resultative is a kind of minor variant derived from it. From the preterite, three types of verb forms are built. Each of them follows the same pattern: the lexical verb occurs in the preterite form and takes one of three further suffixes. These suffixes are the perfective converb suffix, or one of the two cross-categorical

### 14 Analytic verb forms

suffixes *-ce* and *-il*. The suffixes are followed by person enclitics (first and second person) or the copula (third person), which yields one type of forms. The temporal reference of this type of verb forms can be further shifted to the past by means of the past enclitic =*de*.

At least diachronically all three suffixes used after the preterite belong to the same class of cross-categorical suffixes that are added to words of different lexical classes and form either referential attributes with the syntactic properties of nouns (*-ce* and *-il*) or adverbials (§9.6.3). Thus, the perfective converb suffix is identical with the adverbializer -*le*. But for the sake of clarity and readability of the grammar I will gloss it as perfective converb suffix and treat it as a separate item (the same was done for the imperfective converb). The syntactic properties of the cross-categorical suffixes are preserved in the verb forms containing them: the perfective converb is functionally equivalent to adverbials when it is used without the person enclitics, past enclitic or copula; the verb forms with the suffixes -*ce* and -*il* (experiential and experiential past forms) are functionally equivalent to nominalized participles that form relative clauses.


Table 14.4: Forms based on the preterite

### 14.2 Forms based on the preterite


Table 14.5: Exemplary paradigms based on the preterite for the verb 'do, make'

### **14.2.1 The imperfective preterite and imperfective preterite resultative**

The imperfective preterite is formed from the preterite stem of imperfective verbs to which person enclitics for first and second persons are added; for the third person no markers are used. In addition to the ergative and the affective construction, the imperfect also allows for the antipassive construction (41). The imperfective preterite expresses past time reference in combination with imperfective aspect. It can be negated by means of the prefix *a-*. The imperfective preterite is barely attested in the Sanzhi corpus (42); examples (39–41) have been elicited. The imperfective preterite resultative, which is restricted to third person agreement controllers, has been obtained only through elicitation (43).


14 Analytic verb forms


Examples (44) and (45) are from the corpus and show antipassive constructions. In (44), the demoted agent is expressed (clause-final pronoun). The demoted patients, which would have been in the ergative case, are left unexpressed in both examples. See §19.2.1 for antipassive constructions.


### **14.2.2 The preterite**

The preterite is the default past tense with respect to form and function. It is formed from the perfective stem by adding the preterite suffix and for first and second persons the person enclitics; the third person does not have additional marking. It conveys past time reference and is very frequent in the Sanzhi corpus, especially in autobiographical narratives (46) and in daily conversations when speakers report about past events (49). However, it can also occur in traditional narratives (47), (48) and in other narratives about the past that are not related to the personal experience of the speaker (50).

(46) ʡaˁbal three dus year kelg-un=da, remain.pfv-pret=1 du 1sg gaupaχt-le guardhouse-loc a-ka-jč-ib=da neg-down-occur.pfv-pret=1 ca one sːaˁʡaˁt hour 'Three years I remained, I was not one single hour at the guardhouse.' (47) du-l kumek b-arq'-ib=da, tːura-h-aqː-ib=da

1sg-erg help n-do.pfv-pret=1 out-up-take.out.pfv-pret=1 'I helped. I pulled him out.'

14.2 Forms based on the preterite

(48) ca one zamana time bari=ra sun=add wahi-ce evil-dd č'an=ra wind=add čːal argument d-uq-un npl-go.pfv-pret 'Once upon a time the sun and the evil wind argued.'

Negation is expressed through the prefix *a-*:


### **14.2.3 The (perfective) resultative**

The perfective resultative consists of the preterite and the copula. This verb form cannot be used with the first or second person agreement controllers. The presence of the copula conveys perfectivity/resultativity, i.e. the focus is on the result of a situation (51–53). This form is usually not used in personal narratives, but it is very frequent in other texts such as traditional narratives and other third-person perspective narrations. Negation is expressed through the prefix *a-* (53).


' It was at night, there was no fire visible, nobody is there, then he turned and came back.'

Sentence (54) is a corpus example with a first person pronoun in the dative. The predicate in this example is an affective verb which requires a dative experiencer and an absolutive stimulus. In contrast to almost all other bivalent affective verbs the experiencer does not obligatorily control person agreement on the predicate, but the predicate can be used with the copula (i.e. third person). See §19.1.8 for more information

(54) qum.ert-ur forget.pfv-pret ca-d cop-npl na now hetːi those mus-ne place-pl dam 1sg.dat 'Now I have forgotten those places.'

### 14 Analytic verb forms

The focus that the perfective resultative puts on the resulting state can lead to an inferential interpretation that becomes particularly obvious to speakers when they are asked to compare the preterite to the perfective resultative. For example, the following sentence could be uttered in a situation in which Sanzhiat must wash the dishes, she goes to the kitchen and sees that somebody has already washed the dishes (55). This means that she concludes from the result that someone must have washed them.

(55) uže already t'alaˁħ-ne dishes-pl d-irc-ib npl-wash.pfv-pret ca-d cop-npl 'The dishes have already been washed.' (E)

If she then asks *hil dircibe?* 'Who washed (them)?', an appropriate answer of somebody who attended the event could be (56), that is, now the agent is at stake, not the result of the action.

(56) Sanžijat-li Sanzhiat-erg d-irc-ib npl-wash.pfv-pret 'Sanzhiat washed (them).' (E)

Similarly, when looking out of the window the speaker sees a wet road and concludes from this (57).

(57) marka-l rain-erg b-us-ib n-rain-pret ca-b cop-n 'It has rained.' (E)

However, the inferential interpretation can be canceled by a following utterance without leading to a special interpretation (58).

(58) Sanijat-li Sanijat-erg t'alaˁħ-ne dishes-pl d-irc-ib npl-wash.pfv-pret ca-d. cop-npl dam=q'ar 1sg.dat=prt il that či-b-až-ib=da spr-n-see.pfv-pret=1 'Sanijat has washed the dishes. I saw it myself.' (E)

### **14.2.4 The perfect**

The perfect is formed by adding the perfective converb suffix to the preterite, followed by the person enclitics for first and second person and the copula *ca-b* for third person. When the perfective converb is suffixed to the preterite regular assimilation processes take place after the suffixes that end in a sonorant, such that the following allopmorphs result: *-ib-le, -ub-le, -ur-le/-ur-re, -un-ne*. Though it can be elicited with imperfective stems, there are no such instances in the corpus and I will therefore restrict myself to the discussion of perfect forms built with perfective stems.

14.2 Forms based on the preterite

The perfect is not particularly frequent in narratives, but there are enough examples to describe its meaning. Its semantic range primarily covers resulting states; it mostly occurs with verbs such as 'sit', 'lay down', 'die', 'get/become hungry', etc. that denote a change of state and the perfect expresses the resulting state:


This includes transitive verbs of which the agent is then often omitted because the focus is on the resulting state (62).

(62) ik'-i-la dem.up-obl-gen bek' head b-aˁq-ib-le n-wound.pfv-pret-cvb ca-b cop-n hek' dem.up 'Her head has been wounded.'

The following example illustrates one of the traditional greetings for women, used by men and women when the female addressee is seated, for example in front of the house, and the speaker is passing by (63). Example (64) shows a minimal pair illustrating the difference between the preterite and the perfect that formally differ only in the absence vs. presence of the perfective converb. The preterite conveys past time reference with verbs that express changes of state whereas the perfect refers to the state that obtains at the present moment. (64b) is the standard answer to (63).

(63) ka-r-iž-ib-le=de=w? down-f-be.pfv-pret-cvb=2sg=q 'Are you sitting (seated)?'

```
(64) a. ka-r-iž-ib=da
          down-f-be.pfv-pret=1
         'I sat down.' (E)
      b. ka-r-iž-ib-le=da
```

```
down-f-be.pfv-pret-cvb=1
'I am sitting.'
```
As can be seen in (65), the agent can be overtly expressed and the ergative construction is allowed when the perfect is used in Sanzhi, in contrast to the closely related Icari Dargwa variety, which prohibits the perfect with overtly expressed agents inflected for the ergative case.

### 14 Analytic verb forms

(65) nu well hel that priblizitelno approximately nušːa-l 1pl-erg b-urs-ib-le=da=q'al n-tell-pret-cvb=1=mod 'Well, approximately we already said it.'

And in contrast to other Dargwa varieties such as Shiri (Belyaev In Preparation), the Sanzhi perfect can also be used with verbs that do not imply a change of state in the agent (66) even though normally the preterite is preferred in such contexts.

(66) ag-ur-re go.pfv-pret-cvb ca-b cop-hpl sːika-la bear-gen mergʷ-li-šːu lair-obl-ad ' They have gone to the cave of the bear.'

In the right context, the perfect can imply inferentiality/indirect evidentiality similar to the perfective resultative. Example (67) and originates from a narrative about the history of the village of Sanzhi, and the speaker draws a conclusion about the present situation of the village based on past events that he did not witness himself. Similarly, (68) and (69) are inferences about past events that the speakers draw from observed results.

(67) heχ dem.down šːi village imc'a additional b-iχʷ-ij n-be.pfv-inf b-at-ur-re=kːu n-let.pfv-pret-cvb=neg hel-tː-a-li that-pl-obl-erg '(They) have not allowed the village to grow.' (68) sa-jʁ-ib-le hither-come.pfv.m-pret-cvb Gudermec-le, Gudermets-loc ix-tːi dem.up-pl tːura outside

aʁ-ib-il do.pfv-pret-ref b-aχ-ur-re=kːu n-know.pfv-pret-cvb=neg

'When he came to Gudermets he did not know that they had been thrown out (of the village).'

(69) berkʷijce-li-j=ra food-obl-dat=add qːuc touch ∅-ič-ib-le=kːu m-occur.pfv-pret-cvb=neg b-erk-un-ne=kːu n-eat.pfv-pret-cvb=neg ' (He also has not touched the food, he has not eaten.'

Negation of verb forms with first and second person agreement controllers is expressed by means of the prefix *a-*, but there are no corpus examples. For the third person the negative copula *akːu* occurred in its shortened form as an enclitic to the verb (67–69).

### **14.2.5 The past perfect (pluperfect)**

The past perfect is formed by attaching the past enclitic *=de* to the perfective converb. In elicitation, the past perfect is available for perfective and imperfective stems, but there are no corpus examples of the latter. In addition, there is a variant of the past perfect that makes use of the locative copulas to which *=de* is encliticized (see §15.1).

14.2 Forms based on the preterite

The past perfect has the typical pluperfect meaning and also past resultative meaning. It refers to an event (or the resultative state of an event) that occurred before a definite point in past time. In (70) the preceding event is mentioned in the first clause of the utterance. In the other examples (71) and (72) the reference point in the past was mentioned in the preceding context.


Negation is expressed by means of the negative prefix *a-* (73) or the negative past tense copula *akːʷi* (74).


The past perfect also expresses inferentiality. This means that the speaker concludes from an observed result that an event has taken place. Thus, (75) was uttered in a situation when the speaker found out only afterwards when reading the article that the journalist to whom he had talked had written a wrong name. Example (76) is from a narrative about past events that were not witnessed by the speaker himself (namely the grabbing that happened at night). But he inferred from the result and from his knowledge of the general circumstances that the people he is talking about in (76) were the robbers.

(75) Tawlu Tawlu žaˁndaruwič Zhandaruvich b-elk'-un-ne=de n-write.pfv-pret-cvb=pst '(He) had (apparently) written Tawlu Zhandaruvich.'

### 14 Analytic verb forms

(76) šara S. ag-ur go.pfv-pret itːa-la those.obl-gen qul-be house-pl qaˁm grab d-arq'-ib-le=de npl-do.pfv-pret-cvb=pst ' They went to Shara and had grabbed their houses.'

The past resultative meaning can co-occur with the inferential meaning. For instance, in (77) the speaker refers to a state (= the death of his mother) that was obtained before another moment in the past (= his return to the village). At the same time the speaker was not present at the relevant event (= the dying of his mother) such that there is an inferential component.

(77) du 1sg sa-jʁ-ij=satːina hither-come.pfv-inf=until r-ebč'-ib-le=de f-die.pfv-pret-cvb=pst aba mother 'Before I came (home) my mother had already died.'

When speakers are presented with past perfect sentences out of context that contain predicates that do not denote a change of state the inferential meaning is salient and therefore there is a first-person effect with core arguments that denote first persons. This means that (78) can only be uttered if the referent of the first person pronoun did not consciously participate in the situation and therefore did really see Arsen because he did not recognize him.

(78) dam 1sg.dat Arsen Arsen či-w-až-ib-le=de spr-m-see.pfv-pret-cvb=pst

'I (apparently) saw Arsen.' (as it seems, e.g. I did not recognize him) (E)

### **14.2.6 Experiential I and experiential II**

There are two variants of the experiential. They both involve the preterite to which the cross-categorical suffixes are added (-*ce* and -*il*). The cross-categorical suffixes are generally used to form referential attributes from various parts of speech, including verbs (§9.6.1 and §9.6.2). The resulting word forms largely have the syntactic properties of nominals (e.g. they can be inflected for case, they can take over argument positions, etc.), and this leads to very particular syntactic properties of all experiental and experiental past forms that are discussed below.

The first variant, the experiential I, is obtained by suffixing *-ce* (plural *-te*) to the preterite participle, followed by the person enclitics or the copula *ca-b*. The second variant, the experiential II, is formed by adding the suffix *-il* to the preterite, again followed by the person enclitics or the copula. The use of the two different suffixes -*ce* and -*il* does not lead to any semantic differences with respect to the experiential verb forms. Their distribution rather depends on number (this was already explained for the obligative verb forms in §14.1.5). For argument controllers in the singular -*il* is almost exclusively used, although -*ce* is also grammatical. For argument controllers in the plural only -*te* is allowed. The experiential can also be formed with the locative copulas (see §15.1 for an example).

14.2 Forms based on the preterite

The experiential I and II have perfect-like semantics, but are predominantly used when speakers talk about their own experiences and about situations they were personally involved in, so most of the examples contain first person core arguments:


'Then (Abdulkhalik) says, "Have you come here for condolences or for singing songs?"'

(81) w-arq'-ib-il=da m-do.pfv-pret-ref=1 du 1sg azir-lim thousand-num urč'em nine darš-lim hundred-num ʡaˁb-c'anu three-ten xu-ra-ibil five-num-ord

' I (masc.) was born in 1935.'

Somewhat more rarely one finds third person examples that, however, usually relate to the personal sphere of the speaker or, more generally, to the sphere of the Sanzhi people (82), (83). For instance, (82) is from a procedural text in which the speaker explained how Sanzhi women used to make carpets. There are only few examples that are not immediately related to personal experience, mostly occurring in texts from the *Family Problems Picture Task* (San Roque et al. 2012) (84).

(82) a and tak so nuˁq-b-a-cːella hand-pl-obl-comit hel-tːi=ra that-pl=add d-arq'-ib-te npl-do.pfv-pret-dd.pl ca-d cop-npl 'And like this with the hands (they) also made them.' (83) sːema-la pebble.stone-gen χaˁχaˁ, Xaxa nišːa-la 1pl-gen atːa=ra father=add k-ag-ur-il down-go.pfv-pret-ref ca-w cop-m heχtːu-w there.down-m ' Semalla Xaxa (place name). Our father also fell down there.' (84) it that tusnaq-le-r prison-loc-abl sa-jʁ-ib-il hither-come.pfv-pret-ref ca-w cop-m

' He came back from prison.'

### 14 Analytic verb forms

From a morphosyntactic point of view, the experiential and the experiential past are somewhere between a monoclausal and a biclausal structure, which is due to the impact of the cross-categorical suffixes, because the suffixes form words with largely nominal morphosyntactic features. This means that clauses with experiential verb forms resemble clefts with a main copula clause that contains only the person enclitics or the copula and a subordinate relative clause. Thus, instead of person agreement enclitics one finds the copula despite a first or second person agent. For example, the first person agent in (85) is not expressed, but clear from the context of the autobiographical narrative. In the elicited example (86a), the use of a person marker instead of the copula is impossible (86b).

	- b. \* dul 1sg.erg itːi those qːamuš reed ka-d-ičː-ib-te=da down-npl-cut.up.pfv-pret-dd.pl=1 (Intended meaning: 'I cut that reed.') (E)

This suggests that the structure of (86a) is as displayed in (87). In fact, when translating experiential clauses speakers sometimes produce relative clauses in the Russian translation. Thus, a more literary translation that is closer to the structure of (86a) would be 'It is such that the reed was cut by me.'

(87) [itːi those qːamuš reed dul 1sg.erg ka-d-ičː-ib-te] down-npl-cut.up.pfv-pret-dd.pl ca-d cop-npl 'I cut that reed.' (E)

The almost biclausal structure becomes especially salient in term focus constructions when the person enclitic or the copula is not following the verbal complex but an argument or adjunct that is focused (88). In this context, the use of the person marker is allowed, but optional (89). Thus, in the last example we can either employ the person enclitic after the pronoun or the copula, but not both.


14.2 Forms based on the preterite

The biclausal-like structure is also apparent in negation because here always the negative copula *akːu* is used and person agreement is suppressed (90–92). A detailed account of the syntactic structure (i.e. whether it is monoclausal or biclausal or should be analyzed as something else) must be left to future research.


' A cleanliness like theirs I have seen nowhere.'

### **14.2.7 Experiential past I and experiential past II**

Corresponding to the experiential I and II, there are also two variants of the experiential past in which the past enclitic =*de* is used instead of the person enclitics/copula. The lexical verbs appear in the same forms as in the experiential I and II. The experiential past forms are normally used for the narration of personal experiences or of situations that lie within the personal knowledge sphere of the speaker even if s/he did not personally attend it:


These tense forms are often employed in summary-like utterances that do not move forward the main storyline (95) or when providing for background information (96), (97).

(95) hel=ʁuna this=eq cik'al something čujna=ra how.often=add d-arq'-ib-te=de 1/2.pl-do.pfv-pret-dd.pl=pst nušːa-l, 1pl-erg čujna=ra how.often=add

'Things like this, how often did we do them, how often.'

### 14 Analytic verb forms


In negated clauses the negative past copula *akːʷi* is used (98). It can be shortened to the enclitic *=kːʷi* (99) or inflected for person (100). The latter is insofar remarkable as the negative predicate in this case expresses more verbal categories than the affirmative, since person cannot be marked on the predicate in the affirmative because the past enclitic does not encode person. For example, in (100) the person suffix on the copula expresses the first person. By contrast, in affirmative clauses with the same verb form person cannot be expressed (93), (95).


## **15 Periphrastic verb forms**

Periphrastic verb forms are morphologically complex in the same manner as analytic verb forms (Chapter 14) and make use of the same range of non-finite inflectional forms (perfective and imperfective converb, and occasionally participles), but employ different auxiliaries that have, by themselves, particular semantic values. Therefore, the resulting verb forms differ in their meaning from the analytic verb forms. The auxiliaries employed are:


Some of the resulting verb forms have similar meanings although the auxiliaries differ. However, I will take a form-to-function approach and treat all formally distinct combinations of lexical verbs and auxiliaries separately.

There is a very large number of morphologically complex verb forms that can, in theory, be produced and can thus be obtained in elicitation, because various auxiliaries can be employed and partially combined. But since it is impossible to gain an understanding of verb forms if one has only one or two elicited examples, I restrict myself to the examination of commonly attested periphrastic forms and describe the meaning of these forms based on their occurrences in natural texts.

The auxiliaries are inflected according to their morphological possibilities (i.e. the existential copulas have reduced paradigms, see §15.1). Consequently, the auxiliaries can themselves be inflected for verb forms heading subordinate clauses, that is, there are also periphrastic verb forms that occur in subordinate clauses.

### **15.1 Verb forms with locational copulas**

Sanzhi has four locational copulas that are morphologically defective in a way similar to the standard copula. They have locational and existential meaning that includes elevation (§16.2). The use of locational copulas (instead of the standard copula) for the

### 15 Periphrastic verb forms

formation of periphrastic verb forms is not extremely frequent, but it is repeatedly attested. The most widely used locational copula verb is *le-b* 'be located/exist close to the speaker and the hearer', because its semantics is somewhat less specific in comparison to the other three locational copula verbs, and because of its meaning of proximity. The other three locational copulas are *te-b* 'be located/exist away from the speaker', *k'e-b* 'be located/exist above the deictic center', and *χe-b* 'be located/exist below the deictic center'. The semantics of the locational copulas partially determine the meaning of the periphrastic verb forms. For example, the use of *le-w* in (1) implies that the situation took place close to the speaker, and that the speaker consequently saw the event with her own eyes. If *ca-w* had been used instead, then there would be no such implication. Similarly, if *te-b* is used, the situation takes place or took place (far) away from the speaker who did not participate and did not witness the event himself/herself. For instance, the utterance in (2) comes from a report about a woman who was in a hospital in Makhachkala and whom the speaker did not visit there.


'Then, badly like this, she is not able to move.'

The copula *χe-b* refers to events occurring in an area lower than the deictic center, which is often the speaker or a default reference point (3). This example and also (4) were produced during the *Family Problems Picture Task* (San Roque et al. 2012) and the deictic center for (3) is not the speaker who uttered this sentence (his location is irrelevant) but the people on the picture.

(3) ka-d-ič-ib-le=q'ar down-npl-occur.pf=pret-cvb=mod χe-d exist.down-npl heχtːu-d there.down-pl šuš-ne bottle-pl 'There bottles have fallen down.'

The copula *k'e-b* refers to events occurring in an area higher than the deictic center (4). Example (4) is a description of pictures arranged on a table in front of the speaker that were put higher than some other pictures on the same table.

(4) c'il then hek'-tːi, dem.up-pl heχtːu-d there.down-npl sːurrat-le-d picture-loc-npl či-d-iž-aq-ul spr-npl-see.ipfv-caus-icvb k'e-d exist.up-npl maˁjk'a=ra T-shirt=add koftːa=ra jacket=add tːapri=ra shoe=add cin-na refl.sg-gen 'Then those, there on the picture, it shows that there are his T-shirt, jacket, and shoes.'

15.1 Verb forms with locational copulas

If the lexical verb takes the imperfective converb suffix, the resulting verb form corresponds to the compound present (§14.1.1) and has a comparable semantic range covering progressive/continuative (5) and habitual (2).

(5) ca one zamana time b-erčː-ib-le n-drink.pfv-pret-cvb saˁ-q'-uˁnne hither-go-icvb le-w exist-m hel that 'One time he is coming home drunk.'

When the perfective converb (i.e. preterite plus suffix -*le*) is employed, the perfect or other forms are obtained. As the normal perfect (§14.2.4), the perfect with locational copulas mostly expresses states that obtain after a preceding event (3), (6–8).


The locational copulas can be followed by the past enclitic =*de*, so that we get a variant of the past perfect or pluperfect (§14.2.5), usually referring to states that obtained in the past as the result of preceding situations (9), but also occasionally in reference to actions and events that happened before a reference point in the past. Thus, example (10) was uttered when the speaker compared the life of the family before and after an important event that served as a temporal anchoring point in the past.


Another possible periphrastic verb form corresponds to the experiential II (§14.2.6) for which the lexical verb takes the suffixes of the preterite plus the cross-categorical suffix *-il*:

(11) nu well hej=ʁuna this=eq ka-jž-ib-il down-remain.m.pfv-pret-ref te-w exist.away-m ∅-urkː-ar m-find.ipfv-cond.3 het that šːal-le-w=ra side-loc-m=add ' Well, he is like sitting, probably, at the side (of the road).'

### 15 Periphrastic verb forms

As can be seen from the examples in this section, many of the lexical verbs that are used in periphrastic verb forms with locational copulas are position verbs or verbs of movement (3), (5), (6), but other verbs are also allowed (4), (9), (10).

Finally, the use of *le-b* and other locational copulas in periphrasis is more common in other Dargwa varieties such as Mehweb (Daniel 2015), Ashti (Belyaev 2012) and Shiri (Belyaev In Preparation).

### **15.2 Verb forms with** *kelgʷ-* **'remain'**

The verb *kelgʷ-* (pfv) 'remain, stay, be' is used as an auxiliary in constructions conveying continuous, enduring and sometimes habitually occurring situations and actions in the past. The auxiliary verb is either inflected for the preterite *(kelg-un)* or some other verb form derived from the preterite. The imperfective stem *kalg-* is not used in the auxiliary function. This means that the resulting clauses always have past time reference. The periphrastic verb form can be used with verbs of various valency classes, e.g. intransitive verbs (17), transitive verbs (12), or affective verbs (14).

When the lexical verb bears the imperfective converb suffix, the resulting verb forms have habitual or continuative/progressive semantics (12–14). Note that the verb 'see' in (14) has the literary meaning and thus the sentence refers to a continuous situation of seeing.


Periphrasis with *kelgʷ-* is frequently used when talking about, for example, professions and more generally about the kind of work someone is/was doing (15).

(15) Maħaˁmmadħaˁži Mahammadhazhi acːi-l uncle-erg ce what ʡaˁči work b-irq'-ul n-do.ipfv-icvb kelg-un-il=de? remain.pfv-pret-ref=pst ' Which work was uncle Mahammadhazhi doing?'

15.3 Verb forms with *b-el* 'remain, stay'

With perfective converbs the construction is used for the expression of enduring states that obtain during a longer stretch of time (16–18). The verbs in (17) and (18) refer to the actions of getting up and lying down, but when they are used with the perfective converbs they denote the states that obtain after having carried out the respective actions.


However, the use with perfective verb stems is restricted and it is possible that (16) is rather a biclausal sentence that consists of an adverbial clause with the perfective converb ('not having died') followed by a main clause ('the calf remained'). Thus, in (14) the use of the perfective verb stem would lead to ungrammaticality (19). The precise conditions for this periphrastic verb form when it is used with verbs of different aktionsart classes remains open to future research.

(19) \* it-i-j that-obl-dat rurs-be girl-pl či-b-až-ib-le spr-hpl-see.pfv-pret-cvb kelg-un remain.pfv-pret (Intended meaning: ' He was watching at the girls.') (E)

### **15.3 Verb forms with** *b-el* **'remain, stay'**

The defective verb *b-el* 'remain, stay', which refers to enduring states that obtained in the past and still obtain at the moment of speech, is occasionally used in periphrastic constructions that express the continuation of a state similar to periphrasis with *kelgʷ*described above in §15.2. The verb *b-el* is almost exclusively used with lexical verbs inflected for the imperfective converb. When the bare stem *b-el* is used the construction conveys present time reference, referring to an ongoing event (20) or existing state. When the past enclitic is added to the verb, the periphrastic verb form denotes a past state or an ongoing situation in the past (21), (22).

(20) ištːu-w here-m w-isː-ul m-cry-icvb w-el m-remain iž this

'Here he is crying.' (i.e. he continues to cry).

### 15 Periphrastic verb forms


As mentioned above, the forms with *b-el* and *kelgʷ-* show similarities in their semantics, especially when they bear the past enclitic =*de*. For instance, we can replace *kelg-un* in (14) with *b-el=de* and the meaning does not noticeably change (23). As with the periphrastic verb form with *kelgʷ-*, the use of the perfective verb stem is not allowed in this sentence (24).


Semantic differences are only perceptible when *b-el* without the past enclitc is compared to *kelg-un*, because the former has present time reference whereas the latter has past time reference. Thus, if we use *kelg-un* instead of *w-el* in (20), the resulting sentence conveys past time reference (25).

(25) ištːu-w here-m w-isː-ul m-cry-icvb kelg-un remain.pfv-pret iž this 'Here he remained crying.' (E)

The example in (16) with *kelg-un* can be used in a context in which the calf was about to die, but stayed alive (in fact, it fell down a slope but survived). The sentence in (26), which is a modified version of (16), simply means that the calf is alive and has not died. The semantics indicates that the sentence is actually biclausal and the verb 'die' and *b-el* do not form a verbal complex. Periphrastic constructions with *b-el* are only marginally acceptable if the lexical verb has perfective aspect and carries the perfective converb suffix. Combinations of a perfective converb followed by *b-el* are interpreted as separate clauses. Similarly, (27) is a complex clause expressing two situations: the situation that the rye had not been taken away to the neighboring village of Icari and the situation that the rye had remained in the village of Sanzhi.

(26) a-b-ebč'-ib-le, neg-n-die.pfv-pret-cvb b-el n-remain hel that qːačːa calf 'The calf, not having died, is alive.' (E)

15.4 Verb forms with the auxiliary *b-irχʷ-* (ipfv)/*b-iχʷ-* (pfv)

(27) sːusːul gu-r-a-d-erqː-ib-le, d-el=de

rye sub-abl-neg-npl-take.pfv-pret-cvb npl-remain=pst

'The rye was still not taken (to Icari) and had remained (in Sanzhi).'

More frequent than the use of *b-el* in finite periphrastic constructions as discussed in this section is the use in periphrastic constructions that function as heads of temporal adverbial clauses (§18.2.7).

### **15.4 Verb forms with the auxiliary** *b-irχʷ-* **(ipfv)/***b-iχʷ-* **(pfv) 'be, become, be able'**

### **15.4.1 Periphrastic conditionals**

The auxiliary *b-irχʷ-* (ipfv)/*b-iχʷ-* (pfv) 'be, become, be able' is used in periphrastic conditional constructions, where it is inflected for various conditional forms such as the realis conditional (28) or the past conditional (29) (see §18.3.1 and §18.3.2 for more examples and §25.2 for the syntax of conditional clauses). The periphrastic conditionals are functional equivalents of the simple conditionals and according to Sanzhi speakers there is no semantic difference between them. Thus, in (28) *b-ikː-ul ∅-iχ-utːe* could be replaced by *b-ikː-aχː-at(te)*, and in (29) *b-alχ-ul r-iχ-utːel* could be substituted by *b-alχ-aχː-at(te)* without any change in the meaning of the sentences.

(28) at 2sg.dat χabar story b-ikː-ul n-want.ipfv-icvb ∅-iχ-utːe, m-be.pfv-cond.2sg hel=ʁuna that=eq atːa father ∅-irχʷ-i m-be.ipfv-hab.pst hel-i-la that-obl-gen

' If you want stories, (here is one), she had such a father (and these were the stories about him).'

(29) d-ikː-an-ce npl-want.ipfv-ptcp-dd.sg b-alχ-ul n-know.ipfv-icvb r-iχ-utːel, f-be.pfv-cond.pst wallah, by.God ce-k'a what-indef b-iχʷ-ar=ra n-be.pfv-cond.3=add herʔ-adi say.ipfv-hab.pst.1 du=ra 1sg=add cek'u whatchamacallit ' If I had known what you want, by God, I would also have said something.'

### **15.4.2 Epistemic modal constructions**

The same auxiliary is widely used in epistemic modal clauses to convey the meaning 'probably, possibly, presumably'. In such constructions it is mostly inflected for the infinitive (suffix -*ij*) (30) or the modal interrogative *(-ide)* (34), and very rarely also for the realis conditional *(-ar)* (35). Since these are suffixes that are predominantly or exclusively used with perfective stems, it is mostly the perfective stem *b-iχʷ-* that occurs in the epistemic modal constructions. However, it is also possible to use the future in the past, in which case the imperfective stem of the auxiliary must be employed (33). The

### 15 Periphrastic verb forms

auxiliary agrees in gender and number with the absolutive argument, just like the lexical verb. It does not assign case to the arguments; case assignment is determined by the lexical verb. Thus, it behaves just like any other auxiliary. There is, however, one important difference. In normal analytic verb forms and other periphrastic constructions, the auxiliary cannot occur in the form of an infinitive or conditional, since these suffixes are only used in subordinate clauses such as complement or conditional clauses. By contrast, in epistemic modal constructions, such a use is possible. For instance, example (30) contains a lexical verb bearing the imperfective converb suffix and the auxiliary *b-iχʷ-*, to which the infinitive is suffixed. The resulting clause is nevertheless a grammatical independent main clause. The lexical verb appears in a finite or non-finite verb form. A similar construction is attested in Icari Dargwa (Sumbatova & Mutalov 2003: 110).


The auxiliary can be used as the only verb of the clause. It occurs in the form of the infinitive but nevertheless functions as the head of an independent clause (32).

(32) ik' this.up admi person ∅-iχʷ-ij m-be.pvf-inf 'This is probably a man.'

As mentioned above, the auxiliary can also be inflected for future in the past, which itself already has epistemic modal semantics (§14.1.4) (33). Alternatively, the modal interrogative form is attested (34); this form is otherwise only used in questions with first person subjects and epistemic and deontic modality (§17.4). Very occasionally the auxiliary appears in the form of the realis conditional (35).


Together with the infinitive, only third person controllers of person agreement (which are nevertheless suppressed, since the auxiliary is in the infinitive) are allowed. The use of the modal interrogative also permits first and second person subject-like arguments:

15.4 Verb forms with the auxiliary *b-irχʷ-* (ipfv)/*b-iχʷ-* (pfv)

(36) it dem / / du 1sg r-arx-le f-right-advz r-iχʷ-ide f-be.pfv-modq 'She/I was probably right.' (E)

Negation can be expressed on the auxiliary (37) or on the lexical verb (38). In each case it has scope over the entire clause.

(37) nuˁq-be arm-pl aq high d-arq'-ib npl-do.pfv-pret ca-d cop-npl ik'-i-l=ra. dem.up-obl-erg=add b-ikː-ul n-want.ipfv-icvb a-b-iχʷ-ij neg-n-be.pfv-inf

'He also raised his arms. He probably does not want to (be taken away).'

	- b. ʁaj word r-ik'-ul f-say.ipfv-icvb a-r-irχʷ-an=de neg-f-become.ipfv-ptcp=pst heχ dem.down 'Probably she was not scolding.' (E)

The use of the future in the past and the modal interrogative in a construction expressing epistemic modality is not particularly surprising, since (i) these forms have meanings that are similar to epistemic modality, and (ii) they are finite, that is, they can function as heads of main clauses. The use of the realis conditional and the infinitive, however, deserves further explanation. A plausible path of development is conventionalized ellipses of the main clause similar to examples of insubordination that have been investigated by Evans (2007) and Evans & Watanabe (2016). Full conditional constructions consist of an apodosis with the conditional form and a protasis, in which the verb can choose from a rich array of possible morphosyntactic forms (Chapter 18.3). In periphrastic conditionals, Sanzhi makes use of *b-iχʷ*, as was shown above in §15.4.1. If in a periphrastic conditional such as (39) the protasis is omitted, we are left with a clause expressing a likely condition for an unspecified situation ('if X obtains'). The conditional force has been lost and instead the proposition is judged as probable or possible, i.e. 'if X obtains' > 'X probably obtains' (35). In fact, even if there is a protasis, it is nevertheless possible to have two readings for some apodosis clauses, namely a conditional reading and an epistemic modal reading.

	- these Ashura-erg npl-mow.pfv-pret-dd.pl grass neg-npl-be.pfv-cond.3 ' If Ashura has not mowed this grass, …' > 'Ashura has probably mowed this grass.'

### 15 Periphrastic verb forms

A similar development might also be posited for the epistemic modals that are formed with the infinitive of *b-iχʷ-*. They possibly go back to epistemic and perhaps also deontic modal constructions with main predicates such as *belki* 'be possible' or *ʡaˁʁunil* 'necessary, needed, must, should' that take infinitival complements (40). If the main clause is omitted, only the clause with the infinitive remains, which in examples such as (42) has undergone a re-interpretation from deontic to epistemic modality: 'X should obtain' (41) > 'X probably obtains' (43).


### **15.5 Epistemic modality with the auxiliary** *b-urkː-* **'find'**

In addition to the epistemic modal construction described in §15.4.2, there is another construction, which makes use of the verb *b-urkː-* 'find'. The perfective stem of this verb (*b-arkː-*) is only used as an affective verb with the meaning 'find', which means that it requires an experiencer in the dative and a stimulus in the absolutive case. The imperfective stem *b-urkː-* is used both with the meaning 'find' and as an auxiliary with the epistemic meaning 'probably, be possible'.

The *b-urkː-* constructions bears a strong similarity to the other epistemic modal construction because the lexical verb can be finite or non-finite. The default position of *b-urkː-* is the default position for auxiliaries, namely following the lexical verb. It can be the only verb in the clause and still have the epistemic meaning. The auxiliary can be negated, then behaving like any other auxiliary: it expresses the negation of the predication, that is, it has scope over the lexical verb (44). But negation can also be expressed on the lexical verb with the same semantic effect (45).

(44) vmešiwatsa mingle iχʷ-ij be.pfv-inf b-ikː-ul n-want.ipfv-icvb a-b-urkː-ar neg-n-find.ipfv-prs '(He) probably does not want to get involved.'

15.5 Epistemic modality with the auxiliary *b-urkː-* 'find'

(45) c'il then a-b-irq'-an-ne neg-n-do.ipfv-ptcp-fut.3 ∅-urkː-ar m-find.ipfv-prs ' Then he will probably not do (this again).'

The gender/number agreement can follow the ergative pattern and thus be with the absolutive argument (46), or it can follow the accusative pattern. In the latter case it is controlled by the subject-like argument, which can be in the ergative or dative case ('deviant gender agreement') (47). Such behavior is not attested for all auxiliaries, but the copula allows for it (§20.2.4).


In all examples discussed so far, the auxiliary is inflected with the suffix *-ar* (44–47). The lexical verb (if there is any) is responsible for the temporal reference. The suffix *-ar* is also used in the epistemic modal construction with *b-iχʷ-* (§15.4.2), and it looks like the realis conditional suffix for the third person. However, the realis conditional is normally only formed from perfective stems (§18.3.1) and the form (*b*)-*urkː-ar* never expresses conditional semantics. Therefore, although we can suppose that there is a diachronic relationship with the conditional, synchronically the form cannot be analyzed as conditional, but is glossed with prs. Instead of the suffix *-ar*, it is also possible to inflect the auxiliary regularly for the habitual present (48) or the habitual past, resulting in regular person agreement (49), (50).


### 15 Periphrastic verb forms

It is possible that the construction goes back to a complement construction with *b-urkː*as matrix predicate; it is currently grammaticalizing and therefore one finds variation between those sub-constructions that show person agreement and those that do not, and between the locus of negation and the expression of temporal reference (i.e. whether the auxiliary or the main verb conveys the temporal reference). Similar epistemic modal constructions involving a verb 'find' are attested in many other Dagestanian languages including other Dargwa varieties, Hinuq and other Tsezic languages, Avar, and Archi (Forker 2018a,b).

In summary, the epistemic modal constructions with *b-iχʷ-* 'be, become, be able' and *b-urkː-* 'find' have approximately the same range of meanings, and further research is needed to clarify if it is possible to establish semantic differences between them. The only difference observed so far pertains to morphosyntax. The verb *b-iχʷ-* is most commonly used in the form of the infinitive, which only allows for third person subject-like arguments, as the corpus examples in §15.4.2 illustrate. By contrast, *b-urkː-* is also attested in inflected forms that have first or second person agreement controllers (48), (50).

### **15.6 Indirect evidentiality with the auxiliary** *b-ug-* **'be, be at, stay, remain'**

The verb *b-ug-* 'be, be at, stay, remain', in addition to its use as the only predicate of a main clause, occurs as an auxiliary with evidential/inferential meaning. The auxiliary predominantly has the form of the resultative (preterite + copula), but the perfect or pluperfect are also attested. In other words, it is inflected for verb forms that by themselves express resultativity. The lexical verb appears in the form of the imperfective or perfective converb. The alignment of the auxiliary is identical to that of the lexical verb, that is, it agrees in gender with the absolutive argument.

The use of this auxiliary for conveying indirect evidentiality is a common strategy in many (if not all) Dargwa varieties, especially in traditional stories. For instance, the introductory formula for tales in Sanzhi is *b-už-ib ca-b b-už-ib-le=kːu* (n-be-pret cop-n n-be-pret-cvb=cop.neg) 'once upon a time', with the second occurrence of *b-už-* being optional.

The construction expresses non-firsthand evidentiality, in particular propositions based on inferences from traces or results (51), (52) or reasoning (57c). For instance, example (51) occured in a fairy tale in which the villain ate all the people of a village. After she was killed and her belly was opened, the people could be rescued because they were still alive.

(51) il-i-l that-obl-erg č'aˁm chew b-irq'-ul hpl-do.ipfv-icvb b-už-ib-le=kːu, hpl-be-pret-cvb=neg qurt' swallow iʁ-ul do.pfv-icvb b-už-ib hpl-stay-pret ca-b cop-hpl χalq' people '(It turned out that) she did not chew the people, but swallowed them.' 15.6 Indirect evidentiality with the auxiliary *b-ug-* 'be, be at, stay, remain'

(52) c'il then il-i-la that-obl-gen šljaˁp'a hat kelg-un-ne remain.pfv-pret-cvb b-už-ib-le=de n-stay-pret-cvb=pst 'Then (apparently) his hat remained there.'

Negation is expressed on the auxiliary, but negated lexical verb forms that have scope over the auxiliary are also possible. Thus, in elicitation, both ways of negating can be obtained, although with a slight semantic difference that becomes apparent if both verbs are negated (55). In the latter case we can see that the scope of the negation prefix is the evidential auxiliary together with the whole clause if the auxiliary bears the prefix (53), (55). By contrast, the scope is the lexical verb together with its arguments, but excluding the evidential auxiliary, if the prefix appears on the lexical verb (54).


'Then it did not turn out that his hat did not remain there, (but by contrast) we found it (there).'

Indirect evidentiality can also include surprise about the inference if it contradicts the expectations of the speaker (56).

(56) ix-tːi dem.up-pl bahsar, first ix-tːi dem.up-pl qːuʁa-l beautiful-advz er life b-irχ-ul hpl-be.ipfv-icvb b-už-ib hpl-stay-pret ca-b cop-hpl hex-tːi, dem.up-pl er look r-erč'-e! f-look.pfv-imp '(It turned out) they lived well first, look!' (said to a woman)

Sometimes only evidential meaning is expressed, for example in narrations about past events of which no traces remained. In other cases, the speakers acquired their knowledge from the narrations of other people including their ancestors, such that the auxiliary expresses hearsay evidentiality. For example, (57) is part of a longer account about the history of the Sanzhi people, and the speaker speculates about other people who are

### 15 Periphrastic verb forms

said to have lived close to Sanzhi, and others who are said to have come to Sanzhi and destroyed the village. There are no visible results of these events. Instead, the speaker, based on his knowledge of the topography of Sanzhi and of stories about assaults on the village, hypothesizes from where enemies could have reached Sanzhi.

	- b. kerx-ul kill.ipfv-icvb b-už-ib hpl-stay-pret ca-b cop-hpl nišːa-lla 1pl-gen 'They apparently killed our (people).'
	- c. het that šaˁrʡaˁ-rka Shari-abl sa-b-ax-ul hither-hpl-go.ipfv-icvb b-už-ib hpl-stay-pret ca<b>i cop<hpl> 'They came from Shari (to us).'

In general, the use of *b-už-* can be considered to represent a stylistic device for traditional narratives and other traditional stories about the past, including funny and fictional anecdotes that Sanzhi people recite about their ancestors (58).

(58) il-tːa-lla that-obl.pl-gen kʷiriž, machete il-tːa-lla that-obl.pl-gen beretːa ax cik'al something b-a-b-už-ib; n-neg-n-be-pret a-b-alχ-ul neg-n-know.ipfv-icvb b-už-ib n-stay-pret ca-b cop-n 'They had no machete nor ax; they did not know (these tools).'

When the auxiliary is used with the first person we get the reading that the speaker does not consider himself as an active, conscious participant in the event, and was rather informed about its true properties and implications afterwards, in other words, we obtain the first-person effect (59).

(59) [When I was a small child my father took me to Moscow to a meeting of the Party.] dam 1sg.dat Stalin Stalin či-w-až-ib-le spr-m-see.pfv-pret-cvb už-ib-le=de stay.m-pret-cvb=pst '(It turned out, that) I (masc.) had seen Stalin.' (E)

## **16 The copula and other auxiliaries**

The copula function in copula clauses as well as the formation of periphrastic verb forms is fulfilled by predicative particles (enclitics), a copula verb (§16.1) and other auxiliaries (§16.3). In addition, Sanzhi has a number of specialized copulas for locational and existential clauses (§16.2). The syntactic properties of copula clauses with examples of predicative particles and verbs in the copula function are treated in §22.2.

Predicative enclitics are *=da* (first person singular and plural, second person plural), *=de* (second person singular), *=de* (past time reference), *=q'al* (modal particle), *=e*/*=ja* (marker for content questions), *=w*/*=uw*/*=ew* (marker for polar questions) and *=l*/*=jal*/*=el* (marker for embedded questions). They are not verbs and are therefore treated separately in §9.1.

### **16.1 The copula**

The affirmative copula is *ca-b* or *ca<b>i* with a gender/number agreement affix (the longer variant is much less used than the shorter one). The variant *ca-b* is homophonous with the singular reflexive pronoun in the absolutive case, and they seem to be cognates. The copula is morphologically defective, as it cannot be inflected like other verbs. The only verbal category it expresses on its own is gender/number agreement; and it can be inflected for the masdar (*ca<b>ni*). If no further predicative particles are encliticized it conveys present tense reference with third person arguments and affirmative polarity. It has the same functions as the predicative particles, i.e. it heads copula clauses (1) and it is used in analytic tenses of main clauses (2), but only when the person agreement controller is third person. In copula clauses the agreement is always controlled by the subject (see example (7) in §22.2).


Predicative particles such as the person markers, the past marker, the modal particle or the interrogative markers can be added to the copula (3). When person markers are used together with the copula they are obligatorily attached to it as in (3) and can never be encliticized to another constituent (4). Furthermore, first and second person subjects require the use of the person marker (5). The use of the copula as in (3) is optional and cannot replace the person marker (6).

16 The copula and other auxiliaries


The past tense enclitic can also be used with (7) or without the copula (8) without any difference in the semantics. The use of the copula alone conveys present time (1), so it is the past enclitic that expresses the past time reference.

(7) cet'-le how-advz atːa father ʡaˁč-le-r work-loc-abl s-ax-an=qːel hither-go-ptcp=when durħuˁ boy razi-l happy-advz ca-w=de cop-m=pst ' When the father came back from work, the boy was happy.'

(8) durħuˁ boy razi-l=de happy-advz=pst 'The boy was happy.' (E)

The copula has a masdar form built with the normal masdar suffix -*ni*. As with other verbs, the masdar occurs in complement clauses:

(9) di-la 1sg-gen juldašːe friend.pl [urk'i heart ač-te open-dd.pl ca-b-ni] cop-hpl-msd gu-r-b-uq-un down-abl-hpl-go.pfv-pret 'My friends turned out to be open-hearted.'

For all other functions that verbs fulfill when heading independent or dependent clauses, e.g. the use as participles in relative clauses or as infinitive in complement clauses, the auxiliaries described in §16.3 below are used.

The copula is also used as an auxiliary for a number of analytic verb forms with third person arguments that control the agreement (compound present, obligative present, resultative, perfect, experiential I & II) (12). It can never be used in such verb forms with first or second person agreement controllers, not even when person markers are encliticized (10); in such clauses the person markers on their own must be used (11). However, it is possible to encliticize the past marker *=de* to the standard copula and then use it in clauses with subjects of all persons, although this has only been attested in elicitation (13). In general, the past enclitic =*de* is incompatible with the person enclitics, but not with the copula (§9.1).

16.1 The copula


The stem of the negative copula is *(b-)akːʷ-*. It occurs in the following forms:


The full paradigms of the negative copula in the present and the past tense are given in Table 16.1 and Table 16.2. The present tense has a short variant that appears as an enclitic *=kːu* and the past tense has the enclitic *=kːʷi*. The enclitic variants are only used for third person. Examples are found in Chapter 14.

Table 16.1: The negative copula in the present tense


Table 16.2: The negative copula in the past tense


### 16 The copula and other auxiliaries

As can be seen from Table 16.1 and Table 16.2, as well as from the examples, there are two complications. The first is the syncretism of the present and the past tense in the first and second person forms, which is due to the general syncretism of the simple present and past. The third person form of the simple past can, however, also be used for the first and second person, so that in this tense person marking can be avoided and confusion with the simple present circumvented (Table 16.2). The second complication concerns the gender prefix. In principle, the verb can agree, but an agreeing negative copula can only have an existential or locational interpretation; it never has the normal copula meaning. Thus, in (14–16) gender agreement is prohibited because the clauses have identificational semantics, close to the equals sign (*=*). For instance, in (14) the unexpressed copula subject is female, but the copula does not exhibit feminine agreement. Similarly, in (16) the copula subject is male, but the copula does not show agreement (and copula predicates never control agreement).


'He was not one of us, he was Icari, the head (of the kolkhoz).'

As mentioned above, when the negative copula is used with a gender agreement prefix the meaning is existence or location. For this type of meaning the use of the prefix is obligatory. For instance, in (17) the subject is female and the gender prefix is the one for the feminine gender; in (19), by contrast, the agreement controller is male. In principle, the negative copula with gender prefixes can be treated as a separate word that is functionally analogous to the negated forms of the locational copulas described in §16.2, which consist of the negative copula with the gender prefix and the roots of the locational copulas (32).


16.1 The copula

When the copula functions as auxiliary gender agreement is prohibited (20), (21).<sup>1</sup> In tag questions, the negative copula is always used without the gender agreement prefix (22). This is what one would expect, since in affirmative tag questions also only the standard copula and not a location copula is used (see §28.3 on tag questions).


The participle of the negative copula is *(b-)akːʷ-ar(re)*. It translates as 'not having, without' (23), (24) and fulfills the function of a postposition (§8.2.4). It can take further suffixes such as the cross-categorical suffixes *-te/-ce* and *-il*, the concessive marker *=xːar* (25), the suffix *-dex* that derives abstracts nouns, and others. Again gender agreement is, in principle, possible, but very rare in texts (25) and the semantic differences between the omission of agreement prefixes and their occurrence are identical to what was said before: no gender agreement prefix means copula function; gender agreement prefix means locational and/or existential function (25).


'as if his arrival (was) unexpected (lit. without news), yes unexpected ...'

(25) hel-itːe, that-advz du 1sg w-akːʷ-ar=xːar m-cop.neg-ptcp=conc aʁʷc'alla 40.days d-arq'-ij 1/2pl-do.pfv-inf d-irχ-an=da 1/2pl-be.able.ipfv-ptcp=2pl ušːa-l 2pl-erg nawerna probably

<sup>&#</sup>x27;Like that, even me not being there, you should be able to do the 40 days, probably.' (i.e. the religious ceremony held 40 days after the death of a person)

<sup>1</sup>The only exception are the occasional use of affirmative locational copulas, which have gender prefixes, as auxiliaries in periphrastic verb forms (§15.1).

### 16 The copula and other auxiliaries

The masdar of the negative copula is *(b-)akʷ-ri/akʷ-ni*. The latter form does not have an agreement prefix (not even when it encodes existential or locational meaning as in (26)). It mainly occurs in complement clauses:


### **16.2 Locational copulas**

There are four locational copulas that share a consonant bearing a deictic meaning with the demonstrative pronouns (§4.2). However, in the case of the copulas this is the initial consonant, whereas with the demonstratives it is the stem-final consonant. Furthermore, the copulas agree in gender/number, whereas the demonstratives lack agreement. Except for the first copula (*le-b*)/pronoun (*hel*) the semantics of the verbs perfectly match the semantics of the pronouns (Table 16.3).

Table 16.3: Locational copulas and demonstrative pronouns


The copulas can attach further suffixes (e.g. participles, temporal markers such as *=qːella* or *=er*, cross-categorical suffixes, the masdar *-ni*) (28) and predicative enclitics (past marker, person marker) (29–31), just like the copula. But – like the copula – they are defective in comparison to standard lexical verbs because most of the verbal suffixes cannot be added (e.g. suffixes for the habitual present and habitual past, conditional suffixes, the infinitive, etc.). The most frequent copula is *le-b*, which fulfills a kind of default function.

16.2 Locational copulas


'That is the woman, the one who was keeping the boy in her hands.'

(31) χadižat, Khadizhat čina-r=de where-f=2sg u? 2sg du 1sg Sanži-r=da Sanzhi.loc-f=1 / / Sanži-r Sanzhi.loc-f le-r=da exist-f=1 ' Khadizhat, where are you? I am in Sanzhi.' (E)

All locational copulas except *le-b* can be negated by suffixing the negative copula, and the gender agreement follows the standard rules (which means that it is controlled by the subject), i.e. *te-b-akːu, k'e-b-akːu,* and *χe-bakːu* (23) in the present tense and *te-b-akːʷi, k'e-b-akːʷi,* and *χe-b-akːʷi* in the past tense (32).

(32) urq'l-e board-pl te-d-akːʷ-i exist.away-npl-cop.neg-hab.pst hitːu-d there-npl 'There were no boards there.'

The existential copulas, in particular *le-b* due to its less specific meaning, are occasionally used as auxiliaries in periphrastic verb forms together with lexical verbs that bear the perfective or the imperfective converb suffixes (33), (34) (§15.1).

(33) [Talking about how a particular plant grows.] di-la 1sg-gen Asijat-la Asiyat-gen ca one qal-la house-gen baˁʡ fassade ka-b-uc-ib-le down-n-catch.pfv-pret-cvb k'e-b exist.up-n ' [It can become large], at my Asiyat's place it covered one wall of the house.'

(34) du 1sg b-ax-ul n-go.ipfv-icvb le-b=de exist-n=pst

'(At the time when the fox was born), I was walking around (said the wolf).'

<sup>2</sup>Regarding *le-b-te*: the agreement on the locational copula should better be *le-d-te* because *arc* 'money' normally controls neuter plural agreement, but neuter singular is also possible.

16 The copula and other auxiliaries

### **16.3 Other verbs used in copula-functions and as auxiliaries**

There is another copula-like verb *b-el* with the meaning 'stay, remain'. It conveys past time reference, although it does not carry any overt marking (35), (36). The verb is defective and has a reduced inflectional paradigm. In main clauses usually only the bare stem is used. It is the most frequently used base verb in the compound 'remember' (36).


It is possible to add person markers (37) or the past enclitic (38), but it cannot be combined with the copula (e.g. *\*b-el ca-b*). The use of person markers with first and second agreement controlling arguments is obligatory, i.e. in (37) the person enclitic *=da* cannot be omitted.

(37) xːunul-li-cːella woman-obl-comit hana now ižal today busːaˁʡaˁt=ra this.time=add canille together w-el=da, m-remain.pfv=1 c'aq' strong xːunul

woman

'I stayed together with my wife even until today; a good (strong) woman.'

(38) žaˁbraˁʔil=ra Zhabrail=add w-el=de m-remain.pfv=pst 'Zhabrail remained (i.e. stayed alive).'

It can be negated in a way that strongly resembles the locative copulas (namely by employing the negative copula), but with a small change in the stem vowel, i.e. *b-il akːu* and *b-il-akːʷi* (39), (*\*b-el akːu*). When the verb *b-el* bears the perfective converb suffix *-le* the stem vowel remains unchanged under negation (40).


### 16.3 Other verbs used in copula-functions and as auxiliaries

…

It can be inflected for some verb forms that occur in subordinate clauses, namely for the perfective converb (41), the referential attributive form with *-ce*/*-te* (e.g. *d-el-te*), the referential attributive form with *-il* (42), the temporal enclitic *=qːella* (*b-el=qːella*) and the masdar (*b-el-ni*).


In conditional clauses it must occur in a periphrastic verb form together with *b-iχʷ*since it cannot itself be inflected for any conditional form (43).

(43) hu well šalakbluk-me=ra concrete.block-pl=add d-el npl-remain.pfv d-iχʷ-ar, npl-be.pfv-cond.3 'If the concrete blocks remained (are left over), …'

It can serve as an auxiliary in periphrastic verb forms that head main clauses (§15.3) or, more commonly, temporal adverbial clauses (§18.2.7). The latter function is, alongside with the use in main clauses as illustrated in the examples above, the most frequently attested use of this verb.

For all tenses or subordinate clause types, in which the predicative enclitics/negative copula cannot be used, the verb *b-irχʷ-/b-iχʷ-* 'be, become, occur, can' is employed. This verb has the full inflectional paradigm including conditional forms (44) and future forms (45) and is negated like any other lexical verb. In addition to its use as a copula, as a normal lexical verb and in compound verbs (§12.2), it also occurs as an auxiliary in epistemic modal constructions (46) (§15.4.2) and in realis conditional and irrealis conditional clauses (§15.4.1).


There are four more verbs that are also used in copula function and as auxiliaries. The verb *b-irk-/b-ik-* 'be, occur, become, get, receive' is used in copula constructions (47), especially with predicates that are marked with the in-essive, and in many compound verbs (48).

16 The copula and other auxiliaries


The imperfective verb *b-urkː-* 'find' is regularly used in epistemic modal constructions similar to those formed with the verb *b-irχʷ-/b-iχʷ-* 'be, become, occur, can' just mentioned (§15.5). In this function it can be used together with a lexical verb or as the only verb in a copula clause:

(49) kulpat family b-urkː-ar hpl-find.ipfv-prs.3 heχ-tːi dem.down-pl 'They are probably a family.'

The verb *b-už-* (pfv) 'be, stay, remain' is used in copula clauses with evidential semantics (50) and, more generally, as an auxiliary in evidential constructions (§15.6). It is not used in compounding and not for analytic verb forms.

(50) ha uh ʡaˁrʡaˁ chicken b-už-ib n-stay-pret ca-b cop-n 'Ah, it turned out to be a chicken.'

The verb *k.elgʷ-* (pfv) 'remain, stay, be' is used in copula clauses and as an auxiliary conveying habitual or continuative/progressive meaning (51). It is also not used in compound verbs (see §12.2.5 for more examples).

(51) d-aqe npl-long čːal argument d-ik'-ul npl-say.ipfv-icvb kelg-un remain.pfv-pret ' They were arguing for a long time.'

## **17 Non-indicative verb forms**

Non-indicative (or non-declarative) verb forms occurring in Sanzhi are imperative (§17.1), prohibitive (§17.2), optative (§17.3) and the modal interrogative (§17.4). The imperative, the prohibitive, and the modal interrogative are restricted in their use to the second person for the first two forms and the first person for the last form. The imperative and the prohibitive share the (partial) distinction between intransitive and transitive verbs expressed through the use of dedicated stem-augment vowels in the suffixes. The same distinction and the same formal means of expressing it are found with synthetic verb forms and conditionals (§11.4).

### **17.1 Imperative**

The form of the imperative depends on the inflectional class and on the transitivity of the verb. The suffixes are given in Table 17.1. Verbs that have the preterite suffix *-un* have the suffixes *-en* and *-ene(ja)* for singular and plural imperatives respectively, independently of their transitivity. The other three verb classes distinguish (almost) always between intransitive and transitive verbs in the formation of the singular imperative: intransitive verbs employ the suffix *-e*; transitive verbs make use of *-a*. The distinction is absent in the plural imperative, which has the suffixes*-ene(ja)* and *-aj(a)* (and *-ere* as an alternative that is not frequently used).<sup>1</sup> The suffix *-(j)a*, which is optionally found with all plural imperatives as well as the plural forms of the prohibitive and the second person plural optative, can be analyzed as a plural addressee marker following the suggestion by Sumbatova & Lander (2014: 163–165). See §9.5 for some other contexts of its use.

For those verbs that have an imperfective and a perfective stem, the imperative is mostly formed from the perfective stem. Regular exceptions to this rule are the imperfective stems of the verbs 'eat' and 'drink' that often behave differently from other imperfective verbs. They have the following imperatives: for the imperfective stem *r-učː-e* (sg, f)/*d-učː-aja* (pl) 'drink (several times)' and perfective *b-erčː-a* (sg)/*b-erčː-aja* (pl) 'drink (once)'; imperfective stem *r-uk-en* (sg, f)/*d-uk-ene(ja)* (pl) 'eat (several times)' and perfective stem *b-erkʷ-en* (sg)/*b-erkʷ-en(ja)* (pl) 'eat (once)'. Other verbs that allow for the imperfective and the perfective stem to serve as the basis for the imperative are, for example, *k.alž-* (ipfv)/*k.elg-* (pfv) 'remain, stay'. Not all morphosyntactically affective verbs allow for an imperative, but some do, such as *b-aχ-e* (n-know.pfv-imp) (3). Similarly, the verb in (2) has experiental/affective semantics (though its subject appears in the dative) and allows for the imperative. With those types of verbs the imperative has rather the meaning of a wish of the speaker or a deontic flavor similar to 'you should'.

<sup>1</sup>There is one verb with the meaning 'go, leave' that is exceptional because it also allows for the suffixes -*aˁn* and -*aˁne*, cf. the last two lines of Table 17.1.

### 17 Non-indicative verb forms


Sentences (1–3) illustrate the use of the imperative. Though it is not particularly common, imperative clauses can contain the second person pronoun referring to the addressee, which can be an argument in the absolutive (2), in the ergative (15), or in the dative.

(1) hel that ceʁuna which b-urs-a! n-tell.pfv-imp χabar story b-urs-aχː-atːe n-tell.pfv-cond-cond.2sg hitːi behind ka-jž-e! down-remain.pfv-imp

' Tell it like this! Sit down if you (sg) narrate the story!'


As in some other Dargwa varieties (e.g. Icari, Sumbatova & Mutalov 2003: 98, Shiri, Belyaev In Preparation), the imperative cannot be used when the P argument of a transitive verb is first person. In this case, the optative is used instead (4). With second and third person P arguments the imperative is allowed, cf. (6) below.

17.1 Imperative

```
(4) w-at-ab-aja
     m-let.pfv-opt-2pl
                        du!
                        1sg
     'Leave me!'
```
The imperative can be used in combination with the antipassive construction, in which case the verb is intransitive and takes the suffix *-e* if the addressee is singular:

(5) χabur-t-a-l story-pl-obl-erg ∅-ux-e m-tell.ipfv-imp / / r-ux-e! f-tell.ipfv-imp 'Tell stories!' (E)

With labile verbs the use of *-e* and *-a* in the imperative singular is possible, where *-e* indicates that the verb is used intransitively and *-a* that it is used transitively (6–7).

(6) ust'ul chair b-it'kak'-a! n-move.pfv-imp ' Move the chair!' (E)

(7) w-it'kač'-e! m-move.pfv-imp 'Move (yourself)! (said to a man)' (E)

The imperative of the verb *b-ax-* (n-go-) is exceptional because it can be formally zero (i.e. *waš* instead of *waše*), and functionally it can be used as a cohortative, i.e. to encourage or discourage to perform an action together with the speaker. Thus, in (8) *w-aš-e* is not meant as a command to the addressee to perform the action alone, but is intended as an invitation to go together and have a look at the shop. Morphosyntactically, however, the utterance is addressed to a man because the verb agrees in gender with the addressee (masculine singular) but not with whole group. Similarly, in (9) the verb 'go' is used in the function of a cohortative. However, syntactically the periphrastic cohortative constructions are either asyndetic conjunctions of two independent main clauses as in (8) ('Go!' and 'We will look.') or complex clause constructions consisting of a main clause in the imperative followed by a purpose clause in the infinitive or subjunctive as in (9) ('Go in order to eat!').

(8) "w-aš-e=kːʷa!" m-go-imp=prt ∅-ik'ʷ-ar, m-say.ipfv-prs "ce what k'e-b-il exist.up-n-ref er look d-urk'-an=da", 1/2pl-look.ipfv-ptcp=1 ∅-ik'ʷ-ar, m-say.ipfv-prs "tuken-ne-b" shop-loc-n

'" Let's go (together) and look what is there in the shop!", he says.'

(9) xːunr-a-l woman.pl-obl-erg waˁw shout haʔ-ib, say.pfv-pret "d-ax-aj" 1/2pl-go-imp.pl b-ik'ʷ-ar, hpl-say.ipfv-prs "d-uk-ij, 1/2pl-eat.ipfv-inf d-uk-utːaj!" 1/2pl-eat.ipfv-subj.2

'The women shouted, "Let's go eating!"'

### 17 Non-indicative verb forms

### **17.2 Prohibitive**

The prohibitive is formally independent of the imperative. It consists of the prefix *ma*and a suffix. The prefix is positioned between the orientation and the deixis/gravitation preverbs if there are any. The suffixes are similar to the habitual present (§13.1) because they make use of the same stem augment, which depends on the transitivity of the verb. Intransitive verbs take *u*; transitive verbs take *i*, and one verb behaves exceptionally (§20.3.1). In the singular the suffixes are *-ut*/*-utːa*, *-it*/*-itːa*, and *-aˁt*/*-aˁtːa*; in the plural they are *-utːaj(a)*, *-itːaj(a)*, and *-aˁtːaj(a)*. The short and long variants in the singular and plural seem to be in free variation. The prohibitive is only formed from the imperfective stem in case a verb has both stems. Exemplary verbs in the prohibitive singular are shown in the last column of Table 17.2.

Table 17.2: The prohibitive suffixes


As with the imperative, the prohibitive is only used with second persons. The second person pronoun functioning as the addressee is mostly omitted, but it can be overtly expressed. Examples (10) and (11) show intransitive verbs. Sentences (12), (13) illustrate transitive verbs. In addition, affective verbs with dative addressees (experiencers) are allowed (14).


17.3 Optative



As mentioned above for the imperative, with first person P arguments the prohibitive cannot be employed. Instead, the negative optative must be used (15).

(15) ma-jt-aba proh-beat.up.m-opt.1 du! 1sg u-l 2sg-erg u 2sg w-it-a! m-beat.up-imp ' Do not thrash me (masc.)! Thrash yourself (masc.)!' (E)

When the ergative construction or the antipassive construction occur together with the prohibitive, the difference in transitivity is reflected in the different stem augment vowels, that is, the antipassive construction requires *u* (16a), whereas the ergative construction requires *i* (16b).

```
(16) a. Antipassive construction
         hin-ni
         water-erg
                    ma-d-učː-utːaja!
                    proh-1/2pl-drink.ipfv-proh.pl
         'Do not drink water (regularly)!'
```
b. Ergative construction hin water ma-d-učː-itːaja! proh-npl-drink.ipfv-proh.pl 'Do not drink the water!'

### **17.3 Optative**

The optative is formed from perfective verbal stems by means of suffixes (Table 17.3). The suffixes are complex, and *-ab* can be identified as the optative marker to which markers that express person agreement are added. The optative seems to obey the same rules of person agreement that obtain in indicative clauses (§20.3.2). The paradigm has a structure that is similar to other person paradigms, that is, syncretism of first singular and plural with the second plural and zero marking in the third person (§20.3). There is an optional variant *-arte* when the agreement controller is plural. The optative is negated by means of the prefix *ma-*, which is also used for the prohibitive (§17.2).

### 17 Non-indicative verb forms


Table 17.3: The optative

The functions of the optative cover:

	- (17) ašːi-j 2pl-dat / / at 2sg.dat bari day ʡaˁħ good d-iχʷ-ab! npl-be.pfv-opt.3 'A good day to you!' (lit. 'May the day be good for you/to you.')
	- (18) Maˁħaˁmmad-la Mahammad-gen šːal-li-cːe-r side-obl-in-abl d-iχʷ-ab npl-be.pfv-opt.3 ašːi-j 2pl-dat salam-te! greeting-pl
		- 'May there be greetings to you from the side of Mahammad!'
	- (19) ʡaˁpa commemoration b-arq'-ab n-do.pfv-opt.3 cin-na refl.sg-gen 'May God bless him/her!'
	- (20) du 1sg daˁʡle as ma-ka-jsː-ab, proh-down-sleep.pfv.m-opt.3 ja oh Allah! Allah 'May (nobody) lay down (sleep) like me, oh Allah.' (i.e. with so many sorrows)
	- (21) hik' dem.up bari sun ruˁħ disappear b-uq-ij=sat n-go.pfv-inf=as.much ma-kelg-ab-a proh-remain.pfv-opt-1sg du 1sg ' May I not remain until the sunset.' (i.e. May I die before the sunset, I don't mind.)

17.3 Optative

(22) limonad lemonade d-iχʷ-ab npl-be.pfv-opt.3 'Let it be lemonade.' (i.e. The bottle on the picture could be lemonade or something else, I don't care.)

	- (24) Maˁħaˁmmad-li Mahammad-erg at 2sg.dat macːa sheep b-ikː-ab! n-give.pfv-opt.3 ' May Mahammad give you the sheep!' (E)
	- (25) "dam 1sg.dat kumek help b-arq'-aja!" n-do.pfv-imp.pl ∅-ik'-ul, m-say.ipfv-icvb "w-erc-aq-ab-aja!" m-save.pfv-caus-opt-2pl

'"Help me, save me!" he says.'

Especially the first and second functions are used in situations where the speaker does not have control over what is going to happen.

The suffix *-arte* can only be used when the agreement controller is plural (26). In (27) the addressee is a not further specified group of people of whom the speaker wishes that one beloved (masculine) person may die, that is, one man<sup>2</sup> per addressee. Thus, there is a group of people for whom the speaker wishes that they would die, which explains the use of *-arte* and the plural demonstrative pronoun. The addressee is also plural (reflected in the plural possessive pronoun). If the speaker had wished that more than one beloved one should die, the verbs would change to *b-ikː-an-te b-ebk'-arte*.


<sup>2</sup>This explains the masculine singular agreement on the two verbs.

### 17 Non-indicative verb forms

There is also the possibility of using the bare verbal stem in the optative function (28–30) with singular and plural addressee. There is no observable semantic difference between the use of the bare stem and the use of the optative when expressed by the suffixes given in Table 17.3.


The bare optative can even be used like a noun and inflected without the need of adding any derivational morphology. Thus, in (31) the complete optative phrase *urk'i b-ac'* (heart n-thaw.pfv) 'May your/his/her/their heart thaw' has been nominalized and then the dative suffix has been added because the nominal functions as the addressee of the verb 'telephone'. The phrase is used with the idiomatic meaning 'idiot'.

(31) ucːi-li-j brother-obl-dat tilipun telephone d-arq'-ib-le npl-do.pfv-pret-cvb urk'i heart b-ac'-li-j n-thaw.pfv-obl-dat k-ač'-e down-come.pfv-imp d-arq'-ib-le, npl-do.pfv-pret-cvb … 'When they called the brother, this idiot, if he had told me to come, …'

### **17.4 Modal interrogative**

Sanzhi has a suffix *-ide* (with the allomorph *-ida*), which is only used in content questions with first person subject-like arguments of verbs of all valency classes. These questions have a modal meaning covering possibility, deontic modality and future (similar to English *can*, *should*, *will*). The questions are sometimes more like rhetorical questions to which an answer is not expected (32), but they can also have real interrogative illocutionary force as questions that are uttered to solicit answers (34).


17.4 Modal interrogative

(34) "mašin car čina where b-ič-ide, n-lead.ipfv-modq rucːi?" sister ∅-ik'-ul m-say.ipfv-icvb ca-w cop-m '" Where should I bring the car, sister?" he is asking'

The suffix can be added to perfective as well as to imperfective stems with the usual difference in meaning: habitual/iterative/generic if the verb is imperfective (35) vs. specific singular event if the verb is perfective (36).


The suffix is also obligatorily used when a second person absolutive argument of a transitive verb occurs (37). This deviates from the general rule about person agreement because normally in clauses with two speech-act participants both arguments can control person agreement (§20.3.2). Thus, in an indicative clause we could and often would have a second person controlling agreement, as the answer in (37) shows. This is impossible for the modal interrogative. From this we can conclude that the modal interrogative marker is not a person agreement marker, although its use is restricted by person.

(37) u 2sg čina where r-uč-ide, f-lead-modq rucːi? sister du-l 1sg-erg u 2sg r-uk-ul=de f-lead-icvb=2sg qːala fortress qːurejš-le Kurejsh-loc 'Where should I bring you, sister? I will bring you to Kala-Kurejsh (place name).' (E)

Occasionally, *-idel* instead of *-ide* is used (the *l* at the end is the embedded question marker/complementizer *=l*, §28.4). It seems that there is a slight difference in meaning between *-ide* and *-idel*, which reflects the fact that *-idel* is a kind of insubordination, i.e. a use of an originally subordinate form in a main clause. In the Russian translations this is reflected by the use of an additional adverb *interesno* 'interesting', which seems to stand for an omitted matrix clause 'it would be interesting to know'.


### 17 Non-indicative verb forms

However, the form *-idel* is far more common in real embedded questions (40) (see §28.4 for more examples).

(40) iž this ceqːel=el when=indq iž-itːe this-advz pikri thought ∅-ik'-ul=el m-say.ipfv-icvb=indq [d-iʡ-ij npl-steal.pfv-inf ∅-uˁq'-idel m-go.pfv-modq a-w-uˁq'-idel] neg-m-go.pfv-modq ∅-ik'-ul m-say.ipfv-icvb le-w exist-m ' This is when he is probably thinking, should I steal or not.'

The modal interrogative of the verb *b-iχʷ-* (pfv) 'be, become, be able' is also used in epistemic modal constructions (§15.4.2).

## **18 Non-finite verb forms**

There are three types of verb forms that function as heads of subordinate clauses:


Plain non-finite verb forms are simple converbs, participles, the infinitive, the subjunctive, and the masdar. The specialized converbs convey more specific temporal and causal relationships. Plain non-finite verb forms (except for the subjunctive) and constructions with specialized converbs lack person agreement. Only the plain non-finite verb forms are part of the inflectional paradigm of the verb and thus formed by suffixes. Some of them are also used for the formation of analytic TAM forms. By contrast, verb forms functioning as specialized converbs mostly employ enclitics, which can also be added to other parts of speech than verbs. Conditional and concessive clauses have person agreement expressed by suffixes that strongly resemble the suffixes used in synthetic verb forms of main clauses. They are treated here as non-finite verb forms because their basic use is restricted to dependent clauses that cannot syntactically function as main clauses. All non-finite verb forms are normally negated by adding the negative prefix *a-*.

Conditional and concessive forms as well as specialized converbs and some of the plain non-finite verb forms occur in adverbial clauses (Chapter 25). Complement clauses (Chapter 24) are mostly headed by plain non-finite forms such as the infinitive and the masdar, and relative clauses are formed with participles (Chapter 23).

### **18.1 Plain non-finite verb forms**

The following verb forms are considered to be plain non-finite verb forms:


### 18 Non-finite verb forms

### **18.1.1 Simple converbs**

Like all Dargwa languages, Sanzhi has two simple converbs, an imperfective converb (§18.1.1.1) and a perfective converb (§18.1.1.2). For the syntax of adverbial clauses in which these converbs occur see Chapter 25.

### **18.1.1.1 The imperfective converb**

The imperfective converb has the suffix *-ul* (occasionally also *-ule*) or with a few verbs, which have -*un* as the suffix for the perfective converb, *-un(ne)*. For other Dargwa varieties *-ul* has been analyzed as diachronically composed of *-u* (the stem augment that is used for the habitual present) and a converb suffix *-l* (Sumbatova & Mutalov 2003; Belyaev In Preparation), and it is probable that this analysis can be applied to Sanzhi Dargwa as well. The suffix *-unne* is also diachronically complex, consisting of *-un-ne*. The second part -*ne* is an allomorph of the converb suffix -*le* that is also used for the formation of the perfective converb (§18.1.1.2) and, more generally, as an adverbializer (§9.6.3). However, synchronically the imperfective converb is not complex anymore and I will therefore gloss it as one single morpheme.

The imperfective converb can only be formed from imperfective stems and from stems of which the aspect is not specified. The functional range of the imperfective converb is as follows:

	- (1) žimiška sunflower.seeds d-ic-ij npl-sell.pfv-inf arg-ul=da go.ipfv-icvb=1 'I am going to sell sunflower seeds.'
	- (2) ala 2sg.gen xːunul woman ce what r-ik'-ul=e? f-say.ipfv-icvb=q saˁ-q'-unne=kːu=w? hither-go-icvb=neg=q 'What is your wife saying? Is she not coming?'

18.1 Plain non-finite verb forms


'Then Patima appeared outside running. (i.e. ran outside)'

There are plenty of examples in which it is not easy or even impossible to unambiguously identify the main clause to which the converbal clause belongs such that it may seem that the converb can head independent main clauses. In fact, such usages have been reported for Mehweb Dargwa (Kustova (2015); see also Mithun (2008) for a more general account of how and why non-finite verb forms develop into finite verb forms and the use of dependent clauses as independent sentences). However, in elicitation converb clauses are always judged as dependent clauses that need to be related to a superordinate clause in order to form a grammatical sentence. Thus, converb clauses that seem to occur on their own in independent utterances can probably be treated as a feature of colloquial language. In (6) only the first clause is morphosyntactically unambiguously an independent main clause followed by two expressions that indicate the lack of knowledge of the speaker (*ce ca-d=de=l, aχːu*) and two clauses with imperfective converbs without accompanying main clauses.

For instance, the utterance in (7) is part of a characterization of a person, but there is no preceding or following main clause that could serve as a syntactic anchor for the adverbial clause.

(6) ca one ca one t'ult'=de=w? bread=pst=q ce what ca-d=de=l, cop-npl=pst=prt aχːu, not.know [ču-la refl.pl-gen le-b-il exist-n-ref sa-b-iqː-ul] hither-n-carry.ipfv-icvb [lukː-unne] give.ipfv-icvb

' There was one bread for everyone? Whatever, I don't know, what they had they were bringing (to the soldiers) and giving it to them.'

### 18 Non-finite verb forms


### **18.1.1.2 The perfective converb**

The perfective converb is formed by adding the suffix *-le* to the preterite participle. The resulting complex suffixes are *-ib-le, -ub-le, -un-ne,* and *-ur-re* (or *-ur-le*), and their distribution is lexicalized (see the tables in §11.2 for many example verbs). The suffix -*le* is a cross-categorical suffix that forms adverbials from various parts of speech (§9.6.3). The perfective converb is mainly built from perfective verb stems. Thus, what is treated in this grammar under the label "perfective converb" is a conflation of several components that come with their own properties. It is therefore not ideal to gloss only the suffix -*le* as perfective converb (cvb) because this suffix only contributes to the morphosyntax, but not to the semantics. The meaning of the perfective converb originates from the combination of a (usually) perfective verb stem with the preterite suffix.

The functions of the perfective converb are:


18.1 Plain non-finite verb forms

(8) [razi happy ∅-iχ-ub-le], m-be.pfv-pret-cvb [b-uc-ib-le n-catch.pfv-pret-cvb ʡaˁt'a=ra frog=add ca-w=ra], refl-m=add ∅-uq'-ij m-go-inf b-ikː-ul n-want.ipfv-icvb ca-b cop-n qili home

' He got happy, caught his frog, and wants to go home.'


'We plastered the house, hung the wallpaper, put the windows; the boards were already prepared.'

	- (12) [ca-w refl-m urči-j horse-dat murtːa-l rider-advz ha-jʁ-ib-le] up-come.m.pfv-pret-cvb han remember le-w exist-m '( I) remember that he came riding on a horse.'

### **18.1.2 Participles**

Participles and functionally related forms occur in relative clauses and partially also in other constructions. Sanzhi has three participle: (i) the preterite participle (§18.1.2.1), (ii) the modal participle (§18.1.2.2), and (iii) the locative participle (§18.1.2.4). To the first two participles the cross-cateogorical suffixes -*ce* and -*il* can be added (§18.1.2.3). For the syntactic properties of relative clauses see Chapter 23.

### **18.1.2.1 The preterite participle**

The preterite is used for a range of verb forms with past time reference that are used in main clauses. This includes the preterite itself (§14.2.2) but also many more analytic

### 18 Non-finite verb forms

(§14.2) and periphrastic verb forms based on it (Chapter 15). The preterite is also employed in relative clauses (Chapter 23). Formally we deal with one and the same suffix, and I will therefore use only one single gloss for it (pret), although functionally and with respect to morphosyntactic properties the finite verb form 'preterite' differs from the participle. The finite verb form is used together with person enclitics, the past enclitic or the copula, which, by contrast, is impossible for the participle in a relative clause. The participle, in turn, attaches further nominalizing suffixes (16) and can then be case marked (§23.4). Relative clauses that are formed with the preterite participle obligatorily have a nominal head (13), (14).

(13) hex-tːi dem.up-pl [sa-r-ha-aq-ib] ante-abl-up-hang.pfv-pret paltar clothes ca-d cop-npl

'These are clothes that have been hung up there.'

(14) han remember d-irč-aq-ul npl-occur.ipfv-caus-icvb ca-d cop-npl cin-na refl.sg-gen [b-it-ag-ur], n-thither-go.pfv-pret [ag-ur] go.pfv-pret ʡuˁnru life ' He is remembering his gone by, past life.'

The preterite participle cannot directly take case suffixes or similar grammatical markers used with nominals (15). In order to nominalize the preterite participle, one of the cross-categorical participles *-il* (16) or *-ce* needs to be added (see §18.1.2.3 below for more details).


The preterite participle also attaches a number of temporal enclitics, suffixes and other subordinating enclitics such as *-er* 'when', *-la* 'since, after' (17), *=qːel(la)* 'when, because' and *=xːar* 'although' and is then used in adverbial clauses with various specialized converbs (§18.2).

(17) il that ag-ur-ra go.pfv-pret-post hitːi=ra after=add ʡaˁχuˁl-dex guest-nmlz taman end a-b-iχ-ub neg-b-happen.pfv-pret 'After that had happened their friendship did not finish.'

18.1 Plain non-finite verb forms

### **18.1.2.2 The modal participle** *-an*

The modal participle *-an* is only added to imperfective stems. Its semantics covers modality (obligation, deontic necessity) and future time reference. However, in relative clauses the modal meaning is often absent. The modal participle is used for the formation of a range of finite analytic verb forms, namely future (§14.1.3), future in the past (§14.1.4), obligative (§14.1.5), obligative present (§14.1.6), and obligative past (§14.1.7). The second functional domain of the modal participle is the formation of relative clauses. They mostly have habitual semantics and refer to stable properties of the referent of the head noun (18–23). Thus, there are some participles that have been lexicalized into adjectives expressing characteristic properties, e.g. *b-uz b-ik'ʷ-an barcːik'ʷ* (n-tear n-aux.ipfv-ptcp chudu) 'chudu filled with cheese that can be expanded and stretched when it is melted', *dircan* 'trader, seller' (21).


The relative clauses with the participle -*an* can be headless (21–23). The participle can take further case suffixes (23).


' For the one who wants to quarrel there is always an option.' (lit. 'a/the world')

When the modal participle is followed by the cross-categorical suffix *-ce* it also occurs in complement clauses (24) (§24.2.3), in addition to the possible use in headless relative clauses (§23.4). The second cross-categorical suffix *-il* also often co-occurs with the modal participle in various relative clauses.

18 Non-finite verb forms

(24) dam 1sg.dat b-alχ-a-d n-know.ipfv-hab.pst-1 [it that s-erʁ-an-ce] hither-come.ipfv-ptcp-dd.sg ' I know that he will come.' (E)

Finally, a number of temporal enclitics and other subordinating enclitics such as *-er* 'when' (25), *=qːel(la)* 'when, because' and *=xːar* 'although' attach to the modal participle yielding adverbial clauses (§18.2).

(25) c'il then di-la 1sg-gen sub=ra husband=add du=ra 1sg=add ag-ur-re go.pfv-pret-cvb ka-d-iž-ib-le, down-1/2pl-be.pfv-pret-cvb d-uk-an-er, 1/2pl-eat.ipfv-ptcp-when suk meet b-ič-ib hpl-occur.pfv-pret aʁʷal four admi person

'Then my husband and I were sitting and when we were eating we met four men.'

### **18.1.2.3 The cross-categorical suffixes** *-il* **and** *-ce***/***-te* **in combination with the participles**

The preterite and the modal participle can combine with both types of cross-categorical suffixes, *-il* and *-ce* (*-te* in the plural). The general function of these suffixes can be described as the formation of referential attributes or definite descriptions that have the morphosyntactic properties of nominals (§9.6.1 and §9.6.2). When the suffixes are added to the participles we can form relative clause with heads and headless relative clauses. Two participles and two types of cross-categorical suffixes yield four possible combinations that are not all equally common. There seem to be no semantic differences between the two cross-categorical suffixes when occurring in headless relative clauses. But there is a morphosyntactic difference: the suffix *-il* is only used with referents that are not morphologically overtly marked for plural, i.e., the relative clause needs to refer to a singular object or a mass noun such as *χalq'* 'people' or *sungul* 'the community of the Sanzhi people' (even though both nouns control human plural agreement) or something similar as in (28). The referent can be overtly expressed (relative clause with a head) or not (headless relative clause). For overtly marked plural referents or for headless relative clauses denoting a plurality of referents only *-te* can be used (26–28).


18.1 Plain non-finite verb forms

(28) il-tːi that-pl [di-cːella 1sg-comit b-alli hpl-together b-ax-an-te] hpl-go-ptcp-dd.pl b-abq'i hpl-half 'the half (of the people) who were going with me'

By far most common in the Sanzhi corpus are headless relative clauses in which the verb bears the preterite participle suffix plus the suffix *-il* (29). When the dative case is used, the meaning of the nominalized relative clause can be causal (due to the semantics of the dative case) such that these clauses rather function as adverbial clauses expressing cause or reason (30).

(29) [b-učː-ul hpl-drink.ipfv-icvb ka-b-iž-ib-il] down-hpl-be.pfv-pret-ref ka-b-išː-ib=da down-n-put.pfv-pret=1 heltːu there čina=del where=indef

'The one where they are sitting and drinking, I put this (picture) somewhere.'

(30) hej this pašman sad ∅-iχ-ub m-be.pfv-pret ca-w cop-m ʡaˁħ-le good-advz hel that b-arq'-ib-il-li-j n-do.pfv-pret-ref-obl-dat ' He got very (lit. well) sad, because of what he had done.'

The preterite participle of the verb 'say' to which *-il* is added and which is used without the spatial preverb *ha*- (i.e.*ʔ-ib-il*, also written as*ibil*<sup>1</sup> ) functions as marker for ordinal numerals (31) (§6.2).

(31) xu-c'anu five-ten urč'em-ra nine-num ibil ord dusːi-cːe-w year.obl-in-m 'in the year (19)59'

The co-occurrence of both participles (preterite and modal) with *-ce*/*-te* is also common and it is easy to find examples with (32) and without case suffixes (33). There is a clear difference in meaning due to the participles and the aspectual properties of the verb stems. In (32) the preterite participle is used to refer to people who accomplished an action in the past (i.e. they gathered). The modal participle in (33) refers to the Sanzhi people by means of a stative characterization as the ones who eat three times at night and stay hungry during the day, i.e., fulfilling the duties of Muslims during the month of Ramadan.

(32) d-erč-ib-t-a-l npl-collect.pfv-pret-pl-obl-erg ču-j refl.pl-dat d-ut'-ib npl-divide-pret d-urkː-ar npl-find.ipfv-prs

' The ones that gathered (the food) probably divided it among themselves.'

<sup>1</sup>The usual participial form of this verb is *haʔ-ib-il* with the spatial preverb.

### 18 Non-finite verb forms

(33) "abdal fool sungul, Sanzhi.people baˁħ crazy sungul, Sanzhi.people dučːe at.night ʡaˁj-na three-time b-uk-an-te, hpl-eat.ipfv-ptcp-dd.pl ari during.day kːuš-le hungry-adv b-ug-an-te" hpl-stay.ipfv-ptcp-dd.pl b-ik'-ul hpl-say.ipfv-icvb '" Foolish Sanzhi people, crazy Sаnzhi people, at night they eat three times, and during day they stay hungry," they said.'

The combination of the modal participle with *-il* is not particularly frequent; it occurs mostly together with case suffixes as in (34).

(34) hana now cek'u whatchamacallit r-irq'-an-il-li-j f-do.ipfv-ptcp-ref-obl-dat xujal five azir thousand xu-c'al five-ten azir thousand

' now for the one who operates her 5000, 50000 rubles (need to be given)'

Finally, the suffix *-il* can also be added to the existential copulas, which do not inflect for any of the participles, in order to form headed and headless relative clauses (6). In (35) the existential copula with its suffix -*il* is inflected for a spatial case.

(35) durqa-ce, dear-dd.sg u 2sg le-r-il-le exist-f-ref-loc taliħ-la happiness-gen marka rain b-arq'-ab n-do.pfv-opt.3 '( My) dear, may onto the place where you (fem.) are rain the rain of happiness.'

### **18.1.2.4 The locative participle** *-na*

The locative participle has the suffix *-an*. It is only available for imperfective verb stems. It has a spatial meaning that corresponds to the semantics of the lative case (§3.4.2). Similar to spatial adverbs it can take further spatial case suffixes, i.e. the essive and the ablative. It most commonly functions as the head of spatial relative clauses (36), (37), but it is also possible to add a head noun (38).


18.1 Plain non-finite verb forms

(38) durħ-ne boy-pl ag-ur go.pfv-pret [q'ʷal cow luχ-na] cut.ipfv-ptcp.loc musːa place.loc

' The boys went to the place where cows are slaughtered.' (this refers to a specific place in Sanzhi)

In addition, the locative participle can be fully case marked. In order to add case suffixes (other than the suffixes for the essive and the ablative) the participle appears in its oblique forms, just like any other nominal. After suffixing the oblique marker *-l* (which is identical to the ergative), case suffixes follow (39). But as (37) and the second variant in (39) show, it is also allowed to directly suffix markers that express the spatial cases essive, lative and ablative, because the locative participle has inherent spatial meaning. In elicitation, the suffixation of other than spatial cases leads to a broader variety of relative clauses (40). The case-marked participle also occurs in adverbial clauses with causative semantics (103).

(39) [Napisat Napisat ka-r-ils-na-l-le-b down-f-sleep.ipfv-ptcp.loc-obl-loc-n / / ka-r-ils-na-b] down-f-sleep.ipfv-ptcp.loc-n te-b=de exist.away-n=pst maˁlʡuˁn snake ' Where Napisat was sleeping there was a snake.' (E) (40) [Maˁʡaˁ ka-r-ils-na-l-la] χabar le-b=de

Maja down-f-sleep.ipfv-ptcp.loc-obl-gen story exist-n=pst ' There was the gossip that Maja was lying (in the hospital).' (E)

### **18.1.3 The infinitive**

The suffix for the infinitive is *-ij*. It is very likely that the final *j* diachronically goes back to the dative case, and the formal identity of infinite and dative case is also attested in other Dargwa languages (e.g. Tanti, Icari) and other East Caucasian languages such as Hinuq. In principle, the infinitive can be formed from imperfective and perfective stems (41) but in natural texts it is almost exclusively used with perfective stems. There are a number of imperfective verbs for which the infinitive is at least very marginal if not ungrammatical (see the last two examples in (41)).


### 18 Non-finite verb forms

The functions of the bare infinitive are:

	- (42) hel that sa-ka-b-iršː-id hither-down-n-put.ipfv-1.prs b-erkʷ-ij n-eat.pfv-inf ' We put it (on the table) to eat.'
	- (43) [hak' shake d-arq'-ij] npl-do.pfv-inf a-r-iχ-ub=da neg-f-be.able.pfv-pret=1 ' I (fem.) was not able to move (lit. shake) (my legs).'
	- (44) le-b=q'al exist-n=prt [dam 1sg.dat [b-urs-ij] n-say.pfv-inf b-ikː-an] n-want.ipfv-ptcp 'There exists what I want to say.'

In addition, the infinitive can take a number of suffixes and enclitics:

	- (45) na now d-erčː-ij-te npl-drink.pfv-inf-dd.pl d-erkʷ-ij-te npl-eat.pfv-inf-dd.pl li<d>il all<pl> cik'al something le-d exist-npl ' There is something to eat and to drink.'

(46) it-i-j that-obl-dat aχːu not.know han remember b-el n-remain.pfv akːu cop.neg [ce what b-ik'ʷ-ij=al] n-say.ipfv-inf=indq 'I do not remember how to call it.'

18.1 Plain non-finite verb forms

	- (48) "u-l 2sg-erg b-urs-a!" n-tell-imp haʔ-ib=da say.pfv-pret=1 qːant'-le short-advz caj-na one-time cara other daˁʡaˁn-ne secret-advz [Q'urban-ni Kurban-erg t'am sound a-d-aq'-ij-li-j neg-npl-hear.pfv-inf-obl-dat heχ-itːe] dem.down-advz '" You tell him!" I said briefly once more such that Kurban could not hear it.'

### **18.1.4 The subjunctive (i.e. agreeing infinitive)**

Sanzhi Dargwa has another verb form that is functionally equivalent to the infinitive, but shows person agreement, and will be called "subjunctive" in this grammar. Person agreement of the subjunctive is reduced in comparison to other verb forms such as the habitual present or the habitual past. There is no suffix for the first person and instead the normal infinitive is used. The suffixes for the second and the third person, which do not distinguish number, are displayed in Table 18.1. The second person makes use of the same stem augment vowels (*i*, *u*) that occur with other verb forms such as the habitual present or the prohibitive.<sup>2</sup> Diachronically, the suffixes *-itːaj*/*-utːaj* consist of the stem augment, followed by a second person suffix -*tː*, and the last part -*aj*, which represents the actual subjunctive marker. This becomes clear when we compare the subjunctive to other verb forms (§20.3). However, for the sake of understanding I treat the suffixes as units and use a single gloss.



<sup>2</sup>The second person subjunctive suffixes are identical in form with variants of the plural prohibitive.

### 18 Non-finite verb forms

There is a strong correlation between the stem augment vowel and transitivity, i.e. intransitive verbs mostly take *-u* (49) and transitive and affective verbs usually take *-i* (50). Thus, the subjunctive behaves as other verb forms that have person agreement suffixes (§20.3).

(49) a. *r-isː-utːaj* 'cry'

	- b. *či-b-až-itːaj* (pfv) 'see'
	- c. *r-ič-itːaj* (pfv) 'bring, lead'

In the third person, the suffixes *-ar-aj* and *-an-aj* are used, which diachronically consist of *-ar* (homophone to one allomorph of the the habitual present and the realis conditional), or*-an* (homophone to the modal participle) and *-aj*. The choice between *-araj* and *-anaj* is mostly lexicalized (51), (52) but there are a few verbs to which in elicitation both suffixes can be attached, e.g. *či-ha-b-uq-anaj*/*či-ha-b-uq-araj* (spr-up-n-go.pfv-subj.3) 'climb'. In general, *-anaj* is more common both in terms of types (i.e. verb stems to which the suffix is added) as well as in terms of token frequency in my corpus.

	- b. *b-arq'-araj* (pfv) 'do'
	- c. *b-iχʷ-araj* (pfv) 'be, become, can'
	- d. *kaxʷ-araj* (pfv) 'kill'
	- b. *b-elk'-anaj* (pfv) 'write'
	- c. *b-uk-anaj* (ipfv) 'eat'
	- d. *ha-b-irq'-anaj* (ipfv) 'support, bring up, make high'

The subjunctive, just like the infinitive, is mainly obtained from perfective stems, though a number of imperfective stems can also be inflected for it (53).


18.1 Plain non-finite verb forms

The functions of the subjunctive are identical to the functions of the normal infinitive and it is always possible to replace the subjunctive with the infinitive. Thus, the subjunctive heads purpose and complement clauses:

(54) c'il then uškul-la school-gen hextːu-b, there.up-n musːa=ra place=add k'e-b exist.up-n b-iχʷ-ar n-be.pfv-cond.3 ka-d-isː-utːaj down-1/2.pl-sleep.pfv-subj.2

' Then there at the school, there is a place, if it still exists, for you to sleep.'

(55) d-uq'-ij 1/2.pl-go.pfv-inf xːun-be way-pl wahi-l bad-advz d-určː-i=q'al npl-be.ipfv-hab.pst=mod ixtːu there.up d-uq'-aˁtːaj 1/2.pl-go.pfv-subj.2

' The roads to go there were probably bad, for you to go.'

(56) hel that rucːi-l sister-erg r-aʔ f-begin r-išː-ib f-become.pfv-pret ca-r cop-f b-ukː-un-ne n-eat.ipfv-icvb kːalkːi tree ka-b-ik-araj down-n-occur.pfv-subj.3

' The sister started to eat up the tree so it would come down.'

As with the infinitive it is possible to suffix the cross-categorical suffix *-ce* to the subjunctive (57). In this example, *b-arq'-araj-ce* could be replaced by *b-arq'-ij-ce* and the meaning would not change.

(57) uškul school b-arq'-araj-ce n-do.pfv-subj.3-dd.sg balnicːa hospital b-arq'-ib-le n-do.pfv-pret-cvb b-iχʷ-ardel n-be.pfv-cond.pst ' It would be better to build a hospital instead of building a school.' (E)

Furthermore, subordinating enclitics for the formation of adverbial clauses can be attached, in particular *=sat*/*=satːin*/*=satːinna* 'until' (58), *=sar* 'before, until' and *bahandan* 'because of' (§18.2):

(58) k'ʷah silent ∅-ič-ib-le, m-occur.pfv-pret-cvb ʁera dusk ag-araj=sat go.pfv-subj.3=until a-b-urs-ib neg-n-tell.pfv-pret

' He remained silent, until it became dark he did not tell.'

It seems that the subjunctive, which is absent from the more innovative north Dargwa varieties (e.g. from Akusha/Standard Dargwa), is gradually disappearing from south Dargwa varieties. In Icari, it lacks a cell in the transitive paradigm that is replaced with *-ij* (which is not the Icari infinitive, but another suffix.). In Sanzhi, it entirely lacks first person forms. Verb forms similar to the Sanzhi subjunctive are found in other south Dargwa varieties such as Qunqi, and Xuduc, but, e.g., not in Tanti (Sumbatova & Lander 2014: 136). Sumbatova & Mutalov (2003: 107) write that the Icari subjunctive (-*aj*/ -*j*), which is clearly cognate with the Sanzhi subjunctive, is historically and structurally related to the Standard Dargwa infinitive -*es*.

### 18 Non-finite verb forms

### **18.1.5 The masdar**

The suffix for the masdar is *-ni*. It has an allomorph -*ri*, which is only used with a handful of verbs such as the negative form of the copula (*(b-)akʷ-ni* and *akːʷ-ri*). The masdar is available for imperfective and perfective stems as well as for the copula (*ca-b-ni*), and the locative copulas, e.g. *le-b-ni*. The functions of the masdar are:

	- (59) il this at=cun 2sg.dat=only dalaj song w-ik'-ud, m-say.ipfv-1.prs [w-arc-ni=ra] m-get.tired.pfv-msd=add a-b-alχ-ul neg-n-know.ipfv-icvb 'Only for you I sing this song, not knowing whether you got tired.'

If the masdar is formed from a stem with imperfective aspect, the temporal reference of the complement clause is non-past (60), and if it is formed from a stem with perfective aspect the temporal reference is past time (61).

	- (62) mig-le-r ice-loc-f qus slip r-ik'ʷ-ni f-say.ipfv-msd b-irχʷ-i n-be.ipfv-hab.pst ' There was skating on the ice.' (referring to a woman skating)
	- (63) "iš-tːi this-pl at 2sg.dat ca-d, cop-npl di-la 1sg-gen juldaš, friend du 1sg b-erc-aq-ni-li-j" n-save.pfv-caus-msd-obl-dat b-ik'-ul n-say.ipfv-icvb ca-b cop-n ' These (gems) are for you, my friend, because you saved me, said (the bear).'

18.2 Specialized converbs

(64) čina=k'u where=emph ʡaˁlħaˁm-le condolence-loc d-uˁq'-ni-la 1/2pl-go-msd-gen χabar story b-urs-idel n-tell-modq aχːu not.know

' I don't know if I should tell you the story about how we went somewhere, to the condolences.'

(65) značit, thus ce what d-arq'-ij npl-do.pfv-inf uˁq'-ni-lla go.m-msd-gen ʡaˁči=q'al work=mod it that 'What to do, this pasturing (lit. the work of going after the animals) is work.'

### **18.2 Specialized converbs**

Constructions with special converbs occur in adverbial clauses that express temporal or causal relationships. Formally, they mostly consist of an enclitic or suffix bearing the specific temporal/causal meaning or of a postposition/adverb used to express temporal relationships. The converbial marker is attached either to the preterite participle and the modal participle or to the subjunctive and the infinitive (or follows the respective verb forms). Other ways of obtaining specialized adverbial clauses involve the locative participle and the noun *zamana* 'time'. Sanzhi possesses the specialized converbs that are given in (66). There are two enclitics among them (=*qːel(la)* and =*sat*/=*satːin*/=*satːinna*), which are not only used with verbs in order to form adverbial clauses, but also with nominals. Their use with nominals is described in §9.3. The syntax of adverbial clauses is analyzed in Chapter 25.


### 18 Non-finite verb forms

### **18.2.1 The temporal/causal enclitic** *=qːel(la)* **'when, while, because'**

The temporal *=qːel(la)*, which exists in a more commonly used short form and in a less frequently occurring long form, translates as 'when, while, because'. As the other two temporal enclitics it can be hosted by verbs and other parts of speech (§9.3). When used to form adverbial clauses it is added to the preterite participle (18.1.2.1) or to the modal participle (18.1.2.2), and also to forms of the negative copula (usually the participle, but also other forms) (69). It expresses the temporal simultaneity (67), and rarely the temporal anteriority (68) of the situation referred to in the adverbial clause with respect to the situation described in the main clause. In many cases, this also implies a causal link between the two events (69).

(67) či-b-a up-hpl-dir arg-an=qːella go.ipfv-ptcp=when debʁul-m-a-l prayer-pl-obl-erg b-irq'-ul hpl-do.ipfv-icvb b-už-ib-le=de hpl-be-pret-cvb=pst

' When they went up there they were praying.'


When the enclitic is hosted by nominals (nouns, demonstrative pronouns, adjectives) it also means 'when'. Examples are provided in §9.3.

### **18.2.2 The temporal marker** *-er* **'when, as'**

The temporal meaning of simultaneity is also expressed by *-er*, which is, just like *=qːel(la)*, added to the preterite participle or to the modal participle (70–72). This suffix does not imply any causal relationships. In fact, in some examples there is no relationship whatsoever between the situations expressed in the two clauses (72). The suffix is only added to verbs, never to nominals or other parts of speech.

(70) Družba-le Druzhba-loc ka-jʁ-ib-er, down-come.m.pfv-pret-when Družba-le Druzhba-loc sowχoz-la sovkhoz-gen ʡaˁči work b-irq'-ul=de n-do.ipfv-icvb=pst 'When I moved to Druzhba, I worked in the sovkhoz in Druzhba.'

18.2 Specialized converbs

(71) heχ dem.down b-erčː-ib-le n-drink.pfv-pret-cvb haˁ-q'-aˁn-er up-go-ptcp-when cinna pause.filler tːartːar shake uq-un-ne go.m.pfv-pret-cvb heχ dem.down

' When he was drunk and went there, he staggered.'

(72) u 2sg ∅-ik'ʷ-an-er, m-say.ipfv-ptcp-when šːatːir walk arg-ul go.ipfv-icvb ca-b cop-hpl ' As you (masc.) say, they are walking around.'

### **18.2.3 The temporal enclitic** *=sat***/***=satːin***/***=satːinna* **'until, before as much/long as'**

This enclitic, which also belongs to the category of enclitics that can be used with nominals (§9.3), occurs in three variants of different lengths that can be ordered according to their increasing frequency as *=satːin* < *=satːinna* < *=sat*. It is a cognate of the spatial/temporal adverb *satːi* 'at/along the front, as soon as', which, however, occurs before verbs rather than following them. The adverb can be further decomposed into the postposition *sa* 'in front, ago' and the adverbializer *-tːi*, which is part of a few spatial adverbials (§7.1.2).

The enclitic follows the infinitive and the subjunctive and expresses the meaning of temporal posteriority of the situation denoted by the adverbial clause with respect to the situation referred to in the main clause, i.e. 'before, until':

	- 'Until you take your stick out, you have to stand there.'

When the enclitic is attached to the modal participle the meaning is 'as much as, as long as' (75). Two more examples are given in §9.3.

(75) xːunul-li-j woman-obl-dat r-irχʷ-an=satːin f-be.able.ipfv-ptcp=as.much jaˁħ patience b-irq'-ul=de n-do.ipfv-icvb=pst ' The woman was patient as long as she could.'

The same meaning is also attested when the enclitic follows nominals (35), (36). Finally, the enclitic can be attached to demonstrative pronouns forming manner demonstrative pronouns that are used in comparison 'like this, like that, such' (37).

### 18 Non-finite verb forms

### **18.2.4 The temporal adverb/postposition** *sar(ka)* **'until, before'**

The temporal adverb/postposition *sar(ka)* 'before, until' (which also has the spatial meaning 'in front', §8.1.2 and §7.1.2) follows the infinitive or the subjunctive. The resulting meaning is purely one of temporal posteriority, i.e. 'before, until' (76) and occasionally more like an apprehensive (77).


### **18.2.5 The temporal/causal postposition** *(h)itːi* **'after, because'**

The postposition/adverb *hitːi* 'after, behind' (see §8.1.4 for the use in postpositional phrases) can occur in temporal adverbial clauses with the meaning 'when, after'. In the majority of occurrences the postposition is shortened to its enclitic variant *=itːi* and directly attached to the preterite participle (78). However, it is always possible to replace the enclitic with the full form *hitːi*.

(78) c'il then helka from.there ag-ur=itːi go.pfv-pret=after c'il then a-ka-r-ač'-ib=da neg-down-f-come.pfv-pret=1 ' After we left from there, I did not return.'

Occasionally, the temporal relationship also implies a causal relation between the situation referred to in the adverbial clause and the situation expressed in the main clause:

(79) b-erčː-ib=itːi n-drink.pfv-pret=after urk'i heart b-at n-free b-uq-un-ne, n-go.pfv-pret-cvb …

'After drinking the heart opened, (and he molested his wife).'

Instead of using the bare preterite participle it is also possible to suffix another marker *-la* to the participle, which is, in turn, followed by the postposition (80), (81). The suffix -*la* goes most probably back to the genitive case suffix since *hitːi* governs the genitive. The suffix -*la* undergoes assimilation after the sonorants /n/ and /r/ (> -*na*, -*ra*). It can also be employed on its own without the following postposition *hitːi* (§18.2.6).

(80) il-tːi that-pl daˁwi war taman end d-iχ-ub-la npl-be.pfv-pret-post hitːi after xadi married ka-r-iž-ib down-f-sit.pfv-pret hel that

'After the times of the war finished he married her.'

18.2 Specialized converbs

(81) deč-la drinking-gen učː-ul drink.m.ipfv-icvb w-aʔ-ač'-ib-la m-begin-come.pfv-pret-post hitːi, after … 'after he started drinking, …'

In elicitation it is possible to place the postposition *hitːi* after the modal participle -*an* to which again *-la* (in its assimilated allomorphic form *-na*) can be suffixed (82). In this construction, however, the meaning diverges and instead of the sequential meaning we get a causal/conditional meaning. Now the clause containing the postposition expresses a condition or cause for the situation that the main clause denotes.

	- (E)

### **18.2.6 The temporal marker** *-la* **'since, after'**

This marker is suffixed to the preterite participle. As mentioned in §18.2.5, it goes back to the genitive and the construction most probably arose as a simplified variant of the use of the same marker followed by the postposition *hitːi*. The meaning of both constructions is very similar expressing temporal posteriortity of the situation that the adverbial clause refers to with respect to a second situation that is normally expressed by the main clause. However, the adverbial clauses that contain only *-la* without a following postposition mean 'since then, ever since, from then on, after that'.

(84) Sanži-b Sanzhi-n b-akːʷ-i, n-cop.neg-hab.pst hana now ca-b cop-n il that nušːa 1pl ka-d-eʁ-ib-la down-1/2pl-go.pfv-pret-post

' In Sanzhi there was no (such plant), now there is, since we moved here.'

The postposition can always be added and this results in a slight change of the meaning. In (85) the speaker uses first the construction with *-la* and then the construction containing the encliticized postposition, but lacking *-la*. Both clauses have similar, but not completely identical, semantics.

(85) hej this sud trial b-arq'-ib-la, n-do.pfv-pret-post sud trial b-arq'-ib=itːi, n-do.pfv-pret=after hej this ka-jž-ib down-remain.m.pfv-pret ca-w cop-m uže already tusnaq-le prison-loc hež this 'Since the trial, after the trial was made, he is already sitting in prison.'

### 18 Non-finite verb forms

To sum up, in certain contexts the three options (only *-la*, only *=itːi*, or *-la* + *hitːi*) have very similar or even identical meaning ('after, when'). In other contexts when *-la* is used alone it means rather 'since'.

### **18.2.7 The periphrastic adverbial construction with** *b-el-le* **'while, as long as, as soon as, until, when'**

The defective verb *b-el* 'remain, stay' when inflected as perfective converb heads periphrastic adverbial clauses. The same type of periphrastic verb form is attested in independent clauses (§15.3), but the use in dependent clauses is far more common. The verb can occur together with a lexical verb that bears either the imperfective or the perfective converb suffix. It does not assign case to any arguments and therefore shows the same gender/number agreement as the lexical verbs (86). However, it can also be used in the invariant form *b-el-le* (89) in which gender agreement is lost in favor of the petrified prefix *n-* (neuter singular).

When *b-el* co-occurs with a lexical verb inflected for the imperfective converb the meaning of the adverbial clause is 'while, until, as long as' (86). In combination with the past perfect in the main clause the adverbial clause refers to a situation that obtained at a reference point in the past (= the situation expressed in the main clause) and continued to a later point in time at which the other situation had finished (87).

(86) a-w-čː-ul neg-m-drink.ipfv-icvb w-el-le m-remain-cvb […] li<d>il all<npl> iš-tː-a-la this-pl-obl-gen ʡaˁħ-le good-advz ca-d cop-npl

' While until/as long as he is not drinking, everything is good for them.'

(87) er=či look=on ∅-ik'-ul m-look.at.ipfv-icvb w-el-le m-remain-cvb il-i-la that-obl-gen qːuʁa-dex-li-j, beautiful-nmlz-obl-dat iltːi these žaniwar-te animal-pl li<d>il all<npl> ag-ur-re=de go.pfv-pret-cvb=pst 'While/as long as he was looking at its beauty, all animals had already left.'

If the lexical verb appears in the form of the perfective converb, the adverbial clause expresses immediate anteriority that can be translated with 'as soon as, immediately when'. It refers to the point in time when an event is completed or was completed or to the moment when a state obtains or obtained rather than to an enduring situation. The relevant state or event immediately precedes the situation denoted by the main clause.


18.2 Specialized converbs

(89) helka from.there tːura outside ka-w-q-un-ne down-m-go.pfv-pret-cvb b-el-le, n-remain-cvb murtːa rider ∅-iž-ib-le m-be.pfv-pret-cvb urči-li-j horse-obl-dat […] gu-r-ag-ur-il sub-abl-go.pfv-pret-ref ca-w cop-m 'When/as soon as he went out from there, he mounted the horse, [singing a song he pretended to be drunk] and left.' (90) Aminat-li t'ult'-e d-arq'-ib-le b-el-le, nišːi-j k'ʷel

Aminat-erg bread-pl npl-do.pfv-pret-cvb n-remain-cvb 1pl-dat two s-aqː-a! hither-carry.pfv-imp

' As soon as Aminat makes bread/finishes making bread, bring us two (loaves of bread)!' (E)

### **18.2.8 The concessive enclitic** *=xːar(e)* **'although, even if'**

The concessive enclitic *=xːar(e)* is attached to the preterite participle (91), (92), to the modal participle (93), (94), or to the participial form of the negated copula, and expresses concession 'even if, although, though, even though'. In the main clause following a concessive clause optionally the particle *ja* can occur (92), which is probably a cognate of the disjunctive particle *ja*.

(91) q'ʷila a.little b-alχ-an, n-know.ipfv-ptcp itwaj like.this gramatni educated ∅-irχʷ-iri, m-be.ipfv-hab.pst a-b-elč'-un=xːar neg-n-learn.pfv-pret=conc

'Knowledgeable, he was educated even if he did not study.'


' Though here they give (food) for debts, there they do not give him (food) for debts.'

Note that the last example in (94) is an independent clause with the modal participle -*an* suffixed to the verbal head of the main clause. Because of the use of the modal participle the sentence has a habitual meaning without any specific temporal reference. If the suffix *-ne* were added, we would obtain future tense (§14.1.3) with a modal meaning and temporal reference to future events.

### 18 Non-finite verb forms

Frequently concessive clauses are copula constructions without a copula item ('although X is Y'), in which case the temporal reference of the concessive clause depends on the main clause. For instance, in (95) the main clause refers to the past and therefore the concessive clause also refers to a past event even though it does not contain any morpheme expressing temporal reference (i.e. no preterite or modal participle). The host of the enclitic in such concessive phrases is the copula predicate, which can for instance be an adjective (95), and adverbial, (96), or a noun (97).


though it was a small house.'

There is another way of formulating concessive clauses in Sanzhi, namely the use of conditional forms to which the additive is encliticized (§18.3.6).

### **18.2.9 Constructions with** *zamana* **'time'**

The noun *zamana* 'time', ultimately an Arabic loan word, is used in temporal adverbial clauses that are relative clauses from a syntactic point of view (Chapter 23). The noun *zamana* is the head, and the relative clause contains a verb in the form of the modal or the preterite participle. Clauses with the modal participle refer to events that were ongoing during a reference point in time or a reference period, which is expressed in the main clause ('while, when') (98), (99).

(98) Uc'ari-r Icari-abl haˁ-q'-aˁn up-go-ptcp zamana, time lampučka torch ca-w=ra refl-m=add dučːi night ha-aš-i=q'al up-go-hab.pst=mod qili home ʡaˁči-le-r work-loc-abl

' While he came from Icari, he went with a torch at night home from work.'

(99) ʡaˁħ-ʡaˁħ-le good-good-advz gulik'-an listen.ipfv-ptcp zamana, time t'am sound b-aq'-ib-le n-hear.pfv-pret-cvb ca-b cop-n ʡaˁt'-n-a-lla frog-pl-obl-gen t'ama sound

' When he was attentively listening he heard the sound of frogs.'

18.2 Specialized converbs

Clauses with the preterite participle have a comparable meaning of simultaneity:

(100) nuˁq-be arm-pl sa-ka-d-uc-ib ante-down-npl-keep.pfv-pret zamana, time žaˁpar-ra Zhapar-gen xːunul woman qili-rka home-abl tːura outside sa-r-uq-un hither-f-go.pfv-pret ca-r cop-f ' While he presented the condolences (lit. kept his hands in front), Zhapar's wife entered.'

(101) hel that gu-r-ha-jcː-ur down-abl-up-stand.m.pfv-pret zamana, time di-la 1sg-gen ul-be eye-pl het that kaš-le-d porridge-loc-npl kelg-un remain.pfv-pret ' While I got up, my eyes remained on the porridge.'

The *zamana*-construction can be combined with the particle *bah* 'immediately when' that occurs in the initial position of the relative clause (102). The precise origin of *bah* needs further investigation, but we might suggest that it is related to the superlative particle *bah* 'most' and to the adverbs *bahsala, bahsar* 'first', which can be decomposed into *bah-* and a following postposition. The particle can also co-occur with the enclitic *=qːel(la)*.


### **18.2.10 Minor ways of forming adverbial clauses**

The locative participle and the masdar can occur in adverbial clauses expressing causes when they take the dative suffix (103), (104), because the expression of causes is one of the functions of the dative (§3.4.1.4):


### 18 Non-finite verb forms

When the postposition *bahanne/bahandan* 'because of' (§8.2.3) follows the masdar, the resulting clause also expresses causation (105). By contrast, when it follows the infinitive or the subjunctive we get purpose clauses (106):


### **18.3 Conditional and concessive verb forms and clauses**

Conditional and concessive clauses are adverbial clauses that contain specialized verb forms expressing realis and irrealis conditional and concessive meaning. All verb forms here are obtained by means of suffixes that are added to a stem augmentation vowel. The vowel is the same that is used for synthetic verb forms (Chapter 13) and some nondeclarative verb forms (Chapter 17) and not separately glossed in the examples. The suffixes express conditional meaning and person agreement and bear resemblance to the suffixes of the synthetic tenses. The following forms are treated in this section:


All conditional forms head dependent clauses, thus they are normally followed by a main clause. The conditional suffixes alone suffice to convey conditional meaning, but optionally the conjunction *raχle* 'if' can co-occur in conditional clauses (107). However, the use of the subordinating conjunction is rare.

(107) raχle if uc-arre catch.m.pfv-cond.3 het that k-ercː-an down-stand.ipfv-ptcp ca-w cop-m heštːu here 'If (he) caught him, he must stand there.'

For more information on the general syntactic properties of adverbial clauses see Chapter 25.

18.3 Conditional and concessive verb forms and clauses

### **18.3.1 Realis conditional**

The realis conditional is formed from perfective verb stems (for those verbs that occur in pairs of imperfective and perfective stems). To the verbal stem the stem augments (vowels *-u* or *-i*) are added, followed by the conditional suffixes (Table 18.2). In the second person singular, there are two variants possible, *-tːe* and *-tːel*, but the first is clearly preferred. In the third person, there is again largely lexically determined allomorphy between the suffixes *-an* and *-ar*. The latter suffix has a longer variant *-arre*, but the shorter variant is more common. In negative realis conditional clauses the verb bears the negative prefix *a-* (110).

Table 18.2: The realis conditional


Table 18.3: Some illustrative paradigms of the realis conditional


The function of the conditional is the expression of real non-past conditions:


'" If you (= masc.) do not calm down, I make you calm," he says.'

### 18 Non-finite verb forms

It also occurs in utterances in which the conditional is not a condition for the apodosis because there is no conditional connection between the two clauses. This includes the common idiomatic expression 'to be honest' (lit. 'if I tell correctly') (112).


As with the two indicative analytic verb forms, the habitual present and the habitual past, in conditional clauses ergative alignment is, in addition to the dative construction, possible with some affective verbs.

### **18.3.2 Past conditional**

The past conditional bears strong formal resemblances to the realis conditional and the second person is almost identical for both conditional forms (Table 18.4). Before the conditional past suffixes the stem augment vowels occur that are the same as for the realis conditional and for a number of other verb forms such as the subjunctive. In the third person, the first part of the two allomorphic suffixes *-ar-del* and *-an-del* is identical with the suffixes used in the realis conditional (Table 18.2). The second part probably originates from the past enclitic *=de*. Negation is marked with the prefix *a-*. Only perfective verb stems can function as the basis for the past conditional.


Table 18.4: The past conditional

The semantic range of the past conditional comprises the expression of realis conditions that were obtained in the past.

…

(113) raχle if q'an late d-iχ-utːel, 1/2pl-be.pfv-cond.1 'If we were too late, …'

18.3 Conditional and concessive verb forms and clauses


Furthermore, it conveys irrealis conditional meanings, i.e. conditions with low probability and counterfactual conditions (116), and those sentences can lack the apodosis (117), (118). The apodosis of past conditional clauses often contains a verb marked for future in the past (§14.1.4) or habitual past (§13.2) (114).


### **18.3.3 Imperfective realis conditional**

The imperfective realis conditional is formed from imperfective verb stems (for those verbs that occur in pairs of imperfective and perfective stems) by means of the suffix *-aχː*, followed by the vowel *-a* that functions as a stem augment without expressing transitivity, and finally by (almost) the same person suffixes that are used for the realis conditional (§18.3.1). As with all conditional forms treated in this Section, negation is marked by means of *a-* (119). Some affective verbs can occur in the dative experiencer construction and in the ergative construction when inflected for the imperfective realis conditional.

### 18 Non-finite verb forms

Table 18.5: The imperfective realis conditional


Table 18.6: Some illustrative paradigms of the imperfective realis conditional


The imperfective realis conditional is basically the imperfective counterpart of the realis conditional. According to Sanzhi speakers, it covers the same meanings, with the only difference being the aspectual value that the stem carries. Thus, we have realis conditional semantics with present and future time reference (119), (120) and occassionally in utterance in which no genuine conditional semantics is expressed (121).

(119) "hej this ha-r-iq'-ij up-f-bring.up-inf a-r-irχʷ-aχː-alle," neg-f-be.able.ipfv-cond-cond.1 r-ik'ʷ-ar, f-say.ipfv-prs "dam 1sg.dat ʡuˁmru ħaˁžat-le=kːu"

life need-advz=neg

'"If I cannot educate (i.e. bring up) her (myself)," she said, "then life is of no need for me."'

(120) hel-tːi that-pl ha-qː-ij up-carry.pfv-inf a-r-irχʷ-aχː-an neg-f-be.able.ipfv-cond-cond.3 il that ce what r-irq'-an=e f-do.ipfv-ptcp=q dam? 1sg.dat

' If she is not able to carry those (sacks), of what use is she for me?' (i.e. a wife that is unable to carry the sacks of flour is useless)

(121) qus slip tːura-k-aˁq-ib-le, out-down-drag.pfv-pret-cvb er look ∅-ik'ʷ-aχː-an, m-look.at.ipfv-cond-cond.3 il that bek' head b-akːu n-cop.neg

'After having pulled (him) out, if they look, there is no head.'

The verbs that do not have an aspectual distinction can form the realis conditional as well as the imperfective realis conditional without any noticeable semantic difference 18.3 Conditional and concessive verb forms and clauses

between the two forms (122). For verbs with two aspectual stems the semantic difference is restricted to the aspectual difference between imperfective and perfective aspect; the conditional meaning is identical for both forms (123).


### **18.3.4 Imperfective past conditional**

There is also a past version of the imperfective conditional formed only from imperfective verbs. The precise formal make-up is still to be clarified since the form is only very rarely used. There are no corpus examples and elicitation is hard due to the insecurity of the speakers. It seems that the suffix *-aχː-an-del* can be used with all persons. It expresses irrealis conditional (124), (125) and past conditional meaning (126), depending on the sentence and the further context.


'If I would see a nice car, I would want it.' (E)

(126) dam 1sg.dat šahar-ri-cːe-b town-obl-in-n het=qːel dem=when qːuʁa beautiful mašin car či-b-ig-aχː-andel, spr-n-see.ipfv-cond-cond.pst b-ikː-ul n-want.ipfv-icvb hajq-i be.enough.ipfv-hab.pst.3 'At that time when I saw a nice car in the city, I wanted it.' (But now I do not care about cars anymore) (E)

### **18.3.5 Periphrastic conditional clauses**

As shown in (127–131), conditional clauses can be periphrastic, i.e., make use of the additional auxiliary *b-iχʷ-* (pfv) 'be, become, can'. In such clauses, the lexical verb bears a converb or occasionally a participial suffix and the auxiliary *b-iχʷ-* takes one of the conditional forms, e.g. realis conditional (127) or past conditional/irrealis conditional (129–131). More examples can be found in §15.4.1, which describes all uses of *b-iχʷ-* as auxiliary.

### 18 Non-finite verb forms

(127) du 1sg b-arx-le n-direct-advz ∅-ik'-ul m-say.ipfv-icvb a-jχ-ulle neg-be.pfv-cond.1 raχle if u-l 2sg-erg ʡaˁħ good či-b-arq'-a! spr-n-do.pfv-imp

' If (masc.) I do not tell it correctly, correct me!'


### **18.3.6 Concessive conditionals**

The conditional forms presented in the preceding sections can acquire a concessive conditional meaning ('even if') when the additive is encliticized to the conditional suffixes. For instance, the realis conditional (132) or the past conditional (133) can serve as the base for concessives.


### 18.3 Conditional and concessive verb forms and clauses

The third person concessive conditional form of the auxiliary *b-iχʷ-* (n-be.pfv) used in combination with interrogative pronouns lexicalized into a universal indefinite free choice pronoun similar to the English *-ever* series (134) (§4.6.3). Similarly, the verb *bikː-* 'like, want, love' can function as universal indefinite free choice when it takes a concessive conditional form and co-occurs with an interrogative pronoun (135).


However, conditional forms with an additional additive enclitic do not always express conditional concessive meaning. For present conditional forms the concessive semantics can be very weak (136) or even absent, in which case only the conditional meaning is conveyed. For past conditionals the meaning is irrealis conditional instead of concessive (118).

(136) hel=de that=pst hel that pepel ashes p'aq' shake.off ka-b-arq'-itːe=ra, down-n-do.pfv-cond.2sg=add "uberi!" take.away b-ik'-ul=de hpl-say.ipfv-icvb=pst "hetːi" those '(Even) if you let the ashes of your cigarette fall down, they said, "Put it away!"' (i.e. make it clean)

There is another way of forming concessive clauses by means of the enclitic *=xːar* (§18.2.8).

# **Part IV Syntax**

## **19 Valency classes and modification of valency patterns**

### **19.1 Valency classes**

### **19.1.1 Introduction**

Valency classes cross-cut the morphological classes of verbs. This means that the morphological classes (underived verb stems with or without preverbs, derived verbs, compound verbs, see §11.1 and Chapter 12) distribute over the valency classes with probably a preference for the simple underived verbs to occur in the intransitive, the transitive and to a somewhat lesser extent the affective valency class.

I will categorize verbs into valency classes according to two main criteria: (i) the number of arguments and (ii) the case marking of the subject-like argument. By 'subjectlike argument' I refer to the argument of the simple clause that has the most subject properties as opposed to all other arguments (see §22.3 for more details). Subject-like arguments are marked with one of the three cases absolutive, ergative or dative. I use the terms "one-place" or "monovalent", "two-place" or "bivalent", and "three-place" or "trivalent" for referring to the number of semantic arguments required by the verbs. The basic valency classes and the case marking of the subject-like argument are summarized in Table 19.1.


Table 19.1: Valency classes and case marking of subject-like arguments

The term "extended intransitive" refers to two-place predicates that, in addition to an argument in the absolutive, have a further argument in the dative or another case; "extended transitive" verbs are three-place verbs that besides having two arguments bearing the cases that are also used for transitive verbs, have an additional argument marked with the dative or in some other way Dixon (1994: 122–123). Thus, extended transitive

### 19 Valency classes and modification of valency patterns

verbs are ditransitive verbs. Furthermore, I use the term "affective predicates" for a clearcut class of mostly experiential predicates that express the experiencer argument in the dative and the stimulus argument, if there is one, in the absolutive. Affective verbs typically form their own valency class in East Caucasian (see, e.g. Comrie & van den Berg 2006; Ganenkov 2006; Comrie et al. 2018). One might hypothesize that they belong to the class of extended intransitive verbs. However, if one applies the commonly used test for subjecthood to the extended intransitive verbs and the affective verbs it immediately becomes clear that with the former class it is the absolutive argument that exhibits most subject properties whereas with the latter class it is the dative argument. For more information on grammatical relations in Sanzhi see §22.3 and Forker 2019b.

Table 19.2 provides an overview of the major valency classes discussed in this chapter; some minor classes are not listed, but discussed below. All verbs in the table and in the following subsections are presented in the order imperfective/perfective if they have two stems. Otherwise the single stem that is unspecified for aspect is given. In the table and in this chapter as well as elsewhere in the grammar I will use the following letters as mnemonics for macro roles (see Bickel 2011 and Bickel et al. 2015):


Subject-like arguments are of the type S or A. Note that S occurs with monovalent and bivalent verbs, which might seem slightly unusual. My reason for using the label S in this way is case marking, because all arguments falling under this label are marked by the absolutive case, which leads to a range of common morphosyntactic properties. For more details on grammatical relations see §22.3.

As Table 19.1 shows, monovalent verbs have three possibilities for marking their single argument. The majority of the monovalent verbs assign the absolutive case to the single argument (§19.1.2), though dative or, in case of one verb, ergative are also possible (§19.1.3).

Table 19.3 summarizes the case-marking patterns available in constructions with bivalent predicates, because they are the largest and most heterogeneous group. The columns represent the possible cases for subject-like arguments, which can be absolutive (S) or ergative (A), or dative (A). The rows display the possible cases for P arguments (absolutive, dative, genitive, spatial cases, ergative). As the table shows, the absolutive case is the most versatile case that can be combined with all other cases and encodes S, A or P, but the ergative is also quite flexible.

19.1 Valency classes

### case marking number of predicates patterns arguments and examples monovalent predicates intransitive (absolutive) S-abs 1 *b-ubk'-/b-ebk'-* 'die'; *či-r-ha-b-ulq-/či-r-ha-b-uq-* 'vomit' (2), (3) monovalent affective verbs (dative) S-dat 1 *ʡaˁħ-le ca-b* 'feel good, be well'; *c'aχ ka-b-ircː-/c'aχ ka-b-icː-* 'feel ashamed' (9), (10) bivalent predicates extended intransitive (absolutive + dative/spatial case) S-abs, P-dat 2 *kːač b-irk-*/*kːač b-ik-* 'touch', *gu-lik'-* 'listen to' (15) S-abs, P-in-lative/-dat 2 *b-ik'ʷ-* 'talk to'; *xʷit' b-ik'ʷ-* 'whistle at' (17), (21) S-abs, P-ante-ablative 2 *uruχ b-irχʷ-/uruχ b-iχʷ-* 'become/be afraid of'; *uruc b-irχʷ-/uruc b-iχʷ-* 'be/become ashamed/embarrassed of' (24) bivalent affective verbs (dative + absolutive/other) A-dat/erg, P-abs 2 *či-b-ig-/či-b-ag-* 'see'; *b-irʁ-/b-arʁ-* 'understand' (56), (60) A-dat, P-ante-ablative 2 *c'aχ-le ca-b* 'to feel/be ashamed in front of'; *b-irt'-/b-et'-* 'long for, miss' (74) transitive (ergative + absolutive) A-erg, P-abs 2 *b-irc-/b-ic-* 'sell'; *b-urχ-/b-arχ-* 'sew' (32), (33) other bivalent verbs (ergative + dative) A-erg, P-dat 2 *b-aˁq-/b-uˁrq-* 'hit'; *zaˁnʁ d-aˁq-/zaˁnʁ d-uˁrq-* 'phone' (49) trivalent predicates extended transitive (ergative + absolutive + other) A-erg, T-abs, G-dat/-in-lative 3 *lukː-/b-ikː-* 'give'; *či-b-iž-aq-/či-b-až-aq-* 'show'; *haʔ-/herʔ-* 'say, tell'; *b-urs-* 'say, tell'; *xar b-irʁ-/xar b-eʁ-* (36), (39), (44)

### Table 19.2: Major valency classes

### 19 Valency classes and modification of valency patterns

Table 19.3: Case-marking of arguments in constructions with bivalent predicates


Note that Table 19.3 conflates basic valency classes for bivalent predicates (i.e. extended intransitive, transitive, affective) with a number of other special constructions, which are available for some predicates of the basic valency classes (antipassive, constructions with absolutive, dative or genitive Ps).

The first row in Table 19.3 lists all constructions that consist of an S argument in the absolutive and a further P argument. These clauses with extended intransitive predicates, but also antipassives (§19.2.1), and two minor constructions with absolutive and genitive P arguments are described in §19.1.4. In all clauses with bivalent verbs and absolutive S arguments it is the S that controls gender agreement. Gender agreement with any other arguments is ungrammatical.

The second row in Table 19.3 contains all constructions with ergative A arguments and P arguments with various cases. First of all, the P argument can have the absolutive case (standard transitive verbs including causativized intransitive verbs, and, in certain TAM forms, affective verbs, §19.1.5, §19.1.8, §19.2.2). A few bivalent verbs with an ergative agent (A) require a goal or beneficiary argument marked with the dative or in-lative, or even an experiencer in the genitive (51), which represents the P argument. These verbs commonly have lexicalized direct objects in the absolutive case that are invariable parts of the compound verb and therefore do not count as arguments (§19.1.7).

Third, bivalent verbs with A arguments taking the dative are, as mentioned above, mostly affective verbs that have a P (stimulus) in the absolutive.

### **19.1.2 Intransitive verbs**

Intransitive verbs are one-place verbs. The single argument occurs in the absolutive and controls the gender/number agreement and the person agreement on the verb. Example verbs are given in (1) and examples sentences in (2–6).

(1) a. *ʡaˁlħ-*/*ʡaˁħ-* 'fly'


19.1 Valency classes


There are a variety of intransitive verbs that are compounds and contain a nominal part (§12.2.2). The nominal part, however, does not function as argument of the verb. It most frequently appears in the absolutive case (6), but the genitive case is also possible (7), or the loc-lative or spatial postpositions/adverbials. Note that in (6) the absolutive argument that controls the agreement has been omitted and only its genitive modifier appears in the clause.

	- b. *čːal b-ik'ʷ-; čːal b-ulq-/čːal b-uq-* 'argue, quarrel'
	- c. *sːiħ b-ik'ʷ-* 'breath'
	- d. *qal-la b-iχʷ-ij/qal-la ka-b-iž-ij* 'get married'
	- e. *waˁʡda-la b-iχʷ-ij* 'negotiate, conspire'

### **19.1.3 Monovalent affective verbs and exceptional monovalent constructions**

Sanzhi Dargwa has a few constructions with monovalent predicates and a single argument fulfilling the role of a dative-marked experiencer (8). Such constructions can be copula constructions with adverbs (9) or contain compound verbs (10). Gender/number agreement is frozen (prefixes *b-* or occasionally *d-*) and the person agreement is invariably third person.

### 19 Valency classes and modification of valency patterns

	- b. *wahi-l ca-b* 'feel bad'
	- c. *ʡaˁħ-le ca-b* 'feel good, be well'
	- d. *buruš hitːi sa-b-irk-/buruš hitːi sa-b-ik-* 'sleep, fall asleep'
	- e. *beža hitːi d-irk-/beža hitːi d-ik-* 'catch a cold'

The affective verbs 'see' and 'hear' can also be used as monovalent verbs with the meaning 'be/become visible, show off' and 'be/become audible' (see §24.5 for two example sentences and §19.1.8 below for a discussion).

There is a special predicate denoting weather phenomena that has one single argument marked with the ergative (11), (12). The verb does not have an aspectual distinction and always shows neuter gender and third person agreement. The same phenomenon is observed in the neighboring Icari Dargwa variety (Sumbatova & Mutalov 2003: 155), but apparently not in Standard Dargwa.

	- b. *mig-li b-us-* 'hail'
	- c. *duˁħi-l b-us-* 'snow'

### **19.1.4 Extended intransitive verbs and other constructions with bivalent predicates and absolutive S arguments**

The major class of predicates falling into this category are extended intransitive verbs. They are bivalent and have one S argument in the absolutive and another argument in the dative or in a spatial case. Gender/number as well as person agreement on extended intransitive verbs is always controlled by the absolutive argument.

The largest number of extended intransitive verbs have a goal argument in the dative (13). Example sentences are provided in (14–19). For the verb *xadi ag-/argʷ-* 'marry' it is always the nominal referring to the woman that occurs in the absolutive while the dative argument denotes the man (18). Some of the verbs given in (13) can be used as one-place verbs with reciprocal meaning, e.g. *qaˁb lus b-ilk-/qaˁb lus b-ik-* 'embrace each other, hug each other'.

19.1 Valency classes

	- b. *hitːi ka-b-ig-* 'wait for'
	- c. *kːač b-irk-/kːač b-ik-* 'touch'
	- d. *gu-lik'-* 'listen to'
	- e. *xʷit' b-ik'ʷ-* 'whistle at'
	- f. *waˁw b-ik'ʷ-* 'shout at, call, cry'
	- g. *b-iχči(t) argʷ-/b-iχči(t) ag-* 'believe'
	- h. *xadi argʷ-/xadi-ag-*/'marry'
	- i. *er b-ik'ʷ-* 'look at'
	- j. *paˁq b-ik'ʷ-* 'hit at, strike'
	- k. *b-urʁ-* 'shoot'
	- l. *qaˁb lus b-ilk-/qaˁb lus b-ik-* 'embrace, hug'

'The boy began to whistle and to shout at the dogs in the village.'


Verbs of speech and verbs with similar meanings may mark their addressee argument with the dative (17), but much more common is the use of the in-lative (21), (22). There is only one extended intransitive verb of speech, *b-ik'ʷ-* 'say', which is, however, also widely used in compound verbs (20).

19 Valency classes and modification of valency patterns


There are a number of compound verbs and copula constructions with experiential semantics that belong to the extended intransitive class and mark the second argument with the ante-ablative (23), (24) or take a clausal complement (25) (see Chapter 24).

	- b. *uruχ b-irχʷ-/uruχ b-iχʷ-* 'be/become afraid of, fear'
	- c. *uruc ca-b, uruc b-ik'ʷ-* 'be ashamed of, be embarrassed of'
	- d. *uruc b-irχʷ-/uruc b-iχʷ-* 'be/become ashamed of, be/become embarrassed of'

'"Are you afraid of your wife?" he says.'

(25) χːʷalle much ʁaj word d-ik'ʷ-ij=ra 1/2pl-say.ipfv-inf=add uruc embarrassed d-ik'-ul=de 1/2pl-aux.ipfv-icvb=pst 'We were embarrassed to talk a lot.'

Extended intransitive verbs expressing location, position, or movement combine with various spatial cases. Which spatial case is used depends on the semantics of the spatial reference point (i.e. the ground) and on the type of localization or motion (e.g. in or on a reference point, movement to a goal or from a source). The most common spatial series employed in these functions are the spr-series (26), the in-series (27), and the ad-series. Many more examples can be found in §3.4.2 on the spatial cases.

(26) ca-la one-gen antːa-le forehead-loc či-ka-b-iž-ib spr-down-n-sit.pfv-pret ca-b cop-n zija horsefly 'A horsefly sat down on the forehead of one (man).'

19.1 Valency classes

(27) il-tːi that-pl dubur-t-a-cːe-d mountain-pl-obl-in-npl er life d-irχ-ul npl-become.ipfv-icvb d-už-ib npl-be-pret ca-d cop-npl

'(Apparently) they (= the animals) lived in the mountains.'

Other constructions with an absolutive S and a bivalent predicate are instantiated by verbs that assign absolutive or genitive to the P argument, depending on the meaning of the construction. The combination absolutive S plus absolutive P is rare (28), (29). These clauses syntactically strongly resemble copula clauses but make use of verbs that express meanings other than the simple copula meaning. The verb agrees in person, number and gender with the subject-like argument, which can be distinguished from the second argument in the absolutive by reference to prominence properties such as animacy and person.


Example (30) illustrates a clause with two arguments that also resembles copula clauses. The S argument in the absolutive case functions as subject-like argument (e.g. it controls agreement on the verb). The nominal bearing the genitive is not a possessor of an omitted head noun, but an argument of the verb. Note that it is possible to replace the genitive by the absolutive with no salient change in the meaning of the clause.

(30) gacbe hips c'aq' strong darman-na medicine-gen d-irχʷ-an-te npl-become.ipfv-ptcp-dd.pl ca-d cop-npl 'The hips will be a strong medicine.' (E)

### **19.1.5 Transitive verbs**

Simple transitive verbs have two arguments: one is marked with the ergative and functions in a subject-like manner, and the other one bears the absolutive. Gender/number agreement is triggered by the absolutive argument. Person agreement can be controlled by the ergative or the absolutive argument and mostly follows the hierarchy 1,2 > 3 (see §20.3 on person agreement for more details and examples). Transitive verbs can be simple underived verbs as the six first verbs in (31) and the examples in (32), (34), verbs containing various preverbs, compounds containing transitive light verbs (33), (91), or causativized intransitive verbs (111), (113).

### 19 Valency classes and modification of valency patterns

	- b. *b-irq'-/b-arq'-* 'do, make'
	- c. *b-irc-/b-ic-* 'sell'
	- d. *b-urχ-/b-arχ-* 'sew'
	- e. *b-alš-/b-aš-* 'knead'
	- f. *b-it-* 'beat up'
	- g. *ač b-irq'-/ač b-arq'-* 'open'
	- h. *qaˁš k-aʁ-* 'cut into pieces'
	- i. *aq (či-ha-)b-irq'-/aq (či-ha-)b-arq'-* 'lift up, take up'
	- j. *t'aš b-ircː-aq-/t'aš b-icː-aq-; t'aš-aʁ-* 'stop'
	- k. *bursːi b-arq'-/bursːi b-irq'-* 'teach'

' I did not steal the stone. I did not kill anyone.'

### **19.1.6 Extended transitive verbs (i.e. ditransitive verbs)**

Extended transitive verbs have three arguments bearing the ergative, the absolutive and a further case. They follow the same agreement rules as simple transitive verbs. This means that the absolutive argument triggers the gender/number agreement. Person agreement is controlled by the absolutive or the ergative argument, but never by the third argument (reciepient, addressee, etc.). It normally follows the person hierarchy 1, 2 > 3. The extended transitive verbs in (35) all have dative arguments in addition to the ergative and absolutive arguments.

	- b. *qar b-irq'-/qar b-arq'-* 'charge, entrust with'
	- c. *xadi lukː-/xadi b-ikː-* 'marry off'
	- d. *či-b-iž-aq-/či-b-až-aq-* 'show'

19.1 Valency classes

(37) šːan-ni-j fellow.villager-obl-dat xadi married r-ičː-ib=da f-give.pfv-pret=1 di-la 1sg-gen ucːi-l brother-erg 'My brother married (me) off to a fellow villager.'

To this group belong a number of verbs expressing violent physical contact (38). These verbs have an absolutive argument denoting the instrument of the action (39). The instrument is usually omitted such that we are left with two arguments, the ergative agent and the goal that takes the dative or the in-lative (40). The valency frame is typical for this semantic type of verbs and has been described for other East Caucasian languages (Khalilova 2009: 332–334; Forker 2013a: 476).

	- b. *b-urh-/b-erh-* 'knock, strike, bang'
	- c. *irx-/ixʷ-* 'throw at, shoot'

However, the verb *b-erh-* (pfv) 'knock, strike, bang' takes only instruments in the ergative or the comitative case that do not control the agreement, such that the resulting clauses lack absolutive arguments (42). The agreement trigger is not overtly present in the clause and cannot be retrieved by speakers. The difference in gender agreement goes hand in hand with a difference in the meaning of the clauses: when the neuter singular prefix *b-* is used the event occurred only once; when the neuter plural suffix is used the knocking-event occurred repeatedly such that the meaning is rather 'beat off'.

(42) ʡaˁli-l Ali-erg weliχan-ni-j giant-obl-dat naˁq-li hand-erg / / naˁq-li-cːella hand-obl-comit b-erh-ib n-strike.pfv-pret / / d-erh-ib npl-strike.pfv-pret ' Ali knocked/beat off the giant with the hand.' (E)

Another group of extended transitive verbs are verbs of speech that take an addressee argument in the in-lative (43–45). Their absolutive argument is either a clause (44), (45), or a noun that refers to the speech event such as *χabar* 'story' (108).

### 19 Valency classes and modification of valency patterns


Verbs denoting movement and positioning of objects or animate entities combine with various spatial cases, e.g. the ad-lative or the loc-lative.

	- b. *sa-b-ik-/s-ak-, h-ak-, k-ak-* 'bring, lead'
	- c. *b-uk-/b-erč-* 'take, collect, bring'
	- d. *(či-)(ka-)b-irxː-/(či-)(ka-)b-ixː-* 'put'

### **19.1.7 Bivalent verbs with lexicalized objects and other rare constructions with bivalent verbs**

There are a couple of verbs with ergative subject-like arguments that have fixed, lexicalized objects that control the gender/number agreement on the verb, but cannot be exchanged with other nominals because they are constitutive for the meaning of the predicate (48). For such verbs, it is impossible to add another nominal in the absolutive, because the object-like position is already occupied by the lexicalized object. However, it is not always clear whether the lexicalized object controls the gender agreement on the verb or whether the gender agreement is frozen. From a semantic point of view, the verbs can be analyzed as bivalent or trivalent. The bivalent verbs have additional arguments in the dative (49) or in-essive (50) or occasionally in other cases (51). Verbs of speech preferably make use of the in-essive whereas other verbs mostly employ the dative. Some of these verbs take additional complement clauses.

	- b. *kumek b-irq'-/kumek b-arq'-* 'help'
	- c. *tamaša b-arq'-/tamaša b-irq'-* 'wonder'

19.1 Valency classes


'His wife begged him: "Don't go!"'


Some of the frozen objects occur in more than one construction. For instance, *urk'ec'i*, when combined with a verb, can occur in the ergative construction with an agentive experiencer (53), in a construction with a dative experiencer (71), and together with an experiencer in the genitive (72). The stimulus is always a goal or beneficiary-like argument and therefore takes the dative.

(53) du-l 1sg-erg ʡaˁjib-kar-t-a-j guilt-nmlz-pl-obl-dat urk'ec'i pity b-irq'-id n-do.ipfv-1.prs 'I pity the convicted.' (E)

Note that Sanzhi also has a range of compound verbs with nouns marked by the genitive or by spatial postpositions. However, the nouns used in such constructions are not lexicalized objects but nominal parts of compound verbs. For examples see §12.2.2.

### **19.1.8 Bivalent affective verbs**

Bivalent affective verbs are a relatively small class of two-place verbs with epxeriential/affective semantics (54). They express unintentional and uncontrollable perception, emotion, volition, cognitive activities and other non-agentive events and situations.

19 Valency classes and modification of valency patterns

	- b. *t'am b-iq'-/t'am b-aq'-* 'hear'
	- c. *b-ikː-* 'want, like, love'
	- d. *b-irʁ-/b-arʁ-* 'understand'
	- e. *qum urt-/qum ert-* 'forget'
	- f. *b-alχ-/b-aχ-* 'know'
	- g. *b-urkː-/b-arkː-* 'find'
	- h. *han b-irk-/han b-irk-* 'remember, seem'
	- i. *b-ičː-aq-* 'like, love'
	- j. *či-b-b-irt'-/či-b-b-et'-* 'be bored'
	- k. *b-irt'-/b-et'-* 'be bored'

Most bivalent affective verbs have an experiencer argument in the dative and a stimulus argument in the absolutive. They follow the same agreement rules as transitive verbs, i.e. gender/number agreement with the absolutive argument and person agreement is ruled by the hierarchy 1, 2 > 3 (55), (56), and/or it is the experiencer that controls the agreement (55), (56), or it is invariably third person (61), (70).


When inflected for some tenses such as the habitual past, the compound present (57) or the future (58), (59) and in some types of subordinate clauses certain affective predicates allow for the experiencer to bear the ergative instead of the dative case. The ergative alignment pattern is more common in other Dargwa varieties such as Icari Dargwa, and has been investigated from a diachronic perspective in Ganenkov (2013). In Sanzhi Dargwa, it is less common and the precise conditions that allow for ergative experiencers still need further investigation. In any case, it follows the same agreement rules as dative experiencers, e.g. in (57) person agreement is controlled by the ergative experiencer and in (58) by the absolutive stimulus.

(57) či-w-ig-ul spr-m-see.ipfv-icvb akːʷa-di cop.neg-1 du-l 1sg-erg heχ dem.down admi, person xːunul=el woman=indq murgul=el masculine=indq

'I do not see this person (on the picture), whether it is female or male.'

19.1 Valency classes


It is not always possible to determine if a specific examples follows the person hierarchy or if it is the experiencer, who controls the agreement (which can also be formulated as semantic role hierarchy: experiencer > stimulus). For instance, in (60) the person agreement enclitic on the verb =*da* expresses first person singular or plural and second person plural agreement, such that it could be either the experiencer (in accordance with the experiencer controlling agreement independently of person) or the stimulus (in accordance with the hierarchy) that functions as controller. Similarly, in both (55) and (60), a first person experiencer controls the agreement suffix, which can be explained by the person hierarchy or by the semantic role hierarchy).

(60) či-d-d-et'-ib-le=da, spr-1/2pl-1/2pl-bore.pfv-pret-cvb=1/2pl haʔ-ib=da, say.pfv-pret=1 ušːa 1pl dam 1sg.dat '"You (pl.) bored me," I said.'

In general, experiencer verbs seem to allow for a higher degree of variation concerning person agreement than transitive verbs. This includes the fact that under certain circumstances the person agreement is third person although the clause contains a first or second person dative pronoun in the semantic role of experiencer. For instance, with the verb 'forget' both person agreement enclitics and third person agreement are found in the Sanzhi corpus, but third person agreement prevails. Thus, in (58) we find second person singular controlled by the stimulus and (55) the verb agrees with the experiencer in the dative (first person singular). By contrast, in (61) and (62) the agreement is third person instead of the expected first person agreement.


As the following minimal pair shows, the variation that the verb 'forget' shows between person agreement and invariably third person does not imply any differences in meaning and is not tied to certain TAM forms (as it is the case for ergative experiencers, which are only available for a restricted number of TAM forms, but for all affective verbs). The variation includes forms with person suffixes and forms with person enclitics alike.

### 19 Valency classes and modification of valency patterns

One and the same tense form can show variation as the following two examples of the preterite demonstrate. The first sentence (63) shows person agreement with the first person experiencer whereas the second sentence (64) has a verb form that corresponds to the third person preterite (i.e. no person enclitic, no copula).


Based on the data collected so far I am not able to explain the variation by means of linguistic or extralinguistic factors. Another, more general question concerns the nature of the third person forms in (61), (62), (64), and other affective verbs below, for which three different hypotheses could be suggested. First, we can perhaps analyze it as third person agreement controlled by the absolutive patient that overrules the agreement hierarchies stated above. It would then follow the ergative pattern analogously to the ergative agreement attested in certain TAM forms and discussed in §20.3.2. Alternatively, we can claim that we deal with 'suspended person agreement' in the sense that the verb shows the default person agreement form, namely third person, but this form does not underlie control but shows actually the lack of an agreement controller.

A third alternative would be to suggest that the verbs in (61) and (62) are one-place verbs and the dative pronouns are not genuine arguments of the verb but something like adjuncts and can therefore not control the agreement.<sup>1</sup> This argumentation could be supported by the fact that even the verbs 'see' and 'hear', which are normally used as two-place affective verbs can be used as one-place bivalent verbs with the meanings 'be visible' and 'be audible'. In that case normally the dative experiencer can be omitted. Thus, (65) can be used with a dative pronoun, in which case two translations are possible ' I began to see the mountains.' or 'The mountains started to be visible to me.' If the pronoun is omitted, then the only translation is 'The mountains started to be visible.'

(65) (dam) 1sg.dat dubur-te mountain-pl či-d-ig-ul spr-npl-see.ipfv-icvb d-aʔ npl-begin ašː-ib begin.pfv-pret

' I began to see the mountains.' OR 'The mountains started to be visible to me.' (E)

Similarly, the verb 'remember' is a compound verb in which the verbal part consists of the otherwise intransitive light verb *b-ik-* 'occur'. I found only third person agreement in all corpus examples as well as in elicitation, which suggests that the dative experiencer is syntactically not an argument but an adjunct such as a goal (66).

<sup>1</sup> If this approach can be corroborated by further research, then the discussed verbs and examples have to be classified as monovalent affective verbs. For the sake of the argumentation and because I am unable to draw a conclusion at the present moment I prefer to leave this part of the section where it is.

19.1 Valency classes

(66) heχ-tːu-b dem.down-loc-hpl han remember b-ič-ib hpl-occur.pfv-pret dam 1sg.dat kulpat=ra family=add 'There I also remembered my family.'

There are some more predicates that can be classified as two-place affective predicates because they come with two semantic roles, an experiencer and a stimulus, but which differ from the predicates discussed so far in this section. First of all, there are two copula constructions with adverbials that mean 'needed' (67). In these constructions the absolutive stimulus functions as copula subject and thus person and gender agreement controller (68), or alternatively complement clauses can be used. The dative can be classified as copula predicate and its use is optional. The predicates therefore behave in the same way as what has been said above about 'see' and 'hear', i.e., they can be used as monovalent predicates without an experiencer in impersonal constructions or as bivalent affective verbs.

	- b. *ħaˁžat-le ca-b* 'need, be necessary'

There are two bivalent affective verbs that do not show person or gender agreement, but invariable third person forms and the default gender agreement prefix *d*- (69).<sup>3</sup> The verb *simi d-uq-/simi d-ulq-* 'be angry', already mentioned in §19.1.4, is a one-place verb that can be changed into a two-place verb with a further experiencer/goal argument in the dative by adding the spatial preverb *či-* to it (70). This experiencer/goal argument can never control person agreement (i.e. first person agreement in the examples below is ungrammatical) and thus the person agreement is always third person. The identical lexical verb with the same preverb con also occur in a compound with *urk'ec'i* 'pity' with exactly the same morphosyntactic properties (71). Note that in (72) the dative pronoun has been replaced by a genitive possessor that now encodes the semantic role of experiencer. This examples is an indication that the dative pronouns in the other examples (70) and (71) are not arguments but adjuncts, perhaps comparable to external possessor that can be expressed in the dative or in the genitive.

	- b. *urk'ec'i d-uq-/urk'ec'i d-ulq-* 'pity'

<sup>2</sup>This predicate can also occur with an experiencer argument in the absolutive that takes over the role as copula subject and controls agreement. See example (33) in §20.2.1.

<sup>3</sup>The same lexical item *urk'ec'i* 'pity' is used in another semantically very similar predicate together with the lexical verb 'do, make'. In that construction the lexical verb has the agreement prefix *b*- for neuter singular (53). Therefore, the agreement prefix *d*- in (69) cannot be controlled by the items preceding the verbs but must be a default prefix. In general, both *b*- and *d*- function as default agreement exponents in a number of different constructions (§20.2.1).

19 Valency classes and modification of valency patterns


Finally there are a few constructions with dative experiencers and a source-like or cause-like stimulus arguments that bear the ante-ablative (73). This is the same case that is used by some monovalent experiential verbs for marking the source/cause-like arguments (24). In these constructions, there is again invariable third person agreement and default neuter singular gender agreement that is frozen and not controlled by any of the constituents (74–76).


### **19.1.9 Labile verbs**

Sanzhi Dargwa has a number of labile verbs that can be used as intransitive verbs (with the corresponding morphology) or as transitive verbs. Because in Sanzhi arguments can be omitted and are often omitted if their reference is clear from the context, at times it can be difficult to identify labile verbs. Furthermore, occasionally transitive verbs occur in impersonal constructions without arguments that could syntactically be defined as subjects or semantically identified as agents (77), (78). Outside of the constructions shown in (77) and (78), the verb *b-irq'-* (ipfv)/*b-arq'-* (pfv) is transitive and I therefore do not include it in the list of labile verbs.

19.1 Valency classes


Sanzhi Dargwa makes use of different suffixes for the imperative of many intransitive and transitive verbs, and the stem augment vowels in the prohibitive and the habitual present also differ according to transitivity. Thus, the verbal morphology provides decisive clues for deciding whether a verb is used intransitively or transitively.

The majority of the labile verbs are S=P-labile, preserving the argument with the patientive semantic role (79). The first example sentence in (80) shows the intransitive use, and the second one in (81) illustrates the transitive use.

	- b. *b-elq'-* (pfv)/*luq'-* (ipfv) 'break, shatter, smash'
	- c. *b-erc'-* (pfv)/*b-uc'-* (ipfv) 'fry, roast, bake'
	- d. *b-ic'-* (pfv)/*b-irc'-* (ipfv) 'fill'
	- e. *b-aˁč* (pfv)/*b-aˁlč* (ipfv) 'squeeze, break, crush, crack, trample'

'The barn burnt.' (E)

(81) χalq'-li people-erg qal house c'a-l fire-erg b-ikː-ul n-burn-icvb ca-b cop-n 'The people burn the house with fire.' (E)

The prohibitive of the intransitive clause is given in (82), and the prohibitive of the transitive can be found in (83).


I found a few S=A labile verbs that preserve the subject-like argument, namely:

(84) a. *b-elč'-* (pfv)/*b-uč'-* (ipfv) 'read, learn, study, sing', (85), (86) b. *b-erkʷ-* (pfv)/*b-uk-* (ipfv) 'eat'

19 Valency classes and modification of valency patterns


Translational equivalents of 'read' are also labile in a number of other East Caucasian languages (e.g. in Icari Dargwa, Sumbatova & Mutalov 2003: 154–155, and in Hinuq, Forker 2013a: 492). Note that lability surfaces only with the imperfective aspect of those verbs that can be used intransitively or transitively. This means that the perfective stems always occur in transitive constructions.

### (85) intransitive construction

	- a. du-l 1sg-erg kːazat newspaper b-uč'-an=da n-read.ipfv-ptcp=1 'I will read the newspaper.' (E)
	- b. kiniga book b-elč'-en n-read.pfv-imp / / kiniga book ma-b-uč'-it! proh-n-read.ipfv-proh.sg 'Read the book!/Do not read the book!' (E)

Furthermore, an optional P argument can be added in the intransitive use. This argument needs to be semantically plural and indefinite, and is marked with the ergative case (87). This construction is called "antipassive" in Dargwa languages and treated in more detail in §19.2.1.

(87) hana now Sanijat Sanijat kiniga-b-a-l book-pl-obl-erg r-uč'-unne f-read.ipfv-icvb ca-r cop-f 'Now Sanijat reads books.' (E)

### **19.2 Modification of valency patterns**

Sanzhi Dargwa has two major means of modifying the valency patterns of verbs, the antipassive as a detransitivizing operation (§19.2.1), and the causative as an argumentincreasing operation (§19.2.2). There are no potential constructions, which in other East Caucasian languages are used to detransitive verbs. There is also no biabsolutive construction, which would allow for the use of two absolutive arguments with a transitive verb.

19.2 Modification of valency patterns

### **19.2.1 Antipassive**

Sanzhi Dargwa has an antipassive that is formed by reversing the case marking of A and P in a clause with a canonical transitive predicate (88–89). Since both A and P are obligatorily arguments in the antipassive construction, it is not an argument-decreasing operation, although the A argument is frequently covert in examples from natural texts. The verb remains unmarked, but the gender/number agreement on the verb changes. Due to the lack of formal marking on the verb the antipassive in Sanzhi is not a typical antipassive from a typological perspective (Polinsky 2005).


Apart from being restricted to only one predicate class, namely canonical transitive verbs, the antipassive is additionally constrained in other ways:


### 19 Valency classes and modification of valency patterns

would be equally grammatical in the two adverbial clauses. In other words, the antipassive is not needed for pivot modulation. On the contrary, it is used for purely semantic reasons.

(90) [baliqː-a-l fish-obl-erg d-ucː-ul 1/2pl-bake.ipfv-icvb hel-itːe] that-dd [c'idex-li fruit-erg d-uk-unne] 1/2pl-eat.ipfv-icvb d-už-ib=da 1/2.pl-be-pret=1

'Frying fish, eating fruits we stayed.'

	- (94) a. ergative construction Rašid-li Rashid-erg mašin car qːurt push b-irq'-ul n-do.ipfv-icvb ca-b cop-n 'Rashid is pushing a/the car.' (E) b. ergative construction Rašid Rashid mašin-ni car-erg qːurt push ∅-irq'-ul m-do.ipfv-icvb ca-w cop-m 'A/the car is pushing Rashid.' (NOT: 'Rashid is pushing a/the car.') (E)

<sup>4</sup>An anonymous reviewer pointed out that the unavialability of the antipassive reading in (94b) might also be due to the fact that the antipassive expresses repeated or habitual situations. This is possible and more research is needed to give a conclusive answer for why this example cannot be interpreted as antipassive.

19.2 Modification of valency patterns

	- (95) du 1sg t'ult'-li bread-erg r-uk-un-ne=da f-eat.ipfv-pret-cvb=1 'I (fem.) am eating bread.' (E)
	- (96) a. aždaha-l monster-erg du 1sg ∅-ukː-unne=da m-eat.ipfv-icvb=1 'The monster is eating me.' (E)
		- b. du-l 1sg-erg aždaha monster b-ukː-unne=da n-eat.ipfv-icvb=1 'I am eating the monster.' (NOT: 'The monster is eating me.') (E)

Syntactically, the antipassive is a detransitivizing operation. The main proof for this is, of course, that the A argument occurs in the absolutive case and controls the gender agreement, whereas the P argument takes the ergative case. The functional range of the ergative comprises not only the expression of agents, but also of other semantic roles with a more peripheral status (adjuncts), most notably instruments (§3.4.1.2). The ergative P of the antipassive largely fits into this range. Furthermore, the distinction between suffixes for intransitive and for transitive verbs that is made in the imperative and in the prohibitive shows that verbs in the antipassive construction are detransitivized. Thus, the prohibitive suffixes for intransitive verbs are *-ut* (sg)/*-utːaja* (pl) with the stem augment vowel *u*, whereas the transitive verbs have *-it* (sg)/*-itːaja* (pl) with the stem augment *i* (§17.2). The antipassive construction requires the same prohibitive suffix as intransitive verbs (97), which is ungrammatical in the ergative construction (98).

	- a. ušːa 2pl hin-ni water-erg ma-d-učː-utːaja! proh-1/2pl-drink.ipfv-proh.pl 'You do not drink water (regularly)!' (E)
	- b. dig-li meat-erg ma-w-k-ut! proh-m-eat.ipfv-proh.sg 'Do not (always) eat meat!' (said to a man) (E)

19 Valency classes and modification of valency patterns

(98) ergative construction


The major problem in the analysis of antipassive constructions concerns the closeness to S=A labile verbs that can be used intransitively and transitively, thereby preserving the agent argument (§19.1.9). For instance, the imperfective stem of the verb *b-elč'-* (pfv)/ *b-uč'-* (ipfv) 'read, learn, study, sing' can be used in an intransitive construction. When adding the ergative adjunct *student-li* (student-erg) to (99) the translation is unambiguously 'she studies (at a university as a student)'. The same verb can be used in a transitive construction with an ergative agent and an absolutive patient (100).


In the antipassive construction, to the intransitive clause in (99) a P argument in the plural marked with the ergative case is added (101). The presence of the P argument is the only difference between the two sentences (99) and (101). Thus, instead of speaking of an antipassive construction we can also say that Sanzhi has a number of S=A labile verbs that are used intransitively with an optional nominal in the ergative that has syntactically rather the status of an adjunct.

(101) it that turk-m-a-l nasheed-pl-obl-erg r-uč'-unne f-read.ipfv-icvb ca-r cop-f 'She reads (i.e. sings) nasheeds.' (E)

The verb *b-irq'-* (ipfv)/*b-arq'-* (pfv) 'do, make, be busy' belongs to the verbs that frequently occur in antipassive constructions (92) and can also be used intransitively without any P argument (102). For this verb, there is a further possibility of use in weather constructions in which there is no A argument (77). The latter construction thus resembles S=P-labile verbs (§19.1.9).

(102) it that r-irq'-ul f-do.ipfv-icvb ca-r cop-f 'She is busy.' (E)

19.2 Modification of valency patterns

In sum, we can divide verbs in Sanzhi into three classes:


The use of antipassives is semantically rather than syntactically motivated. It has habitual semantics, which is typical for antipassives in general and antipassives in East Caucasian languages in particular (89) (see, e.g., van den Berg 2003a, Tatevosov 2011, Comrie et al. Forthcoming). Most notably, in all corpus examples the P argument is indefinite and usually in the plural or it has the meaning of a mass noun. Morphologically singular P arguments are only allowed if they can have mass noun readings. The P argument does not refer to a particular, specified object, but is semantically demoted. The sentences refer to repeatedly or habitually occurring actions. For instance, in (103) the speaker was talking about the life of her grandfather and how he used to be, which types of work he used to do.


' Saying "I drive that, here", he was writing letters to us.'

19 Valency classes and modification of valency patterns

(107) ʡaˁžib surprising qːuʁa beautiful ʡaˁjur-t-a-l dance-pl-obl-erg b-irq'-ul=de, hpl-do.ipfv-icvb=pst nišːa-la-te 1pl-gen-dd.pl daˁʡle as akːʷ-ar, cop.neg-prs.3 χːula big xːun-re=ra woman-pl=add ' They were dancing surprisingly beautifully, not like our (women), also the elderly women.'

By contrast, the P argument in the ergative construction can have a definite interpretation, referring to specific object. Thus, compare (108) in which the subject referent is telling a specific story<sup>5</sup> to (109), which refers to the action of storytelling without specifying the stories further, but could rather be a characterization of the person as a story-teller.<sup>6</sup>


### **19.2.2 Causativization**

Sanzhi has a very productive derivational process for the formation of causativized predicates by means of the suffix *-aq*. The derived causativized verbs behave like any other underived verbs, i.e., there are no differences in the range of verbal forms and constructions in which they may appear. The suffixation of *-aq* does not have any impact on the aspectual properties of the verb, such that the differences between imperfective verbs and perfective verbs are preserved. In addition, there are other formal means for making causative constructions such as auxiliary change.

Causative constructions are very widespread among the East Caucasian languages, though not all languages have dedicated derivational suffixes. In Sanzhi Dargwa, causativization normally applies only once to the verbal stem, but in elicitation it can also be added twice. When it is added to the verb, usually the number of arguments of the verb is augmented by one. This means that a monovalent verb becomes bivalent whereby S changes to P and a second argument, the ergative A in the role of the causer is introduced.

<sup>5</sup>This is clear from the context of the example. Without a context the same sentence could also be translated as 'He tells a story.'

<sup>6</sup>The two verb in the examples represent two distinct lexemes, which are partially in complementary distribution because of their aspectual properties. The verb in (109) is used as the imperfective counterpart of the verb *b-urs-ij*, which occurs in (108). It is morphologically defective because it can only be inflected for the imperfective converb and the modal participle, whereas *b-urs-ij* can be inflected for all verb forms and is aspectually neuter. The exact relationship between the two verbs requires further investigation.

19.2 Modification of valency patterns

(110) intransitive heχ dem.down urχːab mill lus around b-ik'-u n-move.ipfv-prs.3 'This mill spins around.' (E)

(111) intransitive heχ-i-l dem.down-obl-erg heχ dem.down urχːab mill lus around b-ik'-aq-u n-move.ipfv-caus-prs.3 'This makes the mill spin around.'

(112) transitive Sanži-b Sanzhi-n b-ik'-u=w n-grow.ipfv-prs.3=q ij=ʁuna? this=eq

'Does something like this grows in Sanzhi?'

(113) transitive

Marijam-li Marijam-erg χijal-te cucumber-pl d-ač'-aq-ib npl-grow.pfv-caus-pret ' Marijam was growing cucumbers.' (E)

Similarly, after causativization the S argument of bivalent extended intransitive predicates (114) becomes P and thus does not change its case marking, the second argument also remains unchanged and a third argument, the causer in the form of an ergative A is added (115).

(114) extended intransitive χalq' people qːarq-ne stone-pl arž-i go.ipfv-hab.pst.3 'The people turned into stones.' (E)

(115) extended transitive

hek' dem.up ħaˁšukː-a-d pot-loc-npl rurčː-an boil-ptcp χːink'-e khinkal-pl ʡaˁt'-ne frog-pl arž-aq-i go.ipfv-caus-hab.pst.3 heχ dem.down Irbihin-ni Irbihin-erg ' Irbihin turned the khinkal that was boiling in the pot into frogs.' (E)

Bivalent transitive predicates become trivalent extended transitive predicates when they are causativized, and the former As become Gs whereas Ps are unaffected (116). The G argument, that is, the causee, must be marked with the in-lative case. This case is frequently used in valency patterns of various predicates for semantic roles such as addressee, goal or beneficiary, which explains its use in causative constructions. The causee has semantic properties close to these roles since it is the argument, at which the action is directed and that might profit from it.

19 Valency classes and modification of valency patterns

(116) a. transitive Madina-l Madina-erg kaš porridge b-uk-unne=de n-eat.ipfv-icvb=pst 'Madina was eating porridge.' (E)

> b. extended transitive aba-l mother-erg Madina-cːe Madina-in kaš porridge b-erk-aq-un n-eat.pfv-caus-pret 'Mother made Madina eat porridge.' (E)

In the Sanzhi corpus, causativized transitive verbs are rather rare. Sentences (117) and (118) show two instances. Many corpus examples of causative constructions have intransitive base verbs such as (111) and (113) above, but causativized affective verbs also occur frequently (122).


With bivalent experiential predicates there are two possibilities: either one argument is added or the number of arguments is preserved. In the first case, the experiencer (the former A) becomes G without changing its case marking, but an additional A is added to the clause because the derived verb is trivalent (119b).

	- b. Pat'ima-l Patima-erg Madina-j Madina-dat jangi new kːurtːi dress či-b-iž-aq-ib spr-n-see.ipfv-caus-pret 'Patima showed Madina a new dress.' (E)

The same option is available for the causative of 'know', which translates as 'tell, inform, make know' (120). It is also possible for the experiencer argument to change its case marking from dative to in-lative because the latter case is regularly used for addressees with verbs of speech, but also for causees of causativized transitive and extended transitive verbs (121), (122).

(120) il-i-l that-obl-erg bari-li-j sun-obl-dat barkalla thanks b-aχ-aq-ur n-know.pfv-caus-pret ' He thanked the sun.'

19.2 Modification of valency patterns


The second option for affective verbs is not to have any change in the argument structure of the predicate such that both grammatical relations (A and P) as well as semantic roles remain unaltered. Only the semantics of the predicate slightly changes when the verb is causativized (123a–123b) and acquires a more agentive reading. This becomes especially obvious when the ergative instead of the dative is used to encode the experiencer of a causativized affective predicate. Verbs that choose this strategy are *b-ikː-* 'want, like', *b- arkː-* (pfv) 'find', and *han d-irk-* (ipfv) 'remember' (>*han d-irč-aq-*).

	- b. Murad-li-j Murad-obl-dat Madina Madina r-ičː-aq-ib f-want.ipfv-caus-pret 'Murad loved Madina.' (E)

If trivalent predicates are causativized, then A becomes the causee with the appropriate case suffix (in-lative) and a new causer in the ergative is added to the clause (125). Since the verb *b-ikː-* (pfv)/*lukː-* (ipfv) 'give' assigns not only the dative case to the recipient, but alternatively also the in-lative, it is possible to have two arguments with the same case marking in a clause with the causativized verb 'give' (126). Due to the identical case marking such clauses are ambiguous.


In sum, if an additional argument is added by means of causativization, it is always a causer marked with the ergative, independently of the valency class of the base predicate. Because the causer takes the subject position, the original subject (S or A) is demoted

### 19 Valency classes and modification of valency patterns

into a non-subject position (S > P, A > G), taking over the highest free position on the hierarchy of grammatical relations. For S this is the direct object position (P); for A this is the indirect object position (G) since the direct object position (P/T) is already occupied. It is never P or T that is affected when bivalent or trivalent predicates are causativized such that causativization can perhaps be taken as a weak indicator of an accusative pivot (see the discussion of grammatical roles in §22.3).

Double causativization seems to be possible, as (125) shows, and can lead to the addition of two arguments (i.e. the two-place verb 'see' becomes a four place verb). However, it can also be used for emphasis only such that the second causativization does not result in the addition of a second argument (129). In the corpus I found only one example of this (130). The precise properties of double causative constructions are hard to determine because speakers have divergent intuitions about the acceptability and meaning of elicited examples and the only corpus example (130) is difficult to understand and to judge, even within its context.


The meaning of causative constructions can be described as the expression of "a causal relation between two events, one of which is believed by the speaker to be caused by the other" (Kulikov 2011). Depending on the semantics of the predicate and on the context, the meaning of the causative construction can be close to force ('make do X', 'cause to X'), but it can also be 'soft causation', i.e., asking, requesting or begging (131), or sometimes even quite idiosyncratic and unpredictable. Thus, the meaning of the causativized intransitive verb *b-ucː-* 'work' is 'support, sustain' (134) in addition to the expected causative meaning of 'make work'.

(131) χabar story b-urs-aq-an=da n-tell-caus-ptcp=1 ' We will ask for a story to be told.' 19.2 Modification of valency patterns


Another way of forming causative constructions is by means of transitive light verbs. This operation is applied with compound verbs that contain intransitive light verbs (§12.1).

## **20 Agreement**

Sanzhi Dargwa has gender, number and person agreement. Formally, there are several systems of agreement exponents that act completely independently from each other and are therefore treated separately. We can distinguish between pure number agreement, combined gender/number agreement and person agreement. Pure number agreement occurs noun-phrase internally and at the clausal level with a restricted number of TAM forms (§20.1). Combined gender and number agreement is attested for the vast majority of East Caucasian languages, including Sanzhi Dargwa (§20.2). It is often found within the noun phrase and at the clausal level with all TAM forms, including verb forms such as converbs and participles. Person agreement is rather rare for East Caucasian languages. Among the languages that have it are Dargwa languages such as Sanzhi (§20.3), Lak, Tabasaran, Batsbi (Tsova-Tush), Udi, and to a lesser extend Hunzib, Akhvakh, and some Avar varieties (see Helmbrecht 1996; van den Berg 1999; Schulze 2011). It only occurs at the level of the clause.

I will use the terms "agreement", "target", and "controller" in the sense of Corbett 2006 to describe the properties of the three types of agreement in Sanzhi.

### **20.1 Pure number agreement**

Pure number agreement is found in the noun phrase and at the clausal level. Within the noun phrase, demonstrative pronouns (§4.2) and definite descriptions formed by means of the cross-categorical suffix -*ce* (§9.6.1) agree with the head noun in number. If the head noun is in the plural the demonstrative pronoun must occur in the plural and the cross-categorical suffix must change to *-te* (or be omitted) (1), (2).


Noun phrases modified by numerals other than *ca* 'one' are semantically plural and thus require demonstratives to appear in the plural and prohibit the use of the singular cross-categorical suffix<sup>1</sup> although no overt plural marking on the noun occurs (3). Some mass nouns also require plural agreement even though they are not overtly marked for plural, e.g. terms for ethnic groups of inhabitants of villages (6).

<sup>1</sup>The suffix can also be omitted because the adjectives in attributive function can generally occur with or without it.

### 20 Agreement

(3) hel-tːi that-pl ʡaˁbal three d-uqna(-te) npl-old-dd.pl q'ʷal cow 'those three old cows' (E)

At the clausal level pure number agreement is expressed by means of the special plural suffix of the optative, *-ar-te*, which is only used for plural addressees (4) (§17.3), and through the cross-categorical suffixes -*ce* (plural -te) and -*il* in those periphrastic verb forms, which make use of the suffixes (experiential I, experiential II, obligative present). Singular agreement controllers require -*ce* (5) or -*il* (7); plural agreement controllers require -*te* (6), (8). This type of agreement follows ergative alignment. For one-place verbs and extended intransitive verbs the number agreement controller is the single argument in the absolutive (5), (6); with transitive verbs and affective verbs the number agreement controller is the absolutive patient or stimulus (7), (8). More examples are given in §20.3.2.


### **20.2 Combined Gender/number agreement**

### **20.2.1 General remarks on gender/number agreement**

Combined gender/number agreement is a pervasive feature of East Caucasian languages including Sanzhi Dargwa. It is possible that within one clause three, four, or even more linguistic items agree with one and the same agreement controller. Sanzhi has three 20.2 Combined Gender/number agreement

genders that have a transparent semantic basis: masculine, feminine, and neuter (§3.1). Agreement targets for gender/number agreement can be divided according to the same two agreement domains that have been mentioned for pure number agreement in the previous section, i.e. (i) the clausal domain (§22.1), and (ii) domain of the noun phrase (§21.1). Within the domains the various targets can co-occur, depending on the morphosyntactic context (i.e. a noun in the essive case can but need not to be accompanied by an agreeing postposition). Example (9) illustrates agreement within a clause. Four targets (lexical verb, copula, noun and postposition) agree with the agreement controller (a nominal with a masculine singular referent), which is not overtly expressed. The noun phrase in (10) contains two agreeing modifiers, a quantifier and an adjective.

### **Clausal domain**


### **Domain of the noun phrase**


Furthermore, a small number of nouns (e.g. *b-ah* 'owner, master') (§3.1) and reflexive pronouns in the absolutive (§4.3) and one reciprocal pronoun (§4.4) contain gender exponents that express the gender of the referent.

### 20 Agreement

The agreement affixes are given in Table 20.1. (Almost) all forms can occur as prefixes, suffixes, and infixes.<sup>2</sup> The only exception to this rule is the zero marking for masculine singular agreement, which is only possible in the prefixal position (see below for examples). Verbs (except for copulas) and adjectives have prefixes; the other agreement targets have suffixes or infixes. The agreement slots for prefixes, suffixes, and infixes are obligatorily filled for all targets that have them (i.e. all agreement targets with agreement slots always exhibit agreement).

Table 20.1: Agreement affixes in Sanzhi


As Table 20.1 shows, there are fewer distinctions in the plural than in the singular, because masculine and feminine are united in human plural agreement. In addition, human plural is conditioned by person: first and second person plural agreement controllers are marked with *d*, third person with *b*. This phenomenon is also found in other Dargwa varieties, Archi, Ingush, and Chechen (see, e.g. Chumakina et al. 2007 and Corbett 2012: 239–251 for analyses of Archi) (25).

The prefix for masculine singular is *w-*, but it is (optionally) deleted when it occurs between vowels or in initial position when followed by the vowels /i/ or /u/. Deletion of /w/ between two vowels leads to vowel lengthening when the two vowels have the same quality, e.g. *a-w-ax-an=da* (neg-m-go-ptcp=1) > *aːxanda* 'I will not go' (vs. *a-r-ax-an=da* for female speakers), or the vowel quality changes according to the standard sandhi rules. For instance, *a-w-irχ-ud* (neg-m-be.able.ipfv-1.prs) > *a-irχud* > *erχud* 'I cannot' (vs. *a-rirχ-ud* for female speakers) (see §2.6 for morphophonological rules). When occurring in initial position before *i* the prefix *w-* is optionally omitted, e.g. *∅-ik'-ud*/*w-ik'-ud* (msay.ipfv-1.prs) vs. *r-ik'-ud* (f-say.ipfv-1.prs) 'I say'. Before *u* the deletion is obligatory, e.g. *∅-uq-un*/*\*w-uq-un* (m-go.pfv-pret) vs. *r-uq-un* (f-go.pfv-pret) 'I went'.

There are two agreement domains for gender agreement, the noun phrase and the clause, which follow two different rules. Within the noun phrase, modifiers agree with the head in gender and number independently of the case marking on the head (11)- (14) (see §21.1 for the syntax of noun phrases).


<sup>2</sup>There are only two agreement targets that have infixes, namely the quantifier *li<b>il* 'all' (10) and a variant of the standard copula *ca<b>i*. The form *ca<b>i* is used by a few speakers of Sanzhi in free variation with the much more common form *ca-b*. The quantifier *li-b-il* is diachronically complex and the gender marker is rather a suffix added to a stem *li*- and followed by the referential attributive suffix -*il*.

20.2 Combined Gender/number agreement


Note that within the noun phrase as well as within the clause, gender agreement with a noun modified by a numeral other than *ca* 'one' is semantically based, i.e. it is plural, although the noun itself does not bear an overt plural suffix. See §20.2.2 below for another example and §21.1.2 for number marking and agreement within the noun phrase.

Within the clause, the agreement controller is most commonly the argument in the absolutive, though it is not necessarily overtly present in the clause. This rule applies independently of polarity, TAM features, and clause types, i.e. it is found with all finite and non-finite verb forms including various nominalized verb forms (participles, masdars). Examples (15–18) illustrate monovalent predicates agreeing with the S argument.


In (19–22) bivalent predicates are presented. Example (19) contains a canonical transitive predicate. The agreement on the verb is controlled by the P argument. Other predicates behaving the same as canonical transitive verbs with respect to agreement are affective predicates with experiencers arguments in the dative or ergative and stimulus arguments in the absolutive case (20) (see also §19.1.8 for more information on bivalent affective predicates). Sentence (21) illustrates an extended intransitive predicate whose argument in the absolutive is the agreement controller. In (22) a ditransitive predicate is given that agrees with its T argument.

20 Agreement


In the antipassive construction, agreement is also controlled by the absolutive, which is now the agent (23) (see §19.2.1 for a detailed account).

(23) it that χabur-t-a-l story-pl-obl-erg r-ux-ul f-tell.ipfv-icvb ca-r cop-f ' She is telling stories.'

Gender agreement with other than absolutive arguments is also attested. It is not very common, but corpus examples can be found. The non-absolutive arguments controlling the agreement are either ergative agents or experiencers in the dative. This phenomenon is discussed in detail in §20.2.4.

In complement constructions in which the complement clause functions as the absolutive argument of the matrix predicate the agreement affix *b* is used in case of local agreement of the matrix predicate with the complement clause (24). This can be interpreted as default agreement, because in Sanzhi predicates that do not govern any argument in the absolutive case and therefore do not have a syntactic agreement controller predominantly take the agreement marker *b* (see below). Alternatively, we can say that the matrix verb agrees with the nominalized complement clause. Nominalization of any linguistic items results in nominals belonging to the neuter gender and therefore the matrix predicate must take *b-*.

(24) [nišːa-la 1pl-gen baliqː-e fish-pl le-d-ni exist-npl-msd nišːa-la 1pl.obl-gen erk'ʷ-li-cːe-d] river-obl-in-npl b-alχ-ul=de n-know-icvb=pst '(S/he/they) knew that there were our fish in our river.'

Sanzhi Dargwa, like many other Dagestanian languages, also has the option for longdistance agreement where the gender/number agreement on the matrix verb is controlled by the absolutive argument of the complement clause. Long-distance agreement occurs rather infrequently in the Sanzhi corpus because there are only few agreeing

20.2 Combined Gender/number agreement

matrix predicates and the respective complement constructions are not very often used. Therefore, the precise rules specifying its distribution still need to be studied. In (25) the complement clause contains an intransitive predicate whose single argument is suppressed due to co-reference with the overt argument of the main clause. Nevertheless, it controls agreement on both predicates. More examples of long-distance agreement and references to the literature on East Caucasian languages can be found in §24.4.

(25) nišːij 1pl.dat d-ikː-ul=de 1/2pl-want.ipfv-icvb=pst [d-isː-ij] 1/2pl-cry-inf 'We wanted to cry.' (E)

If the clause does not contain an agreement controller because it is lacking an argument in the absolutive, then mostly the default affix *b* is used:


The same happens with the verb *b-us-* denoting precipitation phenomena (e.g. rain, snow). This predicate governs one single argument marked with the ergative (28). The identical phenomenon is observed in the neighboring Icari Dargwa variety (Sumbatova & Mutalov 2003: 155), but apparently not in Standard Dargwa.

(28) marka-l rain-erg b-us-ul n-rain.ipfv-icvb ca-b cop-n 'It is raining.'

Occasionally, not *b-* but *d-* is used as default agreement exponent. This mainly concerns some compound verbal predicates that consist of a bound stem that is not a nominal, and a light verb (§19.1.7). For instance, in (29) the verb is a compound consisting of the verbal part *b-ulq-* with the meaning 'direct' and a first part *simi*, and the agreement is always *d-*. Another example is the phrase with which one wishes a good day (17).

```
(29) dam
      1sg.dat
               simi
               anger
                     d-ulq-u
                     npl-direct.ipfv-prs.3
      'I am angry.'
```
In addition to verbs also items bearing the essive case and the directional are agreement targets within the clausal domain. All essive cases in Sanzhi Dargwa as well as in other Dargwa varieties are expressed by adding a gender/number suffix to one of the spatial suffixes (§3.4). Thus, in the verbless sentence in (30), the noun bearing the spatial case suffix *-cːe* in the second clause agrees with the omitted absolutive argument

### 20 Agreement

that is identical to the argument in the preceding clause. Both clauses represent copula constructions with an adverbial predicate (first clause) and a nominal predicate (second clause) respectively. Similarly, (31) shows two spatial adverbs agreeing with the absent absolutive argument.


However, it is possible and occasionally attested in the corpus that gender markers of spatial adverbials show default agreement rather than agreement controlled by the absolutive. For instance, in (32) the omitted absolutive argument is female, as can be seen from the agreement on the verb, but the directional adverbial exhibits default agreement. Similarly, in (33) the agreement controller is the masculine singular noun phrase at the end of the clause, but the adverb in clause-initial position has the neuter singular suffix.


'That person needs to be there in the morning at 8 o'clock.'

Another agreement target is the concessive converb of *b-iχʷ-* (pfv) 'be, become, be able, can', which is used in concessive clauses and, when the verb follows interrogative pronouns, for the formation of free-choice indefinite pronouns (see §4.6.3). Since *b-iχʷ*is a verb with an agreement slot, the indefinite pronouns can, in principle, agree. Mostly they have default agreement, but they can also deviate from this pattern, for instance by being controlled by the absolutive argument. Thus in (34), *biχʷarra* could be replaced by *∅-iχʷarra* which would represent agreement controlled by the omitted absolutive subject. At the present moment I do not have enough data to explain this variation.

(34) daˁʡaˁna secret w-irx-ul m-become.ipfv-icvb hi-la-k'a who.obl-gen-indef b-iχʷ-ar=ra n-be.pfv-cond.3=add qili home '(He was) hiding at the house of whomever.'

20.2 Combined Gender/number agreement

### **20.2.2 Semantic agreement and other peculiarities**

Semantic agreement refers to cases in which the morphosyntactic feature values of the agreement target do not match the formal features of the controller. Instead, the agreement matches some semantic properties of the controller. Many instances of semantic agreement are number or gender mismatches. In Sanzhi Dargwa, such examples are found with gender/number agreement exponents on verbs where the agreement controller is an NP containing numerical quantifiers. A noun modified by a numeral normally does not take a plural suffix, but it requires plural agreement on the verb (35), (36).


A further example of semantic agreement occasionally occurs in fairy tales in which the acting personas are animals. In such cases mostly the inherent gender of the nouns is used, i.e. neuter, but sometimes the referents are treated as if they were human beings and thus masculine agreement affixes appear. In (37), the verb of speech has the neuter singular prefix in accordance with the natural gender of the referent, a wolf. But the verb in the quote shows masculine singular agreement and thus the referent has been humanized.

(37) "du", 1sg b-ik'-ul n-say.ipfv-icvb ca-b cop-n "uškul-le school-loc w-aš-ib-il m-go-pret-ref akːʷ-a-di cop.neg-hab.pst-1 du" 1sg ' (The wolf) said, "I did not go to school."'

Agreement with conjoined noun phrases can partially also be treated as semantic agreement (§20.2.3).

Another deviation that cannot readily be explained as semantic agreement is represented by a special construction for words denoting time spans such as *dus* 'year', *bac* 'month', *saˁʡaˁt* 'hour', or *minut'* 'minute'. These words belong to the neuter gender (38) and all of them except for *bac* have a plural form. However, when they are used to express periods of time with the verb *b-ič-* (pfv) 'occur, be', then agreement is neuter plural (39).


### 20 Agreement

### **20.2.3 Gender/number agreement with conjoined noun phrases**

Agreement with conjoined noun phrases follows two strategies: either the conjoined noun phrase is treated like a noun marked for plural and thus controls plural agreement or there is agreement with the closest conjunct.

The first case can be treated as an instance of semantic agreement since the nouns are not morphologically marked for plural. The rules for this type of agreement with conjoined noun phrases are as follows: two nouns denoting human beings control human plural agreement (40), and two nouns denoting animals or objects control neuter plural agreement (41).<sup>3</sup>


' The owner took the horse and the donkey and put the load on them.'

When the first and second person singular or plural pronouns are conjoined with nouns the agreement for first and second person plural is used, i.e. *d* (42), (43).


When a noun denoting a human being occurs in conjunction with a noun denoting

an animal or object the agreement is human plural:

(44) zija=ra horsefly=add kax-ub kill.pfv-pret ca-b, cop-n χudec'an=ra Khuduc.person=add kax-ub kill.pfv-pret ca-w, cop-m b-ebč'-ib hpl-die.pfv-pret ca-b cop-hpl

' (The Sanzhi man shot) and killed the horsefly and he killed the Khuduc man, and (they both) died.'

<sup>3</sup>The same rules apply if one or both of the conjuncts are plural nouns.

20.2 Combined Gender/number agreement

(45) er look b-ik'-ul hpl-look.at.ipfv-icvb ca-b cop-hpl '(The boy and the dog) are looking.'

Plural agreement is sometimes even found in comitative constructions. There are two ways of expressing comitative roles. One is via the use of the comitative case (46), (48) and the other is via the use of the reflexive pronoun (47), (49) (§30.3). In both constructions normally the absolutive argument controls the agreement as the following two sentences show:


However, there are very few examples in which the comitative phrase is treated as a plural noun phrase and therefore controls plural agreement. Example (48) illustrates this for the comitative case, and example (49) shows the comitative construction with a reflexive pronoun. In the first example, the human plural agreement could be replaced with masculine singular *w*-. In the second example, the agreement is first/second person plural *d*- since the author of the quote is referring to himself and his wife, thus the sentence is a quote with an omitted matrix clause.

(48) a and hel that ʡaˁħ good durħuˁ boy cin-na refl.sg-gen qːuʁa beautiful xːunul-li-cːella woman-obl-comit er life b-iχ-ub hpl-be.pfv-pret ca-b cop-hpl hana-li-j=sat=ra now-obl-dat=as.much=add ' The good boy with this beautiful wife lived until now.'

(49) [he probably thinks] heχ-itːe dem.down-advz xːunul=ra woman=add ca-w=ra refl-m=add qːuʁa-l beautiful-advz ka-d-iž-ib down-1/2pl-be.pfv-pret d-iχ-utːel, 1/2pl-be.pfv-cond.pst ʡaˁħ-le good-advz b-určː-i n-find.ipfv-hab.pst.3

'If he would sit together with his wife well like this, this would be good.'

The alternative to plural agreement in conjoined noun phrases is called "closest conjunct agreement". Closest conjunct agreement has been demonstrated to exist in a number of East Caucasian languages (see e.g. Gagliardi et al. 2009 on Tsez, and Chumakina 2014 on Archi). It is possible with conjoined noun phrases that follow or precede the

### 20 Agreement

verb. In each case, the member of the conjunction that happens to occur closer to the agreement target controls the agreement instead of agreement with the noun phrase as a whole:

(50) duˁrħuˁ-la boy-gen b-už-ib n-stay-pret ca-b cop-n χːʷe=ra dog=add ʡaˁt'a=ra frog=add ' The boy had a dog and a frog.'

Gagliardi et al. (2009) show that in Tsez, agreement with the closest conjunct is only possible when the agreement controller is adjacent to the verb. This is not the case in Sanzhi. Example (51) shows that the noun *kulpat* 'family' controls the agreement on the preceding verb (hpl) even though the personal pronoun intervenes.

(51) heχ-tːu-b dem.down-loc-hpl han remember b-ič-ib hpl-occur.pfv-pret dam 1sg.dat kulpat=ra, family=add li<d>il all<npl> cik'al something hel-tːi=ra that-pl=add … ' And there I remembered my family and everything, …'

In the following two examples, the agreement affix *b* can either be interpreted as neuter and thus as instantiating closest conjunct agreement or as human plural agreement, i.e. semantic agreement with a noun phrase that is treated as a noun in the plural (52), (53). More specifically, in example (52), the verb shows closest conjunction agreement with the following noun *duˁrħuˁ* 'boy', and the agreement of the clause-final spatial adverb *b-i* 'into' is ambiguous. Similarly, in (53), the agreement suffix of the copula *ca-b* is also ambiguous and both noun phrases are equally close to the verb in terms of linear adjacency.

(52) il that alen-ni deer-erg qi-m-a-cːe-r horn-pl-obl-in-abl lak' throw w-arq'-ib m-do.pfv-pret ca-w cop-m duˁrħuˁ=ra boy=add χːʷe=ra dog=add hin-ni-cːe water-obl-in b-i n/hpl-in

' The deer threw the boy and the dog into the water on its horns.'

(53) zija=ra horsefly=add kax-ub kill.pfv-pret ca-b cop-n/hpl il that admi=ra person=add '(He) killed the horsefly and the man.'

### **20.2.4 Gender agreement with arguments in other than the absolutive case ("Deviant agreement")**

Surprising for East Caucasian languages is the fact that, in Sanzhi Dargwa, under certain circumstances the agreement in a simple clause can be controlled by arguments not in the absolutive case, but in the ergative or dative. These arguments can be present or absent from the clause. In the following, I will refer to this phenomenon as "deviant gender

20.2 Combined Gender/number agreement

agreement" or simply "deviant agreement". The agreement targets for which agreement with the ergative or dative is attested are the standard copula (54–57) and the existential/locational copulas (63) when they are used as auxiliaries in periphrastic verb forms and also the exponents of the essive case (61). It is mostly found in clauses with a number of analytic verb forms such as the compound present (54) or the resultative (57).

In my Sanzhi corpus agreement with non-absolutive arguments is not particularly frequent, but there are a few clear examples. The majority contains verbs of speech or cognition, in particular *b-urs-* 'n-tell' (54–56), but also a few other verbs (57). In all examples (54–57) the standard copula *ca-b* has an agreement suffix that differs from the agreement prefix of the lexical verb with which the copula forms an analytic verb form.


There are also a number of examples with dative experiencers that control gender agreement (58–60).


### 20 Agreement

There are very few corpus examples in which it is a spatial adjunct in the essive case that shows deviant agreement with an argument that is not marked for absolutive case. In example (61) the ergative first person pronoun is omitted, but it controls the masculine singular agreement on the clause-initial adverbial. More examples can be elicited; in (62) the lexical verb does not have an agreement prefix, so the agreement mismatch is not immediately obvious, but the absolutive patient *kiniga* 'book' is neuter singular and would require the suffix *-b* on the copula in case of non-deviant agreement.


Deviant agreement with the existential copulas can be elicited:

(63) χabar story b-urs-ul n-tell-icvb le-w/te-w/χe-w/k'e-w exist-m/exist.away-m/exist.down-m/exist.up-m il-i-l that-obl-erg 'He is telling a story.' (E)

Deviant agreement never occurs with agreement exponents that belong to the lexical part of the predicate (lexical root, preverbs), but only with copula-auxiliaries and clausal adjuncts. Furthermore, the controller is in the ergative or dative and functions as a agent or experiencer argument of the predicate. It cannot be in any other case. Ergative agents and dative experiencer arguments of transitive and affective verbs share many subject properties with absolutive arguments of intransitive verbs (Forker 2017; 2019b). By contrast, arguments that do not function as agents or experiencers and are marked by other cases lack subject properties and cannot function as agreement controllers.

Deviant agreement is unusual for East Caucasian languages, but has been documented for a number of Dargwa varieties, most notably Akusha (Standard) Dargwa (van den Berg 1999; Ganenkov 2018), Tanti Dargwa (Sumbatova & Lander 2014: 450–493) and Shiri Dargwa (Belyaev 2016; 2017a,b). The different authors have put forward various explanations and hypotheses concerning the syntactic and semanto-pragmatic properties of the construction. According to all authors, gender agreement with the ergative argument (but also with the absolutive or dative) is conditioned by information structure. In her account, van den Berg (1999) states that deviant agreement with ergative agents does not require any specific pragmatic conditions whereas agreement with patients in the absolutive highlights them. She further claims that absolutive patients controlling agreement are topical ("themes" in her terminology). Sumbatova & Lander (2014) refine this analysis and claim that topical arguments independently of their case marking control gender agreement. Sumbatova (2010) and Sumbatova & Lander (2014) write that deviant agreement with ergative agents is frequent in Tanti Dargwa narratives. They further show that deviant agreement can also occur in cleft constructions that express constituent focus.

20.2 Combined Gender/number agreement

When discussing sentences with ergative agreement with Sanzhi speakers and eliciting new examples, an effect on the information structure is noticeable. Absolutive agreement is always possible, so it is the deviation from this pattern that requires an explanation. Absolutive agreement is preferred in answers to constituent questions regarding the agent or the patient that have narrow focus (64a). By contrast, ergative agreement is readily available when the question is, for example, about the place in which the agent is located (64b).

(64) a. [Who is tearing out the carrots?]

Aminat-li žit'a gu-r-ha-b-ilt'-unne ca-b

Aminat-erg carrot sub-abl-up-n-tear.ipfv-icvb cop-n

'Aminat is tearing out a carrot from under the earth.' (E)

b. [Where is Aminat?] Aminat-li Aminat-erg žit'a carrot gu-r-ha-b-ilt'-unne sub-abl-up-n-tear.ipfv-icvb ca-r cop-f 'Aminat is tearing out a carrot from under the earth.' (E)

Constituent order and closeness to the agreement controller also play a role for deviant agreement. In sentence (65a) the controller occurs in sentence-initial position whereas the target, the copula, appears clause-finally. With such a constituent order agreement with a dative (or ergative) controller is highly marginal (although available in elicitation as (64b) proves). It becomes possible when the controller occurs next to the target, more specifically when it is following the target (65b). In fact, in all but one instance of agreement with an ergative or dative argument attested in the Sanzhi corpus the controller immediately follows the copula (58), (60). Furthermore, the controllers are expressed by pronouns (55), (60), or absent from the clause (57), (59)

(65) a. Rasul-li-j Rasul-obl-dat cin-ni refl-erg d-arq'-ib-te npl-do.pfv-pret-dd.pl han remember d-irčaq-ul npl-occur.ipfv-icvb ca-d cop-npl 'Rasul remembers what he had done.' (E) b. cin-ni refl-erg d-arq'-ib-te npl-do.pfv-pret-dd.pl han remember d-irčaq-ul npl-occur.ipfv-icvb ca-w cop-m Rasul-li-j Rasul-obl-dat 'Rasul remembers what he had done.' (E)

This seems to point to point to an explanation based on topicality as formulated by Sumbatova & Lander (2014), i.e., topical items control agreement. And more specifically, deviant agreement is only possible by topical controllers, because (i) pronouns and zero arguments are usually topical, and (ii) the position after the predicate is a frequent position for topical subjects in Sanzhi, and can also be used for contrastive topics (§27.2).

### 20 Agreement

Yet, this analysis must be rejected. My Sanzhi data are in accordance with Ganenkov (2018), who notices a number of problems with the "topic controller hypothesis". Most importantly, the hypothesis implies that in the majority of transitive clauses the patient must be topical, because it is far more common for the absolutive patient to control the gender agreement than for the ergative agent to control it. Such an assumption seems implausible. Ganenkov further shows that focal arguments or indefinite pronouns that cannot be topical nevertheless control gender agreement. His arguments can be replicated for Sanzhi Dargwa.

Furthermore, it is not clear for all corpus examples that the controller is really topical. For instance, in (66) the referent of the omitted ergative argument that controls the agreement has not been mentioned in the preceding context, apart from the use of the indefinite pronoun, because the speaker had a specific person in mind, but could not remember her name. Thus, the absent agreement controller in this sentence cannot really be called "topical".

(66) [To someone (= a woman whose name the speaker forgot) I said, well I will wash (my legs)] kʷi-r-sawtː-ul ca-r

in.the.hands-abl-tear.off.ipfv-icvb cop-f

' She takes (the jug) out of my hands (and washes my legs).'

Moreover, most topical agents or experiencers do not control gender agreement, as in (67). This sentence is the first main clause with a transitive predicate in the narrative. The agent argument, which refers to the protagonist of the story, is the sentence topic and has been omitted. It does not control agreement. Instead, the agreement in the main clause is controlled by the newly introduced patient argument, which is not topical under any account of topicality.

(67) [Once upon a time there was a girl called Patima. She was the oldest within her family. Once after the rain (she) went up to sweep in front of the house.] qʷaˁrš sweep b-irq'-an=qːel, n-do.ipfv-ptcp=when b-arčː-ib n-find.pfv-pret ca-b cop-n qix nut

'When she was sweeping, she found a walnut.'

Therefore, the topicality hypothesis as formulated by Sumbatova & Lander (2014) needs to be rejected, and for a final conclusion about the pragmatic functions of deviant agreement more research is needed.

From a syntactic point of view, the sentences discussed in this section lead to the question whether they are really counterexamples to the claim that gender agreement can only be controlled by nouns in the absolutive case. This would not be the case if it were possible to analyze them as biclausal. This means that the copula is the head of the superordinate clause and agrees with a non-overt absolutive argument that is coreferent with the ergative or dative argument in the subordinate clause. Such an analysis would motivate the pragmatic differences between absolutive and ergative agreement (64a), (64b), and it would also be consistent with the generalization that the prefixes can only agree with the absolutive argument.

20.2 Combined Gender/number agreement

This idea has been proposed by Sumbatova (2010) and Ganenkov (2018). Ganenkov observes that certain characteristics of deviant gender agreement, namely that it is restricted to the copula-auxiliary (as opposed to agreement prefixes of lexical verbs) and thus found only in periphrastic tenses, resemble biabsolutive constructions. In biabsolutive constructions, the agent agrees with the copula-auxiliary and the patient with the lexical verb (68). They have been described for many East Caucasian languages (see Forker (2012) and Gagliardi et al. (2014) for recent accounts) including Sanzhi's neighbor Icari Sumbatova & Mutalov (2003: 156), but are not attested in Sanzhi. In biabsolutive constructions, the agent is generally topicalized whereas the patient is pragmatically demoted and backgrounded. For biabsolutive constructions a biclausal analysis has been proposed (Kazenin 1998; Kazenin & Testelets 1999; Kazenin 2001): the agent in the higher clause controls agreement on the copula-auxiliary just like other intransitive predicates; the patient is located in the subordinate clause and thus only controls the agreement of the lexical verb (68).

(68) Icari Dargwa (Sumbatova & Mutalov 2003: 156) Murad Murad [mura hay d-utː-a-tːi] npl-mow.ipfv-prog-prog.cvb ca-w cop-m 'As to Murad, he is mowing hay.'

Ganenkov (2018) adopts the biclausal analysis for deviant agreement and poses an unexpressed absolutive argument higher in the clause that is co-referent with the ergative (or dative) argument and controls the agreement on the copula auxiliary. In other words, the initial ergative subject raises to the position of the higher absolutive subject (subjectto-subject raising) and controls gender agreement. By contrast, the expressed ergative argument is located in the subordinate clause as schematized in (69).

(69) \_i*(abs)* [Aminat-li<sup>i</sup> Aminat-erg žit'a carrot gu-r-ha-b-ilt'-unne] sub-abl-up-n-tear.ipfv-icvb ca-r cop-f

'Aminat is tearing out a carrot from under the earth.'

Ganenkov's proposal goes back to the suggestion by Sumbatova (2010) to analyze deviant agreement as backward control and thus also as having a bipartite structure. In backward control constructions, the subject-like argument of a complement-taking predicate is expressed in the complement clause and thus receives case marking from the embedded lexical verb. Nevertheless, the matrix verb shows agreement with the embedded controller (see (42) in §24.2.6 and also (116), (117) in §24.5). On the surface this looks like agreement with an oblique argument, but as Polinsky & Potsdam (2002; 2006) have shown for Tsez, it can be argued that the matrix verb contains a covert controllee in the absolutive case.

However, the biclausal approach seems to be problematic. As Forker (2012) demonstrated, a synchronic biclausal analysis for many biabsolutive constructions is not tenable. The same can be said for constructions with deviant agreement. As explained above, in natural texts the default position of the agreement controller is after the predicate.

### 20 Agreement

Since subordinate clauses cannot be discontinues or split up by constituents from the main clause, we thus would have to claim that the pronoun in (70) occurs to the right of the clausal boundary. However, topical pronouns following the predicate are common and there is no reason to assume that they are extraclausal constituents (e.g. no intonational break). Furthermore, in examples such as (56) not only the subject pronoun but also the complement would have to be treated as extraclausal.

(70) [χabar story b-urs-ul] n-tell-icvb ca-w cop-m il-i-l that-obl-erg 'He was talking (telling stories).'

In sum, neither the functional-pragmatic properties nor the syntactic properties of deviant agreement are settled. It seems that a synchronic biclausal analysis for deviant agreement poses problems (though a diachronic analysis may still be possible). Alternatively, we can argue that a binary opposition of monoclausal vs. biclausal constructions is too limited. We should instead refine our notion of clause by applying the model of Multivariate Typology (Bickel 2011; 2015). This would mean breaking up the notion of clause into a sensible number of variables by means of detailed language-specific studies. In a second step we can then check our data for clusters around potential categories such as "monoclausal" and "biclausal" constructions and determine whether the Sanzhi deviant agreement construction fits into one of these.

A typologically-informed account of deviant agreement needs to not only take into account what we know so far, but also non-verbal agreement controllers, more detailed information about word order and further aspects that have not been investigated yet. To the latter belong referential properties of the agreement controller such as animacy or humanness, since we know that some languages do not allow inanimate agents in biabsolutive constructions (Forker 2012).

### **20.3 Person agreement**

### **20.3.1 Introduction**

Like all Dargwa varieties, Sanzhi Dargwa has person agreement enclitics and agreement suffixes. When suffixes are used and when enclitics are used depends on the TAM forms of the verbs, which means that all verbs can, in principle, be used with person suffixes and with person enclitics (with the exception of the morphologically defective copula verbs, which can only attach person enclitics). Suffixes and enclitics follow the same agreement rules, but differ in their form and morphosyntactic characteristics. The origins of the Dargwa agreement systems including Sanzhi Dargwa remain opaque. Pronouns and auxiliaries have been proposed as possible sources but there are no reliable proofs (Sumbatova 2011: 147–158).

The form of the agreement suffixes varies depending on the TAM form. There are a number of different sets. They mostly resemble each other because (i) the third person is either unmarked or differs from the other persons in morphological make-up, (ii) the first

20.3 Person agreement

and third person are not differentiated for number, and (iii) only the second person has two distinct suffixes for the singular and the plural. Thus the person systems are rather reduced, with a clear opposition of speech act participants (first and second person) vs. third person.

The use of the suffixes is restricted to verbs, i.e. only verbs can serve as targets. The most common sets of person suffixes are given in Table 20.2, Table 20.3, and Table 20.4. Imperative and prohibitive suffixes are given here because of their resemblance with the optative paradigm (imperative) and the habitual present, haitual past and conditional paradigms (prohibitive), which suggests a diachronic relationship.

Person agreement is subject to clause-level conditions because not all verb forms of main clauses have person agreement markers. Certain forms with past time reference (e.g. the past progressive, the evidential past, and the evidential pluperfect) make use of the past enclitic, which is in complementary distribution with the person enclitics. Another factor is finiteness: almost exclusively verb forms in finite main clauses and in conditional clauses can be marked for person agreement. Thus, the masdar, converbs, and participles, when used in subordinate clauses, do not contain agreement markers (see (116) below for the subjunctive, which represents the exception to this rule).

In the habitual present, the realis conditional, and the past conditional, the person suffix for the first and second person is preceded by a stem augment vowel that is indicated with V in the Tables above. The vowel is either *i* or *u*. The same vowels are also used in the subjunctive and the prohibitive (116), and the same distinction (though without the stem augment vowels) is attested in the imperative. For one-place verbs *u* is the only vowel that is used. For two-place verbs the following distribution is observed:


This has been summed up in Table 20.5. The stem augment vowels are treated as part of the suffixes. They are not part of the stem. Therefore, they are not separately glossed, but written together with the TAM suffixes. As the Table shows, there is variation when both core arguments are speech act participants (i.e. first and second person). Based on my corpus data and on elicitation I do not have an explanation for the variation and thus my analysis is only preliminary and requires further research before a conclusion can be reached.

In the following, I will briefly illustrate the use of the stem augment vowels. Sentence (71) shows the habitual present first person of an intransitive verb (see also (75) below for another intransitive verb with the stem augment vowel *u*).

(71) ixʷle fast dawlači-w rich-m w-irχ-ud m-become.ipfv-1.prs ' I (masc.) become rich fast.'


Table 20.2: Person agreement suffixes in the habitual present and habitual past

Table 20.3: Person agreement suffixes in conditional forms


Table 20.4: Person agreement in the optative, imperative, and prohibitive


Table 20.5: Stem augment vowels for transitive and two-place affective verbs


20.3 Person agreement

Examples (72) and (73) illustrate the realis conditional with a person marker for second singular. In the first sentence, the stem augment is *u* but *i* would also be possible). In (73) there is a second person agent acting upon a third person, hence only *i* is allowed.


In sentence (74a), the habitual present illustrates a first person experiencer with a third person stimulus with the stem augment *-i* and (74b) shows the reversed scenario with the stem augment vowel *u*.

	- b. it-i-j that-obl-dat du 1sg či-w-ig-ud spr-m-see.ipfv-1.prs 'S/he will see me (masc.)'

Table 20.6 displays the agreement enclitics. As can be seen in this table, only the second person singular has a unique marker. For the third person there are no person markers. Instead, depending on the time reference of the clause and on the context, the third person is left unmarked, or some other marker appears filling the gap in the paradigm (e.g. the copula *ca-b*, which exhibits gender/number agreement or the suffix *-ne*). Person agreement enclitics are widely used throughout the verbal paradigm, e.g. in the compound present and past, the perfect, the preterite, the future, etc.

Table 20.6: Person agreement enclitics


The person enclitics belong to the predicative particles (§9.1). They are normally added to the predicate, but, just as other predicative particles, can also be used to express term focus (also called "constituent focus"). In this case, they are encliticized to the item in focus, which can be an argument or adjunct, such that agreement targets are not only verbs but can be also nominals, adverbs, or other items (Kalinina & Sumbatova 2007, Sumbatova 2013, Forker 2016b).

### 20 Agreement

Person suffixes and person enclitics are subject to the same syntactic alignment rules: S, A, P, and T (i.e. the theme argument of a ditransitive verb) control person agreement. Person agreement is obligatory and it is freely combinable with gender/number agreement because both agreement systems operate independently of each other. Only one argument can control the agreement. The alignment patterns for person agreement among the Dargwa languages vary to a substantial extent (see Sumbatova 2011 and 2013 for overviews). They are determined by the ranking of absolutive vs. ergative arguments, and in a number of varieties also by person hierarchies. The person hierarchies found are either 2 > 1 > 3 (e.g. Icari, Kajtag, Qunqi, and Xuduc) or 1, 2 > 3 (e.g. Akusha and Standard Dargwa, Chirag). In many varieties the hierarchies are combined with a ranking of grammatical roles: patient argument (absolutive) > agent argument (ergative) is found in Akusha and Standard Dargwa, whereas agent argument (ergative) > patient argument (absolutive) has been documented for Chirag, Kubachi, and Mehweb. In Shiri Dargwa, in contrast to the above mentioned varieties, there is a considerable amount of variation within the speech community, and Belyaev (2013) distinguishes three slightly different alignment systems. A similar conclusion can be drawn for Sanzhi. There is also a certain degree of intra- and inter-speaker variation.

### **20.3.2 Person agreement rules**

In clauses with monovalent predicates, only the S argument serves as controller. Examples of first and second person are given in (75–77) for verbal predicates and (78) for a copula construction.


celij whole d-aqil npl-much k'e-d, exist.up-npl či-d-uˁq'-uˁtːal spr-1/2pl-go.pfv-cond.2pl 'There is much there (i.e. the graveyard is large), if you go there.'


In the following examples, third person agreement with intransitive predicates is illustrated. The agreement exponent can be a suffix as in the examples of the habitual past

20.3 Person agreement

in (79). Example (88) shows the compound present for which the copula is used for third person agreement (whereas in the third or second person a person enclitic would occur, see (77)). Other analytic tenses such as the preterite do not make use of the copula for the third person (but employ person markers for the first and second person) (80). In the copula construction in (81) also the copula is used. Finally, verb forms such as the compound present that in declarative main clauses require a copula for the third person omit the copula in questions with interrogative enclitics (82). This is possible because the interrogative enclitics belong to the predicative particles, which fulfill copula-functions, among other things (§9.1).

(79) Habitual Past

daže even hex-tːi dem.up-pl dubur-t-a-cːe mountain-pl-obl-in t'ama sound ha-d-aš-iri up-npl-go-hab.pst.3

' The sound went even to the mountains.'


The same rule applies to extended intransitive verbs, i.e. verbs that have one argument in the absolutive and another one marked with the dative or a spatial case. Thus in (83) and (84), verb forms with first and second person markers occur; in (85), the preterite is used, which lacks a marker for the third person.

(83) Habitual past, conditional past tiliwizur-ri-j television-obl-dat er look r-ik'ʷ-a-di, f-look.at.ipfv-hab.pst-1 či-d-ig-ul spr-npl-see.ipfv-icvb r-iχ-utːel f-be.able.pfv-cond.1sg ' I would watch TV if I were able to see.'

(84) Compound present xːunul-li-sa-r woman-obl-ante-abl uruχ fear ∅-ik'-ul=de=w? m-say.ipfv-icvb=2sg=q 'Are you afraid of your wife?'

20 Agreement

(85) Preterite iltːi those qːačuʁ-e bandit-pl kːač touch a-b-ič-ib neg-hpl-occur.pfv-pret il-i-j that-obl-dat ' The bandits did not touch him.'

There are a number of monovalent predicates that lack absolutive arguments and have only dative arguments. In §20.2.1the consequences for gender agreement were discussed. These predicates cannot control person agreement, and instead the third person is always used (86), (87). A number of weather predicates only have ergative arguments, and likewise they only exhibit third person agreement (88).


There are other monovalent predicates that are compound verbs, and that from a morphological point of view contain petrified nominal arguments which in some cases control gender agreement and in others do not. These behave just like any other monovalent predicate, i.e. the single argument controls the person agreement (89); see also (29) above.

(89) Future

dawaj let's (nušːa) (1pl) dum eating d-alt-an=da 1/2pl-let.ipfv-ptcp=1 ' Come on, we will eat.' (modified corpus example)

In clauses with bivalent verbs that are either genuine transitive verbs or affective verbs both arguments (i.e. agents/experiencers, and patients/stimuli) can control person agreement, but only one argument at a time.

In clauses with only third person arguments we find the respective agreement markers for the third person:

(90) Habitual present it-i-j that-obl-dat it that či-w-ig-u spr-m-see.ipfv-prs.3 'S/he sees him.' (E)


If we have one third-person argument and one first or second-person argument the latter controls the agreement, independently of the grammatical relation, i.e. these clauses are governed by the person hierarchy 1, 2 > 3.

(93) Realis conditional 1 > 3 di-la 1sg-gen w-at m-send k-aʁ-ille down-do.pfv-cond.1 ... ' If (I) send my (brother) ...' (94) Realis conditional 2 > 3 wot well tak so het that hetːu-b-a there-n-dir sa-qː-itːel hither-carry.pfv-cond.2sg urkːa between ... ' like this, if (you) put this here, in the middle ...' (95) Habitual present 2 > 3 iž this di-la 1sg-gen ucːiq'ar cousin χe-w, exist.down-m Mamma-la Mamma-gen Q'urban Kurban b-ik'-ul, hpl-say.ipfv-icvb ašːi-j 2pl-dat w-alχ-atːa m-know.ipfv-prs.2pl ' I have a cousin called Mamala Kurban, you know him.' (96) Habitual past 3 > 1 šːamχal Shamxal acːi-l uncle-erg r-ik-a-di f-lead.ipfv-hab.pst-1 ' Uncle Shamkhal led me (fem.).' (97) Preterite 3 > 2 tːura outside ha-qː-ib=de=w up-carry.pfv-pret=2sg=q u 2sg iž this miskin-ni? poor-erg

'Did the poor man pull you out?'

In clauses with two speech act participants, in principle either participant can control agreement independently of its grammatical role. All four logically possible combinations can be obtained in elicitation with male and female Sanzhi speakers of various ages:

### 20 Agreement

	- b. 1 > 2, patient controls agreement du-l 1sg-erg u 2sg kʷi in.the.hands urc-an=de keep.m.ipfv-ptcp=2sg ' I will keep you (masc.) in my hands.' (E)
	- b. 2 > 1, patient controls agreement u-l 2sg-erg du 1sg kʷi in.the.hands urc-an=da keep.m.ipfv-ptcp=1 ' You will keep me (masc.) in my hands.' (E)

There is only one example of such a scenario in my corpus (100), and it shows agreement controlled by a second person agent.

(100) Realis conditional 2 > 1 du 1sg w-erc-aq-utːe m-save.pfv-caus-cond.2sg ' (I give you a lot of money) if you save me.'

It seems that there is a slight tendency in elicitation for speakers to prefer the examples in which the second person controls the agreement, be it a second person agent, patient, experiencer, or stimulus (101). Nevertheless, Sanzhi Dargwa is unlike Icari in having also first person agreement controllers in clauses with only speech act participants (102), (103). The same variation in person alignment has also been attested for Shiri Dargwa in Belyaev (2013).


20.3 Person agreement

(103) Habitual present 1 > 2 nušːa-l 1pl-erg ušːa 2pl d-uˁrq-itːa 1/2pl-hit.ipfv-prs.2pl ' We hit you.' (E)

To sum up scenarios with two speech act participants functioning as agents and patients, I can only state that my preliminary analysis did not yield more precise results and that the variation is an interesting problem, which requires further testing.

The alignment patterns, including the described variation, seems to slightly change for predicates with three arguments. As said above, recipients, addressees, beneficiaries, and other arguments that are not agents or patients never control person agreement. In sentences with first person agent arguments and second person patient arguments, both agent and patient can control the agreement. This means we either have hierarchical agreement with 2 > 1 as in the second version of (106), or agreement with the agent as in the first version of (104) and in (105). If the agent is a second person pronoun, only this argument can control the agreement (106). Agreement controlled by the first person patient argument is ungrammatical. This is in contrast to examples with twoplace predicates such as (99b) which has a first person patient argument controlling the agreement. At the present moment I do not have any explanation for why the agreement patterns of three-place verbs seem to diverge from those of two-place verbs and the few examples I was able to elicit do not allow me to draw and further conclusions or to develop hypotheses, so this topic must be left for future research.

> / /

```
(104) Preterite 1 > 2
```
du-l 1sg-erg u 2sg Madina-j Madina-dat či-w-až-aq-ib=da spr-m-see.pfv-caus-pret=1 či-w-až-aq-ib=de spr-m-see.pfv-caus-pret=2sg ' I showed you to Madina.' (E)


u-l 2sg-erg du 1sg Madina-j Madina-dat či-w-iž-aq-an=de=w spr-m-see.ipfv-caus-ptcp=2sg=q / / \* či-w-iž-aq-an=da=w? spr-m-see.ipfv-caus-ptcp=1=q ' Will you show me to Madina?' (E)

### 20 Agreement

As soon as a speech act participant co-occurs with a third person agent or patient, it is the speech act participant that controls the agreement (107), (108). In (108) the verb also has a gender/number agreement prefix that is controlled by the absolutive argument. Thus, we can clearly see that person and gender/number agreement function independently. In clauses with only third person agents and patients we find third person agreement, even if we have first or second person recipients (109).

(107) Preterite 1 > 3

du-l 1sg-erg a-cːe 2sg-in cik'al-la thing-gen tiladi request a-b-arq'-ib=da neg-n-do.pfv-pret=1 'I did not ask you anything!'

(108) Realis conditional 2 > 3 hetː-a-la those-obl-gen durħuˁ-li-j boy-obl-dat hej this rursːi girl r-ičː-itːe f-give.pfv-cond.2sg ' if you give the girl to their son'

(109) Preterite 3 > 3 sumk'a bag di-cːe 1sg-in b-ičː-ib n-give.pfv-pret '(He) gave me his bag.'

The obligative (§14.1.5), the obligative present (§14.1.6) and the experiential I and II (§14.2.6) diverge from the TAM forms discussed so far in their agreement rules because they do not make use of any person hierarchy, but person agreement is always controlled by the patient (in clauses with two-place verbs). Thus, example (110) shows the experiential I with the third person patient serving as agreement controller. The use of the first person enclitic is ungrammatical. Sentence (111) from the corpus illustrates the experiential II and does not have an overt agent, but an overt first person patient, which controls the agreement on the verb. All examples also show that the patient also controls the gender marking on the lexical verb (and on the copula if there is any), which is expected and in accordance with the gender agreement rules. Furthermore, the cross-categorical suffix on the lexical verb agrees in number with the patient: a singular patient requires the suffix -*ce* (112) or -*il* (111), (7); a plural patient requires -*te* (8), (110), (113).


20.3 Person agreement

(112) du-l 1sg-erg julʁan-ni-gu-w blanket-obl-sub-m w-arq'-ib-ce m-do.pfv-pret-dd.sg ca-w cop-m ' I gave birth to (my son) under a blanket.' [modified corpus example]

In (113) the modal enclitic *=q'al* functions as a predicative marker such that the copula can be omitted (which is accordance with the general rules for the omission of the copula).

(113) arc money lukː-an-te=q'al give.ipfv-ptcp-dd.pl=mod du-l 1sg-erg ' I have to give the money back.'

The four TAM forms are analytic and make use of either the cross-categorical suffixes *-ce* or *-il*, which, among other things, are used for the formation of referential attributes that have the morphosyntactic properties of nominals (e.g. headless relative clauses). These constructions therefore resemble biclausal constructions, but a detailed investigation is needed before any conclusions can be made.

In sum, person agreement in Sanzhi is conditioned by person, by grammatical relations, and by TAM forms. Only agent, experiencer and patient arguments control person agreement, and the relevant hierarchy is 1, 2 > 3. In clauses with two speech act participants either argument can control agreement, even if it seems that there is a small preference for second person controllers because in elicitation speakers seem to accept them more readily. Variation in clauses with speech act participants is also found with respect to the stem augment vowels (Table 20.5), whereas in all other scenarios no variation is allowed.

Person agreement does not interact with polarity. However, the form of the verb and therefore the form of the agreement marker may change, e.g. in a copula clause with a first or second person subject and present time reference, the person enclitics given in Table 20.6 are used; if the same clause is negated, the negated forms of the copula to which person suffixes are added occur (114).

(114) du 1sg kːuš-le hungry-advz akːʷa-di cop.neg-1 'I am not hungry.' (E)

As was mentioned above, non-finite verb forms mostly cannot take person markers. For example, the adverbial clause in (115) headed by a converb lacks an agreement marker. Only the finite verb in the main clause shows person agreement.

(115) [hel-tːi that-pl d-ičː-ib-le] npl-give.pfv-pret-cvb qili home sa-ač'-ib=da hither-come.pfv-pret=1 '(They) gave them (to me) and (I) went home.'

The only exceptions are certain complement clauses exhibiting control. They can be headed by an infinitive or alternatively by the subjunctive which has the suffix -*Vtːaj* for the second person and *-anaj*/*-araj* for the third person (see §18.1.4). There is no suffix for the first person and instead the infinitive is used. Relevant examples are:

	- b. ašːij 2pl.dat b-ikː-ul=de n-want.ipfv-cvb=pst [d-isː-utːaj 1/2pl-cry-2subj / / d-isː-ij] 1/2pl-cry-inf 'You wanted to cry.' (E)
	- c. il-tːa-j that-pl.obl-dat b-ikː-ul=de n-want.ipfv-cvb=pst [b-isː-araj hpl-cry-3subj / / b-isː-ij] hpl-cry-inf 'They wanted to cry.' (E)

## **21 Noun phrases and postpositional phrases**

This chapter addresses the properties of noun phrases (§21.1) and postpositional phrases (§21.2) including their constituent order. Nominal modifiers that occur outside the noun phrase (i.e. so-called "floating modifiers") are only briefly discussed (§21.1.4). For a detailed treatment of floating modifiers see §27.1.2 and §27.1.3.

### **21.1 Noun phrases**

### **21.1.1 Introduction**

The noun phrase (NP) with an overt head noun minimally consists of a nominal head that can be optionally modified. Nominals that occur as heads of NPs are all sorts of pronouns, common nouns, personal names, or nominalized items. Noun phrases can be coordinated (§26.1), and it is possible to have noun phrases with nominalized modifiers instead of head nouns (§21.1.5). Noun phrases are head-final and thus modifiers precede the head. Usually, nouns are most prone to be modified, but occasionally other nominals can also take modifiers. Noun phrases admit the following types of modifiers:


Noun phrases occur in core argument and adjunct position (e.g. as instruments or temporal adjuncts). They can also be used as predicates in copula clauses (§22.2) and as complements in postpositional phrases (§21.2).

Within the noun phrase, there is gender and number agreement. Targets for gender/ number agreement are a number of vowel-initial adjectives (1) and adjectival quantifiers that have agreement affixes (8), (36), any items bearing essive cases (5), (9), and participles of verbs with gender prefixes (20). Number agreement without gender agreement is found with demonstrative pronouns (12), and with modifiers that have the crosscategorical suffix -*ce* (singular) (4) vs. -*te* (plural) (7) (§9.6.1). There is no case agreement between modifiers and the head noun within the noun phrase, and case suffixes can only occur on the head noun.

21 Noun phrases and postpositional phrases

(1) ca one b-uqen n-long q'aˁli branch 'a/one long branch'

### **21.1.2 Lexical, phrasal, and clausal modifiers in noun phrases**

Sanzhi does not have a special class of articles. Instead, demonstrative pronouns (§4.2) and the numeral *ca* 'one' can be used in the function of definite and indefinite articles respectively, but often their interpretation is ambiguous between definite article and demonstrative pronoun, or indefinite article and numeral (2).

(2) hel-tː-a-lla that-pl-obl-gen hel-tːu-w that-loc-m le-w=de exist-m=pst ca one žahil young durħuˁ boy 'There with them was also a/one young boy.'

Nouns frequently occur without the numeral *ca* or a demonstrative pronoun and receive an indefinite or definite interpretation from the context. Personal names can take demonstrative pronouns when they occur as topical noun phrases (3), but normally they occur without demonstratives.

(3) hila=de whose=pst il that Ražab? Razhab 'Whose was that Razhab?' (i.e. from which family)

There are no special possessive pronouns. Personal pronouns (first and second person), demonstrative pronouns (third person) or reflexive pronouns (third person) marked by the genitive are used instead (9), (16). Most adjectives distinguish between a short bare form and a long form with the suffix *-ce* (plural *-te*) (§5.2). The use of the suffix is obligatory for adjectives in predicative function and for attributive adjectives that do not occur in their canonical prenominal position (see §21.1.3 below). Adjectives used as attributes to nouns can occur with (4) or without the suffix (5), the omission of the suffix being far more frequent than its presence.


Nouns modified by numerals are not marked for plural (as are nouns modified by the interrogative word *čum* 'how many'), although they trigger plural agreement on demonstrative pronouns (12), adjectives (7), and also within the clause, i.e. on verbs, postpositions or adverbs (6). This means that not only semantically, but also syntactically, the

21.1 Noun phrases

noun phrase is plural. Modifying adjectives in noun phrases can occur in the stem form or with the cross-categorical suffix, which has a singular form -*ce* and a plural form -*te* (§9.6.1). In noun phrases with numerals as modifiers, the plural form must be used when the noun has plural reference (7)


Younger speakers occasionally use the plural suffix on the noun in noun phrases with numerals as it is done in Russian (8). In Sanzhi noun phrases that contain quantifiers such as *b-aqil* 'much, many' (36), the noun has also to be marked for plural.

(8) ʡaˁbal three durħ-ne boy-pl sa-b-eʁ-ib-le, hither-hpl-go.pfv-pret-cvb quˁr-be pear-pl li<d>il all<npl> d-alc'-un npl-gather.pfv-pret 'Three boys came and gathered all the pears.'

Nominal modifiers in NPs can be appositions, nouns bearing the suffixes *-il* or *-ce*, and nouns marked for spatial cases (9) or for the genitive case (6). If plural nouns bear the genitive they can have a non-specific interpretation, not referring to a specific possessor but restricting the meaning of the head noun to a certain type (10).


Appositions consist of two (or more) nouns with the same referents immediately following each other. As indicated by their modifiers and case marking, appositional phrases behave like a single noun phrase. They most frequently consist of a proper name and a kinship term, namely *acːi* 'uncle' (11) or *azi* 'aunt' or of nouns denoting different types of roles such as social roles (e.g. *zunra* 'neighbor', *saldat* 'soldier', *tuχtur* 'doctor', *busurman* 'Muslim', *jatim* 'orphan') (12) or gender roles (*xːunul* 'woman', *murgul* 'man'). The role-denoting nouns modify more general terms such as *admi* 'person, man', *insan*

### 21 Noun phrases and postpositional phrases

'person' (12), and some other nouns. Appositions not involving proper names resemble compounds (12) since the meaning of the second nouns is restricted through the meaning of the preceding noun.


Furthermore, the nominals cannot be separated, their order is rather fixed, and only the second nominal is marked for case (13)

(13) du 1sg tuχtur doctor Maˁħaˁmmad-li-cːe Mahammad-obl-in ∅-ik'-ul=da m-say.ipfv-icvb=1 ... 'I say to Doctor Mahammad ...'

Other nominals occurring in appositive phrases are reflexive pronouns that function as emphatic reflexives (§29.1.2) or pronouns with quantifiers, e.g. *nušːa lidil* 'we all'.

The noun phrase can contain an equative expression that either contains the adjective *miši*, which governs the dative (14), or the particle *ʁuna* (15), (16) that changes to *ʁunab* when the suffix *-ce* is added.


Phrasal and clausal modifiers of noun phrases are postpositional phrases (15), (17), (18), relative clauses (20), (22) (see also Chapter 23), purpose clauses (19) and other clauses (20). Although not obligatory, the suffix *-il*, which is used to form nominal modifiers from various parts of speech (§9.6.2), has been added to the postpositional phrase in (17). The suffix can be omitted as example (15) shows, but then the postpositional phrase can either function as a modifier of the noun or as a clausal modifier. Sentence (20) illustrates the use of a commemoration formula that is headed by a verb in the unmarked optative. The unmarked optative can be nominalized and take further case markers. Therefore, the optative clause can be interpreted as a clausal modifier to the noun *xːunul* 'woman'. But it can also be interpreted as a parenthesis that is not syntactically related to the noun. The optative clause is followed by a short relative clause consisting only of a participle.

21.1 Noun phrases


### **21.1.3 The structure and order of constituents within the noun phrase**

Noun phrases can be complex consisting of several modifiers, but in natural texts three or more modifiers are not very common. Because the modifiers themselves can be complex, the actual number of words in noun phrases might easily reach five or more. The usual order of modifiers is displayed in (21), which shows that the noun occupies the right-most position in the noun phrase.

(21) demonstrative/genitive

numeral / quantifier phrase or clause adjective demonstrative / genitive / quantifier appositive noun **head**

Testelets (1998a: 654) has proposed for other East Caucasian, particularly Avar-Andic and Tsezic languages, that the order of modifiers in the noun phrase reflects "the degree of their contribution to the identification of the NP's referent." If this generalization is taken to express a tendency, rather than a strict rule, it can be applied to the Sanzhi noun phrase as well. For example, genitive modifiers and demonstrative pronouns demonstrate this tendency since the former, typically closer to the head noun than the latter (see (23) above), make a larger contribution to identification of referents, although the reverse order is possible (29).

The modifiers can be divided into two groups: (i) quantifiers, demonstratives, and genitives, which specify the quantity, definiteness, and referentiality of the noun phrase and thus anchor it in the discourse, and (ii) adjectives, nominals, phrases, or clauses, which denote qualities and provide further information about the properties of the referent. The two groups are not only distinguished by their semantics, but also by their position within the noun phrase. Members of the first group, i.e. quantifiers, demonstratives, and

### 21 Noun phrases and postpositional phrases

genitives, can occur in two different positions as (21) shows: either in phrase-initial position or (almost) immediately before the head noun. Furthermore, they can float off from the head noun and occur outside the noun phrase. This will be discussed in detail below.

The examples in (22–26) illustrate complex noun phrases. Relative clauses are given in square brackets. More examples in this chapter illustrate other constellations of nominal modifiers in complex noun phrases: numeral + adjective (1), demonstrative + numeral + noun (12), adjective + adjective + genitive (10), and relative clause + relative clause + numeral (6).


Short adjectives are subject to a positional restriction: they can be separated from the head noun only by other adjectives (short or long ones, which bear the attributive suffix -*ce*) (36) or by appositive nouns (27). All other modifiers need to precede short adjectives (36), every other order being ungrammatical (28).


21.1 Noun phrases

There are two positions in which genitives, especially genitive pronouns, occur (21). They are either placed in phrase-initial position (29) or, more frequently, directly preceding the head (22). As mentioned above, the common order of demonstrative pronouns and genitive pronouns is for the demonstrative to precede the genitive (23), but the reverse order is also attested (29).

(29) genitive pronoun + demonstrative hež-tː-a-la this-pl-obl-gen hej this χabar story 'this story of theirs'

There can be scope differences associated with certain constituent orders. For instance, the genitive pronoun in (30) can scope over the entire noun phrase or it can be restricted to the immediately following noun. Similarly, the interpretations of (31a) and (31b) differ slightly.

	- three 1sg-gen book ' three books of mine' (E)

If the genitive is a genuine possessor rather than a genitive which denotes the material and can be retrieved from the context, the head noun is frequently omitted. The second genitive in (32) lacks an overt head. Headless genitive-marked nominals can also, just like other modifiers in noun phrases, be nominalized and take case suffixes (§21.1.5).

(32) šːan-t-a-lla fellow.villager-pl-obl-gen deq'a grain d-elq'-un npl-grind.pfv-pret ca<d>i; cop<npl> tːura outside šːan-t-a-lla fellow.villager-pl-obl-gen d-elq'-un npl-grind.pfv-pret ca<d>i cop<npl> ' The villagers' grain has been ground; (The grain) of the people from other villages has been ground.'

In content interrogative questions with the meaning 'which other X', the order is adjective-first (*cara* 'other'), followed by the interrogative pronoun *ce* 'what', which belongs to the modifiers of the first group (33):

(33) cara other ce what χurejg food d-irχ-u=ja npl-be.ipfv-prs=q nišːa-lla? 1pl-gen 'What other food of ours exists?'

### 21 Noun phrases and postpositional phrases

### **21.1.4 Floating modifiers**

In principle, all modifiers except for demonstrative pronouns can float off to positions outside the noun phrase (i.e. they can be extraposed). We find genitive pronouns (33) and genitive nouns (34) as well as quantifiers (38), adjectives (42) and relative clauses (43) outside the noun phrase. However, the extent to which floating is possible and which positions in the clause are common or available for floating modifiers depends on the type of modifier. The greatest freedom is enjoyed by floating genitives because they do not require any special marking when they are extraposed. Floating quantifiers require overt case marking, and floating modifiers of the second group (nominals, adjectives, relative clauses, etc.) need special additional marking (in addition to case).

Floating modifiers are semantic modifiers of nouns, but do not occur within the corresponding noun phrase; they are separated from the noun phrase by other constituents. In the following, the morphosyntactic properties of different types of floating modifiers will be discussed in more detail. The information-structural properties are mainly discussed in §27.1.2 and §27.1.3.

The most common modifier that occurs detached from the noun is the genitive, and this has been noticed for other East Caucasian languages, the first detailed analysis of floating genitives being made by Creissels (2013) for Akhvakh. Floating genitives often follow the noun and occur at the end of the clause after the verb (6), (34), but sometimes the reverse order is found in which case the genitive precedes a clause-final noun (35). In the latter example it seems that it is the head noun which has been extraposed while the two genitives simply remain in their canonical position.

(34) lamusː-e=ra respect-pl=add ha-d-erχː-ur up-npl-fulfill.pfv-pret ca-d cop-npl χalq'-la people-gen

'The people's respect also finished. (i.e. people do not show respect any more.)'

(35) nišːa-lla 1pl-gen sunglan-t-a-lla Sanzhi-pl-obl-gen le-b exist-n χabar story ' We Sanzhi people have a story.'

Example (36) shows a relatively complex noun phrase that functions as the subject of the existential copula clause. The genitive pronoun and the quantifier following the copula are semantically associated with the noun phrase, but have been dislocated to the right of the copula. A possible explanation for this might be that the noun phrase would be otherwise quite complex and difficult to interpret. More examples of floating genitives and a detailed discussion of their information-structural interpretation can be found in §27.1.2.

(36) [hel that kumek-le help-loc ha-b-ač'-ib up-hpl-come.pfv-pret žahil] young durħ-ne boy-pl le-b=de exist-hpl=pst hel-tː-a-lla that-pl-obl-gen b-aqil hpl-much ' They had many young boys who had come to help.'

21.1 Noun phrases

Some corpus examples of floating quantifiers can also be found: (8), (36–38). Example (37) illustrates that postnominal modifiers can be interpreted contrastively, in particular in elicited, context-free sentences, but as (38) shows, a contrastive reading is not obligatory.


Quantifier floating will be illustrated through the use of the quantifier 'all', which contains a gender/number infix agreeing with the noun it modifies. If the modified noun bears the absolutive, the quantifier can occur in various positions other than the canonical position before the noun (39b–39d).

	- b. *sːa rursbe libil agur uškulle*
	- c. *sːa rursbe agur libil uškulle*
	- d. *sːa rursbe agur uškulle libil*

If the modified noun occupies the A function and has non-absolutive case marking, then the quantifier can, when it bears the same case suffix as the noun it modifies, only occur in positions other than the canonical position preceding the verb. Otherwise the sentence becomes ungrammatical (40c). The reason is that in a noun phrase, case marking can only occur once, namely on the head noun. If the quantifier appears directly before the noun, it is part of the noun phrase and can not be case-marked.

	- 'All girls used to make carpets.' (E)
	- c. \* *libilli rursbal t'amsne dirq'i*

For other grammatical relations, the restrictions are more severe. Quantifiers of addressee arguments floating away from their canonical position are rarely acceptable even if they bear the same case marking as the noun they modify. Sentences such as (41) are marginal. It is possible to make a short break before the quantifier, which is then interpreted as right dislocated, and the translation is 'Mother told the stories to the women, to all (of them).'

### 21 Noun phrases and postpositional phrases

(41) ?? aba-l mother-erg χabur-te story-pl xːun-r-a-cːe woman-pl-obl-in li<b>il-li-cːe all<hpl>-obl-in d-urs-i npl-tell.pfv-hab.pst

'Mother told the stories to all women.'

Floating quantifiers are occasionally found in texts, but all examples have head nouns in the absolutive case (8), (36).

In contrast to genitives and quantifiers, modifiers of the second group, that is, adjectives, postpositional phrases, or relative clauses, need special marking when they occur in a position outside the noun phrase, either immediately following it or further extraposed to the right. There are only a few corpus examples of floating modifiers in a position before and at the same time outside a noun phrase (35) (see also §27.1.3).

There are two markers: the suffix *-ce* (plural -*te*) (§9.6.1) and the suffix -*il* (§9.6.2). These markers optionally occur on modifiers within the noun phrase as (4), (7), (20) and (26) show, but for extraposed modifiers the use is obligatory. Both markers are crosscategorical suffixes and are used for the formation of referential attributes, which morphosyntactically behave like nominals. Their use is similar, but -*ce* and -*il* can only express singular referents, whereas -*te* requires plural referents. The suffix *-ce* is used with adjectives (42), spatial modifiers in the essive case (=postpositional phrases), and occasionally relative clauses; the suffix *-il* is mainly used with relative clauses (43). The same two markers are employed when no head noun is present in the clause and the items, which would otherwise be used as modifiers, are nominalized and take over the function of nouns (§21.1.5). Furthermore, adjectival roots are obligatorily marked with -*ce* when they are employed in predicative function (§5.2).


As was mentioned for postnominal quantifiers above, all corpus examples of floating modifiers occur in clauses in which the noun to which they semantically belong appears in the absolutive case. This means that it does not have overt case marking. If the noun is case-marked, it is not sufficient to add the special marking in form of the suffix -*ce* or -*il* to the modifier (44), but the modifier must also take case marking identical to the case of the noun (45).

(44) \* it that sa-jʁ-ib hither-come.m.pfv-pret tuχtur-ri-šːu doctor-obl-ad ʡaˁħ-ce, good-dd.sg wahi-ce-lli-šːu bad-dd.sg-obl-ad a-ag-ur neg-go.pfv-pret (Intended meaning: 'He went to a good doctor, he did not go to a bad one.') (E)

21.1 Noun phrases

(45) it that sa-jʁ-ib hither-come.m.pfv-pret tuχtur-ri-šːu doctor-obl-ad ʡaˁħ-ce-lli-šːu, good-dd.sg-obl-ad wahi-ce-lli-šːu bad-dd.sg-obl-ad a-ag-ur neg-go.pfv-pret 'He went to a doctor who is good; he did not go to a bad one.' (E)

Case marking of modifiers that occur within the noun phrase is ungrammatical (see §9.6.1.2 for an example).

Kazenin (2002) analyzes floating modifiers in the East Caucasian language Lak and in the unrelated languages Circassian and Nogai. He distinguishes between simple splits in which the nominal modifier is fronted and thus occurs in a position to the left of the noun phrase, and inverted splits when the modifiers follow the noun phrase. Inverted splits always require nominalization of the modifier independently of case marking; simple splits require nominalization whenever the head noun of the noun phrase and the modifier are overtly case-marked, which concerns all cases except for the absolutive. He concludes that nominalized modifiers do not form a single constituent with the noun phrase, which is in line with the analysis adopted in this grammar.

Floating adjectival modifiers have restrictive semantics and imply the existence of another, contrasting referent. For instance, the shampoo in (42) is identified because the modifying adjective restricts the interpretation to expensive shampoos, which might at the same time be contrasted with cheap ones although this remains open within the context of the utterance. However, not every extraposed modifier is contrastive. A discussion of the pragmatic functions of floating modifiers and more examples can be found in §27.1.3.

### **21.1.5 Nominalized modifiers used as head nouns**

When the noun is absent, the item occurring in the rightmost position in the noun phrase undergoes nominalization and functions as the head noun. In principle, all modifiers listed in §21.1.1 can be nominalized and then used as heads of a noun phrase in argument position or as adjunct. Depending on their use in the clause, they take further case suffixes.

Demonstrative pronouns, numerals, and other quantifiers do not require additional morphology when used as nouns (except for oblique stem marking and case suffixes if required by their function in the clause). Adjectives (46), pronouns (47), and nouns marked for the genitive (48) or spatial cases, postpositional phrases, and relative clauses (49), and other clauses take either the suffix *-ce* (*-te*) or the suffix *-il*. Thus, (47) shows a nominalized genitive pronoun being suffixed with *-ce* and thus able to take over the function of the S argument of the verb *b-ič-ij*. Contrast this with (32), which illustrated the genitive whose head noun has been omitted. Nominalized modifiers take case suffixes and often also oblique stem markers just like any other nominals (46), (48).

(46) χːula-ce-li-j big-dd.sg-obl-dat ʡaˁħ-le good-advz ʡaˁq'lu mind b-alχ-u=w? n-know.ipfv-prs.3=q

'Does an older (person) know it better (or a young guy)?'

21 Noun phrases and postpositional phrases


However, headless relative clauses in which the verb bears the modal participle suffix *-an* and that function as absolutive arguments without any further case marking, frequently occur without *-ce* or *-il* (see §23.4 for examples).

Nominalized modifiers can themselves be modified. In (50), a nominalized adjective is modified by a participle, and in (51) the relative clause, which consists of only one participle, is preceded by a demonstrative pronoun.


### **21.2 Postpositional phrases**

Postpositional phrases consist of a postposition that is preceded by a noun phrase. The noun phrase can be complex containing modifiers, quantifiers or determiners (52). Postpositions govern various cases, most notably the genitive and a few spatial cases (52), (53). Non-spatial postpositions mostly govern the absolutive case. See Chapter 8 for more examples of postpositions and postpositional phrases.


Postpositional phrases are always head-final, so it is ungrammatical for a postposition to precede the noun. However, all spatial postpositions also occur as adverbials and/or 21.2 Postpositional phrases

spatial preverbs without any additional morphology. Thus, it is not always easy to identify which part of speech a relevant item belongs to. For instance, the postposition *b-i-b* 'in, inside' governs, among other cases, the loc-essive, loc-lative or loc-ablative (§8.1.9), as shown in (53). However, there is also a spatial preverb *b-i* and an adverbial *b-i* with the same meaning. In example (54), what looks like a stranded postposition *w-i(-w)* is instead the spatial adverb. Example (55) illustrates the preverbal use. No case marked noun phrase is preceding *w-i*.


## **22 Simple clauses including copula clauses and grammatical relations**

This chapter addresses the basic structural properties of simple main clauses, including declarative, interrogative, imperative and optative clauses that express different illocutionary acts, namely assertions, questions and directives. These three clause types contrast mainly by their morphological marking (verbal suffixes, predicative particles), rather than by differences in constituent order or intonation. Simple clauses can be divided into simple verbal clauses with verbs other than the copula (§22.1) and copula clauses (§22.2). This division cross-cuts the division into declarative, interrogative, imperative and optative clauses. Grammatical relations and the notion of subjecthood are analyzed in §22.3. Word order at the clausal level and information-structural properties of simple clauses and other types of clauses are analyzed in Chapter 27.

### **22.1 Simple clauses headed by verbs other than copulas**

This section discusses simple clauses with verbal predicates, which can be declarative, interrogative, imperative and optative clauses. The main focus of the section is on declarative clauses. The other clause types will be briefly mentioned at the end of the section. The constituent order in simple clauses is discussed in §27.2.

Verbal clauses obligatorily consist of a verbal predicate. All other constituents (arguments and adjuncts) can be left out. By contrast, declarative copula clauses can be formed with non-verbal predicates and without finite verb forms if they contain predicative particles (§22.2).

In general, the number of arguments, their semantic roles as well as their case-marking, depend on the valency classes of the verbs and on whether further valency-changing operations such as antipassive or causativization have been applied. Valency patterns of predicates can be divided into one-place (monovalent), two-place (bivalent) and threeplace (trivalent) patterns (see §19.1 for a detailed overview). I will follow the macrorole approach as presented in Bickel (2011) and Bickel et al. (2015) and distinguish between bivalent predicates between an A argument (the argument with the most agentive properties) and a P argument (the argument with the least agentive or most patientive properties). Three-place verbs have, in addition to the A argument, a G argument and a T argument. The G (goal-like) argument is more stationary than the T (theme-like) argument, and, in contrast to T, might be receiving an experience or being exposed to an experience. The label S will be used for the single argument of intransitive verbs and for absolutive arguments of extended intransitive verbs.

### 22 Simple clauses including copula clauses and grammatical relations

Cases used to encode the arguments are absolutive, ergative, dative, and to a lesser extent, genitive and spatial cases such as the ante-ablative or the in-lative. Gender agreement is controlled in most cases by the absolutive argument of the clause (§20.2). However, certain verb forms allow the ergative or the dative argument as controller (§20.2.4), while other clauses lack absolutive arguments and resort to default agreement (§20.2.1). Person agreement, which works independently of gender agreement, follows the hierarchy 1, 2 > 3, whereby only S, A, P, and T arguments can function as controllers. From the person hierarchy follows that speech act participants control the agreement if the clause has any speech act participants. In clauses with only first and second person arguments, either person can control agreement independently of their grammatical role, although there might be a small tendency for second person arguments to outrank first person arguments (see §20.3 for more information).

Declarative verbal clauses obligatorily contain finite verb forms, which exhibit the full range of TAM marking and gender and person agreement. Negation is expressed through a prefix or, if the verb form includes the copula as auxiliary, by means of the negative copula (§11.7). Arguments and adjuncts can be pronouns or full noun phrases (4), (5). Arguments that can be interpreted through the context are often omitted (1). Adjuncts can be instruments, companions or express temporal, spatial or other circumstances (§3.4.1, §3.4.2). The following examples illustrate simple verbal declarative clauses with verbs of the major valency classes: intransitive, extended intransitive, affective, transitive, and extended transitive.

Intransitive verbs (§19.1.2) have one single argument (S) in the absolutive case controlling gender, number and person agreement (1).

(1) cellij why r-isː-ul=de? f-cry-icvb=2sg 'Why are you (fem.) crying?'

Extended intransitive verbs are bivalent and have an S argument in the absolutive and a further argument either in the dative, in-lative, ante-ablative, or in other spatial cases (2) (see §19.1.4 for further examples).

(2) bec' wolf kːurtːa-j fox-dat er look b-erč'-ib-le, n-look.pfv-pret-cvb … 'The wolf looked at the fox, …'

Bivalent affective verbs (§19.1.8) have a dative-marked experiencer (A) and a stimulus argument in the absolutive case (P) (3).

(3) d-aqil npl-much qːihin-dex difficult-nmlz či-d-ič-ib spr-npl-occur.pfv-pret nik'a small ucːi-j brother-dat ' The little brother experienced many difficulties.'

Transitive verbs (§19.1.5) require an agent in the ergative (A) and a patient in the absolutive (P) (4).

22.2 Morphosyntactic properties of copula clauses

(4) hel-i-l that-obl-erg kːurtːa fox a-kax-ub neg-kill.pfv-pret 'He did not kill the fox.'

Extended transitive verbs (§19.1.6) add a further G argument marked by the dative or by spatial cases to the ergative A and the absolutive T (5).

(5) kːurtːa-l fox-erg b-ičː-ib n-give.pfv-pret hel-i-j that-obl-dat cin-na refl.sg-gen ʁez hair 'The fox gave him his hair.'

More detailed information on these and other valency classes can be found in §19.1.

Interrogative clauses are marked by interrogative enclitics, which also belong to the class of predicative particles. They are often but not always signaled by rising intonation. They usually contain finite verb forms, but it is also possible to encode interrogative clauses with certain non-finite verb forms and interrogative enclitics. See Chapter 28 for more information on their morphosyntactic properties and §27.2.2 for the constituent order, which is largely identical to the constituent order of declarative clauses (except for differences that are due to the information structure). Sanzhi has another type of interrogative clause with modal semantics. This type, which I call the *modal interrogative* contrasts to all other interrogative clauses since it makes use of a special verbal suffix. This suffix exists only for first person subject-like arguments (§17.4) and does not express tense. The modal interrogative is thus more like imperative and optative clauses than interrogative clauses.

Imperative, prohibitive and optative clauses contain verbs inflected with suffixes which do not express tense. Imperative and prohibitive clauses are restricted in their use to second person subject-like arguments, which can be overtly expressed as in declarative or interrogative clauses (§17.1 and §17.2). The optative does not show person restrictions, but cannot be formed from imperfective verb stems and is thus constrained in its aspectual value (§17.3).

### **22.2 Morphosyntactic properties of copula clauses**

Copula clauses are constructions prototypically containing a copula, a copula predicate and a copula subject. The copula can, but does not have to be, a verb. Genuine copulas are considered to be semantically empty (Pustet 2003: 5). According to this definition, Sanzhi Dargwa *ca-b* can be considered a copula as well as the person enclitics *=da* and *=de* and the past enclitic *=de*. In addition, Sanzhi has four locational copulas and a number of auxiliaries that also head copula clauses. Table 22.1 provides an overview of all items used in copula constructions with references to the sections in which more information about the morphology of the items is given. In the following section I will refer to all items in Table 22.1 as copulas whenever they are used in copula constructions. The items listed in the last part of the table are auxiliaries that are not entirely "semantically empty". They are included in the table and discussed in this section because they are used to express

### 22 Simple clauses including copula clauses and grammatical relations

TAM forms of copula clauses for which the copulas and predicative particles cannot be used since they lack those forms. Some of the items in Table 22.1 regularly combine in copula clauses. (I) The predicative particles can occur as enclitics on the copula *ca-b* (see below for examples), but not on the negative copula. They can also occur on the locational copulas (§22.2.2) and on the other auxiliaries because they are a regular part of certain TAM forms. (II) The copula *ca-b* and its negative counterpart are also used in a few TAM forms and thus can combine with most auxiliaries given in Table 22.1 (30). (III) The same applies to the locational copulas, which are also occasionally used in certain periphrastic TAM clauses and therefore can combine with the auxiliaries in copula clauses (§15.1).

Which copula item is chosen depends on the person of the copula subject, on the meaning of the copula construction, on the temporal reference of the clause, and on further categories such as mood, epistemic modality, illocutionary force, and clause type. The copula *ca-b*, the locational copulas, and the verb *b-el* are defective, and in the case of *ca-b*, also suppletive under negation. This means that they form only a very small set of verb forms compared with the forms available for regular verbs (including the other auxiliaries listed in Table 22.1).

Under certain circumstances, one of the three constituents can be absent. Copula clauses without a subject are, for instance, weather constructions (45) or other impersonal clauses (47) (§22.2.3). The copula predicate is present in most types of copula clauses, but often lacking in existential copula clauses (34) or when possession is expressed (41). The copula can only be omitted in copula clauses with third person subjects and predicative particles (§22.2.4).

The copula subject occurs in the absolutive case and controls agreement. It may be any kind of noun phrase (common noun, proper name, pronoun, etc.) or nominalized clause. The copula predicate, which bears the absolutive case or an overt case marker, may be a noun phrase, but it may also be an adjective, an adverbial, a postpositional phrase, or a nominalized clause; this partially depends on the copula item. Sentence (6) illustrates a copula construction with the first person pronoun in the absolutive as copula subject and a noun with its adjectival modifier as predicate. The past enclitic *=de* fulfills the copula function.

(6) nik'a small durħuˁ=de boy=pst du 1sg ' I was a little boy.'

The gender/number and person agreement is always controlled by the copula subject, not by the predicate (7), (12a), (13), (16), but not all copulas have gender agreement slots, for example the negative copula lacks an agreement prefix. In copula clauses that lack arguments in the absolutive case, the copula bears the default agreement suffix -*b* and there is invariable third person agreement (see §22.2.3 below for more details). This mainly concerns experiential constructions with affective predicates and experiencers in the dative case (8). One item used as a copula, the past enclitic =*de*, does not encode person (6) (see also §22.1 below for the general structure of simple clauses with other verbs than the copula).

### 22.2 Morphosyntactic properties of copula clauses


### Table 22.1: Items functioning as copulas


In copula clauses, where both the subject and the predicate are in the absolutive and are identical in gender (12), and in copula clauses, where the copula function is taken over by an item that does not exhibit agreement (6), it is impossible to determine the copula subject by means of agreement. But based on general information-structural properties, we can assume that the topical noun phrase is usually the subject, which in many cases is a pronoun (6), (18), (19). In a similar fashion, constituent order can be indicative.

### 22 Simple clauses including copula clauses and grammatical relations

The person enclitics, as well as the past enclitic, can optionally be accompanied by the copula (*ca-b*), and in such constructions the copula always serves as the host for the enclitic (9).

(9) iž-i-la this-obl-gen xːunul woman ca-r=de=w? cop-f=2sg=q ' Are you his wife?'

Furthermore, the copula and the person enclitic or past enclitic need to occur on the same host; they cannot be separated (10). If in the example below the copula at the end is omitted, the clause becomes grammatical with the meaning given in brackets below.

(10) \* du=da 1sg=1 ala 2sg.gen rucːi sister ca-r cop-f (Intended meaning: 'It is me who is your sister.') (E)

In contrast to the neighboring Dargwa variety of Icari (Sumbatova & Mutalov 2003: 138), the copula and the interrogative enclitics can co-occur in Sanzhi (11). According to Nina Sumbatova (p.c.), Icari is the only Dargwa variety found so far where the copula *ca-b* excludes all other predicative particles; in contrast, the situation that we have in Sanzhi Dargwa is common and attested in many other Dargwa varieties (e.g. Akusha).

(11) miši-l similar-advz ca-w=uw cop-m=q iχ dem.down iχ-i-j? dem.down-obl-dat 'Is this similar to him?'

In line with other clause types, the copula most frequently occurs in clause-final position. Subjects predominantly precede the predicate, such that we can assume that the first absolutive constituent is the copula subject and the second one, which is often the host of enclitical copula items, functions as the predicate in clauses with two absolutive constituents (12), (15a), (18). The standard third person copula *ca-b* is comparably strict in its requirement to occur in clause-final position in elicited assertions (12), though it might be followed by additional demonstratives that refer to the same item as the subject, and in questions from the corpus we also find copula subjects and predicates following *ca-b* (11). In principle, *ca-b* can occur on its own and make up a full clause, such that phonological dependency is ruled out as an explanation for the ungrammaticality of (12b).

	- b. \* *ca-w Murad ustːa*
	- c. \* *Murad ca-w ustːa*

22.2 Morphosyntactic properties of copula clauses

For other copula items, it is easier to find utterances with other than clause-final position of the copula, in particular when the predicative particles are used in copula function (6). Locational copulas can also occur in clause-initial position:

(13) a. ištːu-w here-m Murad Murad le-w exist-m 'Murad is here.' (E) b. *le-w ištːu-w Murad*

When used in copula clauses, the predicative particles can only occur on the head of the phrase that functions as the subject (14) or on the head of the phrase that functions as the copula predicate (14), (15a). Alternatively, they in case of coordinated constituents on the leftmost member of the coordination (17). They cannot be encliticized to any constituent that modifies the head (15b).

(14) [it that ʡaˁħ-ce]=de good-dd.sg=pst dalaj-či, song-nmlz cara other wahi-ce=de bad-dd.sg=pst

' The good one was a singer, not the bad one.' (lit. the other was the bad one) (E)

	- b. \* *Madina [darkːʷan=de rursːi]*

The only exception to this rule that I found so far are genitive modifiers: they can host predicative particles in copula clauses even if they do not occupy the functions of copula subject or copula predicate. The genitive noun in (16) modifies the following copula predicate. This is only possible in term focus constructions in which the host of the enclitic is the focal part of the clause. Furthermore, in term focus constructions, the predicative particles can, in principle, also be attached to adverbial modifiers such as spatial adverbials, but the properties of these constructions require further research.

(16) du 1sg sawχuz-la=da sovkhoz-gen=1 dajark'a, milkmaid kalχuz-la kolkhoz-gen akːʷa-di cop.neg-1 ' I am milkmaid of the SOVKHOZ, not of the kolkhoz.' (E)

In (17), the nominal predicate is a coordinated noun phrase that consists of three members, each bearing the additive enclitic as required for nominal coordination (§26.1). The person enclitic follows the last member of the nominal predicate.

(17) u 2sg [sːunkuˁq'=ra, liar=add deč-la drinking-gen χʷe=ra, dog=add bilʡuˁt'=ra]=de thief=add=2sg 'You are a liar, a drinking dog, and a thief!'

Copula constructions in Sanzhi can express identity, group membership, attribution, possession, benefaction, and also location and existence (see, e.g. Curnow 2000, Dixon 2010: 159–188).

22 Simple clauses including copula clauses and grammatical relations

### **22.2.1 Copula constructions expressing identity, group membership and attribution**

In copula constructions that express identity, group membership, and attribution, all items displayed in Table 22.1 except for the locational copulas are used. Predicates are mostly nominals, adjectives, or adverbials. Sentences (18) and (19) show copula constructions expressing identity, including deictic identificational clauses, and group membership.


If the subject is first or second person and the temporal reference is present time or there is no temporal reference because of habituality, then the person enclitics are used (17), (20). If the clause has past time reference, the past enclitic occurs (21).


Adjectives distinguish a short form from a long form. The long form contains the crosscategorical suffix *-ce* (plural *-te*) (§9.6.1). The short form is reserved for the attributive usage within noun phrases (in addition to compounding) (22); the long form is required for the predicative use (23).

(22) hel that r-uqna f-old xːunul=de woman=pst hel that ' She was an old woman.'

(23) ca one r-uqna-ce f-old-dd.sg ʁuna eq ca-r cop-f iχ, dem.down ca one žahil-ce young-dd.sg ʁuna eq 'One (fem.) is like old, one like young.'

Adverbials can also serve as predicates in copula clauses. Example (24) shows a nominalized participial clause in the subject position and an adverb as predicate of the copula clause.

22.2 Morphosyntactic properties of copula clauses

(24) hej this značit, thus hana now [nušːa-l 1pl-erg ka-b-išː-ib-il] down-n-put.pfv-pret-ref b-arx-le n-correct-advz ca-b cop-n ' This means, how we put (the pictures) down is correct.'

Often either the adverb or the adjective can be used as predicates in a copula clause. This leads to a slight difference in meaning that can be illustrated with the following examples (25) and (26). In case of an adjectival predicate, the copula clause refers to the quality denoted by the adjective that is ascribed to the referent of the subject (25). By contrast, if an adverb is used, it is the situation denoted by the copula clause that is attributed a quality and not the subject referent (26). Furthermore, in (25) the noun phrase *iž dus* functions as the subject, whereas in (26) it is a temporal adjunct and there is no overt subject.


The predicate can also be any kind of pronoun, for example a personal pronoun in the genitive (27), a demonstrative pronoun, or a personal pronoun in the absolutive (28), (29). For instance, example (27) demonstrates a situation in which the referent of the copula subject does not belong to the referent of the predicate, which is a genitive pronoun. It is also possible to add the cross-categorical suffix -*ce* to the genitive pronoun (*ala-ce*).


Negation is expressed by means of the negative copula *akːʷ-*, which does not have a gender prefix, but inflects for person and tense (27), (28). As was mentioned in the introduction (§22.2), the copula *ca-b*, the person enclitics, and the past enclitic can only express a restricted range of TAME forms. For instance, the imperfective verb *b-urkː-* has the meaning 'find', and is also used as an auxiliary with the epistemic meaning 'probably, be possible'. The latter use includes copula clauses (30). More examples of copula constructions with the verbs labeled "other auxiliaries" in Table 22.1 are given in §16.3.

22 Simple clauses including copula clauses and grammatical relations

(30) ču-la refl.pl-gen qu field b-urkː-ar n-aux.ipfv-prs.3 hel that 'That is probably their field.'

Note, however, that there are clauses with other auxiliaries that superficially look like copula clauses, but represent intransitive clauses This is because the auxiliary functions as an intransitive verb and not as a copula. For instance, the sentence in (31) contains an S argument in the absolutive case, *χalq'* 'people'. The item following it, *mic'ir* 'alive', is a short adjectival stem which cannot be used as adjectival predicate in copula clauses. As mentioned above, short forms of adjectives only occur as attributes within noun phrases or as part of compound verbs. The complex *mic'ir b-irχʷ-* has to be treated as one compound predicate with the meaning 'be/become alive', and therefore the construction does not contain a copula verb.

(31) χalq' people mic'ir alive b-irχʷ-i hpl-be.ipfv-hab.pst.3 'The people were alive.'

### **22.2.2 Location, existence, and possession**

Location and existence are generally expressed by specialized locational copulas (32), (34), (35), and in negative clauses by the negated copula with the gender prefix (33), which can optionally be preceded by a locational copula (34) (Table 22.1). If the negated copula does not have the gender prefix, it cannot express location or existence but only identity, group membership or attribution (27), (28).


Sentences expressing location contain spatial adverbials such as adverbs or nominals bearing spatial cases, which can precede or follow the locational copulas (35), (36). The standard copula *ca-b* can also be used (38), although locational copulas are normally preferred.

(35) χalil-la Khalil-gen ʡuˁmer Omar b-ik'-ul hpl-say.ipfv-icvb te-w exist.away-m Kaspisk-le-w Kaspisk-loc-m 'Omar's Khalil is in Kaspisk.'

22.2 Morphosyntactic properties of copula clauses

(36) tut-la mulberry-gen kːalkːi-l-gu-b tree-obl-sub-hpl le-b=de exist-hpl=pst hek' dem.up šːal-le-rka side-loc-abl ' They were under the mulberry tree, from that side.'

The person enclitics (37) and the past enclitic (21) can also be used in locational or existential clauses. In addition, the location copulas can attach the person enclitics:

(37) χadižat, Khadizhat čina-r=de where-f=2sg u? 2sg du 1sg Sanži-r=da Sanzhi-f=1 / / Sanži-r Sanzhi-f le-r=da exist-f=1 ' Khadizhat, where are you? I am in Sanzhi.' (E)

For locational copula clauses the standard copula *ca-b* can also be used (38), although locational copulas are normally preferred. The exact distribution of existential/locational copulas vs. the standard copula *ca-b* needs to be determined by future research.

(38) hež this tusnaq-le-w prison-loc-m ca-w cop-m ' He is in prison.'

The expression of possession implies the existence of the possessed item. This means that when talking about any types of possession that one has, be it objects or relatives, the locational copulas are used. In the unmarked case this is *le-b* (39). The possessor occurs in the genitive case.

(39) durħ-ne boy-pl le-b exist-hpl wec'al, ten weral seven rursːi girl ca<b>i cop<hpl> ʡaˁbal three durħuˁ boy ' (I) have ten children; they are seven girls and three boys.'

The following minimal pair illustrates the difference between the two types of copulas. The first sentence in (40) requires an identificational interpretation. It can, for instance, be used when showing and identifying the house. The more literal translation of the second sentence (41) would be 'With/at me there is a house.' or 'My house exists.' If the genitive pronoun is a predicate instead (27) or if other semantic components play a role, the other copulas are used.

	- 1sg-gen house exist-n 'I have a house.' (E)

Less common ways of constructing locational and existential copula clauses or copula clauses expressing possession are available by means of the other auxiliaries given

### 22 Simple clauses including copula clauses and grammatical relations

in Table 22.1. The example in (42) represents the traditional opening formula for fairy tales and is thus not a normal existential clause. The verb *b-ug-* can express indirect evidential semantics, which is often found in fairy tales. The verb *b-iχʷ-* (pfv)/*b-irχʷ-* (ipfv) 'be, become, happen, can' is used, among other things, to express epistemic modal constructions including different subtypes of copula clauses with a modal meaning (43). The sentence in (44) expresses not only past time reference, but also habituality and therefore also contains the auxiliary *b-irχʷ-*, because neither the standard copula *ca-b* nor the past enclitic =*de* can express this specific combination of temporal and aspectual meanings.


### **22.2.3 Copula clauses without a subject**

Copula clauses with temporal or spatial adverbials can occur without an overt copula subject. They only contain a copula predicate:


More generally, copula clauses with predicates expressed by manner adverbs do not require a subject, but can be impersonal (47), (48). The gender agreement affix in such clauses is invariably *b*, since this is the default agreement affix (§20.2). It is possible to add a dative argument fulfilling the semantic role of experiencer or beneficiary (49).


22.2 Morphosyntactic properties of copula clauses

(49) dam 1sg ʡaˁħ-le good-adv ca-b cop-n 'I am well.'

### **22.2.4 Copula clauses without a copula**

Copula clauses obligatorily require a copula item (Table 22.1), otherwise they are ungrammatical:


In copula constructions that have third person subjects and present time reference or habitual meaning, the copula can be omitted when one of the pragmatic predicative particles is used. This can be either one of the three interrogative enclitics if the copula clause is a question (polar question, content question, embedded question) (Chapter 28), or the modal enclitic *=q'al* (§9.4.2). This is possible because the modal enclitic and the interrogative enclitics belong, just like the person enclitics =*da* and =*de* and the past enclitic =*de*, to the predicative particles that can head finite clauses (§9.1). The following examples show the use of *=q'al* (52), a content question (53) and a polar question (54).


Similarly, in copula clauses that function as embedded questions or assertions expressing epistemic modality (uncertainty), the embedded question marker is used (55) and the copula is absent. Sentence (56) does not show an embedded question, but an epistemic uncertainty construction (§28.4).

(55) či-d-až-aq-a spr-npl-see.pfv-caus-imp hari let's [ce what q'ar=el]! herbs=indq ' Show what herbs these are!'

22 Simple clauses including copula clauses and grammatical relations

(56) hana now k'ʷel two dus=el year=indq 'Now it is probably two years (that have passed by).'

The sole use of a pragmatic predicative particle is impossible if the copula subject is first or second person. In such cases, the predicative person marker needs to occur before the question enclitic and cannot be omitted:

(57) uruχ-le=de=w? fear-advz=2sg=q 'Are you afraid?'

When two copula clauses are coordinated, the copula can be omitted in one of the clauses, usually the second clause (58) (see §26.2 for one more example).

(58) du 1sg ʡaˁħ-ce=da, good-dd.sg=1 u 2sg wahi(=de) bad(=2sg) ' I am good, you (are) bad.' (E)

### **22.3 Grammatical relations**

There are only a few studies on grammatical relations in Dargwa varieties so far by Nina Sumbatova (Sumbatova 2014; 2017), but there is a considerable amount of literature on grammatical relations in other East Caucasian languages, and there are works from a comparative perspective that include Dargwa (see Forker 2017 for a recent overview). Case studies of individual languages are often centered on the question whether the investigated language(s) is only morphologically ergative, or whether it also shows indications of syntactic ergativity (cf. Nichols 1980, Crisp 1983, Comrie et al. 2013). The majority of scholars state that ergativity is mostly restricted to morphology. Kibrik (1985; 1997; 2003) concludes that East Caucasian languages belong to the so-called "role-dominated" languages (Foley & Van Valin 1984: 123) in which the marking of arguments is semantically motivated.

In this section, I will briefly discuss the constructions (or variables) displayed in Table 22.2. The section does not present and discuss the data, but contains only crossreferences to the relevant sections in this grammar that contain data for most of the constructions given in the table. See Forker 2019b for data and analysis of more constructions such as complement control or quantifier floating and a detailed account of grammatical relations in Sanzhi. I will not analyze word order because word order on the clausal level strongly depends on the information structure and not on grammatical relations (§27.2).

We can identify three alignment types in Sanzhi Dargwa: ergative alignment, accusative alignment, and neutral alignment. Additionally, there are a number of constructions in Sanzhi that are not sensitive to grammatical relations. The most important constraint is the case-defined predicate class, that is, the distinction between canonical transitive,

### 22.3 Grammatical relations


Table 22.2: Grammatical relations in Sanzhi Dargwa

*<sup>a</sup>*But the evidence for affective verbs is inconsistent because imperative formation of affective verbs is often impossible for semantic reasons.

affective, extended intransitive, and other verbs. These valency classes of verbs are defined on the basis of case assignment patterns to the arguments, and not so much on the basis of the meaning of the predicates. In other words, cases have a high semantic load and the choice of one case suffix over the other largely depends on the semantic contribution of the cases. Thus, Sanzhi Dargwa confirms once more the fact that the semantic impact of cases for East Caucasian languages should not be underestimated.

Ergative alignment, labeled as S=P vs. A in Table 22.2, is basically found in the morphology, namely in the gender agreement and the case marking. There is a large number of bivalent and trivalent verbs that assign ergative case to their A argument, although not all bivalent and trivalent verbs belong to this class. Additionally, there are even more verbs whose S and P arguments trigger gender/number agreement because the arguments bear the absolutive case. Outside the realm of morphology there are almost no indications for ergativity, apart perhaps from quantifier floating (§21.1.4) and causativization (§19.2.2). Instead, accusative alignment (symbolized with S=A vs. P), neutral alignment (S=A=P) and no alignment (no grammatical roles identifiable) are found. Person agreement and reflexivization/reciprocalization are neutral since S, A, P and T are not distinguishable. They only behave differently from G, but this is not relevant for

### 22 Simple clauses including copula clauses and grammatical relations

the determination of grammatical roles. All four macroroles S, A, P and T can control person agreement (§20.3) or reflexive and reciprocal pronouns (Chapter 29) and thus we have neutral alignment. In contrast, relativization largely depends on pragmatics and a suitable context and is not sensitive to grammatical relations because a large variety of positions (S, A, P, G, T, other) can be relativized. Accusativity is found with imperatives because both S and A can be subjects in imperative clauses, but not P or any other position (§17.1). This is not surprising and frequently found in ergative as well as in accusative languages, and some authors do not consider imperatives to represent suitable test constructions for establishing grammatical roles, e.g. Dixon (1994: 131). Furthermore, complement control (§24.5) and conjunction reduction in clauses with the preterite converb show some accusative traits because S and A are always suitable controllers of arguments in complement clauses or converbal clauses, but P is largely excluded. Similarly, causativization can be analyzed as distinguishing between S/A on the one side and P on the other side because it is never the P or the T that is affected when bivalent or trivalent predicates are causativized (§19.2.2). P arguments remain unchanged (because P and T essentially have the same morphosyntactic properties), whereas S changes to P, and A changes to G under causativization, such that causativization can perhaps be taken as a further indicator of an S/A pivot. The antipassive is not a suitable test construction because its application is restricted to the class of transitive verbs, excluding all other valency classes, such that we cannot check how S would be treated.

To sum up, there is no justification for establishing a category of ergative subject that would comprise S and P, and thus Sanzhi Dargwa is only morphologically ergative. This claim is not surprising but supports what has been previously stated for the East Caucasian languages. The only indications for syntactic accusativity are complement constructions and causativization, which is not enough for establishing a category of subject comprising S and A as we know it from European languages. However, simple reflexive constructions and imperative could be viewed as further, though weaker indications for singling out S and A in contrast to P. At the basis of textual frequency even person agreement shows a tendency to occur predominantly with S and A controllers in natural texts because P arguments that are second person are relatively rare.

In this grammar, I use the term "subject-like" or even sometimes "subject" in order to refer to S and A arguments, whereas P arguments are called "object-like" or "object". This terminology has been chosen for reasons of convenience and familiarity. It has to be viewed against the background of the discussion of grammatical roles in Sanzhi as given in this Section.

## **23 Relative clauses**

### **23.1 Introduction**

Sanzhi Dargwa uses participles for the formation of relative clauses. Like other modifiers, relative clauses normally precede the head. There are two simple participles that consist of a suffix added to the verbal stem, and complex participles that make use of additional suffixes. The simple participles are the preterite participle (§18.1.2.1), which is identical in form to the preterite, and the modal participle -*an* (§18.1.2.2). For verbs whose stems exhibit the aspectual distinction, the preterite participle is almost exclusively used with perfective stems, whereas the modal participle occurs only with imperfective stems. The complex participles consist of the simple participles plus the cross-categorical suffixes *-ce* or suffix *-il* (§18.1.2.3). Furthermore, Sanzhi has a locative participle that is used when the head of the relative clause denotes a location (§18.1.2.4).

The simple participles and the complex participles derived from them express temporal relations. The modal participle occurs in relative clauses with non-past time reference (e.g. present, future, habitual) (1); the preterite participle occurs in relative clauses with past time reference (2).


Since the preterite participle is identical to the preterite itself, in a number of cases two interpretations are possible: a main clause that precedes another clause with argument sharing between the two clauses or a relative clause:

(3) raχːaz chain b-ertː-ib n-tear.pfv-pret χːʷe dog sa-r-b-uq-un ante-abl-n-go.pfv-pret ' The dog tore off the chain and left.' OR 'The dog who tore off the chain left.' (E)

The suffixes *-ce* (plural *-te*) and *-il* are used to form attributes that can denote referents. This means that items that bear these suffixes can be used as modifiers in noun phrases but also as predicates or as nominals. With respect to relative clauses, they are used whenever the relative clause occurs in a position that diverges from its canonical prenominal position. See §23.3 below for a discussion.

### 23 Relative clauses

him).'

Moreover, purpose clauses with nominal heads are structurally similar to relative clauses, but must contain an infinitive (or subjunctive) and can also be marked with *-ce* (4), (5). Participles are not allowed if the clause has a purposive meaning.

(4) [du 1sg qili home uq'-ij] go.m-inf pikri thought le-b exist-n 'There is the thought to go home.' (i.e. I have the thought) (E)

(5) sa-ka-b-išː-ib-le, ante-down-n-put.pfv-pret-cvb [cini-j refl.sg.obl-dat b-erkʷ-ij-ce]=ra n-eat.pfv-inf-dd.sg=add χe-b exist.down-n ca-w=ra refl-m=add ' He is there and (something) to eat for him, which has been placed in front (of

In the following, I will first show which positions can be relativized (§23.2), then analyze further semantic and syntactic properties of relative clauses (§23.3), and then briefly discuss headless relative clauses (§23.4).

### **23.2 Positions that can be relativized**

The following examples illustrate the positions that can be relativized. They are labeled with the semantic roles and the cases that nominals in that role bear in a main clause.

	- (6) il-tːi that-pl [bahla-l slow-advz ag-ur] go.pfv-pret durħ-ne boy-pl 'the boys who went slowly'
	- (7) [ħaˁšukː-a-r pot-loc-f rurq-aˁn] boil-ptcp hin-ni-cːe-r water-obl-in-f il that r-ebč'-ib f-die.pfv-pret ca-r cop-f ' In the water that boiled in the pot, she died.'
	- (8) sːa yesterday hextːu there.up [r-emkː-un] f-give.birth.pfv-pret xːunul-la woman-gen rursːi-šːu girl-ad ag-ur-re=de go.pfv-pret-cvb=pst ' Yesterday I went to the daughter (baby girl) of the woman who gave birth.'
	- (9) ca one ca one le-b exist-hpl [xːun-re-ja-r woman-pl-loc-abl naˁq hand či-r-a-ha-jsː-an] spr-abl-neg-up-take.pfv-ptcp murg-le man-pl 'There are those, the men who do not take away their hands from their wives.' (i.e. who beat them constantly)
	- (10) hež this [paltar clothes ic-an] wash.ipfv-ptcp mašin machine 'the washing machine' (lit. the machine that washes clothes)
	- (11) [du 1sg w-alχ-an] m-know.ipfv-ptcp rursːi girl 'the girl who knows me (masc.)' (E)
	- (12) [ħaˁq'-le very-advz qːuʁa-l beautiful-advz ca-l-li one-obl-erg ca one urk'i heart ha-b-eʁ-ib] up-n-go.pfv-pret sub-xːunul-li-j husband-woman-obl-dat miši-l similar-advz ca-b cop-hpl heštːi these ' They look like husband and wife who very well understand each other.'
	- (13) il-tːi that-pl [a-jkː-an] neg-want.m.ipfv-ptcp admi-li-j person-obl-dat 'to those people who do not love you (masc.)'

### 4. agent in the antipassive construction (absolutive)

	- (16) [iž-i-l this-obl-erg d-alc'-un] npl-gather.pfv-pret q'ar plant[npl] le-d=de exist-pl=pst ' There were plants that she gathered.'
	- (17) [umqːa-l whetstone-erg b-irq'-an] n-do.pfv-ptcp dis knife ' a knife that has been sharpened' (E)
	- (18) [ču-la refl.pl-gen χu-d-a-j dog-pl-obl-dat lukː-an] give.ipfv-ptcp χurejg food 'the food that is given to their dogs'
	- (19) [a-b-alχ-an] neg-n-know.ipfv-ptcp šahar-ri-cːe-w town-obl-in-m 'in a town that (you) do not know' (E)
	- (20) [du-l 1sg-erg kiniga book b-ičː-ib] n-give.pfv-pret durħuˁ boy 'the boy to whom I gave the book' (E)
	- (21) [du 1sg er=či look=on w-erč'-ib-il] m-look.pfv-pret-ref rursːi girl 'the girl at whom I looked' (E)
	- (22) [atːa-l father-erg qal house b-arq'-ib-il] n-do.pfv-pret-ref durħuˁ boy razi happy ∅-iχ-ub m-become.pfv-pret 'The son for whom father built the house got happy.' (E)
	- (23) caj-na one-time arg-ul go.ipfv-icvb le-d=da exist-1/2pl=1 [sːika bear b-ebč'-ib] n-die.pfv-pret qːatːa-r canyon.loc-abl či-d-a on-1/2pl-dir ' One time we went there through the canyon where the bear had died.'
	- (24) durħ-ne boy-pl ag-ur go.pfv-pret [q'ʷal cow luχ-na] cut.ipfv-ptcp.loc musːa place.loc ' The boys went to the place where cows are slaughtered.' (this refers to a specific place in Sanzhi)
	- (25) [paˁχ.paˁχ-le pakh.pakh-loc či-w-ax-an] spr-m-go-ptcp musːa-r place.loc-abl ' across the place where (you) go to the Pakh-Pakh (place name)'

12. spatial goal (lative cases)

	- (27) [du 1sg sa-jʁ-ib] hither-come.m.pfv-pret šahar town haraq-le far-advz ca-b cop-n 'the town from which I (masc.) came is far away.' (E)
	- (28) [du 1sg uruχ fear ∅-ik'ʷ-an] m-say.ipfv-ptcp durħuˁ boy 'the boy that I am afraid of' (E)
	- (29) d-iqː-a npl-carry.ipfv-imp dam 1sg.dat [χalq' people c'erx fat a-b-irχʷ-an] neg-hpl-become.ipfv-ptcp q'ampit'-e! chocolates-pl

'Give me the chocolates from which the people do not get fat!' (E)

### 23 Relative clauses

	- (30) [Madina-l Madina-erg χabar story b-urs-ib-il] n-tell-pret-ref xːunul woman di-la 1sg-gen rucːi sister ca-r cop-f 'The woman about whom Madina talked is my sister.' OR 'The woman to whom Madina talked is my sister.' (E)
	- (31) [Aminat Aminat ʁaj=či-r word=on-abl ka-r-ik'ʷ-an] down-f-say.ipfv-ptcp xːunul woman ' the woman about Aminat talks (gossips)' (E)

### 17. comitative

	- (33) [sub husband w-ebč'-ib-il] m-die.pfv-pret-ref xːunul woman r-isː-ul f-cry-icvb ca-r cop-f 'the woman whose husband died is crying' (E)
	- (34) [du-l 1sg-erg t'ult' bread ka-b-ičː-ib-il] down-n-cut.up.pfv-pret-ref dus knife qːut'a-l blunt-advz ca-b cop-n 'The knife with which I cut the bread is blunt.' (E)

Complements of postposition heading relative clauses are not found in the corpus and somewhat hard to elicit, but (35) shows an example. Another one is (30) above in the interpretation in which the head of the relative clause is the topic of conversation (which is normally expressed by a postposition, see §8.1.6).

(35) [durħuˁ boy sala front sa-ka-jcː-ur-il] ante-down-stand.m.pfv-pret-ref qal house nišːa-lla 1pl-gen ca-b cop-n 'The house in front of which the boy is standing is ours.' (E)

More complex constructions are also possible. For instance, the argument of a complement clause can function as the head of a relative clause (36). Similarly, arguments of adverbial clauses can be extracted in order to serve as heads of relative clauses (37). We can have multiple relative clauses embedded into each other (38).

<sup>1</sup> If in (33) the simple instead of the complex participle is used, the first clause is interpreted as a main clause preceding another main clause ('The husband died. The wife is crying.').

23.3 Other syntactic properties of relative clauses


It is possible to find examples in which the head noun does not bear any syntactic relation to the relative clause, i.e., in which it is impossible to argue that the relative clause contains a gap from the extracted head. This is widespread in case of head nouns with a very broad semantics such as *zamana* 'time' but there are also sentences with other head nouns. For instance, (39) illustrates a common construction that explicates the name of a person. The verb *b-ik'ʷ-* 'say, call', that is used in the relative clause, normally requires an absolutive argument that denotes its subject. However, in (39) the subject is absent because it is an impersonal construction, retrievable only from the context and from the fact that the human plural gender prefix is used. The gender agreement prefix is frozen and cannot be replaced by any other prefix. Instead of a complement clause, which is usually used together with the verb *b-ik'ʷ-* 'say, call', the name *Maħaˁmmadħaˁži* occurs. The head noun *durħuˁ* 'boy' does not fulfill an argument or adjunct role in the relative clause. This issue is discussed further in the following §23.3. Example (40) is similar in that the head noun *ħaˁz* 'game' is also not in a syntactic relationship with the verb in the relative clause 'hide'.


### **23.3 Other syntactic properties of relative clauses**

Relative clauses can have a nominal head or be headless (see §23.4 below). The head is normally a common noun, but it can also be a personal pronoun, an indefinite pronoun, a demonstrative pronoun, or a personal name. Thus, relative clauses can be restrictive or non-restrictive without any difference in their morphosyntactic form.

### 23 Relative clauses

	- (41) nu well [ašːi-j 2pl-dat či-b-ig-an] spr-n-see.ipfv-ptcp hel that b-urs-an n-tell-ptcp ca-b cop-n ' Well, you (pl.) will say what you see.'
	- (42) heštːi these akːu=w cop.neg=q heštːi these [b-iħ-ib-te]? hpl-wrestle.pfv-pret-dd.pl ' Are these not the ones who wrestle?'
	- (43) [Rassija-le Russia-loc ag-ur-il] go.pfv-pret-ref du-l 1sg-erg arc money d-irq'-ul=de npl-do.ipfv-icvb=pst 'I, who went to Russia, made money.' (i.e. earned money) (E)
	- (44) [t'ut'u throw.out b-arq'-ib-te] hpl-do.pfv-pret-dd.pl ča-qal who-assoc ∅-ik'ʷ-a-tːe? m-say.ipfv-hab.pst-2sg ' The ones who were thrown out, who did you say this was?'
	- (45) du-l 1sg-erg [mar truth haʔ-ib-il] say.pfv-pret-ref ča-k'al who-indef a-jt-an=da neg-beat.up-ptcp=1 'I will not beat anybody who told the truth.' (E)
	- (46) [irig last.year xːunul woman ka-r-iž-ib-il] down-f-be.pfv-pret-ref di-la 1sg-gen ucːi brother Zamir Zamir er life ∅-irχ-u m-be.ipfv-prs Agni-le-w Ogni-loc-m

'My brother Zamir, who married last year, lives in Ogni.' (E)

The verbal categories expressed in relative clauses are fewer than those expressed in main clauses. Due to the participles employed, the expression of tense is possible to a certain degree (1), (2) and negation is available (107), but person agreement and the marking of illocutionary force are excluded.

Within the noun phrase, to which a relative clause belongs, the relative clauses can be preceded and followed by other nominal modifiers such as demonstrative pronouns (42), possessive pronouns (46), quantifiers, adjectives, and others. See §21.1 for further information on the structure of noun phrases. They can be modified by adverbials just like adjectives (47).

23.3 Other syntactic properties of relative clauses

(47) [c'aq'-le very-advz b-al n-fit b-ič-ib] n-occur.pfv-pret musːa place ' a very fitting (= good) place'

In elicitation, Sanzhi speakers report that there is no difference in the meaning between a relative clause with a simple participle and a relative clause with a participle to which one of the the suffixes *-ce* or *-il* is added. Relative clauses with complex participles in the default position preceding the nominal head are relatively rare in the Sanzhi corpus. The following three sentences (48–50) illustrate the use of both suffixes with the preterite participle. In (49) the speaker is talking about the tools with which Sanzhi women used to cook and compares them with new appliances. Example (50) shows that the cross-categorical suffixes are also added to other parts of speech. In this sentence, -*il* appears on the preterite participle which because of the suffix acquires nominal properties can can function as referent with the meaning 'the one that has been put down'. The second appearance on the noun that is inflected for the in-essive case makes a referent with the meaning 'the one that is in the hand' out of a spatial adverbial


'The one had been put down was the one the one in the hand.'

The head noun in the vast majority of corpus examples follows the relative clause, but other positions are possible, too. Whenever the relative clause occurs after the head or separated from the head (preceding it or following it), the cross-categorical suffixes or the modal participle need to be employed and it can be argued that the relative clause is morphosyntactically not part of the noun phrase anymore. Note that such examples are not particularly frequent and most of the following examples stem from staged narrations and poems. Sentence (51) comes from the translation of a fairy tale from Standard Dargwa/Russian. Russian has postnominal relative clauses and the Russian word order has simply been copied. Similarly, (52) and (53) are translations from Russian. Example (54) and (55) come from spontaneous narratives, and (56) is part of a poem.

	- (51) dam 1sg.dat b-ičː-ib n-give.pfv-pret iž this maˁlʡuˁn-ni snake-erg [ca one kur-re pit-loc ka-b-iž-ib-il down-n-be.pfv-pret-ref dawla-či-w wealth-adjvz-m Ismaˁʔil-li-cːella] Ismail-obl-comit ' The snake that sat in a pit together with the rich Ismail gave it to me.'
	- (52) uc-be brother-pl [čar back b-iχ-ub-te hpl-be.pfv-pret-dd.pl d-ac' npl-empty nuˁq-b-a-cːella] hand-pl-obl-comit 'the brothers who came back with empty hands'
	- (53) dam 1sg.dat rursːi girl r-ikː-ul=da f-want.ipfv-icvb=1 [t'ams-ne carpet-pl d-irq'-an] npl-do.ipfv-ptcp ' I love the girl who makes carpets.' (E)
	- (54) il that c'ikuri bride r-arčː-ib f-find.pfv-pret ca<r>i, cop<f> [hana now xadi married ka-r-iž-ib-il] down-f-be.pfv-pret-ref 'They found this bride who had just married'
	- (55) [nušːa-l 1pl-erg hež this b-arq'-ib-il] n-do.pfv-pret-ref χabar, story [hextːu there.up ag-ur-il] go.pfv-pret-ref a-cːe 2sg-in b-urs-ul=da n-tell-icvb=1 χabar story daˁʡle as 'We tell you the story like the story that we put together, how we got there.'
	- (56) [[čum-k'al=ra how.many-indef=add dus year halk-un-te] catch.fire.pfv-pret-dd.pl di-la 1sg-gen c'a fire d-iš-aq-un-ce] npl-die.out.pfv-caus-pret-dd.sg 'the one that turned out my fire that was burning for so many years'

23.3 Other syntactic properties of relative clauses

Relative clauses are verb-final with very few exceptions, which can be obtained in elicitation or when translating stories from Russian. This property distinguishes them from other subordinate clauses that can more easily place the verb in positions other than the right edge of the clause. For instance, as just mentioned, (52) is part of a fairy tale that has been translated from Russian into Sanzhi. The relative clause not only follows the head, since this is the standard constituent order in Russian, but also contains a comitative phrase after the verb, so that the verb does not end up in the final position.

Within the relative clause the head is usually indicated by a gap. As is typical for East Caucasian languages, Sanzhi does not have relative pronouns. However, reflexive pronouns can, in principle, be used to express co-reference between an argument in the main clause and another argument or adjunct in the relative clause. In the corpus such sentences are not particularly frequent, but some examples can be found. In (57) the reflexive pronoun is a goal or experiencer that is coreferent with the omitted agent of the main clause. The participle in the relative clause is case-marked with the genitive because it expresses the topic of a conversation. In (58) the reflexive pronoun encodes the causer, and in (59) the agent. Example (18) above shows a reflexive pronoun functioning as possessor within the relative clause. In all unambiguous examples these pronouns are co-referential with the omitted subject argument in the main clause, and the head functions as an object in that main clause.

(57) [cin-i-j refl.sg-obl-dat či-d-sa-d-ač'-ib-t-a-lla] spr-npl-hither-npl-come.pfv-pret-dd.pl-obl-gen χabar story b-urs-ul n-tell-icvb urkː-ar find.m.ipfv-prs

'He is probably telling the story about what he had experienced.'


Examples in which the nominal head itself is expressed by a reflexive in the relative clause were judged as not very well-formed sentences by Sanzhi speakers:


### 23 Relative clauses

Relative clauses with semantically empty or light head nouns can be found, and in most cases it is the noun *zamana*, which takes over this function. These clauses have been grammaticalized into adverbial clauses expressing temporal simultaneity (62), (63) (§18.2.9). Relative clauses with *musːa* 'place' as head can be interpreted in a similar fashion as adverbial clauses referring to the location of an event (24), (25).

(62) [ʡaˁħ-ʡaˁħ-le good-good-advz gu-lik'-an] down-listen-ptcp zamana time 'at the time (when they) were listening carefully' (63) [ʁaˁni-le-r gu-d-a hin d-at-aʁ-ib] zamana time

aperture-loc-abl down-npl-dir water npl-send-do.pfv-pret 'at the time (when) you send the water from the aperture'

In general, relative clauses in Sanzhi Dargwa are part of a larger family of constructions that can be classified as 'noun-modifying clause constructions.' They include, apart from genuine relative clauses in which the head has a position in the relative clause, also constructions with 'light nouns' such as *zamana* 'time' (62), (63) and other sentential complements of nouns. Sentences (39), (40) above already showed that the same formal means that are employed to formulate relative clauses are also used when there is no syntactic relationship between the head noun and the preceding noun. In such cases the hearer is expected to establish the semantic link between the noun and the clause that modifies the noun on the basis of the context and of general knowledge. The sentences in (64–67) provide more examples of such sentential modifiers. Such versatility of the relative clause construction is typical for East Caucasian languages and has been repeatedly discussed in the literature (Daniel & Lander 2008; 2010; Comrie et al. 2017).


Instead of relative clauses it is also possible to have a nominalized clause with the masdar suffix that is marked for the genitive (65), (68). Such constructions are semantically equivalent to the noun-modifying construction above (57), (64).

23.4 Headless relative clauses

(68) [čina-k'u where-indef ʡaˁlħaˁm-le condolence-loc d-uˁq'-ni-la] 1/2pl-go-msd-gen χabar story 'the story how we went somewhere for condolences'

### **23.4 Headless relative clauses**

Headless relative clauses can be formed in four different ways: (i) with the modal participle (*-an*) (69–71), (ii) by using the locative participle (see §18.1.2.4 for further information and examples), (iii) by attaching the cross-categorical suffixes *-il* (50), (72) or *-ce* (plural -*te*) to the preterite or the modal participle (56), (75), and (iv) occasionally by means of the nominalized optative (80). The types differ with respect to their function and morphosyntactic properties. Headless relative clauses with the modal participle can only be used when the nominalized relative clause takes over the function of an absolutive argument and therefore does not require further case marking. For instance, in (69) the relative clause functions as P argument, and in (70) and (71) as copula predicate.


'There is not even anybody who is talking. '

Relative clauses with the locative participle can only express spatial meaning, and the locative participle can be marked with directional case suffixes (essive, ablative), but not with any other cases.

The use of the cross-categorical suffixes *-il* and -*ce* is a major strategy for the formation of headless relative clauses. The two suffixes are used in a variety of contexts and their overall function can be roughly described as forming referential attributes/definite descriptions. Items marked with the suffixes acquire the morphosyntactic properties of nouns (see §9.6.1 and §9.6.2 for detailed accounts). There is a functional distribution between the two suffixes. Both suffixes are used when the headless relative clause denotes a singular referent and when it is used without any further case marking, i.e., when it is an argument in the absolutive case in the main clause (56), (72), (73), but the suffix -*il* is more common. Note that the headless relative in (73) contains a further adverbial clause that is embedded into the relative clause. I could find only a handful examples of headless relative clauses bearing *-ce* in the corpus. The example in (56) comes from a poem.

23 Relative clauses

(72) hej this [ka-jcː-ur-il] down-get.up.pfv-pret-ref ∅-ik'-ul m-say.ipfv-icvb ca-w, cop-m "ħaˁsan Hasan ʡaˁli Ali hel-itːe," that-advz ∅-ik'-ul m-say.ipfv-icvb ca-w, cop-m "uruχ-le=de=w?" fear-advz=2sg=q

'The one who is standing says, "Like Hassan Ali," he says, "Are you afraid?"'

(73) [[b-učː-ul] hpl-drink.ipfv-icvb ka-b-iž-ib-il] down-hpl-be.pfv-pret-ref ka-b-išː-ib=da down-n-put.pfv-pret=1 heltːu there čina=del where=indef

' The one (picture) where they are sitting and drinking, I put this (picture) somewhere.'

When the referent of a nominalized relative clause is plural, only the suffix -*te* is allowed (74), (75).


When the headless relative clauses take case markers, the suffix -*il* is used for reference in the singular and *-te* (in its oblique stem form *-ta*) for reference in the plural. Examples with singular referents are not very common in the corpus (76) (see (139) in §9.6.2 for one more instance). Example (77), in which the nominalized relative clause functions as agent, has been elicited. When the dative case is added, the resulting clauses can have the semantics of adverbial clauses expressing causes (due to the meaning of the dative case). One example is (30) in §18.1.2.3.


Examples of headless relative clauses with plural referents and further case markers are comparatively frequent in the Sanzhi corpus. As (78), (79) show, the nominalized relative clauses can occur in various argument and adjunct positions in the main clause. 23.4 Headless relative clauses


The suffix *-ce* (but not -*il* or -*te*) is also used as a nominalized verb form taking over an argument position in a clause with a complement-taking predicate. This means that -*ce* functions as a complementizer in complement clauses of the fact-type (see §24.2.3). In some cases the nominalized clause, which occurs together with a complement-taking predicate, does not express a proposition, but refers to an entity such as a human being or an event or to abstract entities such as thoughts, wishes, etc. In that case the nominalized verb does not function as a complement, but as a headless relative clause (§24.6.2).

In addition to the just discussed types of nominalized relative clauses, Sanzhi has a nominalized optative that functions like a headless relative clause in the sense that it can take over arguments or adjunct positions in the clause and can be inflected. It preserves the semantics of the optative (§17.3). In example (80), the nominalized verb *w-ebk'-* 'die' is inflected for the ergative because it functions as the agent of the verb *kax-* 'kill'.

(80) ah, ah w-ah m-owner w-ebk'-ar-t-a-l m-die.pfv-opt-pl-obl-erg di-la 1sg-gen durħuˁ boy kax-ub-le kill.pfv-pret-cvb už-ib-le=q'al be.m-pret-cvb=mod

'Ah, may the ones die who have (a beloved one), since they apparently killed my son!'

## **24 Complementation**

Complement clauses are subordinate clauses that function as arguments of verbs. Complement taking predicates can be divided into several semantic subgroups (§24.1). Complementation strategies vary according to these subgroups. However, more important for the choice of the formal marking is the semantics of the complement clause (e.g. potential vs. activity vs. fact type) as well as co-reference and control relations between the subject of the matrix predicate and the arguments in the complement clause. Therefore, I will start with a list of complement-taking predicates (§24.1). Then I will discuss the semantic types of complement clauses and how the formal strategies are distributed across the semantic types (§24.2). Due to their high overall frequency in the corpus, reported speech constructions will be treated separately in §24.3, although they do not exhibit many peculiarities that distinguish them from other complement constructions. In §24.4, I analyze the syntactic properties of complement constructions and in §24.5 I discuss in more detail complement control.

The chapter closes with a short discussion of constructions that syntactically do not represent complementation, but semantically resemble complement constructions. Parentheticals (§24.6.1), nominalized relative clauses (§24.6.2), and adverbial clauses (§24.6.3) belong to these constructions.

I use square brackets throughout this chapter in order to indicate the complement. Note, however, that the complement is not always syntactically a clause, but can also be a nominalized verb form or an associated clause in case of parenthetical constructions.

### **24.1 Complement-taking predicates**

Not all predicates listed in this section are complement-taking predicates in the strict sense that they always require a clausal complement (as an alternative to a nominal argument) in order to build a complete grammatical sentence. Copula constructions with adverbs are by themselves independent main clauses, but the clauses that can be added to them formally behave like genuine complement clauses and are therefore included.

### **24.1.1 Utterance verbs**

Sanzhi has the following simple verbs of speech:

	- b. *haʔ-* (pfv)/*herʔ-* (ipfv) 'say'
	- c. *b-urs-* 'tell'
	- d. *b-ux-* (ipfv) 'tell'

### 24 Complementation

More specific verbs are compounds consisting of a first part that can be a noun, an ideophone, or a bound stem, and a following light verb. There are especially many compounds with the noun *ʁaj* 'word, language, talk' (see §12.2.2 for more examples); and the simple verbs of speech listed above also occur frequently together with *ʁaj*. Examples include:

	- b. *ʁaj (ka-)b-ik'ʷ-* (ipfv) 'quarrel, scold, argue, discuss, talk'
	- c. *ʁaj b-arq'-* (pfv)/*ʁaj b-irq'-* (ipfv) 'say, tell'
	- d. *ʁaj b-ikː-* (pfv)/*ʁaj lukː-* (ipfv) 'promise'
	- e. *čːal b-uq-* (pfv)/*čːal b-ulq-* (ipfv) 'argue, quarrel'
	- f. *anru b-ikː-* (pfv)/*anru lukː-* (ipfv) 'command, order'
	- g. *bursːi b-arq'-* (pfv)/*bursːi b-irq'-* (ipfv) 'teach'
	- h. *waˁw b-ik'ʷ- (ipfv); waˁw haʔ-* (pfv)/*waˁw herʔ-* (ipfv) 'call, cry'
	- i. *ʁaˁʁ b-ik'ʷ-* (ipfv) 'shout'
	- j. *t'irt'ir b-ik'ʷ-* (ipfv) 'chat'

Not all utterance verbs take complement clauses that represent reported speech. Some rather denote actions that involve speech (e.g. 'teach', 'command, order') or they denote specific sounds that imitate speech sounds (e.g. *t'irt'ir b-ik'ʷ-* 'chat'). In §24.3, I will only discuss constructions containing quotes.

### **24.1.2 Liking and fearing verbs and other verbs denoting emotions and volition**

The following verbs belong to this group:

	- b. *b-ičː-aq-* 'want, like, love' (the causativized variant of *b-ikː-*)
	- c. *razi b-iχʷ-* (pfv)/*razi b-irχʷ-* (ipfv) 'be happy, agree'
	- d. *xul b-ik'ʷ-* (ipfv) 'dream, wish, hope'
	- e. *urk' b-uq-* (pfv)/*urk' b-ulq-* 'be frightened, astonished, wonder'
	- f. *tamaša b-arq'-* (pfv)/*tamaša b-irq'-* (ipfv) 'wonder'
	- g. *uruχ b-ik'ʷ-* (ipfv) 'get afraid', *uruχle ca-b* 'be afraid'
	- h. *uruc b-iχʷ-* (pfv)/*uruc b-irχʷ-* (ipfv) 'get embarrassed, ashamed', *uruc ca-b* 'be embarrassed, ashamed'
	- i. *c'aχ ka-b-icː-* (pfv)/*c'aχ ka-b-ircː-* (ipfv) 'be embarrassed, ashamed'
	- j. *pašman b-iχʷ-* (pfv)/*pašman b-irχʷ-* (ipfv) 'be sad about, regret'
	- k. *q'as b-arq'-* (pfv)/*q'as b-irq'-* (ipfv) 'decide'
	- l. *sa-b-ag-* (pfv)/*sa-b-arg-* (ipfv) 'imagine, envisage, see'

24.1 Complement-taking predicates

Some of these verbs denote emotions that are cognitively based feelings and that are semantically close to verbs of cognition. Other verbs denoting volition have some semantic overlap with modality.

### **24.1.3 Cognition predicates**

Based on their semantics, I will divide the cognition predicates that take complements into three groups:


### **24.1.3.1 Verbs of knowledge and acquisition of knowledge**

Verbs expressing knowing and the acquisition of knowledge include:

	- b. *arʁ-* (pfv)/*irʁ-* (ipfv) 'understand' (can be used together with *urk'i* 'heart')
	- c. *b-elč'-* (pfv)/*b-uč'-* (ipfv) 'read'

In addition, there is a particle *aχːu* 'I don't know, dunno' that takes complement clauses, as in (50). It most probably goes back to the verb 'know'. This particle can be used with the first person singular dative pronoun *dam*, but not with any other dative experiencer and not with nominal stimuli, which indicates its status as a particle (in contrast to the full verb, which can be used with arguments of all persons and numbers and also with nominal stimuli arguments). Furthermore, it is also used parenthetically (§24.6.1).

### **24.1.3.2 Propositional attitude predicates**

These predicates express a kind of propositional attitude toward the truth of the complement.

	- b. *b-iχː-* 'believe'
	- c. *b-iχ-(b)-it-ag-* (pfv)/*b-iχ-(b)-it-arg-* (ipfv) 'believe' (compound verb, containing *b-iχː-* 'believe')
	- d. *b-iχ-či ag-* (pfv)/*b-iχ-či arg-* (ipfv) 'believe' (compound verb, containing *b-iχː-* 'believe')
	- e. *šak b-ik-* (pfv)/*šak b-irk-* (ipfv) 'guess, suspect, feel'

In addition, there is a phrase *dila pikri ħaˁsible* 'in my mind' (1sg.gen thought following) that also expresses a propositional attitude, but syntactically represents a parenthetical, not a complement-taking predicate, as in (127) and (128).

### 24 Complementation

### **24.1.3.3 Other cognition predicates**

These predicates are achievement verbs for positive (e.g. 'remember') and negative achievement (e.g. 'forget') in the domain of cognition.

	- b. *han b-ik-*, *b-ičaq-* (pfv)/*han b-irk-*, *b-irčaq-* (ipfv) 'remember, seem to, imagine, think'
	- c. *han b-el, han ca-b* 'remember'
	- d. *urk'i-le sa-b-eʁ-* (pfv)/*urk'i-le sa-b-irʁ-* (ipfv) 'descend on the heart, remember'
	- e. *urk'i-le-b le-b* 'think, have thoughts' (lit. 'be on the heart')
	- f. *ħaˁsib b-arq'-* (pfv)/*ħaˁsib b-irq'-* 'test, check'

### **24.1.4 Manipulative verbs**

Manipulative verbs typically have subjects that differ from the subjects in the complement clause and the semantics of the complement clause is irrealis. Utterance verbs belong to this group, such as 'command, order' and 'teach' and the basic verb of speech *b-ik'ʷ-* (ipfv) 'say', which is frequently used with a manipulative meaning. Other manipulative verbs are:

	- b. *b-at-* (pfv)/*b-alt-* (ipfv) 'let, leave'
	- c. *iχtijar b-ikː-* (pfv)/*iχtijar lukː-* (ipfv) 'give permission, right'
	- d. *χajri a-b-iχʷ-* (pfv)/*χajri a-b-irχʷ-* (ipfv) 'forbid'

### **24.1.5 Phasal verbs**

Sanzhi phasal verbs include:

	- b. *b-aʔ b-axː-* (pfv)/*b-aʔ b-irxː-* (ipfv) 'begin, start'
	- c. *či-ka-b-iħ-* (pfv) *či-ka-b-irħ-* (ipfv) 'begin, start'
	- d. *taman b-iχʷ-* (pfv)/*taman b-irχʷ-* (ipfv), *taman b-arq'-* (pfv)/ *taman b-irq'-* (ipfv) 'stop, finish'

There are two verbs that can express the meaning 'continue', *kelgʷ-* (pfv) 'remain, stay, be' and the defective verb *b-el* 'remain, stay'. Both are used in periphrastic verb forms, which are not complement constructions (see §15.2 and §15.3).

24.2 Complementation strategies and their semantics

### **24.1.6 Modal predicates**

Modality comprises epistemic modality (likelihood, certainty), deontic modality (necessity, obligation, permission), and ability. In Sanzhi, modality is typically not expressed through modal verbmodal verbs such as English *must*, *should,* or *may*, but instead by means of analytic and periphrastic verb forms. The obligative tenses denote obligation in addition to future (§14.1.5–§14.1.7). There are a number of periphrastic epistemic modal constructions that express likelihood and certainty (§15.4.2, §15.5). In addition, it is possible to use the embedded question marker together with the optional adverb *belki* 'be possible' (§9.1). The only complement-taking predicates that express modality convey the meaning of ability or necessity:

	- b. *ʡaˁʁunil ca-b* 'be needed, necessary'
	- c. *ħaˁžatle ca-b* 'be needed, necessary'

### **24.1.7 Evaluation**

Evaluation is expressed by adverbs together with a copula or the verb *ag-* (pfv) 'go':

(10) a. *ʡaˁħle ca-b*, *ʡaˁħle ag-* (pfv)/*ʡaˁħle arg-* (ipfv) 'be good' b. *wahil ca-b*, *wahil ag-* (pfv)/*wahil arg-* (ipfv) 'be bad'

### **24.2 Complementation strategies and their semantics**

The following complementation strategies are available in Sanzhi and will be treated in this section.

	- zero strategy (§24.2.1)
	- quotative particles *bik'ul, haʔible* (§24.2.2)
	- cross-categorical suffix *-ce* (§24.2.3)
	- masdar *-ni* (§24.2.4)
	- perfective converb *-le* (§24.2.5)
	- infinitive *-ij*/subjunctive (§24.2.6)
	- embedded question marker *=el* (§24.2.7)
	- imperfective converb *-ul/-unne* (§24.2.8)
	- the *pretend*-construction (§24.2.9)

### 24 Complementation

A number of the grammatical markers listen above also occur in other types of subordinate clauses: the perfective and the imperfective converb head adverbial clauses (§25.1). The cross-categorical suffix *-ce* occurs in relative clauses (Chapter 23). Therefore, it is not always easy to tell apart complement constructions from adverbial or relative clauses.

Based on their semantics, we can distinguish four types of complement clauses (Hengeveld 1989: 130; Dik 1997: 93; Dixon 2006):

**potential type:** refers to the potentiality of the subject of the complement clause becoming involved in an activity

**activity type:** refers to some ongoing activity, relating to its extension in time

**fact type:** refers to the fact that something took place

**speech act type or direct speech type:** refers to a particular speech act

For the linguistic encoding of each semantic type one or more complementation strategies are used (Table 24.1). The strategies will be discussed in detail in the following sections.

Table 24.1: Complementation strategies and the semantic types of complements


### **24.2.1 The zero strategy**

No formal marking of the complement clause is a major strategy in reported speech constructions (§24.3). It is also found, though less commonly, with those emotion and cognition verbs that denote activities that presuppose or imply speech and that have therefore a linguistic component that makes them similar in their behavior to verbs of speech (11–14). The zero strategy can be viewed as an alternative to the use of the quotative particles since the employment of quotative particles is possible with every verb that allows for the absence of formal encoding of its complement.

24.2 Complementation strategies and their semantics

(11) na now dam=ra 1sg.dat=add han remember b-ič-ib n-occur.pfv-pret [w-ax-an=da m-go-ptcp=1 q'ʷila a.little arc money d-irq'-an=da] npl-do.ipfv-ptcp=1

' I also thought, I should go to make a little money.'


' The poor boy got frightened: Does the fox know the human language or what?'

(14) [ʡuˁrus Russian ʁaj-la language-gen ce=jal what=indef b-ik'-ul hpl-say.ipfv-icvb ca-b cop-hpl it-i-j] that-obl-dat dam 1sg.dat qum.urt-ul forget.ipfv-icvb ca-b cop-n ' It (i.e. the plant) is called something in Russian, I forgot.'

Due to the absence of any formal marking, it is alternatively possible to analyze the above examples as juxtaposition of two main clauses without a syntactic link between them, but with a clear semantic relationship, which follows from the meaning of the emotion and cognition verbs and the interpretation of the clauses in brackets as expressing thoughts. In example (14) still another approach suggests itself, namely the analysis of the cognitive predicate as parenthetical, which means that this is not a complement construction, but simply an independent sentence followed by another independent sentence that makes a comment on the previous one and functions as a kind of stance marker to inform the hearer that the speaker is unsure about the validity of some of her utterances about plant names. See §24.6.1 for more information about parentheticals.

### **24.2.2 The quotative particles**

Sanzhi has two quotative particles, *bik'ul* and *haʔible*. They are discussed in more detail in §24.3 in reference to reported speech, because they constitute a major encoding strategy for complements of utterance verbs. Although *bik'ul* and *haʔible* have preserved their verbal properties (e.g. inflectional morphology, gender agreement, position) I will refer to them as "particles" when they occur in addition to matrix verbs of speech in order to differentiate them from the matrix verbs that take complement clauses. They can therefore be called "parentheticals" that do not realize syntactic subordination but pragmatically mark a clause as a speech report.

The particle *haʔible* is far less common than *bik'ul*, and occurs only with verbs of speech (including cases in which they are used as verbs of cognition) and occasionally in purposive clauses (21). The particle *bik'ul*, in contrast, occurs also in complements

### 24 Complementation

of emotion ('be afraid' (15), 'dream' (16)) and cognition verbs that denote activities that heavily rely on the (implicit) use of language, most notably verbs meaning 'think' (17), (18). Verbs of knowledge normally do not mark complement clauses with the quotative particle.


The particle agrees in gender with the subject or subject-like argument of the matrix clause even in those cases in which the matrix predicate takes other cases than the absolutive. For instance, in (17) the experiencer in the matrix clause is marked by the dative, and the predicate that governs this argument shows local agreement with the complement clause (see §24.4 below on the difference between local and long-distance agreement in complement clauses). By contrast, the quotative shows feminine singular agreement because the experiencer has a female referent. Example (18) shows that even a possessor functioning as experiencer can control gender agreement on the quotative particle. The obvious reason for this behavior is the valency pattern of the verb *b-ik'ʷ-* 'say', from which the particle originates. It requires an absolutive argument controlling its gender agreement prefix (in addition to the complement clause).


' The father thinks about whether these clothes are to be washed.'

(19) il=ra that=add šak feel ∅-ič-ib m-occur.pfv-pret ca-w cop-m [itːi those χalq' people b-ik'ʷ-an hpl-say.ipfv-ptcp mar truth b-urkː-ar n-find.ipfv-prs.3 ∅-ik'-ul] m-say.ipfv-icvb

'He also guessed that the people had probably said the truth.'

It might be combined with the modal interrogative suffix (see §17.4 for more information) in the complement clause if the complement represents a question which has deontic modality and in which the subject is co-referential with the author of the quote (20).

24.2 Complementation strategies and their semantics

(20) [d-iʡ-ij npl-steal.pfv-inf uq'-ide=l go.m.pfv-modq=indq a-w-q'-idel neg-m-go.pfv-modq ∅-ik'-ul] m-say.ipfv-icvb pikri thought ∅-ik'-ul=el m-say.ipfv-icvb=indq '(He) is probably thinking, Should I go stealing or should I not go?'

In (21) the particle *haʔible* is followed by the verb *b-ik'ʷ-* used with the meaning 'think'.

(21) [cul-b-a-la tooth-pl-obl-gen tuχtur doctor arg-an=da] go.ipfv-ptcp=1 haʔ-ib-le say.pfv-pret-cvb ik'-ul neg-find.ipfv-prs.3 a-urkː-ar not.know aχːu

' I will become a dentist, he is probably thinking, I do not know.'

### **24.2.3 The cross-categorical suffix** *-ce*

In complement clauses, the cross-categorical suffix -*ce* is added to the preterite participle (22–24) or to the modal participle (131). This suffix is used for the formation of definite descriptions (see §9.6.1 for detailed accounts of its complex array of functions). When it is suffixed to verbs the verbs take over the function of attributes (i.e. as participles in relative clauses) or as nominalized referentially independent clauses. The latter function is relevant for the occurrence of -*ce* in complement clauses of the fact type. More specifically, the proposition in the complement clause, which is marked by -*ce*, is considered to be true and treated as a fact by the speaker. Therefore, only certain verbs denoting emotions, cognition verbs, as well as evaluative predicates express their complement clauses with the attributive marker.


The meaning of these complement clauses is very close to complement clauses formed with the masdar (25) (§24.2.4) and with the perfective converb (24) (§24.2.5) and the strategies can usually be replaced by each other.

24 Complementation

(25) wahi-l=de bad-adv=pst [Rašid-li Rashid-erg ust'ul chair ʡaˁħ good a-b-arq'-ib-le/a-b-arq'-ib-ce/a-b-arq'-ni] neg-n-do.pfv-pret-cvb/neg-n-do.pfv-pret-dd.sg/neg-n-do.pfv-msd 'It was bad that Rashid did not repair the chair.' (E)

Note that the suffix -*ce* also occurs in nominalized relative clauses that semantically resemble complement clauses of the activity type. These constructions are discussed in §24.6.2.

### **24.2.4 The masdar**

The masdar is a deverbal noun that is used not only in complement clauses, but also in other argument and adjunct positions (§18.1.5). Complements expressed by the absolutive (i.e. not further case-marked) masdar denote facts. Therefore, basically the same matrix predicates that make use of the cross-categorical suffix -*ce* also allow for the masdar, i.e., cognition predicates, evaluation predicates and emotion predicates if they refer to factual complements that have propositional meaning:


' Then the wind understood that he would not be able to take off the coat of this person.'

(28) [it that Maˁħaˁmmad-la Mahammad-gen xːunul woman r-iχʷ-ni] f-be.pfv-msd dam 1sg.dat han remember b-akːu n-cop.neg ' I don't remember that she was Mahammad's wife.'

As mentioned above and shown in the elicited example (25) shown in the previous section, the masdar is often semantically equivalent to the preterite participle and the cross-categorical suffix.

### **24.2.5 The perfective converb**

The perfective converb is used to form the same types of complement clauses as the cross-categorical suffix -*ce*, that is, fact complements. Thus, in elicitation it is given as an alternative to *-ce* (§24.2.3), and the types of matrix predicates with which it occurs are the same as for that suffix, namely cognition verbs (29) and emotion verbs (30), and evaluation predicates (32), (25). The complement clauses refer to situations in the past.

24.2 Complementation strategies and their semantics


' I (fem.) regretted that I did not stay for the holidays.' (E)

(32) "[salam-le greeting-loc a-s-ač'-ib-le] neg-hither-come.pfv-pret-cvb ʡaˁħ-le good-advz a-arg-u=q'al," neg-go.ipfv-prs.3=mod ∅-ik'-ul m-say.ipfv-icvb ca-w cop-m

' He said, "It is not good that I did not go to the meeting."' (lit. having come to the greetings)

Clauses with the perfective converb also express activity complements when they are used, e.g., with certain emotional predicates (33). Similarly, (31) could also be translated as 'I regretted when I did not stay for the holidays.'

(33) it-i-j that-obl-dat [xʷit' whistle haʔ-ib-le] say.pfv-pret-cvb a-b-ičː-aq-i neg-n-want.ipfv-caus-hab.pst 'He did not like when one whistled.'

This makes clear that some constructions, which at the first glance look like complement clauses formed with the perfective converb, could also be analyzed as adverbial clauses occurring together with a main clause, which contains one of the complementtaking predicates given in §24.1. The preterite converb is a regular means of forming adverbial clauses that refer to events and situation occurring prior to or at the same time as the situation referred to in the main clause i.e. 'while, when, after, and' (§18.1.1.2). In contrast to complement clauses, adverbial clauses do not fulfill argument positions but serve as clausal adjuncts. For some of the examples in this section further research is needed in order to decide if the subordinate clause is a true complement or if it is an adjunct, as in (30),<sup>1</sup> (31), (32). In example (29) an adverbial-clause interpretation seems rather unlikely.

Finally, the perfective converb occurs in complements of 'finish' as an alternative to the infinite or subjunctive, as in (35) and (34). Such complements are of the activity or the potential type. Example (34) shows the verb 'finish', which contains an intransitive

<sup>1</sup>This example could probably be translated as 'You came and I am happy'.

### 24 Complementation

lexical verb *b-iχʷ-ij* of which the nominal part *taman* 'time' functions as the subjectlike argument of this verb. The clause preceding this verb contains a verb bearing the perfective converb suffix just as the complement clause in example (35) and the two verbs 'finish' form a pair of which the two members differ with respect to the lexical verbs (intransitive *b-iχʷ-ij* 'be, become' vs. transitive *b-arq'-ij* 'do, make') (see §12.2 for many more pairs of verbs of this kind). Thus, it seems reasonable to treat both constructions in (34) and (35) analogously as complement constructions with complements expressed by perfective converbs. However, in (34) an analysis as adverbial clause construction seems again to be possible. In that case the translation would rather be 'Having built (the mill) and the time finished, ...'. Such an analysis cannot be applied to (35). Further testing of the syntactic properties and whether the interpretation as adverbial clause construction in (34) is in fact possible or necessary or perhaps incorrect must be clarified by future research.


### **24.2.6 Infinitive and subjunctive**

The infinitive and the subjunctive are very widely used in complement clauses of the potential type that occur with complement control. Complement control means that the subject in the complement clause is obligatorily omitted because it is identical to the subject or another argument (typically the object) of the matrix clause (§24.5), and when the matrix clause is an impersonal construction, as it is the case with evaluative predicates. Note, however, that it is possible to express an overt subject in an infinitival complement clause containing a trivalent verb when the matrix clause has a different subject (113).

The infinitive can be used with all persons, whereas the subjunctive has the suffix *-tːaj* for the second person and *-araj/-anaj* for the third person (see §18.1.3 for a more detailed account of the infinitive and §18.1.4 for the subjunctive). There is no subjunctive suffix for the first person. The subjunctive can always be replaced by the infinitive without any change in the meaning of the sentence.

Emotion and cognition predicates take complement clauses headed by the infinitive or the subjunctive:

24.2 Complementation strategies and their semantics

(36) a-b-ikː-ul neg-n-want.ipfv-icvb [gu-r-uq'-aˁnaj], sub-abl-go.m-subj.3 bah most hila-r last-abl gu-r-ag-ur sub-abl-go.pfv-pret Sanži-r Sanzhi-abl

' Not wanting to leave, he left Sanzhi as the very last.'


' He forgot to write and I remained like this.'

Due to the nature of manipulative predicates and modal predicates their complements belong to the potential type and they exhibit complement control. Therefore, the use of the infinitive or the subjunctive is the only possible strategy for complementation (40– 44). Other complementizers such as the attributive suffix or the masdar are ungrammatical. As example (42) demonstrates, modal predicates with infinitival complements allow for backward control: the matrix predicate shows feminine singular agreement because the ergative agent *Paitu* in the complement clause has a feminine singular referent. See §24.5 below for more details.


### 24 Complementation

Phase predicates have complement clauses of the potential or of the activity type. In the first case, they make use of the infinitive and subjunctive. Thus, in elicitation, when translating narratives from Russian or Standard Dargwa, or when telling prepared stories, the complements of 'begin' contain the infinitive or subjunctive, as in (45) and (46). Otherwise, the imperfective converb is employed (§24.2.8). Similarly, with 'finish' we find either the perfective converb (§24.2.5) or the infinitive/subjunctive, as in (47).

(45) durħuˁ=ra boy=add kac'i=ra puppy=add či-ka-b-iħ-ib spr-down-hpl-begin.pfv-pret [umc'-anaj search.ipfv-subj.3 ʡaˁt'a-j] frog-dat

' The boy and the puppy began to search for the frog.'


Evaluative predicates also employ the infinitive/subjunctive if the complement clause has potential semantics:


### **24.2.7 The embedded question marker**

The embedded question enclitic has three allomorphs: *=jal* after vowels, *=el* after consonants, and *=l* after some suffixes ending in /e/. It is used in all types of embedded questions and occurs in complementary distribution with the interrogative enclitics used in independent questions (§28.4). It belongs to the class of predicative particles (together with the other two interrogative enclitics and a few other particles, see §9.1). The particle encliticizes to the head of the complement, i.e., the verb, or occasionally to interrogative pronouns.

Complement clauses with the embedded question marker are of the potential type, as in (53) and (62) as well as of the activity type (58), but not of the direct speech type,

### 24.2 Complementation strategies and their semantics

because for direct speech the other two interrogative enclitics have to be used. Matrix predicates that employ the embedded question markers are utterance predicates (see §24.1.1 for examples) and cognition predicates (§24.1.3). It co-occurs with the quotative marker (20).

Examples (50–52) show embedded polar questions. The matrix predicates are negated or they imply the use of an embedded question such as 'know' (§28.4). With affirmative matrix predicates we could alternatively have fact complements ('know that') and consequently other complementation strategies. The matrix clause can be a statement or a question.

(50) [han remember d-irč-aq-ul=el] npl-occur.ipfv-caus-icvb=indq aχːu not.know

'I don't know if (he) is remembering.'

(51) dam 1sg.dat b-alχ-ad n-know.ipfv-1.prs sa-b-irʁ-u=jal hither-hpl-come.ipfv-prs.3=indq itːi they (ja=ra or=add a-sa-b-irʁ-u=jal) neg-hither-hpl-come.ipfv-prs=indq amma but a-cːe 2sg-in a-b-urs-an=da neg-n-tell-ptcp=1 ' I know whether they will come (or not), but I am not going to tell you.'

(52) at 2sg.dat b-alχ-atːe=w n-know.ipfv-prs.2sg=q ʡaˁli Ali panedelnik-le-w Monday-loc-m s-erʁ-u=jal hither-come.m.ipfv-prs.3=indq 'Do you know whether Ali will come on Monday?' (E)

The following two corpus examples illustrate embedded disjunct polar questions. They have the same structure as the embedded polar questions with the only difference that there is only one embedded clause and not two.


With embedded content questions the enclitic mostly appears on the verb (55–58), as it is also common for the interrogative enclitics in independent questions. The matrix clause can be affirmative or negative.

### 24 Complementation


If the complement does not contain a verb the enclitic appears on the question word (59). In utterances with verbs and questions words it is possible to attach the enclitic to the latter (60), but the variant with the verbal host is judged as preferable (61).


' I don't know when aunt Zhamilat was born.' (E)

If the subject of an embedded question is first person, the verb in the complement clause takes the modal interrogative suffix *-ide(l)*, which most probably goes back to a person marker *-id* plus the petrified marker for embedded questions =*el* (§17.4), as in (62). The suffix can occur in combination with the quotative marker (§24.2.2).

24.2 Complementation strategies and their semantics

(62) [ce what b-arq'-ide=l] n-do.pfv-modq=indq a-b-alχ-ul=da neg-n-know.ipfv-icvb=1 'I do not know what to do.'

### **24.2.8 The imperfective converb**

The imperfective converb belongs to the minor complementation strategies. It expresses potential or activity complements with the verb 'begin', for which it represents the most common way of marking complements (63) (alternatively, the infinitive/subjunctive is used, see §24.2.6).

(63) kːurtːa fox b-aʔ n-begin b-išː-ib n-put.pfv-pret [šajt'an devil rucːi sister r-irʡ-uˁl] f-betray-icvb ' The fox began betraying/to betray the devil sister.'

Another possible matrix predicate for complements heading the imperfective converb is the perception verb 'see' whose complement clauses are either of the fact type as the translation in (64) suggests or of the activity type (65).

(64) it-i-j that-obl-dat či-d-až-ib spr-npl-see.pfv-pret [aba-l mother-erg q'uˁl-e cow-pl icː-ul] milk.ipfv-icvb ' S/he saw that mother was milking the cows.' (E)

(65) [Nursijat Nursijat ʡaˁħ good r-iχ-ub-le, f-be.pfv-pret-cvb r-ax-ul] f-go-icvb či-r-až-ib-le, spr-f-see.pfv-pret-cvb razi happy r-iχ-ub=da=q'al f-be.pfv-pret=1=mod ' I (fem.) got happy when I saw that Nursijat recovered and is (already) walking

around.'

### **24.2.9 The** *pretend***-construction**

The cross-categorical suffix *-il*, which forms referential attributes that can modify nouns (e.g. relative clauses) or occur referentially independent in argument or adjunct positions, can also take the genitive case to express activity complements in the pretendingconstruction (66–68) (see §9.6.2 for a summary of all functions of *-il*). Depending on whether *-il* is suffixed to the modal participle (67) or to the preterite participle (69) the complement clauses have past time reference or non-past time reference. Syntactically, the dependent clause does not represent the object argument of the matrix verb 'do, make' because of the genitive case. Nevertheless, we can analyze this construction as a complement construction because there are a number of similar examples with nouns instead of clauses that semantically, but not syntactically, are arguments of the verb 'do, make' despite bearing the genitive case (§3.4.1.3).

### 24 Complementation


The agreement prefix on the matrix verb can be *b-* or *d-* with no difference in semantics (69). The prefix does not seem to be governed by an agreement controller, because of the two available options one is not attested (namely long distance agreement with the absolutive argument of the complement clause), whereas the other (local agreement with the entire complement clause) is implausible since it allows only for the *b-* prefix, but not for the *d-* prefix. Furthermore, in constructions without a complement only *d-* is possible (70).


### **24.3 Reported speech constructions**

### **24.3.1 General characteristics of reported speech**

Reported speech constructions usually contain an utterance verb and a quote. The relationship between the clause containing the verb of speech and the quote can be marked or unmarked. The utterance verb precedes the quote (80), interrupts it (71), or follows it (72). Sometimes it is repeated and occurs in more than one position (74). The quote itself does not bear any specific grammatical marking apart from the optional use of quotative particles to pragmatically mark quotes.

The verb *b-ik'ʷ-* is the most frequently occurring verb of speech that is also used as a quotative particle in reported speech constructions and other complement clauses 24.3 Reported speech constructions

(see also §24.2.2). The basic meaning of this verb seems to be 'say', but it is often used with the meaning 'think', i.e., expressing mental activities such as thinking, considering, or reflecting. The verb has only the imperfective stem. Its subject argument takes the absolutive case and controls the gender agreement prefix. It is very widely used as a light verb in compounding, as shown by some examples above. The compounds can denote activities related to speech and language such as *pikri b-ik'ʷ-* 'think', *xul b-ik'ʷ-* 'wish, dream', *ʁaj b-ik'ʷ-* 'scold', *ʁumku b-ik'ʷ-* 'swear', *iχtilat b-ik'ʷ-* 'chat', etc., but they can also have totally different meanings such as *qus b-ik'ʷ-* 'slide' or *duc' b-ik'ʷ-* 'run' (see §12.2 for more examples). The verb is used as a matrix verb in reported speech constructions, either in the form of the compound present (71) or with the suffix *-ar* for past time reference (72).

(71) "dam 1sg.dat ʡaˁʁuni-l needed-advz akːu" cop.neg ∅-ik'-ul m-say.ipfv-icvb ∅-ik'-ul m-say.ipfv-icvb ca-w cop-m "hel-tːi that-pl cik'al" something

'He says, "I do not need these things."'


This verb is also used when mentioning the name of something or somebody or the word for something in another language or dialect , e.g. *Saliħaˁt b-ik'-ul* '(a person) called Salihat', or as in (75).

(75) nišːa-la 1pl-gen "daˁqaˁ-lla wound-gen q'ar" plant b-ik'ʷ-ar. hpl-say.ipfv-prs [ʡuˁrus Russian ʁaj-la word-gen ce=jal what=indq b-ik'-ul hpl-say.ipfv-icvb ca-b cop-hpl it-i-j] that-obl-dat dam 1sg.dat qum.urt-ul forget.ipfv-icvb ca-b cop-n ' In our (language) it is called "plant of the wound." I forgot what it is called in Russian.'

The verb *b-ik'ʷ-* has grammaticalized into a quotative particle (see below). Moreover, it can express hearsay evidentiality.

Another very frequent utterance verb is *-ʔ-* (pfv)/*-erʔ-* (ipfv) 'say', which is almost always used with the spatial preverb *ha-* 'upwards', that is *haʔ-*/*herʔ-*. This is a transitive verb that marks the subject, i.e. the speaker, with the ergative. It is mainly used in reported speech constructions with past time reference. Besides that it functions as a quotative particle (see below).

(76) "er look b-erč'-e!" n-look.pfv-imp haʔ-ib say.pfv-pret č'an-ni wind-erg '" Look!" said the wind.'

The imperfective stem is used, among other things, for meta-comments on how you express what you want to say, which words you use:

(77) "sabrat gather d-arq'-ib" npl-do.pfv-pret herʔ-an say.ipfv-ptcp akːu=q'al .cop.neg=mod darkːʷan Dargwa ʁaj-la language-gen '"Gather did," you should not say in Dargwa.'

Other common simple utterance verbs are the transitive verbs *b-urs-* and *b-ux-*, which both can be translated with 'tell', and the transitive verb *xar b-eʁ-* (pfv)/*xar b-irʁ-* (ipfv) 'ask', which occur, like all verbs of speech, with or without a quotative particle.

(78) "du-l 1sg-erg b-aˁq-ib-le" n-hit.pfv-pret-cvb ∅-ik'-ul m-say.ipfv-icvb ca-w cop-m "xːunul-li-j, woman-obl-dat ce what b-arq'-ide=l" n-do.pfv-modq=indq ∅-ik'-ul m-say.ipfv-icvb χabur-t-a-l story-pl-obl-erg ux-ul tell.m.ipfv-icvb ca-w cop-m heχ dem.down

'"I hit my wife, what should I do," he says; he is telling the stories.'


24.3 Reported speech constructions

A minor strategy for expressing reported speech is the use of the verb 'begin' (82) and other non-utterance predicates (83).

(82) w-aʔ m-begin ač'-ib, come.pfv-pret qili home sa-jʁ-ib=er hither-come.m.pfv-pret=when […] "du-l 1sg-erg hel=ʁuna this=eq cik'al something imc'a anymore a-b-irq'-an=da" neg-n-do.ipfv-ptcp=1 ∅-ik'-ul m-say.ipfv-icvb 'He began when he came home, […] "I will not do things like this anymore."' (83) "čina-r where-abl sa-k-ul=de?" hither-lead.pfv-icvb=2sg r-ik'-ul f-say.ipfv-icvb r-irχʷ-an=de f-become.ipfv-ptcp=pst het that durħuˁ-la boy-gen xːunul woman

' The daughter-in-law must have asked "From where do you bring the body?"'

Finally, the topic of a conversation or a thought can be expressed by using the postposition *qari=či-b* 'on.top=on-n' together with a complement clause bearing the genitive case suffix (see (142) and §8.1.6 for another example).

### **24.3.2 Formal marking in reported speech constructions**

The distinction between direct and indirect speech as we know it from European languages cannot be applied to Sanzhi because it relies on deictic shift, but in Sanzhi the original speaker's deictic frame is usually retained. Sanzhi does not have any special verb forms or sequences of tense. The only formal marking that is available for reported speech are quotative particles occurring at the end of the quote, and very occasionally simple reflexive pronouns. These quotative particles are also used with other matrix verbs that are not utterance verbs (§24.2.2), and their use is mostly optional. In fact, unmarked quotes are as common as quotes marked by quotative particles.

Sanzhi has two quotative particles for reported speech that transparently derive from the two most frequently used verbs of speech. The first is *b-ik'-ul*, the imperfective converb of *b-ik'ʷ-*. The second is *haʔ-ib-le*, the perfective converb of *haʔ-*. Both are, in fact, formally indistinguishable from the respective converbs. They have not undergone any phonological reduction so far, and the gender prefix of *b-ik'-ul* follows the same agreement rules as the matrix verb of speech from which it is derived. Therefore, it is often impossible to say whether a certain occurrence of them represents the use as a matrix verb or a quotative particle. If the quotative particles co-occur with framing verbs in a matrix clause we can be sure that we are dealing with the quotative-particle use, as in (71), (85) and (86). Sometimes it looks as if the quotative particles alone can mark an utterance as a quote (84). Such an analysis naturally suggests itself if we remember that the marker of embedded questions can also be used without a matrix clause in epistemic modal constructions (§28.4) (Forker 2019a). However, despite the relative frequency of examples such as (84) used in what looks like an independent utterance, these clauses are dependent clauses that cannot occur on their own (see §25.1.8 for a discussion of the apparent use of converbs in what seem to be main clauses).

### 24 Complementation

(84) "xːunul-li-sa-r woman-obl-ante-abl uruχ fear ∅-ik'-ul=de=w?" m-say.ipfv-icvb=2sg=q b-ik'-ul, … hpl-say.ipfv-icvb '"Are you afraid of your wife?" they say and …'

In elicitation, *b-ik'-ul* cannot apparently be used when the matrix verb of speech occurs in the preterite and instead the particle *haʔible* is employed. Thus, if we replace *haʔible* with *ik'-ul* in (87), the sentence is rejected by Sanzhi speakers. However, in the Sanzhi corpus one can find examples of matrix verbs of speech in the preterite used together with *b-ik'-ul*, as in (79) and (80).

The quotative particle *haʔible* is only rarely used and herefore ambiguous examples, such as (85) and (86), are harder to find. In (86) it could either be analyzed as a quotative particle that follows the first part of the quote or as matrix verb that heads the preceding complement clause.


'" It is necessary that he must come, at 8 he must be there. Tell him this!" I said.'

In elicitiation, the quotative particle *haʔible* occurs when the matrix clause has past time reference because it developed from the perfective converb construction that is derived from the preterite participle (87). The quote together with *haʔible* looks exactly like an adverbial clause that follows the main clause and into which a complement is embedded.

(87) Ramazan Ramazan ʁaj word b-ičː-ib n-give.pfv-pret "dam 1sg.dat kumek help b-irq'-an=da" n-do.ipfv-ptcp=1 haʔ-ib-le say.pfv-pret-cvb ' Ramazan gave me his word "I'll help."' (E)

The same converb is used with the meaning 'because, in order to' to express reasons or purpose clauses. The expression of reason or cause is shown in (88). It might have developed from an adverbial construction in which *haʔible* functions as a verb of speech and the converb clause, which precedes *haʔible*, represents a quote that explains or provides reasons for the situation referred to in the main clause. In other words, (88) could alternatively be translated as 'After (they) said that he beat up his family and (they) said that the boy was in the arms (of the mother), they led him away.'

24.3 Reported speech constructions

(88) kulpat family b-it-ib-le hpl-beat.up-pret-cvb haʔ-ib-le, say.pfv-pret-cvb nik'a-ce small-dd kʷi-lle in.the.hands-advz naˁq-li-cːe-w arm-obl-in-m haʔ-ib-le, say.pfv-pret-cvb w-erč-ib-le m-lead.pfv-pret-cvb 'Because he beat up his family, because the boy was in the arms (of the mother), they led him away.'

In elicitation, the quotative particle *bik'ul* is not used when *b-ik'ʷ-* is the matrix verb. But this restriction has purely stylistic reasons and is only apparent. In the corpus, counter-examples can readily be found.

The use of quotative markers together with the infinitive in purpose clauses with the meaning 'in order to' has been noted in a number of other East Caucasian languages such as Ingush, Godoberi, Hinuq, Tsez, and probably also Tsakhur (Forker 2016a). For this construction, it is plausible to assume that it goes back to a reported speech construction with *haʔible* originally functioning as the framing verb to a quote which might have contained another verb with volitional semantics. In other words, (89) might have developed from a construction like 'They said, we want to drink.'.

(89) deč-li drinking-erg b-učː-ij hpl-drink.ipfv-inf haʔ-ib-le say.pfv-pret-cvb ka-b-iž-ib-te down-hpl-be.pfv-pret-dd.pl b-iχʷ-ij hpl-be.pfv-inf heštːu, here aχːu not.know dam 1sg.dat 'They probably sit down in order to drink here, I do not know.'

If the quote is an utterance with non-declarative mood, be it a command or a question, then the mood markers such as the imperative suffix (86) or the enclitics for content questions (79) and polar questions (84) are normally kept and can co-occur with the quotative markers. Otherwise, it is possible to use the special enclitic for embedded questions that does not co-occur with the interrogative enclitics for independently used questions, but can co-occur with the quotative markers (see §24.2.7 above and §28.4). This enclitic is added to the head of the interrogative clause or to the item in focus. In embedded disjunctive polar questions such as (90) it encliticizes to each member of the disjunction. The embedded question marker does not normally occur in independent clauses (except for when it is used to express epistemic modality). Therefore, the complement clauses containing it are marked as dependent, although they have at their disposal the full range of TAM forms as well as person agreement.

(90) xar ask b-irʁ-an=da n-ask.ipfv-ptcp=1 [bek' head le-b=de=l exist-n=pst=indq b-akːʷ-i=jal] n-cop.neg-hab.pst=indq 'We will ask whether he had a head or not.'

The only further peculiarity that reported speech construction show, and which they share with other subordinate clauses, most notably other complement clauses, is the use of reflexive pronouns as logophors (see Forker 2019c for a detailed account of logophoric

### 24 Complementation

reflexives and other properties of non-direct speech constructions in Sanzhi). When the author of the quote, which must be third person, is identical to an argument or adjunct in the quote, the reflexive pronoun can be used instead of the first person pronoun (91). The use of demonstrative pronouns is impossible since they would express disjoint reference with the author of the quote.

(91) wallah by.God ∅-ik'ʷ-ar m-say.ipfv-prs.3 wec'al ten ʡaˁraˁq'i-la vodka-gen šuša bottle ∅-ik'ʷ-ar m-say.ipfv-prs.3 cin-ni refl.sg-erg asː-ib-le, buy.pfv-pret-cvb d-alli npl-together h-aqː-ib-te up-carry-pret-dd.pl d-už-ib npl-be-pret ca-d cop-npl ' By God, he said that he himself had bought ten bottles of vodka; and apparently he had brought them with him.'

The use of personal pronouns is also possible. The first person pronoun is employed when the referent in the quote is identical to the author (78) and the second person pronoun is used when the referent is identical to the addressee (81). Furthermore, the agreement on the verb in clauses with subject-like arguments expressed by reflexive pronouns is not third person, as would be expected when a reflexive pronoun occurs, but first person instead, as in (92).

(92) "cin-ni refl.sg-erg d-arq'-ib-te npl-do.pfv-pret-dd.pl cik'al, something ʡaˁħ good d-irq'-an=da, npl-do.ipfv-ptcp=1 b-arx n-direct xːun-ne road-loc k-ercː-an=da" down-stand.ipfv-ptcp=1 haʔ-ib-le say.pfv-pret-cvb ' (He said,) "The things that I have done, I will repair (make better), I will be on the right road."'

The use of reflexive pronouns in quotes referring to overtly expressed speakers that are first person or second person pronouns is ungrammatical:


With respect to the position of the quote in relation to the utterance predicate we can state that there are four options available:


24.4 The syntactic properties of complement clauses


The first and the second option prevail among the examples from the Sanzhi corpus that have been presented in this section. Instances of a matrix utterance verb followed by the quote can be found in (74), (81), (80), and (93), and the reverse order is illustrated in (71), (72), and (76). The third option means that the quote is interrupted by the verb of speech. The constituent that follows is typically a focused item that is newly introduced, as in (73), or, more frequently, a contrastive topic that is stressed and emphasized, as in (86) and (95). This type of constituent order is unattested for all other kinds of complement clauses that have been discussed in the previous sections and only found with reported speech.

(95) "hextːu there.up uq'-ij go.m.pfv-inf zamana time b-akːu" n-cop.neg haʔ-ib=da say.pfv-pret=1 "heχ-i-la" dem.down-obl-gen 'I said "He does not have the time to go there."'

The position of the quotative particles is mostly at the right edge of the quote, which can easily be explained by their origin. Since they are transparently derived from converbs, they occupy the most common position of converbs in adverbial clauses, that is, the final position (see §25.1 for the constituent order in adverbial clauses). However, occasionally one finds examples in which the quotative particle occurs within the quote, as in the following sentence (96). Example (96) can be analyzed in analogy to (95) with the only difference being that in (95) the matrix predicate separates the contrastive topic from the rest of the quote, whereas in (96) it is the quotative particle that is followed by the contrastive topic.

(96) hek'-i-l dem.up-obl-erg b-urs-ul n-tell.pfv-icvb ca-b cop-n "ce=jal what=indef te-d" exist.away-npl ∅-ik'-ul m-say.ipfv-icvb "hetːu-d" there-npl

'He is telling that there is something there (i.e. to steal).'

### **24.4 The syntactic properties of complement clauses**

All complement taking predicates in Sanzhi that have been analyzed so far occur in the position of objects, i.e., patients or stimuli. So far I did not find complement-taking predicates for which the complement clause is required to function as a subject-like argument.

Complement clauses show many overlaps in their structure with the other types of subordinate clauses (relative clauses, adverbial clauses), as the following paragraphs will

### 24 Complementation

make clear. The argument structure of complement clauses is like that of main clauses: all arguments are retained and adjuncts can be freely expressed. Verbs in complement clauses retain the distinction between imperfective and perfective aspect because this is expressed through the stem and there are no restrictions on negation or on word formation, i.e., all types of derived or compound verbs can be used (27).

The number of verbal categories expressed depends on the complementation strategy. Zero-marked complements and those bearing the embedded interrogative enclitic or containing quotative particles express the same number of categories as main clauses, i.e., person marking, TAM marking, and illocutionary force marking are fully retained. For all other strategies (converbs, cross-categorical suffixes*-ce* and *-il*, infinitive, subjunctive, and masdar) the number of categories expressed in the complement clause is smaller than in the main clause. For instance, the marking for illocutionary force and for person is excluded (except for the subjunctive with its rudimentary person paradigm). Tense marking is largely impossible except for the opposition preterite participle vs. modal participle, which functions as a basic distinction between past time reference and everything else as the following elicited minimal pair illustrates:

(97) dam 1sg.dat b-alχ-a-d n-know.ipfv-hab.pst-1 [it that s-erʁ-an-ce hither-come.m.ipfv-ptcp-dd.sg / / sa-jʁ-ib-ce] hither-come.m.pfv-pret-dd.sg ' I know that he will come/came.' (E)

The constituent order in complement clauses is more frequently verb final than in main clauses, but this is not a strict requirement, e.g. (45). In order to make some preliminary generalizations with respect to the position of the complement clause, I counted all non-elicited complement constructions in this section whose structure is unambiguous and which do not represent reported speech (see §24.3.2 above for the position of the quote in reported speech constructions). The total number is 54, among which one half has the order *matrix verb-complement*, and the other half has the reverse order. Within this data, there is a very small tendency to have the order *matrix verb-complement* the longer the complement is, but this needs further research. It is rare for the complement clauses to be center-embedded into the matrix clause, but two sentences in this section belong to this category, e.g. (33).

What concerns co-reference across the complement and the main clause, Sanzhi has complement control constructions with obligatory subject omission in the complement clause if the latter is headed by the infinitive (or subjunctive). For the details see §24.5 below. In case of co-referential arguments, the overt argument normally occurs in the matrix clause (e.g. (37) among many others). Occasionally, one can find examples that might look like they are contradicting this claim (98). The matrix clause in (98) contains an adverbial *ʡaˁʁunil* 'necessarily, needed', and if we assume that there is an absent argument in this clause that shares the reference with the subject in the complement, then this argument bears the semantic role of a beneficiary or some other role similar to an ethical dative. In other words, it is not a subject or subject-like argument.

24.4 The syntactic properties of complement clauses

(98) ca one zamana, time durħ-ne, boy-pl [nušːa 1pl cellij why ʡaˁlħaˁm-li-j condolence-obl-dat d-uˁq'-ij] 1/2pl-go.pfv-inf ʡaˁʁuni-l=de needed-advz=pst ' One time, guys, we had for some reason to go to offer condolences.'

In a contrastive context, in which arguments are compared to each other, it is possible to add a subject to an infinitival clause in a control construction:

(99) dam 1sg.dat b-ikː-ul=da n-want.ipfv-icvb=1 [du-l 1sg-erg tort tart b-arq'-ij n-do.pfv-inf cara-lli-ja-r] other-obl-loc-abl ' I want to make the tart (rather) than another person (making the tart).' (E)

Co-reference between third person arguments, most notably between the subject in the matrix clause and any argument or adjunct in the complement clause, is expressed by the use of reflexive pronouns. For example, the omitted subject in (100) shares the referent with the goal argument in the complement, which is encoded by the reflexive pronoun in the dative case. Other instances can be found in (132), in which the agent in the complement is co-referential, and in (133), in which the possessor is co-referential.

(100) [cini-j refl.sg.obl-dat d-aˁq-ib-te=ra] npl-hit.pfv-pret-dd.pl=add han remember d-irk-ul, npl-occur.ipfv-icvb ca-w=ra refl-m=add ka-jž-ib down-be.m.pfv-pret ca-w cop-m 'He is sitting and remembering how (they) beat him up.'

Multiple embeddings are possible though rare in natural texts. Relevant examples are (27), which is a translation from Russian, and (102). Both examples have 'be able' in the complement in the middle and therefore an infinitive in the most deeply embedded clause. Other examples in this section illustrate reported speech which is itself complex containing a complement clause (37). The elicited example (101) shows that multiple embedding is allowed with complementation strategies other than the infinitive.

(101) dam 1sg.dat a-b-alχ-ul=de neg-n-know.ipfv-icvb=pst [at 2sg.dat b-alχ-an-ce n-know.ipfv-ptcp-dd.sg [nišːa-la 1pl-gen qili-w home-m ʡaˁbdul Abdul le-w-ce]] exist-m-dd.sg

' I did not know that you know that Abdul was at our place.' (E)

Matrix predicates that have agreement prefixes and non-absolutive arguments usually exhibit local agreement in which the matrix verb agrees with the complement clause as a whole and therefore has the prefix *b-* (neuter singular) (102). This prefix can also be considered to be the default prefix when there is no agreement controller (see §20.2).

(102) dam 1sg.dat han seem b-ič-ib n-occur.pfv-pret [[r-aš-ij] f-go-inf r-irχ-ul f-be.able.ipfv-icvb akːu cop.neg r-ik'-ul] f-say.ipfv-icvb ' I thought that she cannot walk.'

### 24 Complementation

Sanzhi has, in principle, long-distance agreement in gender and number between the matrix predicate and the absolutive argument in the complement clause, as in (103), as well as in (35) and (65) (see also §20.2.1). But in contrast to other East Caucasian languages in which this is a relative common construction (e.g. Tsezic languages, see Polinsky & Potsdam 2001 and Polinsky & Comrie 2003 for Tsezic, and Forker 2013a: 628–639 for Hinuq and further references to the literature), long-distance agreement is almost unattested in the Sanzhi Dargwa corpus.

(103) ašːij 2pl.dat b-ikː-ul=de n-want.ipfv-icvb=pst [d-isːu-tːaj 1/2pl-cry-subj.2 / / d-isː-ij] 1/2pl-cry-inf 'You wanted to cry.' (E)

As in other varieties of Dargwa (Serdobolskaya 2010), only a few complement-taking predicates allow for long-distance agreement, most notably 'want' (103), 'know' (104), 'finish' (35), and 'see' (65), (64). The complement clauses must be of the potential type or of the activity type and can only contain the infinitive/subjunctive (103), the imperfective converb (64), or the perfective converb (35).

(104) [ceʁuna which χurejg food d-arq'-ij] npl-do.pfv-inf b-alχ-atːe n-know.ipfv-prs.2 / / d-alχ-atːe npl-know.ipfv-prs.2 at? 2sg.dat ' Which food do you know how to cook?' (E)

Serdobolskaya (2009; 2010) argues that in Xuduc and Qunqi Dargwa long-distance agreement can be analyzed as clause reduction (clause union) that shares many properties with raising constructions in other languages. She shows that complement constructions with embedded subjunctives/infinitives or converbal clauses have some monoclausal properties. This seems to be true for Sanzhi as well. For instance, arguments of embedded infinitival clauses can easily occur in a clause-final position that can hardly belong to the embedded clause (105).

(105) atːa-j=ra father-dat=add aba-j=ra mother-dat=add [darman medicine b-arq'-ij] n-do.pfv-inf b-ikː-ul n-want.ipfv-icvb ca-b cop-n durħuˁ-la, boy-gen durħuˁ-li-j boy-obl-dat 'The father and the mother want to give medicine of the son, to the son.' (the speaker corrected herself)

The pragmatic effect of long-distance agreement is sometimes described as highlighting the argument that serves as agreement controller, but before being able to make more specific claims about its impact on information structure in Sanzhi Dargwa more research is needed.

24.5 Argument control in complement constructions

### **24.5 Argument control in complement constructions**

Complement constructions in Sanzhi show heterogeneous behavior with respect to control of the obligatorily omitted argument. Complements of the verb *b-aʔašː-* 'begin' can be headed by the imperfective converb (§24.2.8) or by the infinitive/subjunctive (§24.2.6). The controller, i.e., the one who begins something, must be in the absolutive. The controllee can be the single argument of an intransitive verb or the most prominent argument of a two-place verb as the following examples show:



The controllee can never be the second argument of a two-place verb, such as the patient (107a) or the stimulus (107b).

	- b. \* maˁʡaˁlim teacher w-aʔ.ašː-ib m-begin-pret [Murad-li-j Murad-obl-dat \_ abs čirʁ-ij] understand-inf (Intended meaning: 'The teacher began to be understood by Murad'.) (E)

But if we look at bivalent complement-taking predicates, the situation is slightly different. With the matrix verb *b-ikː-* 'want' the complement clause contains either an infinitive or a subjunctive. The controllee can be the subject of an intransitive verb (36). But it can also be any of the arguments of a two-place verb (e.g. agent or patient), depending on the verb form in the complement clause. With subject-like controllees the embedded verb takes the infinitive suffix (108), (138).

(108) Murad-li-j Murad-obl-dat a-b-ikː-ul=de neg-n-want-icvb=pst [ \_ erg ʡaˁli Ali qːurt push w-arq'-ij] m-do.pfv-inf 'Murad did not want to push Ali.' (controllee = A) (E)

However, if the controllee is the second argument of a two-place predicate, then the verb form in the complement clause cannot be the infinitive, but must be the perfective converb (109). The infinitive can only occur when the experiencer of 'want' is controlling a subject-like argument in the complement clause. In (109b) both verbs have different arguments, and the embedded verb cannot bear the infinitive suffix.

### 24 Complementation

(109) a. Murad-li-j Murad-obl-dat b-ikː-ul n-want-cvb ca-b cop-n [Madina-j Madina-dat \_ abs či-w-až-ib-le] spr-m-see.pfv-pret-cvb 'Murad wants Madina to see him .' (controllee = P) (E) b. it-i-j that-obl-dat b-ikː-ul n-want.ipfv-icvb ca-b cop-n [du-l 1sg-erg kaʁar letter b-elk'-un-ne n-write.pfv-pret-cvb / / \* b-elk'-ij] n-write.pfv-inf 'He wants that I write the letter.' (E)

The same phenomenon is observed with another complement-taking predicate *uruχle cab-* 'fear'. If the controllee is a subject-like argument, the complement clause is headed by an infinitive (110a). Otherwise the attributive suffix *-ce* is employed, as in (110b) and (110c).

(110) a. χamis Khamis uruχ-le fear-advz ca-r cop-f [ \_ dat sːika bear či-b-až-ij] spr-n-see.pfv-inf 'Khamis fears to see the bear.' (controllee = A) (E) b. χamis Khamis uruχ-le fear-advz ca-r cop-f [Madina-l Madina-erg \_ abs qːurt push r-irq'-an-ce] f-do.ipfv-ptcp-dd.sg 'Khamis fears that Madina pushes her.' (controllee = P) (E) c. ʡaˁli Ali uruχ-le=de fear-advz=pst [Madina-j Madina-dat \_ abs a-w-aχ-ur-ce] neg-m-know.pfv-pret-dd.sg 'Ali feared that Madina would not recognize/know him.' (controllee = P) (E)

It seems that with trivalent matrix verbs there is no such difference between the treatment of subject controllees on the one hand and object controllees on the other hand. Both types are allowed and the embedded verb forms are identical (111–113).


### 24.5 Argument control in complement constructions

However, this again can be interpreted as a difference in the treatment of subject-like vs. object-like arguments, but now regarding the controller, not the controllee. If the controller is the subject, then the verb form in the complement clause depends on whether the controllee is the object or also the subject. If the controller is the object, then, in contrast, no such difference in the verb form is noticed. To sum up, in complement control we have some indication of an S/A pivot. There are no clause level conditions and at least for the tested complement-taking predicates no difference in the treatment of embedded predicates could be observed. The predicate class of the embedded verb is possibly a decisive feature that needs to be studied in more detail in the future since for other East Caucasian languages it has been observed that intransitive, canonical transitive, and affective verbs are treated differently in some complement constructions (Kibrik 2003).

Finally, I will briefly discuss backward control. This term refers to complement constructions in which the overt controller appears in the embedded clause, and thus its case is assigned by the embedded verb. Nevertheless, the matrix verb shows agreement with the controller. On the surface these constructions look as if the verb is agreeing with a non-absolutive argument. But instead it is argued that the matrix verb contains a covert controllee in the absolutive case that is co-referential with the overt nominal in the non-absolutive case. Backward control is found in other East Caucasian languages, see, e.g. Polinsky & Potsdam (2002; 2006) on Tsez, and Serdobolskaya (2010) on Qunqi Dargwa, and is typically restricted to a few modal and phasal predicates.

In Sanzhi Dargwa, there are two verbs that allow for backward control, *-b-iχʷ-* (pfv)/*birχʷ-* (ipfv) 'can, be able' and *b-aʔ axː-* (pfv), *b-aʔ b-išː-* (pfv)/*b-aʔ b-irxː-* 'begin, start'. The verb 'can, be able' is far more readily available. In the Sanzhi corpus, backward control is only attested with the verb 'can, be able', but can be obtained with 'begin, start' in elicitation. In standard forward control constructions, the two verbs require subject-like arguments in the absolutive case that control gender (and person) agreement just like intransitive verbs:


In backward control constructions the clauses contain subject-like arguments in the ergative case that has been assigned by the embedded verb. This means that in both (116) and (117) the subject argument bears the ergative case because the embedded verbs are transitive.

(116) [rursːi-l girl-erg kːurtːi dress b-arχ-ij] n-sew.pfv-inf r-irχʷ-an-ne f-be.able.ipfv-ptcp-fut.3 ' The girl will be able to sew the dress.' (E)

### 24 Complementation

(117) [rursːi-l girl-erg palaw pilau b-uk-unne] n-eat.ipfv-icvb r-aʔ f-begin r-išː-ib f-become.pfv-pret / / r-aʔ.ašː-ib f-begin.pfv-pret ' The girl began to eat the pilau.' (E)

Backward control is only available with embedded transitive verbs. Affective verbs do not allow for this construction. There are two cases that look like apparent exceptions. In example (118), the experiencer argument occurs in the dative, assigned by the affective verb *b-ikː-* 'want, like, love' appearing as the imperfective converb, whereas the finite verb is *haq-*, which usually translates as 'manage, be enough'. Thus, one might suspect that *haq-* functions as a matrix complement-taking verb into which a complement clause headed by *b-ikː-* has been embedded together with both the experiencer and the stimulus argument. However, *b-ikː-ul haq-* rather functions as a lexicalized periphrastic predicate and the construction is monoclausal. The verb *b-ikː-ul* cannot be replaced by any other verb and the semantics of the periphrastic predicate is not transparently composed of the semantics of the individual predicates.

(118) ca one qːuʁa beautiful žahil young durħuˁ-li-j boy-obl-dat r-ikː-ul f-want.ipfv-icvb haq-ib manage.pfv-pret ca-r cop-f žahil young rursːi girl ' One beautiful young man fell in love with a young girl.'

The second apparent exception is the use of affective predicates that are usually bivalent as monovalent predicates. This is possible with 'see', which then has the meaning 'be/become visible' and 'hear', which then means 'be/become audible'. Thus, the dative nominals in the following two examples can be left out, such that we end up with intransitive constructions. The agreement on the verb 'begin' is controlled by the absolutive arguments, not by the dative experiencers, which is an unambiguous indication that we do not deal with backward control with an embedded bivalent affective verb, but with an intransitive complement clause in a construction to which a dative adjunct has been added.


(122) dam 1sg.dat dubur-te mountain-pl či-d-ig-ul spr-npl-see.ipfv-icvb d-aʔ.ašː-ib npl-begin.pfv-pret ' I began to see the mountains.' OR 'The mountains started to be visible to me.' (E)

If the constructions were truly biclausal, we would expect restrictions on the constituent order, since items of one clause should normally not be allowed to appear within the other clause. This is precisely what we find with 'begin, start'. In a backward control construction, the ergative argument must occur within the complement clause; it cannot be positioned clause-initially if it is followed by the matrix predicate. This is only possible in forward control since then the argument is governed by the matrix predicate:


For the verb *b-iχʷ-* (pfv)/*b-irχʷ-* (ipfv) 'be, become, happen, can, be able' the data are not so clear. Some examples show a very flexible word order, which points towards a monoclausal analysis with a periphrastic predicate:


Other examples have been rejected by speakers. Further research is needed to provide a more detailed account of the properties of backward control in Sanzhi.

### **24.6 Constructions that semantically resemble complement clauses**

### **24.6.1 Parentheticals**

There are three particles and phrases that refer to cognitive activities and are used in a way that resembles complement clauses with cognition predicates. The first is the frequently used phrase *(possessor) pikri ħaˁsible* 'according to the thoughts of (somebody)', which consists of a possessor followed by the two borrowed items *pikri* 'thought' and

### 24 Complementation

the postposition *ħaˁsib-le* (test-advz) 'according to'. The entire construction is a kind of calque that partially consist of loans and partially of Sanzhi morphemes. Zero marking is the only possible usage option for the clauses with which the phrase occurs, as in (127), (128); the use of quotative particles or other complementation markers together with the phrase is ungrammatical. In contrast to the common positions of complement clauses as either following, preceding or occasionally being center-embedded into the matrix clause (see §24.4 below), the phrase frequently occurs in the middle of the complement. Taking all these peculiarities together, the phrase has to be characterized as a parenthetical.

(127) hej, this di-la 1sg-gen pikri thought ħaˁsible, following tusnaq-le-r prison-loc-abl tːura outside uq-un go.m.pfv-pret ca-w cop-m hež

this

' In my mind, he left prison.'

(128) hež-i-la this-obl-gen xːunul, woman di-la 1sg-gen pikri thought ħaˁsible, following χʷe=ʁuna dog=eq wahi-ce bad-dd.sg ca-r cop-f 'His wife is, in my mind, bad like a dog.'

Note that other complement-taking predicates can also occur in the middle of the complement clause, in which case additional overt marking of the complement by means of particles is forbidden. For instance, in (128) we can replace *dila pikri ħaˁsible* by *dam han birkul cab* 'seems to me'. This complement-taking predicate normally requires the use of the quotative particle when it occurs before the complement clause.

The phrase *... pikri ħaˁsible* is sometimes replaced by its Russian equivalent *po-moemu*, which is used in the same manner (129).

(129) pomoemu, my.opinion atːa-l father-erg aba mother r-it-ib f-beat.up-pret ca-r cop-f 'In my mind, the father beat up the mother.'

Moreover, the particle *aχːu* 'I don't know, dunno' also occasionally occurs as a parenthetical. In (130) there is not only no formal sign of subordination, but not even a clear semantic relationship between *aχːu* and the surrounding clauses, so this is an example of its parenthetical use. However, in the majority of examples the clause accompanying the particle is marked by the embedded question marker =*el* (§24.2.7).

(130) amma but ʁaˁʁ scream ∅-ik'-ul m-say.ipfv-icvb ca-w cop-m ik', dem.up aχːu, not.know nuˁq-be arm-pl aq high d-arq'-ib npl-do.pfv-pret ca-d cop-npl ik'-i-l=ra dem.up-obl-erg=add 'But he is screaming, I don't know, he also rose up his arms.'

24.6 Constructions that semantically resemble complement clauses

### **24.6.2 Nominalized relative clauses resembling complement constructions**

There are a variety of constructions with the predicates listen in §24.1 above as "complement-taking predicates", which are syntactically not complement clauses. They function as a core argument of a higher clause and have the internal constituent structure of a clause. But they are usually shorter than real complement clauses and contain only one core argument in addition to the verb. Most importantly, they do not refer to propositions, but to entities such as persons, events, etc. They are, therefore, not complement clauses but headless relative clauses, which have been nominalized. Headless relative clauses can be formed with (i) the cross-categorical suffix -*ce* (plural -*te*), the crosscategorical suffix -*il*, and the modal participle -*an*. In this section, I will only discuss headless relative clauses that occur together with complement-taking predicates and show how they differ from true complementation. For general information about headless relative clauses see §23.4.

As has been described in §24.2.3 above, the suffix -*ce* marks complements of the facttype (131). In this function, -*ce* can never be replaced by -*te*, which otherwise function as the plural equivalent of -*ce*. Thus, replacing -*ce* by -*te* in (131) would result in an ungrammatical sentence.

(131) du-l 1sg-erg razi-l=da happy-advz=1 [c'il then dus year [nušːa-l 1pl-erg basːejn pool b-arq'-ib-le] n-do.pfv-pret-cvb ha-b-urχː-an-ce] up-n-finish.ipfv-ptcp-dd.sg

' I am happy that during the next year we will finish building the pool.' (E)

By contrast, when the suffix occurs in headless relative clauses, the use of the plural suffix *te* is possible when the referent of the nominalized clause is plural (132), (133). In (133) the two nominalized relative clauses are conjoined by means of the additive enclitic =*ra*, which is regularly used to conjoin noun phrases (§26.1). Examples (132) and (133) share with genuine complement clauses their occurrence in the argument position of verbs of cognition and their ability to preserve their internal argument structure. For instance, in (132) the agent of the embedded verb is expressed by means of a reflexive pronoun in the ergative case, which is in accordance with the usual case frame required by transitive verbs. However, the embedded clauses are propositions (e.g. they cannot be expressed through a *that*-clause in English). This is particularly clear in (133) because in this example the nominalized clauses are, first of all, conjoined like ordinary noun phrases, and second, modified by a possessor, which refers to the agent, and by a quantifier. In a complement clause the agent of the embedded verb is not expressed by a possessor in the genitive, but by a nominal in the ergative case.

(132) [cin-ni refl.sg-erg d-arq'-ib-te] npl-do.pfv-pret-dd.pl han remember d-irč-aq-ul npl-occur.ipfv-caus-icvb ca-w cop-m uže already 'He remembered what he had done.'

### 24 Complementation

(133) pikri thought uq-un-ne go.m.pfv-pret-cvb [cin-na refl.sg-gen li<d>il all<npl> [d-urs-an-te=ra] npl-tell-ptcp-dd.pl=add [d-irq'-an-te=ra]] npl-do.ipfv-ptcp-dd.pl=add ... ' and thought of what he had said and done ...'

The same kind of reasoning applies to nominalized clauses with the cross-categorical suffix -*il*. The following two sentences show a minimal pair. The first example illustrates the headless relative clause, which refers to an animal (134a). The second clause is a complement construction with a clausal complement (134b).

	- this-obl-dat n-know.pfv-pret cop-n npl-eat.pfv-msd 'He got to know that (they) ate (them).' (E)

The use of nominalized clauses with the -*il* suffix is ungrammatical in constructions that require clausal complements:

```
(135) * du-l
        1sg-erg
                 razi-l=da
                 happy-advz=1
                                 [c'il
                                 then
                                       dus
                                       year
                                            [nušːa-l
                                            1pl-erg
                                                    basːejn
                                                    pool
                                                            b-arq'-ib-le]
                                                            n-do.pfv-pret-cvb
        ha-b-urχː-an-il]
        up-n-finish.ipfv-ptcp-ref
        (Intended meaning: ' I am happy that during the next year we will finish
        building the pool.') (E)
```
As has been shown in (133) for nominalized clauses with -*ce*, nominalized clauses can also be conjoined:

(136) na now [[b-urs-ib-il=ra] n-tell-pret-ref=add [b-arq'-ib-il=ra]] n-do.pfv-pret-ref=add qum.urt-u forget.ipfv-prs.3 dam 1sg.dat ' I forget what I say and what I do.'

The modal participle regularly occurs in headless relative clauses §23.4. When the main clause contains a complement-taking predicate as in (137), the structure seems to be ambiguous between an interpretation as a complement clause of the activity type ('studying') and a nominalized verb that functions as action noun ('the studies'). This type of construction requires future research in order to be able to decide whether the construction is, in fact, ambiguous, or whether we can exclude one of the two potential analyses.

(137) ucːi-l brother-erg taman end b-arq'-ib n-do.pfv-pret [b-uč'-an] n-learn.ipfv-ptcp ' Brother finished studying/the studies.' (E)

24.6 Constructions that semantically resemble complement clauses

In elicitation, I also obtained example (138). This looks like a complement clauses of the potential type with the modal participle. At the same time this sentence instantiates a constituent focus construction with a floating predicative particle (the past enclitic =*de*, see §27.3.2 for more information).

(138) dam 1sg.dat b-ikː-an n-want.ipfv-ptcp [χadižat-li-j=de Khadizhat-obl-dat=pst kiniga book lukː-an] give.ipfv-ptcp ' I wanted to give the book to KHADIZHAT.' or 'It was Khadizhat to whom I wanted to give the book.' (E)

Interestingly, it is impossible to place the predicative particle =*de* on its usual host, which would be the verb in the main clause (139). This is only allowed if we simultaneously replace the modal participle with the infinitive, which is the default marker for complement clauses with potential meaning (140). At the same time the use of the infinitive instead of the modal participle in (138) is ungrammatical because the constituent focus construction requires the use of participles.


The cross-categorical suffixes *-il* and -*ce* (as well as the masdar) can take case suffixes. Occasionally these nominalized verbs can occur in the argument position of complementtaking predicates. For examples, the verb 'believe' regularly requires the dative case on its goal argument. In sentence (141) the goal argument is a nominalized clause with its own arguments and adjuncts.

(141) ča-k'al who-indef w-iχčit m-believe ag-ur-il go.pfv-pret-ref akːʷ-i cop.neg-hab.pst [du 1sg ce what ∅-iχʷ-ni-li-j m-be.pfv-msd-obl-dat hek' dem.up kuzaw-le-w=či-w] coachwork-loc-m=on-m ' Nobody believed in what had happened to me there up on the coachwork (on the car).'

For the utterance verbs and cognition verbs whose complement clauses denote speech acts or similar types of activities that require the use of language it is possible to use the postposition *qari-či-b* followed by a participial clause with the appropriate case marking to denote the topic of the speech act (§8.1.6) or the topic of the cognitive act (142). Syntactically this construction is not a relative clause but a nominalized case-marked verb, which is governed by a postposition.

24 Complementation

(142) pikri thought ∅-ik'-ul m-say.ipfv-icvb ca-w cop-m [cin-ni refl.sg-erg d-arq'-ib-t-a-lla npl-do.pfv-pret-pl-obl-gen qari=či-b] on.top=on-n 'He thinks about what he had done.'

### **24.6.3 Adverbial clauses used with emotion and cognition predicates**

The possibility of an analysis adverbial clauses instead of a complement constructions has been amply discussed in §24.2.5 for the perfective converb. Sentence (143) shows another example of a construction for which, however, only the adverbial-clause interpretation is available because the modal participle is followed by temporal/causal enclitic *=qːel* 'when, while, because', which regularly occurs in adverbial clauses (§18.2.1). Precisely to what extent these constructions are used, as well as their semantic and morphosyntactic properties, needs to be clarified by future research.

(143) han remember d-irk-ul npl-occur.ipfv-icvb d-urkː-ar npl-find.ipfv-prs.3 [heχ dem.down cin-ni refl.sg-erg it-an=qːel] beat.up-ptcp=when ' He is probably remembering when he was beaten.'

## **25 Syntactic properties of adverbial and conditional clauses**

This chapter analyzes the syntax of adverbial and conditional clauses in Sanzhi and compares them to the syntactic properties of similar clauses in other East Caucasian languages.

### **25.1 The syntax of adverbial clauses**

Sanzhi has different types of adverbial clauses that can be distinguished by the morphological make-up of the verb forms in the subordinate clause and by their semantics. Semantically, we can distinguish between simple converbs with a fairly general meaning and specialized converbs with a rather specific temporal or non-temporal meaning. The first group consists of the imperfective (§18.1.1.1) and perfective converb (§18.1.1.2). The second group contains temporal, causal, and other converbs (§18.2). A similar distinction is found in many East Caucasian languages (e.g. in Tsezic, see Comrie et al. 2012, and in Dargwa varieties, see Belyaev 2010). The syntactic characteristics of constructions with general converbs have repeatedly been discussed in the literature because they exhibit a mixed behavior, showing features of subordination as well as of coordination (see, among others, Kazenin & Testelets 2004; Haspelmath 1995; Belyaev 2010; Comrie et al. 2012; Creissels 2010; 2012; Forker 2013c). Sentences in Sanzhi can be fairly complex, containing a number of adverbial clauses that are combined with one main clause. Semantically, these clauses either resemble coordination, as in (1), or subordination, when the meaning of the adverbial clause is causal (2).

(1) amχa donkey [b-arcː-ur-re] n-get.tired.pfv-pret-cvb [ka-b-ič-ib-le] down-n-occur.pfv-pret-cvb b-ebč'-ib n-die.pfv-pret ca-b

cop-n

' The donkey got tired, fell down, and died.'

(2) bahsar first [heχ dem.down cin-na refl.sg-gen atːa-la father-gen jurt-la house-gen qːarqːa stone ʡaˁbal three qal-la house-gen xːari down k-ag-ur-re] down-go.pfv-pret-cvb [qːaq-li-j back-obl-dat či-ka-d-irxː-ul] spr-down-npl-put.ipfv-icvb [ha-d-iqː-ul] up-npl-carry.ipfv-icvb qːarqːa=ra stone=add gu-r-h-aqː-ib=da sub-abl-up-carry-pret=1

' First, (because) the stones of father's house had fallen down three floors, we put them on the back and carried them, carried the stones.'

### 25 Syntactic properties of adverbial and conditional clauses

The perfective converb is widely used in procedural texts, such as the description of how to prepare dishes. These texts consist of a list of actions that are expressed by verbs bearing perfective converb suffixes with a main clause at the end. The actions are supposed to occur in the order in which the clauses follow each other, i.e., there is iconicity, and the order of the clauses cannot be changed without changing the meaning of the whole sentence. This is generally interpreted as a semantic feature of coordination, as opposed to subordination, where the order of the clauses does not reflect the temporal order of the events and can therefore be changed without a concomitant change in the meaning. Linear order will be discussed in more detail in §25.1.4 below.

Converbs are non-finite in the sense that they head only subordinate clauses. The two general converbs (imperfective, perfective) also occur in analytic tenses in main clauses (Chapter14), but only when combined with a copula or a predicative particle (§9.1). Therefore, they are unable to express illocutionary force or absolute temporal reference but share those properties with the verb form in the main clause (see §25.1.1 below). They are also not marked for person by person suffixes or enclitics, in contrast to the verb forms in the superordinate clause. However, they express aspect, because aspect is mainly conveyed through the verbal stem and there are no restrictions concerning the use of perfective or imperfective stems in adverbial clauses. Moreover, they can have their own arguments that fulfill the same grammatical roles as arguments in main clauses, i.e., case marking patterns in adverbial clauses and main clauses do not differ. Furthermore, gender agreement is present in adverbial clauses. In contrast to main clauses, it is strictly controlled by the absolutive argument. By contrast, in main clauses copulas can exhibit gender agreement with ergative or dative arguments. However, these copulas cannot occur in subordinate clauses.

The constituent order in adverbial clauses shows a far greater tendency for verb-final order than is observed for main clauses (2), but adverbial clauses in which the verb is followed by other constituents can be found as well (3), (20).

(3) [hel=ʁuna that=eq musna-w place.loc-m ink meet w-aq-ib=qːel m-go.through.pfv-pret=when du=ra] 1sg=add dam=ra 1sg.dat=add ʡaˁħ-le=kːʷi good-advz=neg.pst

' [When I was in that situation], when I also was in a place like this, I also did not feel well.'

In the following discussion, I will adopt the typology of Bickel (2010) for the investigation of clause-linkage patterns. Bickel's typology consists of eleven variables, which are reproduced in the first column of Table 25.1. A short description is given in the second column of the same table.

I will additionally use a number of other criteria that have been proposed in order to differentiate between coordination and subordination, namely co-reference and expression of shared arguments, morphosyntactic locus, and relativization of constituents of adverbial clauses.

25.1 The syntax of adverbial clauses


Table 25.1: Syntactic variables for the analysis of adverbial clauses

I will mainly analyze the two general converbs as well as the temporal converb *=qːel(la)* 'when, while, because', which expresses temporal simultaneity and anteriority as well as causality, because these converbs show the largest semantic overlaps and are semantically close to coordination.

### **25.1.1 Scope properties**

Adverbial clauses do not contain markers for illocutionary force, such as the imperative, optative suffixes, or the interrogative particles ("banned"). Those markers can only occur in the main clause. Their scope can be restricted to the main clause ("local"), but, in the appropriate context, it can also extent across the adverbial clause ("extensible"). However, the latter possibility is noticeably less common in texts. Examples (4–6) illustrate local scope restricted to the main clause.

(4) [hej this šːal-li-cːe side-obl-in cːiχːin roof.beam ka-b-alt-an=qːel] down-n-put.ipfv-ptcp=when het that šːal-la side-gen ʡaˁnčːi earth a-ka-d-ax-u=w? neg-down-npl-go-prs.3=q

'When you put the roof beam at this (at one) side, does the clay of that (the other side) not fall down?'

25 Syntactic properties of adverbial and conditional clauses


Some converbs seem to fully ban joint scope of illocutionary operators. For instance, interrogative markers (4) or imperative markers (7) cannot scope over the temporal converb *=qːel*, although tense suffixes can.

(7) [t'ult' bread asː-ib=qːel] buy.pfv-pret=when nisːe-cːella cheese-comit b-erkʷ-en! n-eat.pfv-imp ' When you buy bread, eat it with cheese!' (NOT: 'Buy bread and eat it with cheese!') (E)

But at least with the perfective and the imperfective converbs it is also possible that the two clauses have joint scope:


Similarly, adverbial clauses can only express aspectual distinctions because this is a property of the verbal stem. Other semantic categories of verbs such as tense and evidentiality are only available to verb forms in main clauses. The converbs have relative temporal reference. This means that they refer to situations that take place before, after or during the situation that is expressed by the matrix clause. For instance, in (10) the verb form in the main clause has future/modal meaning, which is extended to the adverbial clause with the preterite converb. Sentence (11) conveys past time reference due to the preterite in the main clause, and (12) conveys present time reference because of the compound present tense. Both sentences contain adverbial clauses with the imperfective converb that only expresses that the situation in the adverbial clause took place during the situation described in the main clause.

<sup>1</sup>Within the contexts from which this example originates the subjects of the adverbial clause and the main clause differ. The speaker who was guiding a truck full of people urged them to jump off the car because he had problems controlling it. This means that the full translation is 'While/because my legs are trembling, jump!' Out of context, however, the most natural reading is rather: 'While your legs are trembling, jump!' with a same-subject interpretation.

25.1 The syntax of adverbial clauses


' Very slowly making small steps we went across (the river).'

(12) [qili-r home-f du=gina 1sg=only r-irχ-ul] f-be.ipfv-icvb [a-r-isː-ul] neg-f-cry-icvb r-ug-ul=da f-stay.ipfv-icvb=1 'Being alone at home, I (fem.) stay not crying.'

Similarly, the past perfect used in the main clause of (13) expresses not only past time reference but also indirect evidentiality, which extends to the meaning of the full sentence including the adverbial clause with the preterite converb.

(13) [it-i-sa-r that-obl-ante-abl s-asː-ib-le hither-take.pfv-pret-cvb wec'-nu ten-ten urek-ra six-num azir] thousand it-i-l=ra that-obl-erg=add d-ataʁ-ib-le=de npl-send.pfv-pret-cvb=pst 'From her (he) took 16 000, and he sent (that money to us).'

In short, fewer categories are expressed in adverbial clauses than in main clauses, because person agreement, tense, evidentiality, and illocutionary force are absent. This means that Sanzhi adverbial clauses are, in Bickel's terms, "asymmetrical" and non-finite (Bickel 2010).

### **25.1.2 Focus and question words**

Most but not all focus-sensitive particles can appear in adverbial clauses attached to the converbs. The following examples show the enclitic *=cun* 'only' and the emphatic modal particle *=q'ar* in clauses together with the perfective converb and the *=qːel* converb. The modal particle *=q'al* can also be employed in certain types of adverbial clauses, but in general its use in subordinate clauses is subject to many restrictions (16), (17). The restrictions are specific to this particle and therefore not relevant for a discussion of the morphosyntactic properties of adverbial clauses.


### 25 Syntactic properties of adverbial and conditional clauses


As mentioned in §25.1.1 above, interrogative particles (which also belong to the focussensitive particles) cannot be used in adverbial clauses. However, adverbial clauses with various converbs can contain interrogative pronouns as the following examples with the perfective converb (18) and the converb *=qːel* (19) show.


### **25.1.3 Co-reference and expression of shared arguments**

Converb clauses can almost always have their own subjects that do not need to be coreferential with the subject in the main clause. Examples of adverbial clauses with differing subjects can be found in (20) for the perfective converb, in (21) for the imperfective converb, and in (4) and (16) for constructions with *=qːel*. However, for the sentence in (21) there is no alternative possibility of using a same-subject construction because the two weather verbs grammatically require different subjects. Thus, syntactically (21) is a complex clause with two different subjects, but semantically there is a clear relationship between the two clauses.


### 25.1 The syntax of adverbial clauses

If the subjects differ, it is possible that other arguments are co-referential instead. In (6) the subject of the first clause with the imperfective converb is not identical to that of the following, but can be identical to the omitted possessor (see the comment in the footnote). In (16), the omitted dative subject of the adverbial clause shares the referent with the possessive pronoun in the main clause. Similarly, in (22) the omitted subject of the adverbial clause is identical to the referent of the possessive pronoun in the main clause. It can also be the case that a string of adverbial clauses shares the subject with an adjunct in the main clause.

(22) [can meet ka-b-iž-ib=qːel] down-hpl-be.pfv-pret=when ču-la refl.pl-gen jašaw-li-cːe-b being-obl-in-n zamana time ca-b cop-n 'When they got married, they had a good life.' (lit. When they met it is the time of their well-being.)

The sharing of the subject argument is clearly preferred for the perfective converb and can be seen in most examples in this section. Even in example (20) there is at least a causal relationship between the described events: because of the death of Stalin the trains tooted and honked. If no such causal relationship can be found, a complex clause with different subjects is impossible (23).

(23) ?? [ʡaˁli Ali qili home w-i-ha-w-q-un-ne] m-in-up-m-go.pfv-pret-cvb Indira-l Indira-erg kːurtːi dress b-urχ-ul=de n-sew.ipfv-icvb=pst (Intended meaning: 'When Ali came home, Indira was sewing a dress.')

The requirement for shared subjects is even stronger for the imperfective converb, for which it is almost the only attested possibility in natural texts. By contrast, for *=qːel* it is easy to find examples with differing subjects (25), but still around half to two third of the examples share the subject (24), (7)

(24) [tːura outside sa-w-q-un=qːel] hither-m-go.pfv-pret=when heχ dem.down Allah-li-cːe Allah-obl-in ulkː-un-ne pray-icvb-cvb ' When he left, he prayed to Allah.'

In clauses with disjoint subjects, normally at least one of the subjects (20), (22), if not both are overt. However, even in those cases it is possible that both subjects are absent, as in example (25), in which it is clear from the context that the referent of the subject of the first clause is the children, and that the referent of the subject in the main clause as well as in the following adverbial clause is the main character of the story.

(25) [a-b-ug-an=qːel] neg-hpl-be.calm.ipfv-ptcp=when b-i-ka-b-at-ur hpl-in-down-hpl-leave.pfv-pret ca-b cop-hpl [q'ʷani-l-cːe box-obl-in uruχ fear b-arq'-ib-le] hpl-do.pfv-pret-cvb ' When they did not calm down, (he) put (the children) into the box, frightening

them.'

### 25 Syntactic properties of adverbial and conditional clauses

Co-referential arguments are omitted, so zeroes commonly occur in the subordinate clause. Therefore, cataphora is very frequent. In example (26) the omitted argument in the first clause corresponds to the agent in the second clause.

(26) [bari-la sun-gen gʷana-dex-li-j warm-nmlz-obl-dat šak feel ič-ib-le] occur.m.pfv-pret-cvb il-i-l that-obl-erg bari-li-j sun-obl-dat barkalla thanks b-aχ-aq-ur n-know.pfv-caus-pret 'When he felt the warmth of the sun, he thanked the sun.'

But anaphora is also attested (27). In this example, we find G=S=S=A, with only the first G argument being a full noun phrase and all other occurrences of the same argument left implicit, so that no grammatical relations are involved.

(27) [hitːi after b-uq-un-ne n-go.pfv-pret-cvb č'aka eagle χːʷe-j=ra dog-dat=add hel-i-j=ra] that-obl-dat=add [sa-r-b-uq-un-ne, ante-abl-hpl-go.pfv-pret-cvb sa-r-b-uq-un-ne] ante-abl-hpl-go.pfv-pret-cvb [waˁw call b-ik'-ul] hpl-say.ipfv-icvb b-arčː-ib-le=kːu n-find.pfv-pret-cvb=neg ʡaˁt'a frog

'The bird runs (i.e. flies) after him and his dog, and they run and run, and shout, but they did not find the frog.'

Another strategy commonly employed is to have the co-referential NP in clause-initial position, syntactically belonging to the main clause, but separated from the rest of the main clause in terms of linear order. The controlee is in the embedded clause, resulting in center embedding. In (10), the adverbial clause contains an intransitive predicate; therefore, the pronoun *dul* '1sg.erg' must be part of the main clause. If both clauses have the same valency frame, it is in principle impossible to decide to which of the two clauses the overt argument belongs. In general, arguments whose referents the speaker assumes to be known to the hearer are left implicit such that often none of the clauses contains an occurrence of the shared arguments.

Though shared arguments are very common, this is not a necessity. In (20) the first adverbial clause contains an overt S, *Istalin*, which is not shared in the subsequent adverbial and main clause.

The adverbial clause mostly precedes the main clause, but the reverse order is also attested (§25.1.4). Shared S and A arguments in either order are frequently found in texts (26), (10), and are easily provided in elicitation (28a), (28b). The situation gets more complicated if P arguments are also involved. An overt S argument in the first clause can correspond to a covert P in the second clause but not if the verb in the subordinate clause bears the converb suffix *-le*. Instead, the more specific construction with *=qːella* must be used such that the first clause is not only syntactically but also semantically an adverbial clause (28c). According to my Sanzhi consultants, the more general converb *-le* can only be used if the S in the converbal clause corresponds to an S or A in the main clause.

25.1 The syntax of adverbial clauses

	- b. [Murad-li-j Murad-obl-dat Madina Madina či-r-až-ib-le] spr-f-see.pfv-pret-cvb \_ abs ag-ur go-pret 'Murad saw Madina and went away.' (A = S) (E)
	- c. [rursːi daughter sa-r-eʁ-ib=qːella] hither-f-come-pret=when aba-l mother-erg \_ abs r-aχː-un f-feed-pret 'When the daughter came, the mother fed (her).' (S = P) (E)

If the first clause contains two arguments A and P, then an implicit S in the second clause can, in principle, be co-referential with any of these two arguments. However, co-reference with P is less preferable, i.e. in example (29), the S argument in the second clause can be co-referential with P in the first clause, or with another argument previously established in the context. In natural texts the co-referential argument would rather be expressed as S in the main clause and left implicit in the adverbial clause. In (28b), co-reference between the A in the first clause and S in the second clause is the preferred reading, and co-reference with a third person is rather unlikely.

(29) [atːa-j father-dat Madina Madina či-r-až-ib-le] spr-f-see.pfv-pret-cvb \_/ abs razi happy r-iχ-ub f-become-pret 'Father saw Madina and (she) got happy.' (P = S) (E)

If we exchange the predicate in the second clause in (28c) with a transitive predicate, we again encounter the same situation. If the shared argument occurs as P in the adverbial clause, the whole sentence becomes rather marginal because out of context the referent of the omitted A in the main clause could be either the mother or the daughter. Therefore, speakers prefer to express the shared argument as A in the main clause (30).

(30) [aba-l mother-erg \_ abs až-aq-ur-re] go.pfv-caus-pret-cvb rursːi-l girl-erg qal house qʷaˁrš sweep b-arq'-ib n-do.pfv-pret 'Mother called her daughter and she (= the daughter) swept the house.' (P = A) (E)

Thus, there is some evidence that shared arguments are preferably expressed as S or A instead of P. However, co-reference is never a grammatical necessity. In each of the sentences an implicit argument can always be co-referential with other referents in the contexts that do not occur in the sentence to which the omitted argument belongs.

Pronouns (demonstrative or reflexive) in combination with co-referential noun phrases are usually not employed to express shared arguments, because the use of pronouns often leads to disjoint reference as the only available interpretation. Adverbial clauses preceding the main clause never allow for pronominal cataphoras as we know them from European languages. This means that the demonstrative or reflexive pronoun in (31) cannot be co-referential with a following noun phrase.

### 25 Syntactic properties of adverbial and conditional clauses

(31) [cin-ni refl.sg-erg / / it-i-l that-obl-erg t'ult' bread asː-ib=qːel] buy.pfv-pret=when Zajnab-li Zainab-erg nisːe-li-cːella cheese-obl-comit b-erk-un n-eat.pfv-pret ' When s/he (i.e. not Zainab) bought bread, Zainab ate (it) with cheese.' (E)

If we reverse the order of pronoun and noun we also have disjoint reference for the demonstrative pronoun (32). However, with the reflexive pronoun the situation is more complicated because this pronoun can be interpreted as fulfilling a purely emphatic function, which means that the main clause actually lacks an overt subject. This makes it possible, in turn, to arrive at a co-referential reading (33), (34). If we exclude the emphatic interpretation of the reflexive, then in clauses with the *=qːel* converb, disjoint reference is the only possible interpretation, but perfective converbs still seem to allow co-reference.


We can also swap around the order of the clauses. In sentences in which the main clause precedes the adverbial clause, no cataphora whatsoever is allowed (35), (36). This means that neither zeroes nor pronouns can express co-reference with subject arguments in the following subordinate clauses. A pronoun (or a zero anaphora) may not both precede and c-command its antecedent (Langacker 1969: 185; Reinhart 1976: 8). Note that if we use demonstrative pronouns or zero, the person reference in the first clause remains unspecified. By contrast, the reflexive pronoun would be used if we continue to talk about a person who already was the topic of the conversation.


'While Marko was singing a song, s/he (i.e. not Marko) dug the field.' (E)

25.1 The syntax of adverbial clauses

### **25.1.4 Linear order and iconicity**

This criterion concerns the linear order of adverbial clause and main clause ("position" and "layer" in the terminology of Bickel 2010). Although the adverbial clauses most frequently precede the main clause, they may also follow it (37–39), (25), and they may be separated by other subordinate clauses from the main clause, e.g. by other adverbial clauses.

In (37), the imperfective converb clause follows the main clause and shares with the main clause the subject referent and the past time reference. In (38) the converbal clause with *=qːel* also follows the main clause and most probably shares the subject-like argument. In (39) we again have a converbal clause with *=qːel* that follows the main clause and has a causal interpretation.


'Then, in the maternity hospital, here you do not give money when you go to take (the child) out (of the hospital and home)?'

(39) wallah, by.God haʔ-ib=da, say.pfv-pret=1 [a-cːe 2sg-in hel that b-arx-dex n-right-nmlz b-aχ-ij n-know.pfv-inf bahanne] because.of sa-r-ač'-ib-il=da hither-f-come.pfv-pret-ref=1 [ik'-i-l dem.up-obl-erg a-b-urs-ib=qːel] neg-n-tell-pret=when ' I came to know the truth from you, I said, because he (the other doctor) did not tell me (the truth).'

Examples (40–42) show center-embedding, i.e. adverbial clauses that occur within the main clause. That it is in fact center-embedding and not adverbial clauses preceding the main clauses is indicated by the case-marking on the shared argument. The verb in the adverbial clauses differs from the verb in the main clause in transitivity, and the case of the shared argument is assigned by the predicate in the main clause. Note that in all examples the only interpretation available is the shared subject interpretation.


### 25 Syntactic properties of adverbial and conditional clauses

(42) Madina-l Madina-erg [qili home sa-r-eʁ-ib=qːel] hither-f-go.pfv-pret=when t'alaħ-ne dishes-pl d-irc-ib npl-wash.pfv-pret ' Madina, having come home, washed the dishes.' (E)

It has been observed for the perfective converb in other Dargwa varieties and other East Caucasian languages that when the subjects are not identical, the order of main clause and adverbial clause can be changed, but then only the causal interpretation is possible (Belyaev 2010; Kustova 2015; Kazenin & Testelets 2004). In other words, when the adverbial clause precedes the main clause, we can have both a same-subject and a different-subject reading (43). However, the different-subject reading is rather marginal and only available in the right context (see the discussion in §25.1.3 about example (23)).

(43) [Murad Murad ʡaˁrkːa ill ∅-iχ-ub-le] m-be.pfv-pret-cvb lac fence a-b-arq'-ib neg-n-do.pfv-pret ' When/Because Murad got ill he (= Murad or some other person) did not build the fence.' (E)

If we reverse the order, interpretations with shared subjects are more frequently disapproved, e.g. (44) means that an unspecified person is digging the field while Murad is singing. For the perfective converb, a reversal of the order means that a causal interpretation between the two described situations is required (45), whereas in the default order, in which the adverbial clause precedes the main clause, a causal interpretation is possible, but not necessary. Sentences such as (43) can also simply express the temporal order of the events as occurring simultaneously or sequentially without implying a causal relationship.


This means that the order of the clauses in constructions with perfective and imperfective converbs cannot be changed without a concomitant change in the interpretations. This property makes the respective converb constructions slightly similar to clause coordination, which also depicts the order of the events if they do not occur simultaneously: the first clause refers to the first event, the second clause to the second event. By contrast, for other converbs such as the temporal converb *=qːel*, it is possible to reverse the order of the clauses without changing the interpretation, which makes them more similar to subordination (38), (39).

25.1 The syntax of adverbial clauses

### **25.1.5 Morphosyntactic locus**

In addition to the properties discussed, I also tested for morphosyntactic locus (Kazenin & Testelets 2004), i.e. the locus of marking a complement clause as dependent on the main clause. For coordination embedded into a complement clause, the formal marking of embedding is expected to occur on each member of the coordination. By contrast, in case of subordination we can expect the formal marking to occur only on the head or within the head constituent of the complement, but not within another adverbial clause that is part of the complement. This is the case for Sanzhi adverbial clauses that can occur in complement constructions. For instance, in (46) and (47) the masdar suffix that marks the complement clause as dependent occurs only on one verb, whereas the other verb in the complement retains its converbal suffix. In (48) complementation is achieved by means of the cross-categorical suffix -*ce* added to the preterite.


' I am happy when Murad got healthy and finished building the fence.' (E)

(48) du 1sg razi-l=da happy-advz=1 [[Fat'imat Fatimat dalaj song r-ik'-ul] f-say.ipfv-icvb qu garden b-urqː-ib-ce] n-dig.pfv-pret-dd.sg ' I am happy that Fatimat dug the field while singing a song.' (E)

### **25.1.6 Island constraints: relativization and extraction**

The data concerning extraction out of relative clauses varies depending on the converb used and on the interpretations available. The converb *=qːel* blocks extraction, as example (49b) shows. By contrast, the perfective converb allows for extraction (50b). Although the data in (49a–50b) generally fits what has been observed for other East Caucasian languages (e.g. Kazenin & Testelets 2004; Creissels 2012; Bickel 2010), two divergent examples are not enough to understand whether Sanzhi adverbial clauses show the behavior of coordination or of subordination and to what extent this depends on the converbs themselves or on the available interpretations.

(49) a. [šupir-ri driver-erg mašin car b-ik-an=qːel] n-lead.ipfv-ptcp=when rurs-be girl-pl uruχ fear b-iχ-ub hpl-be.pfv-pret 'When the driver was guiding the car, the girls became afraid.' (E)

### 25 Syntactic properties of adverbial and conditional clauses

	- b. [[ħaˁžimurad-li-j Hazhimurad-obl-dat \_ abs b-ičː-ib-le] n-give.pfv-pret-cvb nušːa 1pl razi happy d-iχ-ub-il] 1/2pl-be.pfv-pret-ref ajpun i-phone b-iqː-a! n-take.out.ipfv-imp ' Give me the i-phone that when it was given to Hazhimurad we got happy.' (E)

### **25.1.7 Summary**

Table 25.2 summarizes some of the morphosyntactic properties of perfective and imperfective converb clauses as well as adverbial clauses with *=qːel* that have been discussed in the previous sections. The table shows that the three converbs by and large share most of their properties. If we compare the behavior of Sanzhi adverbial clause constructions with adverbial clauses in other East Caucasian languages, we also find that Sanzhi converb constructions strongly resemble their counterparts in other languages of the family (e.g. Forker 2013c on Tsezic; Creissels 2010; 2012 on Akhvakh; Bickel 2010 on Chechen).

### **25.1.8 Adverbial clauses as independent utterances?**

When examining natural texts it is striking to notice that adverbial clauses headed by perfective and imperfective converbs occur sometimes without a main clause that is obviously connected to it. Example (51) illustrates a perfective converb clause, followed by an imperfective converb clause, and then the speaker concludes his narrative about his military service with a comment that is not directly related to the two preceding adverbial clauses. The utterance in (52), which consists of three adverbial clauses with preterite converbs, describes what the speaker's uncle Abdulkhalik did in order to build himself a house. It is followed by a comment that explicitly states the name of the uncle, but not by a main clause referring to the building of the house, which would be expected based on the general rules of use for the perfective converb.

### 25.1 The syntax of adverbial clauses


Table 25.2: Morphosyntactic properties of adverbial clauses


npl-be-pret-cvb Abdulkhalik-erg

' The brother of my father (= Abdulkhalik) tore down the wall, apparently took the good (materials) in order to build his house (=room), the bad (materials) Abdulkhalik threw away there, (My fathers brother was called Abdulkhalik.)'

Therefore, we might wonder if we perhaps observe an ongoing change in which subordinate verb forms develop into forms that can head independent main clauses. For Mehweb Dargwa, it has been observed in elicitation that some speakers allow perfective and imperfective converbs to head main clauses (Kustova 2019), although the corpus does not contain any examples. In Sanzhi, the situation is reversed: in elicitation, examples such as (51) and (52) are clearly judged as subordinate clauses, but in narrations we find again and again subordinate clauses with a missing main clause. The following excerpt from a discussion between two speakers illustrates the phenomenon. The conversation starts with a question by speaker A (53), which is then answered by speaker B. About half of the utterances by speaker B are formally subordinate clauses.

### 25 Syntactic properties of adverbial and conditional clauses

	- (53) c'il then cellij why w-ebč'-ib-le=de? m-die.pfv-pret-cvb=pst ' Why did he die?'
	- (54) cin-na refl.sg-gen hetːu there qːanaw-t-a-cːe-w ditch-pl-obl-in-m kːiši-l hunger-adv w-ebč'-ib-le, m-die.pfv-pret-cvb … ' in the ditches, he died because of hunger, …'
	- (55) ∅-irχʷ-an=de m-be.ipfv-ptcp=pst cellij why ubk'-an-ne die.m.ipfv-ptcp-fut.3 c'il then 'Something must have happened to him, why should he die (i.e. what other reasons were there to die at that time).'
	- (56) zaˁʡip ill ∅-ič-ib-le, m-occur.pfv-pret-cvb w-ataʁ-ib-le, m-let.pfv-pret-cvb … 'He got ill, and they let him go, …'
	- (57) nuz-b-a-l louse-pl-obl-erg b-ukː-unne, hpl-eat.ipfv-icvb χalq' people nuz-b-a-l louse-pl-obl-erg t'ut'u spread ka-b-ik'-ul, down-hpl-move.ipfv-icvb … ' The lice were eating (the people), lice were all over (the people), …'

	- (58) b-ubč'-i hpl-die.pfv-hab.pst.3 naχadu. without.break Sanži-b Sanzhi-hpl b-ebč'-ib hpl-die.pfv-pret ' They died without stopping. In Sanzhi (people) died.'

Mithun (2008), examines the development of subordinate clauses into main clauses in Navajo, Central Alaskan, Yup'ik, and a few other languages, and notes that the respective sentences contain background information, evaluations or comments that do not advance the storyline. However, this does not seem to be the case in Sanzhi. In both examples (51) and (52), it is rather the other way around. The adverbial clauses drive

25.2 The syntax of conditional clauses

forward the narrative and the main clauses that follow them provide background information or evaluations. And when we compare the main clauses with the subordinate clauses in (53) to (58), there is no obvious division into story line and background information that correlates with the use of converbs and finite verb forms. Only a more detailed study of the Sanzhi corpus can help to clarify whether we really observe an ongoing change, or whether utterances such as the ones discussed in this Section can simply be explained as natural, unprepared spoken text or perhaps performance errors.

### **25.2 The syntax of conditional clauses**

Conditional clauses behave syntactically like adverbial clauses, but also show some differences; for the morphological structure and their functions see §18.3. Firstly, conditional clauses have person agreement. Secondly, conditional clauses express the difference between present/future time and past time reference, and they can also express irrealis mood. Thirdly, an imperative marker in a main clause does not have scope over the conditional clause (59), such that the illocutionary scope is always "local". Conditional clauses may share their subject or other arguments with the main clause, but this is not a requirement. They mostly precede the main clause, but can also follow it (60).


The past conditional occurs recurrently without an apodosis (61). Such sentences can also express wishes (62).


' In which year it was, beloved Allah, if I would remember the years, …'

Interrogative pronouns (63a) and focus-sensitive particles (63b) are allowed to occur in conditional clauses. Extraction out of conditional clauses is blocked (63d):

(63) a. [Zapir-ri Zapir-erg ce what asː-ar] buy.pfv-cond.3 Zajnab Zainab xadi married r-ax-an-ne? f-go-ptcp-fut.3 'If Zapir buys what Zainab will marry him?' (E)


(Intended meaning: 'The car that if Zapir buys it Zainab will marry him is a foreign car.') (E)

## **26 Coordination**

This chapter describes the coordination of phrases (§26.1) and clauses (§26.2.2), including adversative (§26.2.3) and disjunctive coordination (§26.2.4).

### **26.1 Coordination of noun phrases and other phrases**

For the coordination of noun phrases the additive enclitic *=ra* is used (§9.4.1). It is encliticized to the head of every member of the conjunction, which can consist of more than two noun phrases (1).

(1) qːajakent-li-j, Kajakent-obl-dat hej-ka this-abl gu-d-a down-1/2pl-dir ag-ur=da go.pfv-pret=1 ʡaˁbdulχaliq'=ra Abdulkhalik=add Isaq'adi=ra Isakadi=add du=ra 1sg=add nišːa-la 1pl-gen qːulluqː-a-j matter-obl-dat

' To Kajakent, from here up downwards we traveled, Abdukhalik, Isakadi and me, for our matters.'

The enclitic can also coordinate other types of phrases or modifiers within a noun phrase. In (2) two adverbials are conjoined, in (3) two participles and in (4) two extraposed genitives:


' Such a story happened to Tawlu and him.'

The additive enclitic is used in comitative constructions formed with reflexive pronouns (§30.3). Syntactically, they have the structure of coordinated noun phrases (NP*=ra* refl*=ra*), e.g. (5).

26 Coordination

(5) kulpat=ra family=add ca-w=ra refl-m=add 'with his family'

Occasionally, noun phrase conjunctions or conjunctions of other phrases occur without overt marking by simple juxtaposition of the phrases (6–7). With respect to nouns there is only a very limited number of kinship terms – such as the ones in the example (6) – that allow for coordination by juxtaposition. Sanzhi also has other coordinators that are loans, namely *wa* 'and', and the Russian conjunctions *i* 'and', *a* 'and, but', but these loan words are only used in clause coordination (§26.2). That the two nouns in (6) are coordinated is not only clear from the meaning of the clause and their juxtaposition, but also from the gender/number agreement on the verb, human plural (hpl), which is used for conjoint noun phrases with human referents.


It is possible to form disjunctive noun phrases, either by means of the disjunction *ja* (or *ja* bearing the additive enclitic) (8) or by means of the polar question marker *=w* (9) that is used in disjunctive polar questions (see §28.1 for details on disjunction encoded by the polar interrogative enclitic and more examples).


Coordinated noun phrases are semantically and syntactically plural and therefore trigger plural agreement (6). For agreement resolution with coordinated noun phrases, see §20.2.3.

### **26.2 Coordination of clauses**

### **26.2.1 General remarks on the conjunctive coordination of clauses**

Sanzhi Dargwa, like many East Caucasian languages, does not have native words or special syntactic strategies for the coordination of independent main clauses, except for 26.2 Coordination of clauses

simple juxtaposition. Instead, the main way of combining clauses such that they are semantically equivalent to coordinated clauses in European languages is the use of simple converbs, predominantly of the preterite converb (10). Those clauses sometimes contain the additive enclitic *=ra*, which is used for the coordination of phrases (§26.1), on a constituent such as the object or in some other position. Nevertheless, with respect to their morphosyntactic properties, these constructions do not represent coordination in the strict sense, since they contain dependent clauses and they show some other properties of subordination.

(10) [nejg=ra milk=add dabawit add d-arq'-ib-le] npl-do.pfv-pret-cvb [č'irič'an flat.bread b-arq'-ib-le] n-do.pfv-pret-cvb č'irič'an, flat.bread hel-tːi that-pl d-uk-a-di npl-eat.ipfv-hab.pst-1 nušːa-l 1pl-erg

' We added milk, prepared flat breads and ate them.'

Juxtaposition of clauses is illustrated in (11–12). Again in the second sentence in (11) we find the additive *=ra* encliticized to the object, which emphasizes the semantic relationship between the two sentences, but does not function as a syntactic means of clause conjunction. The sentence in (12) shows that simple juxtaposition is also possible.


(12) qili-b home-n asː-ib. buy.pfv-pret b-arq'-ib=da. n-do.pfv-pret=1 ' They bought (the medicine) at home. I did (the medication).'

Coordinated copula clauses are normally only juxtaposed, and the copula item occurs only once in the first clause. This is possible even in those examples in which the two copula subjects do not share person/number values:


Sanzhi has a set of conjunctions ultimately borrowed from Arabic and Persian of which *wa* 'and', *amma(ki)* 'but', and *ja(ra)* 'or' are used for the coordination of main

### 26 Coordination

clauses (for a full list see §9.2). Moreover, it has borrowed the same conjunctions again from Russian: *i* 'and', *a* 'and, but', *no* 'but', and *ili* 'or'. The Russian conjunctions are far more frequently used than the older borrowings. In particular *wa* is almost absent from the corpus (see below for the number of occurrences).

### **26.2.2 Conjunctive coordination of clauses**

The Arabic loan *wa*, although commonly used in written Standard Dargwa for clause conjunction (van den Berg 2004), is not widespread in the other varieties such as Icari, Ashti, or Sanzhi. In the Sanzhi corpus it is only attested in one text that is a translation from Russian. It is a monosyndetic medial conjunction normally occurring between two main clauses (15). However, since its use is so rare and it can have been acquired only via formal education in Standard Dargwa, Sanzhi speakers do not fully adapt to the manner in which *wa* is used in the Standard. Thus (16) shows the conjunction used between a preterite converb clause and a main clause.


' He took off his coat, put it together well, and tied it to his horse's saddle.'

The Russian conjunction *i* is far more frequently used than *wa*, predominantly in translations from Russian (17), but also occasionally in natural discourse (18). In addition, it occurs as clause-initial conjunctional adverb 'and then' (26) that connects longer stretches of discourse (see §9.2 for examples). The total number of occurrences of *i* in the Sanzhi corpus is 45, whereas *wa* appears only three times in one and the same text, which had been translated from Russian into Sanzhi and intended to represent a non-colloquial, written text.

(17) [caj-na one-time durħuˁ-l boy-erg b-arčː-ib n-find.pfv-pret ʡaˁt'a] frog i and [sa-qː-ib hither-carry-pret qili] home ' Once a boy found a frog and brought it home.'

26.2 Coordination of clauses

(18) [di-la 1sg-gen xazajstweni household ʡaˁči work d-irq'-ul=da] npl-do.ipfv-icvb=1 i and [pensija pension ha-jsː-ul=da] up-take-icvb=1 ' I do my house work and receive retirement pay.'

### **26.2.3 Adversative coordination of clauses**

The Arabic conjunction *amma(ki)* 'but' is very rarely employed as a genuine conjunction for independent clauses (19–20). In the majority of instances, it appears as an adversative adverbial in clause-initial (20) or clause-final position (19) (for both cases see §9.2). Though in this function it also has the adversative semantics, it rather connects larger episodes of texts that are contrasted to each other.


' There is this medical cow-parsnip, but if these (plants) get on (the skin), it is bad.'

The same functional range is found with the Russian conjunction *a* 'and, but'. It predominantly occurs as sentence-initial marker of topic switch (§9.2), but there are also examples of adversative coordination of clauses (21). The second adversative coordination from Russian, *no* 'but', differs in its semantics from *a* because its meaning is more specific. In the Sanzhi corpus there is only one instance of *no* occurring between an independent clause and an ellipsis (22).

(21) [hextːu-b there.up-hpl šːatːir visit sa-b-ač'-ib-te hither-hpl-come.pfv-pret-dd.pl li<b>il=ra all<hpl>=add χalq' people b-ikː-ul hpl-want.ipfv-icvb ca-b cop-hpl gu-r-b-uq'-aˁnaj] sub-abl-hpl-go-subj.3 a but [gu-r-b-uq'-ij down-abl-hpl-go-inf mašin-te car-pl d-akːu] npl-cop.neg

' There all the people who came want to leave, but there are no cars to leave.'

(22) [latiši Latvian wredni harmful χalq'=de, people=pst wahi evil χalq'=de] people=pst no but [parjadok, order amzu-dex clean-nmlz χʷal-le heχ-tː-a-lla]

big-advz dem.down-pl-obl-gen

' The Latvian people were harmful, they were bad people, but order, cleanliness (was) great among them.'

### 26 Coordination

### **26.2.4 Disjunctive coordination of clauses**

The particle *ja(ra)* (*jara* is composed of *ja* and the additive *=ra*) is ultimately a loan from Persian. It used as a bisyndetic particle in clause initial position. In affirmative clauses it means 'either … or' and in negative clauses it means 'neither … nor'. Although the use in affirmative clauses can be obtained in elicitation, all corpus examples show negative clauses. Sentence (24) illustrates that more than two clauses can be coordinated.


For the disjunctive coordination of affirmative clauses the Russian disjunction *ili* is used, which occurs between the members of the disjunction or in clause-initial position (27).


' Or it must probably be this, or when he is set free, or when is taken (into prison).'

## **27 Constituent order and information structure**

This chapter addresses constituent order at the phrase level, in particular within the noun phrase (§27.1); constituent order at the clause level in main clauses and subordinate clauses; as well as the information-structural patterns that are associated with certain orders (§27.2). It also provides a short overview of other ways of manipulating the information structure, most notable term focus constructions and cleft-like constructions and focus-sensitive particles (§27.3–§27.5).

### **27.1 Constituent order at the phrase level**

### **27.1.1 General remarks on the constituent order at the phrase level**

Noun phrases are strictly head-final. A template of the structure of noun phrases and the internal order of modifiers is given in §21.1.3. Noun phrases can be quite complex, but mostly they have between zero and two modifiers. Example (1) illustrates a noun phrase consisting of a demonstrative, a genitive pronoun, a short relative clause, and the head noun. All modifiers occur in the order which is most common in the Sanzhi corpus (i.e. demonstrative before genitive before relative clause).

(1) het that di-la 1sg-gen [r-alχ-an] f-know.ipfv-ptcp xːunul-li-šːu woman-obl-ad r-ax-ul=da f-go-icvb=1 ' I go to my woman who I know.'

Within postpositional phrases and adjective phrases, the word order is also head-final. Thus, adverbial modifiers always precede and never follow adjectives (2), and postpositions always follow their complements (3). However, since some postpositions also occur as adverbs, one might come across examples that seem to contradict this claim because they contain postpositions used adverbially without a complement, or with what seems to be a postpositional complement, but does not occur in the expected position (see §21.2 for example (54)).


### 27 Constituent order and information structure

All modifiers except for demonstrative pronouns and numerals occasionally occur in positions detached from the noun phrase and in such cases they syntactically do not belong to the noun phrase anymore, but form a separate constituent (§21.1.3). I will use the term "floating" for these items. In the Sanzhi corpus, floating modifiers sometimes immediately follow the head noun ((7) below), but mostly they occur after the verb (4). There are only few examples of floating modifiers in positions to the left of the noun phrase (see §21.1.4 for one example with a genitive and §27.1.3 below for two examples with adjectives). In general, the position after the verb can have special properties with regard to its contribution to the information structure of the sentence (§27.2), and it seems that when modifiers are extraposed to this position they often receive a contrastive interpretation. This is not unique for floating modifiers, but also regards other constituents that occur after the verb.

Floating modifiers in the East Caucasian languages have been analyzed by a number of researchers. Kazenin (2002) investigates their syntactic structure in Lak and the question of whether they form one constituent with the noun phrase or not (see §21.1.4). Other authors such as Testelets (1998a), Creissels (2013), Komen (2014), and Lander (2014; 2016) treat their impact on the information structure, in particular contrastivity. In the following two sections, I will present the accounts that can be found in the literature and discuss their applicability to Sanzhi.

### **27.1.2 Floating genitives**

The most common modifiers occurring in an extraposed position are genitive modifiers (4–7). In the majority of cases, these postponed genitives are personal pronouns or demonstratives used as personal pronouns. This means that their referents are human and highly topical. As was just mentioned, the floating genitive most frequently follows the verb (4–6). However, genitives occupying the position directly after the noun are also attested. In (7) this can be explained by the fact that the genitive phrase is part of a subordinate clause for which verb-final constituent order is strongly preferred (§25.1).


27.1 Constituent order at the phrase level

There are various explanations for why genitive modifiers can follow the head noun, but contrastiveness does not seem to be the best one. Creissels (2013) analyzes floating genitives in the East Caucasian language Akhvakh. Like in Sanzhi, the floating genitives predominantly follow the verb, denote human referents, and occur in one of the three functions that are also attested for prenominal genitives:


As in Akhvakh, possessive predications in Sanzhi, i.e. copula clauses with nominals in the genitive functioning as predicates, can be treated as a subtype of floating genitives when they have genitive nominals that follow the copula (8).

(8) hetːi those weral seven ucːiq'ar cousin le-b=de exist-hpl=pst di-la 1sg-gen aba-la mother-gen ' My mother had seven cousins.'

In contrast to Akhvakh, however, the floating genitive in Sanzhi can also have a head noun in a case other than the absolutive, though only one such example has been found so far in the Sanzhi corpus. Thus, example (9) contains the postponed genitive *nišːala* 'our', which functions as possessor of the noun *učitilla* 'teacher.gen'. Furthermore, the sentence contains a relative clause that follows its head noun *zamana* 'time'.

(9) hel that zamana time [duc' run uq-un-il go.m.pfv-pret-ref du 1sg [heχ-i-la dem.down-obl-gen t'uˁ-me leg-pl gu-r-sa-d-ertː-ij]], sub-abl-hither-npl-tear.pfv-inf hek' dem.up učitil-la teacher-gen w-arx-le m-direct-advz nišːa-la 1pl-gen […] dukla onto.lap w-i-w-aˁq-ib=da m-in-m-strike.pfv=pret=1

' At the time when I (masc.) ran to pull down his legs, I fell (lit. hit) directly on the lap of our teacher.'

Creissels (2013: 346) describes the semantic properties of floating genitives for Akhvakh by noting that postposed genitives have an "empathy effect" and "consider the situation from the point of view of the possessor." He writes that the floating genitive construction "has a possessive framing function, in the sense that the floating genitive identifies the personal sphere of its referent as the frame within which the predication expressed by the clause holds" (Creissels 2013: 333). He further compares them to other framing adjuncts such as spatial and temporal expressions and external possessors. The account given by Creissels fits the Sanzhi data well. Like the genitives in Akhvakh, floating genitives in Sanzhi denote affected participants similar to beneficiaries or maleficiaries that are not functioning as arguments, but whose referents are either strongly

### 27 Constituent order and information structure

involved in the situation expressed by the verb and/or are in physical proximity to that situation (see also Shibatani 1994 and Seržant 2016 on external possessor constructions). Thus, the floating genitives can often be interpreted as referring to a location (4), and it is relatively common for adjuncts denoting locations to follow the verb in a locational copula construction (§22.2.2). The following minimal pair illustrates the difference between preposed and postposed genitive modifiers (10).

	- b. heštːi these xujal five macːa-l sheep-erg di-la 1sg-gen li<d>il all<npl> mura hay d-erkː-un npl-eat.pfv-pret ' The five sheep of mine ate all the hay.' (E)

In general, preposed genitives are definite (11).

(11) di-la 1sg-gen χatːaj grandfather ca-w, cop-m ala 2sg.gen itwaj like.that χatːaj grandfather ca-w cop-m ' (He) is my (real) grandfather. For you he is only an old man.' (lit. He is like a grandfather of yours.)

In the following example (12) with the verb *b-ic'-* (pfv) 'fill', the noun denoting with what the house is filled has human reference and bears the genitive case. Although the genitive in (12) cannot be analyzed as a modifier of the preceding noun *qal* 'house', the construction expresses the affectedness and the involvement of the referent of the noun *χalq'* 'people' similar to the other extraposed genitives discussed so far. By contrast, if the filler is inanimate, the ergative case has to be used instead of the genitive. The ergative also denotes instruments, and instruments are, in general, not affected by an action or a situation (§3.4.1.2).

(12) ʡaˁħ-le good-advz qal house b-ic'-ib n-fill.pfv-pret χalq'-la people-gen 'the house filled well with people'

Similarly, sentence (13) shows a personal pronoun in clause-final position following the verb. This pronoun can be considered an floating modifier of the quantified noun phrase *k'ʷel admi* 'two persons' in clause-initial position. Again it has a human referent that is depicted as being affected by the situation.

(13) k'ʷel two admi person itwaj like.this ka-d-at-ur-te down-1/2pl-let.pfv-pret-dd.pl akːʷa-di cop.neg-1 nušːa 1pl ' (She) did not put the two of us like this (i.e. in vain).'

27.1 Constituent order at the phrase level

In a few cases the use of floating genitives may have other pragmatic reasons. One factor is probably to avoid interpretative ambiguities. If the noun phrase contains other nominals as modifiers, then the modifiers of the head noun that precede the genitive could be interpreted as modifiers of the genitive (see §21.1.3 for more examples) instead of belonging to the head noun. In some examples the postponed genitive might be a kind of afterthought, providing more information about the referent. This could be the case in the following example (14).

(14) hel that qːatːi hat hel-i-la, that-obl-gen hel that durħuˁ-la boy-gen 'the hat of him, of the boy'

Finally, contrast might occasionally be a reason when the extraposed noun is inanimate and cannot be analyzed as an affected participant of the situation. In the dialogue from which (15) was taken, the speaker is talking about two types of medical treatments that were proposed to her, and she perhaps uses a postponed genitive in order to express contrast.

(15) "hana now awgust-le-r August-loc-f sa-r-eʁ-e," hither-f-go.pfv-imp r-ik'ʷ-ar, f-say.ipfv-prs "lečenie treatement b-elk'-ij n-write.pfv-inf ukul-t-a-lla!" injection-pl-obl-gen 'She said "Come now in August, to prescribe the treatment with injections!"'

### **27.1.3 Floating adjectives, postpositional phrases, and relative clauses**

Floating adjectives, postpositional phrases, and relative clauses are less frequent than floating genitives, but they are also attested in the corpus. They obligatorily bear the cross-categorical suffix *-ce* (adjectives, postpositional phrases, and occasionally relative clauses, §9.6.1) or the cross-categorical suffix *-il* (relative clauses, §9.6.2), which both are used for nominalizations. As with genitives, the floating modifiers often do not follow the noun immediately, but occur after the verb. In most examples they occur in copula constructions, but other clause types can also be found.

For other East Caucasian languages, particularly Avar-Andic and Tsezic languages, Testelets (1998a) claims that "modifiers […] may be postponed only under contrast." This statement and the examples, which follow it, suggest that it is a bi-conditional connection, to the extent that every modifier that is postponed is necessarily contrastive. Similar claims have also been made for Tanti Dargwa (Lander 2014; 2016, Sumbatova & Lander 2014) but with the caveat that in Tanti Dargwa modifiers bearing a cognate of the Sanzhi suffix *-ce* are said to be always restrictive, but not necessarily contrastive, and that contrastive modifiers are less likely to follow the head noun.

The Sanzhi sentences containing floating modifiers seem to support the hypothesis that those modifiers normally receive a restrictive interpretation, which can be contrastive, but does not have to be. Thus, in (16), the speaker adds more information about

### 27 Constituent order and information structure

the river without contrasting it to other rivers. Similarly, the stories mentioned in (17) are not contrasted to other stories that were not interesting. The man in (18) is also not contrasted to another man that was not on the tree, but the speaker simply mentions a feature of the man that unambiguously identifies him. The fact that the man is standing in the crown of a tree is encoded by means of a postpositional phrase that modifies the noun *admi* 'man' and is marked with -*ce*.


'About them there were, are many stories, interesting.'

(18) či-w-až-ib-le spr-m-see.pfv-pret-cvb admi person kːalkːi-cːe-w tree-in-m či-w-ce on-m-dd.sg 'and then (he) saw the man on the tree'

Sometimes it is unclear if the floating modifier has only a restrictive meaning or if the existence of another, contrasting referent is also implied. For instance, the shampoo in (19) is identified by means of its description as being expensive, but from the context it remains unclear whether it is also contrasted with other shampoos that are not expensive.

(19) detski children's šampun shampoo le-b exist-n durqa-ce expensive-dd.sg 'There is shampoo for children, expensive.'

Unambiguous examples of contrastive postposed modifiers can be found. In (20) the speaker compares the watermelon that he bought with the other watermelons that were sold but were smaller, and (21) contrasts the thin khinkal to other types of khinkal. In (21) the contrastive interpretation results from the postnominal position, not from the use of the suffix *-ce*.


27.1 Constituent order at the phrase level

It can be argued that the examples in (16–18) represent instances of right dislocation in which the dislocated element is a nominalized expression conveying an afterthought that provides more information about the referent of the head noun. A similar case can be made for examples in which the nominalized adjectives precede the nouns to which they refer. In those sentences it is the nominal that is right-dislocated (22–23) (see also (35) in §21.1.4).


Clear examples of floating relative clauses are even less frequently found. They mostly have the form of identificational copula clauses and thus resemble cleft constructions. Usually relative clauses contain a gap in the position of the head of the relative clause (Chapter 23), but the floating relative clauses can also be free relative clauses with a pronoun that is co-referential to a nominal outside of the relative clause. The relative clauses are restrictive, although two of them have personal names functioning as heads. In examples (25) and (26) the speaker helps the addressee to identify the referents by giving more information about them. Since Patimat is the most common female name in Dagestan and Rasul is also a common name, the hearer cannot be expected to immediately know about whom the speaker is talking.

(24) iž=ra this=add het=ra, that=add het that ʡaˁχːuˁl guest ∅-iχʷ-ij m-be.pfv-inf [xːunul-la woman-gen qajqaj-li-cːe jaw-obl-in b-aˁq-ib-il] n-hit.pfv-pret-ref

' This also and this also is probably the man who hit the woman on the jaw.'


'One was the son of my Cibac, his name is Rasul, the one that lived in Sanzhi, do you know him?'

In (27) the head is an indefinite pronoun from Russian that is followed by two relative clauses, restricting the reference of the pronoun. Note that the indefinite pronoun functions as recipient in the clause. It is a borrowing from Russian and unmarked for case

### 27 Constituent order and information structure

(both in Sanzhi and in Russian), although a Sanzhi nominal in this position would have required dative case. This is remarkable because it is one of the few corpus examples of floating modifiers with a head noun in a position, which normally requires case marking (see also (9) for another examples with a floating genitive). Almost all examples of floating modifiers discussed in this and other sections belong to nouns that are in the absolutive case (e.g. S and P arguments).

(27) ʡaˁħ-le good-advz χːula big ħuˁrmat respect b-irq'-u n-do.ipfv-prs lubuj any [hej-ka this-down sa-jʁ-ib-il] hither-come.pfv-pret-ref [cin-na refl.sg-gen w-alχ-an] m-know.ipfv-ptcp ' He gives respect to anybody who comes here, whom he knows.'

Komen (2014) analyzes floating relative clauses in the East Caucasian language Chechen. Like in Sanzhi, floating (also called "extraposed" by him) relative clauses are rare. Furthermore, if they are restrictive then their head always occurs in the preverbal position, which is the focus position in Chechen. In Sanzhi, almost all examples are copula clauses with the head of the relative clause immediately preceding the copula, a position which is normally used for focal items (25). However, in (27) the head follows the verb. Due to the lack of more examples it is impossible to clarify at the moment whether the condition that Komen (2014) established for Chechen also holds for Sanzhi.

### **27.2 Constituent order at the clause level and information structure**

In this section, the constituent order of clauses, including that of interrogative clauses, will be analyzed. I will also discuss how this order reflects the information structure status of the constituents.

I adopt the definition of focus given by Dik et al. (1981) as "what is relatively the most important or salient information in the given setting." Focus is opposed to topic, which is defined as "the entity 'about' which the predication predicates something in the given setting." Sentences can be fully-focused when the entire sentence conveys the most important information. However, normally only a sub-part of the utterance carries focus, which means that we have constituent focus (or "term focus" in the terminology of Dik et al. 1981). Constituent focus can be divided into (i) completive focus and (ii) contrastive focus. Completive focus fills a gap in the pragmatic information of the addressee, as is the case with answers to content questions (§28.2). Contrastive focus constitutes a reply to the addressee's contrary belief, for instance by correcting and replacing it. Dik et al. distinguish three subtypes of contrastive focus:


### 27.2 Constituent order at the clause level and information structure

Predicate-centered focus, in which the verb is in focus, represents a type of constituent focus that can also be either completive (in answers to questions about the kind of action that was carried out) or contrastive (e.g. when the truth value is at issue).

Topical information and topical referents are usually familiar in the given context, and as such they can be used as anchors for providing new information. Therefore, they are usually unmarked and often omitted from the utterance in Sanzhi Dargwa. However, topics can also be new or contrastive when they differ from the topic(s) in the previous utterances.

The third notion relevant for information structure is contrast. It is independent of the notions "topic" and "focus" (e.g. Vallduví & Vilkuna 1998; Neeleman et al. 2009). Contrast is relational because there must be a relation between the contrasted item and at least one other identifiable alternative in the context, and both the contrasted item and the alternative must be explicitly verbally mentioned (Malchukov 2004; Izutsu 2008; Repp 2010). The two items must be comparable to each other with respect to a shared domain. At the same time they are different and can therefore be contrasted with each other.

### **27.2.1 Declarative clauses**

The constituent order in main clauses is basically free, at least in the sense that every logically possible order can be found in texts and it can easily be elicited. The most frequent order is SOV, but SVO is also very common. Verb-initial orders (VSO and VOS) are rare, but not ungrammatical. For intransitive clauses we find SV and occasionally VS. In general, the constituent order is heavily influenced by pragmatics and the information structure of the utterance. Contextually retrievable arguments are frequently omitted. This includes subject-like arguments as well as object-like arguments. Therefore, many monovalent predicates in the corpus occur without a subject and most bivalent predicates have only one overt argument. For clauses with one (monovalent verbs) and two (bivalent verbs) overt arguments, the tendencies for linking certain constituent orders with specific pragmatic values of the constituents listed in Table 27.1 can be observed. With the label "neutral", I refer to predicate focus, which is commonly assumed to represent a universally unmarked type of focus in which the subject is the topic and the predicate is focused.

The pragmatically neutral order, in which none of the constituents is particularly emphasized, is SV in intransitive clauses and SOV/SVO in transitive clauses, including in elicited sentences without any context. In such clauses, the subject usually represents given information, while the predicate alone or the predicate together with the object carries the new information.

(28) SV: neutral information structure

[talking about a woman whose husband is taken away by the police] hel that r-isː-ul f-cry-icvb ca-r cop-f 'She is crying.'


Table 27.1: The relationship between constituent order and information structure in main clauses


27.2 Constituent order at the clause level and information structure

(32) SVO: neutral information structure [finishing the description of how to prepare *khinkal*] dam 1sg.dat b-ičː-aq-id n-want.ipfv-caus-1.prs burt sour.cream sːerži garlic ʡaˁħ-le good-advz b-erʁ-ub n-dry.pfv-pret dig=ra meat=add ' I like garlic, sour cream, and also well dried-meat.'

For a number of SOV languages such as Urdu, Turkish, Armenian, Georgian (e.g. Butt & King 1996, Comrie 1984, Testelets 1998b) and also Chechen (Komen 2007), a very strong association between focus and the preverbal position has been observed. In many if not all East Caucasian languages, there is also a clear tendency for putting focused items immediately before the verb (Testelets 1998a,c; Forker & Belyaev 2016). In general, Sanzhi behaves alike, but focused constituents are not always and exclusively placed directly in front of the verb. The two neutral orders SOV and SVO can be used to focus the subject (33) or the object (31).

(33) SVO: subject focus

[(talking about doctors and treatments) I go to the woman who I know.] Eštːa Ashti.loc Rasul-li Rasul-erg b-irq'-u n-do.ipfv-prs lečenie cure ca one ca=qːel one=when ' In Ashti, Rasul makes treatments from time to time.'

However, subject or object focus is mostly expressed by other constituent orders that cannot be considered pragmatically neutral. The order OVS has been noticed to be used when the object or the object together with the predicate is in focus (Testelets 1998a,c; Forker & Belyaev 2016), and this is confirmed by the following Sanzhi examples (34–36). In (35) the speaker uses two clauses to describe basically the same event, the arrest of her husband by the police. The first clause has the constituent order OV and no overt subject, and the second clause has VS without an overt object.

(34) OVS: object focus ce what ∅-ik'-ul=de m-say.ipfv-icvb=2sg u? 2sg 'What do you (masc.) say?' (35) OVS: object + predicate focus [talking about a woman holding her little son in her hands while her husband is arrested by the police] sub husband uk-ul gather.m.ipfv-icvb ca-w, cop-m kʷi-r-s-asː-ib-le in.hands-abl-hither-take.pfv-pret-cvb ix-tː-a-l dem.up-pl-obl-erg ...

'(They) are taking her husband, they took him away from her hands, …'

27 Constituent order and information structure

(36) OVS: object + predicate focus [ Then at this time in order to show his masculinity, while she had the child in her arms, ...] xːunul-li-j woman-obl-dat b-aˁq-ib n-hit.pfv-pret ca-b cop-n hel-i-l that-obl-erg 'He hit his wife.'

OVS order can also be used when the subject is a contrastive topic, which, in principle, does not need to exclude the possibility of focusing the object. In (37) the object is an aboutness topic, in addition to the subject being a contrastive topic.

(37) OVS: contrastive topic (I vs. they)

[A: They had a five-liter canister of wine. B: Are you saying that you had the canister?] xujal five litru-la liter-gen kanister canister čaˁʁir-la wine-gen b-alli n-together b-erqː-ib=da n-carry.pfv-pret=1 du-l 1sg-erg čːaˁʡaˁl-la morning-gen

' C: I had taken the five-liter canister of wine in the morning.'

OSV constituent order is not particularly common in texts, but its pragmatic value seems to be relatively clear. It is mainly used for topicalizing objects, in particular for contrastive topics (38), (39). It is also used to focus the subject together with the predicate (40), (41).

(38) OSV: topical object + predicate focus [One student whom I took with me in the car said at the end of the journey] u 2sg du-l 1sg-erg nikagda never qum.a.art-an=de forget.ipfv.neg-ptcp=2sg 'You, I will never forget!'


27.2 Constituent order at the clause level and information structure

(41) OSV: object topicalization, subject + predicate focus na now it that du-l 1sg-erg r-alχ-an=q'al f-know.ipfv-ptcp=mod 'Well, I know her [that Salikhat who married into my family.]'

Although in principle two different verb-initial orders can be elicited, in texts only VSO is attested. As stated in Forker & Belyaev (2016), VSO consistently expresses verb focus in combination with topical subjects and topical objects (42–45). The first item in (44), *t'am*, forms a compound verb together with the following verbal lexeme and thus does not function as argument. Example (45) shows an antipassive construction in which the subject appears as pronoun in the absolutive and the object as ergative-marked noun


whistle=add

'and the dogs didn't hear his sound or whistling.'

(45) VSO: verb focus, topical subject [Then they went to drink with the money they made from the theft.] b-učː-ul hpl-drink.ipfv-icvb ka-b-iž-ib down-hpl-be.pfv-pret ca-b cop-hpl hetːi those čaˁʁir-li wine-erg 'They sit and drink wine.' (lit. 'Drinking they sit down with wine.')

The answer to the following question illustrates verb focus with an intransitive predicate and a topical subject (46):

	- a. ce what b-irq'-ul=e? n-do.ipfv-icvb=q ' Q: What are (they) doing?'

### 27 Constituent order and information structure

b. šːatːir walk tːura outside b-uq-un hpl-go.pfv-pret ca-b cop-hpl hex-tːi, dem.up-pl či-b-b-axʷ-araj spr-hpl-hpl-relax.pfv-subj.3 ' A: They went for a walk, to relax.'

To sum up what has been observed so far, we can state that focal arguments and adjuncts with various semantic functions and grammatical roles most commonly precede the verb, occurring immediately before it. The only regular exceptions are presentational sentences with newly introduced arguments, which follow the verb (§27.3.1). Topical constituents appear to the left of focal constituents, but contrastive topics also normally occur at the right edge of the clause. This is in line with the studies by Testelets (1998a,b) and Forker & Belyaev (2016).

In Sanzhi, there is no clear tendency for the relative placement of direct object (theme = T) vs. indirect object (recipient/goal = G). Both G-T-V and T-G-V are found. Which order is chosen depends on the pragmatic value of the arguments within the information structure of the utterance, and there does not seem to be a pragmatically neutral order. Thus, the G arguments in (48), (49) seem to be contrastive, and together with the verb form part of the new information. Moreover, like other arguments, T and G arguments can also occur after the verb.

(47) G-T-V

hi-cːe-k'al who.obl-in-indef cik'al something lukː-an-te=de=w, give.ipfv-ptcp-dd.pl=pst=q aχːu, not.know hel-i-l=ra that-obl-erg=add

' (He) also had to give something to somebody, I don't know.'

(48) T-G-V

[Did it die, he asked. No, I said. From there he ran away.] sumk'a bag di-cːe 1sg-in b-ičː-ib n-give.pfv-pret '(He) gave me his bag.'

(49) T-G-V

 [May your beloved stay alive, dear brother!] hana now hel that cik'al something dam 1sg.dat či-ma-sa-b-iršː-itːa! spr-proh-hither-n-put.ipfv-proh.sg 'Do not put that thing on me!' (i.e. that piece of work)

Goal-like arguments (recipients, addressees), just like subjects and objects, most commonly occur before the verb (G-T-V, T-G-V) (47–49). However, they also seem to have a relatively high probability to follow the verb, which does not depend on their status within the information structure of the utterance (50), (51). The tendency includes goallike adjuncts (directional adverbials, and possibly also beneficiaries, see below). It has

### 27.2 Constituent order at the clause level and information structure

been observed for verb-final languages of other language families, most notably Western Iranian languages, but also Iraqi Turkmen and Azerbaijani spoken in the wider area (Anatolia, south Caucasus; see Haig 2015). Since the postverbal placement does not extend to objects, it cannot be explained by Russian influence, but might be due to contact with Kumyk (Turkic) speakers. However, before we can attribute the use of postverbal goals in Sanzhi to the impact of Kumyk it needs to be clarified if Kumyk belongs to the Turkic languages with postverbal goals and if other East Caucasian languages that are not in contact with Kumyk do not have postverbal goals to the same extent as Sanzhi. An alternative explanation might resort to iconicity. The goal is the spatial endpoint of the situation. Thus, a postverbal goal is iconic in that the destination of the transfer follows the entity to be transferred (the T argument) and it also follows the action expressed by the predicate.


Adjuncts like comitative noun phrases, instruments, and manner adverbials are most frequently positioned after the subject, if there is one, and before the verb, but sometimes they can be found after the verb (37), (52), (54).

Temporal and locational adjuncts normally occur at the beginning of clauses and precede any arguments and other adjuncts; especially the short temporal adverbs *ha* and *hana* 'now' (15), (49), (52), and locational adverbs such as *heštːu* 'here' and *hetːu* 'there' have a strong tendency to occur clause-initially. However, they can be placed postverbally when representing new information or when they are contrastive. By contrast, directional adjuncts are goal-like and behave similarly to goal-like arguments (54). Though a position left of the verb at the beginning of the clause is common for directional adjuncts, postverbal placement is roughly equally common, not only when they encode new information. This behavior is part of a general tendency for all goal-like arguments and adjuncts, and can be explained by means of iconicity.

Examples of temporal, locational, and directional adverbs are (33), (52), and (53). If adjuncts are focused, they immediately precede the verb (54), (55).

(52) hel that zamana time il-i-l that-obl-erg hati=ra more=add qːuʁa-l beautiful-advz d-al npl-together d-ič-aq-ib-le, npl-occur.pfv-caus-pret-cvb …

' At that time he embellished (his story) even more, …'

### 27 Constituent order and information structure


As discussed in §27.1.2, Sanzhi has floating genitives for which the referents of the genitives are highly topical and affected. In the great majority of cases, those genitives take over the clause-final position and can possibly be regarded as framing adjuncts.

Interjections and addressee particles occur at the edge of clauses (53), either preceding all other items in the clause or following them (see §9.5 for examples).

Clauses fulfilling argument positions in complement constructions can precede or follow the verb. For more information see §24.4. In a complex sentence consisting of a main clause and at least one subordinate adverbial clause, the neutral order is for the adverbial clause to precede the main clause or to be center-embedded within the main clause, though the order in which the main clause precedes the adverbial clause is also attested (§25.1).

Subordinate clauses have a strong tendency to be verb-final (52). This is true especially for complement clauses (§24.4) and relative clauses (§23.3). It is easier to find adverbial clauses, in particular those headed by the general converbs, which show other than verbfinal orders. Due to the rather fixed word order and the readiness with which arguments are dropped, which for subordinate clauses is higher than for main clauses, the manipulation of the information structure in subordinate clauses by means of the constituent order is not readily available. However, it is possible to switch the order of subject and object, and occasionally postverbal arguments can be found. In such cases, the same connections between focal or topical elements and certain positions in the clause can be observed that were summarized for main clauses in Table 27.1. For instance, (56) shows an adverbial clause with the copula complement preceding the verb because it represents the focus and the topical subject following the verb. Example (57) illustrates VSO order in an adverbial clause and resembles the finite VSO clauses in (43) because the verb carries the new information and the subject is topical.

(56) nacijonalist patriot gu-r-sa-jč-ib-le sub-abl-hither-occur.m.pfv-pret-cvb hel, that … ' That (journalist) was a patriot, …'

27.2 Constituent order at the clause level and information structure

(57) [the protagonists ran out of alcohol and sent a boy to the car that was filled with bottles] mašina-l-cːe-r car-obl-in-abl k-aqː-ib-le down-carry-pret-cvb durħuˁ-l boy-erg cara=ra other=add čaˁʁir, wine … ' The boy brought another (bottle of) wine from the car, …'

### **27.2.2 Information structure in interrogative clauses and question-answer pairs**

Polar questions are obligatorily marked by the enclitic *=w*. The enclitic is added to the verb or another constituent, e.g. noun, pronoun, adjective. Polar questions seem to have a greater tendency for verb-final constituent order than declarative utterances have (58), (61), but nevertheless postverbal topical constituents can be found (59a). Answers to polar questions consist mostly only of the verb, which can be the affirmative copula *ca-b*, the negative copula *akːu*, or a full verb form (58–60). Polar questions are analyzed in detail in §28.1.

	- b. cara other musːa-b, place-n absalut'na completely cara other musːa-b place-n b-arq'-ib-il=de n-do.pfv-pret-ref=pst ' A: In another place, they built it in a completely other place.'
	- b. e, yes ag-ur=da go.pfv=pret=1 'A: Yes, we went.'

Content questions contain interrogative words and are also marked by a special enclitic *=e*/*=ja* that is normally attached to the predicate (§28.2). If there is no predicate, then the interrogative pronoun functions as head. The interrogative pronoun mostly appears immediately before the verb, that is, in the same position in which most focus items occur in declarative utterances. Sentence topics regularly precede the interrogative pronoun (61), (62a) or, more rarely follow it (63a). It is also possible to have one clause-initial topic and another clause-final topic.

### 27 Constituent order and information structure

	- b. iχ dem.down selsawet-le village.head-loc kelg-un remain.pfv-pret birgadir-le brigadier-loc kelg-un remain.pfv-pret c'ili then selsawet-le-r village.head-loc-abl ag-ur go.pfv-pret ' A: He was the village head, brigadier, then he stopped being village head (lit. left).'

' Q: Then from where did you go?'

b. aq high d-arq'-ib-le npl-do.pfv-pret-cvb heštːi, these či-r-d-ax-ud spr-abl-1/2pl-go-1.prs erk'ʷ-le-r river-loc-abl

'A: We put up these (trousers) and go through the river.'

Fronting of interrogative words is also possible, but very rare in natural texts. The only pronoun that is repeatedly fronted is *cellij*/*cel* 'why' (64) (see §4.5.2.4 for more examples).<sup>1</sup> The pronoun *ceqːel* 'when' also occasionally occurs in clause-initial position (65), but usually it is preceded by spatial adverbs. Thus, sentence (65) was repeated by the speaker and in the second occurrence the order of the interrogative pronoun and the directional adverbial were swapped.


<sup>1</sup>As one reviewer pointed out, the fronting of a pronoun with the meaning 'why' is common. For instance, in Hungarian it is the only interrogative word that can be found in a position other than the focus position immediately to the left of the verb. In the Austronesian language Pohnpeian 'why' must be initial; other interrogative phrases do not have to be, see Dryer (2005) for examples and references.

27.2 Constituent order at the clause level and information structure

For other pronouns corpus examples are non-existent, but available in elicitation. Sentence (66) can be uttered in a situation in which we know that the shop is empty and we wonder what Batichaj can bring if there is nothing to buy. The utterance in (67) represents the pragmatically neutral constituent order for this type of question with a postverbal topical subject, see also (34).


Topic-comment sentences are answers to questions such as *What is X doing?* Example (68) shows this type of question and the respective answer that has a neutral pragmatic structure, that is, the subject is topical and has been omitted in the answer. The predicate together with the object and the goal represents the focal information and occurs in front of the verb (object) and directly after the verb (goal), the typical positions for focal objects and goals. The constituent order in the question is SOV, with the interrogative pronoun occurring *insitu*.

(68) Isaq'adi-l Isakadi-erg ce what b-irq'-ul=e? n-do.ipfv-icvb=q qus slip ha-b-aˁq-ib-le, up-n-drag.pfv-pret-cvb hel that qːalpuz watermelon ka-b-išː-ib down-n-put.pfv-pret het that xːun-r-a-j woman-pl-obl-dat naprotiw in.front 'What is Isakadi doing? He dragged the watermelon and put it in front of the women.'

In sum, the information structure of the vast majority of content questions is *(topic) question word-verb-(topic)* with the interrogative enclitic attached to the verb. Answers to content questions can mirror this structure by placing the item that answers the question in the preverbal position as well, with optional topical elements placed at the edges of the clause (62b). Alternatively, they can also contain the item in focus in another position, as (63b) shows, in which the relevant noun follows the verb. Short answers consisting only of the focus are also common.

It is possible to use the constituent focus construction in interrogative clauses. In this construction the interrogative enclitic is attached to the item in focus and the verb must take the form of a participle. See §27.3.2 below for more details and examples.

### **27.2.3 Right and left dislocation**

Because of its free constituent order, it is not always easy to identify dislocation in Sanzhi. Topicalization by means of placing arguments or adjuncts at the edges of clauses cannot be equated with dislocation. We can be sure we are dealing with dislocation when we find a pronoun in the clause that is co-referential with the dislocated noun phrase and

### 27 Constituent order and information structure

when the dislocated noun phrase does not correspond to any arguments or adjunct of the clause and is therefore unlinked.

Dislocation is not particularly frequent, but when it occurs it has the same structure and the same functions that have been attested for dislocation in other East Caucasian languages (Forker & Belyaev 2016). Thus, left dislocation is a topicalization strategy. The dislocated noun phrase occurs in the absolutive case. It can be linked or unlinked (69), (70). In the following two examples, the dislocated items are given in square brackets and they are unlinked to the following clauses.


' The fingers, the hands, I do not see that far.'

Right dislocation expresses afterthoughts that either extend the reference of the doubled item, make it more explicit, or re-phrase it in order to help the addressee to arrive at a correct understanding. The dislocated item bears the same case marking as its doubled counterpart in the clause. In (71) the recipient, which is encoded as reflexive pronoun in the clause, is also expressed as a full noun phrase after the clause. In (72) the temporal adjunct has been repeated (though it is not an exact repetition).


### **27.3 Other types of focus constructions**

### **27.3.1 Thetic sentences and presentational constructions**

Thetic sentences are fully-focused sentences that can be uttered out of the blue or as answers to the question *What happened?* In Sanzhi, they have the same structure with 27.3 Other types of focus constructions

respect to the division of the utterance into topic and focus as utterances with the default information structure (73) (§27.2).

(73) [the beginning of a story] ca one šːi-l-cːe-b village-obl-in-hpl ca one kulpat, family χːula big kulpat family b-už-ib hpl-stay-pret ca-b cop-hpl ' In one village there lived a family, a big family.'

Moreover, there is a specialized construction for presentational and existential sentences that introduce new and mostly animate referents into discourse. This construction can be either a thetic sentence if it occurs, e.g., at the very beginning of a story (74), but it can also contain some topical material. In all cases the newly introduced referent is an argument that follows the verb and occurs as the subject if the verb is intransitive and as the object if the verb is transitive. Thus, the constituent order is VS (74), (75) or VO (76).<sup>2</sup> The optional topical items precede or follow the verb with its argument.

(74) daršːi friendship ka-b-iž-ib down-n-be.pfv-pret ca-b cop-n sːika=ra bear=add bec'=ra wolf=add kːurtːa=ra fox=add waˁlur=ra camel=add ' The bear, the wolf, the fox, and the camel were friends.'

(75) nišːa-la 1pl-gen b-irχʷ-i n-be.ipfv-hab.pst ħaˁz game '(When we were little) we had a game.'

(76) Allah-li Allah-erg ma take hana now dučːi-la night-gen paˁq strike či-ka-b-arq'-ib spr-down-n-do.pfv-pret duˁħi snow 'Allah sent snow during the night.'

### **27.3.2 Contrastive focus and floating predicative particles**

Selective and corrective focus, which belongs to the category of contrastive focus, is expressed by placing the predicative particles (§9.1) or the copula (§16.1) immediately after the focused constituent. This construction exists in many East Caucasian languages, but its frequency in texts seems to greatly differ from language to language. In this section, I will analyze the Sanzhi construction by first presenting data about the floating predicative particles and then about the floating copula. All predicative particles have the (verbal) predicate as their default host. In contrastive focus constructions, the particles float off from their normal host and appear on arguments and adjuncts, on phrasal heads and on modifiers of heads. The lexical verb must take the form of a participle (77–80). It can be either a complex participle with the suffixes*-il* or*-ce*, or it can simply be the modal

<sup>2</sup>This sentence contains the particle *ma*, which used when giving things to other people and inviting them to take the things (§9.5). In this examples, it is Allah who gave snow to the people who had to 'take' it, i.e., live with it.

### 27 Constituent order and information structure

participle. In contrast to sentences without contrastive focus, it cannot be a converb or the simple preterite participle.


As examples (81) and (82) show, floating predicative particles can even occur in certain types of subordinate clauses such as infinitival complements.


Most notably, with transitive and affective verbs it is possible to drop the ergative or dative marking of the subject if the predicative particle is attached to it.


When Sanzhi speakers are asked to formulate questions to which sentences with floating predicative particles are suitable answers, they produce content questions in which the relevant item that is focused in the answer is replaced by a question word serving as the host for the particles (85).

27.3 Other types of focus constructions

(85) a. ušːa 2pl ceqːel=da=ja when=2pl=q Družba-le Druzhba-loc d-ax-an? 1/2pl-go.ipfv-ptcp 'WHEN will you/do you have to go to Druzhba?' (E)

b. nušːa 1pl ižal=da today=1 d-ax-an 1/2pl-go.ipfv-ptcp 'We will/have to go TODAY.' (E)

In an interrogative clause with a floating person enclitic the interrogative enclitic – which is also a focus-sensitive particle – must be added to the same item that serves as the host for the person enclitic. Encliticizing the two markers to two different items (86) is ungrammatical.

(86) \* ušːa 2pl ceqːel=da when=2pl Družba-le Druzhba-loc d-ax-an=e? 1/2pl-go.ipfv-ptcp=q (Intended meaning: 'WHEN will you/do you have to go to Druzhba?')

Person enclitics or the past enclitic used as contrastive focus markers are not very frequent in the corpus although the constructions are readily available in elicitation as many examples in this section prove. In example (90) the past enclitic occurs in combination with another focus-sensitive predicative particle, the modal particle =*q'al*.


'It was Latvia that seemed to me to be really the most beautiful country.'


'But you were the one who said that they had taken only Alibatir to the ditch.'

### 27 Constituent order and information structure

It is slightly easier to find examples with interrogative enclitics (91–93) or the modal particle *=q'al* (94–96); see also the example (90) above. All corpus examples discussed so far can be classified as selective focus, because the focus selects one item from among a presupposed set of possible alternative values (Dik et al. 1981: 62).


Constituent focus constructions with the copula have the same pragmatics as constructions with predicative particles, that is, they express contrastive focus, most commonly selective focus. Constituent focus constructions with the copula are a bit more flexible in the sense that the lexical verb can not only occur as a participle, but appears occasionally in the form of the general converbs, which are normally used to form analytic tenses (101). In most of the examples, there is a weak pitch accent on the item preceding the auxiliary. In (97) the existential copula serves as host for the modal particle *=q'al*, which also belongs to the class of predicative particles.

(97) a-rurg-an neg-burn-ptcp le-b=q'al exist-n=mod it, that birikːalla.ʁut' cow.parsnip le-b=q'al exist-n=mod het that 'There is the one that does not burn, the cow-parsnip.'

27.3 Other types of focus constructions

(98) [A: They are sitting, enjoying themselves, passing the time. B replies:] e, yes ču-la refl.pl-gen dard-ane sorrow-pl šišːim-te suffering-pl istikan-na glass-gen hila-cːe-d behind-in-npl ca-d cop-npl d-irq'-an-te npl-do.ipfv-ptcp-dd.pl

'It is behind the glass that resolve their sorrows.' (i.e. by drinking alcohol)

(99) "ala 2sg.gen gawhar pupil ca-b," cop-n r-ik'ʷ-ar, f-say.ipfv-prs "zaja spoil b-iχ-ub-il" n-be.pfv-pret-ref '"It is your pupil that is spoiled," she said.'

It is not only the standard copula *ca-b* that can be used in focus constructions, but also the existential/locational copulas (§16.2):

(100) χːuˁrbe graveyard k'e-d exist.up-npl heχtːu-d there.down-npl či-d-ig-ul spr-npl-see.ipfv-icvb nišː-ala 1pl-gen ' Up there the graveyard of ours is visible.'

In the next two examples, the constituents followed by the copula are already mentioned and thus given in the context. These examples can count as confirmative or approving focus. The speakers do not correct some previous statements, but they acknowledge the selected alternative and reassure the selection. For instance, (101) is part of a description of a picture. The speaker is at first unsure about one of the depicted objects. With the utterance of (101) he affirms that those objects must be bunches or bundles of something unidentified. With the sentence in (102) the speaker sums up his argumentation about a possible sequence of events. The demonstrative pronoun and the verb bearing the participle suffix were uttered with a rising tone, and there was no noticeable accent on the pronoun preceding the copula.

(101) [What are these, not bundles? Bundles.]

cel-la=jal what.obl-gen=indef χːala bundle ca-b cop-n b-iχ-un-il. n-be.pfv-pret-ref hež=ra this=add χːala bundle ca-b cop-n ka-b-išː-ib-le down-n-put.pfv-pret-cvb

'This is a BUNDLE of something. This BUNDLE also has been put down (there).'

(102) [Discussing which events pictures might show] hel that ca-b cop-n hel-i-j that-obl-dat ag-ur-il go.pfv-pret-ref '[This exactly, this has happened]. This is it what has happened to him.'

Contrastive focus constructions that make use of the copula or of other particles such as person enclitics are found in a number of other East Caucasian languages. In particular, floating person enclitics have been investigated for Udi (Harris 2001; 2002), Lak (Kazenin 2009), Standard Dargwa (Xajdakov 1986) and Tanti Dargwa (Sumbatova 2013).

### 27 Constituent order and information structure

The impact of the modal particle on the information structure and its use in focus constructions has been analyzed by Forker (Accepted). Kazenin (2009) proposes an analysis of such constructions as synchronic cleft constructions because they have the pragmatic value of cleft construction: the predicative particle or copula identifies or specifies the argument in a presupposed open proposition. The structure can be simplified as follows:

### (103) (topic) [FOCUS]=particle/copula (topic)

Clefts functionally and formally resemble copula clauses (§22.2) when the focal part is followed by a clause headed by a participle that has the properties of a relative clause (83), (102) §23.3. The constructions express term focus. They can be formed not only with the help of the modal particle, but also with all other predicative particles or auxiliary verbs that function as copulas.

If we adopt the definition of cleft proposed by Lambrecht (2001), we also notice the morphosyntactic similarities between the constructions in Sanzhi and other Dagestanian languages and clefts in European languages. Lambrecht defines a cleft as a syntactically biclausal structure consisting of two simple clauses, that is, a main clause headed by an auxiliary, and a subordinate clause that is a (free) relative clause or relative-like clause. Thus, the auxiliary is the predicative particle or copula, i.e. items that are also used in copula clauses, and the lexical verb appears as participle. In most corpus examples, the focused constituent occurs at the left or right edge of the clause, so that the construction looks like a biclausal construction with a copula clause and a preceding or following relative clause that provides the information needed to identify the copula subject (101), (102) (for more details see Kazenin 2009; Forker 2016b). However, the topical part is also frequently simply a demonstrative pronoun (97).

Since it is possible to place the focused item in the middle of the clause, between what is supposed to represent a free relative clause, a synchronic cleft analysis as proposed by Kazenin (2009) becomes questionable. For a detailed discussion of the cleft approach and its problems see Forker (2016b).

### **27.3.3 Corrections**

Corrective focus is found in utterances that correct a previous statement, e.g. by replacing the relevant information. This is frequently done by means of the negative copula *akːu* that indicates constituent focus on the preceding item (104–106). The adversative particle *=n(u)* is optionally encliticized to the copula.


27.4 Verb doubling

(106) iž this hin-na water-gen akːu=n, cop.neg=prt iž this č'an-na wind-gen urχːab mill ca-b cop-n ' This is not a water, this is a wind mill.'

In (107) the order of rejecting clause and correcting clause is reversed in comparison with the above examples. The first clause, which represents a cleft-like constructions as discussed in §27.3.2, expresses the correction and the second clause the rejection.

(107) ala 2sg.gen mašin car ca-b cop-n zaja spoil b-iχ-ub-il, n-be.pfv-pret-ref ʡaˁli-la Ali-gen akːu cop.neg 'It is your car that is spoiled, not Ali's.' (E)

### **27.4 Verb doubling**

Another construction that is found in Sanzhi and in other East Caucasian languages is verb doubling. This construction consists of a verb that is repeated. The first part occurs either in the infinitive or in the form of the stem without any inflectional suffix, but preserving derivational morphology, spatial preverbs, or morphemes used in compounding. To this part the additive *=ra* is encliticized. The second part consists of a finite verb form of the same lexical verb (108–110). All corpus examples from Sanzhi make use of the verbal stem, but examples with the infinitive that have the same pragmatics can be elicited (see below).


(110) [But the Estonians, I think, are very similar to the Icari people, no difference, how often I went there] b-uč=ra n-drink.ipfv=add b-učː-u n-drink.ipfv-prs a-cːella 2sg-comit ka-b-iž-ib-le, down-hpl-be.pfv-pret-cvb qːuʁa-ce beautiful-dd.sg iχtilat=ra talk=add b-irq'-i n-do.ipfv-hab.pst 'As for drinking, they sit and drink with you, and (we) had nice conversations.'

It is also possible to employ the particle *=q'ar* instead of the additive. This particle is frequently used to topicalize the items to which it is added (§9.4.3). In (112) only the preverb in a slightly modified form is repeated, not the verbal stem.

### 27 Constituent order and information structure


are stairs from here and there.)'

Maisak (2010) discusses verb doubling constructions in a number of Dagestanian languages including Xuduc Dargwa and Icari Dargwa under the heading of "predicate topicalization". He concludes that the constructions cannot be used in thetic utterances, but only in reference to previously mentioned or otherwise already known situations, and that they mostly involve polarity focus, because there is an associated adversative or concessive clause that is following the verb doubling construction or implicitly present in the context. Maisak's interpretation seems to fit the Sanzhi data well, since in all examples at least part of the information conveyed in the verb doubling construction has been provided in the previous context. As for the adversative or concessive clause following the verb doubling construction, there is one corpus example in which the following clause has an adversative (112) marking, but the other examples (108–110) lack adversative clauses, and there is no example with a following concessive clause. Again, such examples can be elicited, but due to the general absence of associated adversative or concessive clauses in the corpus, the expression polarity focus cannot be regarded as the main function of verb doubling in Sanzhi. Instead, the main function is predicate topicalization.

### **27.5 Focus-sensitive particles**

Sanzhi Dargwa has a number of focus-sensitive particles. The most common particles are:


Further particles are *=cun* 'only', *=gina* 'alone, only', *malle* 'even', *arrah* 'at least', and *akːʷar* 'except, without, only'. In addition, all predicative particles, not only *=q'al*, can be used in focus construction and thus also belong to the class of focus-sensitive particles. 27.5 Focus-sensitive particles

The general functions of these particles are analyzed in more details in §9.1 and §9.4 including examples for every particle.

The focus-sensitive particles can be encliticized to focal items in term focus-constructions as was described in §27.3.2 above. Focal items are frequently nominals or adverbials, but verbs can also be focused. If the focal item is a verb, the enclitics are either added to the finite inflected verb or to some other part of the predicate such as the preverb (113), or the verb appears in a non-finite form such as the infinitive (114).

(113) [Thanks to Allah for the life he gave]

ʡaˁħ good r-al f-fit r-iχː-ib=da, f-guard-pret=1 qːuʁa beautiful r-al=ra f-fit=add r-iχː-ib=da, f-guard-pret=1 ucːi brother rucːi sister akːʷ-ar=xːar

```
cop.neg-prs.3=conc
```
'He guarded me (fem.) well, he also guarded me safely, although (I was) without brothers and sisters.'

(114) qili home w-iteʁ-ij=q'ar m-reach.pfv-inf=mod wahi-l bad-advz akːu cop.neg ' Coming home is not bad.'

For example, when *=q'al* is attached to the finite verb, it is the verb phrase or the whole clause that is in its scope. It is possible that the verb alone is in the scope of the particle and represents the focus (115), but it can also be the verb with its overtly expressed arguments (97), which sometimes amounts to the entire clause. If the entire clause is in its scope, we can have verum focus: e.g. in (116) the polarity of the verb is at stake, not the lexical meaning conveyed by the verb.


Particles can be combined and are either encliticized to the same host (89), (90) or two different hosts. Example (117) contains the additive and both modal particles occurring in one and the same utterance.

(117) nik'a-ce=xːare, small-dd.sg=conc du-l=q'ar, 1sg-erg=mod it=ra that=add ʡaˁči=q'alle, work=mod unc-e ox-pl sa-r-d-uc-ij hither-abl-npl-keep.pfv-inf uˁq'-ni-lla go-msd-gen ' Although I (masc.) was little, I myself went behind the oxen, this is also work.'

### 27 Constituent order and information structure

The particle *akːʷar* 'except, without, only' differs from the other particles because it is morphosyntactically the negative copula inflected for the third person of the habitual present. It follows the item in focus and occurs in a clause with negative polarity. Its literal meaning is '(X) not being' and the negative clause refers to a situation that only takes place when X is present. For instance, (118) can literary be translated as 'There were no cars not being trucks.'


Modal particles such as *=q'al* or *=q'ar*, but also additives, scalar additives, or exclusive particles are often compared to or sometimes even equated with focus particles. For instance, the Standard Dargwa particle *q'alli* is cognate with Sanzhi *=q'al*. It has been called a "sentence focus particle" by van den Berg (2001: 74–75).

However, all particles discussed in this section participate in the pragmatic structuring of the utterance, but they are not plain markers of focus. They all have a lexical meaning, although the semantics of modal particles is comparatively hard to access. Whenever they are used, this meaning is expressed, but their use is not restricted to focus constructions. There are a number of corpus examples in which the referent of the item bearing the particle is not only not new, but has been mentioned in the immediately preceding utterance (54), (70), (120). It is taken up again in the utterance containing a focus-sensitive particle. In such examples the host is frequently not a verb, but a nominal, and occurs in clause-initial position. The host is not focal, but topical, and the construction is used to topicalize patients (54), (120) or other semantic roles that are typically non-topical (70). For a more thorough discussion of the functions of *=q'al*, see Forker (Accepted).

(120) [Talking about Sanzhi dishes, the daughter of the speaker reminds her mother that there is also milk soup. Then the mother continues to talk about this topic.] nejg-la milk-gen nerʁ-be=q'al, soup-pl=mod cara-la=ra other-gen=add d-irχ-u npl-be.ipfv-prs

'There is milk soup and soup from other things, (we are also preparing milk soup…)'

## **28 Interrogative clauses**

Interrogative clauses are marked by interrogative enclitics and by rising intonation, but the latter is not always particularly salient. The interrogative enclitics belong to the class of predicative particles (§9.1). This means that interrogative enclitics can co-occur with certain non-finite verb forms in analytic tenses, and they turn the verb plus enclitic complex into a finite verb form used in main clauses. Thus, in many questions there is no copula, auxiliary or other predicative particle (person enclitic, past enclitic), but only a non-finite lexical verb and an interrogative enclitic, but the clause is nevertheless a full grammatical question. This chapter covers


### **28.1 Simple polar questions and disjunctive polar questions**

Polar questions are marked by means of an enclitic that has four allomorphs: *=w* after vowels, and *=uw* after consonants, with additional restrictions. The allomorph *=ew* occurs after the imperfective converb suffix *-ul* and the allomorph *=aw* after the first person suffix *-id* and after certain words ending in /aw/, e.g. *qːanaw* 'ditch'. Its use is obligatory. It is encliticized to the predicate, i.e., to the verb if there is a verb (1), (2), otherwise to the nominal predicate which bears a person or past enclitic (3). Elliptic polar questions can consist of nominals or other items without an accompanying verb or other auxiliary (4).


28 Interrogative clauses

```
(4) [Why did they not kill Osman?]
     qːačuʁ-e=w?
     bandit-pl=q
    ' The bandits?'
```
It cannot be added to another constituent if the clause contains a verb:

```
(5) a. t'ult'
         bread
               le-b=uw
               exist-n=q
                          ala?
                          2sg.gen
         'Do you have bread?' (E)
     b. * t'ult'uw leb ala?
```
Very frequently polar questions are followed by the interrogative pronoun *ce=ja* (6), which functions as a kind of hesitation marker or expresses doubts on the part of the speaker similar to a phrase such as 'I wonder'.

(6) kinu film b-aʔ n-begin ač'-ib-le=w come.pfv.pret-cvb=q ce=ja? what=q ' Did the movie start or what?'

Examples (6–8), (10) illustrate that the interrogative enclitics are sufficient to guarantee the finiteness of the clause even when no copula, person enclitic or past enclitic is present. In (6) the lexical verb appears as perfective converb, and in (7) and (10) as imperfective converb. In (9) the interrogative enclitic follows the past enclitic, which cannot be left out.

There are three basic ways of answering polar questions: (i) the full verb with or without arguments and adjuncts can be used (7); (ii) only the appropriate copula or auxiliary occurs (8), or (iii) only the particles *e* 'yes' or *a*, *aʔa* 'no' are used. However, the use of only these particles is not very common as Sanzhi speakers told me and as the data from my own corpus show. It is also possible to combine the strategies.


If the question is headed by a non-finite verb form without a copula, only the negative copula can stand alone in an answer. If the answer is affirmative either the particle *e* 'yes' or the whole predicate including the lexical verb must be used. The copula alone cannot serve as a grammatical answer (11).

28.1 Simple polar questions and disjunctive polar questions


If the question already contains the affirmative copula, then the copula alone is enough for making up a complete answer (12). However, in an interrogative clause such as the one in (12) normally no copula is used because the interrogative particle is sufficient (compare examples (7), (12) above).

(12) it that w-ax-ul m-go.ipfv-icvb ca-w=uw cop-m=q Derbent-le? Derbent-loc ca-w cop-m 'Is he going to Derbent? Yes, he is.' (E)

Answers to negative polar questions normally contain a predicate because the particles alone could lead to misunderstanding (13), (14). However, negative polar questions are rare and most of them are rather rhetorical questions to which no real answer is expected, but they express surprise or disbelief on part of the speaker (15).

	- b. ca one dučːi night d-už-ib=da 1/2.pl-be-pret=1 'We stayed one night.'

' Don't the Sanzhi people know at least that (story) to tell?'

Polar interrogative clauses are mostly verb-final (2), (13a), but it is also possible to find examples with verbs occurring in other positions (5a), (12), (16).

(16) hel-i-la that-obl-gen ʁaj-li-j word-obl-dat qili home arg-ul=de=w go.ipfv-icvb=2sg=q u? 2sg 'Are you going home because of what she said?'

### 28 Interrogative clauses

Polar interrogatives are frequently combined with a following phrase that bears the marker for embedded questions (17) since that marker is also used for expressing epistemic uncertainty (§28.4).

(17) ač'i=de=w wheat=pst=q ce what ca-d=de=l? cop-npl=pst=indq 'Was it wheat or something else.'

In disjunctive polar questions the interrogative enclitic appears on the predicate in each member of the disjunction:


' They say to the wife, "Did your husband have a head or not?"'

The same construction can also be used as an assertive disjunction without any interrogative illocutionary force (20). A similar multifunctionality of interrogative particles covering polar and content questions as well as in declarative disjunctions and some other contexts is attested in a number of other East Caucasian languages (e.g. Hinuq, see Forker 2013b) as well as unrelated languages such as Japanese and Malayalam (Slade 2011: 2, Uegaki 2018).

(20) hek' dem.up bari-li-j=uw sun-obl-dat=q bac-li-j=uw moon-obl-dat=q miši-l similar-advz ca-b cop-n 'This is similar to the sun or the moon.'

### **28.2 Content questions**

Content questions contain interrogative pronouns (see §4.5 for a list of the pronouns) and an enclitic that has two allomorphs. It is *=e* after consonants and *=ja* after vowels. The use of the interrogative enclitic is obligatory. The only two contexts where its use is prohibited are questions containing the second person enclitic *=de* or the past enclitic *=de*, which both end with /e/.<sup>1</sup> It is usually added to the predicate (verbal or nominal), unless there is no verb or other predicates, then it is added to the question word.

<sup>1</sup>One can argue that the enclitic *=e* is used in such cases, but because its phonological form is identical to the vowel in the past enclitic and the second person enclitic, its presence is not noticeable. However, in many instances a sequence of identical vowels leads to long vowels (§2.6.5), which is not the case for the respective interrogative clauses. The long vowel [eː] does not occur very often, but it is attested. Thus, it is reasonable to assume that in interrogative clauses an underlying long vowel [eː] has been shorted, but for a conclusive argumentation more research is needed.

28.2 Content questions

The constituent order in questions is such that topical items mostly precede the interrogative pronoun and the verb appears in clause-final position.


Often genitive phrases are split up if the head is part of the interrogative phrase and the genitive then follows the host of the interrogative enclitic (24). Such extraposed genitives are not restricted to questions but also frequently found in assertions (§27.1.2).

(24) cara other ce what χurejg food d-irχ-u=ja npl-be.ipfv-prs=q nišːa-la? 1pl-gen ' What other dishes do we have?'

If there is no verb then the respective predicate or even the interrogative pronoun can end up in clause-final position bearing the interrogative enclitic.


It is also possible to put the interrogative pronoun in clause-initial position, but this is far less common (27–30). Equally possible and slightly more common is the occurrence of material following the verb.


28 Interrogative clauses

(29) čina-b=e where-hpl=q hana now heštːi? these ' Where are these now? (these = people)'

(30) cellij why it-i-l that-obl-erg salam greeting a-lukː-an=e neg-give.ipfv-ptcp=q nišːi-cːe? 1pl-in 'Why will he not greet us?'

The only constituent order that is strictly forbidden is for the interrogative pronoun to follow the verb (31) or to follow the constituent that bears the person enclitic or past enclitic (32). In the latter examples possible orders are *u ča=de?* and *ča u=de?*:


As in polar questions the lexical verb can appear in non-finite forms (30) and the interrogative particle can take the place of the copula (29). Except for verbs, any constituent can be questioned, be it an argument or an adjunct, e.g. absolutive (24), ergative, spatial adjunct (29), manner adjunct (22), causal adjunct (30). Genitive modifiers or other modifiers of nouns can also be questioned (see §4.5 for examples). It is also possible to question constituents of subordinate clauses (33), (34) or of postpositional phrases (35).


Interrogative clauses can contain more than one interrogative pronoun (36). The order of the interrogative pronouns in (37) can also be switched around.

(36) hi-l who.obl-erg ce what padarit make.present b-arq'-ib=e? n-do.pfv-pret=q 'Who made which present?' (E)

28.3 Tag questions

(37) a and it that ce what darman=e medicine=q cellij? what.dat ' And this is what medicine for what?'

Answers to content questions can either consist of only the focus part (41) or they can be whole sentences. In the answers that are full sentences the nominal part of the constituent that constitutes the focus can be absent (i.e. in (39) the inflection on the verb is enough to convey the meaning of the first person pronoun that represents the answer to the question). In (40) the item in focus follows the verb which is rather unexpected if one embraces the position that the focus position in East Caucasian is immediately before the verb.


### **28.3 Tag questions**

Tag questions are very common in Sanzhi and occur after assertions. The form of the tag depends on the polarity of the assertion because it reverses the polarity. In the vast majority of corpus examples tag questions follow clauses with affirmative clauses and the tags are formed by means of the negative present tense copula to which the polar question marker *=w* is attached, i.e. *akːu=w?* (lit. 'Is it not?') (42), (43). This copula can even be used when the assertion contains a verb with past time reference (44). However, in such a case the negative past tense copula can also be used (45). Note that the copula in (42) that functions as short answer to the tag question has the default neuter agreement although the copula in the assertion before the tag question is inflected for masculine singular.

### 28 Interrogative clauses

(42) ag-ur go.pfv-pret d-iχ-ub-t-a-lla npl-be.pfv-pret-pl-obl-gen qari=či-b at.top=on-n ʁaj word ∅-ik'-ul m-say.ipfv-icvb ca-w cop-m heχ, dem.down akːu=w? cop.neg=q ca-b cop-n 'About what had happened he is talking, right? Yes.'

(43) ij kampanija ca-b, akːu=w?

this company cop-hpl cop.neg=q 'This is the company (group of friends), right?'

(44) di-la 1sg-gen pikri thought ħaˁsible following ka-jž-ib-le down-be.m.pfv-pret-cvb het-ka that-down čar back ∅-iχ-ub m-be.pfv-pret ca-w, cop-m akːu=w? cop.neg=q

' In my opinion, he was imprisoned and then came back from there, right?' (45) žaq'-ne a-d-uk-i nušːa-l, akːʷ-i=w?

boar-pl neg-npl-eat.ipfv-hab.pst 1pl-erg cop.neg-hab.pst=q ' We did not eat boars, right?'

The tense of the copula in the tag question reflects the tense of the verb (copula or other) or the tense of the copula if it is an analytic inflection form in the assertion preceding the tag. For instance, in (45) the verb in the assertion appears in the habitual past and the copula in the tag also appears in the habitual past.

When the assertion is negative the tag marker normally has positive polarity as, e.g., the copula *ca-b* (46) or the verb in (47), which is simply the negation of the predicate in the assertion.<sup>2</sup>

(46) baliqː-e fish-pl akːu=q'al, cop.neg=mod ca-d=uw? cop-npl=q aχːu not.know wallah by.God

'(These) are not fish, are they? By God, I don't know.'

(47) [a but iš-tːi that-pl juldašː-e friend-pl ce what b-ik'-ul=el] n-say.ipfv-icvb=indq tolko only hel that sːurrat picture ħaˁsible following b-aχ-ij n-know.pfv-inf a-w-irχʷ-ar, neg-m-be.able.ipfv-prs w-irχʷ-an-ne=w? m-be.able.ipfv-ptcp-fut.3=q e-rχʷ-an-ne, neg-be.able.ipfv-ptcp-fut.3 e yes

'And these friends, what they are saying, only by means of the picture, (one) cannot know, can one? One cannot know, yes.'

<sup>2</sup>Note that this is not the case in example (45), and I do not have an explanation for this example.

28.4 Subordinate questions

### **28.4 Subordinate questions**

Subordinate questions are marked with an enclitic that has three allomorphs: (i) *=jal* after vowels, (ii) *=el* after consonants, and (iii) *=l* after some suffixes ending in /e/, in particular the perfective converb suffix. The enclitic occurs in all types of embedded interrogatives, i.e. content questions, polar and disjunctive polar questions. It is also used as a complementizer with verbs of speech and cognition (§24.2.7), and for the formation of specific indefinite pronouns (§4.6.1). The rules for the placement of the enclitic are the same as for normal questions. This means that in embedded polar questions the marker occurs on the verb if there is any (48), and otherwise on the non-verbal predicate (49).


In embedded disjunctive polar questions it occurs on all members of the disjunction:


In embedded content questions the enclitic appears on the verbal or non-verbal predicate:


' We are thinking where to meet Abdulkhalik.'

28 Interrogative clauses

(54) nu well [ceʁuna which qːihin-dexː-e difficult-nmlz-pl či-d-až-ib=da=jal] spr-npl-see.pfv-pret=1=indq nišːi-j 1pl-dat b-alχ-ul=de n-know.ipfv-icvb=pst ' Well, we knew which difficulties we had seen.'

The Sanzhi enclitic is not only used in embedded questions, but also in the protasis of realis and irrealis conditional clauses (55) (§18.3). The apodosis of irrealis conditionals can be omitted, in which case the construction expresses wishes similar to an optative (56).


The marker for embedded questions has also evolved into a general marker of epistemic uncertainty occurring in rhetorical questions for which the speaker does not expect an answer (57).

(57) cet'le how herʔ-an-ne=l, say.ipfv-ptcp-fut.3=indq aj oh Allah! Allah 'How should this be said, oh, Allah!' (i.e. 'How should I say this?')

And more importantly, it is used when speakers are expressing thoughts whose truth they do not vouch for and when they advance hypotheses, i.e. for the expression of epistemic modality, more specifically, epistemic uncertainty. In such cases there is no matrix complement-taking predicate (58), (59), but optionally the borrowed adverb *belki* may occur (60). Since the marker for embedded questions belongs, just like the other interrogative enclitics, to the class of predicative particles, it can be used with a converb, but the resulting clause is finite (§9.1). Such examples might be interpreted as "insubordination" (Evans 2007; Evans & Watanabe 2016, see also Mithun 2008) because a clause with a dependency marker is used as an independent sentence.


28.4 Subordinate questions

(60) belki it.is.possible ij this tusnaq-le-r prison-spr-abl sa-jʁ-ib-il hither-come.m.pfv-pret-ref ca-w=el cop-m=indq ' Maybe this is when he came from prison.'

This use has been conventionalized in the frequently occurring phrase *ce ca-b=el* (*ce ca-d=el*) 'what it might be, whatever' lit. 'what cop-n-indq' (61).

(61) masliʡaˁt reconciliation b-arq'-ij n-do.pfv-inf saˁ-q'-un-ne=jal hither-go.pfv-pret-cvb=indq cara other ce what ca-b=el cop-n=indq 'He came out to reconcile them or so or there is something else.'

Evans (2007: 367), who introduces the term "insubordination", defines it as "the conventionalized main clause use of what, on prima facie grounds, appear to be formally subordinate clauses." He explains the diachronic development of insubordination by means of four steps:


Evans' four-step model provides a very plausible path of the diachronic change: the Sanzhi particle underwent all four stages, but at the same time preserved its function as complementizer. Thus, the syntactic development can be summarized as (62), accompanied by the semantic extension schematized in (63).


Examples for the first step are (48–54). Embedded interrogatives occur with matrix verbs that denote cognitive activities (which can include perception verbs such as 'see'). Embedded interrogatives are often of irrealis modality and therefore not asserted as factual or actual events or situations (49). During the third step the interpretation was restricted from various possible main clauses to an omitted main clause as general as 'It is probable that X'. Language-internally, the reanalysis (step 4) might have been supported by the presence of the other predicative particles. This means that by analogy with the person markers, the past enclitic or the modal particle the embedded question marker received its syntactic ability to express finiteness of clauses.

However, there is one general problem with the insubordination analysis of the enclitic *=(e)l*/*=(j)al*. In principle, it is possible that the diachronic development occurred in

28 Interrogative clauses

the reverse order, i.e. that the particle was originally a marker of epistemic modality that subsequently came be to be used in embedded interrogatives due to its epistemic modal meaning. Because we lack data of older stages of Sanzhi Dargwa, this question cannot be resolved with certainty.

Insubordination of the type just described is also found in other East Caucasian languages. Examples in case are irrealis markers in the Tsezic languages Bezhta, Hunzib, and Hinuq, the potential infinitive in Bagvalal, and the potentialis (i.e. the infinitive) in Tsakhur (see Comrie et al. (2016) for examples and references). The study by Kalinina (2011) provides many examples of insubordinated exclamative utterances in Agul, Archi, Avar, Bagvalal, and Bezhta (East Caucasian) as well as Adyghe (West Caucasian). Of the surveyed languages only Agul employs an irrealis conditional form in embedded questions and exclamatives in a similar fashion as the Sanzhi example (56).

### **28.5 Other uses of questions**

Some questions are normally used as greetings. Thus female addressees are not greeted with the Arabic phrase *As-salam ʡaˁlaykum*, but with the sentences below; the first variant is for single addressees (64a), the second for situations when the addressee is more than one person (64b). The answer is again a clause (64c).

	- b. ka-d-iž-ib-le=da=w! down-1/2pl-be.pfv-pret-cvb=2pl=q 'Hello!' (lit. 'Are you sitting?') (plural addressee)
	- c. ka-r-iž-ib-le=da down-f-be.pfv-pret-cvb=1 'Hello!' (lit. 'I am sitting.')

For greeting people who have just got up in the morning, the sentences in (65) are used. The addressee can be female or male.

(65) r-alh-un=de=w! f-wake.up.pfv-pret=2sg=q / / w-alh-un=de=w! m-wake.up.pfv-pret=2sg=q 'Good morning!' (female/male addressee)

When Sanzhi speakers meet during the day they might ask when greeting each other:

(66) ce what ʡaˁči work b-arq'-ib=de? n-do.pfv-pret=2sg 'What (work) have you done?'

This question can be taken literally, but it is also a general question of the type "How do you do?"

## **29 Reflexive and reciprocal constructions**

This chapter discusses the syntax of reflexive and reciprocal constructions. The morphological paradigms of reflexive and reciprocal pronouns are given in §4.3 and §4.4.

### **29.1 Reflexive constructions**

Sanzhi has morphologically simple and complex reflexive pronouns. The pronouns are *ca-w*/*ca-r*/*ca-b* in the singular (oblique stem *cin-*) and *ca-b*/*ca-d* (oblique stem *ču-*) in the plural. The relevant case paradigms are given in §4.3. Their functions are summarized in Table 29.1. There are two types of complex reflexive pronouns, which always consist of reduplication of the simple reflexive pronoun. Both types contain one reflexive that bears case marking according to the function of the reflexive pronoun in its clause and often appears as the second part of the complex reflexive pronoun. The first part either copies the case marking of the controlling noun phrase ("case copying") or it invariably bears the genitive case ("complex genitive reflexive"). The functional range of both types is roughly identical, but there are three restrictions that all concern reflexives marked by the genitive case. First of all, complex genitive reflexives do not have a genitive case form, because this would lead to two identical pronouns used together, which is ungrammatical (see Table 4.8 in §4.3). Therefore, in the respective constructions only case-copying complex reflexives can be used (e.g. local reflexivization and reciprocal constructions). Second, in emphatic reflexivization the complex reflexives are morphologically rather a mixture of genitive reflexive and case-copying reflexive, because they consist of a first part in the genitive and a second part that has the same case as the nominal to which the emphatic belongs (§29.1.2). Third, only genitive forms of simple reflexives occur as pause fillers; all other case forms cannot be used.

In this chapter, local, non-local, and emphatic reflexivization are treated, as well as coreference across clausal boundaries, and reciprocalization. For the other functions see the respective sections (references to them are given in Table 29.1).

In all types of local reflexive constructions, the pronouns are only used with third person referents. For reflexivization of first and second person, the personal pronouns are used. In long-distance reflexivization with logophoric function, simple reflexive pronouns are occasionally used to refer to speech act participants (i.e. first or second person referents).

### 29 Reflexive and reciprocal constructions

Table 29.1: Functions of simple and complex reflexive pronouns


### **29.1.1 Local reflexivization**

In local reflexive constructions, pronouns are bound by an antecedent within the same clause. With first and second person, the normal personal pronouns occur:


Only for third person reflexivization the reflexive pronouns are used. Almost all corpus examples contain simple reflexive pronouns, but (4) shows a complex reflexive in the function of beneficiary.


Reflexive pronouns are in complementary distribution with personal or demonstrative pronouns (5), (6). The c-command requirement holds, i.e., a possessor cannot control a reflexive pronoun (7).

(5) it-i-l that-obl-erg cin-i-j refl.sg-obl-dat cik'al something isː-ul buy.ipfv-icvb ca-b cop-n 'S/he buys herself/himself something.' (E)

29.1 Reflexive constructions


The complex reflexive pronouns must be locally bound (8) and therefore cannot have an antecedent in another clause. The example in (8) is fully grammatical if the pronoun is bound by the noun *aba* 'mother', which occurs in the same clause.

(8) Madina-j Madina-dat b-ikː-ul n-want.ipfv-icvb ca-b cop-n [aba mother cinij refl.dat ca-r refl-f či-r-až-ib-le] spr-f-see.pfv-pret-cvb 'Madina wants that the mother sees herself∗/.' (E)

The simple pronouns can occur as arguments and adjuncts of various types, e.g. patients, experiencers (7), stimuli (23), beneficiaries (5), goals (9), possessors (10), and complements of postpositions (11).


The same is true for the complex reflexive pronoun except for the possessor function and the use in postpositional phrases. The complex genitive can never be used as possessor, and the case-copy pronoun is judged as marginal or interpreted as an emphatic reflexive and thus not as part of a complex reflexive pronoun. Thus, the preferred and entirely acceptable reading of (12) is 'Rashid himself loves his mother'. In the possessor function normally the simple reflexive pronouns are used (10). The same point is illustrated in example (13): the reflexive bearing the ergative case functions as emphatic particle whereas the genitive reflexive occupies the possessor position.

(12) ?? Rašid-li-j Rashid-obl-dat cin-i-j refl.sg-obl-dat cin-na refl.sg-gen aba mother r-ičː-aq-u f-love-caus-prs.3 'Rashid loves his mother.' (E)

29 Reflexive and reciprocal constructions

(13) Rašid-li Rashid-erg cin-ni refl.sg-erg cin-na refl.sg-gen qal house b-arq'-ib n-do.pfv-pret 'Rashid himself/alone built his house.' (E)

Similarly, a complex reflexive within a postpositional phrase is judged as possible but less felicitous than a simple pronoun unless the first part can function as emphatic reflexive (11).

The simple and the reflexive pronouns are interpreted as bound variables that can be bound by indefinite noun phrases (14), (15).


The internal order of the two parts of the complex reflexive pronoun that exhibits case copying is free to some degree. Thus, in example (16) the two pronouns can be switched around and they can also be positioned before the controller.

	- b. *Rašid [cinij caw] erči wik'ul caw*
	- c. *[caw cinij] Rašid erči wik'ul caw*

This is not possible in the ergative construction. In contrast to the extended intransitive construction in (16), the complex reflexive cannot precede the controlling noun (17). Due to time constraints, I did not systematically test verbs from different valency classes and their use with complex reflexives in local reflexivization. This topic must be left to future research.

(17) \* [cin-ni refl.sg-erg ca-w] refl-m Rasul-li Rasul-erg gap praise w-irq'-ul m-do.ipfv-icvb ca-w cop-m (Intended meaning: 'Rasul is praising himself.') (E)

We can speculate a bit about the origin of complex reflexives. It is possible to elicit examples in which it seems that the pronoun can be split up (18a), (18b). In these examples, the two parts of the case-copying reflexive pronoun are independent of each other and do not form one constituent. The part that copies the case functions as emphatic reflexive, which is co-referent with the controlling noun phrase, and enforces the reflexive interpretation. The second part is a simple reflexive pronoun. It is likely that such sentences represent the diachronic source of the case-copying complex reflexive

29.1 Reflexive constructions

constructions.<sup>1</sup> In the following examples (18) the emphatic reflexive is given in bold. The other reflexive pronoun functions as goal argument of the predicate 'look at' in a standard local reflexive construction.

	- b. *caw Rašid cinij erči wik'ul caw*

As (19) shows, the emphatic reflexive cannot be separated from the noun it accompanies by an intervening verb, which is a general rule that applies to all emphatic reflexives.

(19) \* *Rašid cinij erči wik'ul caw caw*

The situation is different in case of the complex genitive reflexive which consists of a genitive reflexive pronoun and a second reflexive pronoun that takes the appropriate case-marking, because they do not allow for varying orders of the two pronouns. The reason for the restriction is probably a grammaticalization path, which differs from that of the case-copying complex reflexives. Genitive reflexives are not used in the emphatic reflexive function, but only as pause fillers, and I do not want to suggest that their function as pause fillers forms the basis of the complex genitive, although I lack an alternative hypothesis. In the complex genitive reflexive construction, the internal order of the genitive reflexive pronoun cannot be changed (20). If we switch the order around to *cinna caw*, (20) becomes grammatical.

(20) \* Rasul-li Rasul-erg [ca-w refl-m cin-na] refl.sg-gen gap praise w-irq'-ul m-do.ipfv-icvb ca-w cop-m (Intended meaning: 'Rasul is praising himself.') (E)

Within a ditransitive construction, the direct (21) or the indirect object (22) can function as binder, although simple reflexive pronouns would be preferred in such examples.

(21) Pat'imat-li Patimat-erg Rašid Rashid surraticːe-w picture.in-m cin-na refl-gen cinij refl.dat či-w-iž-aq-ul=de spr-m-see.ipfv-caus-icvb=pst 'Patimat was showing Rashid to himself on the picture.' (E)

(22) Pat'imat-li Patimat-erg či-w-iž-aq-ul=de spr-m-see.ipfv-caus-icvb=pst Arsen-ni-ji Arsen-obl-dat surrat-le-w picture-loc-m či-w on-m cinij refl.dat ca-w refl-m

'Patimat was showing to Arsen himself on the picture.' (E)

<sup>1</sup> I am grateful to an anonymous reviewer for this suggestion. Note that the root of the absolutive reflexive pronouns, *ca*-, is very likely a cognate of the standard copula *ca-b*.

### 29 Reflexive and reciprocal constructions

Most notably, the case marking of the antecedent and the reflexive pronoun can swap. This phenomenon, which is cross-linguistically rare, has been observed in a number of East Caucasian languages, among them Sanzhi Dargwa and other Dargwa varieties (see Forker 2014 for a typological study). It is constrained by three interacting factors: morphological complexity of the pronouns, constituent order, and valency class of the predicate.

For morphologically simple reflexive pronouns, case swapping is generally unavailable. For instance, if in clauses with transitive or affective verbs the cases are distributed such that the controlling noun bears the case marking of the agent or experiencer (ergative or dative) and the reflexive appears in the absolutive, then a local reflexive (14), (15) and a non-reflexive reading are possible (23). In the non-reflexive reading, the pronoun refers to a contextually salient referent that, for instance, had been mentioned in the preceding discourse:

(23) Madina-j Madina-dat ca-r refl-f r-ikː-ul f-want.ipfv-icvb ca-r cop-f 'Madina likes/wants/loves herself.' OR 'Madina likes/wants/loves her.' (E)

If we swap the case marking, only the non-reflexive reading remains. With swapped case marking, it is more natural to position the pronoun in the dative case before the noun in the absolutive (24), although the reversed order is also possible.

(24) cin-i-j refl.sg-obl-dat Madina Madina r-ikː-ul f-want.ipfv-icvb ca-r cop-f 'She likes/wants/loves Madina.' (E)

With complex reflexive pronouns, affective as well as transitive predicates exhibit a reversal of case marking (25–26), but all other positions including co-arguments of extended intransitive verbs are excluded (27). With transitive and affective predicates the distribution of the case marking in reflexive constructions is free, i.e. either the controller or the pronoun take the ergative or the dative case suffix (25–26).

	- b. Rasul Rasul ca-w refl-m cin-ni refl-erg / / cin-na refl-gen cin-ni refl-erg gap praise w-irq'-ul m-do.ipfv-icvb ca-w cop-m 'Rasul is praising himself.' (E)
	- b. Rasul Rasul cinij refl.dat ca-w refl-m / / cin-na refl-gen cinij refl.dat čiːg-ul praise.m-cvb ca-w cop-m 'Rasul sees himself.' (E)

29.1 Reflexive constructions

The reversal of the marking is forbidden for co-arguments of extended intransitive verbs that are marked with spatial cases. For example, 'believe' requires an argument in the absolutive and another argument in the dative. The controller of the reflexive pronoun must be in the absolutive case; any change in the marking leads to ungrammaticality (27).

	- b. \* Šamil-li-j Shamil-obl-dat či-w-w-irχː-ul spr-m-m-believe.m.ipfv-icvb ca-w cop-m cin-na refl-gen ca-w refl-m (Intended meaning: 'Shamil believes in himself.') (E)

Out of context, there is no semantic or pragmatic difference between reflexive constructions with standard case marking and reflexive constructions with reversed marking. This stands in contrast to other East Caucasian languages for which such differences have been reported (see Forker 2014 for references). There are some restrictions on the word order concerning both standard reflexive constructions (see above) and those with reversed marking, but this topic requires further research.

To sum up the discussion on complex reflexive pronouns in Sanzhi, we can state that these anaphoras can function as A (agents and experiencers) as the examples in (25b) and (26b) have shown. The constructions illustrated in both sentences are plain reflexive constructions with basic transitive and affective verbs. Under certain circumstances, the complex reflexive pronouns can also precede their antecedents (16c), but this topic requires more testing. Since the complex reflexive pronouns are not personal or demonstrative pronouns, but must be locally bound, the Sanzhi data look like a violation of a commonly assumed constraint, namely that anaphors cannot fulfill the grammatical role of subjects. However, such an argumentation crucially depends on the notion of subject and whether it can be applied to Sanzhi Dargwa. Forker (2014) argues that Sanzhi does not have subjects in exactly the same sense in which this term is applied to European languages, and that therefore the examples in (25b), (26b) do not represent a violation of the subject anaphora constraint.

### **29.1.2 Emphatic reflexive and intensifying uses**

Simple and complex genitive reflexives serve emphatic and intensifying functions, in which case they are not bound by a co-referential argument in the clause but simply occur together with a nominal co-constituent. Complex reflexive pronouns that copy the case of the controller cannot be used in this function. The nominal co-constituent can be a pronoun (32) or a noun (28), or it can be absent (35). In the Sanzhi corpus, most emphatic reflexives occur without the nominal co-constituent, but if they co-occur, then the reflexive follows the nominal, which is cross-linguistically the most common position for emphatic reflexives. Furthermore, the simple reflexives are preferred over the complex reflexives in this function. The simple and complex genitive reflexives can

### 29 Reflexive and reciprocal constructions

only function as emphatic reflexives with third person co-constituents. As is the case with local reflexivization, emphatic reflexivization of first and second persons is done with first and second person pronouns. These pronouns occur in the genitive case and are usually used without the pronominal co-constituents (see the examples below).

König & Gast (2006) list the following four functions of emphatic reflexives:


Reflexive pronouns in Sanzhi predominantly occur in the first function, in which two situations or two referents are contrasted with each other. This can be done by means of parallel structures in which the items follow each other and are explicitly contrasted. In (28) the speaker contrasts herself with a friend called Hurija. In example (29) the contrast is expressed in a parallel structure that is marked by means of the additive enclitic =*ra* on both members, the speaker herself and the friends.


brothers), the guys, I also did not show a lack of money, things (i.e. supported them), and (they) themselves also did not, so I have lived (my life).'

In most cases the contrast is rather indirect and resembles topic switch constructions in which the sentence topic switches from one sentence to the next (30), (31). As examples (32), (33) show, among the complex reflexives only genitive reflexives occur in the emphatic reflexive function.

(30) [He says, "Hello, Asja Iwanowna."] ca-r refl-f ka-r-icː-ur down-f-stand.pfv-pret ca-r be-f er look či-ka-r-ik'-ul spr-down-f-look.ipfv-icvb heχ-i-j dem.down-obl-dat ' (She) herself is standing and looking at him.'

29.1 Reflexive constructions


'They themselves were carrying the sacks and when they were filled ...'

(33) tem.bolee moreover nišːa-la 1pl-gen priezd-li-j=ra arrival-obl-dat=add ču-la refl.pl-gen ca-b=ra refl-hpl=add padgatuwleni=de prepared=pst

'Moreover, (the Icari people) themselves were prepared for our coming.'

The second function of emphatic reflexives, the adverbial-exclusive function 'alone, without help', is also attested for Sanzhi. Example (34) illustrates this function with a complex genitive reflexive. Example (35) originates from a fairy tale and here the reflexive can be interpreted as adverbial-exclusive and/or as adnominal-contrastive.


The third function (adverbial-inclusive 'also, too') is not common in East Caucasian languages, including Sanzhi, because the languages have additive enclitics that already serve this function (§9.4.1). The fourth function (attributive 'own') is covered by pronouns in the genitive case, i.e., by personal pronouns for first and second person and by simple reflexive pronouns for the third person.

The first and second person genitive pronouns, when used as intensifiers, slightly differ in their morphosyntactic properties from the reflexive pronouns. First of all, they normally do not occur together with the pronominal co-constituent (36a). It would not be natural to add the pronoun in the absolutive to this sentence in order to overtly fill the position of the copula subject (36b), and there are no examples of this kind in the corpus.

### 29 Reflexive and reciprocal constructions

(36) a. [He was a relative) of those people up there. I don't know of whose family he was.] nik'a small durħuˁ=de boy=pst di-la 1sg-gen '(I) myself was a little boy.' b. ?? du 1sg nik'a small durħuˁ=de boy=pst di-la 1sg-gen '(I) myself was a little boy.' (E)

Second, in all corpus examples, which are only a small handful, the genitive pronoun appears at the right boundary of the clause after the verb (36a), (37), (38). This position is typical for contrastive topics (§27.2.1). But as the example in (39) proves, this position is not obligatory. As to the function, the first and second person genitive pronouns used as intensifiers fulfill the first function as adnomial intensifiers (36a), (39) and the second as adverbial-exclusive particles (37), (38).


### **29.1.3 Long-distance reflexivization**

The simple reflexive pronouns are also bound across clausal boundaries. This means that they can occur in various types of subordinate clauses with the antecedent belonging to the main clause. Such a usage is impossible for complex reflexives. Sentence (40) illustrates long-distance reflexivization with a relative clause. In (42) we find the reflexive pronoun in a complement clause.

(40) il=qːel that=when juldaš-li friend-erg [juldašː-a-l friend.pl-obl-erg cin-i-j refl.sg-obl-dat sa-qː-ib-te] hither-carry.pfv-pret-dd.pl xunul-be gift-pl ʡaˁli-j Ali-dat d-ičː-ib npl-give.pfv-pret 'At that, the friend gave to Ali the gifts that his friend had brought to him (= to the friend).'

29.1 Reflexive constructions

The vast majority of instances of long-distance reflexivization are logophoric contexts. This includes longer stretches of discourse that are framed by verbs of speech and cognition. Normally it is the author of the quote, who serves as the antecedent of the reflexive:


'When the man was in prison he remembered a lot, how the police beat him up, how he beat up his wife.'

(43) ca one zamana time bari=ra sun=add wahi-ce evil-dd.sg č'an=ra wind=add čːal argument d-uq-un npl-go.pfv-pret [kutːi which ču-cːe-rka refl.pl-in-abl c'aq'-ce=de=l] mighty-dd.sg=pst=indir.q

' Once the sun and the evil wind argued about who is stronger.'

It is also possible that the speaker of the quote is included in the group of referents of the reflexive pronoun, i.e., pronoun and antecedent are not identical in their denotation, but the domain of reference of the reflexive is larger (44).

(44) hel-i-l that-obl-erg b-urs-ib n-tell-pret ca-b cop-n [ču-j refl.pl-dat ag-ur-il] go.pfv-pret-ref ' He told what had happened to them.'

Reflexive pronouns are also used in non-logophoric contexts of co-reference across clauses and across sentences. They can occur when the speaker wants to refer to a topical referent in the discourse that is not necessarily used as an argument in the preceding clause, but simply a salient discourse topic at the moment of utterance. This is the case with optative phrases used to commemorate dead relatives and friends or acquaintances (45).

(45) hel that ʡuˁrus Russian xːunul woman er look r-ik'-ul f-look.at.ipfv-icvb r-už-ib-le; f-be-pret-cvb alžana heaven b-ikː-ab n-give.pfv-opt.3 cin-i-j refl.sg-obl-dat

'It turned out that the Russian woman had looked (at the events), may heaven be given to her.'

### 29 Reflexive and reciprocal constructions

Discourse topics expressed with reflexive pronouns are also found outside of optative phrases (46). In (47), it seems that the speaker used first the reflexive because he assumed that the referent of the pronoun would be topical enough to be interpretable, but then he changed his mind and added the full noun phrase as an afterthought in order to reassure the reference of the pronoun.


In cases of discourse topics the referent of a reflexive pronoun can even be inanimate (48).

(48) [ There is a tall summit.] muza summit arilla during.day muza summit b-ik'ʷ-ar hpl-say.ipfv-prs cin-i-j=ra refl.sg-obl-dat=add 'It is called the midday summit.'

### **29.2 Reciprocal constructions**

Reciprocal constructions are built either with plural reflexive pronouns (54) or with specialized pronouns that make use of the numeral *ca* 'one'. There are three types of reciprocal pronouns whose paradigms are given in §4.4. All pronouns occurring in reciprocal constructions are morphologically complex with the same patterns that the morphologically complex reflexive pronouns exhibit.

Reciprocal pronouns are always clause-bound. They occur in various argument and adjunct positions and are controlled by a suitable plural antecedent that can be omitted. In the following examples, they function as patient (60a), as experiencer or stimulus (49), as beneficiary, as addressee (50), as goal (63a), and as genitive possessor fulfilling the semantic role of an experiencer (51), (52).

29.2 Reciprocal constructions

(49) ca-lla one-gen ca one ħuˁrmat respect b-irq'-ul=de, n-do.ipfv-icvb=pst calli-j one.obl-dat ca one b-ikː-ul=de hpl-want.ipfv-icvb=pst

' They respected each other, loved each other.'


' The main point is that they understand each other, the couple.'

(52) ca-lla one-gen ca-lla one-gen urk'i heart hitːi-dex behind-nmlz b-akːu n-cop.neg 'There are no bad feelings between each other.'

In all the above corpus examples, the first part of the reciprocal pronoun copies the case of the antecedent, which is absent from the clause, and the second part takes the case marking appropriate to its role in the clause. It is also possible, just like with complex reflexive pronouns, to mark the first part invariably with the genitive (52), (53).

(53) Murad=ra Murad=add Rašid=ra Rashid=add ca-lla one-gen calli-j one.obl-dat qːurt push b-ik'-ul hpl-aux-icvb ca-b cop-hpl 'Murad and Rashid are pushing each other.' (E)

Other variants of reciprocal constructions involve the plural reflexive pronouns (54) and the group numeral form of *ca* 'one', which is *ca-b-a* (55) (§6.4). The latter item means 'the ones, some' and therefore (55) has, in addition to the reciprocal interpretation, another reading in which one person loves another one, who in turn, loves a third person, and so on, such that there are no reciprocal feelings of love between any of the involved persons.


### 29 Reflexive and reciprocal constructions

Reciprocal pronouns can also be marked with spatial cases (50) or be governed by postpositions (56), (57).


Syntactically, reciprocal constructions show the same properties as local reflexivization. The c-command requirement holds. Therefore, possessors cannot bind reciprocal pronouns. For instance, in (58) the conjoined possessor noun phrase cannot serve as an antecedent for the reciprocal pronoun, but only the head of the genitive phrase can. The pronouns are interpreted as bound variables and can thus be controlled by non-specific noun phrases (59).


As has been shown for complex reflexive pronouns above, the reciprocal pronouns can also occur in the position of the ergative agent controlled by an antecedent that fulfills the role of the absolutive patient.

	- Murad=add Rashid=add one.obl-erg one n-hit.pfv-pret ' Murad and Rashid hit each other.' (E)

Similarly, experiencers can be expressed by reciprocal pronouns that are bound by absolutive stimuli. In other words, case marking can swap from the standard distribution to the reverse non-standard distribution. Note that in (61) this does not lead to any change in the form of the reciprocal pronoun because this is the case-copying variant and the two cases involved are the same independently of which case appears on the antecendent.

29.2 Reciprocal constructions

(61) a. Musa-j=ra Musa-dat=add Murad-li-j=ra Murad-obl-dat=add calli-j one.obl-dat ca one b-alχ-u hpl-know.ipfv-prs.3

'Musa and Murad know each other.' (E)

b. Musa=ra Musa=add Murad=ra Murad=add calli-j one.obl-dat ca one b-alχ-u hpl-know.ipfv-prs.3 'Musa and Murad know each other.' (E)

In fact, it seems that the reversed case marking pattern is sometimes preferred with affective constructions. Thus, the standard case marking has been rejected or judged as very marginal for a similar clause with the same type of reciprocal pronoun (62).

(62) ?? Madina-j=ra Madina-dat=add Pat'imat-li-j=ra Patimat-obl-dat=add calli-j one.obl-dat ca one či-b-ig-ul spr-hpl-see.ipfv-icvb ca-b cop-hpl (Intended meaning: 'Madina and Patimat see each other.') (E)

Transitive verbs and affective verbs are the only valency types that permit the cases to be switched around. As we have noticed for complex reflexive pronouns, swapping of case marking is ungrammatical for extended intransitive verbs (63b).

	- those-obl-dat one.obl-dat one look spr-hpl-say.ipfv-prs.3 (Intended meaning: 'They look at each other.') (E)

Again, there is some freedom concerning the word order both with the standard case marking pattern and when the cases have been switched around. Nevertheless, there are word orders that are forbidden, most notably when the pronoun is split apart and the part that copies the case precedes its antecedent from which the case has been copied. More generally, complex reciprocal pronouns, just like complex reflexive pronouns, cannot be split into two parts, and none of the individual parts could be interpreted as fulfilling another function (e.g. as emphatic particle, intensifier or as pause filler). Thus, they must occur next to each other as one constituent.

(64) standard case marking

a. Madina-l=ra Madina-erg=add Dinara-l=ra Dinara-erg=add [cal-li one.obl-erg ca] one gap praise b-irq'-i hpl-do.ipfv-hab.pst.3 'Madina and Dinara (regularly) praised each other.' (E)

29 Reflexive and reciprocal constructions


Note that again the case marking of the reciprocal pronouns is identical for the standard patterns as well as for the reversed patterns as is obvious when comparing examples above with the following sentences (65).

	- a. Madina=ra Madina=add Dinara=ra Dinara=add cal-li one.obl-erg ca one gap praise b-irq'-i hpl-do.ipfv-hab.pst.3 'Madina and Dinara (regularly) praised each other.' (E)
	- b. *Madinara Dinarara [ca calli] gap birq'i*
	- c. *[ca calli] Madinara Dinarara gap birq'i*

Finally, reciprocal pronouns can only have antecedents within the same clause. For instance, in (66) the pronoun is bound by the conjoined noun phrase 'Patimat and Murad' and cannot be controlled by the compound noun *atːa-aba* 'parents' in the higher clause. The pronoun consists of a part in the dative in accordance with its function in the clause, and a first part that either copies the case of the controller (66a) or occurs in the absolutive (66b).

	- father mother-dat n-want.ipfv-icvb cop-n Patimat-erg=add Murad-li=ra Murad-erg=add ca one calli-j one.obl-dat kumek help b-arq'-ib-le] n-do.pfv-pret-cvb ' The parents want that Patimat and Murad help each other.' (E)

## **30 Minor constructions**

### **30.1 Comparative constructions**

In comparative constructions two or more items are examined in order to note similarities and differences in degree between them (Dixon 2008: 787). Inequality between two items is expressed by means of one of the spatial cases (§3.4.2.2). In superlative constructions, degree adverbs occur. Equative constructions and the expression of similarity are realized by means of several particles (§30.2).

In Sanzhi comparative constructions we find a comparee, the standard of comparison, and the parameter of comparison. The standard of comparison is marked with the loc-ablative case that has the suffixes *-ler(ka)*, *-ar(ka)* or *-jar(ka)* (§3.4.2.2). It is crosslinguistically common to mark the standard of comparison with an ablative (or locative) case (Dixon 2008: 791), and East Caucasian languages including Dargwa varieties nicely confirm this tendency. Neither the comparee nor the parameter of comparison bears any special marking. Consequently, if the standard of comparison were to be omitted, the construction would be a simple clause and not a comparative construction. Most commonly the standard precedes the comparee. The parameter is a gradable adjective or adverb that occurs in its plain form without any additional index (as, e.g., English *more*).


' The animals had apparently more conscience than our rich (people).' (lit. 'their conscience was bigger')

### 30 Minor constructions

Superlative constructions contain a comparee, a standard of comparison, and a parameter. They basically have the same structure as the constructions described so far in this section. The only differences are the case marking of the standard, which is now the in-lative, and the additional degree adverb modifying the parameter. The standard of comparison can be omitted if it is inferable from the context


### **30.2 Equative constructions and the expression of similarity**

For equative constructions and the expression of similarity Sanzhi has two particles, *ʁuna* and *daˁʡle* 'as, like', and the adjective *miši* 'similar'.

The particle *ʁuna* 'as, like' immediately follows the parameter of comparison that it has scope over like, e.g., focus-sensitive particles. The parameter can be a pronoun, an adjective (9), an adverbial (10), or a noun (12). Very often it is simply a demonstrative pronoun, and the combination of demonstrative and equative particle means 'like this, such' (11). Depending on the parameter, the particle thus appears, e.g., within noun phrases (9) or in the position of adverbial modifiers.


30.2 Equative constructions and the expression of similarity

It can also occur as a predicate in a copula clause without a head noun and it can be nominalized by suffixing -*b* (unclear origin) and the cross-categorical suffix in the plural form -*te* (*ʁunabte*).

(12) heχ dem.down kuruškːa mug ʁuna eq b-irχʷ-i n-be.ipfv-hab.pst.3 'This was like a mug.'

The particle *daˁʡle* 'as, like', which diachronically seems to be an adverbial derived with the adverbializing suffix -*le*, has a meaning very similar but not identical to *ʁuna*. It indicates only that some situation or some item resembles another situation or item. Both particles slightly differ in their distribution. The particle *daˁʡle* follows the parameter of comparison over which it has scope. As with *ʁuna*, the parameter can be expressed by nouns (13), adverbials (14), or adjectives (15). But in contrast to *ʁuna*, *daˁʡle* is most frequently used in non-finite clauses headed by participles (16) or the infinitive (17).


Finally, the adjective *miši* 'similar' assigns the dative case to its complement that represents the standard of comparison (18). In copula clauses, in which it is used in the copula complement, the adverbializing suffix *-le* is added, as it regularly happens with adjectival stems in copula construction.

(18) tusnaq-li-j prison-obl-dat miši-l similar-advz akːu cop.neg 'This is not similar to a prison.'

### 30 Minor constructions

The differences between the three comparative constructions lie mostly in their morphosyntactic behavior, with an additional semantic distinction between *ʁuna* and *daˁʡle* on the one side, and *mišil* on the other (19), (20). The particles *ʁuna* and *daˁʡle* have the distribution of focus-sensitive particles and can therefore occur within certain types of phrases as, e.g., noun phrases, but do not assign case to the items they scope over, in contrast to the case-assigning adjective *miši*.


### **30.3 Comitative constructions**

Sanzhi has two ways of expressing comitative meaning: case marking (in combination with optional postpositions) and a construction involving the use of reflexive pronouns.

The cases used are the comitative case (*-cːella*, §3.4.2.1) or, more rarely, the in-ablative case (§3.4.2.4). They can occur together with the postposition *b-alli* (§8.2.1) or the postposition/adverb *canille* (§8.2.2). These constructions can be used with animate and inanimate nouns. In the latter case they can express instruments (21).


There does not seem to be a clear semantic difference between *b-alli* and *canille* (23). The two items can only be distinguished by means of their morphosyntactic behavior, because *b-alli* agrees in gender with the argument in the absolutive (23) and it always implies a complement even when the complement is not overtly expressed. For instance, (24) entails that there were other people with whom we came, whereas in (25) there is no such implication and *canille* only functions as an adverb that expresses the fact that Madina and the speaker came together:

(23) burta lid ħaˁšak-li-cːella pot-obl-comit b-alli n-together / / canille together b-irχʷ-an n-be.ipfv-ptcp ca-b cop-n ' The lid should be together with the pot.' (E)

30.3 Comitative constructions


The second construction is the use of a reflexive pronoun in what looks like a coordination of noun phrases. This construction has been described for Standard Dargwa by van den Berg (2004). The structure is [Y*=ra* X*=ra*] 'X with Y'. X refers to an animate (usually human) entity and is formally expressed through the reflexive pronoun. Y is a nominal that can be animate or inanimate and takes case suffixes. It can be a common noun, a pronoun, a personal name or any other type of noun phrase. Both X and Y are marked with the additive *=ra* and are often adjacent to each other, which makes them look like a coordinated noun phrase. However, the argument referred to by the reflexive can be expressed independently. Furthermore, the coordinated noun phrase usually does not take an argument position in the clause. It is rather one of the individual members that functions as argument. For instance, in (26) the pronoun *heχ* that is following the comitative phrase represents the subject of the following verb as the agreement on the verb shows (masculine singular).


Example (28) shows that the two items bearing the additive enclitic *=ra* can be separated by other material. The agreement on the existential copula is controlled by the first noun phrase *ca ca šuša,* which is semantically plural and functions as the copula subject of the existential/locational copula *χe-d*, and the reflexive pronoun appears in a kind of right-dislocated position, such that it is syntactically not part of the subject constituent.

(28) harkil-li-cːe-d every-obl-in-npl ca one ca one šuša=ra bottle=add χe-d exist.down-npl ca-b=ra refl-hpl=add ' Everybody is with a bottle in their hands.' (lit. 'There is one bottle each in everybody's (hand), and they also.')

It is also possible to elicit examples in which the semantically coordinated items function as a coordinated noun phrase. The coordinated noun phrase controls plural agreement on intransitive verbs if it functions as subject (29). However, masculine singular would also be possible in this type of construction as (27) shows. In example (30) the two coordinated items are marked for the ergative case. Again the coordinated noun phrase rather looks like an adjunct in the clause in which *Musal* is the agentive argument.

30 Minor constructions


The construction has probably evolved from the emphatic use of reflexive pronouns (§29.1.2) in combination with the additive meaning of the enclitic *=ra*. Thus, in (31) the two parts occur in independent clauses that follow each other as arguments of their respective verbs. The reflexive pronoun in the second clause doubles an omitted subject argument and conveys the emphatic meaning 'she herself'. The whole construction can be rephrased as 'both the backpack and she herself' and has a comitative reading ('she went away with the backpack') that has to be inferred from the structure.

(31) ca one xːunul-li woman-erg χːap grab b-arq'-ib-le n-do.pfv-pret-cvb hil-i-la that-obl-gen wešimišuk'=ra backpack=add b-erqː-ib n-carry.pfv-pret ca-b, cop-n ca-r=ra refl-f=add ag-ur go.pfv-pret ca-r cop-f

'There one woman grabbed his backpack, took it, and went away with it.' (lit. 'She also took [the backpack], and [she herself also] went away.')

### **30.4 Possession**

Possession is either expressed by cases or by means of the *b-ah* construction. In the first case the possessor is marked with the genitive case and most commonly preceding the possessed item (32), but other positions are available, too (33) (see §21.1 on noun phrases and §27.1 on the constituent order of phrases). There is no grammaticalized distinction between alienable and inalienable possession. Clauses expressing possession are copula clauses (§22.2) containing locational copulas (§16.2).


If the respective item is not permanently possessed but only temporarily in the custody of the possessor, the in-essive case is used (34) (§3.4.2.4).

(34) hež-i-cːe-b this-obl-in-n šuša bottle ca-b cop-n deč-la drinking-gen ' He has a bottle with a drink.'

30.4 Possession

The second way of expressing possession is the *b-ah* construction. The noun *b-ah* means 'owner' (plural *b-ahin-te*, also translates as 'parents'). It is one of the few nouns that have a gender prefix expressing the gender of the owner. The possessed item appears in the genitive with *b-ah* as the head noun of the genitive phrase. If the possessor is overt it occurs after *b-ah*. Both noun phrases together form an appositive phrase (§21.1). The possessed items in this construction are normally inanimate objects (35–38). Often they refer to clothes (37) or body parts (35), (36) that are used to characterize and identify the owner. From this noun the adjective-like item *wahwalla* 'own, everybody's own' with frozen gender agreement has been derived.


' On the upper side you smear […] water, egg, whatever you have.'

## **Appendix A: List of affixes and enclitics**

This section gives a list of all prefixes, suffixes, infixes, and enclitics, including those suffixes that do not have glosses. The list includes the name of the item, their glosses (if any), and the number of the chapter or section where the item is treated. Predictable allomorphs are separated by comma or given in brackets, and occasionally within separate entries. Allomorphs that occur in free variation and cannot be predicted are separated by forward slashes.











## **Name index**

Abdullaev, Said, 8, 9, 12 Abdullaev, Zapir G., 12, 82, 124, 126, 188 Belyaev, Oleg, 11, 12, 257, 264, 274, 296, 306, 386, 394, 398, 487, 498, 521, 523, 524, 530 Bickel, Balthasar, 342, 390, 417, 488, 491, 497, 499, 500 Butt, Miriam, 521 Chumakina, Marina, 376, 383 Comrie, Bernard, 11, 342, 365, 430, 444, 476, 487, 521, 552 Corbett, Greville G., 373, 376 Creissels, Denis, 410, 487, 499, 500, 512, 513 Crisp, Simon, 430 Curnow, Timothy J., 423 Daniel, Michael, 12, 257, 274, 444 Dik, Simon C., 454, 518, 534 Dixon, R. M. W., 341, 423, 432, 454, 569 Dobrushina, Nina, 9 Dryer, Matthew S., 528 Evans, Nicholas, 279, 550, 551 Foley, William A., 430 Forker, Diana, xiii, 6, 7, 11–13, 65, 92, 99, 101, 105, 106, 132, 168, 172, 174, 221, 282, 342, 351, 360, 386, 389, 390, 393, 430, 469, 471, 476, 487, 500, 521, 523, 524, 530, 536, 540, 544, 558, 559 Gadzhimuradov, Gadzhimurad, xiii, 7, 12, 13 Gadžieva, O. Š., 12

Gagliardi, Annie, 383, 384, 389 Ganenkov, Dmitry, 11, 12, 197, 217, 342, 354, 386, 388, 389 Gasanova, Saida M., 9, 12 Gast, Volker, 560 Gigineishvili, B. K., 11 Grenoble, Leonore, 8, 9 Haig, Geoffrey, 525 Harris, Alice C., 535 Haspelmath, Martin, 11, 132, 244, 487 Helmbrecht, Johannes, 373 Hengeveld, Kees, 454 Izutsu, Mitsuko Narita, 519 Jusupov, Khizri A., 12 Kadibagomedov, Amirbek A., 12 Kalinina, Elena, 11, 12, 161, 174, 393, 552 Kazenin, Konstantin, 389, 413, 487, 498, 499, 512, 535, 536 Khalilova, Zaira, 351 Kibrik, Aleksandr E., 7, 11, 430, 479 King, Tracy Holloway, 521 Komen, Erwin R., 512, 518, 521 König, Ekkehard, 560 Koryakov, Yuri, 7, 9, 11 Kulikov, Leonid, 370 Kustova, Marina, 307, 498, 501 Lambrecht, Knud, 536 Lander, Yury, xiii, 12, 65, 70, 124, 161, 162, 174,188, 219, 295, 319, 386–388, 444, 512, 515 Langacker, Ronald, 496 Magomedov, Rasul M., 9

### Name index

Magometov, Aleksandr A., 11, 12 Maisak, Timur, 538 Malchukov, Andrej L., 519 Mithun, Marianne, 307, 502, 550 Musaev, Magomed-Said M., 9, 12 Mutalov, Rasul, xiii, 12, 161, 162, 180, 190, 278, 296, 306, 319, 346, 360, 379, 389, 422 Neeleman, Ad, 519 Nichols, Johanna, 44, 217, 430 Polinsky, Maria, 7, 361, 389, 476, 479 Potsdam, Eric, 389, 476, 479 Pustet, Regina, 419 Reinhart, Tanya, 496 Repp, Sophie, 519 San Roque, Lila, 14, 94, 151, 267, 272 Šaxbanov, Marko, 9 Schilling, E. M., 12 Schulze, Wolfgang, 92, 373 Serdobolskaya, Natalia, 11, 12, 476, 479 Seržant, Ilja A., 514 Shibatani, Masayoshi, 514 Slade, Benjamin Martin, 544 Sumbatova, Nina, xiii, 9, 11, 12, 65, 70, 161, 162, 174, 180, 188, 190, 219, 278, 295, 296, 306, 319, 346, 360, 379, 386–390, 393, 394, 422, 430, 515, 535 Tatevosov, Sergei, 174, 217, 221, 244, 365 Temirbulatova, Sapijakhanum M., 11, 12 Testelets, Yakov G., 389, 407, 487, 498, 499, 512, 515, 521, 524 Uegaki, Wataru, 544 Uslar, Peter K., 9, 12 Vallduví, Enric, 519 van den Berg, Helma, 11, 12, 69, 70, 124, 132, 174, 188, 217, 342, 365, 373, 386, 508, 540, 573

Van Valin, Robert D., 430 Vilkuna, Maria, 519

Watanabe, Honoré, 279, 550 Wixman, Ronald, 9

Xajdakov, Said M., 535

ablative, 64, 65, 65<sup>5</sup> , 68, 69, 72, 73, 76, 103, 104, 108, 128, 140–143, 150, 153, 155–157, 198, 217, 218, 221, 314, 315, 348, 358, 415, 418, 437, 438, 445, 569, 572, 580 absolutive, 11, 43, 54–56, 64, 89–91, 97, 104–108, 115, 131, 147, 149, 150, 154, 155, 157–159, 189, 195, 214, 230, 232, 261, 278, 280–282, 285, 296, 303, 341–346, 349– 352, 354, 356, 357, 357<sup>2</sup> , 360, 361, 363, 364, 374, 375, 377– 380, 383–386, 388, 389, 394– 396, 400, 411–414, 417–422, 425, 426, 431, 434–436, 439, 445, 456, 458, 466, 467, 475– 477, 479, 480, 488, 513, 518, 523, 530, 546, 557<sup>1</sup> , 558, 559, 561, 566, 568, 572 additive, 169–173, 178, 217, 222, 328, 336, 507, 510, 537–539, 561, 573, 574, 580, *see also* enclitic, additive adjectives, 22, 46, 78, 85<sup>11</sup> , 86, 89, 119, 121–127, 129, 137, 146, 147, 162, 169, 183–185, 187–191, 194, 195, 197, 206, 229, 237, 311, 322, 373<sup>1</sup> , 375, 376, 403–405, 407, 408, 410, 412, 424, 426, 440, 511, 512, 515, 517, 571, 577, 578 adjectivizer, 127, 197 adverb degree, 122, 128, 139, 146, 147, 569, 570 manner, 102, 127, 139, 146, 147, 428, 579

147, 187, 217, 314, 375, 380, 384, 415, 528 temporal, 104, 144, 145, 150–152, 323, 324, 525 adverbializer, 31, 37, 107, 133, 195, 250<sup>2</sup> , 258, 306, 323 agent, 28, 50, 57, 65, 73, 77, 78, 80, 127, 260, 262–264, 268, 344, 351, 363, 364, 378, 386–389, 392– 394, 398–401, 418, 435, 443, 446, 447, 461, 475, 477, 483, 494, 558, 566, 578–581 agreement closest conjunct, 383, 384 deviant, 385–390 person, 11, 162, 163, 165, 205, 214, 215, 243, 259, 261, 264, 268, 269, 278, 281, 282, 286, 299, 303, 305, 317, 318, 330, 344– 346, 349, 354–358, 373, 390, 391, 394–396, 398–401, 418, 420, 432, 440, 471, 491, 503 agreement controller, 215, 251, 285, 288, 299, 301, 356, 357, 374, 375, 377–381, 384, 387–390, 400, 466, 475, 476 agreement prefix, 34–36, 44, 80, 81<sup>9</sup> , 184, 207–212, 214, 215, 220, 222, 224, 232, 288–290, 357, 357<sup>3</sup> , 385, 386, 389, 400, 420, 439, 456, 466, 467, 475 agreement target, 380, 381, 384 antipassive, 57, 58, 257, 259, 260, 297, 299, 344, 360–362, 362<sup>4</sup> , 363– 366, 378, 417, 432, 435, 523

spatial, 44, 65, 77, 103, 104, 139–144,

apposition, 84, 191 aspect imperfective, 243, 259, 320, 360, 361 perfective, 11, 207, 210, 227, 276, 320, 335, 474 assimilation, 30, 37, 262, 324 causative, 28, 37, 205, 212, 215, 225, 227, 315, 317, 320, 360, 368, 370, 577, *see also* construction, causative clause adverbial, 80, 152, 163, 169, 275, 277, 293, 305–308, 310, 312, 313, 315, 317, 319, 321–326, 328– 330, 361, 362, 401, 438, 444– 446, 449, 454, 459, 460, 470, 473, 486–490, 490<sup>1</sup> , 491–500, 502, 503, 526 conditional, 116, 163, 167, 178, 243, 254, 277, 278, 293, 330–333, 335, 391, 487, 503, 550 interrogative, 107, 191, 197, 306, 419, 471, 518, 529, 533, 543, 544<sup>1</sup> , 581 purposive, 455 relative, 86, 163, 184–188, 190–195, 256, 258, 268, 286, 305, 309– 316, 328, 329, 361, 401, 403, 406, 408, 410, 412–414, 433, 434, 438–447, 449, 454, 457, 458, 465, 473, 483–485, 499, 511, 513, 515, 517, 518, 526, 536, 562 simple, 341, 384, 417, 420, 536, 569 clause union, 187, 476 cleft, 257, 386, 517, 536 comitative, 54–56, 58, 63, 65, 67, 73, 90, 105, 106, 108, 113, 157, 158, 351, 383, 438, 443, 505, 525, 554, 572–574, 578 complementation, 449, 453, 454, 461, 463, 465, 474, 475, 482, 483,

499 compounding, 43, 83, 123, 125, 197, 206, 215, 227, 228, 234, 294, 424, 467, 537 concessive, 115, 117, 118, 172, 289, 305, 327, 328, 330, 336, 337, 380, 538, 581 conditional, 163, 167, 168, 172, 214, 215, 245, 277–279, 281, 290, 293, 305, 318, 325, 328, 330–337, 391, 393, 395, 397, 398, 400, 503, 515, 552, 577–581 conjunction, 161, 166–169, 171, 172, 297, 330, 382, 384, 432, 505–509 consonant, 11, 19, 20, 22–25, 27, 28, 30, 32, 37–40, 49, 50, 66, 78, 92, 113,194, 206, 225, 227, 257, 462, 541, 544, 549 distribution of, 19 geminate, 24, 49, 78 tense, 48 constituent order, 11, 123, 153, 192, 219, 387, 403, 417, 419, 421, 443, 473, 474, 481, 488, 511, 512, 518–522, 526, 527, 529, 531, 545, 546, 558, 574 construction causative, 228, 366–368, 370, 371 comparative, 69, 128, 147, 569, 572 equative, 570 involuntary agent, 73 partitive, 60 reciprocal, 106, 391, 553, 564–566 reflexive, 61, 104, 105, 180, 432, 553, 554, 556–559 control backward, 11, 389, 461, 479–481 forward, 479, 481 converb imperfective, 32, 147, 173, 212, 213, 249, 250<sup>2</sup> , 252, 258, 271, 273– 275, 278, 291, 306, 308, 326, 366<sup>6</sup> , 453, 454, 462, 465, 469,

476, 477, 480, 490, 492, 493, 497, 500, 541, 542, 581 perfective, 37, 195, 212, 250<sup>2</sup> , 257– 259, 262–264, 273, 275, 276, 282, 292, 293, 306, 308, 309, 326, 453, 454, 457–460, 462, 469, 470, 476, 477, 486–488, 491–493, 498–500, 542, 549 coordination, 84, 171, 423, 487–489, 498, 499, 505–507, 509, 510, 573 copula, 57, 79, 105, 123, 125, 158, 162– 166, 169, 181, 184, 187, 194, 195, 222, 224, 249–252, 254– 258, 261, 262, 264–266, 268– 272, 281, 282, 285–294, 306, 308, 310, 320, 322, 327, 328, 345, 348, 349, 356, 357, 357<sup>2</sup> , 375, 376<sup>2</sup> , 380, 384–390, 393– 395, 400, 401, 403, 410, 417– 430, 445, 453, 488, 507, 513, 515, 517, 518, 526, 527, 531, 534– 536, 540–543, 546–548, 557<sup>1</sup> , 561, 571, 573, 574, 577 existential, 194, 256, 314, 375, 410, 420, 427, 534, 573 locational, 23, 174, 272, 291<sup>2</sup> , 426, 427, 514, 573 counterfactual, 254, 333 cross-categorical suffix, 120, 126, 131, 158, 161, 191, 197, 255–258, 266, 268, 273, 289, 290, 308, 311, 312, 316, 319, 373, 374, 400, 401, 403, 405, 412, 424, 425, 433, 441, 445, 453, 454, 457, 458, 465, 474, 483–485, 499, 515, 571 dative, 11, 33, 43, 54–56, 58, 61, 69, 80, 89, 90, 102, 105–108, 110, 111, 113, 131, 132, 199, 232, 245, 261, 280, 281, 295, 296, 298, 302, 313, 315, 317, 320, 329, 332, 333, 341–344, 346, 347, 350– 358, 368, 369, 377, 378, 384–

389, 395, 396, 406, 418–420,

428, 435, 436, 446, 451, 456, 474, 475, 480, 485, 488, 493, 518, 532, 558, 559, 568, 571, 579 degemination, 30, 31, 39, 40, 47–51, 225 delabialization, 30, 35, 39, 46 derivation, 43, 79, 80, 85<sup>11</sup> , 103, 125, 139, 172, 197, 207, 227, 577–581 dislocation, 517, 529, 530 ejective, 11, 19, 20, 22, 25 enclitic, 31, 32, 39, 104, 107, 114, 116–118, 127, 153, 162–164, 170, 171, 173– 179, 215, 249–251, 254, 256– 258, 264, 268–270, 273, 275, 276, 286, 287, 292, 306, 310, 316, 321–324, 327–329, 332, 337, 355, 356, 391, 395, 400, 419, 420, 422–425, 427–430, 462–464, 471, 474, 485, 486, 491, 505, 506, 527, 529, 533, 541, 542, 544, 544<sup>1</sup> , 545, 546, 549– 551, 574, 581 additive, 84, 112, 117, 118, 135, 166, 171, 337, 423, 483, 505–507, 560, 573 modal, 163, 401, 429 temporal, 104, 107, 293, 578, 580, 581 epenthetic vowels, 26 ergative, 11, 31, 37, 40, 43, 54–58, 60, 65– 67, 89, 90, 95<sup>1</sup> , 105–108, 110, 113, 131, 132, 149, 199, 232, 233, 245, 246, 257, 259, 260, 263, 281, 296, 299, 315, 332, 333, 341–344, 346, 349–356, 360– 367, 369, 374, 377–379, 384– 389, 394, 396, 418, 419, 430– 432, 435, 438, 447, 461, 468, 479, 481, 483, 488, 514, 532, 546, 555, 556, 558, 566, 573, 580 experiencer, 11, 61, 261, 280, 333, 342, 344, 345, 353–357, 357<sup>2</sup> , 368, 369, 386, 393, 398, 401, 418,

428, 435, 443, 451, 456, 477, 480, 558, 564 experiential, 62, 76, 184, 187, 194, 195, 257–259, 266–269, 273, 286, 342, 348, 358, 368, 374, 400, 420, 431 finiteness, 11, 161, 169, 391, 542, 551 floating modifier, 403, 410, 412, 413, 512, 514–516, 518 focus, 162, 174, 262, 263, 268, 387, 393, 423, 471, 473, 489, 511, 518–523, 525–527, 528<sup>1</sup> , 529, 531, 532, 534–536, 538–540, 547 completive, 518, 519 constituent, 386, 393, 485, 518, 519, 529, 536 contrastive, 177, 518, 531–534 corrective, 518, 531 focus-sensitive particle, 119, 169, 171, 173, 177, 491, 492, 503, 511, 533, 538–540, 570, 572, 578 gender, 11, 24, 31, 34–36, 39, 43, 44, 51, 54, 60, 64, 65, 77, 80, 81, 81<sup>9</sup> , 89, 90, 103, 104, 106, 108, 119–122, 126, 127, 127<sup>1</sup> , 133, 136, 155, 157, 162, 184, 188, 205–212, 214– 217, 220–222, 222<sup>1</sup> , 225, 230, 232, 240, 249, 249<sup>1</sup> , 251, 278, 281, 282, 285, 288, 289, 289<sup>1</sup> , 290, 291, 297, 300, 326, 344, 346, 349–352, 354, 357, 358, 361, 363, 373–376, 376<sup>2</sup> , 377– 381, 384–386, 388, 389, 393, 394, 396, 400, 403, 405, 411, 418, 420, 421, 425, 426, 428, 431, 439, 455, 456, 467, 469, 476, 479, 488, 506, 572, 575, 577–581 genitive, 11, 23, 37, 43, 54–56, 58, 60, 60<sup>4</sup> , 61, 65, 79, 81<sup>9</sup> , 83, 86, 89, 90, 105, 106, 108–110, 113, 124, 127, 131, 132, 135<sup>1</sup> , 144, 149–156, 180,

230, 233, 234, 320, 324, 325, 342, 344, 345, 349, 353, 357, 403–405, 407–410, 413, 414, 418, 423, 425, 427, 438, 443, 444, 465, 469, 483, 511–515, 518, 545, 553, 555, 557, 559–562, 564–566, 574, 575, 580 goal, 64, 67, 70, 344, 346, 348, 351, 353, 356, 357, 367, 436, 437, 443, 475, 485, 524, 525, 529, 557, 564 grammatical cases, 11, 43, 54 grammatical role, 370, 394, 397, 418, 431, 432, 488, 524, 559 habitual present, 39, 162, 163, 212, 214, 215, 243, 245–247, 251, 281, 290, 298, 306, 317, 318, 332, 359, 361, 391–393, 540, 577– 579, 581 hearsay evidentiality, 283, 468 ideophone, 19, 21, 28, 228, 236, 450 imperative, 32, 163, 181, 212, 214, 295– 299, 301, 359, 363, 391, 392, 417, 419, 431, 432, 471, 489, 490, 503, 577, 578 infinitive, 169, 170, 184, 187, 212, 230, 277–280, 282, 286, 290, 297, 305, 315–319, 321, 323, 324, 330, 401, 434, 453, 454, 460– 462, 465, 471, 474–478, 485, 537, 539, 552, 571, 579 information structure, 11, 179, 386, 387, 419, 430, 476, 511, 512, 518–521, 524, 526, 529, 531, 536 insertion glide, 30 glottal stop, 30, 33 insubordination, 113, 279, 303, 550, 551 interjection, 26, 161, 183 juxtaposition, 455, 506, 507

184, 187, 193, 194, 197, 199, 200,

166, 183, 191, 205–207, 210, 212, 215, 216, 218, 224, 227, 228, 235,

labialization, 22, 30, 31, 38 masdar, 37, 38, 40, 80, 162, 186, 212, 224, 225, 285–287, 290, 293, 305, 317, 320, 329, 330, 391, 444, 453, 454, 457, 458, 461, 474, 485, 499, 580 microtoponym, 77, 201 mirative, 174 modal interrogative, 212, 277–279, 295, 303, 304, 419, 456, 464, 579 negation, 28, 31–36, 116, 123, 177, 214, 215, 217, 222, 224, 246, 251, 252, 254, 269, 280, 282, 283, 292, 333, 420, 421, 440, 474, 548, 577 nominalizer, 127, 188 noun phrase, 84, 123, 129, 143, 149, 157, 169,171,187–190,192, 373, 375– 377, 380, 382–384, 403, 405– 415, 420, 421, 423, 425, 440, 441, 494, 495, 505, 506, 511, 512, 514, 515, 529, 530, 553, 556, 564, 566, 568, 573 nouns, 11, 22, 28–31, 34, 35, 39, 40, 43, 44, 44<sup>1</sup> , 45, 46, 48–52, 54, 56, 58, 60<sup>4</sup> , 65–70, 74, 75, 77–85, 85<sup>11</sup> , 86, 89–91, 97, 111, 112, 117, 119– 121, 124–127, 127<sup>1</sup> , 129, 133, 135, 137, 142, 147, 154, 156, 159, 169, 171, 184, 185, 187–191, 195, 197– 200, 206, 229, 230, 232–234, 258, 289, 300, 312, 317, 320, 322, 353, 373, 375, 381, 382, 382<sup>3</sup> , 388, 403–406, 408, 410, 412, 413, 439, 440, 444, 445, 465, 495, 506, 517, 518, 546, 571, 572, 575, 577–581 number, 7–9, 11, 12, 20, 22, 27, 30, 39, 40, 43, 49, 50, 54, 60, 64, 67– 69, 73, 76, 77, 80, 82, 83, 89, 90, 103, 104, 106, 108, 119, 120, 122, 124, 126, 127, 136, 155, 157, 162,

240, 249, 266, 271, 278, 281, 285, 286, 290, 306, 310, 312, 315–318, 326, 332, 341, 343– 346, 348–352, 354, 355, 357<sup>3</sup> , 358, 360, 361, 364, 366, 368, 373–379, 381, 383, 385, 386, 388, 390, 391, 393, 394, 396, 400, 403, 407, 411, 417–420, 430, 431, 433, 453, 454, 465, 471, 474, 476, 487–489, 506– 508, 512, 521, 535, 538, 540, 544, 558, 577 numerals cardinal, 129, 135, 172, 190 collective, 129, 172 group, 129, 133, 375, 565, 577 multiplicative, 32, 107, 111, 129, 135, 580 ordinal, 124, 129, 132, 194, 195, 313, 579 object direct, 60, 60<sup>4</sup> , 232, 257, 370, 524 indirect, 370, 524, 557 oblique stem, 37, 66, 67, 85<sup>11</sup> , 90, 91, 169, 185, 413, 446, 553, 579, 580 optative, 163, 173, 181, 214, 216, 222, 295, 296, 299, 300, 302, 374, 391, 392, 406, 417, 419, 445, 447, 489, 550, 563, 564, 577, 578 orthography, 8, 13, 19, 22, 24 palatalization, 30, 225, 227 parentheticals, 455 participle, 28, 30, 68, 79, 82, 86, 123– 125, 158, 169, 184, 185, 187, 206, 249, 250, 253–256, 258, 271, 286, 287, 289, 290, 305, 309– 315, 322, 324, 373, 377, 391, 403, 406, 414, 433, 438<sup>1</sup> , 440, 441, 443, 457, 485, 505, 529, 531, 534–536, 571, 577

locative, 40, 83, 125, 197, 305, 309, 314, 315, 321, 329, 433, 445, 577 modal, 169, 170, 184, 212, 249, 309, 311–314, 316, 318, 321–323, 325, 327, 328, 366<sup>6</sup> , 414, 433, 441, 445, 457, 465, 474, 483–486, 531 preterite,124,185–187,192,194, 266, 308–310, 313, 321, 322, 324, 325, 327–329, 433, 441, 457, 458, 465, 470, 474, 532 particle, 22, 23, 32, 112, 161–166, 168, 169, 173–175, 178–183, 222, 251, 285, 327, 329, 393, 395, 406, 417, 419–423, 429, 430, 451, 453– 457, 462, 466, 468–471, 473, 474, 481, 482, 485, 488, 489, 491, 492, 501, 510, 526, 531, 531<sup>2</sup> , 532–544, 546, 550–552, 555, 562, 567, 569–572, 579, 580 past, *see also* tense, past experiential, 222, 258, 259, 268, 269 habitual, 162, 163, 212, 214, 243, 245–247, 281, 290, 317, 332, 333, 354, 361, 392, 394, 548, 577–579, 581 patient, 56, 58, 60, 323, 356, 363–365, 374, 386–389, 392, 394, 398– 401, 418, 436, 477, 564, 566 pause filler, 181, 183, 567 pharyngealization, 19, 23, 28–30, 34 pluperfect, 222, 258, 259, 265, 273, 282, 391 possession, 59, 124, 420, 423, 427, 574, 575 predicate cognition, 451, 458, 460, 463, 481 complement-taking, 316, 389, 447, 449, 451, 453, 459, 473, 476– 479, 482–485, 550 preverb, 11, 23, 24, 26, 28, 31–39, 65, 75, 149, 153, 156, 157, 178, 205–207, 209, 212, 214–224, 227, 228,

240, 298, 313, 313<sup>1</sup> , 341, 349, 357, 375, 386, 415, 468, 537, 539, 578–581 prohibitive, 181, 212, 214, 222, 295, 298, 299, 301, 317, 317<sup>2</sup> , 359, 363, 391, 392, 419, 578–581 pronoun demonstrative, 54, 67, 89, 91, 94, 100–102, 104, 107, 135, 139, 145, 146, 151, 156, 169–171, 192, 217, 257, 290, 301, 322, 323, 373, 403, 404, 407, 409, 410, 414, 425, 439, 440, 472, 496, 512, 535, 536, 554, 559, 570 demonstratives, 581 indefinite, 32, 89, 109, 110, 112–119, 129, 133, 134, 170, 172, 179, 188, 191, 380, 388, 439, 440, 517, 549, 578–580 reciprocal, 79, 89, 106, 133, 375, 431, 432, 553, 564–568 reflexive, 67, 89, 104–106, 180, 188, 285, 375, 383, 404, 406, 443, 469, 471, 472, 475, 483, 495, 496, 505, 530, 553–557, 557<sup>1</sup> , 558, 559, 561–567, 572–574 quantifier, 113, 117, 119, 120, 124, 133, 178, 194, 375, 376<sup>2</sup> , 407, 408, 410, 411, 430, 431, 483 questions, 32, 113, 114, 116, 162, 163, 173– 175, 179, 212, 247, 251, 278, 285, 302, 304, 316, 387, 395, 409, 417, 422, 429, 462–464, 469, 471, 506, 518, 519, 527, 529, 532, 541–547, 549–552, 577–581 content, 429 embedded marker, 173, 303, 316, 429, 453, 454, 462, 471, 482, 551 tag, 289, 541, 547, 548 quotative, 453–456, 463, 464, 466, 468– 471, 473, 474, 482 recipient, 342, 369, 436, 517, 524, 530

reduplication, 43, 82–84, 222, 553 reflexivization local, 105, 106, 553, 554, 556, 560, 566 long-distance, 11, 553, 554, 562, 563 reported speech, 11, 73, 449, 450, 454, 455, 466–469, 471, 473–475 resultative, 222, 257–259, 261, 262, 264– 266, 282, 286, 308, 361, 385 spatial case, 11, 32, 37, 43, 54, 56, 63–67, 69, 70, 75, 83, 143, 145, 147, 149, 152, 153, 171, 187, 195, 197, 201, 217, 218, 221, 234, 314, 315, 342, 343, 346, 348, 352, 379, 395, 403, 405, 413, 414, 418, 419, 426, 559, 566, 569, 577–580 spatial postposition, 64, 75, 126, 143, 149, 217, 230, 234, 345, 353, 414 standard of comparison, 69, 128, 569– 571 stimulus, 56, 76, 261, 280, 342, 344, 353– 355, 357, 358, 374, 377, 393, 398, 418, 436, 477, 480, 564 stress, 19, 30, 176, 190 subjunctive, 169, 170, 297, 305, 317, 317<sup>2</sup> , 318, 319, 321, 323, 324, 330, 332, 391, 401, 434, 453, 454, 459– 462, 465, 474, 476, 477, 577– 579, 581 syllable structure, 19, 24, 31, 206 tense past, 39, 162–164, 173, 245, 252, 260, 265, 286–288, 291, 421, 547 present, 162, 285, 287, 291, 421, 490, 547 theme, 56, 394, 436, 524 valency class, 274, 302, 341, 342, 344, 369, 417–419, 431, 432, 556, 558 verb affective, 56, 227, 245, 246, 261, 274, 280, 295, 298, 318, 332, 333,

342–344, 346, 353–356, 356<sup>1</sup> , 357, 368, 369, 374, 386, 396, 418, 431, 435, 436, 479, 480, 532, 558, 559, 567 analytic form, 161–166, 187, 194, 222<sup>1</sup> , 249, 257, 271, 278, 286, 294, 308, 311, 332, 385 bivalent, 342–344, 352, 356, 396, 519 compound, 28, 33, 44, 60, 78, 80<sup>8</sup> , 120, 123, 152, 205, 206, 215, 228, 230–240, 293, 294, 341, 344, 345, 347, 348, 353, 356, 371, 375, 396, 426, 451, 474, 523 intransitive, 56, 62, 80, 227, 230, 234, 237, 243, 274, 295, 298, 318, 342, 344–349, 358, 363, 370, 374, 386, 391, 395, 417, 418, 426, 434, 436, 477, 479, 558, 559, 567, 573 labile, 297, 358, 359, 364, 365 light, 206, 228, 231, 232, 234–240, 349, 356, 371, 379, 450, 467 manipulative, 452 modal, 453 monovalent, 342, 346, 366, 519 periphrastic form, 249, 271–276, 285, 289<sup>1</sup> , 291, 293, 306, 308, 310, 326, 374, 385, 452, 453 phasal, 452 three-place, 341, 399 transitive, 80, 227, 232–234, 237, 243, 263, 274, 295, 296, 298, 303, 341, 344, 349–351, 354, 355, 358–363, 368, 374, 377, 396, 418, 419, 431, 432, 435, 468, 480, 483, 579 trivalent, 431, 460 two-place, 353, 357, 370, 391, 399, 400, 477 utterance, 450, 455, 466, 468, 469, 473, 485 vocative, 57 vowel deletion, 30, 31, 46–49, 51

vowel mutation, 29, 30, 33, 35, 36, 39, 46–48, 50

word order, 390, 430, 441, 481, 511, 526, 559, 567

# Did you like this book?

This book was brought to you for free

Please help us in providing free access to linguistic research worldwide. Visit http://www.langsci-press.org/donate to provide financial support or register as a community proofreader or typesetter at http://www.langsci-press.org/register.

## A grammar of Sanzhi Dargwa

Sanzhi Dargwa belongs to the Dargwa (Dargi) languages (ISO dar; Glottocode sanz1248) which form a subgroup of the East Caucasian (Nakh-Dagestanian) language family. Sanzhi Dargwa is spoken by approximately 250 speakers and is severely endangered. This book is the first comprehensive descriptive grammar of Sanzhi, written from a typological perspective. It treats all major levels of grammar (phonology, morphology, syntax) and also information structure. Sanzhi Dargwa is structurally similar to other East Caucasian languages, in particular Dargwa languages. It has a relatively large consonant inventory including pharyngeal and ejective consonants. Sanzhi morphology is concatenative and mainly suffixing. The language exhibits a mixture of dependent-marking in the form of a rich case inventory and head-marking in the form of verbal agreement. Nouns are divided into three genders. Verbal inflection conflates tense/aspect/ mood/evidentiality in a rich array of synthetic and analytic verb forms as well as participles, converbs, a masdar (verbal noun), and infinitive and some other forms used in analytic tenses and subordinate clauses. Salient traits of the grammar are two independently operating agreement systems: gender/number agreement and person agreement. Within the nominal domain, modifiers agree with the head nominal in gender/number. Agreement within the clausal domain is mainly controlled by the argument in the absolutive case. Person agreement operates only at the clausal level and according to the person hierarchy 1, 2 > 3. Sanzhi has ergative alignment in the form of gender/number agreement and ergative case marking. The most frequent word order at the clause level is SOV, though all other logically possible word orders are also attested. In subordinate clauses, word order is almost exclusively head-final.