# The Mehweb language

Essays on phonology, morphology and syntax

Edited by Michael Daniel Nina Dobrushina Dmitry Ganenkov

#### 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:


# The Mehweb language

Essays on phonology, morphology and syntax

Edited by Michael Daniel Nina Dobrushina Dmitry Ganenkov

Daniel, Michael, Nina Dobrushina & Dmitry Ganenkov (eds.). 2019. *The Mehweb language*: *Essays on phonology, morphology and syntax* (Languages of the Caucasus 1). Berlin: Language Science Press.

This title can be downloaded at: http://langsci-press.org/catalog/book/225 © 2019, the authors Published under the Creative Commons Attribution 4.0 Licence (CC BY 4.0): http://creativecommons.org/licenses/by/4.0/ ISBN: 978-3-96110-208-2 (Digital) 978-3-96110-209-9 (Hardcover)

DOI:10.5281/zenodo.3374730 Source code available from www.github.com/langsci/225 Collaborative reading: paperhive.org/documents/remote?type=langsci&id=225

Cover and concept of design: Ulrike Harbort Typesetting: Vadim Radionov Proofreading: Ahmet Bilal Özdemir, Andreas Hölzl, Bev Erasmus, Eva Schultze-Berndt, Havenol Schrenk, Ivica Jeđud, Jeroen van de Weijer, Jean Nitzke, Sebastian Nordhoff, Steven Kaye, Sune Rygård Tatiana Philippova Fonts: Linux Libertine, Libertinus Math, 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

## **Contents**


#### Contents


This volume presents several papers on Mehweb, a one-village language spoken in the central part of Daghestan, a republic of the Russian Federation. The language has a relatively low number of speakers (about 800) but is not immediately endangered, as shown in the first contribution by Nina Dobrushina, which is an introduction to the sociolinguistic situation of Mehweb. The contribution covers the geographical position of Mehweb and its economic situation, the official status of the language, the ethnic affiliation of the villagers, the recent history of Mehweb, its neighbours and the patterns of multilingualism observed. While there are no visible signs of first language loss, the paper shows that there is a strong tendency towards the loss of traditional patterns of multilingualism, with Russian replacing all other languages for interethnic communication.

Mehweb belongs to the Dargwa branch of the East Caucasian (Nakh-Daghestanian) language family. It is often considered as a dialect of Dargwa (Magometov 1982), along with many other lects within the Dargwa branch. A different tradition treats Mehweb as a separate language (Khajdakov 1985; Koryakov & Sumbatova 2007). The survey of Dargwa idioms in Sumbatova & Lander (2014) indicates that Mehweb is most often classified as belonging to the northern group of Dargwa dialects. Although the residents of Mehweb presently consider themselves to be the descendants of re-settlers from the village of Mugi, where the Akusha dialect of Dargwa is spoken (Uslar 1892; see also Dobrushina 2019 [this volume]), there is no linguistic analysis that shows any special affinity between Mehweb and Mugi. According to lexicostatistical analysis, Mehweb is a member of the north-central group of Dargwa and shows more similarities to Murego-Gubden than to Mugi (Koryakov 2013).

The first linguistic source on Mehweb is a reference by Uslar (1892). This short grammar describes another dialect of Dargwa, but starts with a brief survey of different Dargwa languages and dialects. Among these dialects Uslar also mentions Mehweb, qualifying it as a dialect spoken in Mugi, but "notably degraded". Two descriptions of Mehweb appeared in the 1980s, both in Russian. The first is a grammar of Mehweb which describes its phonology and morphology but not its syntax (Magometov 1982). This description, extremely clear and explicit, considers only the main morphological forms while excluding some less frequent ones,

and does not provide a detailed analysis of their semantics. The second, a book by Khaidakov, was written at almost the same time as Magometov's grammar. It compares the formal morphology of several Dargwa languages and dialects, including Mehweb.

In 1990, a field team from Moscow State University came to work on Mehweb, but no publications followed. In the aftermath of this trip, in the 2000s, Nina Sumbatova started to work on Mehweb and, among other things, compiled a list of glosses and suggested an analysis of Mehweb verbal inflection, some elements of which are integrated into this volume (Sumbatova manuscript).

The only dictionary of Mehweb which exists to date is a small vocabulary supplement in Magometov (1982). One of the aims of our study was to compile a dictionary and document the main inflectional forms. The dictionary is being developed by Michael Daniel with the participation of many members of the field team, especially George Moroz, and implemented as a web page by Aleksandra Kozhukhar. The current version of the dictionary is available online – https:// linghub.ru/mehwebdict/.

Mehweb texts were first published by Magometov (ibid.) with translation, but without morphological glossing. New texts were recorded and glossed during this project by Michael Daniel, including a sample of Pear Stories (currently transcribed but not yet glossed). The corpus includes 35 texts (including 13 from Magometov 1982) comprising about 1,000 sentences and 10,000 tokens and is also being prepared for open access.

The following brief overview is intended for the reader who is not familiar with East Caucasian languages. It provides background on the most important features of the language.

The consonant inventory includes voiced and voiceless consonants. Stops (but not other consonants) also have an ejective series. Unlike some other Dargwa dialects, Mehweb lacks phonologically distinctive geminate stops. The vocalic system has four members with a gap in the mid back position [i, e, a, u]; [oˤ] only appears as a realization of [u] with the pharyngeal feature. Velar, uvular and radical consonants may be labialized. In addition to [ʔ, ħ, h], Mehweb also has the less common [ʜ, ʡ] which seem to be phonologically secondary, appearing only as pharyngealized counterparts of [ħ, ʔ], respectively. Pharyngealization is strongly – but not exclusively – associated with uvulars, pharyngeal and laryngeal consonants. For further details on phonetic inventories and pharyngealization see the contribution by George Moroz, who discusses details of the inventory, syllable structure, stress placement, morphophonological alternations and pharyngealization.

Mehweb morphology is agglutinative. Mehweb is ergative in terms of both gender agreement and case marking. To start with the latter, the case inventory

includes the nominative (absolutive), the ergative, the genitive, the dative, the comitative and some peripheral case forms. Note that some authors of the volume follow Kibrik (1997) in the use of the term *nominative* for ergative alignment. Spatial forms are bimorphemic, as is typical of East Caucasian. The first category is that of localization, defining a spatial domain with respect to the ground (in Mehweb: 'on', 'near', 'at', 'in(side)', 'among'). The second category is that of orientation, defining the figure's motion with respect to this domain (Goal, Source, Path) or absence thereof (Static location). Unlike other branches of East Caucasian – but as in the other lects of the Dargwa branch – the lative form (Goal) is zero marked and the essive form (location) is marked by the presence of a gender agreement slot controlled by the nominative argument of the clause. The plural is expressed by a number of suffixes, sometimes accompanied by alternations. For more on nominal morphology, see the contribution by Ilya Chechuro, dealing with plural formation, the oblique stem, case formation and formation of irregular locatives. There is also a brief discussion of the use of the case forms.

Mehweb verb inflection is by and large similar to that of other Dargwa languages. It resides upon a fundamental distinction between two stems, perfective and imperfective, from which all other forms are derived. The formal relation between the stems is irregular and can involve alternations, infixation and loss of agreement slots. Most forms are derived from both perfective and imperfective stems, except the prohibitive and the present/habitual, which are only available in the imperfective. The combination of the irregular relation between perfective and imperfective stems and the almost perfectly parallel inflection based on the two stems partly assimilates the Mehweb (and generally Dargwa) aspectual system to that of derivational aspect. Irregular verbs include verbs of motion, the verb 'give', the verb 'say' and some others. For more on verbal morphology, see the contribution by Michael Daniel.

Zooming in on one fragment of the verb morphology, Nina Dobrushina provides a detailed analysis of both form and meaning in the irrealis domain. Several features are typologically infrequent, although common for East Caucasian languages: the formal split between transitive and intransitive imperatives, the expression of the negative imperative by a dedicated inflectional form (the prohibitive), and the presence of a dedicated inflectional optative used in blessings and curses. The presence of a dedicated apprehensive is rare even within East Caucasian. The jussive and the hortative are expressed periphrastically. A detailed analysis of another fragment of verbal morphosyntax is provided in the contribution by Daria Barylnikova. She provides a survey of periphrastic constructions based on 'drive' and 'let' and explains the ways in which these constructions show incipient signs of grammaticalization into expressions of factitive and permissive causation, respectively.

Gender agreement in Mehweb follows strict semantic assignment: in the vast majority of cases, it is enough to know the semantics of the noun to determine its agreement pattern. Mehweb gender (class) agreement distinguishes masculine, feminine and neuter in the singular and human and non-human in the plural. One complication is connected to mass nouns; although morphologically singular (and capable of forming morphological plurals), these nouns control nonhuman plural agreement. While this behavior of mass nouns is typical of Dargwa languages, the next twist is an innovation and probably results from contact with Lak. The majority of feminine nouns have moved from the original Dargwa feminine (*r-*, glossed as f in the book) gender to a gender identical to non-human plural (*d-*, glossed as f1). The distribution is roughly between married/old (f) and unmarried/young (f1) women. The choice between the two agreement patterns is still partly flexible and may become a tool of language game or insults. One could speculate that the source of this development is some kind of indirect reference motivated by politeness. Another development in agreement is that personal agreement on the verb, well attested in Dargwa languages, developed into the typologically rare phenomenon of egophoric agreement; the suffix *-ra* (glossed ego) appears with first person subject in the affirmative and with second person subject in the interrogative. Unlike gender, personal agreement works on an accusative rather than an ergative basis.

Clause subordination is based on dependent verb forms, including action nominals, infinitives, participles and converbs, rather than on finite predication introduced by conjunctions. Converbs include two general (contextual) converbs (perfective and imperfective) whose relation to the main clause is context-determined and several special converbs that specify this relation (in Mehweb, immediate anteriority, gradual accumulation, cause, concession etc. – see the contribution by Maria Sheyanova). Some aspects of the syntax of general converbs are presented in the contribution by Marina Kustova, who covers periphrastic converbs, independent uses of converbs and their use in imperative contexts, and different strategies for how the converb clause may share its arguments with the main clause. In the absence of true clause co-ordination, the respective discourse/narrative function is performed by chains of general converbs. Kustova's contribution attempts to address this issue by considering several tests targeting the subordination – co-ordination distinction.

One apparent exception to the non-use of finite predication in subordination is constituted by reported speech constructions. Reported speech in Mehweb, as generally in East Caucasian, is structurally similar to direct reporting and typologically distant from true subordination. Mehweb has a pronominal stem *sa*‹cl›*i*, used with a wide range of functions, from logophoric function in reported

speech to resumptive to reflexive, considered in the contribution by Aleksandra Kozhukhar. The author suggests that, in Mehweb, there is neither a morphological nor a (sharp) syntactic distinction between logophoric and long-distance uses of the pronoun.

The three other contributions on syntax are the chapters by Dmitry Ganenkov (syntax – case assignment and personal agreement – of the simple clause), Yuri Lander and Aleksandra Kozhukhar (the relative clause) and Yuri Lander (a survey of the uses of the focus particle). Ganenkov shows how the distribution of personal and gender agreement control classifies Mehweb verbs into several morphosyntactic classes, non-trivially connected to their transitivity, and demonstrates how this distribution is linked to conventional subject properties such as control of reflexivization. Lander and Kozhukhar argue that the use of the reflexive pronoun has been specialized as resumptive in relative clauses, taking as evidence the restrictions on its use as compared to the use of simple reflexives. Finally, Lander argues that the focus particle *gʷa*, formally identical to the imperative of 'see', surprisingly does not have to be adjacent to the constituent in the scope of the focus.

\* \* \*

This volume is the result of a collective field research project run by the linguists from the School of Linguistics of HSE University, Moscow. Part of the team consisted of bachelor's students who conducted their research under the supervision of the more experienced members of the team. Collective field research is a practice developed by Aleksandr Kibrik, an eminent Russian typologist who organized more than 40 field trips attracting hundreds of young people to the description of minority languages. Kibrik edited numerous grammars where chapters were contributed by all team participants, including students in their early years at university.

In 1990, Aleksandr Kibrik brought to Mehweb a large group of students which included, among others, Michael Daniel and Nina Dobrushina. This specific field trip produced relatively little in terms of scholarly output, the most important result being a three-page sketch of Mehweb morphology (a list of the major forms and morphemes) by Nina Sumbatova.

The more important legacy of the 1990 expedition was a personal/human one. Anvar Musaev and Maisarat Muslimova (now Musaeva), two teachers at the local school, took an active part in the organization of the life of the expedition. A long-lasting human bond was established with them. In 2010, Michael Daniel and Nina Dobrushina decided to pass by Mehweb on their way from Archib to Makhachkala. Anvar and Maisarat, this time a married couple with grown-up

children, were so open and hospitable, and so full of memories of the 1990 visit, that the idea of working on Mehweb came very naturally. In 2013, five students from the Higher School of Economics accompanied by Michael Daniel, Nina Dobrushina, Dmitry Ganenkov, Yuri Lander and George Moroz came to Mehweb to start working on a description of the language. In the course of four field trips in 2013, 2014, 2015 and 2016, each lasting about two weeks, we recorded texts, compiled a small dictionary, and wrote several papers. The student team was not always the same. Some of the students involved did not participate directly in this volume, but they all made a contribution to the analysis of the data. It is thus our pleasure to list the participants of all field trips over these four years: Ekaterina Ageeva, Darya Barylnikova, Ilya Chechuro, Violetta Ivanova, Aleksandra Khadzhijskaya, Aleksandra Kozhukhar, Marina Kustova, Yevgeniy Mozhaev, Olga Shapovalova, Semen Sheshenin, Aleksandra Sheshenina, Mariya Sheyanova.

Anvar and Maisarat and their family invariably provided us with housing and logistical support and never grew tired of being our primary native consultants, including over email, Skype and now, in the final days of our work on the book, also over WhatsApp, a very useful tool for instant proofreading of examples. We are also infinitely grateful to our friends and consultants Abakar and Zalmu Sharbuzovy, to their daughters Patimat and Kamila, so intelligent and helpful, to the indefatigable Kazim, foe of all tea parties, his wife Munira and his sister Bulbul; to Patimat Tagirovna, who deserves to become the first announcer on Mehweb radio, if it is ever established; to Khavsarat, Magomedzagid, Mariam and many other Mehweb people the limits of whose patience we have been stretching for too many years. We remember the touch of the hand of Aminat, Maisarat's mother.

The authors are very grateful to Samira (Helena) Verhees who proofread most of the papers presented here, to our very patient type-setter, Vadim Radionov, and to the reviewers of drafts of individual chapters of the volume: Aleksandr Arkhipov, Gilles Authier, Oleg Belyaev, Denis Creissels, Francesca Di Garbo, Diana Forker, Martin Haspelmath, Olesya Khanina, Timur Maisak, Nina Sumbatova, Yakov Testelets, as well as to the anonymous reviewers of Language Science Press.

This volume was prepared within the framework of the Academic Fund Program at the National Research University Higher School of Economics (HSE) in 2015–2016 (grant #15-05-0021) and by the Russian Academic Excellence Project «5-100».

*Michael Daniel and Nina Dobrushina*

REFERENCES

### **References**


## **Предисловие**

Настоящий сборник – результат многолетней работы исследовательской группы Школы лингвистики Национального исследовательского университета Высшая школа экономики. В проекте приняли участие студенты бакалавриата, которыми руководили более опытные исследователи. Коллективная полевая работа – практика, введенная Александром Евгеньевичем Кибриком, выдающимся советским российским типологом, организовавшим более сорока лингвистических экспедиций, в ходе которых в полевых исследованиях малых языков приняли участие сотни студентов. А.Е. Кибрик выпустил большое число грамматик, главы которых писались в том числе студентами, лишь недавно начавшими учебу в университете.

В 1990 г. А.Е. Кибрик привез в селение Мегеб (Гунибский район Республики Дагестан) большую группу студентов, участниками которой были, в том числе, М. Даниэль и Н. Добрушина. От этой поездки сохранилось не так много материалов. Важным результатом стал краткий обзор мегебской морфологии (список основных форм и морфем), составленный Н. Сумбатовой.

С точки зрения человеческих отношений самым главным приобретением экспедиции 1990 г. стало знакомство с Анваром Мусаевым и Майсарат Муслимовой (ныне Мусаевой), молодыми учителями мегебской школы, которые приняли активное участие в жизни экспедиции. В 2010 г. мы (Н. Добрушина и М. Даниэль) решили заехать в Мегеб на обратной дороге из Чародинского района в Махачкалу. Майсарат и Анвар, к этому времени – семейная пара с двумя взрослыми детьми, приняли нас настолько радостно и тепло, были так полны воспоминаниями о той давней поездке, что идея возобновить работу над мегебским языком показалась совершенно естественной и даже неизбежной. В 2013 г. пять студентов ВШЭ под руководством М. Даниэля, Н. Добрушиной, Д. Ганенкова, Ю. Ландера и Г. Мороза приехали в Мегеб для работы над грамматикой этого языка. В результате четырех поездок (2013–2016 гг.), каждая продолжительностью около двух недель, мы записали некоторое количество текстов, собрали небольшой словарь и написали несколько черновых статей. Студенческий состав

#### Предисловие

не оставался постоянным. Некоторые из участников этих экспедиций не приняли участие в написании настоящего очерка, но каждый из них внес тот или иной вклад в сбор и анализ данных. Мы приводим полный список участников всех экспедиций: Екатерина Агеева, Дарья Барыльникова, Виолетта Иванова, Александра Кожухарь, Марина Кустова, Евгений Можаев, Александра Хаджийская, Илья Чечуро, Ольга Шаповалова, Семен Шешенин, Александра Шешенина, Мария Шеянова.

Нашими неизменными хозяевами и главными переводчиками были Майсарат и Анвар. Они и их сыновья обустраивали нашу жизнь и неутомимо отвечали на наши вопросы о мегебском языке, в том числе по электронной почте, скайпу, а в последнее время – по WhatsApp'у, совершенно незаменимому инструменту для того, чтобы в последний момент вносить правку в корректуру статей по малым языкам. Кроме того, мы бесконечно благодарны нашим друзьям и переводчикам – Абакару и Залму Шарбузовым, их дочерям Патимат и Камиле, таким умным и всегда готовым поделиться своим временем, неутомимому чаененавистнику Казиму, его жене Мунире и его сестре Булбул; Патимат Тагировне, которая несомненно заслуживала бы роли первого диктора мегебского радио, если таковое когда-нибудь начнет вещание; Исрапилу, Кавсарат, Магомедзагиду, Марьям, Саиде и многим другим мегебцам, границы терпения которых мы испытывали в течение стольких лет. Всем им мы желаем долгих лет жизни и здоровья.

Мы помним рукопожатие Аминат, мамы Майсарат, Муниры и Марьям.

Авторы сборника очень признательны Самире (Хелене) Ферхеес, которая вычитала многие из статей, Вадиму Радионову, который взял на себя сложную верстку тома, рецензентам первых версий статей – Александру Архипову, Жилю Отье, Олегу Беляеву, Дени Кресселю, Франческе Ди Гарбо, Диане Форкер, Мартину Хаспельмату, Олесе Ханиной, Тимуру Майсаку, Нине Сумбатовой, Якову Тестельцу, а также анонимным рецензентам издательства Language Science Press.

Сборник был подготовлен в ходе проведения исследования № 15-05-0021 в рамках Программы «Научный фонд Национального исследовательского университета "Высшая школа экономики" (НИУ ВШЭ)» в 2015–2016 гг. и в рамках государственной поддержки ведущих университетов Российской Федерации «5-100».

*Михаил Даниэль и Нина Добрушина*

### **List of abbreviations**




## **Chapter 1**

## **The language and people of Mehweb**

### Nina Dobrushina

National Research University Higher School of Economics, Linguistic Convergence Laboratory, ndobrushina@hse.ru

This paper describes the sociolinguistic situation of Mehweb, a lect of the Dargwa branch of East Caucasian, spoken in the Republic of Daghestan. In the course of several field trips to the village of Mehweb (officially, Megeb), sociolinguistic interviews were conducted in Mehweb and four neighbouring Avar- and Lak-speaking villages. The paper describes the demographic situation in Mehweb, the villagers' official status, their social and economic life in the past and at present. The multilingual repertoire of Mehwebs and their neighbours is described in both qualitative and quantitative terms. I conclude that, while there are no signs of language loss, the traditional patterns of multilingualism in Mehweb are highly endangered.

### **1 Introduction**

Mehweb belongs to the Dargwa group of the East Caucasian (Nakh-Daghestanian) language family. It is sometimes considered a dialect of Dargwa (Magometov 1982), but more often, it is treated as a separate language (Khajdakov 1985; Koryakov & Sumbatova 2007). Mehweb is spoken in a single village called Mehweb<sup>1</sup> and geographically separated from all other Dargwa languages. While Dargwa languages generally constitute a continuous area, Mehweb is surrounded by speakers of Avar and Lak, which are languages of other branches of the family.

The village of Mehweb is located in Gunibskij region, in the central part of Daghestan at 1800 meters above sea level. The total number of speakers is estimated to be between 800 and 900, 600 to 700 of whom live in Mehweb itself.

<sup>1</sup>Russian *Мегеб* – [megeb], the native term is [mehʷe], while [mehʷeb] is the Avar spelling which includes the final *-b* of the locative form.

Nina Dobrushina. 2019. The language and people of Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 1–15. Berlin: Language Science Press. DOI:10.5281/zenodo.3402054

#### Nina Dobrushina

About 100 to 200 live in the so called *kutan*<sup>2</sup> *kolkhoza imeni Gadzhieva* (located 350 km away from Mehweb and four kilometres away from the sea coast, near the village Krainovka). Kutan was not examined from either a linguistic or sociolinguistic point of view. All data in this paper come only from Mehweb. There are also Mehweb families in Makhachkala, Kizlyar and Bujnaksk, and a few elsewhere. All Mehweb-speaking families originate from the village Mehweb.

Like most Daghestanians, Mehwebs are Muslim. Mehweb has no literacy tradition. The Mehwebs write in Avar or Russian. We have no evidence that Mehweb was ever written in the Arabic or Cyrillic script in the observable past. At least, the residents of Mehweb could not recall any manuscripts in Mehweb (unlike some other minority languages of Daghestan – see Magomedkhanov 2009 about the Archi manuscript).

So far, there are no indications of language loss in Mehweb. All villagers speak Mehweb, and Mehweb is the first language acquired by children.

The Mehwebs often suggest that their idiom is more conservative than other Dargwa lects and contains some archaic features. This opinion is also expressed in some descriptions of Mehweb (Magometov 1982; Khajdakov 1985). Recent studies on Dargwa languages show that at least some phenomena (such as various properties of agreement) are innovative in Mehweb compared to other Dargwa lects (Sumbatova & Lander 2014).

The command of Russian, Avar, and Lak is spread in Mehweb (see §5 for details). The proficiency in standard Dargwa is infrequent. In §2, the official status of the Mehweb language is discussed. §3 and §4 briefly describe social and economic life of the village in the past and at the present time. §5 is devoted to the multilingual repertoires of Mehwebs and the neighbouring villages. A brief conclusion summarizes the paper.

### **2 Mehweb officially**

Mehweb is located in a district where Avars are numerically dominant. As a result, Mehwebs are in some respects considered to be Avars (Tishkov & Kisriev 2007: 98).

Firstly, paradoxically, they are taught Avar at school during lessons called *native language* (Russian *родной язык*, lit. 'native language'), even though Avar belongs to another group of East Caucasian and is genealogically distant from

<sup>2</sup>Originally, kutans were territories for lowland herding in the winter. At the present time, people often prefer to stay in these lowland settlements for the whole year, thus establishing new villages.

#### 1 The language and people of Mehweb

Mehweb. Mehweb children begin learning two foreign languages in the first grade (at 6–7 years old) – Avar and Russian, which, according to their parents, is not easy for them. Another result of learning only Avar at school is that Mehwebs are not acquainted with standard Dargwa, unlike most people who speak other lects of Dargwa.

Secondly, most Mehwebs are registered as Avars in their passports. That continued until the 1990s, when the obligatory indication of ethnicity in passports was cancelled in Russia. The villagers explain that those Mehwebs who got their passports at the village council were registered as Avars, while those who got their passports in the cities were registered as Dargis.

In the 2002 and 2010 censuses of the Russian Federation, Mehwebs were not mentioned. Residents of Mehweb were registered as Dargis or as Avars. In 2002, 747 Dargis and 98 Avars were reported as residents of Mehweb. In 2010 the numbers were 572 Dargis and 124 Avars. The difference between the data of the two censuses has no reasonable explanation. Mehweb is very homogenous both ethnically and linguistically, as are most villages of highland Daghestan. There are no outsiders in the village except for several Avar women taken as wives. Most probably, the ethnic population of Mehweb has not changed in at least the last hundred years, and the census information does not reflect the true ethnic structure of Mehweb in any way.

According to interviews with the villagers, Mehweb residents identify themselves as Dargis. They are well aware of the closeness of their language to Dargwa, and have regular contacts with the Dargwa people from the village Mugi (see §3).

Data from the censuses on native language are again controversial. The Mehweb language is not mentioned. It follows from the 2002 census that 792 residents indicated Dargwa as their first language, while 53 indicated Avar. According to the 2007 census, this has changed: 566 indicated Dargwa as their first language, and 113, Avar. The mention of Dargwa as a first language is most likely because Mehweb is usually considered a variation of Dargwa, and therefore the residents of Mehweb may have referred to their native language as Dargwa. But there are no reasonable explanations for the mention of Avar as a first language: there is no one in Mehweb who speaks Avar as a first language, apart from the two or three women who married in.

Mehwebs are not officially recognised as an ethnic group, nor is Mehweb officially recognised as a language.

Nina Dobrushina

### **3 The past of Mehweb**

There is a common belief that the village of Mehweb was founded by re-settlers from the Dargwa-speaking village of Mugi (Uslar 1892). Mugi is located in the Akushinskij district (in the central part of Daghestan, about 70 km from Mehweb; it takes two to three hours by car). As far as I know, there is no tangible historical evidence for the connections between Mehweb and Mugi, apart from oral testimony. Mehwebs are convinced that Mugi is their ancestral homeland, and have several versions of how they left it. One of the local stories reports that there was an isolated part of Mugi which was in the way of Timur's (Tamerlane's) army. When they realized they could not resist Tamerlane, the residents fled and settled higher in the mountains. According to this version, Mehweb was founded in the 14th century. Khajdakov (1985: 101) dates the migration of Mehwebs to somewhere between the 8th and 9th centuries, reporting the opinion of a respected Mehweb resident. An early report by Komarov says that the Mehweb people are (descendants of) refugees from another village, but Mugi is not mentioned (Komarov 1868) 3 .

According to lexicostatistical analysis, Mehweb belongs to the Northern-central group of Dargwa languages, and is closer to Murego-Gubden lects than to the dialect of Mugi (Koryakov 2013).

Although it is not clear if this view on the origin of Mehwebs has historical grounds, the residents of Mehweb and Mugi are quite positive. They have established intensive contacts: they practice reciprocal group visits, and invite each other to important festivities. Most of the Mehwebs I had spoken to said they did not understand the dialect of Mugi and preferred communicating with the Mugis in Russian.

The relations of the Mehwebs with Avars were much more intensive. The main road to Mehweb was through a big Avar village, Chokh, and through another, smaller Avar village, Obokh. In the 19th century, Mehweb was a part of the socalled Andalal free association which mainly consisted of Avar villages. After the revolution of 1917, Mehweb became a part of the Charoda district. In 1928, it was transferred to the Gunib district. Both districts are dominated by Avars. Between 1929 and 1934, it was transferred to the Lak district, and then was re-transferred to Gunib. Therefore, from the administrative point of view, the Mehwebs were mostly connected with Avars.

<sup>3</sup>«Недалеко от Чоха есть большое селение Меге, по преданию, основанное даргинцами,

в разное время искавшими спасения от кровомщения».

#### 1 The language and people of Mehweb

Avars were, and still are, the closest neighbours of the Mehwebs – it takes about 40 minutes to walk to Obokh. Although the more distant Lak neighbours were also important for Mehweb, because the Mehwebs would regularly go to the Kumukh market where the communication was in Lak. The distance is about 15 kilometres from Mehweb to Kumukh, taking four to five hours to get there by foot. Some women would go there every Thursday. Visits to the market in Kumukh gradually became less frequent after the 1950–60s.

Mehweb was and still is one of the biggest villages in the neighbourhood. This number has remained stable over the 20th century: 710 in 1926, 780 in 2007.

The main occupation of Mehwebs was breeding sheep and cattle. They also grew corn and potatoes. The specialty of Mehweb was cultivating black peas which usually yielded a good harvest. There were no fruit trees before the 1950s, although at the present moment Mehwebs grow apples, pears and apricots. Mehwebs were neither rich nor poor in comparison to other settlements of highland Daghestan.

As Mehwebs had enough corn, they did not need to look for jobs outside the village. According to the recollections of local people, seasonal employment outside the village was not customary. Only a few Mehweb people are reported to have practiced tinsmithing, like their Lak neighbours. We were also told by the residents of the neighbouring village of Shangoda that Mehwebs were good stone masons and builders, and were invited to other villages. Another reason for inter-ethnic contact was shepherding on remote pastures (transhumance), which resulted in irregular contact with Avars and Kumyks. In general, the Mehweb people did not migrate a lot.

Mehweb people rarely married out. As in all of Daghestan (Comrie 2008; Wixman 1980), a marriage partner from Mehweb was preferable. Often the spouse was chosen from the same patrilineal clan (*tukhum*). In the infrequent cases of mixed marriages the wife was taken from one of the neighbouring Avar villages. The tradition of endogamic marriages started to die away only in the beginning of the 21th century.

### **4 The present of Mehweb**

Today, Mehweb has between 600 and 700 residents. The population has not decreased as much as in many other neighbouring villages. For example, the Avar villages Obokh and Shangoda were twice as populated in the past. The Lak villages Mukar and Uri are on the verge of complete abandonment. However, several families still live in the Lak villages Palisma and Kamakhal, which were re-

#### Nina Dobrushina

cently among the biggest in the neighbourhood. The Avar village of Shitlib (Shitli) has been abandoned. The only village in the neighbourhood which did not lose a significant part of its population, apart from Mehweb, is the Avar village Bukhty. Mehweb is the biggest and the most vital village in the vicinity, with a large school and a sizeable population of children. Still, the locals report a slight population decrease: the school had more pupils in the 1980s than now.

Apart from the regular school, Mehweb has a special boarding school for training boys in freestyle wrestling. There are usually about 10–15 boys from other parts of Daghestan who live in Mehweb and study with local children. These boys have different native languages (most often Avar) and communicate with the locals in Russian. There is a kindergarten where local teachers communicate with the children in Mehweb and in Russian. The village boasts a large social centre. It has a billiard room and, on occasions, hosts concerts and dances. A small medical centre employs three nurses.

As elsewhere in Daghestan, the Mehwebs complain about local unemployment. Those who are not employed at the school, kindergarten, social centre or the medical centre, can make their living only by going away for construction jobs, or by selling meat and cheese. There are also several small shops run by local families.

People in Mehweb, as in all other villages in the neighbourhood, have had TV since the 1980s. Regular access to broadcasts became possible from the 1990s when a transmission tower was constructed in Sogratl. The broadcasts are mainly in Russian. Apparently this has influenced the level of bilingualism in Russian.

The Mehwebs take pride in the fact that several of its residents distinguished themselves during WW2. Two men were decorated as Hero of the Soviet Union for their military service during the war. Mehweb has a war memorial, and Victory Day (May 9) is also of special importance to the village.

### **5 Neighbours and language contact**

The level of multilingualism was studied in Mehweb and in four neighbouring villages: Obokh and Shangoda (Avar) and Uri and Mukar (Lak) – see Figure 1. During fieldtrips in 2012–2015, a series of sociolinguistic surveys was conducted to study the multilingual repertoires of the residents<sup>4</sup> .

<sup>4</sup> Sociolinguistic study of multilingualism in Mehweb and neighbouring villages is a part of a larger project documenting patterns of multilingualism in Daghestan (https://multidagestan. com/).

#### 1 The language and people of Mehweb

Figure 1: Mehweb and neighbouring villages (map courtesy of Yuri Koryakov)

#### **5.1 Data and methodology**

In order to obtain quantitative data about the command of other languages in each of these villages, the method of retrospective family interviews (introduced in Dobrushina 2013) was applied<sup>5</sup> . The dynamics of multilingualism is accessed through, and based on, short interviews with speakers of different generations, thus resembling apparent time studies<sup>6</sup> (Cukor‐Avila & Bailey 2013). The impor-

<sup>5</sup> I am deeply grateful to Darya Barylnikova, Ilya Chechuro, Michael Daniel, Violetta Ivanova, Aleksandra Khadzhijskaya, Marina Korshak, Aleksandra Kozhukhar, Marina Kustova, Yevgenij Mozhaev, Olga Shapovalova, Marija Sheyanova, Semen Sheshenin, Aleksandra Sheshenina who ran the interviews on multilingualism in Mehweb together with me.

<sup>6</sup>Apparent time studies of language change use surveys of people of different ages, with an assumption that the speaker's speech reflects the speech patterns acquired in the childhood.

#### Nina Dobrushina

tant difference from the apparent time method is that data are obtained not only about the respondents themselves, but also about their deceased relatives.

The method aims at capturing multilingual repertoires of the speakers of the recoverable past in order to reconstruct traditional (i.e. pre-Soviet) patterns of language contact. It was typical for highland Daghestani to have large families where parents lived together with their youngest son and communicated with other children on a daily basis, looked after their grandchildren and helped to run the household. The younger generation was usually well acquainted with their grandparents. By asking 60 to 80 year old villagers about language repertoires of their grandparents, the data collected sometimes dates back to the end of the 19th century, and even to the mid-19th century. Table 1 provides an example of a questionnaire completed for one person.


Table 1: Example of a filled in sociolinguistic questionnaire

The choice of respondents was more or less random. The aim of the study is to reconstruct the multilingualism of the past; so the eldest possible respondents were preferred, and younger generations were included for the sake of comparison. The controlled parameters of the sample were thus the respondents' age and gender.

The shortcomings of this method include, first of all, the subjective character of judgments about language proficiency. No test of proficiency of the respondent was undertaken (and obviously no such test was possible for his or

#### 1 The language and people of Mehweb

her late relatives). Estimations of the level of bilingualism were based on the respondents' judgments. The second shortcoming is the fact that the respondent's memories of e.g. his mother and father were limited to their older period of life. Third and probably most importantly, judgments may reflect stereotyped notions about past multilingualism widespread in the village, rather than being based on personal memories of individual linguistic repertoires. For a further discussion, see Dobrushina (2013).

Multilingualism is a social behaviour developed through interaction. Hence sociolinguistic surveys were conducted not only in the village of Mehweb but also in four neighbouring villages. The data from retrospective family interviews in neighbouring villages helped us to better understand how the communication between neighbouring villages was performed. Were both languages used for communication or was one of them preferred? For example, if we only found that most Mehwebs spoke Avar and Lak, we still would not know whether Avar and Lak neighbours of Mehwebs could speak Mehweb or not, and therefore could not estimate the role of the Mehweb language in the area.

The closest neighbours of Mehweb are the Avar villages Obokh and Shangoda (Figure 1).

Obokh villagers speak a dialect of Avar. In their opinion, this variety differs from the dialects of other villages in the area. At school, the Obokhs learn standard Avar. There is an opinion among them that their village is the oldest in the neighbourhood. They support this idea by the size of the cemetery. Another fact which might prove that Mehweb is younger than Obokh is that Obokh possesses more land than Mehweb, although the village itself is smaller.

Shangoda, another Avar village, is further away from Mehweb than Obokh. The track goes up and down, and it takes about 90 minutes to reach Shangoda. Slightly closer than Shangoda was the Avar village Shitlib, which is now abandoned. After Shangoda, there are what were earlier the Lak villages Palisma and Kamakhal (about 30 minutes walk from Shangoda). They are also abandoned now. In the 19th century and at the beginning of the 20th century, Shangoda belonged to the Kazikumukh district, dominated by Laks. It was connected to Kumukh by a mountain path. Until the 1930s, when Shangoda was transferred to the Gunib district, the inhabitants of Shangoda had their administrative centre in the village of Palisma. Therefore, relations with Laks were more important for Shangoda than relations with Mehwebs or with Avar villages.

Lak villages are further away from Mehweb than Obokh or Shangoda, but the contacts with them were essential for Mehwebs because of their regular visits to the Lak market in Kumukh. In Lak villages, the Mehweb people had friends with whom they could stay on their way to the Kumukh market.

#### Nina Dobrushina

All five villages (Mehweb, Obokh, Shangoda, Mukar, Uri) are located at more or less the same height above sea level (1500–1800 meters). In the observable past, the economic life and the standards of living in all these villages were similar.

In Mehweb, the sociolinguistic survey was the most extensive. Our database contains 240 entries, including 90 people who are deceased. The databases for other villages have less entries: 80 in Shangoda, 80 in Uri, 103 in Obokh, 110 in Mukar (note that these villages are presently much less populated than Mehweb).

People were divided into two groups: those who were born before and those who were born after 1919. The reason for establishing 1919 as a cut-off point was that in the 1930s, Soviet schools were opened in all villages. The teaching was done in Russian. The generation born after 1919 therefore usually had a secular education, often had some level of literacy, had less opportunities to learn Arabic script (because of the atheistic politics of the USSR), and most often spoke some Russian. The generation born before 1919 was closer to what we consider traditional patterns of multilingualism, as will be shown in the next section.

#### **5.2 Multilingualism among the residents born before 1919**

According to our study, Mehwebs communicated with the Avars and Laks in Avar and Lak respectively. This follows from the level of mutual bilingualism of the Mehwebs and their neighbours. Almost 100% of Mehwebs born before 1919 spoke Avar and Lak (see Table 2). Their neighbours from Avar and Lak villages had no command of Mehweb at all. Only 8% of the people from Obokh, the closest Avar village, were reported to speak Mehweb (Table 2).

Mehwebs acquired Avar through their communication with the neighbouring Avar villages, Obokh and Shangoda, and bigger villages which were more distant but important economically and socially, including Sogratl, Chokh, and Gunib. There were no Lak villages located as close as Obokh and Shangoda to Mehweb and the main source of knowledge of Lak was the market in Kumukh. The role of this market in the area was important enough for the Mehwebs to acquire Lak.

Occasionally, Mehwebs also mentioned their command of Kumyk. Kumyk was acquired by those who brought sheep to the lowlands where Kumyks lived. This practice was apparently not very common – only 2–3% of the people born before 1919 spoke Kumyk.

About 45–50% of the Mehwebs born before 1919 could read the Quran<sup>7</sup> . Note that the reported ability to read does not imply ability to understand Arabic, but only to recite the text. The knowledge of Arabic was usually limited to the knowledge of the phonetic meaning of letters. If a person was reported to be able

<sup>7</sup> See also Kozhukhar & Barylnikova (2013) about the dynamics of literacy in Mehweb.

#### 1 The language and people of Mehweb

to read Arabic, the researchers asked more specific questions about the ability to translate (understand) Arabic text. According to our study, only 6% of Mehwebs could understand and translate the Quran.

About 20% of Mehwebs of this generation spoke Russian. The command of Russian was much more common among men who travelled in order to earn money.

As for the residents of Avar villages, the knowledge of Lak was reported significantly more often in Shangoda (93%), than in Obokh (22%). This is not surprising. Lak villages were very close to Shangoda (30 minutes walk), and the residents of Shangoda and the Lak villages were socially and economically connected. For both Shangoda and Obokh, the market in Kumukh was very important, but Kumukh was much closer to Shangoda. There was a striking difference between Obokh and Mehweb. The villages were almost at the same walking distance from Lak villages, but the difference in the level of Lak is huge: 95% in Mehweb and 22% in Obokh. There is only one plausible explanation for this discrepancy. Mehwebs as speakers of a minor language were disposed to speaking other languages, while Avars, being the majority in the district, were in general oriented to use their own language in all circumstances.

The residents of Lak villages also had some command of Avar, but the level of their bilingualism was lower than in Avar villages (Table 2).


Table 2: The level of multilingualism in five villages: generations born before 1919

Mehwebs were the most multilingual people of the villages in the area. The language contact between Mehwebs and their neighbours was asymmetrical. They spoke the languages of their neighbours, while the neighbours did not speak Mehweb. Presumably, Mehweb was never used as a second language (we cannot be positive about this because we have no information about the more distant past). The reason for this asymmetry in the linguistic relations between neighbours was obviously the fact that Mehweb was spoken only in one village and had no importance at the supralocal level.

#### Nina Dobrushina

#### **5.3 Multilingualism among the residents born after 1920**

In the second half of the 20th century, knowledge of local languages decreased, while knowledge of Russian increased significantly. People in Mehweb and Obokh spoke virtually no Lak (Table 3). In Shangoda, the command of Lak persisted longer, but it was almost lost in the generation born after 1960. The command of Avar in Lak villages Uri and Mukar was also practically lost.

Table 3: The level of multilingualism in the generation born after 1920


There are several factors which triggered the drastic changes in local multilingualism. The first reason is the spread of Russian as a lingua franca across Daghestan. The command of Russian substituted local bilingualism. Secondly, the relations within the neighbourhood started to lose their economic significance, being substituted by connections with bigger towns. At present, the Mehwebs prefer shopping in Makhachkala rather than in Kumukh. Villagers also ceased cultivating fields, the borders with the neighbours have lost their significance, and communication became rarer.

There are rare cases of some Obokhs speaking Mehweb among those born in the 1960s. This is because, until the 2000s, there was no senior school in Obokh, and some children continued their education in Mehweb. Several people reported the ability to understand Mehweb, acquired during their school years.

In Mehweb, people born after the 1950s speak almost no Lak, but the command of Avar is still very high. Avar was supported by schooling and communication with neighbours and with the Avar administration. Mehwebs born after 1990, however, do not speak Avar. This might be a manifestation of the same process of the loss of local multilingualism as in other villages, but it could also be a pattern of age-based multilingualism, whereby a neighbouring language is acquired when people start to work. In the latter case, this generation will speak Avar after their professional socialization, at the age of 30–40. Only later research will show what pattern the now young Mehwebs will follow.

#### 1 The language and people of Mehweb

Some Mehwebs reported a command of the Akusha dialect of Dargwa. In the 1950–1970s, Mehweb did not have enough shepherds, and the Dargis from the Akusha district worked in the Mehweb kolkhoz as sheepherders. The Mehwebs remember communicating with these shepherds and with their wives, who came to see their husbands when they returned to Mehweb with the sheep. As a result, some of the Mehwebs acquired the Akusha dialect, which is otherwise not intelligible to them.

Another change concerned literacy. The atheistic politics of the USSR resulted in a dramatic loss of Arabic literacy. Only 5% of Mehwebs born after 1920 knew the Arabic script (as compared to the 48% in the generation born before 1919). A similar change happened in other villages. At the same time, most villagers became literate in Cyrillic and could read and write Russian and Avar.

### **6 Summary**

Mehweb is a minority language, spoken in only one village. As mentioned in the introduction, there are no signs of language shift in Mehweb. In the village, everybody speaks Mehweb, and since the 19th century the number of speakers has not decreased. There is, however, a strong tendency towards the loss of traditional patterns of multilingualism. Over the 20th century, knowledge of neighbouring languages in highland villages was substituted by knowledge of Russian, because Russian spreaded all over Daghestan and started to serve as the lingua franca (the level of bilingualism is shown in Figure 2). A good command of Russian was supported by the arrival of television and by intensive migration to towns. Today, almost every family has relatives who live elsewhere and come to the village for vacation or on some special occasion (such as weddings and funerals). Children who were born in cities usually only speak Russian, and pass Russian to their peers who live in the village (Daniel et al. 2011). Therefore, until recently the languages that could influence the vocabulary and the grammar of Mehweb were Avar and Lak. This role has now been assumed by Russian. In spite of the changes in the multilingualism patterns, the Mehweb community still remains, comparatively, more multilingual than other neighbouring communities.

### **Acknowledgements**

I am very grateful to Olesya Khanina and Francesca Di Garbo who read the paper and gave valuable feedback.

Figure 2: Multilingualism in five villages: before 1919 and after 1920 (map courtesy Yuri Koryakov)

### **References**


1 The language and people of Mehweb


## **Chapter 2**

## **Phonology of Mehweb**

### George Moroz

Linguistic Convergence Laboratory, National Research University Higher School of Economics

In this paper, I describe the phonetic inventory of Mehweb, consonants and vowels, as well as the main productive alternations. Two separate sections treat the rules of syllable structure and give a preliminary treatment of pharyngealization. In Mehweb, pharyngealization is a feature which extends the basic vowel inventory (*i*, *e*, *a*, *u*) to include *oˤ* (the pharyngealized variant of *u*, along with pharyngealized *iˤ*, *eˤ*, *aˤ*, *uˤ*) and the inventory of radical and laryngeal consonants by the process of epiglottalization (where *ʡ* is a pharyngealized variant of *ʔ* and *ʜ* is a pharyngealized variant of *ħ*).

*Keywords:* syllabification, stress, vowels, consonants, pharyngealization, alternations.

### **1 Introduction**

This paper is an overview of the phonology of Mehweb. It is primarily descriptive and is intended to make phonological aspects of Mehweb clear to the reader. The paper is organized as follows. In §2 and §3 I describe the language's consonant and vowel systems. §4 is dedicated to syllable and word structure of Mehweb. §5 deals with stress. In §6 I introduce basic phonological and morphophonological alternations. In the last section I describe pharyngealization and how it affects segments.

### **2 Consonants**

The inventory of consonants is given in Table 1. Sounds provided in parentheses are allophones, distributed either contextually or socially, as described below.

George Moroz. 2019. Phonology of Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 17–37. Berlin: Language Science Press. DOI:10.5281/zenodo.3402056

#### George Moroz


Table 1: Mehweb consonant phonemes<sup>1</sup>

There are 41 consonant phonemes in Mehweb, which are listed in Table 1. Most plosives and affricates form three-way oppositions (voiced vs. voiceless vs. ejective), but there are no radical<sup>2</sup> voiced segments except some rare realizations of *h* as *ɦ*. I don't mark concrete place of articulation for the sonorants *n*, *l* and *r*, since they can be realized as either dental or alveolar. All postvelar consonants and velar plosives have labialized counterparts, which occur in word-initial, medial intervocalic, medial preconsonantal and final position. Some Dargwa languages have voiceless geminate consonants. They correspond to voiced consonants in Dargwa languages lacking geminates. There are no geminates in Mehweb (contra Magometov 1982: 8). Sequences of homorganic consonants, however, are realised as geminates phonetically (cf. example (1)):

#### (1) *it-di-ni* > *itːini* this-pl-erg

The voiced velar fricative *ɣ* is attested only word initially in a few roots and only in the speech of older consultants (cf. examples (2–4)). Younger consultants use the velar stop *g* instead.

(2) *ɣan*

'snake'

$$
\begin{array}{cccccccccc}
\text{CT} & \text{g} & \text{z} & \text{s} & \text{c} & \text{c} & \text{g} & \text{z} \\
\text{IPA} & \text{g} & \text{3} & \text{f} & \text{tf} & \text{ts} & \text{dz} & \text{dy} \\
\text{.} & \text{.} & \dots & \dots & \dots & \text{.} & \text{!} & \text{!} \\
\end{array}
$$

2 I use *radical* after Ladefoged & Maddieson (1996) to denote pharyngeal and epiglottal sounds.

<sup>1</sup> In the table, +v stands for voiced, -v stands for voiceless, ej stands for ejective, lab stands for labialization. Some allophones are presented in brackets. To be consistent with the transcription system used in the other contributions to this collection, I use the following transcriptions:

2 Phonology of Mehweb


The voiced affricates *ʒ* and *ǯ* are allophones of the voiced fricatives *z* and *ž*. They are attested only in the speech of older consultants (cf. example (5a–b)):

(5) a. *ʒe* (older speakers) 'salt' b. *ze* (younger speakers) 'salt'

Some realizations of*s* in intervocalic position seem geminate and are perceived as such by some of our consultants, including *isːes* 'take (ipfv)', cl-*isːes* 'weep (ipfv)', *usːes* 'grind'. These are the only three verbal roots with intervocalic *s* known to us, and we have no comparable evidence for nouns. There is thus no clear evidence that geminate *sː* is phonologically distinct from simple *s*. The issue requires further investigation.

The glottal stop *ʔ* is usually deleted in initial and intervocalic position. Some older speakers occasionally produce the voiced glottal fricative *ɦ* instead of voiceless *h* in intervocalic position.

In non-final position epiglottal *ʡ* and *ʜ* are in most cases followed and/or preceded by a pharyngealized vowel. The segments *ʔ* and *ħ* are never followed or preceded by a pharyngealized vowel<sup>3</sup> . In §7 I will discuss some examples of *ʔ*/*ʡ*and *ħ/ʜ*-alternations triggered by pharyngealization, where I will also consider evidence for the independent and suprasegmental nature of the pharyngeal feature.

<sup>3</sup>The situation is however more complex. First of all, the difference between *ʜ* and *ħ* is not perceived by all speakers; the others blame it on the quality of the preceding or the following vowel. Second, on *a* and *u*, the presence of the pharyngeal feature is very hard to perceive, even if the speakers were able to recognize the few minimal pairs we were able to find. One could then simply assume that *ʔ* and *ħ* only appear in non-pharyngealized contexts and *ʡ* and *ʜ* only appear in pharyngealized contexts. However, in the perceptually clearest cases, which are a combination of a pharyngeal stop *ʡ* with the vowel *a*, in some words, epiglottal *ʡ* can be followed by non-pharyngealized *a*. Some of these are Avar loanwords, including *ʡat'* 'dough, flour' (cf. Avar *ʕat'* 'flour'; providing a pseudo minimal pair *ʡat'* 'dough' vs. *ʡaˤt'a* 'frog'), *ʡaraq* 'haystack' (cf. Avar*ʕaraqχ* 'haystack'), *ʡamal* 'temper' (Avar*ʕamal* 'temper'), *maʡna* 'meaning' ( Avar *maʕna* 'meaning'). But other seem to be native, including *ʡarʁal* 'long', *ʡa‹b›ad* 'behind', *ʡaraʁa* 'last', *beʡʷes* 'seed', and particle *ʡaj*. In combination, all this lead to inconsistencies in our transcription of pharyngeals throughout the book. Pharyngealized vowels other than *a* and *u* may also have been lost in transcription.

George Moroz

### **3 Vowels**

There are four plain vowels and five pharyngealized vowels. Length is not distinctive. Some pharyngealized vowels such as *iˤ*, *eˤ*, *uˤ* are very rare, the phonological status of these sounds thus is not clear, so they are written in brackets. Pharyngealized vowels occur most often adjacent to, or in forms containing, epiglottals (*ʡ*, *ʜ*) or uvulars (*q*, *χ*, *ʁ*). However, *aˤ* is also attested in some stems without those segments:

Figure 1: Vowel system


'handful'

(8) *taˤj*

'foal'

Pharyngealized vowels are not common in Mehweb, and some are rarer than others. For example, pharyngealized *iˤ* and *eˤ* are only attested in very few words. Pharyngealized *oˤ* seems to be a realization of *u* in pharyngealized syllables; however, while in some roots only *oˤ* is attested (9a), in other forms *uˤ* occurs as a variant (9b). This distribution may also depend on individual speakers.

(9) a. *doˤrʜoˤ* 'cub' b. *malʡuˤn, malʡoˤn* 'wolf'

Vowels, as well as pharyngeal and epiglottal consonants, rarely show clear evidence of independent behavior of the pharyngeal feature. Pharyngealized vowels show alternations in e.g. plural stem formation, as shown in examples (10–15); see also Chechuro (2019).

(10) a. *jaˤbu* horse 'horse' b. *jaˤb-ne* horse-pl 'horses' (11) a. *taˤj* foal 'foal' b. *tuˤj-re* foal-pl 'foals'

2 Phonology of Mehweb


Table 2 sums up the vowel alternation patterns shown in (10) to (15). Pharyngealization-related processes are explained at the end of §7.

Table 2: Examples of alternation patterns


Vowel frequencies in a list of 596 noun roots are as follows: *a* – 38%, *i* – 27%, *u* – 23%, *e* – 6%, *aˤ* – 6%, other vowels less than 2%. The most frequent vowel structure in bisyllabic words is *a*–*a*.

The most complex phenomenon in Mehweb phonology is pharyngealization. Pharyngealization seems to be associated with uvular, pharyngeal and epiglottal consonants, but there are some cases where it is not; cf. (6–8). Pharyngealized vowels typically appear after radical or uvular consonants, e.g. (13a–15a); but sometimes they may precede them, e.g. (13b–15b); or occur both preceding and following them; e.g. (12a) and (12b). For a discussion of an approach to pharyngealization as a suprasegmental feature, see §7.

### **4 Syllable and word structure**

Except in some borrowings, the syllable structure of most words can be described as (C)V(C)(C). In other words, possible syllables are: CV, CVC, CVCC, VC, VCC, and V (cf. (16–21)). If the coda is complex, the first consonant is most frequently

#### George Moroz

either a liquid or a nasal, as in examples (16) and (18). Clusters of sonorants in the same syllable are not attested. Consonant sequences cannot be longer than three segments, as in (21), and appear only at morphological boundaries. I treat such sequences as divided between two syllables. All native words can be divided into syllables according to the above schemata, but no experiments with speakers' judgments on the location of syllable boundaries have been conducted.


The two action nominals *w-ilsk'-ri* (m-look:ipfv-nmlz) and *w-ebk'-ri* (m-die: pfv-nmlz) are the only examples known so far to show a deviant syllable structure. Note that there is some evidence from nominal inflection (Chechuro 2019) that *b* may be treated as a sonorant.

In Mehweb, the sonority sequencing principle<sup>4</sup> is thus rarely violated: codas are predominantly sequences of a sonorant and an obstruent. Sequences of sonorants or vowels are not allowed.

Noun stems can have from one to five syllables (cf. (22–26)). Most common are one- and two-syllable roots. Table 3 shows the proportion of one-, two-, three-, four- and five-syllable noun stems, based on a list of over 500 noun entries.

(22) *bec'* 'wolf'

<sup>4</sup>This principle can be formulated as follows: the overall acoustic energy of segments should increase from the beginning of the syllable towards the syllable nucleus, and decrease from the nucleus toward the end of the syllable. I use a shortened version of the Sonority Hierarchy: obstruents < sonorants < vowels.

#### 2 Phonology of Mehweb

Table 3: Distribution of one-, two-, three-, four- and five-syllable noun stems


	- 'wood'

(26) *urʁaˤdiq'aˤni* 'fat tail'

Most verbal stems are monosyllabic. Out of 150 verbs collected so far, only five are disyllabic (cf. (27)).

(27) *usaʔʷ-as* m.sleep:pfv-inf 'sleep'

There are also five irregular verbal stems (cf. (28–32)) which, in some word forms, only consist of one consonant or, in the case of 'say' (cf. (29)), may be considered to be realized as zero morphs. The vast majority of Mehweb verbs have two stems, a perfective stem and an imperfective stem. It is worth pointing out that all irregular mono-consonantal stems are perfective.

(29) *ib*

say:pfv.aor 's/he said'

<sup>(28)</sup> *k-ib* bring.to:pfv-aor 's/he brought something to somebody'

George Moroz


's/he brought'

These examples show a difference in number of syllables in nominal and verbal stems: nominal stems tend to be disyllabic, while verbal stems are mostly monosyllabic. This type of asymmetry is typical for the other Dargwa varieties as well.

### **5 Stress**

As compared with different Dargwa varieties, Mehweb has more or less fixed stress (cf. Moroz 2014). In nearly all polysyllabic forms the stress is on the second syllable.


There are, however, some exceptions and even some minimal pairs distinguished by the position of the stress (cf. (35–38)).

(35) *bek'á* hill 'hill'

<sup>5</sup>The nucleus of the stressed syllable is marked by an acute accent mark.

2 Phonology of Mehweb


(37) *dužé* night 'night'

(38) *d-úž-e* npl-drink:ipfv-imp 'drink (it)!'

When a suffix is added to a monosyllabic root, the stress is placed on the second syllable, as shown in (39–40).


Some verbal forms are more complex. In (41b) and (41c), as compared to (41a), the stress is on the second syllable, as expected. Example (41d), the only type of structure where two initial syllables are added in inflection, shows that the stress may not leave the verbal stem:


A form that goes against the second-syllable stress generalization is the vocative. A special vocative form only exists for two-syllable stems which denote humans. These forms are treated as a special stress pattern, with the stress on the first syllable. However, an acoustic study is necessary to find out whether this salience should be treated as stress or, alternatively, as a special vocative intonation. In these forms stress always on the first syllable (cf. (42–43)).

George Moroz

(42) a. *adáj* father 'father' b. *ádaj* father[voc] 'fatherǃ' (43) a. *urší* brother 'brother' b. *úrši* brother[voc] 'brotherǃ'

Another exception is the optative form: the optative marker is never stressed (cf. (44–45)):


Imperative forms never have the stress in the final position – as in the optative, in the imperative the stem is stressed. Plural forms, however, where the imperative is suffixed with the plural-of-addressee marker *-na*, have the pattern with stress on the second syllable.

(46) *árc-e* fly:pfv-imp 'fly!'

(47) *arc-é-na* fly:pfv-imp-imp.pl 'fly! (to a group of people)'

There are numerous Arabic borrowings and proper names which are stressed mostly as in Arabic (cf. (48–51)):

(48) *amanát* 'assignment' (49) *paraq'át*

'calm'

(50) *ʡáˤq'lu* 'wit'

2 Phonology of Mehweb

(51) *másala* 'for example'

### **6 Some phonological and morphophonological alternations**

In Mehweb hiatus is not allowed, and the underlying forms are changed in various ways whenever such configurations arise. If the verb stem is *i*C or *u*C, *i* becomes *j* (as in (52) and (53)) and the vowel *u* (*uˤ*, *oˤ*) becomes *w* (as in (54) and (55)).


Whenever the verbal stem consists of two consonants, the root-initial vowel deletes after the negation marker (as in (56)).

(56) *ħalʔun /ħa-elʔ-un/* neg-count:pfv-aor 'he didn't count'

The vowel *u*, when followed by a consonant cluster, is deleted and triggers the labialization of the final consonant (compare (54–55) and (57–59)). Most labialized consonants that appear as a result of this rule also occur as independent phonemes (see Table 1), but some labialized consonants, e.g. *zʷ*, only appear as a result of this process.

George Moroz


The alternation from the examples above can be generalized as follows:

$$\begin{aligned} \text{(60)} \quad \text{a.} \quad & (a \text{-} u\text{C} / \longrightarrow [a \text{wC}] \\ \text{b.} \quad & (a \text{-} u\text{C} \text{C} / \longrightarrow [a \text{C} \text{C} \text{w}] \end{aligned}$$

The behavior of the labialization feature can be explained by phonotactic constraints. As stated in §4, if the coda is complex, the first consonant is most frequently a sonorant, no complex onsets are allowed, and clusters of sonorants in the same syllable are not attested. The rule in (60b) provides a resolution of unacceptable consonant cluster (*w*-sonorant-plosive).

The marker of the prohibitive and the negative optative (negvol) *m*(V)- has an unspecified vowel that, when appearing before CVC or cl-VC roots, copies the vowel of the root (cf. (61–63)):


The gender marker *b*- assimilates to the nasality of the preceding negvol marker *m*V-; cf. (64).


2 Phonology of Mehweb

The same segment in the verb root does not undergo assimilation:

(65) *m-ib-adi* (\**m-im-adi*) negvol-sew:ipfv-opt 'don't sew'

There are some assimilations triggered by *l* and involving *n* and *l*. The sequences *nl*V or *ll*V in final position can become *w* or *jj* after *u* (66, 67) and *jj* elsewhere (68, 69).


There is a correlation between the age of the speaker and the preferred type of the alternation in nouns: older speakers tend to use the *w*-variant of the genitive, middle-aged speakers consider both *w*-variants and *jj*-variants as well-formed, and younger speakers tend to use the *jj*-variant only. In the imperfective converb, only *w* is available for all speakers.


In medial position, the sequences *nli* or *lli* become *j* and cause vowel deletion (cf. (72–75)):


George Moroz


The sequences *n***V***l* or *l***V***l* after *u* show deletion of a medial vowel, which feeds the *nl*/*ll* alternations above; cf. (76):

	- b. *hujzé /hun-li-ze ← huní-li-ze/* road-obl-inter[lat]
	- c. *huwá /hun-la ← huní-la/* road-gen

When the clusters *n*V*l* or*l*V*l* follow any other vowel, only an unstressed vowel can be deleted, and this deletion also feeds the *nl*/*ll*/*jj* alternation described above (cf. (77–80)):


There are some exceptions to the vowel deletion rule, illustrated in (76). While (81) shows non-deletion of a stressed vowel, in (82–83) the stressed vowel is deleted:

(81) *culála* tooth-gen


#### 2 Phonology of Mehweb

Finally, *r* can assimilate to *n* and *l* (cf. (84–88)), including after applying vowel deletion (cf. (97) and (98)), which then feeds the *r*-assimilation.


In some cases, this assimilation is optional (cf. (89–91)):


The *r*-assimilation would increase the number of forms to which *nl-* and *ll*mutations would apply. This does not happen, however, so I postulate that *r*assimilation applies after *nl-*/*ll-*mutations (counterfeeding order, see Kiparsky 1968):

Table 4: Interaction of the *nl-*/*ll-* mutation rule and the *r*- assimilation rule


The rule for vowel deletion between the consonants *r*, *l* or *n* can be generalized as follows:

**vowel deletion rule**: V → ∅ / [+cons;+son; DORSAL]\_\_[+cons;+son; DORSAL]

#### George Moroz

Table 5 summarises the rules discussed in this section.

Table 5: Interaction of the *nl-*/*ll-* mutation rule, the *r*- assimilation rule and the vowel deletion rule


### **7 Pharyngealization**

I suggest that pharyngealization is a suprasegmental feature. By this I mean that the pharyngealization is not associated with a specific consonant or vowel but with a whole syllable; under certain conditions, it may spread to other syllables. I will mark the presence of the pharyngeal feature on the nucleus of the syllable by *ˤ*. Phonetically, pharyngealization causes centering of vowels and epiglottalization of the consonants *ʔ* and *ħ*:

Table 6: Effect of pharyngeal feature on vowels and consonants


The evidence that the surface segment *ʡ* and *ʜ* are underlyingly *ʔ* and *ħ* comes not only from the fact that the latter segments do not co-occur with pharyngealization (see note 2 above) but also from different realizations of the same morphological segments in inflection and derivation. Consider the following examples:


(93) *ar-b-uχ-ib* away-n-bring:pfv-aor 'took it away'

2 Phonology of Mehweb

### (94) *ʡaˤr-d-aˤq'-un* < */ʔar-d-aˤq'-un/* away-f1-go:pfv-aor 'she is gone'

As stated in §2, the glottal stop *ʔ* in intervocalic and initial position is often deleted. Glottal stops in initial and intervocalic position can be deleted and appear only in the formal speech styles. I stipulate that at the underlying level vowel initial morphemes have the initial glottal stop. Examples (92–93) show that the pharyngeal feature can spread backward, under which condition an underlying *ʔ* and *ħ* become epiglottal and cease to be affected by the deletion rule. This provides a uniform underlying representation of the prefix, as shown in Table 7.

Table 7: Pharyngealization of underlying initial glottal stop


In the first wordform, there is no lexical pharyngeal feature on the root. Pharyngealization does not spread leftward and does not change the underlying glottal stop to *ʡ*; it can then be deleted. On the contrary, in the second wordform, the lexical pharyngeal feature of the root spreads leftwards and changes the glottal stop to epiglottal, which cannot be dropped.

There is another argument for the *ʔ*-to-*ʡ* pharyngealization hypothesis. Examples of the sequences of the epiglottal *ʡ* and plain vowels are rare and seem to be detectable as Avar borrowings. This interpretation creates some minimal pairs distinguished by the pharyngeal feature alone (cf. (95–98)):

(95) *ʔe*

'winter'

(96) *ʡeˤ* < */ʔeˤ/* 'summer'

(97) *d-irʔ-an* npl-gather:ipfv-hab 'gathers them'

George Moroz

(98) *d-irʡ-aˤn* < */d-irʔˤ-an/* npl-freeze:ipfv-hab 'they are freezing'

Pharyngealization in (98) may be explained as a floating feature (similarly to floating tone in Goldsmith 1976) that attaches to the post-root syllable of *-irʔ*; the ending *-an* becomes pharyngealized.

Evidence for *ħ* becoming *ʜ* in a syllable with the pharyngeal feature is provided by the negation prefix *ħa-* in contexts of the pharyngeal feature spreading backward (cf. (99–100)):


In nouns, some of the plural CV-morphemes may delete the stem-final vowel. If the deleted vowel is pharyngealized, the pharyngeal feature moves to the previous syllable (101–103):


I suggest that, in examples (101a), (102a) and (103a), only the second syllable of the underlying form is pharyngealized. In examples (101b), (102b) and (103b), the plural morpheme deletes the nucleus of the pharyngealized syllable, and the feature spreads to the previous syllable. We thus observe *ʡ* in examples (101b) and (102b).

2 Phonology of Mehweb

Pharyngealization rules in Mehweb represent a complex phonological phenomenon that requires further study. I will summarize its most prominent properties:


### **8 Conclusion**

This paper explored phononological characteristics of Mehweb. The main generalizations are as follows. Most plosives and affricates form three-way oppositions (voiced vs. voiceless vs. ejective). There are epiglottal consonants and pharyngealized vowels that can be described as a result of the realization of suprasegmental pharyngeal feature. The majority of native Mehweb words can be described as (C)V(C)(C). Nearly all polysyllabic forms have the stress on the second syllable. To describe alternations (including vowel deletion, *nl*- and *ll*- mutation rules and *r*-assimilation), I stipulate that vowel deletion feeds all other rules, and *r*-assimilation counterfeeds *nl-*/*ll*-mutations.

### **Acknowledgements**

I would like to thank a number of people for reading this paper and discussing its contents with me; in particular, Michael Daniel and Alexandre Arkhipov, as well as to the anonymous reviewers of Language Science Press.

### **List of abbreviations**



### **References**

Chechuro, Ilya. 2019. Nominal morphology of Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 39–72. Berlin: Language Science Press.


## **Chapter 3**

## **Nominal morphology of Mehweb**

### Ilya Chechuro

Linguistic Convergence Laboratory, National Research University Higher School of Economics

This paper describes the nominal morphology of Mehweb. It deals with the following issues: the nominal paradigm, plural formation, the oblique stem, case formation and use, and irregular locatives. In this paper I analyze both the structure and the semantics of these forms.

*Keywords*: nominal inflection, case, number, locative.

### **1 Introduction**

In this paper, I consider the following aspects of Mehweb grammar:


Since gender is not marked on nouns and is only reflected in verb agreement, this aspect of the grammar is discussed in the chapter on verbal morphology (Daniel 2019).

### **2 Structure of the nominal paradigm**

The the nominative singular form is identical to the nominal root. Mehweb also has two intermediate derivational stems, the oblique stem and the plural stem. The oblique stem is derived from the root by an affix or, much more rarely,

Ilya Chechuro. 2019. Nominal morphology of Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 39–72. Berlin: Language Science Press. DOI:10.5281/zenodo.3402058

#### Ilya Chechuro

through a non-segmental operation, and further derives all inflectional forms other than the nominative and the genitive case in the singular, including the ergative case. The rules of oblique stem formation are described in §4. The plural stem is derived from the root and attaches plural suffixes. The rules of plural stem formation are specific to each of the plural suffixes and are discussed in the sections dealing with the corresponding suffixes. In the plural, case suffixes follow the plural suffix. Figure 1 describes the general mechanism of the formation of the plural and oblique stems, starting from the root of a noun:

Figure 1: Plural and oblique stem formation

Or, in tabular form:



#### 3 Nominal morphology of Mehweb

As one can see from Table 1, the first slot is occupied by case or localization markers, while the second slot is restricted to the orientation markers and can be filled only if there is a localization marker in the first slot.

Henceforth I distinguish between the *stem* and the *root* of a word. The root of a word is the deepest level of underlying representation of the unchangeable part of a noun, which usually coincides with the nominative. The only exception is overt gender marking, which is only characteristic of some nouns of adjectival origin, such as *uqna* 'old man' (plural *b-uqna-r-t* 'old men'). Here the gender markers, which are not part of the root, are present in both singular and plural forms (masculine singular *w-* is assimilated with the [u] in the beginning of the word, *b-* stands for the human plural). In this and similar cases I consider the gender (also called *class*) agreement slot a part of the root and mark it as cl. This definition is slightly different from the canonical one given in Haspelmath & Sims (2010: 19) where the root is defined as the part of a lexeme that remains after all affixes have been removed. I assume the agreement slot (but not the marker itself) to be part of the nominal root.

The stem is a representation of a root, including intermediate phonological and morphological representations. Thus, the root is an abstraction that can correspond to a number of different stems, as in the two forms in Table 1: *ʁarʁa* 'stone:nom.sg' and *ʁarʁ-u-be* 'stone-pl.stem-pl'. In this example the root is *ʁarʁa*, while the stems are *ʁarʁa* and *ʁarʁ-u*. Stems are never used without case suffixes (assuming a zero affix in the nominative) and thus are also an abstraction.

The nominal paradigm of Mehweb consists of two parts (or sub-paradigms): *grammatical*, or *functional*, cases and *locative forms*. The two types differ in their morphology: functional case markers consist of one inflectional morpheme; locative forms include two inflectional slots: *localization* (LOC) and *orientation* (OR).

There are a number of nominal inflectional forms that can be historically analyzed as former locatives but are synchronically monomorphemic. These are *the causal*, *the substitutive*, *the replicative* and probably *the comitative*.

### **3 Plural**

The description of plural formation in this chapter is based on wordlists presented in Magometov (1982) and lexical data collected by George Moroz during the field trips undertaken in 2013–2016 (Moroz 2019 [this volume]).

The category of number distinguishes three values: singular, plural and associative. The singular is not marked. The plural is marked with the following suf-

#### Ilya Chechuro

fixes: *-t*, *-be*, *-me*, *-ne*, *-e*, *-le*, *-he*, *-re*, *-še*, *-nube*, *-tune*, *-urbe*, *-lume*. The associative plural suffix is *-qale*.

The suffixes *-t*, *-be*, *-me*, *-ne*, *-e* are frequent. The suffixes *-le*, *-he*, *-re*, *-še*, *-nube*, *-tune*, *-urbe*, and *-lume* are limited to small classes of nominal stems.

Strictly speaking, the choice of the plural suffix is lexical. In most cases, it cannot be predicted either from the formal properties of the stem or from the semantics of the noun. The plural stem formation is not always predictable, either.

On the other hand, each plural suffix has certain – and sometimes quite strong – constraints on the phonotactic structure of the stems to which it can attach. There are different rules of plural stem formation for different affixes, which, however, involve partially similar patterns. For instance, the suffix *-e* only attaches to one-syllable stems (§3.8) and the suffix *-re* usually changes the root vowel of one-syllable nominative stems to [u] (§3.9). Another almost universal process is final vowel syncope, which affects all stems except monosyllabic words and borrowings. However, though the processes discussed in this chapter often apply to most of the formally eligible nouns, almost none of them is truly obligatory.

Below, I attempt to formalize (to some extent) the rules of plural formation. Each of the subsections deals with a particular suffix. In each subsection, I describe the restrictions observed, based on the dictionary data and the data from Magometov (1982). For the suffixes *-ne*, *-e*, *-le*, *-he*, *-re*, *-še*, *-nube*, *-tune*, *-urbe*,  *lume*, and partly also for the suffix *-me*, I have been able to specify the classes of nouns that take these suffixes. For the other suffixes, I have only been able to specify the stem changes they cause.

I will use the following abbreviations: C for consonants, V for vowels, R for sonorants.

#### **3.1 The plural suffix** *-t*

The plural suffix *-t* is one of the most productive suffixes found with this function. In the presence of this suffix, the stem undergoes the following changes:


<sup>1</sup> If this vowel is pharyngealized, it changes into [oˤ], the phonetic realization of /uˤ/.

#### 3 Nominal morphology of Mehweb

4) The word *uqna* 'old man' forms the plural stem by inserting *-r-* even though it is not borrowed.

The rough generalization is that the suffix *-t* attaches to stems ending in sonorants.

Table 2 illustrates vowel drop and vowel change (Rule 1):


Table 2: Rule 1

Table 3 illustrates the second rule:

Table 3: Rule 2


Table 4 shows how the *-t* suffix interacts with borrowed stems ending in a vowel: the vowel drop does not apply.

#### Ilya Chechuro


Table 4: Rules 3 and 4

Borrowed stems that end in a sonorant attach the *-t* suffix directly, as illustrated in Table 5:


Table 5: Borrowed stems that attach the suffix *-t* directly

The plural suffix *-t* also forms plurals of the words that denote inhabitants of Mehweb and neighbouring villages. In Magometov (1982) this use of the suffix *-t* is described as a separate suffix *-n-t*. However, forms such as *mehʷa-n* 'a Mehweb

<sup>2</sup>The word *uqna* also contains a gender marker, which expresses the number and gender of this word. Thus, in the singular the marker is masculine singular *w-* (dropped before the [u] of the stem), while in the plural the human plural marker *b-* occurs. Several other nouns in Mehweb and other Dargwa dialects also include a gender marker.

3 Nominal morphology of Mehweb

person' and *surʁatla-n* 'a person from the village of Sogratl suggest that *-n* is a nominalizer and therefore not part of the plural morpheme (see Table 38 in §6).

#### **3.2 The plural suffix** *-ne*

With the suffix *-ne*, the stem undergoes the following changes:


Table 6 illustrates the first rule:

Table 6: Rule 1


Table 7 illustrates the mechanism of the plural formation of one-syllable stems attaching the suffix *-ne* (Rule 2):



#### Ilya Chechuro

Table 8 illustrates Rule 3:

Table 8: Rule 3


Rule 3 has one exception: the plural stem of the word *ʁamas* 'box' is formed by syncope of the last vowel of the root:



The nouns given in Table 10 undergo haplology:

Table 10: Haplology


The haplology here applies to the contiguous VR sequences: when after a derivation there are two VR sequences with the same R next to each other, the

#### 3 Nominal morphology of Mehweb

first one is dropped, e.g. *urʁadiqaˤn-u-ne* → *urʁadiq-uˤ-ne*. These words can also be analyzed as attaching the suffix *-e* after dropping the final vowel. However, since the suffix *-e* prefers one-syllable stems, my analysis seems more feasible<sup>3</sup> .

Several words form the plural stem by changing the vowel in the first syllable (which is also the penultimate) into */u/*:

Table 11: Vowel change in the root


#### **3.3 The plural suffix** *-tune*

The words *qašqar* 'bald man', *wakil* 'lawyer', *arab* 'Arab' and *sabab* 'reason' attach the plural suffix *-tune*. Diachronically, these words employed the suffix *-t(e)*, as in some other Dargwa dialects, e.g. Kubachi. Presumably, this plural marking was then reinforced by *-ne*, which required the change of the final vowel to *-u*. Together, these suffixes formed the structure *-tune*, which is synchronically monomorphemic (Table 12):

Table 12: The plural suffix *-tune*


#### **3.4 The plural suffix** *-be*

With the suffix *-be*, the stem undergoes the following changes:

<sup>3</sup>Magometov (1982: 36) treats these cases as cases of apophony rather than haplology. He analyzes the forms *χajqune* and *sersit'une* as follows: "There are cases, albeit rare, when a word ending with *-e* in the plural differs [from the singular] only by a vowel change in the stem. This vowel change, therefore, acquires a morphological meaning" (translation from Russian by the author).

#### Ilya Chechuro


Table 13 illustrates Rule 1:


Table 14 illustrates Rule 2:

Table 14: Rule 2


Note that *liga* 'bone' also forms the plural stem by attaching *-u-* even though the first syllable does not contain [a].

Several nouns form their plural stems by changing the root vowel to [u]. All of these words either have [e] in this syllable or contain a labialized or labial consonant:

Table 15: Vowel change in the root


#### 3 Nominal morphology of Mehweb

An assimilation occurs in stems ending with [n]: /n+be/ → [mbe]:


If a stem ends in a labialized consonant, this consonant is delabialized:

Table 17: Delabialization


#### **3.5 The plural suffixes** *-nube* **and** *-urbe*

The suffix *-nube* forms the plural of five lexemes. The suffix *-urbe* forms the plural of four lexemes. These suffixes are similar to *-tune* in that they may be analyzed as *-ne* and *-re* followed by *-be*. The *-u-* of the suffixes *-nube* and *-urbe* may be considered as resulting from the final vowel change seen in §3.2 above. Synchronically, *-nube* and *-urbe* are monomorphemic suffixes with a very limited lexical distribution (Table 18):

Table 18: The plural suffixes *-nube* and *-urbe*


Ilya Chechuro

#### **3.6 The plural suffix** *-me*

With the suffix *-me*, the following rules apply:


Table 19 illustrates Rule 1:


Table 20 illustrates Rule 2:



Some nouns form plural stems by attaching *-u-* after dropping the last vowel. All of them contain an [u] or a labial/labialized consonant. One may notice that in most cases, after the final vowel drop has been applied, [u] is inserted to avoid

#### 3 Nominal morphology of Mehweb

a phonologically illegitimate consonant cluster. There is, however, no such consonant cluster in *uq'lah-u-me* (cf. *kʷih-me* 'sheep, PL'). The Russian loanword *bidra* 'bucket' also belongs to this group. Table 21 below illustrates this process.

Table 21: Plural stem formation by attaching *-u-*


The words *laˤwlu* and *waq'lu* are also analyzed as dropping their last vowel and then attaching *-u-*:

*laˤwlu + me → laˤwl + me → laˤwl + -u- + -me → laˤwl-u-me*

Under this analysis, the [u] in the plural is not the same as the [u] in the singular.

#### **3.7 The plural suffix** *-lume*

The following words form the plural with the suffix *-lume*, which historically seems to be the plural suffix *-le* with a change of the final vowel before the plural suffix *-me* (Table 22):

Table 22: The plural suffix *-lume*


#### Ilya Chechuro

#### **3.8 The plural suffix** *-e*

The suffix *-e* attaches to one-syllable stems. It can attach directly to CVC(C) stems. In some cases, the rules for plural stem formation derive one-syllable stems from more-than-one syllable stems and are as follows:



Table 23: The plural suffix *-e*

Table 24 illustrates Rule 1:

Table 24: Rule 1


<sup>4</sup>The word *naˤʁ* 'hand' appears to undergo the */a/* → */u/* vowel alternation described in §3.2. Since this alternation does not affect the word *maq'ʷ* 'root' that has a similar phonetic structure, it is possible to hypothesize that the suffix *-e* has originates from several different suffixes that merged in the *-e* form due to phonetic changes.

#### 3 Nominal morphology of Mehweb

Table 25 illustrates the vowel syncope described in Rule 2:


Table 25: Rule 2

#### **3.9 The plural suffix** *-re*

This suffix has a limited lexical distribution. The rules for plural stem formation are similar to the rules for other C*e* suffixes<sup>5</sup> (see also §3.4):


The suffix *-re* prefers one-syllable words and two-syllable stems ending with [i].

Table 26 illustrates Rule 1:



<sup>5</sup> I do not have a satisfactory explanation for this parallel. It is possible that *-e*, which was originally present in all plural suffixes including *-t*, as confirmed by other Dargwa lects, at some point became associated with the expression of plurality, and the consonants came to be interpreted as parts of the plural stem of the noun.

#### Ilya Chechuro

Table 27 illustrates Rule 2:

Table 27: Rule 2


However, there are exceptions to Rule 2. Two roots contain [a] but do not undergo vowel change (Table 28):


The [r] in the suffix *-re* can, but need not, assimilate to the [l] of the stem (Table 29):


#### **3.10 The plural suffix** *-le*

The plural suffix *-le* only occurs with four nouns. If the stem ends in a vowel, the vowel is dropped. The vowel of the stem changes to /u/ (Table 30):

Table 30: The plural suffix *-le*


#### 3 Nominal morphology of Mehweb

#### **3.11 The plural suffixes** *-he* **and** *-še*

The suffix *-he* occurs with two nouns. Both have irregular plural stems, so the plural formation may be considered to be weak suppletion (Table 31):


The plural suffix *-še* occurs with one noun, *qu* 'field' (Table 32):

Table 32: The plural suffix *-še*


#### **3.12 The associative plural suffix** *-qale*

The plural suffix *-qale* most probably results from grammaticalization of the noun *qali* 'house'. In the case of Mehweb, this suffix covers the so-called associative plural meaning 'X and his or her family' (in spontaneous texts also 'X and those with him/her', 'X and his/her group'). For Tanti Dargwa, Lander (2008) observes that the suffix *-qale* has developed a regular plural meaning. This evolution has not been reported for standard Dargwa. In Mehweb Dargwa, regular plural uses of *-qale* are attested on nouns for 'mother' and 'father'; for 'grandmother' and probably 'grandfather', both regular and associative plural readings are attested. Table 33 illustrates the use of this suffix:

Table 33: The associative plural suffix *-qale*


#### Ilya Chechuro

### **4 Oblique stem**

The genitive case suffix attaches directly to the nominative stem (in all nouns but not in all pronouns – cf. *di-la* I.obl-gen 'my'). All other cases (including ergative) require an oblique stem. In the plural, all case suffixes attach directly to the plural marker.

The oblique stem marker has three allomorphs: *-li*, *-j*, and *-i*. The marker *-li* is the default way to form an oblique stem and is applicable to almost any stem.

The marker *-i-* may be considered prothetic (to resolve consonant clusters) and is generally not separated or glossed in this book. The use of the segmental marker *-li-* is a lexical property. With some nouns, the two strategies are in competition:

(1) *muħammad-li-ni* Muhammad-obl-erg *muħammadi-šu* Muhammad-ad(lat)

The oblique stem marker *-li-* may (but does not have to) change to *-j-*. Table 34 shows contexts that license the change. The first column shows the vowel preceding the last consonant. The second column shows the consonant and the vowel that can follow it:



Example (2) illustrates the process (see more in Moroz 2019):

(2) *rasul* Rasul *rasuj-ni* Rasul.obl-erg

### **5 Nominal inflection system**

The nominal inflection of Mehweb Dargwa consists of two parts (sub-paradigms): *grammatical cases* and *locative forms*. The two types of inflectional forms differ in their morphology: grammatical case forms contain one inflectional morpheme (Table 35); locative forms contain two inflectional morphemes. The first morpheme of a locative form designates the *localization*: the spatial area defined

#### 3 Nominal morphology of Mehweb

with respect to a landmark (rows in Table 36 below). The second designates the *orientation* (columns in Table 36 below): the trajectory of the object with respect to the area designated by the localization.

The core function of locative forms is to describe spatial relations between a figure and a ground (Rubin 2001). Grammatical cases are primarily used to express grammatical relations and abstract semantic roles. However, across East Caucasian, this is only a typical division of labour, and both types of inflection can be used in both functions (Kibrik 2003). In Mehweb, grammatical cases do not have any spatial uses (except for the fact that the genitive suffix is identical to the elative suffix) but spatial cases can have (nearly) abstract functions.

In Mehweb, there are five localization morphemes and five orientation morphemes. Each localization can take each of the orientations, forming a system of 25 locative forms. The subsections below are named according to the grammatical case labels and localization markers. One localization morpheme can designate several distinct spatial areas. I thus use the labels written in small-caps (e.g. inter) as a semantic label, not as a gloss of a morphological category (as in the rest of the papers in this collection).

I do not discuss the semantics of the orientation markers in separate subsections. Their spatial functions are introduced in Table 36 and are independent from the semantics of the localization they combine with. In their non-spatial uses, most locative forms cannot be described compositionally by referring separately to the semantics of the localization and orientation markers. I thus discuss these uses among the functions of the individual localization markers in the relevant subsections.

The structure of the case system is shown in the two tables below. Table 35 shows grammatical cases. Table 36 shows locative forms, together with their core


Table 35: Mehweb functional sub-paradigm

#### Ilya Chechuro


#### Table 36: Mehweb locative sub-paradigm

meanings. The abbreviations for the morphemes in the orientation slot are as follows: lat – lative, ess – essive, el – elative, trans – translative, dir – directive, cl – gender agreement marker.

Example (3) illustrates how the locative markers function:

(3) *ʁarʁa* stone(nom) '(a) stone' *ʁarʁa-li-če* stone-obl-super(lat) 'onto the stone' *ʁarʁa-li-če-w* stone-obl-super-m(ess) '(he is) on the stone' *ʁarʁa-li-ze-b* stone-obl-inter-n(ess) '(it is) in the stone'

<sup>6</sup> cont is the functional label of a spatial configuration in which the object is located on the surface of a landmark and stays there because of the nature of the contact between the object and the landmark, or because it is a part thereof. Typical cont contexts are: '(a picture) on the wall', '(a ring) on a finger', '(wings) on the back', '(a birthmark) on the face'. Many East Caucasian languages have a separate localization marker for the cont configuration. In Mehweb, this configuration is divided between *-če-* (labelled super, discussed in §5.9) and *-ze-* (labelled inter, discussed in §5.11).

The lative (lat) is expressed by the absence of an orientation marker. The essive (ess) is expressed by the presence of the gender agreement slot (shown as -cl in the table). The agreement is controlled by the NP designating the trajector. Since the two markers do not have their own dedicated exponency, their glosses are bracketed.

### **5.1 Nominative**

The nominative case marks the S of an intransitive verb and the P of a transitive verb:


The nominative case is also used when addressing someone:

(6) *baba* granny *nab* I.dat *inc'ul* more *uk-es* m.eat:pfv-inf *ħa-d-ig-an.* neg-npl-want:ipfv-hab 'Granny, I don't want to eat any more.'

The nominative is also used in constructions like (6):

(7) *χʷe-li-če-la* dog-obl-super-el *ažda* crocodile *b-uh-ub.* n-become:pfv-aor 'The dog has become a crocodile.'

### **5.2 Ergative**

The ergative case marks the A of a transitive verb and the instrument:

(8) *adaj-ni* father-erg *mašinka-li-ni* hair.cutter-obl-erg *muc'ur* beard(nom) *b-erč-ur.* n-cut.hair:pfv-aor 'The father cut his beard with clippers.'

The ergative case also marks periods of time. The semantics of such constructions can be formulated as 'X did something for two hours', i.e. the result was not necessarily achieved:

#### Ilya Chechuro

(9) *k'ʷi-jal* two-card *saʡaˤt-li-ni* hour-obl-erg *rasul* Rasul(nom) *ħule* look *w-ilz-uwe* m-lv:ipfv-cvb.ipfv *le-w-re* be-m-pst *ši-la* village-gen *surt-me-če.* picture-pl-super(lat) 'Rasul has been looking at the photos of (his) village for two hours.'

### **5.3 Genitive**

The genitive case marker is *-la*. It can undergo the following processes:


The genitive of place names is formed with *-la* or *-ja* (probably derived from *-n-la*; see below), while their *-la* form serves as the elative. Note that place names in Mehweb are a separate part of speech possessing morphological and syntactic properties of both nouns and locative adverbs. They lack an oblique stem and have an irregular genitive form. They attach orientation markers directly, like spatial adverbs. Their quotation form is also the essive form. Hence, the *-la* marker in Table 37 is not only a genitive marker but also an elative marker:


Table 37: The Genitive of Place Names

The main function of the genitive case is to mark a noun that is dependent on another noun (possessive construction):

(10) *rasuj-ni* Rasul.obl-erg *ar-d-uk-ib* away-npl-lead:pfv-aor *muħammad-la* Muhammad-gen *kʷihme.* sheep.pl 'Rasul took away Muhammad's sheep.'

In possessive predication, the possessor genitive is "free" in that it does not form a single constituent with the possessum.

(11) *nuša-la* we-gen *le-b* be-n *ʁarʁ-u-be-la* stone-pl.stem-pl-gen *qali.* house 'We have a stone house.'

In the predicative possessive construction, Mehweb distinguishes two types of possessors: locative possessor and genitive possessor. Locative possession is only possible in predicative constructions, while genitive possession can be either adnominal or predicative (free genitive). The semantic difference between the two constructions is that the locative possessor has an object with/on her, but this object does not necessarily belong to her. The genitive possessor possesses an object, i.e. it belongs to her:


The difference does not apply to adnominal possessive constructions. It is not possible to use the localization marker *-ze* in an adnominal possessive construction:

(15) *\*rasuj-ze-b* Rasul.obl-inter-n(ess) *dis.* knife '(someone else's) knife that Rasul has got.'

### **5.4 Dative**

The dative case marker is *-s*. It attaches to the oblique stem. Its basic function is to mark the recipient in the 'give' construction:

(16) *abaj-ni* mother-erg *gi-b* give:pfv-aor *sadaq'ači-li-s* pauper-obl-dat *t'ult'.* bread 'Mother gave bread to a pauper.'

#### Ilya Chechuro

The dative also marks the benefactive and several other related roles:


The two types of predicative possession described in §5.2 are paralleled by different strategies for encoding the recipient, as shown in (18). The two types of transmission are encoded by the dative vs. inter-lative form. If the rights of possession are transmitted together with the object, the recipient is encoded with the dative case. If they are not transmitted, as in (19), the recipient is marked with *-ze*:

(19) *rasuj-ni* Rasul.obl-erg *gi-b* give:pfv-aor *muħammadi-ze* Muhammad-inter(lat) *dis.* knife 'Rasul lent a knife to Muhammad.'

The dative is also used for some experiencers. Experiential verbs have one of the two case frames: [experiencer = inter(lat), stimulus = nom] and [experiencer = dat, stimulus = nom]. A dative experiencer is only possible with the verb cl*-iges* 'love/want' and complex predicates:


### **5.5 Comitative**

A co-participant is expressed by the comitative:

(22) *rasul* Rasul *urʁes* fight:ipfv-inf *w-ik-ib* m-lv:pfv-aor *muħammadi-ču.* Muhammad-comit 'Rasul fought with Muhammad.'

3 Nominal morphology of Mehweb

This case is also used for instruments, including consumables:


#### **5.6 Causal**

According to Magometov (1982), there is a case that marks the cause of a situation. My consultants did not confirm Magometov's examples and rejected the *-čeble*/*-čible* forms that I constructed. I assume that the case no longer exists in Mehweb. Examples (25) and (26) are cited from Magometov (1982: 49):


#### **5.7 Substitutive**

The morpheme *-čemadal* has substitutive semantics, i.e. it indicates that the actor performs an action instead of someone who was supposed to perform it, the latter being coded by this case form:

(27) *nu* I *adaj-čemadal* father-subst *tukaj-ħe* shop.obl-in(lat) *w-aˤq'-un-na* m-go:pfv-aor-ego 'I went to the shop instead of father'

Diachronically, this form can be analyzed as *-če-m-ad-al*, in which *-če-* marks super localization, *-m-* is an unknown morpheme that occupies the localization slot and *-adal* is the elative marker (cf. Table 36 above).

Ilya Chechuro

#### **5.8 Replicative**

The last non-spatial case suffix is *-sum*. It conveys the semantics of performing an action in the way similar to how someone or something else performs it, or in the way it is usually done. The form attaches to an irregular oblique stem:

(28) *dilaj-sum* I.obl-repl *b-aq'-a* n-do:pfv-imp.tr 'Do as I do'

The following sections deal with spatial forms.

#### **5.9 The locative marker** *-če-*

The basic semantics of the locative marker*-če-* issuper, i.e. by default this marker is used in contexts like the following:

(29) *ustuj-če-b* table.obl-super-n(ess) *ʁadara* plate *le-b.* be-n 'A plate is on the table.'

The locative marker *-če-* is also used to mark the cont configuration. It shares this function with the locative marker *-ze-*, whose basic semantics isinter (§5.11). The instances involving cont semantics seem to be distributed over the two markers, but the rules are difficult to formulate. Examples (30) and (31) show that the two locative markers are not in free distribution in spatial contexts:


The locative marker *-če-* can be used in 'support' contexts like *put against* (*a tree* etc.):

(32) *ʡali-ni* Ali-erg *mažar* rifle *baˤʜi-če* wall-super(lat) *b-ix-ib.* n-put:pfv-aor 'Ali put the rifle against the wall.'

(33) *nu* I *baˤʜi-če-la* wall-super-el *ʡaˤq* far *ʡaˤr-aˤq'-un-na.* away-m.go:pfv-aor-ego 'I stepped away from the wall.'

In comparative constructions, the object of comparison is marked with *-če-*:

(34) *rasul* Rasul *quwati* strong *le-w* be-m *muħammadi-če-w.* Muhammad-super-m(ess) 'Rasul is stronger than Muhammad.'

The morpheme *-če-* is used to mark the target of an oriented action, e.g. with verbs such as 'hit', 'bark', 'shout at', 'be angry at', 'look at', 'laugh at':

(35) *rasul* Rasul *laχu* scream *uk'-uwe* m.lv:ipfv-cvb.ipfv *le-w* be-m *muħammadi-če.* Muhammad-super(lat) 'Rasul is shouting at Muhammad.'

The super-elative *-če-la* is used with verbs of avoidance: 'run away', 'hide', 'fear', etc.:

(36) *rasul* Rasul *w-aˤld-un* m-hide:pfv-aor *muħammadi-če-la.* Muhammad-super-el 'Rasul hid from Muhammad.'

The marker *-če-* is also used to mark periods of time. The semantics of such constructions can be formulated as 'X did something in two hours', i.e. the result was achieved:

(37) *k'ʷi-jal* two-card *saʡaˤti-če* hour-super(lat) *rasuj-ni* Rasul.obl-erg *kung* book *b-elč-un.* n-read:pfv-aor 'Rasul read the book in two hours.'

#### **5.10 The locative morpheme** *-ħe-*

The locative morpheme *-ħe-* expresses the configuration in when one object is inside another one. The ground is, or is conceptualized as, a container.

(38) *ħarši* soup *k'unk'ur-le-ħe-r* pot-obl-in-npl(ess) *le-r.* be-npl 'The soup is in the pot.'

#### Ilya Chechuro

In (38), the morpheme *-ħe-* causes vowel assimilation (i → e) in the oblique stem marker. Between two vowels, [ħ] may be dropped, and the vowels contract. In such cases, the only indication of in semantics is the vowel change:

(39) *ħarši* soup *k'unk'ur-le-r* pot-obl.in-npl(ess) *le-r.* be-npl 'The soup is in the pot.'

This localization does not have any non-locative uses in any of the Dargwa dialects, including Mehweb.

#### **5.11 The locative morpheme** *-ze-*

The morpheme *-ze-* denotes the configuration when an object is within the spatial area of the landmark and the landmark is either a substance or a set of objects (e.g. 'forest'). This configuration in labelled inter:

(40) *k'as* fish *ħark'ʷi-ze-b* river-inter-n(ess) *le-b.* be-n 'The fish is in the river.'

The morpheme *-ze-* is also used in some cont contexts (also see §5.9):

(41) *surat* picture *aqi-le* up-advz *le-b* be-n *baˤʜi-ze-b.* wall-inter-n(ess) 'A picture is hanging on the wall.'

Forms in *-ze-la* (inter-el) express an *involuntary agent* – a participant who becomes the agent or cause of a situation unintentionally. Only the inter-elative forms in *-la* but not its variants are used in this function:

(42) *di-ze-la* I.obl-inter-el / / *\*di-ze-b-adala* \*I.obl-inter-n-el *mašina* car *b-oˤrʡ-oˤb.* n-break:pfv-aor 'I accidentally broke the car.'

The involuntary agent construction seems to combine only with intransitive (labile in 42) verbs and thus is a means of introducing an agent-like participant rather than decreasing control on the part of a true agent. The same locative form is also found in contexts of participant-internal possibility:

3 Nominal morphology of Mehweb

(43) *rasuj-ze-la* Rasul.obl-inter-el *aq* up *b-aq'-as* n-do:pfv-inf *b-uh-es* n-become:pfv-fut *ʁarʁa.* stone 'Rasul will be able to lift the stone.'

The morpheme *-ze-* marks a temporary possessor (cf. §5.3), temporary recipient (cf. §5.4) and the addressee with verbs of speech:


The functional range of *-ze-* shows that its uses are not always related to its spatial meaning, and that the spatial metaphor, when present, may be weak.

#### **5.12 The locative morpheme** *-šu-*

The ad *-šu-* localization is used to express the fact that one object is located in close proximity to another object:

(46) *nuša* we *ustuj-šu-b* table.obl-ad-hpl(ess) *ka-b-iʔ-i-ra.* pv-hpl-sit:pfv-aor-ego 'We are sitting near the table.'

It is also used as a personal locative:

(47) *nu* I *w-aˤq'-un-na* m-go:pfv-aor-ego *aħmadi-šu.* Ahmad-ad(lat) 'I visited Ahmad.'

#### **5.13 The locative morpheme** *-ʡeˤ-*

The apud marker *-ʡeˤ-* denotes an area close to an object, in which the figure must be located to interact with the object (functional proximity). This suffix shows a very restricted distribution. It is only compatible with words designating landmarks that have an area associated with them in this way; e.g. *ustul* 'table', *iniz* 'water source', *qali* 'house'. In different languages, the same landmark may be conceptualized as having such an area or not. In Mehweb the set of words to

#### Ilya Chechuro

which this suffix can be attached varies across speakers. The following examples illustrate the difference between the ad *-šu-* and apud *-ʡeˤ-* localizations:


It also expresses the meaning of an exchange equivalent:

(51) *rasuj-ni* Rasul.obl-erg *bars* exchange *b-aq'-ib* n-do:pfv-aor *q'ʷaˤl* cow *šu-wal* five-card *kʷiha-le-ʡeˤ-b.* sheep-obl-apud-n(ess) 'Rasul exchanged the cow for five sheep.'

With appropriate grounds, the morpheme *-ʡeˤ-* may be used to designate the area not near to but bounded by the landmark:

(52) *škaf* wardrobe *unza-le-ʡeˤ-di* door-obl-apud-trans *b-aˤq'-un.* n-go:pfv-aor 'The wardrobe went through the door.'

It thus becomes semantically similar to *-ħe-*; in (53), *-ħe-* is used in the same context:

(53) *škaf* wardrobe *unza-le-ħe-di* door-obl-in-trans *b-aˤq'-un.* n-go:pfv-aor 'The wardrobe went through the door.'

Like *-ħe-*, *-ʡeˤ-* causes vowel assimilation *i* → *e* in the oblique stem marker (cf. 52 and 53).

#### 3 Nominal morphology of Mehweb

### **6 Irregular locatives**

A limited number of nouns form locatives in an irregular way. Such irregular locatives usually mark the default location associated with the landmark. As with the locative forms discussed above, the presence of a gender agreement slot conveys the meaning of stative location (essive form), and the same form without the slot conveys the meaning of direction towards (lative). Table 38 shows the irregular locatives attested so far.


Table 38: Irregular locatives

### **7 Place names**

Names of local villages form a separate morphological class close to adverbs; they lack functional cases and attach orientation markers directly to the stem. Their unmarked locative (i.e. lative) form also serves as quotation form. They are nominalized by adding *-n* (also used in the nominalization of adjectives) and form plurals in *-t* to designate the inhabitants of the village. While the genitive in *-la* is produced by simple suffixation of the genitive marker, the variant genitive in *-ja* probably derives from the nominalized form in *-n* (*-ja* < *-n-la*, as discussed in Moroz (2019), thus meaning not 'that of the village of Mehweb' but 'that of a Mehweb villager'.

The inflection of local place names is given in Table 39. Declension of *anži* 'Makhachkala' and *maskaw* 'Moscow', which are not local placenames and behave like regular nouns, is given for the sake of comparison in the last lines of each column.


Table 39: Place names

### **Acknowledgements**

The author is grateful to the Mehweb people for being extremely generous in sharing their knowledge of the language, to his fellow fieldworkers, for their support, and to his teachers for their careful guidance and endless patience.

### **List of abbreviations**



### **References**


## **Chapter 4**

## **Mehweb verb morphology**

### Michael Daniel

National Research University Higher School of Economics

The paper describes the morphology of the verb in Mehweb, a Dargwa lect of central Daghestan, Russia. The description is partly based on previous research (Magometov 1982, Sumbatova unpublished) and partly on the field data the author has been collecting from 2009 to the present. Mostly, formal morphology of synthetic verb forms and complex verbs are discussed.

*Keywords*: East Caucasian, Dargwa, Mehweb, verb, inflection, perfective, imperfective, transitivity, complex verbs.

### **1 Introduction**

In this chapter, I provide an overview of the verb morphology of Mehweb, a lect of the Dargwa branch of East Caucasian languages, spoken in the village of the same name in the Gunib district of the Republic of Daghestan. The paper is mostly focused on formal and synthetic morphology. Periphrastic forms are treated only peripherally, and the semantics of the verbal categories is not discussed at all. As a result, labels provided for different inflectional categories are conventional and to a large extent based on previous research. While formation of deverbal nominal forms – nominalizations and participles – is covered, their further inflection as nominals is also left out. The previous treatment of the Mehweb morphology, Magometov (1982), provided the basis for many analytical solutions.

Mehweb verbs agree in gender (noun class) with their nominative argument, distinguishing three primary genders – masculine (M), feminine (F) and neuter (N) in the singular, human plural (HPL) and non-human plural (NPL) in the plural. There is an additional gender for unmarried girls and women. Agreement marking is largely similar to agreement in adjectives, spatial forms, numerals

Michael Daniel. 2019. Mehweb verb morphology. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 73–115. Berlin: Language Science Press. DOI:10.5281/zenodo.3402060

#### Michael Daniel

etc., which are not treated in this chapter. Agreement morphology is discussed in §2. Additionally, and unlike other parts of speech, some verbal forms show special inflection with first or second person subjects, depending on the illocutionary force (with first person in affirmative utterances and with second person in interrogative ones). These are discussed in §3.

The whole inflectional paradigm of the verb is divided into two parallel sets of forms, based on perfective and imperfective stems, whose relation to each other is complex and follows several different formal patterns with most verbs. The relation between the stems of a few verbs is irregular. Many forms are formed from both stems. This is discussed in §4.

In Mehweb, there are three distinct verbal inflectional classes, distinguished by the suffix they take in the perfective past (aorist), *-ib* (*-ub*), *-ur* or *-un*. The aorist stem is used in the participle and the forms derived from it. Other forms, including all forms in the imperfective, are however formed in the same way for the verbs of all three classes. This is discussed in §5, which also provides a table showing all inflectional forms known so far.

Verbal negation is discussed in §6. The structure of the verbal paradigm as a whole is discussed in §7. Some of the forms follow specific rules, independent from the classification into three inflectional classes. These include imperatives and infinitives and are described in §8. Inflection of the auxiliary is discussed in §9. Verbs with irregular morphology, including verbs of motion, are discussed in §10. §11 presents data on transitivity, including regular morphological causativization and lexically constrained phenomena such as lability. §12 explains the morphological makeup of complex verbs, including verbs with vestigial prefixes, light verbs and verbalizers and bound verbal roots.

### **2 Gender agreement**

Mehweb nouns belong to one of the three primary genders – masculine, feminine and neuter, glossed as M, F and N, respectively. Animate non-human nouns belong to the neuter gender. In the plural, all human nouns behave the same, so that only human plural (HPL) and non-human plural (NPL) are distinguished. Additionally, nouns and pronouns referring to girls or unmarried women (glossed as F1) show a special pattern of agreement – in the singular, they require the same marker as non-human plurals. Many mass nouns and some abstract nouns, in the singular, control NPL agreement.

The morphology of gender markers is shown in the following table and is common to all targets of agreement – adjectives and verbs having a prefix agreement

#### 4 Mehweb verb morphology

slot, locative nominal forms – a suffix slot, etc. Verbs may only have gender markers in the prefix position, and not all verbs have this slot (though most do).


Table 1: Gender agreement marking

The marker of the masculine *w-* is lost in forms where it is preceded by a prefix, either grammatical (negation) or derivational. There is some evidence that this process is optional, at least with the prefix of negation. Cf.:


For more information on the morphology of negation see §6.


Note that, synchronically, most combinations of preverbs with the root are not compositional. Thus, the preverb *ar-* etymologically means 'away', while the verb *-ik-* synchronically means 'happen' (etymologically most probably 'fall').

The masculine marker is also lost in stems with initial *u*-, such as:


For more on preverbs, see §12.

### **3 Egophoric forms**

Some categories of the verb vary depending on whether they have a subject in the first or second person or not. The forms signaling that their subjects are speech act participants will be called egophoric forms below. Unlike gender agreement, subject agreement shows an accusative pattern and is controlled by S/A arguments. The peculiar property of subject agreement in Mehweb as compared to

#### Michael Daniel

other Dargwa languages is that it is sensitive to the illocutionary type of the utterance. The subject suffix appears with first person subjects in declarative utterances but with second person subjects in interrogative utterances. This distribution, known in typological studies as egophoric, is sometimes dubbed disjunct vs. conjunct forms and in East Caucasian languages is so far only attested in Akhvakh (Creissels 2008; 2018) and Zakatala Avar (Forker 2018). Below, this inflection will be glossed as ego.

All TAME categories that have egophoric forms are shown in Table 2, in both egophoric (ego) and unmarked (3) forms:


Table 2: Egophoric forms and their unmarked counterparts

In the past, the egophoric forms are marked with the suffix *-ra*, assimilated to *-na* after the nasal auslaut in the aorist. In the imperfective past, the tense suffix *-ib-* irregularly drops its final *-b*. In the future, non-egophoric forms are identical to the infinitive, while the egophoric forms use a special suffix *-iša*. In the present habitual (which also serves as synthetic present for some stative verbs), there is an opposition of two special affixes, *-an* for non-egophoric and *-as* for egophoric forms. Following the idea that the basic distinction is between egophoric forms that are marked and non-egophoric unmarked forms, I gloss *-an* simply as hab and *-as* as hab.ego (similarly with other forms). Egophoric forms are also present

4 Mehweb verb morphology

with the present form of the auxiliary *lewra* (m), *lella* (< *ler-ra*, f and npl), *lebra* (n and hpl) and the negative copula *aħinna* (< *aħin-ra*) – see §9 on inflection of auxiliaries.

### **4 Aspectual stems**

In Mehweb, the vast majority of the verbal categories are formed from two different stems, perfective and imperfective. I will consider verbal inflection as divided into perfective and imperfective paradigms. The two paradigms are largely parallel. Most categories attested both in the perfective and the imperfective paradigms use the same affixes. The exceptions are listed in the following table:


Table 3: Asymmetries between perfective and imperfective paradigms

On the choice of one of the markers in the same category see the relevant sections below. For the different markers of the aorist (perfective past) see §5; for the choice of the vowels in the imperative and the infinitive see §8; the second of the two affixes in the present tense is the egophoric form (see §2 above). For the asymmetries in the system of special converbs see Sheyanova (2019). Other parallel categories in the two paradigms use the same markers.

There are verbs that lack the perfective stem. When asked to produce perfective forms for these verbs, the consultants suggest a combination of the infinitive with perfective verbs, mostly *--aɁes* 'begin'. These defective verbs denote states and some atelic activities, such as *izes* 'be ill', *--iges* 'want', *--ukes* 'itch', *ures* 'rain', *ruržes* 'be shivering' (also 'boil'), *rurqes* 'flow', *--uzes* 'work', *urʁes* 'fight', *--ulqes* 'dance'. Note that some of these verbs show a morphological structure similar to one of the models of the imperfective stem derivation – infixation of *-r-* or *-l-* – and may historically go back to a regular two-stem verb. In fact, *--ulqes* 'dance' is

#### Michael Daniel

identical to the imperfective stem of *--uqes* ~ *--ulqes* 'go, run'. Another defective verb is the bound root \**k'es* (probably related to *uk'es* (ipfv) 'say') that is used in some morphologically complex but unanalyzable verbs.

Some verbs have identical perfective and imperfective stems. These include *umces* 'weigh, measure', *irxes* 'reap', *irc'es* 'weed', *--alces* 'spin (thread)', *--urhes* 'tell', *--uhes* 'scold', *--uʔes* 'be', *--ises* 'weep', *--aˤldes* 'hide' (tr). Note again that some of these verbs have the -V(l/r)C- structure typical of imperfective stems.

There are also several verbs whose imperfective stem is distinct from the perfective stem in that it does not contain the gender prefix slot: (--)*ižes* 'lick', (--)*iˤšqes* 'mow, peel', (--)*ites* 'beat', (--)*igʷes* 'burn'. More generally, there is an asymmetry between perfective and imperfective stem in terms of the presence of the gender agreement slot: imperfective stems may lack it with those verbs whose perfective stems have it, but not vice versa. Cf. the following table:

Table 4: Asymmetries between perfective and imperfective paradigms


The two verbs which exceptionally have gender slots in the imperfective stem but lack it in the perfective stem are *kes* (pfv) ~ *--ukes* (ipfv) 'bring' and *es* (pfv) ~ *--uk'es* (ipfv) 'say, tell', both of which are morphologically irregular. The latter verb may be considered two separate lexical items ('say' and 'tell').

There are several highly irregular verbs, all shown in Table 5. Note that, again, with 'see' and 'give', the imperfective stems show one of the regular patterns of imperfective stem formation (see below) and are similar to their perfective stems, so that they represent a case of weaker suppletion than fully irregular 'say' and 'go'.

Table 5: Aspectual stems of the irregular verbs


The attested patterns of the connection between the perfective and the imperfective stems are summarized in Table 6. The choice of the pattern is not fully

independent of other formal properties of the verb, first of all the perfective past formation and/or the presence of labialization (a labialized final consonant or *u*); see the explanations below the table.


Table 6: Patterns of aspectual stems formation

Infixation of *-l-* (18 verbs) is attested in all inflectional classes, while infixation of *-r-* (seven verbs) is present in five simple verbs, four of which are labialized (aorist in *-ub*). The model VlC ~ luC is typical specifically of the verbs with aorist in *-un*. Vowel alternation in V(C)C roots is usually *a-*/*e-* ~ *i-*, with *i-* changing to *u-* in verbs with the aorist in *-un*, *-ur* or *-ub*.

### **5 Conjugation classes and the issue of labialization**

I group Mehweb verbs into three inflectional classes according to the marker of the perfective past they use – *-ib*, *-ur* or *-un*. Most verbs use the *-ib* suffix, which I will consider to be the default; the same suffix is used by verbs of all conjugations with the imperfective stem as the imperfective past, so in fact it may be considered to be simply a suffix (of the secondary derivational stem) of the past, perfective or imperfective, the choice between the perfective/imperfective

#### Michael Daniel

interpretation being, in these forms, fully determined by the aspectual characteristics of the stem. A small additional fourth class is very similar to the 'default' conjugation except that all verbs in this class have labialization on the final consonant of the stem and the aorist marker is realized as *-ub*; it is shown as 1a in the following table. However, not all inflectional properties of this 1a class may be explained as being a labialized variety of the first class; see below. Here are some representative forms:

Table 7: Verbal inflectional classes


In verbs with lexical pharyngealization, the *-u-* of the aorist marker may be realized as *-oˤ-* (on pharyngealization, see Moroz 2019). Cf.:


Labialized stems also exist in the *-un* and *-ur* classes, where the labialization is, however, lost before (absorbed by) the vowel of the aorist suffix. It is also lost in the imperfective forms if the stem vowel changes to *-u-* – apparently, the root vowel absorbs the labialization of the following consonant, including when there is another consonant that comes between the root vowel and the labialized consonant. Depending on the form and class, labialization of the stem is thus realized as labialization of the last consonant of the stem (e.g. in the imperative), labialization of the stem vowel (in various imperfective forms) or labialization of the suffix vowel (in *-ib* of the aorist).

Most verbs with *-ub* in the aorist also have labialization in other forms, so that one interpretation is that *-ub* results from the *-ib* marker meeting the final labialization of the stem. The two verbs that take *-ub* but do not show labialization in other forms – *--oˤrʡ-* 'break' and *--uh-* 'become' – both have *-u-* as the

#### 4 Mehweb verb morphology

underlying vowel of the root (*oˤ* is the result of pharyngealization of *u*). When comparing this to the fact that the *-u-* in the imperfective stem absorbs the labialization of the final consonant, as shown in Table 8, it seems appropriate to posit the deep form of the perfective stem of these two verbs as having the labialized consonant whose labialization changes the aorist marker *-ib* to *-ub* but is itself always absorbed \**--oˤrʡʷ-*, \**--uhʷ-*. Then, all verbs that take *-ub* in the aorist have final labialization. On the other hand, none of the *-ib* verbs has a labialized final consonant.


Table 8: Labialized stems

Given this evidence, it seems that the *-ub* conjugation should merely be considered a formal subtype of the *-ib* conjugation. However, the conjugation of the *-ub* and *-ib* verbs diverge in two important points. First, both the aorist marker *-ib* and the homophonous imperfective past marker on all verbs lose the final consonant when followed by *-ra* in egophoric forms or the perfective converb marker *-le*. With *-ub*, both forms keep the final *-b*. Second, the *-ib* in the imperfective paradigm does not change to *-ub* after a labialized stem – something which we would expect assuming that *-ub* in the perfective paradigm results from …ʷ+*-ib*.

Table 9: Divergence between the default *-ib* and the *-ub* conjugations


In other words, the suffix *-ub* shows morphophonological behavior which is significantly different from *-ib.*

Whatever the ultimate interpretation of the *-ub* aorist should be, it seems that this inflection type shows a position intermediate between a separate conjuga-

#### Michael Daniel

tion class and a subtype of the default. The full list of the attested labialized stems for all conjugations is as follows (in the aorist form): *--eˤʡub* 'seed', *--erkun* 'eat', *gub* 'see', *ihub* 'throw', *--alk'un* 'take fire', *--igub* 'burn', *ik'ub* 'put on', *--erhun* 'slaughter', *--usaˤʡun* 'fall asleep', *--erʔub* 'dry up', *--aˤʜun* 'get soaked', *--erq'ub* 'become worn', *--erʁub* 'dig out', *--alħun* 'wake up', *--erχur* 'come down'. As explained above, the verbs *--oˤrʡoˤb* 'break' and *--uhub* 'become' are only labialized in their underlying forms.

### **6 Polarity**

Verbal negation is expressed by one of the two prefixes, the standard negation prefix *ħa-* and the volitive negation prefix *m*V-. The latter is only used in volitional moods including the prohibitive (negative imperative) and negative optative, and the former is used elsewhere, both on finite and non-finite forms. Some speakers allow using *ħa-* in negative optative forms. The standard negation *ħa*is, however, never used in prohibitive forms.

In periphrastic verbal forms, both the lexical and the auxiliary verb may be negated. The standard negation *ħa-* is placed immediately before the verbal stem, thus following the preverb with preverbal verbs. The full pre-root template of the verb is shown in the following example:

(6) *har-ħa-d-uq-un*. pv-neg-f1-flee:pfv-aor 'She did not run away.'

Some of the negative forms of the verb *--ak'-as* 'come' are given in Table 10 as an example. As masculine forms morphophonologically interact with the prefix (see below), feminine (more specifically, F1 – girls gender) forms are given instead.

The forms are morphophonologically straightforward except on vowel initial bases, including those resulting from the elision of the masculine prefix*w-*, where the vowel *-a* of the prefix interacts with the initial vowel of the stem. The elision of the masculine prefix *w-* occurs after all prefixal elements, including the standard negation prefix itself. After this, the following processes occur:

(7) initial a- or e- of the base is dropped:

$$\begin{aligned} \text{ha+aC...} &\rightarrow \text{ha-C...} \\ \text{ha+eC...} &\rightarrow \text{ha-C...} \end{aligned}$$

(8) initial i → j:

ħa+iC… → ħa-jC…


Table 10: Some negative forms of *--ak'as* ~ *--ik'es* 'come'

(9) …and then dropped before a consonant cluster:

ħa-jCC → ħa-CC…

(10) initial u → w:

ħa+uC… → ħa-wC…

(11) …and then dropped before a consonant cluster leaving (probably optionally) labialization on one of the consonants:

> ħa-wCC → ħa-C( ʷ )C ( ʷ )

This labialization may only result from the initial *u-* of the root, not from the masculine prefix *w*-, which is dropped after a prefix, leaving no trace. Cf. the following forms with different types of anlaut (masculine forms are given for the verbs with the initial gender agreement slot):

Table 11: Standard negation on verbal stems with and without gender prefix slot


#### Michael Daniel

The same processes apply to the optative forms when they use the standard negation marker, cf.:



Attested forms of negation in periphrastic forms use the negative auxiliary *agʷara*:

(12) negation in periphrasis:


The morphophonology of the forms with the dedicated volitive negation (negvol) marker is different. The prohibitive and the negative optative forms both take the same consonantal prefix *m-* (*m*V- before consonants) but two different suffixes. The masculine prefix *w-* is lost after the negative volitional *m-*. When followed by a consonant, either a gender prefix or the initial consonant of the stem, the negative volitional copies the stem vowel. Finally, the neuter/human plural prefix *b-* is assimilated by the negative volitional and is represented by *m-*.

(13) morphophonology of the negative volitional prefix:


#### 4 Mehweb verb morphology

As (13c) also shows, the process of stem copy (see below) applies before assimilation in nasality takes place.

As to the suffix position, the negative optative and the prohibitive have different suffixes. The negative optative takes the suffix *-ab*, same as the positive optative; the prohibitive takes a dedicated suffix *-adi*, whose final vowel is optionally dropped. In both cases, the initial *-a-* of the suffix is analyzed below as a marker of a secondary derivational stem termed irrealis (see next section). The following table shows the prohibitive of verbs with different stem structures.

Table 13: Volitional negation with different stem structures


The prohibitive frequently appears with what looks like reduplication; more specifically, a full copy of the stem together with the gender marker is placed to the left of the negative volitional prefix. The process is optional.

(14) stem copy in the prohibitive:

```
d-iz ~ mi-d-iz-ad
f1-wash:ipfv ~ negvol-f1-wash:ipfv-proh
(also mi-d-iz-ad)
     negvol-f1-wash:ipfv-proh
'Do not wash her!'
```
Outside its use in the prohibitive, stem copy is relatively common in the context of standard negation and elsewhere with a certain added expressive or pragmatic value (cf. Maisak 2012 on similar processes in other East Caucasian languages). Note that the stem copy shows the underlying form containing the masculine prefix, not the copy of the actual realization of the stem in this specific context. This is seen in standard negation involving stem copies; cf. (15) and (16):

#### Michael Daniel

(15) stem copy in standard negation: *w-ak' ~ ħa-k'-ib-i* m-come:pfv ~ neg-m.come:pfv-aor-atr *d-ak'* ~ *ħa-d-ak'-ib-i* f1-come:pfv ~ neg-f1-come:pfv-aor-atr 'the one who did not come'


The process is not reduplication sensu stricto. I call it stem copying. Structurally, the copy of the stem may be separated from the verb form by other material, especially by the discourse particle, as in (17) and (18), where it forms a separate wordform.


'When we arrived there, Buga Muhammad took me on his lap; having taken me, he examined (me).'

In all contexts stem copying is optional. However, it is in the prohibitive that these forms are very consistently produced as first translations of the Russian stimuli with the relevant meaning. It seems that expressive pragmatics of stem copying is being incipiently grammaticalized in the expression of the prohibitive. 4 Mehweb verb morphology

### **7 Synthetic paradigm**

This section gives an overview of the synthetic paradigm of the Mehweb verb. A summary table is provided at the end of the section. Polarity, gender and egophoric subject agreement and aspectual stem formation have been discussed above.

The derivation of forms is summarized in the following figure. For some more exceptional patterns, including derivation of special converbs from general converbs or from the infinitive stem, see Sheyanova (2019). (An asterisk shows morphologically bound bases.)

Figure 14: Derivation of verbal forms

The aspectual stem immediately derives the past (aorist in the perfective, imperfective past in the imperfective paradigm; note that the forms further derived from this secondary stem, e.g. converbs or participles, do not necessarily have past reference), present habitual (in the imperfective stem only), infinitive, the imperative, the nominalization in *-ri*.

Several other forms are based on a bound (hence the use of the asterisk) base produced by adding *-a-* to the aspectual stem; this base may be considered the base of irrealis (potential in terms of Nina Sumbatova, unpublished), because it produces such forms as optative, conditional, apprehensive, counterfactual and some other (see Dobrushina 2019). Support for this analysis not confirmed diachronically by the data from other Dargwa lects, comes from the counterfactual form in *-are*, one of the irrealis series, segmentable into the irrealis marker *-a-* and the past marker *-re*. The latter is attested elsewhere, including on the auxiliary in the past forms (*lewre* and *agʷire*) and probably elsewhere (*--igibre* from *--igib* 'want' Ipft) – see Dobrushina (2019). Note the morphophonological difference between counterfactual *-re* and the egophoric *-ra* – the latter causes the past marker *-ib* to drop the final *-b*, while in the counterfactual *--igibre* it is preserved, just as in the egophoric forms of the verbs in the *-ub* subtype.

The general converb and the participle are formed differently in the perfective and the imperfective paradigms. In the perfective, the attributive marker *-i*(*l*) and


Table 15: Verbal inflection

the converb marker *-le* are added to the aorist. In the imperfective, the participle marker *-ul* and the converb marker *-uwe* are added directly to the imperfective stem. While the *-l* of the imperfective participle marker*-ul* is always present, that

#### 4 Mehweb verb morphology

of *-i*(*l*) is often dropped, and the distribution of the variants is not clear (though it seems that at least in the predicative use of the participle in *-i*(*l*) the full variant is impossible).

It seems plausible to differentiate between *-ul* as the participle marker proper, used only with the imperfective stem of the verb, and the attributive marker *-i*(*l*), attached to the aorist but also used on infinitives (to form future participles, also used finitely), auxiliaries (to form periphrastic participles) and adjectives. Note that the imperfective converb ending *-uwe* is more or less straightforwardly analyzable into *-ul-le*, where *-le* is a general converb marker (also in the perfective paradigm) and, more generally, is used as a cross-categorial adverbializer, i.e. in forming adverbs from adjective roots.

Special converbs may be based on the general converb form, as the causal converb *-na*, or on the participle, as anterior converb *-*(*j*)*aʁle*; see more on special converb formation in Sheyanova (2019).

Unlike the nominalization in *-ri*, which is formed directly from the aspectual stem, nominalization in *-deš* is formed from many forms, including finite past, future, present (habitual), participles – but not from volitional forms and not from the nominalization in *-ri*. Given that *-deš* is also attached to adjectives and nouns, the generalization seems to be that *-deš* is not a derivational morpheme but a cross-categorial predicate nominalizer. The suffix does not combine with egophoric forms.

Table 15 summarizes synthetic verbal inflection. Forms are given without gender agreement marking; for gender agreement see §1. The negative prefix may attach to all forms in the table (except the imperative); morphology of polarity marking is discussed in §6. The marker *-na* is the marker of the plural of the addressee in volitional forms.

### **8 Imperative and infinitive**

Both the imperative and the infinitive are formed from each of the two stems. While in the imperfective paradigm the suffixes are invariably *-e* and *-es*, respectively, the perfective imperative and the perfective infinitive / perfective nonegophoric future both have two markers (*-e* vs. *-a* in the imperative, *-es* vs. *-as* in the infinitive). The choice of the allomorph in the two categories is independent.

The choice of the imperative vowel depends on the transitivity of the verb: transitive verbs take *-a* and intransitive verbs take *-e*. Cf. *--urs-a* 'pound', *--iʡ-aˤ* 'steal', but *--alħʷ-e* 'wake up', *--uq-e* 'go'. Note that the choice of the marker is primarily based on transitivity rather than control, as e.g. motion verbs all take *-e*.

#### Michael Daniel

Table 16: Imperative and infinitive suffixes


P-labile verbs (i.e. verbs that are used with and without agentive argument) take *-e* or *-a* depending on the interpretation; cf. *w-aˤld-e* 'hide (intr)' (to a man) vs. *w-aˤld-a* 'hide it'. Other labile verbs also show similar behavior; cf. *abx-a* 'open (it)' vs. *abx-e* 'open (intr)'; *b-oˤrʡ-a* 'break (it)' vs. *b-oˤrʡ-e* 'break (intr)'. Although in these cases the intransitive imperative might seem unlikely, it is readily interpreted by my consultants as when talking to something that resists acting on it, does not yield, or seems to take too long to achieve the result. There is evidence that A-labile verbs (i.e. verbs that may omit the patientive argument ascribing nominative to the agentive argument) may also take both markers; cf. *--erq-a* 'suck (e.g. milk)' vs. *--erq-e* 'suck' (implicit, out-of-focus patient).

Experiential verbs do not behave in a unified way. Generally, they prefer the intransitive suffix, but some also allow the transitive one, without a clear meaning shift; cf. *qumart-a* and *qumart-e* 'forget', *--ah-e* and *--ah-a* 'know'. One would expect an interpretation with the imperative subject's increased control over the situation but this is certainly not consistent through all the experiential verbs, though some consultants do report this difference e.g. in the verb *--arg-e* vs. *--arg-a* 'find'. The verb *gʷes* 'see' does not form a generally accepted imperative, but if it does, the form is *gʷ-a*.

There is no alternation in the imperfective imperative. A possible way to account for this would be to consider all imperfective imperatives as using the intransitive imperative suffix, which would amount to transitivity decrease with obligatory promotion of the Agent. Imperfectives are crosslinguistically more Agent-oriented forms. In an ergative language like Mehweb, promoting the Agent would show up as decrease in transitivity. The assumed promotion is, however, internal to verbal morphology and does not change argument marking. P retains nominative case, and A, if present, is marked with ergative.

The imperative of the verb 'give' has two perfective stems, *aga* and *--ega*, depending on the person of the recipient. The first stem is used when the recipient is the first person, otherwise the second stem is used. Both stems are suppletive with respect to the non-imperative stems, and the second stem additionally introduces an agreement prefix slot. This pattern or the verb 'give' is attested elsewhere in Dargwa and in East Caucasian at large (see Comrie 2003, also Daniel

#### 4 Mehweb verb morphology

et al. 2010). Another verb with an irregular imperative stem is *es* 'say' (inf) – *bet'a* 'say' (imperative). The verb *--uˤq'es* 'go' has two imperatives, the regular *--uˤq'-e* and the irregular *--eˤʡ-e*. The semantic distinction is not fully clear but probably has to do with the final point, the first better translated as 'go there' and the second as 'go away, leave'. Irregular imperatives only exist in the perfective paradigm.

The forms*--eg-a* 'give (away)' and *--eˤʡ-e* 'go (away)' contain the expected imperative suffixes (transitive and intransitive, respectively). On the other hand, their stems are not present elsewhere in the paradigm of these verbs, and neither can they be causativized. The second form is also fully suppletive to all other stems of *--uˤq'es* 'go'. They are thus close to the status of separate lexical items – imperative interjections. This becomes clear when they are compared to another suppletive imperative stem, *bet'-a* (from *es* 'say'), which has a clearly different morphological status. First, it is the only imperative available for this verb. Second, the stem *bet'-* is optionally used in other forms (see Table 18), including causatives (see Table 21).

Imperatives show plural marking based on the number of the addressees (thus showing, formally, an accusative pattern of agreement). Unlike in the prefix slot – and, for that matter, anywhere in Mehweb – this marking is independent from the gender. The suffix is *-na* and it is regularly attached to the imperative marker as well as to the irregular imperatives except in the verb *--aš-e* 'come here' vs. *--aš-ina* 'come here' (plural addressee). Cf.:

(19) intransitive imperative


On imperatives in Mehweb, see more in Dobrushina (2019).

The choice of *-es* vs. *-as* in the perfective infinitive/non-egophoric future forms, on the other hand, seems to have a purely formal motivation. The default marker

#### Michael Daniel

is clearly *-es*, while *-as* is only attested in about twenty verbs which (a) have *-a*as a stem vowel that is (b) followed by a stem final glottal, pharyngeal, uvular or velar consonant; cf. *--usaˤʡʷ-as* 'fall asleep', *--aʔ-as* 'begin', *--ah-as* 'know', *--aˤʜʷaˤs* 'get wet', *aq'-as* 'pour', *--aχ-as* 'nurture', *--ak-as* 'smear'. Neither (a) nor (b) alone seem to require *-a-* as the vowel of the infinitive; cf. *--uˤq'-es* 'go' (condition b but not a) or *--ac'-es* 'melt' (condition a but not b).

There is a number of verbs where the consonant of the required place of articulation is separated from the -a- of the stem by another consonant. In these cases, the default seems to be *-es*, including *ask'-es* 'catch on', *--alk'ʷ-es* 'burn', *abx-es* 'open', *--arx-es* 'send', *--arχ-es* 'touch', *--aˤlq'-es* 'rinse', *--alħʷ-es* 'wake up', *--aˤld-es* 'hide'. However, some verbs, including *--aˤlq-aˤs* 'give harvest', *--aˤbʡ-as* 'kill', *--aˤrʡ-as* 'freeze', *--aˤrʜ-as* 'copulate' do choose *-a-* as the vowel of the infinitive.

### **9 Auxiliary**

Mehweb verbal inflection heavily relies on periphrasis. Periphrastic forms are used e.g. to form progressive/durative or resultative/perfective forms (combination of a converb with the auxiliary), future (combination of the infinitive with the auxiliary) and others. There are some periphrastic forms based on auxiliary use of the verb *--uɁes* 'be' (Pfv--Ipfv), but most forms in the periphrastic paradigm use one of the auxiliaries in Table 17. Complex forms (surcomposé) are also attested, using the auxiliary, the second auxiliary in a converb form and yet another converb of the lexical verb.

Periphrastic forms are also used to form jussive (combination of the imperative of the lexical verb with the imperative of the verb 'say'; see Dobrushina 2019) and perfective forms from defective verbs that only have the imperfective stem.

The same verb is also used in locative and existential predication. Inflection of the auxiliary is presented in the following table:


Table 17: Inflection of the auxiliary

The form *sabi* is included in the list but has a very marginal status in Mehweb. If used at all, it has the status of a particle rather than of a true auxiliary/copula. It is clear that the *-b-* of the stem, etymologically a gender marker, has been fossilized.

Some forms, such as the converb of imminence, are not attested. Other special converbs are well-formed: *le--ijaʁle, sabijaʁle, agʷirijaʁle* (but apparently not *aħinijaʁle*), causal *le--lena*, *agʷarlena*, concessive *le--leʡur* and *agʷarleʡur*, additive *le--lera* and *agʷarlera* etc. Nominalizations such as *le--deš*, *le--ideš*, *sabideš, aħindeš, agʷiredeš*, *agʷarideš* etc. are easily produced.

### **10 Irregular verbs**

There is a number of irregular verbs, including especially motion and caused motion verbs. Several irregular verbs show irregularly short roots, consisting only of one consonant. In the case of *es* 'say' it may be argued that it has a zero stem in the perfective. With the exception of the bound verb \**k'es* (cf. *uruχ k'es* 'be afraid of'; the verb itself is probably historically a reduced version of the imperfective of *--uk'es* 'say, tell' Ipfv), all these verbs are irregular in the perfective stem, while their imperfective stem fits one of the regular patterns of stem formation (cf. *lug-* 'give' and *luk-* 'saw', *irgʷ-* 'see' and *irk'ʷ-* 'put on', *uk'-* 'say' and *uk-* 'eat').

Note the marker of nominalization, usually *-ri*, is *-ari* on verbs that lack any vowel of the stem (*ari* for 'say', *gʷari* for 'see', *gari* for 'give'), and the presence of two different imperatives of 'give' – 'give to me' and 'give to someone else'. The inclusion of the stem *-uk'-* as the imperfective counterpart to the verb *es* 'say' is controversial. The two stems differ in transitivity, the former being intransitive and the latter transitive, so that they may be considered as separate lexical items. However, *--uk'es* is not an equivalent of 'talk (with/to)' but is an imperfective counterpart of *es* 'say'. In the perfective, it lacks any segment at all except in the imperative and irrealis series that share the stem *bet'*, which is optional in irrealis forms.

Further, there are several highly irregular motion verbs. The first one is the basic verb of motion, *--aˤq'-*(*un*) ~ *--aš-* 'go', a non-ventive verb. In both perfective and imperfective subparadigms, two different stems are present. In the perfective, these are *--aˤq'-* (the participle and forms based on the participle stem, including aorist and general converb) and *--uˤq'* (imperative, infinitive, future, forms based on irrealis *a*-base and the action nominal). These are stems distributed between different perfective forms.


Table 18: Inflection of the irregular verbs

In the imperfective, in addition to the stem *--aš* that possesses the full range of forms, there are several forms based on the stem *q'ˤ-*. Attested are the synthetic present forms, the conditional form, the action nominal, the participle and the general converb; probably, there are other, unelicited forms. Unlike other stems, these forms lack the gender prefix altogether. The regular perfective *--aˤq'uwe* designates andative situations ('go away from here') and implies absence of the subject at the place of speech ('he is gone'). The converb *q'uˤwe* is imperfective and designates an actual ventive situation ('he is coming'). The converb *--ašuwe* is also imperfective but conveys multiple or habitual situations. The perfective situation 'he has come' is conveyed by the perfective converb of the regular verb *--ak'es*.

A similar meaning (probably implying that the situation of coming is visually attested) is conveyed by the present forms *q'aˤn* (non-egophoric) and *q'aˤs* (egophoric); unlike other synthetic presents that (at least tend to) have nonepisodic (habitual) interpretations, these forms seem to be progressives. The same irregularities are observed in the andative verb *ʡaˤr--aˤq'-*(*un*) (*ʡaˤr--uˤq'-*, *ʡaˤr-q'ˤ-*) ~ *ar--aš-*, which is a derivation of *--aˤq'-*.


Table 19: Inflection of the motion verb *--uˤq'es*

Further, there are two perfective imperatives. The difference between them is not very clear but is probably correlated with telicity (the presence or absence of the final point of motion), as in 'go away, leave!' (*--eˤʡe*) and 'go there!' (*--uˤq'e*). The imperfective imperative is interpreted either as a multiple going event (habitual interpretation, as 'be visiting them!') or as a single ventive imperative event (as 'come here!'). Single andative imperative events require the use of the perfective imperative.

As to the caused motion verbs, there are two series of forms, one based on *k-*, the other on *χ-*. To the best of my knowledge, the two series of forms are strictly parallel and designate bringing/fetching events, the difference essentially being between fetching or bringing animate entities (*k-*) vs. bringing inanimate entities (*χ-*). I will further gloss them conventionally as 'lead' vs. 'bring', though the contrast is not identical to the contrast between *lead* and *bring* in English. In both series, the monoconsonantal base expresses the meaning of ventive (*k-* and *χ-*) and is perfective, the *--u*C- with a gender prefix slot is perfective and elsewhere-oriented (*--uk-*, *--uχ-*), and the *--i*C base with a gender prefix slot is imperfective and orientation neutral (*--ik-*, *--iχ-*). The strictly andative meaning 'lead/bring away from here' is expressed by a verb with a prefix (*ar--uk-* ~ *ar--ik*-; *ar--uχ-* ~ *ar--iχ-*).

In a sense, there are two pairs of stems, C ~ *--i*C and *u*C ~ *--i*C, with two perfective stems sharing one imperfective counterpart. However, similarly to the 'plain' motion verbs (see above), the relation between the stems is probably different from that in other perfective ~ imperfective stems. The *--i*C stem seems to convey the meaning of multiple events while the C and *--u*C stems designate single events. As a result, the monoconsonantal verb behaves irregularly in that it has two converbs, perfective *kile* and several specifically imperfective forms, including the imperfective converb *kuwe*, general present forms (with actual interpretation) *kas* (non-egophoric) and *kan* (egophoric), and the participle (other parallel forms may be present but unelicited). Unlike the non-causative motion verb described above, the supplementary episodic imperfective forms *kas, kan, kuwe* (*χas, χan, χuwe*) in the imperfective share the stem with one of the perfective series. A different look at the paradigm would be to consider each of the verbs of caused motion as including two different verbs, the more or less regular Pfv<sup>2</sup> ~ Ipfv<sup>2</sup> and the highly defective Pfv<sup>1</sup> ~ (Ipfv<sup>1</sup> ), probably with the regular verb used as andative and the irregular as ventive – but this needs further research into semantics and usage of motion verbs.

Another irregularity of the caused motion verbs is morphosyntactic: their imperfective stem is A-labile with an antipassive pattern; see the following section.

#### 4 Mehweb verb morphology


Table 20: Inflection of the caused motion verbs *kes* 'bring (animate)' and *χes* 'bring (inanimate)'

### **11 Transitivity**

In this section, I consider several transitivity related issues, first of all morphological causativization, but also change in argument structure or case assignment which is not marked by morphological means – binominative constructions and related lexical phenomena, labile verbs and antipassive verbs. I also briefly consider another type of verbal derivation, typologically rare, probably even limited to (and within) East Caucasian languages – the category of verificative.

The only regular process of valency change in Mehweb is causativization. Periphrastic causativization is weakly grammaticalized in Mehweb; it is based on the verbs *aʔ*(*ib*) ~ *iʔ-* 'drive, cause to go', *--aq*(*ib*) ~ *--irq-* 'let go' and *--aq'*(*ib*) ~ *--iq'-* 'do', and is discussed in detail in Barylnikova (2019). This section limits the discussion to the causativization in morphological and lexical domains. The discussion of morphological causatives heavily relies upon the data collected by Ekaterina Ageeva in 2012 (unpublished field report).

Mehweb verbs are very productively causativized through the suffixation of *-aq-*. The suffix is identical to the perfective stem of the verb *--aq*(*ib*) ~ *--irq-* 'let go'. Grammaticalization of 'let go' into a causative marker is not surprising, but the suffix does not have the agreement slot present on the verb. Even though the slot might have been lost in the process of grammaticalization, the suggested path remains somewhat speculative. The suffix may combine both with the perfective and imperfective stem, so that each form present in the paradigm of

#### Michael Daniel

the original, non-causative verb, also has its causative counterpart. Note that all causative verbs follow the *-ib* inflectional class in the aorist, independently of the inflectional class of the lexical verb: *--aˤʜun* 'get wet' – *--aˤʜaˤqib* 'cause to get wet', *--arcur* 'get stuck' – *--arcaqib* 'cause to get stuck'; just as *--ac'ib* 'melt' *--ac'aqib* 'cause to melt'; labialized verbs preserve labialization: *--erq'ub* 'tear apart' ~ *--erq'ʷaqib* 'cause to tear apart'. In a periphrastic form, the lexical verb but not the auxiliary is causativized:

(21) *b-aš-aq-u-we* hpl-go:ipfv-caus-cvb.ipfv *le-b-re*. aux-hpl-pst 'He made them go (repeatedly).'

Causatives are formed from verbs with all types of argument structure, including intransitive, experiential and transitive; cf.:


let me go before I ate something.'

The causative from the ditransitive verb *g*(*ib*) ~ *lug-* 'give' is not attested in the corpus but is well-formed. It is, however, morphologically irregular, as with several other verbs with monoconsonantal stems. These verbs form causatives by adding the suffix *-aχ-*.

The verb *es* 'say' forms the imperative from each of its two perfective stems (see Table 18), *a-* (*aqaqib*) and *bet'-* (bet'aqib), both meaning 'caused to say'.

#### 4 Mehweb verb morphology

*g*(*ib*) 'give' *g-aχaq-ib* 'caused to give' *g*(*ub*) 'see' *gʷ-aχaq-ib* 'caused to see' *χ* (*ib*) 'bring' *χ-aχaq-ib* 'caused to bring' *k*(*ib*) 'lead' *k-aχaq-ib* 'caused to lead'

*i-b* 'say' *aqaq-ib* 'caused to say'

Table 21: Irregular perfective causatives

Caused motion verbs with irregular paradigm structure (see Table 20 above) apparently form causatives from all three stems; cf.:

### (25) causatives of caused motion *--uχes* 'bring'


away-m.bring:pfv-caus-fut.ego

'I will cause you to be brought away (by the river).'

c. *ar-m-iχ-aq-adi* away-negvol-m.bring:ipfv-proh

'Let (the river) not bring me away!'

The non-caused motion verb *--uˤq'es* does not form the causative from its short stem *q'-* (see Table 19); the two available forms are formed from the stems *--uˤq'-* (perfective) and *--aš-* (imperfective):

(26) causatives of motion verb *--uˤq'es* 'go'


'cause (them) to go' (perfective causative infinitive)

c. *b-aš-aq-uwe*

hpl-go:ipfv-caus-cvb.ipfv

'making them come again and again' (imperfective causative converb)

Irregular causatives in the imperfective are not attested.

#### Michael Daniel

Morphologically possible and accepted by many speakers are double causatives (noted in Ageeva 2014). In some cases, the forms convey the compositional meaning of double causation (27), but sometimes consultants interpret them as single causatives (28). Double causatives are not attested in the corpus; elicited examples include:

	- a. *b-elʁ-aq-aq-ib* n-eat.full:pfv-caus-caus-aor 'made someone feed (an animal)'
	- b. *b-erc'-aq-aq-ib* n-fry:pfv-caus-caus-aor 'made someone fry (it)'
	- c. *d-aˤʜʷ-aˤq-aq-ib* npl-get.wet:pfv-caus-caus-aor 'made someone get them (feet) wet'
	- d. *b-alk'ʷ-aq-aq-ib* n-burn:pfv-caus-caus-aor 'made someone get (it) burning'
	- e. *b-arχ-aq-aq-ib* n-touch:pfv-caus-caus 'made someone touch (it)'
	- f. *b-ac'-aq-aq-ib* n-melt-caus-caus-aor 'made someone melt (it)'
	- a. *d-alħ-aq-aq-ib* f1-wake.up:pfv-caus-caus-aor 'woke her up'
	- b. *w-aˤrʡ-aq-aq-ib* m-freeze:pfv-caus-caus-aor 'made him freeze'
	- c. *w-aˤbʡ-aq-aq-ib* m-kill:pfv-caus-caus-aor 'made someone kill him'

#### 4 Mehweb verb morphology

The semantic contrast between double causatives in (28) and the respective simple causatives --*alħaqas* 'cause to wake up' etc.) is unclear, if it exists at all. Except (28c), all verbs in (27) and (28) are intransitive. (The verb *--arχ-es* 'touch' means literally 'something touched on something', with a natural interpretation of getting one's hand in contact with something. The full meaning of the form in (27e) is thus 'caused someone<sup>i</sup> to cause one<sup>i</sup> 's hand to contact something'.) These are all double causative forms elicited by Ageeva. From a comparative East Caucasian perspective, all these meanings tend or may be labile; and some are also labile in Mehweb (e.g. (27d). This provides a tentative explanation of why double causatives may be limited to these verbs. A simple causative from a labile root is usually interpreted as a causative of its intransitive rather than transitive meaning (schematically, 'burn (tr/intr)' → 'burn (intr)'-caus (tr)). In such uses, the causative suffix does not derive a new transitive meaning but emphasizes the transitive semantics already present in the lexical meaning of the labile verb as one of its possible interpretations. It may be considered as a disambiguation mechanism for interpreting a labile root as expressing specifically transitive meaning. As this causative suffix does not have exactly the same function as regular causativization, it allows for a second marker which serves as a regular causative derivation.

The semantics of the simple causative forms, on the other hand, is always compositional, unless the whole causative derivation is lexicalized. On the special use of the causative in optative constructions see Dobrushina (2019). Examples of lexicalized causatives are, e.g. *--aʔ-aq*(*ib*) 'bring back' and also 'hit' - cf. *--aʔ*(*ib*) 'reach' (the latter probably from 'reach with hand', lit. 'cause the hand to reach'), *--ik-aq*(*ib*) 'put right' (of a joint etc.) – cf. *--ik*(*ib*) 'happen' (probably from 'fall', thus 'make fall in place') etc.

Some verbs are equally available in transitive and intransitive constructions without any morphological marking of the (de)transitivization on the verb. There are two known types of labile verbs, P-preserving labile verbs and A-preserving labile verbs. Note that lability is strictly lexical and limited to small classes of verbs. Additionally, there is a phenomenon formally similar to A-labiles that includes one verb that may be called lexical antipassive.



In other words, in comparing intransitive uses of these verbs to the transitive ones, P-labiles suppress their A-argument; A-labiles lose their P-argument and re-assign nominative marking to the A-argument; and, finally, antipassives re-assign nominative marking to the A-argument without suppressing their P-argument but demoting it to an oblique slot.

With P-preserving labiles, the problem is that, in an ergative language with pro-drop, it is hard to distinguish between a transitive verb with an omitted A-argument and intransitive use of a labile verb. Cf. their schematic representation in English:


Although, in my experience, the speakers easily distinguish between the availability of the intransitive reading with labile verbs and pro-drop with strictly transitive verbs (e.g. by translating into Russian and using mediopassive for the former and a non-referential third person plural for the latter, or else adding *it happened all by itself* vs. *someone did it*), some kind of formal diagnostic may also be used. This diagnostic is provided by the morphological distinction between transitive and intransitive imperatives in the perfective paradigm. I thus classify a verb as labile if it is judged grammatical with both imperative endings. The following labile verbs are attested (note that the speakers' judgements do not seem to be fully consistent):


(35) *--oˤrʡ*(*oˤb*) ~ *--oˤʡ-* 'break'


#### 4 Mehweb verb morphology

The labile verbs designate situations that may proceed unsupervised (such as cooking events), may both be carried out on purpose or occur spontaneously (such as breaking or opening/closing events) or may involve both non-human/inanimate (thus non-intentional) or human undergoers (such as 'hide' or 'gather'); on the semantics of lability in East Caucasian, see Haspelmath (1993); Daniel et al. (2012).

Another test that could have been applied to Mehweb labiles is marking of egophoricity. Because personal agreement works on the accusative rather than ergative basis, after the A-argument is suppressed, the remaining P-argument controls personal agreement on the verb (see Section 3.1 in Ganenkov 2019). However, I have only applied the imperative test. Note that both tests are applied to labile verbs with some difficulty, or not equally well to all of them. Most labile verbs, in their intransitive uses, typically take inanimate subjects and thus are not compatible with first and second person subjects and are not easily compatible with imperatives. In the latter case, the speakers envisage a situation of urging a process to proceed (see Dobrushina 2019) – and most of them very easily accommodate to this interpretation.

No special study of semantics of the transitive/intransitive pattern alternation with labile verbs has been carried out. The following two examples from the text indicate that, in some cases, it may be connected to the absence of the agent, the usually transitive situation proceeding in a spontaneous way:

(42) intransitive (Corpus)

*urx-ne* key-pl *q'-aˤb-ib* pv-close:pfv-aor *k'ʷan,* qot *unza* door *ʡaj-k'-i-le* lock-lv:pfv-aor-cvb *b-ik-ib*. n-happen:pfv-aor 'The lock has locked itself, the door closed (=locked).'

(43) transitive (Corpus)

*abaj* mother *hil-l-ix-i-le* pv-f-lie.down:pfv-aor-cvb *r-arg-i-ra,* f-find:pfv-aor-ego *unza--ra* door--add *ʡaj-k'-i-le,* lock-lv:pfv-aor-cvb *hil-l-ix-i-le* pv-f-lie.down:pfv-aor-cvb *r-arg-i-ra* f-find:pfv-aor-ego *abaj*. mother

'I found (my) mother already gone to bed – I discovered that, having locked the door, she lay down.'

#### Michael Daniel

Note that, in these examples, there is no direct morphosyntactic evidence of transitive vs. intransitive use. It is only the context that suggests these readings. In (42), the agent is truly absent. In (43), it is omitted in the converb clause ('having locked the door') under co-reference to the subject of the main clause ('mother went to bed'). The first episode describes a situation of spontaneous locking of the door, leaving the master of the apartment, unexpectedly, outside the door and unable to go inside. The second episode tells how the narrator, coming home quite late, discovered her mother already asleep, and the door locked (apparently, by her mother, prior to going to bed). Very often, however, the division of labour between transitive and intransitive constructions with labile verbs in East Caucasian languages is more complex, so this needs further research.

In Mehweb, most experiential verbs are intransitive, with the experiencer marked by the inter-lative case. Some of these verbs take either the transitive or intransitive imperative suffix (e.g. *--arg*(*ib*) ~ *--urg-* 'find'; *--ah*(*ur*) ~ *--alh-* 'know'; *qum-art*(*ur*) ~ *-urt-* 'forget'). For two verbs, this correlates with a change in argument marking – the experiencer changes from inter-lative to ergative, and its agentivity increases ('know' = 'learn (so as to know)', 'forget' = 'try to forget' – see Ganenkov 2019).

A-preserving labiles are less prominent in Mehweb and, generally, in East Caucasian, and were not collected systematically, although, in principle, the same imperative test could have been applied. It seems that the following is an example of a verb that can be used both intransitively and transitively while preserving its A-argument: *--erq*(*ib*) ~ *--uq-* 'suck (intr and tr – e.g. milk)'.

Finally, two caused motion verbs *k*(*ib*) ~ *--uk*(*ib*) ~ *--ik*(*ib*) 'bring (animate object)' and *χ* (*ib*) ~ *--uχ* (*ib*) ~ *--iχ* (*ib*) 'bring (inanimate object)' exceptionally follow the antipassive pattern of valency change. The verb is primarily transitive, but, exclusively (or at least preferably) in the imperfective, it can also be used with the A-argument in the nominative and the P-argument in the ergative.


4 Mehweb verb morphology

This pattern, to the best of my knowledge not documented in other Dargwa varieties, was independently confirmed by several consultants.

Some morphologically simple verbs may be considered to be 'lexical causatives' with respect to other simple verbs – i.e. forming pairs of verbs whose mutual relation is more or less similar to that in causative pairs but whose stems are not morphologically related. The list cannot be exhaustive because it largely depends on what pairs one considers to be in causative correlation, but, in a language with highly productive causative derivation, lexical causatives are not expected to be many. One example is *--ebk'*(*ib*) ~ *--ubk'-* 'die' – *--aˤbʡ*(*ib*) ~ *--iˤbʡ* 'kill'; the other, already much more questionable, is *q'ˤ-* ~ *--aˤq'*(*un*) ~ *--aš-* 'go' – *k*(*ib*) ~ *--uk*(*ib*) ~ *--ik*(*ib*) 'lead'.

The last phenomenon related to transitivity is the binominative (alias biabsolutive) construction. In Mehweb, as in some other East Caucasian languages, including the languages of the Dargwa branch, periphrastic constructions license nominative marking of both A- and P-arguments. Binominative constructions are only available in periphrastic forms based on imperfective converbs (see Ganenkov 2019).

(46) binominative construction (Corpus)


'He (lit. this one) squatted and was tying (his) daughter's shoelace.'

The alternation between the expected ergative ~ nominative and the binominative pattern in the periphrastic transitive construction has been noticed and discussed by Magometov (1982: 84ff.) The semantic effect that the binominative construction brings remains unclear; in fact, Magometov suggests that, in Mehweb, it is the binominative construction that is more natural in imperfective periphrasis. For further discussion of the syntax of binominative constructions in Mehweb, see Ganenkov (2019); Lander (2019).

Finally, I provide some examples of what has come to be called, in recent research on East Caucasian, the verificative construction. This construction has not been controlled in elicitation; the only and few examples that I have come from the corpus. The verificative construction based on a verb P is a complex predicate whose meaning is, speaking formally, 'verify whether P is true' or 'check what/who is x such that P(x) is true', where x is the argument of P – see the examples below. The verbal complex essentially includes two elements – the lexical

#### Michael Daniel

verb followed by the interrogative particle followed by a more or less grammaticalized form of the verb 'see'; literally, 'P-whether-see'. This construction has been previously attested in two distantly related Lezgic languages, Archi (Kibrik 1977: 291) and Agul (Maisak & Merdanova 2004), and later also reported in Chirag by Dmitry Ganenkov. In Daniel & Maisak (2014); Maisak (2016), various properties of the verificative construction are discussed, including that, while various forms may appear in elicitation, the verificative is primarily used in purposive contexts with the infinitive ('in order to check whether…') or in the imperative ('go and check whether…'). These are exactly the forms attested in the corpus; only the copula as the main verb is attested:


In all East Caucasian languages where it has so far been attested, the verificative results from univerbation of the interrogative form of the main verb with the verb 'see'. Our consultants tend to write these forms together in transcription; otherwise, the only formal indication of grammaticalization in Mehweb is the loss of labialization in infinitive verificatives (*gʷ-es* → *-g-es*). In other languages the grammaticalization process is more advanced. To understand the position of the Mehweb verificative with respect to the parameters previously set up for Archi and Agul, further research is needed.

### **12 Complex verbs**

In Mehweb, a verbal stem is a bound morpheme that typically consists of one syllable, followed by one or more inflectional suffixes (an exception being the truncated optative, where no suffix follows; see Dobrushina 2019). Pre-root slots are optional. The presence of a gender prefix is lexically determined – formally identical roots may be different in having or not having a gender agreement

#### 4 Mehweb verb morphology

prefix (cf. *umc-* 'weight (ipfv)' and *--umc-* 'swell (ipfv)'). After the agreement prefix, the next slot to the left is that of the inflectional marker of negation (either standard or volitional). Then may follow a preverbal element. Schematically, this template may be generalized as Preverb-Negation-Gender-Root-Inflection.

I consider the position of the negation prefix to be a diagnostic of a morphologically complex verb – if it is inserted inside what otherwise seems a verbal stem that conveys single verbal meaning, then the morphological element preceding the negation marker is a preverbal part of the verb, however bound it is. For verbs possessing an agreement slot, the position of this slot is another such diagnostic. Cf. the verb *qumartes* 'forget' where neither *qum-* or *-art-* is used without the other part, yet the negation is inserted between them. In *kajʔes* 'sit down', the gender prefix comes after what historically is a spatial preverb.


Unlike negation, positioning of a gender prefix at the beginning of a verbal form does not prove its simplex status, because the preverbal element may have its own gender agreement position. Then, the complex status of a verbal stem is only unambiguously tested by the position of the negation.

(51) 'pull' *bit'ak'ib* (n), *dit'ak'ib* (f1), cf. *bit'-ħa-k'-ib*

There is only one bisyllabic simplex root recorded so far – a root with two syllables not split by negation:

(52) 'fall asleep' *--usaʔ*(*un*) ~ *--usulʔ-*, cf. negative *ħa-wsaʔun*

While many East Caucasian languages use some more or less bound preverbal morphemes, some but not all of them also have a more or less substantial set of true preverbs (derivational verbal prefixes). Preverbs constitute a specific subclass of preverbal elements in that they combine with several verbal stems – first of all, motion and posture verbs, and have an isolatable meaning – often, spatial. While many Dargwa languages possess a considerable inventory of preverbs, in Mehweb they all ceased to be productive, so that many verbs with preverbs ended up with non-compositional meanings. On the other hand, there is a set of verbal stems that are more or less productively used in complex verb formation. Finally, some complex verbs are combinations of a preverbal element and a verbal stem that are only used together, as *qum-art-* above. I will consider them in turn.

#### Michael Daniel

Dargwa preverbs are identifiable in Mehweb first of all on etymological grounds. The only typical preverb formations are the prefix *ar-* 'away' (*ʡaˤr*in roots with pharyngealization, see Moroz 2019) in various motion verbs, in which a prefix with a clear directional meaning combines with a motion verb. All other combinations show a strong degree of idiomatization. The presence of highly idiomatic combinations seems to contradict Magometov's (1982: 74) suggestion that, in Mehweb, the system of prefixes has not been fully developed – rather, it passed away, leaving behind few vestiges. Below, all preverb ~ verb combinations attested so far are given as perfective and imperfective, the perfective also showing the aorist suffix in parentheses; the preverbs are provided with meaning labels suggested by Magometov (1982: 74–80), who based these suggestions on comparison with other Dargwa languages.

	- a. *ʡaˤr--aˤq'*-(*un*) ~ *ar--aš-* 'go away, leave' from *--aˤq'-* 'go'
	- b. *ar--uk-*(*ib*) ~ *ar--ik-* 'lead away'; cf. *--uk-* ~ *--ik-* 'lead'
	- c. *ar--uχ-*(*ib*) ~ *ar--iχ-* 'bring away'; cf. *--uχ-* ~ *--iχ-* 'bring'
	- d. *ar--ik-*(*ib*) ~ *ar--irk-* 'fall down, fall out'; cf. *--ik-* ~ *--irk-* 'happen' (etymologically probably 'fall')
	- e. *ar--ih*(*ub*) ~ *--irhʷ-* 'throw away, out from somewhere'; cf. *--ih*(*ub*) ~ *--irhʷ-* 'throw'
	- f. *ar--as*(*ib*) ~ *ar--is-* 'take away'; cf. *as*(*ib*) ~ *is-* 'take'
	- g. *ar--uʔ-* ~ *ar--ulʔ-* 'lose'; cf. *--uʔ-* ~ *--ulʔ-* 'spoil'
	- a. *ka-lʔ*(*un*) ~ *k-ulʔ-* 'remain'; cf. *alʔ-*(*un*) ~ *ulʔ-* 'cut'
	- b. *ka--at*(*ur*) ~ *ka--alt-* 'leave'; cf. *--atur* ~ *--alt-* 'put on/under (?)' (the distribution of this verbal stem in Mehweb is further discussed below)
	- c. *ka--iʔ*(*ib*)*-* ~ *ka--irʔ-* 'sit down'; the stem is not attested as a free verb
	- a. *har--ik*(*ib*) ~ *har--irk-* 'become first'; cf. *--ik*(*ib*) ~ *--irk* 'happen' (etymologically probably 'fall')
	- b. *har--uq*(*un*) ~ *har--ulq-* 'run away, flee'; cf. *--uq*(*un*) ~ *--ulq* 'come, enter'
	- a. *če--uq*(*un*) ~ *če--ulq-* 'grow (of plants or hair)'; cf. *--uq-* ~ *--ulq* 'come, enter'

4 Mehweb verb morphology

	- a. *q'-aˤbʡ*(*ib*) ~ *q'-ibʡˤ-* 'close'; cf. *ʡaˤbʡ*(*ib*) ~ *ʡibʡˤ-* 'shut someone up; cast someone a spell of not being able to urinate or defecate (?)'
	- b. *q'a--ik*(*ib*) ~ *q'a--irk-* 'become silent, stop'; cf. *--ik*(*ib*) ~ *--irk-* 'happen'

Some preverbs are only attested with one verbal root, and thus synchronically indistinguishable from bound preverbal elements discussed below:


Like many East Caucasian languages, Mehweb has verbs that combine with various elements in preverbal position to form non-compositional (or not fully compositional) complex verbs. Сomplex verbs show different degrees of univerbation, which may be viewed as a decrease in compositionality of the complex and an increase in the boundedness of the preverbal element. The latter includes the loss of categorical transparency of the preverbal element, from autonomous noun, adverb or adjective for which the verbal stem serves as a verbalizer, to a bound morpheme with no clear autonomous semantics or categorical status. Assumedly, intermediate cases are also possible, when the preverbal element is recognized by the speakers as a separate word but is much more often used in a verbal complex, but this issue has not been studied, so the orthographic solutions are somewhat arbitrary. Whenever I have no elicited evidence that the element is only used in this complex, I write it separately below.

The most productive verbs include *--uh*(*ub*) 'become' and *--aq'*(*ib*) 'do'. When combining with adjectives (the short form, lacking the attributivizer *-*(*i*)*l*), the two verbs form inchoative ~ causative pairs. Cf. *ara --uhes* 'recover' lit. 'healthy become', *ara --aq'as* 'heal' lit. 'healthy do' from *ara*(*l*) 'healthy'. Other verbs are only exceptionally attested in inchoative constructions. I have one example: *ʡaˤrʁa --aʔib* 'stretch'; cf. *ʡaˤrʁa*(*l*) 'long' and *--aʔas* 'begin'.

The verbs *--uh*(*ub*) 'become' and *--aq'*(*ib*) 'do' also form less compositional derivations with nouns or elements of synchronically unclear categorical status, e.g. *deħ buh*(*ub*) 'start stinking' (*deħ* 'smell'), *gʷer baq'*(*ib*) 'rock (a cradle)', *χal-baq'*(*ib*) 'seek', *dam-baq'*(*ib*) 'beat up'.

The verb *ib* 'say' (pfv) is used in complex verbs designating sound production or similar (*šʷaˤt' ib* 'whistle', *tu ib* 'spit', *aˤmču ib* 'sneeze' etc.) The recorded complex verbs designating motion are based on the verb *--uq*(*un*) ~ *--ulq* 'come, enter' which has a limited distribution as a free verb but is also used with prefixes (see above), or in combination with an adverbial element *dur*(*a*) 'outside' in *dura --uq*(*un*) 'exit'. The complex verbs with *--uq*(*un*) ~ *--ulq* 'move, enter' include *t'aˤʜ --uq*(*un*) 'jump', *čaˤχ --uq*(*un*) 'slip', *duc' --uq*(*un*) 'run', *tir --uq*(*un*) 'wander' – it seems such verbs tend to designate quick movement. The verb *--aˤq*(*ib*) *~ --irqˤ* 'hit' is used in several complex verbs, from highly compositional *k'ʷama --aˤq*(*ib*) 'churn butter' (*k'ʷama* 'butter') and *urculi --aˤq*(*ib*) 'chop wood' (*urculi* 'firewood') to non-transparent verbs with no common semantic denominator, *kal --aˤq*(*ib*) 'go stale' (*kal* 'stale'), *ʡaˤš--aˤq*(*ib*) 'come back' and *uruχ --aˤq*(*ib*) 'become afraid'. The meaning 'be afraid' in the imperfective may also be rendered by *uruχ k'-*, where *k'-* is a bound verbal stem only attested in the imperfective. It could be that the difference between the two imperfective verbs, *uruχ --irqˤ* (*ib*) and *uruχ k'*(*ib*) is that between multiple episodic events (true imperfective of *uruχ --aˤq*(*ib*)) vs. state, respectively – but the evidence for this is not sufficient.

Other verbs include completely non-compositional combinations with roots which do not serve as productive verbalizers, so that identification of a light verb with a lexical verb is fully formal. These include:


While the common way of univerbation is the increase in boundedness of the preverbal adverb or nominal with the stem of a free verb, several complex verbs contain a stem whose identification is problematic. Attested cases are:


#### 4 Mehweb verb morphology

In (65) and (66), the (*a*) appears before the negative prefix, and is otherwise lost before the vowel of the stem. The verb in (67) has a negative form *ʡajk'-ħajk'-an* 'does not (usually) lock', which suggests an underlying structure of the positive forms looking something like *ʡajk'-k'*(*ib*), with the two occurrences of *k'* merging in one.

Two cases have an especially unclear morphological status in terms of (un)boundedness of the verbal root. First, the verbal root *--at*(*ur*) ~ *--alt* seems to mean 'put' (probably from the original meaning 'leave'), but it is a markedly rare choice in this meaning (the common verb for 'put' is *--ix*(*ib*)). The stem is much more common in several non-compositional structures, including the prefixal verb *ka--at*(*ur*) ~ *ka--alt-* 'leave behind, lose' (also causative *ka--at-aq-* 'kidnap (cause to be lost?)'), with designation of clothes meaning 'take off', the noun *ši* 'sting' (meaning 'sting (verb)'), the apparently bound element *dub* (meaning 'hold fast', cf. also *dub buc*(*ib*) 'hold fast' and *dub aˤʡ*(*ib*) 'eat'), the word *c'urʔa* in the sense 'become/leave orphan' and the spatial form *hune--* 'on the road' meaning 'see off' ('leave/put on the road'?). But it is also used in the construction *--atur --aʔas* 'let (someone pass/go)', where what appears to be a finite form (an aorist *--atur*) is used in apparent subordination to the verb 'begin'/'arrive'. Another probable use is the complex verb *waˤb-aˤt*(*ur*) ~ *waˤb-aˤlt-* 'call out'. The verbal stem is similar, but, first, the putative *--at*(*ur*) ~ *--alt* is irregularly pharyngealized (probably, pharyngealization has spread from the preverbal component, but this is an irregular process, because pharyngealization in Mehweb usually spreads leftwards – see Moroz 2019). And, second, in negative forms, the *b* splits in two (*waˤb-ʜa-baˤt*(*ur*)). This may mean that the former gender marker, now frozen because it was controlled by the lexical noun which was the source of the bound preverbal element *waˤb-*, fused with the final *-b* of this element when there was no intervening negation prefix. But this process, again, is irregular.

Second, the verbal root *--iz*(*ib*) ~ *--ilz-* is attested with a preverb (see *a--iz*(*ib*) 'stand up' above), in *tir --iz*(*ib*) ~ *--ilz-* 'turn around' (cf. *tir --uq*(*un*) 'wander, go in circles' above), and in the expression *ħule --iz*(*ib*), where *ħule* is an unclear form related to the noun 'eye', while the complex verb agrees with the subject – the one who looks). Otherwise, the verb *--iz*(*ib*)/*--ilz-* does not seem to be used alone.

Finally, there are some idiomatic combinations of words of different categories with verbs, showing more or less clear paths of semantic derivation, e.g. *liħi bixes* 'listen' – lit. 'ear put'; *surat diltes* 'draw', lit. 'take out image'; *himi abizes* 'become angry', lit. 'the bill raises', *aqu ihʷes* 'cover', lit. 'throw up'; and less transparent synchronically *žuχ wiʔ*(*ib*) 'urinate' and *k'uč'e wiʔ*(*ib*) 'defecate' – cf. the same root as a bound root in *ka--iʔ*(*ib*) ~ 'sit down'; *ask'es --erχʷes* 'fight' (lit. 'catch/cling go') etc.

#### Michael Daniel

### **List of abbreviations**



### **References**


*Daghestanian languages: Tense, aspect, evidentiality, mood and modality*, 247– 266. Leiden-Boston: Brill.


## **Chapter 5**

## **Moods of Mehweb**

### Nina Dobrushina

National Research University Higher School of Economics, Linguistic Convergence Laboratory, ndobrushina@hse.ru

The paper is a description of moods in Mehweb, a lect of the Dargwa branch of East Caucasian (Nakh-Daghestanian) languages, Republic of Daghestan. The data were collected in the course of several field trips to the village of Mehweb. The forms of non-indicative moods and common constructions where these forms occur are described. Mehweb has inflectional forms for the imperative, prohibitive, optative, irrealis and apprehensive. Hortative and jussive are expressed periphrastically.

*Keywords:* Nakh-Daghestanian, East Caucasian, modality, mood, imperative, hortative, jussive, optative, irrealis, conditional, apprehensive, volitional.

### **1 Introduction**

This paper is a description of non-indicative moods in Mehweb. Mehweb moods are briefly discussed in Magometov (1982); Khajdakov (1985) and in a sketch of Mehweb morphology by Nina Sumbatova (manuscript). The data for this paper were collected in the course of field trips to Mehweb in 2013, 2014 and 2015.

I describe morphological forms of non-indicative moods as well as periphrastic constructions used for the expression of some categories which are rendered by non-indicative moods in many languages of the world.

There are five forms which can be considered as inflectional forms of mood in Mehweb: second person imperative, prohibitive, optative, irrealis, and apprehensive. I also briefly describe the converbs which are used in the subordinate part of conditional clauses, because these forms are functionally close to the nonindicative moods, and in many languages, non-indicative forms are used in these clauses. The hypothetical conditional converb is derived from the same irrealis stem in *-a* as optative, irrealis, and apprehensive, thus manifesting similarity with non-indicative moods.

Nina Dobrushina. 2019. Moods of Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 117–165. Berlin: Language Science Press. DOI:10.5281/zenodo.3402062

#### Nina Dobrushina

I also consider two periphrastic constructions: one is used for the hortative (=first person plural imperative, or inclusive imperative), and the second for the jussive (third person imperative).

The paper is structured in accordance with the semantics of non-indicative forms and constructions. It starts with volitional categories. In §2, the formation of second person imperative is considered, and typical constructions with second person imperative are described. §3 describes the prohibitive – the negative imperative which is expressed in Mehweb, as in most East Caucasian languages, by a dedicated morphological marker. Several interjections with imperative meaning are considered in §4. §5 and §6 describe the form and semantics of periphrastic constructions which are used for hortative and jussive. In §7, the semantics of the optative is discussed, as well as some typical constructions involving the optative. After volitionals, the forms with the irrealis meaning are considered in §8; as in most East Caucasian languages, they occur almost exclusively in conditional clauses. Last, I consider the apprehensive form, used to introduce a situation the speaker is afraid of (§9). In §10 (Discussion), I compare the system of Mehweb non-indicative moods with that of five other Dargwa languages and dialects.

### **2 Second person imperative**

Second person imperative expresses commands and requests addressed to the hearer. In this section, I analyze the formation of second person imperatives in their relation to transitivity and controllability of the verbs, the agreement of imperatives with the addressee, and the forms of address in the imperative constructions.

#### **2.1 Formation of imperatives**

The second person imperative of imperfective verbs is always marked by the suffix *-e* (1, 2), unlike the imperative of perfective verbs. The second person imperative of perfective verbs is marked either by *-e* or *-a* depending on the transitivity of the verb. Intransitive verbs take the suffix *-e*, transitive verbs take the suffix *-a* (see Table 1):

(1) *niʔ* milk *urt'-e!* **pour:ipfv-imp** 'Pour the milk!'


Table 1: Formation of second person imperatives


As *-e* as an imperative marker is an unmarked choice, it is glossed simply as imp.

Labile perfective verbs can form two imperatives, one that follows the transitive pattern, the other that follows the intransitive one. Cf. *abxes* 'open, pfv', (*b*)*aˤldes* 'hide, pfv', (*b*)*erqʷes* 'become worn, pfv':


Nina Dobrushina


Some verbs have irregular and/or suppletive imperative forms. For example the verb *es* 'say' has the imperative *bet'a*; other cases are considered in Daniel (2019) [this volume].

Imperatives from verbs that denote events and situations over which the speaker exerts no control are acknowledged as grammatical by some speakers only. In most cases speakers are able to come up with a special context. For example, one can say *Bemže!* 'Get hot!' as if one was addressing a stove.

Imperatives of some perfective verbs which denote uncontrollable events are presented in Table 2.


Table 2: Imperative of intransitive uncontrollable verbs

Most two-place experiencer verbs have two imperatives, with suffix *-a* and with suffix *-e*. There is no clear difference in meaning between these two forms.

(11) *ħa-ze* you.sg.obl-inter(lat) *arʁ-e!* understand:pfv-imp '[You] understand!'

5 Moods of Mehweb

(12) *ħa-ze* you.sg.obl-inter(lat) *arʁ-a!* understand:pfv-imp.tr '[You] understand!'

Imperatives from experiencer verbs are shown in Table 3. Not all speakers acknowledge both imperative forms of these verbs; the less accepted forms are marked by a question mark.



Notably, verbs that show semantic restrictions on the formation of imperatives easily produce imperatives within the jussive construction. The jussive is built as a combination of an imperative of the main verb with the imperative of the verb *es* 'say' (see §6):

(13) *gʷ-e* **see:pfv-imp** *bet'-a!* say:pfv-imp.tr

'Let him see!' (he should make attempts to see)

Some intransitive verbs that allow just one form of second person imperative have the jussive construction with two imperative forms, the one in *-e* and the one in *-a*. Speakers' first choice is usually the form in *-e*. They do not see any semantic difference between the jussive based on the imperative in *-e* and the jussive based on the imperative in *-a*. Cf. examples (13) and (14):

(14) *gʷ-a* **see:pfv-imp.tr** *bet'-a!* say:pfv-imp.tr 'Let him see!' (he should make attempts to see)

Examples of the jussive constructions with intransitive and experiencer verbs are shown in Table 4.

#### Nina Dobrushina


Table 4: Examples of jussive construction with uncontrollable verbs

#### **2.2 Number and gender of the addressee**

All verbs in the imperative obligatorily add a dedicated imperative plural suffix *-na* to convey the plurality of the addressee.

Intransitive verbs which have a prefixal agreement slot agree in gender and number with the nominative argument. Since this nominative argument and the addressee coincide in intransitive verbs, the plural imperative suffix *-na* agrees with the same argument as the prefix (17).


'Come to me (addressing several people)!'

Transitive verbs with a prefixal agreement slot also agree with their nominative argument. Here, however, the addressee is the agent in the ergative case. The prefixal agreement and the plural imperative suffix are triggered by different arguments (19).

(18) *b-aˤbʡ-a* **hpl-kill:pfv-imp.tr** *urš-be!* boy-pl

'Kill these boys (addressing one person)!'

5 Moods of Mehweb

(19) *w-aˤbʡ-a-na* **m-kill:pfv-imp.tr-imp.pl** *rasul!* Rasul 'Kill Rasul (addressing several people)!'

The suffix *-na* as a plurality of addressee marker is also used on prohibitive forms (see §3).

In some Dargwa dialects (e.g. in Tanti – Sumbatova & Lander 2014: 146) the imperative form is not used if the P of the transitive construction is a first person argument. The optative is used instead. This is not true for Mehweb – there is no restriction on the usage of the imperative with the first person:

(20) *nu* I *dub* eat *aˤʡ-aq-a!* lv-caus-imp.tr 'Feed me!'

### **2.3 Subject and forms of address**

The agent of the imperative is not usually expressed, but it can be indicated by an overt second person pronoun if it is stressed:


Imperative utterances may contain forms of address expressed by a noun phrase in the nominative. The form of address is in the nominative even when referring to the agent of transitive verbs:


Second person pronouns and demonstratives (used as third person pronouns) cannot be used as forms of address:

(25) *\*ħu* you.sg(nom) *deč'* song *b-aq'-a* **n-do:pfv-imp.tr** (26) *\*it* this(nom) *deč'* song *b-aq'-a* **n-do:pfv-imp.tr**

The second person imperative construction can however include a third person NP which is not a form of address. It is marked by the ergative with transitive verbs and by the nominative with intransitive verbs. Although the construction formally includes a third person NP, it is addressed to the hearer whose name is Muhammad:


Speakers often build this construction with the additive particle *-ra*:


The construction with a third person NP and the imperative is primarily used when the speaker addresses several people. The following sentences can be uttered by the teacher who is addressing the whole class and chooses the pupils to perform certain actions:

(32) *pat'imat--ra* Patimat(nom)--add *d-ak'-e,* **f1-come:pfv-imp** *asijat--ra* Asijat(nom)--add *d-ak'-e.* **f1-come:pfv-imp** 'Patimat come, and Asijat come.'

5 Moods of Mehweb

(33) *pat'imat-li* Patimat-erg *deč'* song *b-aq'-a,* **n-do:pfv-imp.tr** *asijat-li* Asijat-erg *deč'* song *bel'č'-a.* **read:pfv-imp.tr** 'Patimat sing the song, and Asijat read the rhyme.'

The following example with the word *ca* as third person imperative subject comes from the corpus:

(34) *mallarasbadij-ni* Molla Nasreddin.obl-erg *ib* say:pfv.aor *iš-di-li-ze :* that-pl-obl-inter(lat) *ca* **one** *udi-di* below-trans *w-iz-e-na***, m-stand:ipfv-imp-imp.pl** *ca* **one** *aqu-di* up-trans *w-iz-e-na,* **m-stand:ipfv-imp-imp.pl** *urga-w* between-m *nu* I(nom) *w-iz-iša,* m-stand:ipfv-fut.ego *nu-ni* I-erg *ħuša* you.pl *k'ʷi-jal-la* two-ord-add *χʷasar* rescue *b-aq'-iša* hpl-do:pfv-fut.ego *ca-ca* one-one *ʁuruši-ze.* rouble-inter(lat) 'Molla Nasreddin told them: one of you stand higher, the other stand

lower, I will stand between you two, I will rescue the two of you for one rouble each.'

A similar phenomenon – the possibility to use 2nd person imperative with 3rd person subject with reference to the addressee - is found in other East Caucasian languages (cf. Dobrushina 2001: 323).

#### **2.4 Imperative with particles**

The imperative can be used with particles *-w* and/or *-ca*. Although the particle *-w* resembles the masculine gender marker, it does not depend on the gender of the addressee:

```
(35) deč'
     song
           b-aq'-a-w!
           n-do:pfv-imp.tr-ptcl
     'Sing a song! (addressing women or men)'
```
The particle *-w* is identical to the question particle *-w*/*-u*. The particle *ca* is formally identical to the word *ca* 'one' and probably originates from it.

#### Nina Dobrushina

(36) *ʜaˤramir-ti-la* Haramirt-pl-gen *ʁuša-ne* house-pl *elʔ-a-ca.* **count:pfv-imp.tr-ptcl** 'List the families of the Haramirt (clan).' (Text 19. Clans, 1.6)

Neither of the particles can be used if the imperative utterance expresses permission:

(37) *abaj,* mother *b-uh-es--u* n-become:pfv-inf--q *nu-ni* I-erg *g-es* give:pfv-inf *rasuj-s* Rasul-dat *k'amp'it'?* sweet '– Mother, can I give a sweet to Rasul?' *b-uh-es* n-become:pfv-inf *b-eg-a* n-give:pfv-imp.tr / / ⁇*b-eg-a-w* n-give:pfv-imp.tr-ptcl / / ⁇*b-eg-a-ca.* n-give:pfv-imp.tr-ptcl '– You can, give it to him.'

The particle *-w* expresses a more categorical demand than that expressed by the particle *-ca*. Therefore, it is not used in situations when the speaker has a status lower than the addressee, or when the speaker has no right to demand. In the following example, the child asks her mother to give her the sweet; with the particle *-w* she is rather too direct, as if her mother must give it to her; with the particle *-ca* the utterance sounds as a mild request.

(38) *Abaj* Mother *ag-a* **give:pfv-imp.tr** / / *ag-a-ca* **give:pfv-imp.tr-ptcl** / / ?*ag-a-w* **give:pfv-imp.tr-ptcl** *nab* I.dat *k'amp'it'.* sweet 'Mother, give me a sweet.'

In example (39), the imperative with the particle *-w* would have been completely inappropriate, since the pupil addresses his request to the teacher. The imperative with the particle *-ca* is better, although it is not the typical way to address the teacher.

(39) ?*Maisarat* Maisarat *Magomedovna* Magomedovna *ag-a-ca* **give:pfv-imp.tr-ptcl** *di-ze* I.obl-inter(lat) *kung.* book 'Maisarat Magomedovna, give me the book please.'

5 Moods of Mehweb

The particles *-w-* and *-ca* can occur together:

(40) *pat'imat* Patimat *ħu* you.sg(nom) *d-ak'-e-w-ca!* **f1-come:pfv-imp-ptcl-ptcl** 'Patimat, [you] come!'

According to the corpus, the particle *-ca* is used very frequently; the particle *-w* was not found in the corpus.

#### **2.5 Coordinated constructions with imperatives**

If several imperatives are combined, the chain of verb forms can either consist of imperatives or combine imperative(s) with converb(s):


'Patimat, peel the potato and make the soup!'

(43) *k'amp'it'-une* sweet-pl *as-i-le* **take:pfv-aor-cvb** *tukaj-ħe-la* shop-in-el *ħu-ni--jal* you.sg-erg--emph *mu-d-uk-adi.* **negvol-npl-eat:ipfv-proh** 'Buy some sweets, (but) don't eat them.'

Further examples and some discussion of the contrast between the chains with imperatives and the chains with converbs can be found in Kustova (2019) (this volume).

### **3 Prohibitive**

The prohibitive is a negative imperative which is expressed by a dedicated affix. It is formed with the prefix *mV-* with an unspecified vowel which assimilates to the next vowel (see discussion in Moroz 2019 [this volume] and Daniel 2019 [this volume]), and the suffix *-adi*, sometimes truncated to *-ad*. In §10, I give

#### Nina Dobrushina

some information on the origin of this marker. The gender agreement marker *b-* (N or hpl) assimilates to the negvol marker *m*V- (see Moroz 2019 [this volume]). Sometimes, prohibitive formation involves reduplication, as in (46) – see the discussion in Daniel (2019) [this volume].

(44) *deč'* song *mi-m-iq'-ad***(***i***)***!* **negvol-n-do:ipfv-proh** 'Don't sing!'

The prohibitive can be derived only from imperfective stems. Therefore, each verb has two imperatives but only one prohibitive. There is no formal distinction between transitive and intransitive prohibitives.

(45) *mu-lug-adi* **negvol-give:ipfv-proh** *d-uk'-a-k'a-ra,* f1-say:ipfv-irr-cond-add *maja* Maja *g-i-le* give:pfv-aor-cvb *le-l-le* aux-f-cvb *hub-li-s.* husband-obl-dat 'Although she said: 'Don't give', they still married Maja off'. (Text 14. Laces, 1.3) (46) *gurda* fox *b-ik'-uwe* n-say:ipfv-cvb.ipfv *le-b* aux-n *sinka-li-ze* bear-obl-inter(lat) *b-is-mi-m-is-adi* **n-cry-negvol-n-cry-proh** *ħu.* you.sg(nom)

'The fox said [to the bear]: "Don't cry".' (Text M. A bear, a wolf and a fox, 1.11)

The prohibitive has the same marker of plurality *-na* as the imperative. The prohibitive suffix cannot be truncated before the plural marker.

	- Intended: 'Don't sing!' (addressing several speakers)

The prohibitive can be used with forms of address in the same way as the imperative (§2.3):

(49) *pat'imat,* Patimat *deč'* song *mi-m-iq'-adi.* **negvol-m-do:ipfv-proh** 'Patimat, don't sing the song.'

Constructions with third person subject are also available for the prohibitive:

(50) *pat'imat-li* Patimat-erg *deč'* song *mi-m-iq'-adi.* **negvol-m-do:ipfv-proh** '[Patimat] don't sing the song.'

The prohibitive can take the particle *-ca*:

(51) *mi-m-iq'-adi-ca* **negvol-m-do:ipfv-proh-ptcl** *hel* this *deč'!* song 'Don't sing this song!'

### **4 Imperative interjections**

There are several words which function as imperatives although they are not related to any verb. They are used to urge the addressee to perform an action, and some of them can attach the imperative plural marker *-na*.

The interjection *ma* 'take, hold' is known in various languages of Daghestan (e.g. Archi, Agul). In Mehweb, it may attach the plural marker *-na*:

(52) *ma!* intj 'Take!'

(53) *ma-na!* intj-imp.pl 'Take (addressed to several people)!'

The interjection *ma* can be combined with other imperative forms:


The imperative interjection *hara* is used to attract the visual attention of the addressee. It also can attach the plural marker *-na*:

```
(56) hara!
      intj
      'Look!'
(57) hara-na!
      intj-imp.pl
      'Look! (addressing several people)'
```
Two imperative interjections are used to urge the addresses to be quiet and keep silence. For example, the teacher can use them in order to make children silent: *q'ah!* 'Shhh!' and *c'it'!* 'Shhh!'. These interjections cannot combine with the plural marker *-na*.

### **5 Hortative (first person inclusive imperative)**

The term *hortative* is used here for the constructions which express the inducement to perform an action together with the speaker, cf. English *Let's go*. There is no dedicated hortative morphology in Mehweb, but the periphrastic construction is widely used to express invitation to common action.

The hortative construction consists of the infinitive of the main verb and the form cl*-aš-e*, where cl is a gender marker.

(58) *w-aš-e* **m-go:ipfv-imp** *χal* seek *w-aq'-as* **m-do:pfv-inf** *ħa-la* you.sg.obl-gen *urtaq'.* friend 'Let's look together for your friend' (Aspectual test 1, 1.121)

The form cl*-aš-e* is an imperative of the verb cl*-aš-es* 'go/come (ipfv)'. Alone, this form can be used as a second person imperative and as a hortative. There are no other words in Mehweb which combine these two meanings in one form; there are also no other hortatives which are expressed lexically, in one word.


5 Moods of Mehweb


This pattern of hortative construction – with an infinitive and a particle originating from an imperative or hortative form of a motion verb – is attested in some other East Caucasian languages (Khwarshi (Khalilova 2009), Lak and Rutul (personal fieldnotes)).

The imperative cl*-aš-e* followed by the plural marker *-na* is used as a second person plural imperative or as an inducement to several addressees to perform an action together. There is an irregular change of *-e* to *-i* when the plural suffix is added: *w-aše – b-ašina*:

(63) *b-aš-ina* **hpl-go:ipfv-imp.pl** *tukaj-ħe!* shop.obl-in(lat)

'Go to the shop!' / 'Let's go to the shop!' (addressing several people)

In the hortative construction, the form cl*-aš-e* agrees with the addressee, while the infinitive of the main verb agrees with the nominative. In the constructions with intransitive imperatives, the addressee and the nominative participant coincide (64, 65). In the constructions with transitive imperatives, the addressee coincides with the ergative participant; therefore, the main verb and the auxiliary form cl*-aš-e* agree with different arguments (66–69).


The plural suffix *-na* is added to the verb cl*-aše* when the hortative construction is addressed to several people and the action is thus meant to be performed by more than two participants, including the speaker:

(70) *b-aš-ina* **hpl-come:ipfv-imp.pl** *deč'* song *b-aq'-as.* **n-do:pfv-inf** 'Let's sing a song (addressing several people)!'

The hortative construction can contain the first person plural pronoun as a subject:

(71) *d-aš-e* **f1-go:ipfv-imp** *nuša* we *tukaj-ħe* shop.obl-in(lat) *b-uˤq'-as.* **hpl-go:pfv-inf** 'Let's go to the shop (addressing a girl)' (72) *b-aš-e* n-go:ipfv-imp *sinka* bear *b-erkʷ-es* n-eat:pfv-inf *nuša-jni!* we-erg 'Let's eat the bear!' (fox addressing wolf) (Text M. A bear, a wolf and a fox)

In the hortative construction, negation is marked on the main verb, since the illocution is not under the scope of negation:


Constructions with the negated verb of motion are not interpreted as hortatives:

5 Moods of Mehweb

(75) *mi-d-ik'-adi* **negvol-f1-come:ipfv-proh** *deč'* song *b-aq'-as.* **n-do:pfv-inf** 'Don't come to sing a song.'

If a hortative occurs in the coordinative construction, one of the predicates can be expressed by a perfective converb (76), or both predicates are expressed by infinitives (77); in the latter case, one hortative auxiliary can belong to both infinitives:

(76) *b-aš-ina* **hpl-go:ipfv-imp.pl** *qali--ra* house--add *b-aq'-i-le,* **n-do:pfv-aor-cvb** *q'ʷaˤl* cow *as-es.* **take-inf** (77) *b-aš-ina* **hpl-go:ipfv-imp.pl** *qali--ra* house--add *b-aq'-as,* **n-do:pfv-inf** *q'ʷaˤl--ra* cow--add *as-es.* **take-inf** 'Let's build the house and buy the cow.'

The motion verb almost always takes the first place in hortative constructions (78), but its final position is not completely ungrammatical (79).


The particle of mild request *-ca* can be used with the hortative:

(80) *w-aš-e-ca* **m-go:ipfv-imp-ptcl** *heč'* that.higher *xunul* woman *ʡaˤχ* good *r-aq'-as.* f-do:pfv-inf 'Let's help that women.' (Text 06. Mahmud Omar who was friends with devils, 1.11)

### **6 Jussive (third person imperative)**

Jussive is a form or construction which is used to express an inducement to a third person, most often transferred via the addressee. Some East Caucasian languages have a dedicated form for this meaning; often, the meaning of jussive is covered by the optative (Dobrushina 2012). In Mehweb, the meanings of the jussive and optative are expressed separately, by a periphrastic construction and

#### Nina Dobrushina

by an inflectional form respectively. In §6.1, the structure of the jussive construction is described. §6.2 discusses the semantics of the jussive construction. The optative is considered in §7.

#### **6.1 Jussive construction**

The Mehweb jussive consists of the imperative of the verb 'say' *bet'a* (irregular form; see Daniel 2019 [this volume]) and the imperative of the main verb. The jussive is conceived as a transfer of a command or request to the non-locutor via the addressee (Tell him "Work!" → Let him work!):


Jussive semantics does not require the verb to designate a controllable action (see §6.2). Therefore, verbs which denote uncontrollable actions can occur in the jussive construction in the form which is morphologically imperative, while normally the second person imperative of these verbs is not used (see also §2.1):


The imperative of the verb 'say' does not have an agreement slot. It can only agree with the addressee in number, as all imperatives:


5 Moods of Mehweb

The jussive construction shows some evidence of grammaticalization. The agent of the jussive construction usually bears A or S marking (ergative with transitive verbs and nominative with intransitive verbs):


Although, as was shown above (§2.3), second person imperative in Mehweb can be used with 3rd person subject and A/S marking, such constructions are clearly peripheral. They do not occur in the texts; they are used in a special pragmatic type of context (addressing several people in distributional meaning); and they cannot apply to non-animate subject. Examples (82–86) hence cannot be interpreted as cases of reported speech.

The addressee of the verb 'say' is normally marked by the inter-lative. The availability of S or A marking in the jussive construction shows that the jussive has developed into a periphrastic form distinct from the complement construction of the verb 'say'. Cf. example (87) with a complement clause-like structure with addressee marking in (88):

(88) *musa-ze* Musa-inter(lat) *uz-e* m.work:ipfv-imp *bet'a.* **say:pfv-imp.tr** 'Tell Musa to work.'

In jussive constructions, the verb 'say-imp' follows the imperative of the main verb. The following sentence is ungrammatical:

(89) *\*musa* Musa *bet'-a* **say:pfv-imp.tr** *uz-e.* **m.work:ipfv-imp**

As with the hortative, negation is marked on the lexical verb of the jussive construction:

(90) *muħammad-ini* Muħammad-erg *deč'* song *mi-m-iq'-adi* **negvol-n-do:ipfv-proh** *bet'-a.* **say:pfv-imp.tr** 'Let Muhammad not sing a song.'

### **6.2 Semantics of the jussive**

The jussive is used in exhortations to actions by third person agents:

(91) *išbari* today *muħammad-ini* Muhammad-erg *t'ult'* bread *b-aq'-a* **n-do:pfv-imp.tr** *bet'-a.* **say:pfv-imp.tr** 'Let Muhammad bake bread today.'

The jussive can also express permission:

(92) *b-uh-es--u* n-become:pfv-inf--q *muħammad-ini* Muhammad-erg *k'amp'it'* sweet *as-es?* take:pfv-inf '– May Muhammad take a sweet?' *b-uh-es,* n-become:pfv-inf *as-a* **take:pfv-imp.tr** *bet'-a.* **say:pfv-imp.tr** '– (He) may, let him take one.'

Jussives can have inanimate subjects. The jussive construction with an inanimate subject expresses the speaker's indifference towards the situation (indifference is semantically close to permission). The implication is that the addressee should not interfere with the realization of the situation; for instance, s/he should not take the boiling soup from the stove:


Constructions with inanimate subjects show again that the jussive construction is highly grammaticalized, because the imperative *bet'a* has lost its original meaning 'say!'.

The jussive is available only in the third person. First and second person pronouns cannot occur in jussive constructions:

(95) *it-ini* that-erg *as-a* **take:pfv-imp.tr** *bet'-a* **say:pfv-imp.tr** *k'amp'it'.* sweet 'Let him take your sweet.'

5 Moods of Mehweb


The semantics of indifference is the source for the constructions where the jussive has a concessive meaning:


is beautiful, she does not work at home).'

Unlike the optative, the jussive is not used to express wishes. Accordingly, example (100) is acknowledged to be grammatical, but semantically inappropriate; one of the speakers suggested that this sentence can be uttered by an atheist who thinks that God can be forced to perform an action. The correct choice would be to use the optative (101).


When the jussive is used do denote uncontrollable situations, it is interpreted as expression of indifference or allowance but not as wish. The following ut-

#### Nina Dobrushina

terance can be pronounced when the speaker does not care about the rain, e.g. because he has already done his work in the field:

(102) *d-aq'-a* **npl-do:pfv-imp.tr** *bet'-a* **say:pfv-imp.tr** *zab.* rain 'Let it rain (I don't care).'

If the speaker wants the rain to fall, she would rather use the form of optative:

(103) *d-aq'-a-b* **npl-do:pfv-irr-opt** *zab!* rain 'May it rain!'

### **7 Optative**

The optative is used to convey good and bad wishes. In Mehweb, as in many other East Caucasian languages, the optative is expressed by a dedicated inflectional form (for a discussion of optatives in languages of the Caucasus see Dobrushina 2011). The formation of the optative is described in §7.1, its semantics in §7.2, and typical constructions involving the optative form – in §7.3.

#### **7.1 Morphology of the optative**

The optative is marked by the suffix *-b* added to the irrealis stem in *-a-*:

(104) *aradeš* health *g-a-b!* **give:pfv-irr-opt** 'May [Allah] make [you] healthy!'

The optative can be derived from both the perfective and imperfective stems: *g-a-b* (give:pfv-irr-opt) – *lug-a-b* (give:ipfv-irr-opt); *d-ic-a-b* (npl-sell:pfv-irropt) – *d-ilc-a-b* (npl-sell:ipfv-irr-opt).

The negative optative is derived from the imperfective stem with the prefix *mV-* (the same negative volitional marker which is used in the prohibitive). The negative optative may also be formed with the regular negative prefix *ħa-*. The negative optative with the prefix *mV-* usually comes as a first choice of the speaker when s/he translates wishes with negation, but the forms with the prefix *ħa-* are also often considered grammatical. Forms in *ħa-* are more easily accepted from perfective verbs, thus filling the gap of the perfective negative optative. Sometimes, however, an imperfective negative optative with the prefix *ħa-* is also accepted by the speakers (see Table 5). Negative optative is not a frequent form, it

does not occur in the corpus. I was unable to compare the actual frequency of these two negative forms.


Table 5: Forms of the positive and negative optative

Some optatives have a reduced form without any suffixes: *w-ebk'-a-b* 'may [he] die!' – *w-ebk'* 'may [he] die!'


(107) *ħa-la* you.sg.obl-gen *abaj* mother *r-ebk'!* **f-die:pfv(opt)** 'May your mother die!' (…can be uttered by a mother of a child, and addressed to the child if something bad is going to happen to her/him –

i.e. may I die in your stead!)

Apart from the verb 'die', the reduced form was attested for the verbs cl-*erʔʷes* 'become dry', *če*-cl-*uqes* 'grow', and cl-*alqaqas* 'grow (causative)'. However, not all speakers accept all these examples (unlike *w-ebk'* which is frequent).

#### Nina Dobrushina


Truncated forms of the optative are also attested in Akusha (van den Berg 2001: 34), Ashty (Belyaev (a), manuscript), Shiri (Belyaev (b), manuscript), Tanti (Sumbatova & Lander 2014), and Sanzhi (Forker, in preparation) lects of Dargwa.

Some optative forms have a causative suffix which is not motivated semantically. Cf. examples (111), (112), (114), (115), (116) and (117). When the speakers discuss the difference between the optative with and without the causative suffix, they usually say that the sentences with causative suffix *-aq-* imply an appeal to God:


5 Moods of Mehweb


This semantic difference between the ordinary and the causative optative is due to the fact that the causative derivation adds a new participant to the situation. The sentences with the causative suffix may include the ergative of Allah (118, 119). If the participant is not overtly expressed in the sentence, this new participant in the causativized optative construction is by default understood as Allah. In another Daghestanian language, Archi (Lezgic), the ergative of Allah can be included even in intransitive optative constructions meaning 'with the help of Allah', where the ergative may be interpreted as the ergative of the cause, one of the known functions of the ergative case (Dobrushina 2011). In Mehweb, most speakers reject intransitive optative sentences with Allah in the ergative (120, 121).


If there is another overt ergative participant in the sentence, the clause is interpreted as an ordinary causative construction; cf. (124):

(122) *rasul* Rasul *w-ebk'-a-b!* **m-die:pfv-irr-opt** 'May Rasul die!'

#### Nina Dobrushina


### **7.2 Optative constructions**

The optative form is available for all persons, but with the first person the construction is pragmatically less felicitous.

Third person optative construction


Second person optative construction


First person optative construction

(129) *nu* I *r-ebk'* **f-die:pfv(opt)** / / *r-ebk'-ab!* **f-die:pfv-opt** 'May I die [but not you – addressing the child]!'

In optative constructions, frozen formulaic expressions are typical, and central participants are often left implicit. Cf. examples (108), (114), (116) where the person affected by the wish is overtly expressed. However, mentioning this person is not ungrammatical, as in the following examples:

5 Moods of Mehweb


Another possible participant of the optative situation is Allah. Most often it occurs in optative sentences as a form of address:

(132) *ja-allah* **ptcl-Allah(nom)** *ħušab* you.pl.dat *taliħ* luck *g-a-b!* **give:pfv-irr-opt** 'May [Allah] give [you] luck!'

In transitive constructions, Allah can also be expressed as an Agent, assuming ergative marking:


The Ergative form of the word *Allah* cannot co-occur with another agent in the ergative case:

(135) *\*allah-li* Allah-erg *ħu-ni* you.sg-erg *b-iz-il* n-tasty-atr *t'ult'* bread *b-aq'-a-b.* **n-do:pfv-irr-opt** Intended: 'May you make good bread with the help of Allah.'

#### **7.3 Semantics of the Optative**

Optative forms are dedicated to the expression of good or bad wishes.

(136) *ʔaq'* intellect *lug-a-b,* **give:ipfv-irr-opt** *balhni* knowledge *g-a-b.* **give:pfv-irr-opt** 'May [Allah] give [you] intellect, may [Allah] give [you] knowledge.'

<sup>1</sup>The ergative forms *Allahlini* ~ *Allahli* are morphological variants.

#### Nina Dobrushina

Unlike the jussive, the optative does not denote an action which is meant to be fulfilled by the addressee or by a third person. If the optative is derived from a verb which typically denotes controllable actions, the sentence is interpreted as a wish that Allah fulfills the action. The following example can be interpreted as a wish which can be made real by Allah, but not as an indirect command to the third person to give money:

(137) *d-aq-il* npl-much-atr *arc* money *g-a-b.* **give:pfv-irr-opt** 'May you be given [by Allah] a lot of money.'

The optative cannot refer to the past, cf. examples (138) and (139):


Optative forms are widely used in everyday life. Below are some traditional optative formulae:


'May [it] grow! (wish formula addressed to the person who is planting something)'

<sup>2</sup>This word occurs only in this formula and so far seems to be unanalyzable.

5 Moods of Mehweb

#### **7.4 Expression of wish by means of forms in** *-q'alle*

The wish of the speaker can also be expressed by forms ending in *-q'alle*. The derivation of these forms is described in §8. Forms in *-q'alle* show some properties of converbs (see §8 and Sheyanova 2019 [this volume]); the wishconstructions with forms in *-q'alle* must be considered as cases of insubordination (in terms of Evans 2007).

The counterfactual conditional converb in *-q'alle* can be used in a main clause in order to express the speaker's wish (similar to the forms of the conditional protasis in many European languages, as well as other languages of the East Caucasian family, cf. Belyaev 2012). Independent converbs in *-q'alle* differ semantically from the optative. While the optative form expresses blessings and curses, constructions with conditional converbs denote dreams and desires of speaker about some uncontrollable events. In Dobrushina (2011), these two types of optative were referred to as performative optative and desiderative optative. East Caucasian languages often have a dedicated inflectional form for the former, but the latter is usually expressed by conditional forms, as in Mehweb.


The speaker's wish can also be expressed by a combination of the infinitive with the counterfactual marker *-q'alle*:

(146) *nu-ni* I-erg *čaj* tea *d-erž-es-q'alle!* **npl-drink:pfv-inf-ctrf** 'I wish I had some tea!'

Unlike other converbs in *-q'alle*, the converb derived from the infinitive is not used in reference to the past:


#### Nina Dobrushina

The hypothetical conditional converb in *-k'a* (see §8) cannot be used in independent constructions.


### **8 Irreal forms**

Cross-linguistically, forms with irreal meaning are most often found in conditional constructions and in complement clauses (Mauri & Sansò 2016). In Mehweb, as in many other languages of Daghestan, complement clauses do not employ irreal forms. Mehweb conditional constructions have non-finite forms in the subordinate clause (conditional converbs), and a finite form in the main clause (irrealis). In this section, the derivation of conditional converbs (§8.1) and irrealis (§8.2) will be discussed. In §8.3, §8.4, and §8.5, conditional constructions of different types will be considered.

#### **8.1 Conditional converbs**

There are two markers of conditional clauses in Mehweb. They are distributed according to the degree of (ir)reality: the suffix *-k'a* is used in conditional clauses which may come true (hypothetical marker), the suffix *-q'alle* designates situations which did not and cannot take place (counterfactual marker).

The suffix *-k'a* presumably originates from the particle *k'a*. The particle *k'a* is used for topicalization of words of different classes. In example (151), it attaches to the noun *sinkala*, in example (152) – to the perfective stem of the verb. In the latter example, the particle is used together with reduplication, typical for predicate topicalization (Maisak 2010): *dargk'a dargira*.

(151) *sinka-la* bear-gen *k'a* **ptcl** *abzul-le* all-advz *ʁʷaˤn-ne* lie-pl *d-elʔ-un-na* npl-tell:pfv-aor-ego *wahaj-le--l* very-advz--emph *ʁʷaˤn-ne* lie-pl *luʔ-es* tell:ipfv-inf *w-aʔ-i-ra.* m-begin:pfv-aor-ego 'As for the bear, I did actually tell fibs.' (Aspectual test 1, 1.89)

5 Moods of Mehweb

(152) *d-arg-k'a* **npl-find:pfv-ptcl** *il-di* this-pl *qulle* house.pl *di-ze* I.obl-inter(lat) *d-arg-i-ra* npl-find:pfv-aor-ego *huni--ra* road--add *b-arg-i-ra.* n-find:pfv-aor-ego 'As for getting there, I did reach those houses and found the street.' (Aspectual test 1, 1.164)

Elicitation gave examples with predicate topicalization marked by the particle *k'a* alone, without reduplication:


That the suffix of conditional clause originates from the topicalization particle is in conformity with the close relation between topic and condition as described in Haiman (1978). It is likely that the suffix of counterfactual condition *-q'alle* also originates from the marker of topicalization. In Mehweb, the only function of *-q'alle* is to mark counterfactual conditional converbs, but in some other Dargwa languages there is a particle *q'al* (*q'alli*) with a wide range of meanings including topicalization (Maisak 2010; Mutalov & Sumbatova 2003; Forker in preparation). The following examples come from two Dargwa dialects:

Dargwa (Khuduts village) (Maisak 2010; example elicited by D. Ganenkov)

(155) *buč'--q'al* read:ipfv--**ptcl** *buč'unni* read:ipfv.cvb *cab* cop *cik'al* nothing *hankalgunnekːu.* remember:ipfv.cvb+cop.neg 'As for reading, he reads (the book), but does not remember anything.'

Dargwa (Icari village) (Maisak 2010; example suggested by R. Mutalov)

(156) *buč'-ni-la* read:ipfv-nmlz-gen *q'alli* **ptcl** *buč'atːa* read.prs.1sg *cacajnaqːilla* sometimes *behelra…* however 'As for reading, I read (books), but…'

#### Nina Dobrushina

Forms in *-q'alle* and in *-k'a* can be embedded. This is an argument in favor of their converbial status.


In §8.1.1 and §8.1.2, the derivation of conditional converbs in *-k'a* and *-q'alle* will be considered in more detail.

#### **8.1.1 Hypothetical conditional converb**

In the Hypothetical conditional converb, the suffix *-k'a* can be added to the Irreal stem of imperfective and perfective verbs. Therefore, every verb has two conditional converbs in *-k'a*: cl-*elč'es* 'read, pfv' – cl-*elč'ak'a*; *luč'es* 'read, ipfv' – *luč'ak'a*.

Conditional clauses with converbs in *-k'a* denote that the situation can come true in the future:

(159) *hel* this *deħ* smell *b-aq'-a-k'a* **n-do:pfv-irr-cond** *sinka-li* bear-obl(erg) *nuša* we *k'ʷi-jal-la* two-card-add *b-erg-es.* hpl-eat:pfv-inf 'If the bear smells this, he will eat us both.' (Text M. A bear, a wolf and a

fox, 1.6)

Followed by the additive particle *-ra*, hypothetical conditional converbs are used in concessive clauses (160). This pattern of marking concessive clauses – by a combination of conditional converb and emphatic or additive particle, also well known in Latin and Romance languages – is attested in the majority of East Caucasian languages (cf. Tanti (Sumbatova & Lander 2014: 138), Agul (Dobrushina & Merdanova 2012); Forker 2016 for generalizations).

5 Moods of Mehweb


#### **8.1.2 Counterfactual conditional converb**

The counterfactual marker *-q'alle* can be added to the past and infinitive forms, but not to the present. The speakers of Mehweb sometimes consider *-q'alle* as a separate word, but it cannot be separated from the verb. In this description, we consider *-q'alle* as a suffix. Table 6 summarizes the combinations of the verbal stems and the suffix *-q'alle*: possible combinations are marked as (+), impossible combinations are marked as (–); the perfective present form does not exist in Mehweb. Examples are presented in Table 7.

Table 6: Stems which can add the counterfactual suffix *q'alle*


Table 7: Examples of the forms with the counterfactual suffix *-q'alle*


#### Nina Dobrushina

Counterfactual converbs in *q'alle* are used in subordinate clauses of conditional constructions (example (162), more details in §8.3), and in independent clauses with the meaning of wish (example (163), more details in §7.4). This latter usage may be considered a case of insubordination, typical for the forms used in conditional clause.


'If only my son got married!'

#### **8.2 Irrealis**

The predicate of the main clause of conditional constructions is expressed by the form with the suffixal cluster *-a-re*: cl*-ubk'are* 'would die'. The cluster consists of the suffix of the Irreal stem *-a-* and the suffix of the Past *-re* (*-a-re* – irr-pst). The marker *-are* is used only for the expression of irrealis, so the form must be considered as a dedicated irrealis. The past suffix *-re* is not productive. Apart from irrealis, the suffix *-re* occurs regularly only in several lexemes: in the past copula *le-*cl*-re*, negative copula *agʷire*, in the lexeme *burgare* 'likely, probably' (originally irrealis), and the form *digibre* 'would like':


Some speakers acknowledge other forms in *-re* derived from the past stem of imperfective verbs as grammatical, such as *luč'ibre* (*luč'es* 'read, study, ipfv'),

5 Moods of Mehweb

*isibre* (*ises*, 'take, buy, ipfv'), *urcibre* (*urces* 'fly, ipfv'). These forms are also interpreted as irrealis:

(166) *? tukaj-ħe-la* shop-in-el *si-k'al* what-indef *is-ib-re* **take:ipfv-ipft-pst** *nu-ni--ra,* I-erg--add *arc* money *d-uʔ-ib-q'alle.* npl-be:ipfv-aor-ctrf 'I would have bought something in the shop, if (I) had some money.'

These forms however are never used spontaneously, do not occur in texts, and many speakers do not recognize them at all. Even the speakers who can come up with an example using one of these forms, tend to replace it by the regular irrealis in *-are*.

The irrealis form in *-are* is used in the main clause of conditional clauses (most often counterfactual) (167) as well as for the expression of irreal situations in independent clauses beyond conditional constructions (168):


'Rasul could have mowed the field yesterday, but he didn't go.'

#### **8.3 Counterfactual conditional clauses**

Counterfactual conditional clauses contain a converb in *-q'alle* in the protasis, and the irrealis in the apodosis. The constructions with the converb in *-q'alle* and irrealis in *-are* denote situations which did not take place in the past (169), and most likely will not take place in the future (170).

(169) *urši-li-ni* boy-obl-erg *xunul* wife *k-ib-q'alle,* **take:pfv-aor-ctrf** *nu* I *iχ-di-li-šu-r* that-pl-obl-ad-hpl(ess) *d-uʔ-a-re.* **f1-become:pfv-irr-pst**

'If my son had got married, I would have lived at their place.'

#### Nina Dobrushina

(170) *c'able* tomorrow *w-ebk'-ib-q'alle,* **m-die:pfv-aor-ctrf** *nu--ra* I--add *d-ubk'-a-re.* **f1-die:ipfv-irr-pst** 'If you should die tomorrow, I would also die.'

A conditional clause with a counterfactual converb derived from an aorist refers to the past; if the converb is derived from an imperfective participle, it refers to the present:


Converbs in *-q'alle* based on infinitives refer to the future, but there is an additional meaning of wish. They are also used in independent clauses (§7.4) to express wish. In conditional protasis, they denote desirable situations (173). Therefore, the converb "infinitive + *-q'alle*" is not appropriate if the conditional construction denotes non-desirable situations (175):


#### **8.4 Hypothetical conditional constructions**

Hypothetical conditional constructions denote situations which can either be true in the present, or can be realized in the future, or are habitual. The protasis

#### 5 Moods of Mehweb

of a hypothetical construction is expressed by the converb in *-k'a*. The apodosis can have different forms depending on the semantics of the clause.

(176) *iχ-ini* that-erg *b-arx-le* n-be.right-cvb *b-urh-a-k'a,* n-tell:ipfv-irr-cond *iχ* that(nom) *w-atur* m-free *aʔ-as-i.* drive:pfv-inf-atr 'If he tells the truth, they will let him go.'

Clauses with perfective and imperfective hypothetical conditional converbs in *-k'a* contrast as denoting single *vs.* multiple actions:

(177) *het* that *kung* book *b-elč'-a-k'a* **n-read:pfv-irr-cond** *nu-ni* I-erg *ħa-ze* you.sg.obl-inter(lat) *b-urh-iša* n-tell:ipfv-fut.ego *hel-li-ja* this-obl-gen *χabar.* story 'If you read this book, I will tell you his story.' (178) *d-aq-il* npl-much-atr *kung-ane* book-pl *luč'-a-k'a* **read:ipfv-irr-cond** *d-aq-il* npl-much-atr *si-k'al nuša-ze d-alh-ul.*

what-indef we-inter(lat) npl-know:ipfv-ptcp

'If we read many books, we know many things.'

#### **8.5 Real conditional constructions**

Real conditional clauses presuppose that the state of affairs in the subordinate clause is true. Real conditionals are sometimes treated as reason clauses, since they lack the main feature of conditionals – the lack of knowledge about the state of affairs denoted in the subordinate clause. In Mehweb, this type of conditionals has a special mode of marking, using an analytic construction with the verb cl*arges* 'find, pfv'. This verb is found in many languages of Daghestan in semigrammaticalised constructions designating direct (visual) evidence (cf. Maisak & Daniel 2018).

Conditional clauses of real conditional constructions have an auxiliary verb cl*-arges* marked by the conditional suffix *-k'a*, and the lexical verb.

The main clause of real conditional constructions can have different indicative forms depending on the semantics of the situation. In example (179), the situation of the matrix clause belongs to the past, in examples (180) and (181) it belongs to the future:


'As [it turned out that] Anwar came, he will go to his mother.'

(181) *rasuj-ze* Rasul.obl-inter(lat) *arc* money *kʷe* in.hands(lat) *d-ik-i-le* **npl-happen:pfv-aor-cvb** *d-arg-a-k'a,* **npl-find:pfv-irr-cond** *il-ini* this-erg *abaj-s* mother-dat *sajʁat* gift *as-es.* take:pfv-inf

'As Rasul [as it turned out] has got the money, he will buy the gift to his mother.'

### **9 Apprehensive**

Mehweb has a dedicated form to express apprehension. When used in independent clauses, the apprehensive means that the speaker is afraid that some undesirable situation may come true. The apprehensive is formed with the suffix *-la* attached to the irrealis stem: *-a-la*.

(182) *d-arʔ-a* npl-gather:pfv-imp.tr *mura,* hay *zab* rain *d-aq'-a-la.* **npl-do:pfv-irr-appr** 'Collect the hay, it might rain.'

The apprehensive has a negative counterpart:

(183) *zab* rain *ħa-d-aq'-a-la* **neg-npl-do:pfv-irr-appr** *hab,* ahead *d-aˤq-a* npl-hit:pfv-imp.tr *šin* water *agarod-le-ħe.* vegetable.garden-obl-in(lat) 'Turn on the water in the garden, [because/in case] it might not rain.'

#### 5 Moods of Mehweb

Apprehensive forms are commonly used to express warnings about something that may happen to the addressee:


Apprehensives are often accompanied by the particle *ʡaj*:

(186) *ħu* you.sg *ʁanq'* drown *uh-a-la* **m.become:pfv-irr-appr** *ʡaj.* ptcl 'Beware not to drown.'

First and third person subjects are also available in the apprehensive constructions:


The apprehensive has an inherent negative value. If it is used with reference to situations which are usually viewed as positive, the situation changes its value from positive to negative. Example (190) is grammatical only if the speaker wants to have a daughter more than a son (which is unusual for Daghestan). Example (191) is only grammatical if the speaker does not want to recover from his illness.

(190) *urši* boy *w-aq'-a-la* **m-do:pfv-irr-appr** *ħu-ni* you.sg-erg *d-aq'-a* f1-do:pfv-imp.tr *dursi!* girl '[I am afraid that] you give birth to a boy, [better] give birth to a girl!' Nina Dobrushina

(191) *ara* healthy *d-uh-a-la!* **f1-become:pfv-irr-appr** '[I am afraid that] I become healthy!'

Apprehensive predicates are regularly used in the complement clauses of verbs of fear followed by the complementizer *ile* (which is the perfective converb of the verb 'say'):

(192) *nu* I *uruχ* be.afraid *k'-uwe* lv:ipfv-cvb.ipfv *le-w-ra* be-m-ego *žanawal-li-ni* wolf-obl-erg *maza* sheep *ar-b-uk-a-la* **away-n-lead:pfv-irr-appr** *ile.* comp 'I am afraid that the wolf steals a sheep.'

(193) *nu* I *uruχ* be.afraid *k'-as* lv:ipfv-hab.ego *ħu* you.sg *iz-es* be.ill:ipfv-inf *d-aʔ-a-la* **f1-begin:pfv-irr-appr** *ile.* comp 'I am afraid that you might fall ill.'

If the subject of the apprehensive complement clause is coreferent to the subject of the main clause, the logophoric pronoun *sa*‹cl›*i* is used (see Kozhukhar 2019 [this volume]). This is a phenomenon common to other cases of subordination with the complementizer *ile*.


'The grandmother is afraid of falling down.'

Apprehensives cannot refer to a situation in the past. The next example is ungrammatical (196), and has to be modified as in (197).

(196) *\*nu* I *uruχ* be.afraid *k'-as* lv:ipfv-hab.ego *dag* yesterday *anwal-li-če* Anwar-obl-super(lat) *χʷe* dog *q'ac'* bite *\*b-ik-a-la.* **n-lv:pfv-irr-appr** Intended: 'I am afraid that the dog bit Anwar yesterday.'

(197) *nu* I *uruχ* be.afraid *k'-as* lv:ipfv-hab.ego *dag-ʔʷanal* yesterday-like *anwal-li-če* Anwar-obl-super(lat) *χʷe* dog *q'ac'* bite *b-ik-a-la* **n-happen:pfv-irr-appr** *ile.* comp 'I am afraid that the dog might bite Anwar as it happened yesterday.'

The clause with the apprehensive and complementizer can be inserted into the main clause:


The apprehensive construction without the complementizer cannot be inserted into the main clause:


Apprehensive is used to express negative purpose:


The purpose converb in *-alis* is also used to express negative purpose. Unlike apprehensive, negation in the purpose converb is overtly marked by prefix *ħa-*:

(206) *w-aˤld-e* m-hide:pfv-imp *adaj-ni* father-erg *ħu* you.sg(nom) *dam* beat *ħa-q'-a-lis.* **neg-m.do:pfv-irr-purp** 'Hide, so that your father does not beat you.'

(207) *c'a-li-če* fire-obl-super(lat) *ħule* look *w-iz-e* m-lv:pfv-imp *ħa-b-uš-a-lis.* **neg-n-die(of.fire):pfv-irr-purp** 'Watch the fire so that it does not go out.'

As some other verbal forms, apprehensives can be part of constructions with topicalizing reduplication.

(208) *it* this *w-erχʷ* **m-enter:pfv** *ħa-rχʷ-a-la* **neg-m.enter:pfv-irr-appr** *nu* I *le-l-la* aux-f-ego *uruχ* be.afraid *k'-uwe.* lv:ipfv-cvb.ipfv 'I worry that he may not enter [the university].'

### **10 Discussion**

In this section, I will compare the Mehweb system of non-indicative forms with that of several other Dargwa lects (languages or dialects): Akusha, Ashty, Shiri, Tanti, and Icari. Akusha is especially interesting for this study, because it is suggested that Mehwebs came to the place where they now live from the areas

#### 5 Moods of Mehweb

where the Akusha dialect is spoken (Dobrushina 2019 [this volume]). If this hypothesis is true, we might expect that Mehweb will show more similarity with Akusha than with other Dargwa lects. Another object for the comparison is Avar – the language which is spoken in the vicinity and which could have influenced Mehweb.

The main prominent feature of Mehweb is the absence of personal endings in all non-indicative forms. In this respect, Mehweb is presumably unique among Dargwa languages and dialects. Akusha, Tanti, Shiri, Ashty, Icari – all distinguish persons in the forms of optative and in conditional forms. The loss of personal endings may be due to the influence of Avar, since the latter has no personal paradigm.

Some traces of the former personal endings are still present in the grammar of non-indicative mood forms. The Mehweb prohibitive ends in *-ad*(*i*). In Akusha Dargwa, *-ad* of prohibitive coincides with the second person Future marker (van den Berg 2001: 36). Shiri, Ashty and Icari use the endings *-t*/*-t*: in the prohibitive, which are the markers of the second person in some other forms of these lects (Belyaev (b) manuscript; Mutalov & Sumbatova 2003). Mehweb, however, has the marker*-ad*(*i*) only in the prohibitive, hence synchronically it does not denote person. Sumbatova suggested that the Mehweb prohibitive marker originates from the second person ending (Sumbatova & Lander 2014: 590).

In other respects, however, the Mehweb prohibitive is similar to that of the other Dargwa lects: it is formed by a special negative prefix *ma-* (used only for the prohibitive and the negative optative) and the suffix *-ad*(*i*). In Avar, the prohibitive is expressed by a suffix.

There are several more features which distinguish Mehweb non-indicative mood forms from what is typical for Dargwa lects.

The system of imperative marking is simpler in Mehweb than in other Dargwa dialects. In Akusha, Tanti, Ashty, Shiri, and Icari, the choice of the imperative marker is triggered by three factors: transitivity, aspect and the formal class of the verb. In Mehweb, the formal class is irrelevant for the choice of the imperative marker. The only relevant factors are transitivity and aspect.

It is interesting that the marker of the imperative itself is formally identical to that of Tanti but not to that of Akusha (which is supposed to be closest to Mehweb). In Akusha, Ashty, Shiri and Icari, the marker for perfective transitive imperatives is *-a*, other types of imperative have *-i* or some other marker depending on the class of verb (van den Berg 2001: 48; Belyaev (a) & (b) manuscripts; Mutalov & Sumbatova 2003). In Mehweb, the second class of imperatives takes*-e*, like the Tanti dialect (Sumbatova & Lander 2014: 142). The marker *-e* in Mehweb could have been supported by the imperative of Avar, but the distribution of

#### Nina Dobrushina

Avar markers is opposite to that of Mehweb: *-e* for transitive imperatives, *-a* for intransitive (Charachidzé 1981: 105).

Mehweb differs from other Dargwa idioms in using the marker *-na* for the plural imperative and prohibitive. Akusha, Ashty, Shiri, Tanti, and Icari also mark the plurality of the addressee by a special ending, but in these dialects this marker is identical to the marker of the second person plural in other forms. The Mehweb imperative/prohibitive plural marker differs from other Dargwa lects even formally. In Mehweb, the plural imperative/prohibitive is *-na*; compare to *-ja*/*-aja* in Akusha (van den Berg 2001: 48), *-a*: in Ashty (Belyaev (a) manuscript), *-aja* in Shiri (Belyaev (b) manuscript), *-a*/*-ja* in Tanti (Sumbatova & Lander 2014: 142), *-aja* in Icari (Mutalov & Sumbatova 2003). Note that Avar has no special endings for the second person plural imperative. For the moment, I have no suggestions as to the origin of the marker *-na*.

Unusual for Dargwa idioms are also Mehweb conditional markers. In Akusha, Ashty, Shiri, Tanti, and Icari, conditional forms are marked by the suffix *-li* or *-le*. Counterfactual conditionals in all these lects are derived from hypothetical conditionals with the marker of the past tense. Mehweb conditionals differ both in terms of content and in terms of structure. Mehweb conditionals have other markers than these Dargwa dialects (*-k'a* for hypothetical conditional converb and *-q'alle* for counterfactual; see §8.1 on the probable origin of these markers). The counterfactual form is not formally related to the hypothetical. It seems therefore that the proto-Dargwa conditonal forms were completely substituted in Mehweb by new forms.

The optative of Mehweb has the same marker *-b* as other Dargwa lects. Another similarity is the presence of truncated optative forms in Mehweb as well as in Akusha, Ashty, Shiri and Tanti (see references in §7.1). The difference from other Dargwa lects is that the Mehweb optative has one form for all persons, as I mentioned before. Mehweb is also simpler than the related idioms is that it does not use the optative for commands which have first person object, as do Tanti, Shiri, Ashty, and Icari (I have no information about this construction in Akusha).

As most other Dargwa dialects, Mehweb lacks a dedicated form for the hortative. The meaning of the hortative is regularly expressed by the combination of the particle based on the imperative of 'go' and the infinitive. Unfortunately, there is no sufficient information on how the hortative is expressed in Akusha, Ashty, Shiri, Tanti, and Icari.

As for the jussive, Mehweb uses a periphrastic construction to express it. The combination of the imperative of the verb with the imperative of the verb of speech (lit. 'verb-imp say') is found in several East Caucasian languages

5 Moods of Mehweb

(Akhvakh (Creissels manuscript), Lak and Archi (Dobrushina 2012)), but not among the Dargwa lects discussed above.

Apprehensives seem to be rare in East Caucasian (as well as in other languages of the world). To my knowledge, apart from Mehweb, the apprehensive is attested only in Archi (Kibrik 1977). These forms however are rarely looked for by linguists, so the reason for the infrequency of these forms can as well be their undocumentedness.

### **11 Conclusion**

As this study has shown, there are several features which show the special position of Mehweb among other Dargwa lects. In several cases, Mehweb differs from the other five lects used for comparison, while those five show affinity between them. The study of non-indicative moods did not show any special similarity between Mehweb and Akusha. The influence of Avar, however, is also not attested in these forms. The only feature of the Mehweb system of non-indicative moods which can result from intensive contact with other languages is that, in several respects, it is simpler than the system of other Dargwa lects.

### **List of abbreviations**




### **References**


Sheyanova, Maria. 2019. Specialized converbs in Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 235–253. Berlin: Language Science Press.

Sumbatova, Nina. The sketch of Mewheb grammar. Manuscript.


## **Chapter 6**

## **Periphrastic causative constructions in Mehweb**

### Daria Barylnikova

National Research University Higher School of Economics

In Mehweb, periphrastic causatives are formed by a combination of the infinitive of the lexical verb with another verb, originally a caused motion verb. Various tests that Mehweb periphrastic causatives do not qualify as fully grammaticalized. But the constructions are not compositional expressions, either. While a clause usually contains either a morphological or a periphrastic causative marker, there are instances where, in a periphrastic causative construction, the lexical verb itself may carry the causative affix, resulting in only one causative meaning.

*Keywords*: causative, periphrastic causative, double causative, Mehweb, Dargwa, East Caucasian.

### **1 Introduction**

The causative construction denotes a complex situation consisting of two component events: (1) the event that causes another event to happen; and (2) the result of this causation (Comrie 1989: 165–166; Nedjalkov & Silnitsky 1973; Kulikov 2001). Here, the first event refers to the action of the causer and the second explicates the effect of the causation on the causee.

Causativization is a valency-increasing derivation which is applied to the structure of the clause. In the resulting construction, the causer is the subject and the causee shifts to a non-subject position. The set of semantic roles does not remain the same. Minimally, a new agent is added. With a new argument added, we have to redistribute the grammatical relations taking into account how these participants semantically relate to each other. The general scheme of the causative derivation always implies a participant that is treated as a causer (someone or something that spreads their control over the situation and "pulls

Daria Barylnikova. 2019. Periphrastic causative constructions in Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 167–187. Berlin: Language Science Press. DOI:10.5281/zenodo.3402064

#### Daria Barylnikova

the trigger"). At the same time, there is someone who is, willingly or not, involved in the situation induced by the causer. With two-place predicates there is also another, undergoer participant who does not interact with the causer directly and does not play a role in the redistribution of grammatical relations. This participant retains the marking that it had in the original sentence. The following English examples illustrate these options:


Mehweb has a morphologically productive category of causative (Ageeva 2014; Daniel 2019). The aim of this study is to identify and investigate the means of building periphrastic constructions with causative semantics, with a verb that functions as a separate cause predicate in the construction ("causative verb" below). As noted in Harris & Campbell (1995: 151–194), biclausal structures may undergo simplification over the history of a language and end up as a fused clause. In this paper I shall briefly discuss the degree of grammaticalization of periphrastic causative constructions in Mehweb by considering their clause structure.

I propose the following research questions:


The paper is divided into five sections. They present the results of syntactic tests applied in order to detect whether these constructions are periphrastic causatives or not. §2 surveys possible ways of non-periphrastic expression of the causative meaning, including synthetic and suppletive causatives. §3 introduces lexical verbs participating in periphrastic causative constructions. §4 considers the syntax of such constructions in more detail, in particular, what types of verbs are allowed to be used with each causative verb. In §5, some aspects of forming negative causative clauses are discussed. Finally, §6 provides some evidence on the double causative construction.

6 Periphrastic causative constructions in Mehweb

### **2 Synthetic and suppletive causatives**

There are three possible ways of expressing causative meaning in Mehweb: synthetic (morphological), suppletive (lexical) and analytic (periphrastic).

Synthetic causatives are formed by adding an affix to the verbal stem. Synthetic means of expressing causation usually produce monoclausal structures, with no lexical predicate added to the syntactic structure. In Mehweb, the causative affix *-aq-* is used. It has an allomorph *-aχaq-* with a very limited distribution. The affix can be added to both perfective and imperfective verb bases.

(1) *abaj-ni* mother-erg *urši-li-ze* boy-obl-inter(lat) *kung* book *b-aˤld-aˤq-ib.* **n-hide:pfv-caus-aor** 'Mother made her son hide a book.'

This way of causative derivation is highly productive in Mehweb. The causative affix can be added to all kinds of verbs. For further discussion of morphological causative formation see Ageeva (2014) and Daniel (2019).

Suppletive causatives are also called "covert" causatives (Kulikov 2001), since they share no morphological material with their non-causative equivalents. The English pair *kill* and *die* is commonly treated as an example of suppletive causativization. In Mehweb, the pair cl<sup>1</sup> *-aˤbʡas* 'to kill' and cl*-ebk'es* 'die' is also an example of lexical causativization.

### **3 Periphrastic causativization**

The constructions considered in this paper (originally) represent complementation with several matrix verbs<sup>2</sup> :


Compare the two causative constructions in (2). Ex. (2a) illustrates the synthetic causative expression. (2b) conveys the causative meaning, but involves two verbs. The main predicate is the verb *aʔib* 'drove', and its dependent argument is the verb of caused action (cl*-aˤldes* 'hide').

<sup>1</sup>Here and further I will use glossing cl- to refer to a gender agreement slot (on verb agreement morphology, see Daniel 2019 [this volume]).

<sup>2</sup> Further, verbal forms from the list will be given with the perfective stem as a quotation form.

#### Daria Barylnikova

(2) a. *abaj-ni* mother.obl-erg *urši-li-ze* boy-obl-inter(lat) *kung* book *b-aˤld-aˤq-ib.* n-hide:pfv-caus-aor b. *abaj-ni* mother.obl-erg *urši* boy *kung* book *b-aˤld-es* n-hide:pfv-inf *aʔib.* drive:pfv-aor 'Mother made her son hide a book.'

The lexical meaning of the verb *aʔas* 'drive, cause to move' involves caused motion, describing the action of driving e.g. a herd. The lexical meaning of the verb cl*-aqas* is 'leave', 'leave behind', 'let stay where it is' and expresses the permissive caused motion. Consider examples of non-causative uses of these verbs:


According to Song (2001), analytic causatives include two predicates. One is the predicate of cause, namely a verb that expresses causative impact. It has two functions: (1) to introduce a new argument (the causer), and (2) to establish the new position of the causee. The other predicate which functions as a lexical argument to the predicate of cause is called the predicate of effect. It fills the slot established by the predicate of cause. For instance, in *The concierge made the lobby boy carry the bags on his own* the predicate of cause is the verb *make* and *carry* is the predicate of effect. Below, I follow this terminology.

I will discuss the causative constructions produced by combining cause and effect predicates. Note that the verbs used as predicate of cause continue to be used in their lexical meaning, and this meaning involves an element of causation. The question is thus whether these verbs should be considered grammaticalized expressions of causation. Below, I argue that there is linguistic evidence to conclude that they are, to some extent, grammaticalized.

#### **3.1 The structure of the periphrastic construction**

In Mehweb the syntactic structure of causative constructions requires using a finite predicate of cause and a non-finite predicate of effect. The predicate of cause functions as the predicate of a simple transitive sentence, with its A (the causer) in ergative case and the causee in the absolutive case. The effect predicates are

#### 6 Periphrastic causative constructions in Mehweb

represented by infinitives, either perfective or imperfective (see (5a–b)). Other verbal forms are ungrammatical, either finite or non-finite; cf. examples (5c–e) with the aorist, imperfective past and perfective converb, respectively.

	- b. *adaj-ni* father-erg *kung* book *urši* boy *luč'-es* **read:ipfv-inf** *aʔ-ib.* drive:pfv-aor 'Father made his son be reading the book.'
	- c. \**adaj-ni* father-erg *kung* book *urši* boy *b-elč'-un* **n-read:pfv-aor** *aʔ-ib.* drive:pfv-aor 'Father made his son read the book.'
	- d. \**adaj-ni* father-erg *kung* book *urši* boy *luč'-ib* **read:ipfv-ipft** *aʔ-ib.* drive:pfv-aor 'Father made his son be reading the book.'
	- e. \**adaj-ni* father-erg *kung* book *urši* boy *b-elč'-i-le* **n-read:pfv-aor-cvb** *aʔ-ib.* drive:pfv-aor 'Father made his son read the book.'

The word order is not strict, but there is a preference for sov. Considering the clausality of the whole construction, we may expect the object 'book' of the embedded verb 'read' to be adjacent to it, but it is not. This is, however, not a good criterion for postulating biclausal structure. Native speakers do not seem to be very sensitive to changing word order of the direct and indirect object in the examples above. The finite verb is typically in the final position, and the infinitive immediately precedes it. These two forms cannot be separated by an additional phrase, e.g. by a temporal adverb (see (6c); the rule is only relevant in case if both verbal forms are located at the end of the phrase).

(6) a. *abaj-ni* mother.obl-erg *rasul* Rasul *q'ar* grass *iˤšq-es* mow:ipfv-inf *iʔ-an* drive:ipfv-hab *har* every *barħi.* day b. *har* every *barħi* day *abaj-ni* mother.obl-erg *rasul* Rasul *q'ar* grass *iˤšq-es* mow:ipfv-inf *iʔ-an.* drive:ipfv-hab

#### Daria Barylnikova

c. \**abaj-ni* mother.obl-erg *rasul* Rasul *q'ar* grass *iˤšq-es* mow:ipfv-inf *har* every *barħi* day *iʔ-an.* drive:ipfv-hab 'Mother makes Rasul mow the lawn every day.'

The scope of the temporal phrase depends on the context. Sometimes the temporal or adverbial phrase belongs to the main clause, sometimes it belongs to the subordinate clause. Both readings are available when the temporal phrase is placed at the border between the two clauses. Consider the next example:


In (7), even though the cause predicate has perfective aspect, there are no restrictions on the aspect of the effect predicate. The same is observed in constructions with the cause predicate in the imperfective, where either imperfective or perfective effect predicates are allowed. In other words, aspectual categories of the cause and effect predicates are mutually independent.

Causative semantics has two major subtypes: (a) something is made/urged to be done/happen (factitive causative), and (b) something is not prevented from being done (permissive causative). The first meaning is associated with the verb *aʔas* 'drive'. The second meaning is associated with the verb cl*-aqas* 'leave'.

#### **3.2 The use of** *aʔas* **'drive'**

Factitive causatives (English constructions with *make*, *force*, *get* or *have* someone (to) do something) are formed by means of the verb *aʔas* 'drive'. The causee usually is an animate object. Inanimate objects are incompatible with the semantics of coercion. They can be urged to do something, but due to their lack of volition, they cannot comply (see below for exceptions). The causer is marked with the ergative, while the causee carries the absolutive. Consider examples (8–10):

(8) *pat'imat-ini* Patimat-erg *anwar* Anwar *uˤq'-es* m.go:pfv-inf *aʔ-ib.* drive:pfv-aor 'Patimat made Anwar go away.'

6 Periphrastic causative constructions in Mehweb


The causer is typically an animate agent. However, it is also possible to have an inanimate causer. These uses seem to be explained through personification, attributing control to natural forces.


In (13a) the snow appears as a human causer, not a natural force. In a more realistic situation, for instance after a meltdown in the mountains, the sentence would be as in (13b).

	- b. *doˤʜi-li-ni* snow-obl-erg *ħark'ʷ* river *χʷala* big *b-aq'-ib.* n-do:pfv-aor 'Snow has made the river big.'

Examples with an inanimate causee are not common, but not very difficult to construct. The consultants produce them freely and do not have troubles in identifying the participants' roles. More about the third kind of causative with 'do' see in §3.5.

(14) *ʡali-ni* Ali-erg *adaj-la* father-gen *sune-če-l* self.obl-super(lat)-emph(lat) *naˤʁ* hand *aq b-aq'-as* up *aʔ-ib.* n-(do):pfv-inf drive:pfv-aor 'Ali made his father raise a hand against him.'

#### Daria Barylnikova

The causative construction with the verb *aʔas* is, thus, flexible. It allows using an inanimate as well as an animate causer. The same applies to the causee. In particular, in example (11), the illness is presented as something physically real which functions as a living creature (fairy tale style). While consultants allow such uses, they do not produce them as first answer in the elicitation task but simply accept a constructed sentence. In any case, it is important that there are no strict constraints on animacy of the participants.

#### **3.3 Permissive causative with cl***-aqas* **'leave'**

In the permissive construction, the causer permits rather than causes the causee to bring about the caused event. In Mehweb, it is usually expressed by means of the verb cl*-aqas* 'leave'. The causer carries ergative marking, while the causee is in the absolutive. Consider some examples with different effect predicates (15– 17):


One of the main contexts for the permissive is a positive response to request. For instance, in (17), it is entailed that, before kissing Patimat, Anwar actually asked permission for this action.

(17) *pat'imat-ini* Patimat-erg *anwar* Anwar *w-aq-ib* m-leave:pfv-aor *umma* kiss *d-aq'-as.* npl-do:pfv-inf 'Patimat let Anwar kiss her.'

On the other hand, there may be no inquiries or requests, and the causer is introduced as an independent agent. Inanimate causees are widespread in such contexts. Consider some examples:

(18) *rasuj-ni* Rasul.obl-erg *šin* water *rurq-es* flow:ipfv-inf *d-aq-ib.* npl-do:pfv-aor 'Rasul let the water flow.' (did not prevent this from happening) 6 Periphrastic causative constructions in Mehweb

(19) *rasuj-ni* Rasul.obl-erg *uq'laha* window *abx-es* open:pfv-inf *b-aq-ib.* n-leave:pfv-aor 'Rasul let the window open.' (did not prevent this from happening)

Examples like (18) and (19) can be described in terms of a physical situation in which the causer does not interfere with what is happening to the causee. There are some other effect predicates that denote natural processes. For instance, verbs like *ulč'es* 'become bald', *miʔ aʔʷas* 'freeze', cl*-ic'es* 'melt' in causative constructions usually are found in combination with the cause predicate cl*-aqas* 'leave'. Cf. the following examples:



The permissive constructions in Mehweb are closely connected to the original meaning of the word cl*-aqas* 'leave'. The causer leaves the causee on its own without taking any part in the change of its state. This is especially visible when the causer is an inanimate object (18–21). In cases where the causee is a person (17), the permissive element is evident. The permissive is then understood in a metaphorical sense of not preventing someone's action. I interpret the construction with cl*-aqas* 'leave' as a permissive causative.

#### **3.4 Agreement in permissive causative construction**

The relation between case assignment and gender agreement is relevant only for the verb cl*-aqas* 'leave', because *aʔas* 'drive' does not carry any gender markers. Periphrastic causative constructions allow two agreement patterns. The first one apparently prevails, with the causee retaining the absolutive case (22a). Note that gender agreement on the verb is controlled by the absolutive participant (the masculine gender marker appears on the verb 'leave'). The second pattern shows

#### Daria Barylnikova

marking of the causee by inter-lative<sup>3</sup> case; the gender agreement changes (from masculine to neutral). There is no absolutive participant in the matrix clause to agree with. What we observe is distant agreement between the matrix predicate and the absolutive argument of the dependent clause. Consultants translate both (22a) and (22b) in the same way.

(22) a. *sovet-ini* administration-erg *direktur* principal *uškul* school *q'-aˤbʡ-es* pv-close-inf *w-aq-ib.* m-leave:pfv-aor b. *sovet-ini* administration-erg *direktur-li-ze* principal-obl-inter(lat) *uškul* school *q'-aˤbʡ-es* pv-close-inf *b-aq-ib.* n-leave:pfv-aor

'The administration let the principal close the school.'

In (23), the causative verb shows plural agreement with the absolutive argument in the dependent clause.

(23) *pat'imat-ini* Patimat-erg *urši-li-ze* boy-obl-inter(lat) *d-aq-ib* npl-leave:pfv-aor *d-ix-es* npl-put:pfv-inf *heš-di* (prox)-pl *karawatu-ne* bed-pl *caj-li* one-obl *quli.* room.in(lat) 'Patimat let the boy carry these beds to another room.'

#### **3.5 Adjectival causative**

Adjectives form causatives by means of 'do'-periphrasis, adding the verb cl*-aq'as* 'do' (24b). In Mehweb, this is one of the rare contexts where the adjective cannot be used with the attributive affix (cf. 24b and 24c).

	- b. *χaj-ni* khan.obl-erg *musa* Musa *zuba* blind *w-aq'-ib.* m-do:pfv-aor 'Khan blinded Musa.'

<sup>3</sup> See Chechuro (2019) on the use of the form.

6 Periphrastic causative constructions in Mehweb

c. \**χaj-ni* khan.obl-erg *musa* Musa *zuba-l* blind-atr *w-aq'-ib.* m-do:pfv-aor 'Khan blinded Musa.'

### **4 The syntax of causatives**

#### **4.1 Biclausality**

While morphological causative constructions are monoclausal, periphrastic causatives are apparently biclausal. This means that they have a main clause that contains the causative predicate that introduces the causer and the dependent clause that describes the caused event. The causee also belongs to the matrix clause. In Mehweb, the dependent clause is headed by an infinitive (25).

(25) *anwal-li-ni* Anwar-obl-erg *rasul* Rasul *abaj-ze* mother.obl-inter(lat) *b-arx-le* n-be.right-cvb *b-urh-es* n-tell:pfv-inf *aʔ-ib.* drive:pfv-aor 'Anwar made Rasul tell mother the truth.'

In order to prove that there are two syntactic clauses in periphrastic causative constructions, I use several tests. The first test is based on the case of the causee. In (26), two agentive participants are present. It is impossible to have two ergative arguments in one clause. The verb cl*-erhʷes* 'slaughter' also requires an ergative agent, but only the verb *aʔas* 'drive' assigns the ergative to its agent. The case of the causee is absolutive and is thus assigned by the predicate of cause.


The second test is based on agreement. The verb agrees in gender with the absolutive participant of its clause. If the analytic causative constituted only one clause, it would be possible for a verbal form which is marked with a gender marker to agree with the sole absolutive argument. In (27), the predicate of cause agrees with the absolutive argument (i.e. the causee) in the main clause, whereas the predicate of effect agrees in gender with the other absolutive argument. Changing agreement so that the predicate of cause agrees with *kung*

#### Daria Barylnikova

'book' is ungrammatical. Based on §3.4, one could expect that distant agreement from the embedded clause is available, because, in principle, the matrix verb may agree with the embedded absolutive argument.


The periphrastic causative construction contains two absolutive arguments. Only one of them controls the agreement of the causative verb. The other triggers agreement on the predicate of effect. It is thus biclausal.

### **4.2 Types of predicates of effect**

The predicate of effect fills the valency of the causative verb. In all periphrastic causative constructions the causer gets ergative marking, while the causee appears in the absolutive or inter-lative case. All other arguments preserve their case marking. Below, different possible types of effect predicates with the verb *aʔas* 'drive' (factitive causatives) are discussed. The permissive causative verb cl*-aqas* 'leave' behaves in exactly the same way.

#### **4.2.1 A-intransitive verbs and Р-intransitive verbs**

In general, intransitive verbs are more frequently causativised. An agentive intransitive verb takes one lexical subject in the absolutive case and represents an action, as *duc'* cl*-uqes* 'run' in (28).


The difference between A- and P-intransitive verbs is the degree of control of the subject. While the subject of A-intransitive controls the situation they are involved in, the subject of P-intransitive does not. cf. (29):

6 Periphrastic causative constructions in Mehweb

	- b. *anwal-li-ni* Anwar-obl-erg *inc* apple *b-erħ-es* n-rotten:pfv-inf *b-aq-ib.* n-leave:pfv-aor 'Anwar let the apple rot.'

#### **4.2.2 Experiential verbs**

In East Caucasian, subjects of experiential verbs are non-canonical subjects and take non-core case marking. In Mehweb, they are coded with the inter-lative case (30a), or with a dative with the verb cl*-iges* 'want'. Under causativization, the causee switches from inter-lative to absolutive, according to the general scheme causee case marking in analytic causative constructions.

	- b. *adaj-ni* father-erg *dursi* girl *urši* boy *qum-art-es* forget-lv:pfv-inf *aʔ-ib.* drive:pfv-aor 'Father made his daughter forget the boy.'
	- c. *\*adaj-ni* father-erg *dursi-li-ze* girl-obl-inter(lat) *urši* boy *qum-art-es* forget-lv:pfv-inf *aʔ-ib.* drive:pfv-aor 'Father made his daughter forget the boy.'

With morphological causatives of experiential effect predicates, the causee retains its inter-lative case. Consider the following example, quoted from Ageeva (2014: 8):

	- b. *pat'imat-ini* Patimat-erg *ʡali-ze* Ali-inter(lat) *χabar* tale *arʁ-aq-ib.* hear:pfv-caus-aor 'Patimat told Ali a tale.'

#### Daria Barylnikova

Unlike what happens in morphological causatives, in the analytic causative construction the original marking of the causee as non-canonical subject is ungrammatical (see 30b).

#### **4.2.3 Transitive verbs**

With originally transitive constructions, case marking of the causee changes. In analytic causatives, the causer takes the ergative, leaving the absolutive slot to the causee (32b). Having two ergative arguments in one utterance is not allowed (32c).

	- b. *rasuj-ni* Rasul.obl-erg *uzi* boy *maza* ram *b-erhʷ-es* n-slaughter:pfv-inf *aʔ-ib.* drive:pfv-aor
	- c. \**rasuj-ni* Rasul.obl-erg *uzi-ni* boy-erg *maza* ram *b-erhʷ-es* n-slaughter:pfv-inf *aʔ-ib.* drive:pfv-aor 'Rasul made his brother cut the ram.'

#### **4.2.4 Ditransitive verbs**

Ditransitive verbs take three arguments that correspond to the subject, the recipient and the theme. As with causativization of transitive verbs, analytic causativization of ditransitive verbs does not license two ergative arguments. The causee is coded by inter-lative.


Causativization of transitive and ditransitive verbs thus follows the same scheme, with the causer in ergative and the causee in a peripheral case.

### **5 Negation**

Formation of a negative clause is one of several possible ways for testing the degree of grammaticalization of causative constructions. The negation in constructions with *aʔas* 'drive' is only allowed on the matrix predicate, that is, the predicate of cause. The dependent infinitive cannot take the negation prefix *ħa*-.

	- b. *abaj-ni* mother-erg *rasul* Rasul *q'ar* grass *iˤšq-es* mow:ipfv-inf *ħa-ʔ-ib* neg-drive:pfv-aor *har* every *barħi.* day
	- c. *\*abaj-ni* mother-erg *rasul* Rasul *q'ar* grass *ʜaˤ-šq-es* neg-mow:ipfv-inf *aʔib* drive:pfv-aor *har* every *barħi.* day
	- d. *\*abaj-ni* mother-erg *rasul* Rasul *q'ar* grass *ʜaˤ-šq-es* neg-mow:ipfv-inf *ħa-ʔ-ib* neg-drive:pfv-aor *har* every *barħi.* day 'Mother did not make Rasul mow the lawn every day.'

Examples (34c) and (34d) are considered ungrammatical by consultants no matter what meaning is implied (whether the negation scopes over the embedded predicate 'makes not to mow' or the matrix verb 'does not make mow'). Another example shows the same effect.


b. \**učitel-t-ini* teacher-pl-erg *nuša* we *meħʷe-la* in.Mehweb-gen *mezi-sum* language-repl *ħa-b-uʜ-aˤq'-as* neg-hpl-talk-lv:ipfv-inf *aʔ-ib.* drive:pfv-aor 'The teachers make us not speak Mehweb [at school].'

On the other hand, in constructions with cl*-aqas* 'leave' it is possible to use the negative prefix both on the predicate of effect and on the predicate of cause, with different resulting meanings.


The next pair of examples illustrates the same.

(38) *abaj-ni* mother.obl-erg *urši* boy *ħa-q-ib* neg-m.leave:pfv-aor *uškuj-ħe* school.obl-in(lat) *w-aš-es.* m-go:ipfv-inf 'Mother did not let her son go to school.' (39) *abaj-ni* mother.obl-erg *urši* boy *w-aq-ib* m-leave:pfv-aor *uškuj-ħe* school.obl-in(lat) *ħa-š-es.*

neg-m.go:ipfv-inf

'Mother let her son not to go to school.'

The examples above show the possibility of placing the negative prefix on either the causative or the effect predicate. On the other hand, it is considered ungrammatical to use the negative form of the infinitive of the verb dependent on *aʔas* 'drive'. The verb cl*-aqas* 'leave' forms a looser connection with its predicate of effect and, thus, seems to be less grammaticalized than *aʔas* 'drive'.

6 Periphrastic causative constructions in Mehweb

### **6 Double causative**

Morphological and periphrastic causatives may co-occur. In other words, if a construction already contains a predicate of cause (i.e. *aʔas 'drive'* or cl*-aqas* 'leave'), the predicate of effect can be additionally marked with a causative affix *-aq-*. In (40a) and (40b), the morphological marker is optional and may be dropped, while the analytic causative predicate remains in the sentence and the meaning of the whole does not change.

	- b. *adaj-ni* father-erg *urši* boy *kung* book *b-elč'-es* n-read:pfv-inf *aʔ-ib.* drive:pfv-aor 'Father made his son read a book.'

Constructions with inanimate causees show the same effect.

	- b. *anwal-li-ni* Anwar-obl-erg *inc* apple *b-erħ-aq-as* n-rotten:pfv-caus-inf *b-aq-ib.* n-let:pfv-aor 'Anwar let an apple rot.'

Examples (40) and (41) illustrate a double causative construction. Ageeva (2014: 10) points out that it is possible to build a double morphological causative by adding a second causative affix (cf. cl*-arʡaˤqaqib* 'freeze'). The meaning of the form remains the same, with no (clear) semantic change as compared to the (simple) morphological causative. Here, a similar phenomenon is observed in periphrasis. Constructions with double causative marking sound natural to native speakers and are produced spontaneously during elicitation. Consultants easily derive double analytic causatives from all analytic causatives discussed previously in the paper.

### **7 Conclusions**

Periphrastic causative constructions co-exist in Mehweb with synthetic causatives. There is no difference in meaning between analytic and morphological markers. It does not matter what syntactic type the predicate of effect is; verbs

#### Daria Barylnikova

of all morphosyntactic classes are allowed. There are however some structural limitations on periphrastic causative formation.

There is a semantic division of labor between the causative predicates. Factitive causativization is expressed by means of the verb *aʔas* 'drive'. The permissive meaning is expressed by cl*-aqas* 'leave'. Both predicates introduce an infinitive expressing the predicate of effect. In adjectival causativization, the cl*-aq'as* 'do' is used.

Cause predicates also show other differences. The verb *aʔas* 'drive' only allows animate causees. The verb cl*-aqas* 'leave' also allows inanimate causees. In both factitive and permissive constructions, the negation marker may attach to the matrix predicate. However, the verb cl*-aqas* 'leave' also allows negation on the infinitive.

These differences are summarized in Table 1.


Table 1: Summary of the causative predicates.

In terms of case assignment, arguments other than the causee behave identically with all morphosyntactic types of predicates. The causer is always marked with the ergative. Other arguments retain their original case marking. As to the causee, the intransitive causee keeps its original absolutive marking and the transitive causee is marked with the inter-lative. No causative construction seems to allow two ergative arguments, marking both the causee and the causer with the ergative. This is similar to what happens under morphological causativization. Morphological and analytic causativization, however, become different if one compares what happens to the causee of experiential predicates with originally non-canonical subjects. Under morphological causatives, the causee keeps its original peripheral case marking (dative or inter-lative, depending on the verb). In analytic causative constructions, however, it obligatorily changes to the absolutive.

Causative constructions in Mehweb may combine morphological and analytic causative strategy together. Apparently, the meaning of such constructions is not different from the usual causative construction with either a synthetic or an analytic causative alone. Double causatives seem to be semantically redundant.

The tests discussed in the paper reveal some divergences between the constructions under consideration. The results of the negation test show that the factitive causative construction, apparently, is more grammaticalized than the permissive causative. It is not possible to apply negation to the dependent verb form in constructions with the verb *aʔas* 'drive', while cl*-aqas* 'leave' allows the negation either on the main verb or on the infinitive.

The negation test and agreement tests diverge. While the negation test in factitive causatives indicates a monoclausal structure, gender agreement indicates two separate clauses. Only negative constructions support grammaticalization of periphrastic causatives in Mehweb.

### **Acknowledgements**

I would like to express my gratitude to all Mehweb consultants who agreed to help in this research. I also thank Nina Dobrushina, Michael Daniel, Dmitri Ganenkov, Denis Creissels and Samira Verhees.

### **List of abbreviations**


#### Daria Barylnikova


### **References**

Ageeva, Ekaterina. 2014. Kauzativy v megebskom jazyke [Causatives in Mehweb]. In E. M. Devjatkina, D. S. Ganenkov, Ju. V. Mazurova, D. V. Makhovikov & A. B. Shluinskij (eds.), *Problemy jazyka: Sbornik naučnyx statej po materialam tret'ej konferencii-školy "Problemy jazyka: Vzgljad molodyx učenyx"*. Moscow: Institut jazykoznanija RAN.

Barylnikova, Daria. 2015. *Quasi-causative constructions in Mehweb*. Working papers by NRU HSE, series WP BRP "Linguistics" 26. HSE.

Chechuro, Ilya. 2015. *Nominal morphology of Mehweb Dargwa*. Working papers by NRU HSE, series WP BRP "Linguistics" 32. HSE.

Chechuro, Ilya. 2019. Nominal morphology of Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 39–72. Berlin: Language Science Press.


## **Chapter 7**

## **Case and agreement in Mehweb**

### Dmitry Ganenkov

University of Bamberg; National Research University Higher School of Economics

The chapter deals with patterns of case marking and agreement in Mehweb. On the basis of morphosyntactic coding and anaphoric binding, a system of five valency classes is described. The chapter covers basic monoclausal structures with verbs of the five valency classes as well as their interaction with several specific constructions, such as reciprocal, causative, and biabsolutive.

*Keywords*: case, personal agreement, gender, transitivity, experiential verbs, dative verbs, subject, reported speech, biabsolutive construction.

The present chapter deals with the morphosyntax of argument expression in Mehweb. In many respects, Mehweb is a fairly typical representative of the Dargwa branch of Nakh-Daghestanian, and of the Nakh-Daghestanian family in general. In certain respects, however, the language displays rare features only attested in a few other languages of the family. Three linguistic phenomena – argument case marking, gender agreement, and person agreement – are the focus of this chapter. The three coding properties are interrelated in many ways and together constitute major surface evidence about grammatical functions (including subjecthood), supported by other diagnostics, such as the binding of reflexive and reciprocal pronouns. They also generally determine how the Mehweb verbal lexicon can be broken down into verb (valency) classes. The notion of *core argument* will be key to capturing the system of valency classes. In this chapter, I define *core argument* as a clausal constituent expressed by a noun phrase that is able to determine at least one type of verbal agreement, either gender or person, or both. Depending on the number of core arguments and their morphosyntactic

Financial support from the Basic Research Program of the National Research University Higher School of Economics (Moscow) is gratefully acknowledged.

Dmitry Ganenkov. 2019. Case and agreement in Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 189–234. Berlin: Language Science Press. DOI:10.5281/zenodo.3402066

#### Dmitry Ganenkov

behavior with respect to coding properties, the Mehweb verbal lexicon is divided into the following valency classes:

	- a. *Intransitive verbs* have a single core argument in the absolutive that triggers both person and gender agreement.
	- b. *Transitive verbs* feature two core arguments. One core argument, the subject, is in the ergative case and triggers person agreement on the finite verb. The other core argument, the direct object, is in the absolutive case and determines agreement features in the gender agreement slot.
	- c. *Locative subject verbs* are also bivalent verbs with two core arguments. Instead of an ergative argument, as with transitive verbs, they possess a core experiencer argument in the spatial case called *inter-lative*, see Chechuro (2019) [this volume] for details of the nominal paradigm. Like the ergative subject of a transitive verb, the inter-lative (henceforth, locative) subject of a locative-subject verb also triggers person agreement.
	- d. *Dative subject verbs* have one core argument in the absolutive, which triggers gender agreement only. No argument of a dative subject verb is able to determine person agreement on its own.
	- e. The *inter-elative subject verb buhes* 'manage, be able' features one core argument in the inter-elative case which optionally triggers person agreement but cannot control gender agreement.

The rest of this chapter provides empirical evidence about the behavior of various types of verbal arguments that motivates the above classification. §1 describes patterns of case marking and provides evidence from reflexive binding about the relative structural prominence of verbal arguments. §2 and §3 deal with rules of gender and person agreement. §4 presents an overview of case marking and agreement in reciprocal constructions. §5 discusses causative constructions. §6 describes basic properties of the biabsolutive construction. The conclusion briefly summarizes the main issues covered in the paper.

7 Case and agreement in Mehweb

### **1 Case marking and structural prominence**

Mehweb is a morphologically ergative language where the sole argument (S) of intransitive verbs is grouped together with the direct object (P) of transitive verbs with regard to morphological case marking, but separately from the subject (A) of transitive verbs: S and P arguments are in the unmarked absolutive case, while A arguments bear the ergative case.


In (2), the DP *ʡali* 'Ali (a man's name)' is in its unmarked form and functions as the core argument of the intransitive verb *bak'es* 'come'. In (3), the same form is used to express the direct object (patient) of the transitive verb *buces* 'catch, seize'. In (4), however, the DP functions as the subject of the transitive verb *baˤbʡes* 'kill' and thus must be in the ergative case.

An absolutive case DP is present in almost every Mehweb clause. In intransitive clauses, the absolutive argument is the highest one from the structural point of view, as seen from the fact that it can bind reflexive pronouns in any other position, but cannot be bound itself by any other argument.<sup>1</sup> Example (5) shows the intransitive verb *ħule* cl*-izes* 'look' with an oblique argument which is diagnosed as structurally less prominent than the clause-mate absolutive argument.

<sup>1</sup> In this paper, to diagnose structural prominence, I employ sentences with *wh*-pronouns serving as antecedents of reflexive pronouns. This is necessary in order to exclude the possibility of the co-reference relation between the antecedent and the reflexive (Reinhart 1981). Co-reference is normally available with referential antecedents and works on pragmatic rather than strictly syntactic grounds in Mehweb. In particular, the "antecedent" can appear in a structurally lower position in co-reference, as in (i), which is not a grammatical option under semantic binding by non-referential antecedents (quantified, *wh*-pronouns), cf. (5b).

<sup>(</sup>i) *sune-la-l* self-gen-emph *urši* son(abs) *madina-če* Madina-super(lat) *ħule* look *w-iz-ur.* m-lv:pf-aor 'Her<sup>i</sup> son looked at Madina<sup>i</sup> (a woman's name).'

#### Dmitry Ganenkov

	- a. *čija* who(abs) *ħule* look *d-iz-ur-a* f1-lv:pf-aor-q *sune-la-l* self-gen-emph *urši-li-če?* son-obl-super(lat) 'Who<sup>i</sup> looked at her<sup>i</sup> son?'
	- b. *\*sune-la-l* self-gen-emph *urši* son(abs) *hi-če* who-super(lat) *ħule* look *w-iz-ur-a?* m-lv:pf-aor-q 'Who<sup>i</sup> did her<sup>i</sup> son look at?'

The linear order plays no role in acceptability of the "reversed" anaphoric binding, thus attesting to the relevance of weak crossover effects in reflexive binding, as shown in example (5c).

	- c. *\*hi-če* who-super(lat) *sune-la-l* self-gen-emph *urši* son(abs) *ħule* look *w-iz-ur-a?* m-lv:pf-aor-q 'Who<sup>i</sup> did her<sup>i</sup> son look at?'

The absolutive argument is not restricted to expressing any particular thematic role. It can denote an agentive participant, a patientive participant, or an experiencer. Unergative and unaccusative verbs in Mehweb are not distinguished by case marking. Some intransitive verbs are listed in (6).

(6) Intransitive verbs

*a--izes* 'stand up', *arces* 'fly', *aqas* 'raise, climb', *--alħʷes* 'wake up', *--ebk'es* 'die', *--erʔʷes* 'become dry', *--ises* 'cry', *--usaʔʷas* 'fall asleep', *--urdes* 'become worn', *--ušes* 'die out (of fire)', *--uzes* 'work', *kalʔes* 'remain', *uruχ --aˤqes* 'get afraid'

Two-place verbs are those verbs that mark their structurally highest argument with a morphological case other than the absolutive. As suggested in (1) above, depending on the particular case of the highest argument, two-place verbs fall into three classes: transitive verbs with ergative subjects, locative subject verbs with inter-lative subjects, and dative subject verbs with dative subjects.

With transitive verbs, the ergative-marked argument is structurally the most prominent, as evidenced by its ability to bind a reflexive pronoun in any other position in the clause, including the absolutive argument, as in (7a-8a). The reverse binding of the ergative reflexive by an oblique or absolutive argument is impossible, as shown in (7b) and (8b).

7 Case and agreement in Mehweb

	- a. *hinija* who(erg) *haraq'e* forward *ihʷ-es-a* throw:pf-fut-q *sune-la-l* self-gen-emph *urši?* son(abs) 'Who<sup>i</sup> will deceive his<sup>i</sup> son?'
	- b. *\*sune-la-l* self-gen-emph *urši-li-ni* son-obl-erg *čija* who(abs) *haraq'e* forward *ihʷ-es-a?* throw:pf-fut-q 'Who<sup>i</sup> will his<sup>i</sup> son deceive?'
	- a. *hinija* who(erg) *sune-la-l* self-gen-emph *urši-li-s* son-obl-dat *kumak* help(abs) *b-aq'-ib-a?* n-do:pf-aor-q 'Who<sup>i</sup> helped his<sup>i</sup> son?'
	- b. *\*sune-la-l* self-gen-emph *urši-li-ni* son-obl-erg *hi-sa* who-dat *kumak* help(abs) *b-aq'-ib-a?* n-do:pf-aor-q 'Who<sup>i</sup> did his<sup>i</sup> son help?'

Apart from agents, the ergative argument of a transitive verb can also denote a non-agentive causer (see also Chechuro 2019 [this volume] on the instrumental function of the ergative).


Ergative case is thus tightly associated with agentive and causative semantics and is not employed to express participants with other thematic roles. Almost every transitive clause contains an absolutive argument. Exceptions are very few and can be summarized as follows.

With verbs of contact like cl*-aʔaqas* 'hit (an inanimate object)' and cl*-aˤqas* 'hit (an animal)', the absolutive argument expresses the instrument. Generally, instruments are never obligatory and can be freely omitted from overt expression.

#### Dmitry Ganenkov

The absolutive argument in the instrumental function thus often does not appear overtly.


The absolutive argument, when omitted from overt expression, is arguably still present in the sentence, as evidenced by the possibility of non-default (neuter plural) gender agreement.

(14) *ʡali-ni* Ali-erg *unza-li-ze* door-obl-inter(lat) *d-aʔ-aq-ib.* npl-hit-lv:pf-aor 'Ali hit the door (with his fists).'

In (14), the plural gender marking on the verb reflects the plurality of the instrumental DP in the absolutive.

With some transitive verbs of speech and thought, the absolutive argument denotes the content of speech/thought.


Likewise, many such verbs alternatively subcategorize for either an absolutive DP argument or a clausal argument. In the latter case, no absolutive argument is present in the clause.

(17) *rasuj-ni* Rasul+obl-erg *abzulaj-ze* all+obl-inter(lat) *b-urh-ib* n-tell:pf-aor *murad-li* Murad-erg *mašina* car(abs) *as-ib* take:pf-aor *ile.* comp 'Rasul told everyone that Murat had bought a car.'

7 Case and agreement in Mehweb

With some complex transitive verbs, a nominal constituent in the unmarked form functions as a non-verbal component.


The morphosyntactic status of such unmarked nominals is not clear. They can be analyzed either as absolutive-cased DPs or as (pseudo)-incorporated caseless NPs. More work is needed to decide on this question.

Some verbs are P-labile, that is, have both a transitive use and an intransitive use where the subject of the intransitive use corresponds to (i.e. expresses the same participant as) the direct object of the transitive use. The verbal lexicon has not been systematically studied for P-lability. The transitive use with an unspecified (omitted) subject of a P-labile verb and the intransitive use of the same verb are distinguished by: (i) different imperative marking (see Daniel 2019 [this volume]), and (ii) the ability of the absolutive argument to trigger person agreement on the finite verb (see §3.1 below).

Two other classes of two-place verbs are locative subject verbs and dative subject verbs. The locative subject class includes the verbs *arʁes* 'hear, understand', *bahes* 'know', *barges* 'find', *gʷes* 'see'.


Dmitry Ganenkov

The dative subject class includes the verbs *biges* 'want, love', *bikes* 'happen', *eba buhes* 'get bored', *určeb leb* 'remember', *urče bak'as* 'recall', *urče bikes* 'recall'.


The verb *qumartes* 'forget' alternatively allows for either locative or dative case marking on its subject.

(29) {*ʡali-ze* Ali-inter(lat) / / *ʡali-s*} Ali-dat *deč'* song(abs) *qum-art-ur.* forget-lv:pf-aor 'Ali forgot the song.'

The inter-lative (locative) and dative arguments are the highest arguments in their respective clauses. Again, this is evidenced by the ability of the locative/dative argument to bind any other argument (including the absolutive), while the reverse binding pattern is ungrammatical.

	- a. *hi-ze* who-inter(lat) *g-ub-a* see:pf-aor-q *sune-la-l* self-gen-emph *urši?* son(abs) 'Who<sup>i</sup> saw her<sup>i</sup> son?'
	- b. *\*sune-la-l* self-gen-emph *urši-li-ze* son-obl-inter(lat) *čija* who(abs) *g-ub-a?* see:pf-aor-q 'Who<sup>i</sup> did her<sup>i</sup> son see?'
	- a. *hi-sa* who-dat *ħa-d-ig-ul* neg-f1-love:ipf-ptcp *sune-la-l* self-gen-emph *abaj?* mother(abs) 'Who<sup>i</sup> does not love his<sup>i</sup> mother?'
	- b. *\*sune-la-l* self-gen-emph *abaj-s* mother-dat *čija* who(abs) *ħa-d-ig-ul?* neg-f1-love:ipf-ptcp 'Who<sup>i</sup> does his<sup>i</sup> mother not love?'

Again, while the absolutive argument generally must be present in a clause with a locative or dative subject verb, it may be absent in case the corresponding semantic argument is expressed by another constituent. Most locative and dative subject verbs allow a clausal complement instead of the absolutive argument.

(32) *arʁes* 'hear' with finite complement *ʡali-ze* Ali-inter(lat) *arʁ-ib* hear:pf-aor [*abaj* mother(abs) *iz-uwe* be.sick:ipf-cvb.ipfv *le-r* aux-f *ile*]. comp 'Ali heard that mother was sick.' (33) *biges* 'want' with infinitival complement *rasuj-s* Rasul+obl-dat *dig-uwe* want:ipf-cvb.ipfv *le-b* aux-n [*anži-li* Makhachkala-in(lat) *uˤq'-es*]. (m)go:pf-inf 'Rasul wants to go to Makhachkala.' (34) *bikes* 'happen' with a finite complement *abzulaj-s* everyone+obl-dat *b-ik-ib* n-happen:pf-aor [*ʡali* Ali(abs) *w-ebk'-i-le* m-die:pf-aor-cvb *ile*]. comp

'Everyone thought (lit. it occurred to everyone) that Ali was dead.'

Finally, the verb *buhes* 'manage, be able' is the only verb in Mehweb that licenses a core argument in the inter-elative case.

(35) *rasuj-ze-la* Rasul+obl-inter-el *ajz-es* (m)rise:pf-inf *ħa-b-urh-an.* neg-n-manage:ipf-hab 'Rasul cannot stand up.'

Dmitry Ganenkov

(36) *rasuj-ze-la* Rasul+obl-inter-el *ħa-b-uh-ub* neg-n-manage:pf-aor *ʁarʁa* stone(abs) *aq* up *b-aq'-as.* n-do:pf-inf 'Rasul did not manage to lift the stone.'

To summarize, Mehweb has five verb classes depending on the case of the structurally highest argument: (i) intransitive verbs with absolutive subject, (ii) transitive verbs with ergative subject, (iii) locative subject verbs with inter-lative subject, and (iv) dative subject verbs with dative subject, and (v) one inter-elative subject verb *buhes* 'manage, be able'. The argument structure of all verbs includes an absolutive argument. As will be shown below, the subject and the absolutive argument (when they are different) play a special role in gender and person agreement, and thus are called *core arguments*. All other arguments are *oblique*.

### **2 Verbal gender agreement**

Two morphological slots for gender agreement are potentially available in the Mehweb clause. One is the prefixal gender agreement marker on lexical verbs. Every verbal stem is specified for whether it hosts the prefixal gender agreement slot. Most verbs are specified to host this agreement marker in their perfective stems. In imperfective stems, the slot is often absent. For more on agreement morphology and its relation to stems, see Daniel (2019) [this volume].

	- b. *urši-li-ni* boy-obl-erg *kaʁar-t* letter-pl(abs) *luk'-an.* write:ipf-hab 'The boy writes letters (every day).'

The verb 'write' has a prefixal slot for gender agreement in its perfective stem, as shown in (37a), but lacks any such slot in its imperfective stem, as in (37b). If a stem features gender agreement, it is obligatory in any verbal form based on this stem, be it finite or non-finite.

The other morphological slot for gender agreement in the verbal complex is the suffix on the auxiliary in periphrastic verbal forms.

(38) *urši-li-ni* boy-obl-erg *kaʁar-t* letter-pl(abs) *luk'-uwe* write:ipf-cvb.ipfv *le-r.* aux-npl 'The boy is writing letters.'

The rule of thumb for gender agreement in monoclausal structures is to agree with the clause-mate absolutive argument. With regard to gender agreement on lexical verbs, this means that agreement is always with the absolutive subject of an intransitive verb or with the absolutive direct object of other verb classes, as shown below.

	- b. *dursi* girl(abs) *d-ak'-ib.* f1-come:pf-aor 'The girl came.'
	- b. *sinka-li* bear-erg *ʡali* Ali(abs) *w-aˤbʡ-ib.* m-kill:pf-aor 'A bear killed Ali.'
	- b. *adaj-ze* father-inter(lat) *dursi* girl(abs) *d-arg-ib.* f1-find:pf-aor 'Father found his daughter.'
	- b. *ʡali-s* Ali-dat *madina* Madina(abs) *d-ig-ib.* f1-love:ipf-ipft 'Ali loved Madina.'

#### Dmitry Ganenkov

If a clause lacks an absolutive argument, as observed with some types of formally transitive verbs, gender agreement on the lexical verb appears as the default singular neuter agreement marker *b*-. This is also observed with intransitive impersonal predicates. See examples in §1 above.

The verb *buhes* 'manage, be able' subcategorizes for an inter-elative subject and an infinitival complement and thus does not have an absolutive argument. This verb, therefore, invariably appears with the default (singular neuter) marker *b-*, as in examples (35) and (36) above.

The second morphological slot for gender agreement appears on the auxiliary within periphrastic verbal forms like Present and Past Progressive, Present and Past Resultative. This slot cross-references the gender-number features of the highest absolutive argument or shows the default (neuter singular agreement) in clauses with no absolutive argument.

	- b. *dursi* girl(abs) *iz-uwe* be.sick:ipf-cvb.ipfv *le-r.* aux-f 'The girl is sick.'
	- b. *rasuj-ze* Rasul+obl-inter(lat) *madina* Madina(abs) *d-alh-uwe* f1-know:ipf-cvb.ipfv *le-r.* aux-f 'Rasul knows Madina.'
	- b. *rasuj-s* Rasul+obl-dat *madina* Madina(abs) *d-ig-uwe* f1-love:ipf-cvb.ipfv *le-r.* aux-f 'Rasul loves Madina.'

7 Case and agreement in Mehweb

In complex verbs that include an adjectival stem specified for prefixal gender agreement as a non-verbal component, the adjective always agrees with the absolutive argument.

	- b. *adam-ule-ni* man-pl-erg *hun-be* road-pl(abs) *d-aˤʡu* npl-wide *d-aq'-ib.* npl-do:pf-aor 'Men widened the roads.'

If a sentence contains two absolutive arguments, as attested in biabsolutive constructions, the auxiliary agrees with the subject (see §6).

### **3 Verbal person agreement**

### **3.1 Intransitive, transitive, and locative subject verbs in synthetic indicative forms**

In synthetic indicative tense-aspect forms (aorist, imperfect, habitual, future), person agreement operates on a nominative-accusative basis and cross-references the person of the subject: the absolutive argument of intransitive verbs, the ergative argument of transitive verbs, or the inter-lative argument of locative subject verbs.


Morphologically, person inflection only distinguishes two options. One is a form overtly specified for person (*-iša* in the Future, *-s* in the Habitual, *-ra* in other indicative tense-aspect forms), the other is a non-agreeing form. A peculiar

#### Dmitry Ganenkov

feature of Mehweb is that person agreement is sensitive to the illocutionary force of the utterance.<sup>2</sup> In declarative sentences, the overt person marker signals a first person subject, whereas non-agreeing forms are observed with second and third person subjects. By contrast, the same person marker indicates second person subject in interrogative sentences, while first and third person subjects do not trigger overt person marking on the verb. The following question-answer pairs illustrate.


Example (51) shows that second person subjects in interrogatives and first person subjects in declaratives obligatorily require overt person marking, whereas subjects in reverse the combinations of person and illocutionary force – first person subjects in interrogatives and second person subjects in declaratives – can never trigger person marking, as example (52) demonstrates. (For discussion of one notable exception see §3.4 below.)

Person marking on synthetic tense-aspect forms is obligatory with intransitive absolutive subjects and transitive ergative subjects and cannot be omitted. Locative subject verbs display variation here. The verb *gʷes* 'see' patterns with the

<sup>2</sup>This type of agreement system is also referred to as *egophoric*, *conjunct*/*disjunct*, or *assertive agreement*, see Creissels (2008) who discusses assertive agreement in another Nakh-Daghestanian language, Akhwakh (Andic branch).

transitive and intransitive verbs in requiring person agreement, whereas with all other locative subject verbs, person marking is optional.


Similar to locative subject verbs, the inter-elative subject of the verb *buhes* 'manage, be able' triggers overt person marking only optionally.


Non-subjects, including absolutive direct objects, inter-lative indirect objects (addressee, causee), inter-elative arguments (including involuntary agents) and other oblique arguments can never trigger person agreement.


Dmitry Ganenkov


This strict subject orientation of agreement allows us to distinguish between transitive and intransitive uses of P-labile verbs, as shown in the following examples.

(63) a. *nu* I(abs) *quli-w* house-m(ess) *w-aˤld-un-na*. m-hide:pf-aor-ego 'I hid in the house.' b. *nu* I(abs) *quli-w* house-m(ess) *w-aˤld-un*. m-hide:pf-aor 'They hid me in the house.'

In (63a), the presence of the agreement marker on the verb indicates that the first person singular pronoun *nu* is in the subject position, and that the sentence therefore instantiates the intransitive use of the labile verb. The absence of agreement in (63b) can only indicate that the absolutive pronoun is in the direct object position and that we are thus dealing with the transitive use of the labile verb.

### **3.2 Dative subject verbs**

Unlike subjects of intransitive, transitive, and locative subject verbs, dative subjects do not trigger overt person agreement.

(64) *nab* I(dat) *rasul* Rasul(abs) {*w-ig-an* m-love:ipf-hab / / \**w-ig-as*}. m-love:ipf-hab.ego 'I love Rasul.'

7 Case and agreement in Mehweb


The contrast between locative and dative subject verbs is clearly seen in sentences with the verb *qumartes* 'forget'. Recall that this verb allows both locative and dative subjects. With a first person locative subject, the verb has optional person agreement, as with other locative subject verbs. With a first person dative subject, the verb cannot show overt person marking, as is usual with dative subject verbs.

(70) a. *di-ze* I-inter(lat) *ʡali* Ali(abs) *qum-art-ur*(*-ra*). forget-lv:pf-aor-ego b. *nab* I(dat) *ʡali* Ali(abs) *qum-art-ur*(-\**ra*). forget-lv:pf-aor-ego 'I forgot Ali.'

In sentences with dative subjects, absolutive direct objects do not trigger person agreement either, as shown in examples (71) to (74).

#### Dmitry Ganenkov


The absence of agreement with the absolutive argument is unexpected given the fact that many of the dative subject verbs clearly go back to intransitive structures where absolutive arguments diachronically go back to intransitive subjects, and thus could act as agreement triggers, contrary to fact.

	- b. *X* dat *Y* abs *urče-b* in.heart-n(ess) *le-b.* be-n 'To X, Y is on heart.'
	- c. *X* dat *Y* abs *urče* in.heart(lat) *b-ak'-as.* n-come:pf-inf 'To X, Y comes to heart.'

The clear contrast between intransitive and dative subject constructions with respect to person agreement is observed in a construction with the verb *haraq'e bak'as* (lit. 'come forward'), which denotes "illusionary seeing", as in dreams or hallucinations, as in (76).

(76) *rasuj-s* Rasul+obl-dat *tamaša-l* surprising-atr *si-k'al-t* what-indef-pl *haraq'e* forward *d-ik'-uwe* npl-come:ipf-cvb.ipfv *le-r.* aux-npl 'Rasul sees something bizarre.' (lit. 'Something bizarre is coming forward to Rasul.')

As in other dative subject structures, neither of the two arguments, the dative subject or the absolutive direct object, is able to trigger person agreement on the verb.

(77) a. *nab* I(dat) *tamaša-l* surprising-atr *si-k'al-t* what-indef-pl *haraq'e* forward {*d-ak'-ib* npl-come:pf-aor / / \**d-ak'-i-ra*}. npl-come:pf-aor-ego 'Something bizarre appeared to me.' b. *rasuj-s* Rasul+obl-dat *nu* I(abs) *haraq'e* forward {*w-ak'-ib* m-come:pf-aor / / \**w-ak'-i-ra*}. m-come:pf-aor-ego

'I appeared to Rasul (in a hallucination).'

Overt person marking on the verb *bak'as* 'come' in the latter example is grammatical only in the literal sense of physical movement.

(78) *rasuj-s* Rasul+obl-dat *nu* I(abs) *haraq'e* forward {*w-ak'-i-ra* m-come:pf-aor-ego / / \**w-ak'-ib*}. m-come:pf-aor 'I came forward to Rasul.' (not: 'I appeared to Rasul (in a hallucination).')

We therefore have a minimal pair: in the same construction with *haraq'e bak'as* 'come forward', person agreement with the first person absolutive argument is obligatorily required when denoting physical movement and completely prohibited when referring to imaginary visions.

To sum up, neither of the two arguments of a dative subject verb – the dative subject or the absolutive direct object – can control person agreement on their own. Strikingly enough, overt person marking on a finite dative subject verb is nevertheless possible in constructions where both the dative subject and the

#### Dmitry Ganenkov

absolutive direct object are first person (i.e. in reflexive constructions with a first person subject).

(79) *nab* I(dat) *nu--wal* I(abs)--emph *w-ig-as.* m-love:pf-hab.ego 'I love myself.'

The syntax of dative subject constructions and the way they interact with person agreement require further syntactic analysis.

#### **3.3 Agreement in the Present Progressive**

Present Progressive forms exhibit a different pattern of person agreement in sentences with transitive and locative subject verbs. Unlike other indicative forms, not only the person feature of the subject is taken into account here, but also the person feature of the direct (absolutive) object.

The descriptive generalization is that overt person agreement with the first person subject is only possible (and obligatory) when the absolutive direct object is a locutor (first or second person). Otherwise, with third person direct objects, person agreement is ungrammatical, and the finite verb is in the unmarked form.<sup>3</sup>

	- b. *nu-ni* I-erg *ħu* you.sg(abs) *ulc-uwe* (m)catch:ipf-cvb.ipfv *le-w-*\*(*ra*). aux-m-ego 'I am catching you (male).'

<sup>3</sup> In transitive clauses with third person direct objects, such as (80a), first person marking is marginally accepted by some native speakers. It is not clear where such marginal acceptability stems from. One option could be that optional person agreement in these configurations is actually a part of Mehweb grammar. Another option, however, is that it arises from confusion with biabsolutive constructions where person agreement with the subject is obligatory in the Present Progressive (see §6). Indeed, many speakers, when accepting person agreement in examples like (80a), tend to rephrase the ergative construction of (80a) into the corresponding biabsolutive construction with the absolutive subject, with subject-controlled person and gender agreement on the auxiliary. Note that with locative subject verbs, which are not easily allowed in biabsolutive constructions, person agreement in the Present Progressive is definitely rejected by all speakers, see (81a).

7 Case and agreement in Mehweb

b. *di-ze* I-inter(lat) *ħu* you.sg(abs) *irg-uwe* see:ipf-cvb.ipfv *le-w*-\*(*ra*). aux-m-ego 'I can see you.'

Examples (80a) and (81a) show that agreement with first person subjects is impossible in the presence of a third person absolutive direct object. By contrast, agreement is obligatory when the direct object is also a locutor. Relative specification of the subject and the direct object for number plays no role in the availability of person agreement.

	- we-erg cauldron(abs) n-wash:ipf-cvb.ipfv aux-n-ego 'We are washing the cauldron.'

### **3.4 Matrix infinitival questions**

One exception to the generalization that only second, but not first, person subjects trigger person agreement in interrogative sentences concerns agreeing Future forms, which may co-occur with first person subjects in interrogatives, yielding questions with modal semantics.

(84) *nu-ni* I-erg *ħad* you.sg(dat) *sija* what(abs) *g-iša?* give:pf-fut.ego+q 'What should I give you?' (not: 'What will I give you?') Dmitry Ganenkov

(85) *nu* I(abs) *uˤq'-iša?* (m)go:pf-fut.ego+q 'Should I go?' (not: 'Will I go?')

Examples like (84) and (85) are remarkable in two respects. First, they are only available in the Future, and not in other tense-aspect forms.

(86) *\*nu-ni* I-erg *ħad* you.sg(dat) *sija* what(abs) *g-i-ra?* give:pf-aor-ego+q intended: 'What should I have given you?' (or 'What did I give you?')

Second, the modal interpretation of the questions in (84) and (85) only arises with first person subjects, but never with second person subjects, cf. the contrast between (87) and (88).

(87) *nu* I(abs) *kuda* where *uˤq'-iša?* (m)go:pf-fut.ego+q 'Where should I go?' (not: 'Where will I go?')

(88) *ħu* you.sg(abs) *kuda* where *uˤq'-iša?* (m)go:pf-fut.ego+q 'Where will you go?' (not: 'Where should you go?')

This contrast raises the question whether the two sentences in (87) and (88) contain the same or two different verb forms. This question is especially relevant in the light of the fact that the infinitive in Mehweb is formally identical to nonagreeing future forms, which appear, for example, in declarative sentences with second/third person subjects, as shown in (89).

	- b. *ʡali-s<sup>i</sup>* Ali-dat [pro<sup>i</sup> abs *šaˤ-baˤʜ* village-dir *uˤq'-es*] (m)go:pf-inf *dig-uwe* want:ipf-cvb.ipfv *le-b*. aux-n 'Ali wants to go to the village.'

The infinitive and the future are normally distinguished in contexts with overt person marking (e.g. declarative sentences with first person subjects). The Future takes overt person marking, while the infinitive never does so, as shown in (90). 7 Case and agreement in Mehweb

	- b. *nab<sup>i</sup>* I(dat) [pro<sup>i</sup> abs *šaˤ-baˤʜ* village-dir *uˤq'-es*] (m)go:pf-inf *dig-uwe* want:ipf-cvb.ipfv *le-b*. aux-n 'I want to go to the village.'

Now note that across Dargwa languages, the modal semantics found in the Mehweb examples in (84), (85), (87) is commonly expressed by a special form with a first person marker added on top of the infinitive, as seen in (91) from Chirag Dargwa.

	- *di-cːe* I-erg *χabar-e* story-pl(abs) *d-urs-i-da-j?* npl-tell:pf-inf-ego-q 'Should I tell the stories?'

The same modal semantics is cross-linguistically characteristic of matrix infinitival questions (cf. English *Where to go?* or German *Wohin gehen?*, Bhatt 2006: 108, 110).

It is natural to propose that Mehweb modal questions as in (84) and (85) actually involve a combination of the infinitive and overt person marking rather than the formally identical agreeing form of the Future, as suggested by (i) the formal identity between the infinitive and the future in non-agreeing forms and (ii) the morphological evidence that the combination of infinitive with first person marking may yield the modal semantics of 'should' in other Dargwa languages.

### **3.5 Indexical shift and agreement shift in embedded reports**

Person agreement as described above is only available in finite clauses: no nonfinite clause can feature a person agreement marker. The following examples show that person agreement is unavailable in complements headed by nominalizations.

	- a. *nu-ni* I-erg *kung* book(abs) {*b-elč'-un-deš* n-read:pf-aor-nmlz / / \**b-elč'-un-na-deš*}. n-read:pf-aor-ego-nmlz '… that I read (past) the book.'

Dmitry Ganenkov

> b. *nu-ni* I-erg *ħu* you.sg(abs) *ulc-uwe* (m)catch:ipf-cvb.ipfv {*le-w-deš* aux-m-nmlz / / \**le-w-ra-deš*}. aux-m-ego-nmlz '… that I am catching you.' c. *nu-ni* I-erg *kung-ane* book(abs) {*luč'-an-deš* read:ipf-hab-nmlz / / \**luč'-as-deš*}. read:ipf-hab.ego-nmlz '… that I read (habitual) books.'

Apart from the independent finite clauses described above, Mehweb also features finite complement clauses with the complementizer *ile*. Etymologically, the complementizer stems from (and is still synchronically identical) the perfective converb of the verb *es* 'say'. It is used with verbs of speech and thought to introduce reported speech (attitude reports).

(93) a. *abaj-s* mother-dat *b-ik-ib* n-happen:pf-aor *ca* one *insan* person(abs) *w-ak'-ib* m-come:pf-aor *ile.* comp 'Mother thought that someone had come.' b. *abaj-ni* mother-erg *b-urh-ib* n-tell:pf-aor *ca* one *insan* person(abs) *w-ak'-ib* m-come:pf-aor *ile.* comp 'Mother said that someone had come.' c. *abaj* mother(abs) *uruχ* be.afraid *d-aˤq-ib* f1-lv:pf-aor *ca* one *insan* person(abs) *w-ak'-ib* m-come:pf-aor *ile.* comp 'Mother feared that someone had come.'

Personal pronouns and person agreement in embedded reports under the complementizer *ile* are subject to *person shift* (*indexical shift* and *agreement shift*, respectively), see Schlenker (2003),Anand & Nevins (2004),Nikitina (2012), Shklovsky & Sudo (2014) on indexical shift in a theoretical and typological perspective.

Indexical shift affects the interpretation of first and second person pronouns and is always optional. Personal pronouns in embedded reports may refer not only to the participants of the actual speech act, as in independent finite clauses, but also to the participants of the speech act denoted by the matrix clause. In

the latter case, the first person pronoun refers to the reporter (attitude holder) expressed as the subject of the matrix clause, while the second person pronoun denotes the addressee of the matrix reporter.

	- a. 'Rasul<sup>i</sup> said that my<sup>j</sup> car was broken.' (unshifted reading of the 1st person pronoun)
	- b. 'Rasul<sup>i</sup> said that his<sup>i</sup> car was broken.' (shifted reading of the 1st person pronoun)
	- a. 'Madina said to Rasul<sup>i</sup> that your<sup>j</sup> car was broken.' (unshifted reading of the 2nd person pronoun)
	- b. 'Madina said to Rasul<sup>i</sup> that his<sup>i</sup> car was broken.' (shifted reading of the 2nd person pronoun)

With matrix verbs selecting for a complement clause with *ile* but lacking an addressee, such as matrix verbs of thought, only first person pronouns can be shifted, while second person pronouns only denote the addressee of the actual speech act.

(96) *rasul* Rasul(m) *uruχ* be.afraid *w-aˤq-ib* m-lv:pf-aor *di-la* I-gen *mašin* car(abs) *b-uˤrʡ-uˤb* n-break:pf-aor *ile.* comp


comp

'Rasul<sup>i</sup> fears that your<sup>j</sup> car was broken.' (only unshifted reading of the 2 nd person pronoun)

#### Dmitry Ganenkov

Person agreement in finite embedded clauses is subject to obligatory *agreement shift*. Only arguments denoting the participants of the reported speech act can control person agreement, while other arguments including those representing the participants of the actual speech act can never trigger agreement. In declarative embedded clauses, only embedded subjects denoting the closest reporter / attitude holder trigger overt agreement on the verb. One possibility is that the embedded subject is expressed by the shifted first person pronoun.

(98) *rasul* Rasul(abs) *uruχ* be.afraid *w-aˤq-ib* m-lv:pf-aor *nu-ni* I-erg *mašin* car(abs) *b-uˤrʡ-aq-i-ra* n-break:pf-caus-aor-ego *ile.* comp 'Rasul<sup>i</sup> feared that he<sup>i</sup> had broken the car.'

In (98), the subject is expressed by the first person pronoun that undergoes indexical shift; that is, it does not refer to the speaker of the actual speech act, but rather to the attitude holder (Rasul) expressed as the subject of the matrix clause. The embedded verb thus shows obligatory overt agreement for person.

Another possibility is that the embedded subject is expressed by the longdistance reflexive pronoun bound by the matrix subject representing the attitude holder. The long-distance reflexive thus ends up being co-referent with the attitude holder, and the verb obligatorily shows overt person marking.

(99) *rasul* Rasul(abs) *uruχ* be.afraid *w-aˤq-ib* m-lv:pf-aor *sune-jni* self-erg *mašin* car(abs) *b-uˤrʡ-aq-i-ra* n-break:pf-caus-aor-ego *ile.* comp 'Rasul<sup>i</sup> feared that he<sup>i</sup> had broken the car.'

No other argument can trigger person agreement on the finite verb in embedded reports, including unshifted first person pronouns denoting the speaker of the actual speech act. Example (100) illustrates.

(100) *rasul* Rasul(abs) *uruχ* be.afraid *w-aˤq-ib* m-lv:pf-aor *nu-ni* I-erg *mašina* car(abs) {*b-uˤrʡ-aq-ib* n-break:pf-caus-aor / / \**b-uˤrʡ-aq-i-ra*} n-break:pf-caus-aor-ego *ile.* comp 'Rasul<sup>i</sup> feared that I<sup>j</sup> had broken the car.'

#### 7 Case and agreement in Mehweb

Kozhukhar (2019) [this volume] reports that overt person marking with an unshifted first person pronoun is also possible in examples like (100). Indeed, consultants sometimes judge such sentences to be acceptable. I maintain, however, that overt person agreement with an unshifted first person pronoun is ungrammatical, and the judgments must stem from confusion. First person pronouns strongly tend to shift their reference in embedded reports, and consultants usually struggle to recognize that the pronoun could refer to the actual speaker. So, when presented with a sentence containing a first person pronoun and overt person marking on the verb, some consultants judge it acceptable due to the fact that they have a different reading in mind. Instead of the reference to the speaker of the actual speech act, they interpret the pronoun as denoting the attitude holder. However, if a suitable example is constructed where the confusion is not possible because of overt morphological marking, overt person marking with unshifted first person pronouns is uniformly judged unacceptable. Consider the following examples.

(101) *abaj-s* mother-dat *b-ik-ib* n-happen:pf-aor *nu* I(abs) *usaʔ-uwe* (m)fall asleep:pf-aor.cvb *le-w*(-\**ra*) aux-m-ego *ile.* comp 'Mother<sup>i</sup> thought that I<sup>j</sup> had fallen asleep.'

(102) *abaj* mother(abs) *uruχk'-uwe* be.afraid:ipf-cvb.ipfv *le-r* aux-f *nu* I(abs) {*arik-es* (m)fall:pf-fut / / \**arik-iša*} (m)fall:pf-fut.ego *ile.* comp 'Mother<sup>i</sup> is afraid that I<sup>j</sup> am going to fall down.'

In (101) and (102), the first person pronoun in the embedded clause is unambiguously interpreted as denoting the actual speaker, since masculine gender marking appears on the embedded verb (both on the converb of the lexical verb and the auxiliary), indicating that the referent of the first person pronoun is a man. Since the attitude holder ('mother') is unambiguously female, the embedded first person pronoun may only receive a disjoint reference, and thus denote the speaker of the actual speech act. In this configuration, overt agreement was unanimously considered grossly ungrammatical.

Agreement shift thus makes possible various mismatches between the "lexical" person feature of an argument and verbal person agreement. On the one hand, third person reflexive pronouns trigger overt person marking, as in (99). On the other hand, first person pronouns referring to the actual speaker can never trigger overt person agreement, as in (100) through (102).

#### Dmitry Ganenkov

The examples above show that the attitude holder can be lexically expressed in the embedded clause by either a shifted first person pronoun or a long-distance reflexive pronoun. However, these two options cannot co-occur within the same embedded clause. In the presence of a long-distance reflexive bound by the matrix subject, first person pronouns are obligatorily interpreted as referring to the speaker of the actual speech act.

	- i. \*'Rasul<sup>i</sup> feared that he<sup>i</sup> broke his<sup>i</sup> car.'
	- ii. 'Rasul<sup>i</sup> feared that he<sup>i</sup> broke his<sup>j</sup> car.'
	- iii. \*'Rasul<sup>i</sup> feared that I<sup>j</sup> broke his<sup>i</sup> car.'

In (103), the embedded clause includes both the first person pronoun in the ergative subject position and the possessive reflexive pronoun that modifies the direct object. The two cannot be interpreted as denoting the same participant, as shown by the ungrammaticality of reading (i). Two further options are logically possible: either the first person pronoun or the reflexive is interpreted as denoting the attitude holder. In the former case, the reflexive must have disjoint reference (long-distance bound by an even higher subject or a free logophor, see Kozhukhar 2019 [this volume]), as indicated in interpretation (ii). In the latter case, the first person pronoun must refer to the actual speaker, which is not possible in this sentence, since unshifted first person pronouns do not trigger verbal person marking, hence the ungrammaticality of reading (iii). Should the finite verb in the embedded report be in the unmarked form *buˤrʡaqib*, reading (iii) becomes available.

In interrogative embedded clauses, a similar distribution is observed: only arguments co-valued with the addressee of the reporter (expressed as the addressee argument of the matrix verb) show overt person marking on the embedded verb, whereas unshifted second person pronouns cannot trigger overt person marking.

(104) *rasuj-ni* Rasul-erg *madina-ze* Madina-inter(lat) *xarba-ib* ask:pf-aor *ħu* you.sg(abs) *kuda* where {*d-aš-as-a* f1-walk:ipf-hab.ego-q / / \**d-aš-an-a*} f1-walk:ipf-hab-q *har* every *barħi* day *ile.* comp 'Rasul asked Madina<sup>i</sup> where she<sup>i</sup> goes every day.'

7 Case and agreement in Mehweb

(105) *rasuj-ni* Rasul-erg *madina-ze* Madina-inter(lat) *xarba-ib* ask:pf-aor *ħu* you.sg(abs) *kuda* where {*w-aš-an-a* m-walk:ipf-hab-q / / \**w-aš-as-a*} m-walk:ipf-hab.ego-q *har* every *barħi* day *ile.* comp 'Rasul asked Madina where you go every day.'

Again, in examples like (105), the second person pronoun in the embedded clause may only be interpreted as disjoint from the matrix addressee argument, due to a gender mismatch between the feminine gender of the matrix addressee and the masculine gender agreement on the embedded verb. When this is the case, overt person agreement is ungrammatical with a second person pronoun in interrogative embedded clauses.

For the sake of completeness, a few words are in order about the availability of indexical shift and agreement shift. As mentioned above, both are only possible in finite complement clauses with the complementizer *ile* under verbs of speech and thought, but not in other types of complements. The examples below demonstrate that indexical shift and agreement shift are possible in the finite complement of the verb *arʁes* 'hear', but not in the factive non-finite (nominalized) complement with the same verb.

	- a. 'Rasul<sup>i</sup> realized that my<sup>j</sup> car was broken.' (unshifted reading of the 1 st person pronoun)
	- b. 'Rasul<sup>i</sup> realized that his<sup>i</sup> car was broken.' (shifted reading of the 1st person pronoun)
	- a. 'Rasul<sup>i</sup> realized that my<sup>j</sup> car was broken.' (unshifted reading of the 1 st person pronoun)
	- b. \*'Rasul<sup>i</sup> realized that his<sup>i</sup> car was broken.' (shifted reading of the 1st person pronoun)

#### Dmitry Ganenkov

Whether or not a matrix verb combines with *ile*-complements is not lexically determined, but rather depends on the semantics of the matrix verb (speech or thought report). This is clearly seen in cases like those shown in the following examples.


Example (108) shows that the factive matrix verb *bahes* 'know' does not combine with finite *ile*-complements. In (109), the causative *bahaqas* of the same verb is normally understood as denoting a speech act ('let know, inform'), and is therefore compatible with an *ile*-complement.

### **4 Reciprocals**

Reciprocal pronouns consist of two instances of the numeral *ca* 'one' adjacent to one another.

(110) *uz-be-ni* brother-pl-erg *ca-li-ni* one-obl-erg *ca-li-če* one-obl-super(lat) *b-aʔ-aq-ib.* n-hit-lv:pf-aor 'The brothers hit each other.'

As can be seen in the example above, the two components of the reciprocal bear independent case marking. One component is always in the case of the subject, while the other component bears the case of the second argument of the reciprocal construction. The distribution of case marking on the two components of the reciprocal pronoun depends on the particular argument/case combination.

Absolutive case, whether it corresponds to the subject or to the direct object, is always marked on the second component of the reciprocal. The first component therefore bears the case of the other argument participating in the reciprocal construction.

7 Case and agreement in Mehweb

(111) *uz-be* brother-pl(abs) *ca-li-če* one-obl-super(lat) *ca* one(abs) *ħule* look *b-iz-ur.* hpl-lv:pf-aor 'The brothers looked at each other.'

(112) *uz-be-ni* brother-pl-erg *ca-li-ni* one-obl-erg *ca* one(abs) *b-aˤbʡ-ib.* hpl-kill:pf-aor 'The brothers killed each other.'

In (111), the intransitive verb *ħule* cl*-izes* 'look' is used in the reciprocal construction. The absolutive case of the subject is marked on the second part of the reciprocal, whereas the case of the oblique argument is marked on the first part. In (112), the transitive verb *baˤbʡas* 'kill' participates in the reciprocal construction. Again, the absolutive case, which is the case of the direct object here, is marked on the second part of the reciprocal pronoun, while the ergative case of the transitive subject is marked on the first part.

When no absolutive argument participates in a reciprocal construction, the case marking on the reciprocal pronoun is determined by structural prominence. The first component is in the case of the higher argument, while the second component is in the case of the lower argument, as in (110) above and in the following examples.


The case of the overt antecedent NP also depends on the presence of an absolutive argument in the construction. As a rule, the overt antecedent bears the case of a more structurally prominent argument. Examples (110), (112), (113), and (114) above show that in the reciprocal construction with transitive verbs, the overt antecedent is in the ergative case. Example (111) shows that the reciprocal construction with intransitive verbs requires an overt antecedent in the absolutive case. Example (115) below illustrates the reciprocal construction with locative subject verbs.

(115) *uz-be-ze* brother-pl-inter(lat) *ca-li-ze* one-obl-inter(lat) *ca* one(abs) {*g-ub* see:pf-aor / / *b-ah-ur* hpl-know:pf-aor / / *b-arg-ib* hpl-find:pf-aor / / *qum-art-ur*}. forget-lv:pf-aor 'The brothers {saw / recognized / found / forgot} each other.'

The only exception to this rule comes with dative subject verbs, where absolutive marking of the overt antecedent is preferred over dative marking.

(116) {*it-ti* this-pl(abs) / / ⁇*it-ti-li-s*} this-pl-obl-dat *ca-li-s* one-obl-dat *ca* one(abs) *b-ig-uwe* hpl-love:ipf-cvb.ipfv *le-b.* aux-hpl 'They love each other.' (117) {*it-ti* this-pl(abs) / / ⁇*it-ti-li-s*} this-pl-obl-dat *ca-li-s* one-obl-dat *ca* one(abs) *eba* bored *b-uh-ub.* hpl-become:pf-aor

'They got bored with each other.'

The absolutive marking of the overt antecedent is also possible in reciprocal constructions with two core arguments of two-place verbs.


Therefore, we have two possibilities for antecedent marking in constructions featuring the two core arguments of two-place verbs. The antecedent can be marked for the morphological case of the higher argument (i.e. the subject) or for the absolutive case, even though the absolutive is the morphological case of the lower argument (i.e. the direct object) in such configurations. With dative subject verbs, the first option is severely disfavored and the second option is pre7 Case and agreement in Mehweb

ferred, while with other two-place verbs (transitive and locative subject), the two options are equally acceptable.

No other reciprocal construction allows the overt antecedent in the case of a lower argument. Example (120) illustrates this for a combination of the intransitive subject and an oblique argument, cf. (111). Example (121) shows a reciprocal construction with a transitive subject and a dative recipient, cf. (113).


In transitive constructions where the absolutive direct object does not participate in the reciprocal relation, the absolutive case cannot be used to mark the overt antecedent either.

(122) \**ul-e* child-pl(abs) *ca-li-ni* one-obl-erg *ca-li-s* one-obl-dat *kumak* help(abs) *b-aq'-ib.* n-do:pf-aor 'The kids helped one another.'

Gender agreement in reciprocal constructions functions according to the general rule of agreement with the absolutive argument. In structures with an overt absolutive NP this is straightforward, as shown in examples (111) and (116) through (119). In structures with no overt absolutive NP, as in (112) and (115), the verb shows the gender and number features of the overt antecedent.

Person agreement also works as usual in constructions where the overt antecedent is in the morphological case of the subject; that is, first person intransitive absolutive, transitive ergative, and locative subjects trigger overt person marking on the finite verb.


Dmitry Ganenkov

(125) *nuša-ze* we-pl-inter(lat) *ca-li-ze* one-obl-inter(lat) *ca* one(abs) {*g-ub-ra* see:pf-aor-ego / / *b-ah-ur-ra*}. hpl-know:pf-aor-ego 'We {saw / recognized} each other.'

In structures with the overt antecedent in the absolutive case, as in (116) through (119), first person pronouns also trigger obligatory person marking.


The reciprocal construction with the absolutive marking of the antecedent thus behaves like an intransitive structure with respect to person agreement.

### **5 Causative construction<sup>4</sup>**

Morphologically, the causative construction is formed by means of the suffix *-aq-* (*-aχaq-*) attached to an aspectual stem of the causativized verb, as described by Daniel (2019) [this volume]. Syntactically, the causative morpheme introduces an additional participant which is interpreted as the causer of the event described by the lexical stem. The causer is always marked by ergative case. Case marking of the causee depends on the class of the causativized verb. Absolutive subjects of intransitive verbs always remain in the absolutive case. The causative construction based on an intransitive verb thus features two arguments: the ergative causer and the absolutive causee, as with regular transitive verbs.

(128) a. *ʡali* Ali(abs) *w-alħ-un.* m-wake.up:pf-aor 'Ali woke up.'

<sup>4</sup>The description of case marking in causative constructions in this section is based on Ageeva (2014).

b. *pat'imat-ini* Patimat-erg *ʡali* Ali(abs) *w-alħ-aq-ib.* m-wake.up:pf-caus-aor 'Patimat woke up Ali.'

Ergative subjects of transitive verbs obligatorily receive locative (inter-lative) marking in the causative construction. Case marking of the causee with transitive causativized verbs does not depend on the degree of agentivity. Both agentive and non-agentive transitive causees are in the inter-lative.


Rasul+obl-erg river-inter(lat) / river-erg wood(abs) *d-erʁ-aq-ib.* npl-sweep.away:pf-caus-aor 'Rasul floated the wood down the river.' (literally: 'Rasul made the river sweep away the wood.')

Locative subjects of the verbs 'see', 'hear, understand', 'find', 'know', and 'forget' are marked with inter-lative case when they occur as a causee in the causative construction. This is the same marking as in the baseline construction.


Dmitry Ganenkov


It is not quite clear whether the locative case of the causee in causative constructions with locative subject verbs reflects the inter-lative subject marking assigned by the lexical verb or the inter-lative causee marking assigned in the causative construction.

Causatives of two locative subject verbs exhibit special behavior as they can denote a situation with no additional causer of the event. Instead, the experiencer subject acquires a higher degree of agentivity and is marked by ergative case, cf. examples (21) and (29) above.


When a dative subject verb is causativized, the experiencer participant can either remain in the dative, as in the original construction, or bear inter-lative marking assigned to the causee in the causative construction.

(138) a. *nab* I(dat) *it* this *dehʷ* word(abs) *urče* in.heart(lat) *b-ik-ib.* n-happen:pf-aor 'I recalled that word.' b. *abaj-ni* mother-erg {*di-ze* I-inter(lat) / / *nab*} I(dat) *it* this *dehʷ* word(abs) *urče* in.heart(lat) *b-ik-aq-ib.* n-happen:pf-caus-aor 'Mother reminded me of that word.'

The difference in interpretation between the two variants of causee marking relates to the degree of control exhibited by the causer over the caused situation. Dative marking implies a lesser degree of involvement of the causer, while interlative marking indicates a more direct causation on the part of the causer.

The causative form of the verb *biges* 'want, love' does not normally have a causative interpretation. Neither the number of arguments nor their case marking changes. The semantics is usually conveyed as 'like' rather than 'love' (as is the case with the underived forms of *biges*).

(139) *nab* I(dat) *it* this *dursi* girl(abs) *d-ig-aq-uwe* f1-love:ipf-caus-cvb.ipfv *le-r.* aux-f 'I like this girl.'

The causative reading of the causative form of the verb *biges* 'want, love' is also accepted by many speakers, though not by all of them, and often not without hesitation. As in causatives of other dative subject verbs, the causee can be marked by either dative or inter-lative case (with no sharp interpretational differences between the two variants).

(140) *adaj-ni* father-erg {*di-ze* I-inter(lat) / / *?nab*} I(dat) *it* this *dursi* girl(abs) *d-ig-aq-uwe* f1-love:ipf-caus-cvb.ipfv *le-r.* aux-f 'Father makes me love this girl.'

Gender and person agreement in the causative construction follows the rules operative in transitive clauses. Gender agreement on the lexical verb is always with the absolutive argument. Gender agreement on the auxiliary in progressive verb forms is also with the absolutive argument.

	- Ali-erg Patimat(abs) f1-wake.up:pf-caus-aor 'Ali woke up Patimat.'

Dmitry Ganenkov

> b. *nu-ni* I-erg *urši-li-ze* boy-obl-inter(lat) *inc-be* apple-pl(abs) *d-uk-aq-uwe* n.pl-eat:ipf-caus-cvb.ipfv *le-r.* aux-n.pl 'I am making the boy eat apples.'

Person agreement is controlled by the ergative causer according to the rules described above in §3.1 and §3.3. This includes the restriction on overt marking in the Present Progressive, as shown in (142). The inter-lative causee or the absolutive argument can never control person agreement.



Note, however, that despite the absence of an overt ergative argument in causative constructions based on transitive verbs, it is possible to show that they do contain an unexpressed ergative subject of the lexical verb. This is seen from case marking that appears on reciprocal pronouns. As explained in §4 above, the two parts of the reciprocal pronoun always bear two different morphological cases corresponding to the case marking of the arguments in the reciprocal relation. When used in a causative construction describing a reciprocal relationship between the causee and the absolutive direct object, one part of the reciprocal pronoun shows up in the ergative case, even though no overt ergative argument appears on the surface.

7 Case and agreement in Mehweb

(146) *madina-jni* Madina-erg {*ul-e* child-pl(abs) / / *ul-e-ze*} child-pl-inter(lat) *ca-li-ni* one-obl-erg *ca* one(abs) *b-az-aq-ib.* hpl-wash:pf-caus-aor 'Madina made the kids wash one another.'

Note that in example (146), the causee in the causativized reciprocal construction of the transitive verb can be expressed by the absolutive or by the inter-lative. This corresponds to two possibilities observed in non-causativized reciprocals: (i) the overt subject is marked by the absolutive, and the whole construction behaves as an intransitive structure, or (ii) the overt subject is marked by the ergative, and the whole reciprocal construction is a transitive structure. Under causativization, the intransitive variant (i) of the reciprocal construction yields absolutive marking of the causee, whereas the transitive variant (ii) of the reciprocal construction yields inter-lative marking of the causee.

### **6 The biabsolutive construction**

Periphrastic verbal forms with durative semantics (present and past progressive) allow for an alternative layout of argument case marking with transitive verbs. Instead of the standard transitive pattern with an ergative subject and an absolutive object, transitive verbs can participate in the *biabsolutive construction*, where both the subject and the direct object are expressed in the absolutive case.<sup>5</sup> Changes in argument case marking are accompanied by a change in gender agreement on the auxiliary, which is controlled by the absolutive subject; gender agreement of the lexical verb is invariably controlled by the absolutive direct object.

(147) Q: *sija* what(abs) *b-iq'-uwe* n-do:ipf-cvb.ipfv *le-w-a* aux-m-q *rasul?* Rasul(abs) 'What is Rasul doing?' A: *rasul* Rasul(abs) *kung* book(abs) *luč'-uwe* read:ipf-cvb.ipfv *le-w.* aux-m 'Rasul is reading a book.'

<sup>5</sup> See Forker (2012) for an overview of the biabsolutive across Nakh-Daghestanian. Gagliardi et al. (2014) present a minimalist analysis of the biabsolutive construction in the Nakh-Daghestanian languages Lak and Tsez. Harris and Campbell discuss the diachrony of the biabsolutive construction (1995: 187–189).

#### Dmitry Ganenkov

Unlike ergative constructions with periphrastic forms, the biabsolutive construction shows no restrictions on person agreement of the absolutive subjects. Overt person marking with the absolutive subject is obligatory, as shown in (148a), cf. the minimally different example (148b), where the ergative subject cannot agree with the finite verb.

	- I-erg book(abs) read:ipf-cvb.ipfv aux-m-ego 'I am reading a book.'

Unlike what is attested in related languages (Forker 2012), there seem to be no observable differences in semantics between the ergative and biabsolutive alignment of the transitive clause. In fact, the biabsolutive construction is often resorted to when person agreement with the subject fails in certain subject-object combinations in periphrastic forms, see §3.3.

Synthetic verbal forms with imperfective semantics do not allow the biabsolutive construction.


Only clauses with agentive subjects normally participate in the biabsolutive construction, whereas clauses with non-agentive subjects are either considerably degraded or completely ungrammatical.


7 Case and agreement in Mehweb

(153) \**zab* rain(abs) *mura* hay(abs) *d-aˤlʜʷ-aˤq-uwe* npl-become wet:ipf-caus-cvb.ipfv *le-r.* aux-npl 'The rain is making the hay wet.'

Similarly, non-agentive subjects of locative-subject verbs are not allowed to participate in the biabsolutive construction for many speakers, though some sentences are judged to be more acceptable than others. The acceptability of locativesubject verbs in the biabsolutive construction may depend on semantic and pragmatic factors and requires further investigation.


'The boy is finding the hidded money.'

(156) ⁇*rasul* Rasul(abs) *het* this *dehʷ* word(abs) *b-alh-uwe* n-know:ipf-cvb.ipfv *le-b.* aux-n 'Rasul knows that word.'

The dative subject verb *biges* 'love, want' can occasionally participate in the biabsolutive construction.

(157) *nu* I(abs) *het* this *urši* boy(abs) *w-ig-uwe* m-love:ipf-cvb.ipfv *le-l-la*. aux-f-ego 'I love this boy.'

Despite initial appearances, the biabsolutive construction contains an unexpressed ergative argument of the lexical verb which can be seen in reciprocal constructions. Similar to what is found in causative constructions, one of the two components of the reciprocal pronoun in the biabsolutive always bears the ergative case licensed by the lexical verb, despite the phonological absence of an ergative argument, compare (158) with (113) above.

Dmitry Ganenkov

(158) *ul-e* child-pl(abs) *ca-li-ni* one-obl-erg *ca-li-s* one-obl-dat *kumak* help(abs) *b-iq'-uwe* n-do:ipf-cvb.ipfv *le-b*. aux-hpl 'The kids help one another.'

Syntactically, the biabsolutive construction may thus be analyzed as consisting of two layers. The lower layer is headed by the lexical verb and contains the lexical verb itself and all of its arguments in their respective cases. The higher layer is headed by the copula and contains the absolutive subject. The biabsolutive construction thus has two important properties: (i) it requires the subject to have the agent theta-role, and (ii) it includes an unexpressed ergative argument which is obligatorily interpreted as having the same reference as the overt absolutive subject. These two properties make the biabsolutive construction look like an obligatory control construction. A schematic representation of the syntactic structure of the biabsolutive construction is given in (159).

	- b. [CopP npabs [vP proerg npabs V] cop]

The causative construction may also be transformed into a biabsolutive construction. With causatives of intransitive verbs, the biabsolutive construction works the same way as with biabsolutives of ordinary transitive verbs: both the causer and the causee are in the absolutive case. The former controls gender and person agreement on the copula, while the latter controls gender agreement on the lexical verb.

(160) *rasul* Rasul(abs) *c'a* fire(abs) *d-uš-aq-uwe* npl-die.out:ipf-caus-cvb.ipfv *le-w.* cop-m 'Rasul is extinguishing the fire.'

With causatives of transitive verbs, there are three case marking options in the biabsolutive construction. One option is to mark the causer with absolutive case, as with causatives of intransitive verbs above. Gender and person agreement on the copula are determined by features of the higher absolutive, in this case the causer. Example (161) shows the baseline causative construction in (a) and the biabsolutive construction with absolutive marking of the causer in (b).

7 Case and agreement in Mehweb

	- b. *abaj* mother(abs) *urši-li-ze* boy-obl-inter(lat) *kung* book(abs) *luč'-aq-uwe* read:ipf-caus-cvb.ipfv *le-r.* cop-f 'Mother makes the boy read the book.'

The second option is to mark the causee with the absolutive case, whereas the causer bears its usual ergative case. Again, gender and person agreement on the copula are determined by features of the higher absolutive, which is the causee in this case.

(161) c. *abaj-ni* mother-erg *urši* boy(abs) *kung* book(abs) *luč'-aq-uwe* read:ipf-caus-cvb.ipfv *le-w.* cop-m 'Mother makes the boy read the book.'

Finally, the third option is to mark both the causer and the causee with absolutive case. We therefore have three absolutive arguments in the same clause. Again, gender and person agreement on the copula is determined by the highest absolutive, that is, the subject causer.

(161) d. *abaj* mother(abs) *urši* boy(abs) *kung* book(abs) *luč'-aq-uwe* read:ipf-caus-cvb.ipfv *le-r.* cop-f 'Mother makes the boy read the book.'

The possibilities of case marking shown in (161c–d) require further investigation. In standard biabsolutive constructions as described in this section, the absolutive marking of the transitive subject apparently becomes available due to the presence of a second clausal layer headed by the copula. It is not quite clear how the copula in the progressive could license absolutive marking of the transitive causee in (161c) and, especially, the absolutive marking of both the ergative causer and the transitive causee in (161d). Any syntactic speculation on this question, however, requires more specific assumptions about the clause structure and mechanisms of case licensing which lie outside the scope of the present work. I therefore leave this issue for another occasion.

### **7 Summary**

In this chapter, I have discussed the major morphosyntactic properties of monoclausal sentences in Mehweb, including case marking, gender and person agreement. The paper describes the system of Mehweb verbal (valency) classes on the basis of their arguments' morphosyntactic behavior and ability to bind reflexive pronouns. I distinguish (i) intransitive verbs with absolutive subjects, (ii) transitive verbs with ergative subjects, (iii) verbs with inter-lative subjects, (iv) verbs with dative subjects, and (v) one verb with inter-elative subject. Gender agreement operates on an ergative–absolutive basis, whereas person agreement has nominative–accusative syntax.

Mehweb person agreement is unique within Nakh-Daghestanian in that it is sensitive to the illocutionary force of the utterance. As in other Daghestanian languages with person agreement, verbal person marking is also sensitive to the syntactically introduced logophoric center, as in finite logophoric clauses with the complementizer *ile*. In such environments, personal pronouns undergo optional indexical shift, whereas person marking is obligatorily shifted to the perspective of the syntactic logophoric center.

Although traditionally Mehweb person agreement is considered to be purely subject-oriented, this chapter argues that several constructions, such as agreement in sentences with dative subject verbs and agreement in the Present Progressive, reveal the sensitivity of person agreement to the person feature of the absolutive direct object.

I also describe case marking and agreement in causative and biabsolutive constructions. Despite overall semantic and syntactic differences between the two, they demonstrate similar behavior with respect to the ergative subject of the lexical verb, which can still be diagnosed when it is absent from the phonological expression, by means of case marking on reciprocal pronouns. Finally, I identify a previously unattested construction with three absolutive arguments.

### **List of abbreviations**



### **References**


## **Chapter 8**

## **Specialized converbs in Mehweb**

### Maria Sheyanova

National Research University Higher School of Economics, Faculty of Humanities, School of Linguistics, masha.shejanova@gmail.com

This paper describes the semantic inventory and morphosyntactic properties of specialized converbs in Mehweb. The data for the description were collected during field trips to the village of Mehweb. Converbs with the following meanings are described: anteriority, immediacy, inceptivity, simultaneity, posteriority, hypothetical conditionality, counterfactuality, concessivity (and another meaning close to concessivity), causality, purpose and graduality. A participle-based form with the semantics of a locative converb is discussed.

*Keywords*: East Caucasian languages, adverbial subordination, converbs, specialized converbs, verbal morphology.

### **1 Introduction**

Specialized converbs are a subtype of converbs that specify the semantic relation between the main and dependent clauses (e.g. purposive or causal); for a fuller definition see §2.2. This paper describes the inventory and morphosyntactic properties of specialized converbs in Mehweb. The structure of the paper is as follows: §2 introduces the subject of this study and defines the terms. §3 describes specialized converbs in Mehweb. §4 summarizes the morphology of the forms in two tables. §5 is the conclusion.

### **2 Defining the terms**

### **2.1 Converb**

According to Haspelmath (1995b: 3), a converb is "a non-finite verb form whose main function is to mark adverbial subordination". In other words, one can un-

Maria Sheyanova. 2019. Specialized converbs in Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 235–253. Berlin: Language Science Press. DOI:10.5281/zenodo.3402068

#### Maria Sheyanova

derstand converbs as "verbal adverbs, just like participles are verbal adjectives". The definition of converb in Nedjalkov (1995) is similar: "as a first approximation, we can define a converb as a verb form which depends syntactically on another verb form, but is not its syntactic actant, i.e., it does not realize its semantic valencies". In an example of a canonical converb provided by Nedjalkov (1), the verb form *uleta-des* 'crossing' is dependent on the main verb *poordus* 'turned', but is not its argument.

(1) Estonian (Nedjalkov 1995) *uleta-des* cross-cvb *joge* river *poordus* turned *ratsanik* horseman *ulati* always *paremale.* right 'Crossing the river, the horseman always turned right.'

The two definitions agree that a converb: 1) is a form of a verb, and 2) marks adverbial subordination (i.e. is not a semantic argument of the main verb).

### **2.2 Specialized converbs**

Some languages distinguish specialized and general (contextual) converbs. As formulated in Haspelmath (1995a), unlike general converbs, which "leave the precise nature of the semantic link between the clauses open", specialized converbs have "a quite specific adverbial meaning", i.e. establish a specific semantic relation between the matrix and converbal clauses. The Lezgian sentences (2) and (3) taken from Haspelmath (1995a) exemplify general (contextual in Haspelmath 1995a) and specialized converbs, respectively.


'My mother Anni worked as a teacher until she got married.'

While the imperfective general converb *k'el-iz* 'reading' simply describes an additional action, the posterior specialized converb *fi-daldi* 'after going' expresses an action immediately preceding the action in the main clause.

The relations expressed by specialized converbs can be of a temporal, locative or logical nature. Converbs of logical relation normally also have a temporal 8 Specialized converbs in Mehweb

meaning. Specialized converbs are never used in clause chaining<sup>1</sup> or periphrasis, which are two other common functions of general converbs. For a discussion of general converbs in Mehweb, see Kustova (2019).

#### **2.3 Problems in defining specialized converbs**

After distinguishing between specialized and general converbs, problems remain with the definition of the category of specialized converbs. This includes distinguishing specialized converbs from other non-finite verb forms which introduce subordinate clauses with adverbial semantics. I discuss three verb forms that are problematic in this resoect, namely infinitives, participles and action nominals inflected for case.

Distinguishing an infinitive clause from converbal clauses is problematic when the infinitive has purposive semantics. Here, the infinitive formally fits the definition of a converb. The issue is discussed in Haspelmath (1995b). According to Haspelmath, prototypical infinitives have a crucial property that converbs lack: they are primarily used in complement clauses, e.g. as arguments of modal or phasal verbs.

The issue of participles and action nominals is not so easy to solve. In this paper, I consider inflected participles heading subordinate clauses, such as *wak'ibičela* in example (4), to be specialized converbs.

(4) *ʡaˤχul* guest *w-ak'-ib-i-če-la* m-come:pfv-aor-ptcp-super-el *ur-uwe* rain:ipfv-cvb.ipfv *le-r.* aux-npl 'From the moment the guest arrived, it was raining.'

However, I do not count case-marked action nominals, like *berχʷrilizela* in example (5), as specialized converbs.

(5) *šahal-li-če* town-obl-super(lat) *b-uħna* hpl-inside(lat) *b-erχʷ-ri-li-ze-la* hpl-enter:pfv-nmlz-obl-inter-el *d-iq'-es* npl-do:ipfv-inf *d-aʔ-ib* npl-begin:pfv-aor *zab.* rain 'As soon as they entered the town, it started raining.'

<sup>1</sup> In Haspelmath (1995b) clause chaining is defined as a sequence in which each converb depends on the verb that immediately follows it and which contains only one fully finite final verb.

#### Maria Sheyanova

There are two main reasons why participles and action nominals are treated differently. The first is the fact that case-inflected participles seem to be a more frequent source of specialized converbs (see e.g. Creissels 2010). Second, in adverbial subordination, case-inflected action nominals are in their prototypical syntactic position, similar to nominal adjuncts. Case-inflected participles, on the other hand, are not. In Mehweb, participles are prototypically used in adnominal position and are not inflected for case. To be used as heads of adverbial clauses, they thus need to change category, from an attribute to a nominal head. Although this is a productive syntactic process, its use to produce adverbial subordination may be considered evidence for grammaticalization. Note that there is no systematic data available on case inflection of headless participles; they appear to be infrequent in the corpus.

### **3 Specialized converbs in Mehweb**

This section provides a description of specialized converbs in Mehweb. For each converb, I provide examples showing that the form can be used both when the subject of the converb is coreferential with the subject of the main clause and when the two clauses have different subjects. Some examples also show that the converb clause can be embedded to the main clause; this is considered to provide evidence for its subordinate status.

The section focuses on temporal converbs and converbs expressing logical relations. There is one occurrence in the corpus of what seems to be a locative converb, discussed at the end of the section, but this form needs further investigation. In §4, I provide a table showing the availability of each converb for perfective and imperfective verbal stems and provide examples of the relevant word forms.

#### **3.1 Temporal converbs**

#### **3.1.1 Anterior converb**

A converbal clause with an anterior converb expresses an event that takes place before the event in the main clause, and can be translated as 'when P happened' or 'after P happened'. There are a number of variant markers of this converb – *-arʁle, -aʁle, -aʁe, -arʁ, -aʁ, -ʁale, -ʁela* – which are added to the participle. Speakers vary in the extent to which they consider each variant acceptable. Only *-aʁle*

is equally accepted by all speakers. It is possible that there are slight semantic differences between these markers, but I was unable to establish any. In general, a speaker accepts several variants, considering them to be interchangeable without any change in meaning. The form is derived from participles based on both perfective and imperfective stems. In perfective forms, a hiatus between the *-i* of the participle and the *-a* of the marker is eliminated by a more or less clearly articulated prothetic *j* (not reflected in the transcription).

(6) *iχi-šu,* this-ad(lat) *barħi* sun *b-uq-un-i-aʁle,* n-enter:pfv-aor-ptcp-ante *dursi* girl *d-ak'-ib.* f1-come:pfv-aor

'When the sun rose, a girl came to him.'


### **3.1.2 Immediate anterior converb**

The immediate anterior converb encodes an event which immediately precedes the event in the matrix clause. Its meaning is comparable to that of the English construction 'as soon as P happened'. The marker of the immediate anterior converb is the suffix *-rijal* attached to the perfective stem followed by the irrealis suffix.

(9) *doˤʜi,* snow *iχ* this *unna-li-šu* neighbour-obl-ad(lat) *w-ak'-a-rijal,* m-come:pfv-irr-imm *b-iq'-es* n-do:ipfv-inf *b-aʔ-ib.* n-begin:pfv-aor

'Just after he got to his neighbour's, it began to snow.'

#### Maria Sheyanova

(10) *sudi-če* court-super(lat) *w-aʔ-a-rijal,* m-arrive:pfv-irr-imm *šalʔu* in.the.bosom(lat) *naˤʁ--ra* hand--add *b-aˤq-i-le,* n-hit:pfv-aor-cvb *sudija-li-s* judge-obl-dat *haraq'e-r* in.front-npl(ess) *ʁarʁ-ube* stone-pl *χʷarčara* shake *d-iz-aq-i-le* npl-lv:ipfv-caus-aor-cvb *le-r.* aux-npl 'As soon as he got to the court, he put his hand under his arm and shook the stones (so they jingled like coins) in front of the judge.' (Magometov

1982: p. 147, sentence 27)

This form cannot be derived from the imperfective irrealis stem, cf. \**wik'arijal* from imperfective *-ik'es* 'come', \**urcarijal* from imperfective *urces* 'fly'.

There is another way to express the same meaning using an action nominal in the inter-elative form:

(11) *šahal-li-če* town-obl-super(lat) *b-uħna* hpl-inside(lat) *b-erχʷ-ri-li-ze-la,* hpl-enter:pfv-nmlz-obl-inter-el *d-iq'-es* npl-do:ipfv-inf *d-aʔ-ib* npl-begin:pfv-aor *zab* rain 'As soon as they entered the town, it started raining.'

#### **3.1.3 Inceptive converb**

The event encoded by the inceptive converb is the initial boundary of the event described by the main clause. It can be translated into English as 'from the moment when' or 'after'. The marker of the inceptive converb is *-čela*, which is attached to the perfective participle. The converb marker originates from a combination of the nominal suffixes *-če-la* (super-el), which literally means 'from above'.


8 Specialized converbs in Mehweb

The form is impossible with the imperfective stem, cf. *\*wik'uličela* (from imperfective *-ik'es* 'come'), *\*urculičela* (from imperfective *urces* 'fly').

#### **3.1.4 Simultaneous converb**

Simultaneity is expressed by the converb marker *-ijadal* attached to the imperfective participle or to the aorist in the perfective. The form is probably related to one of the nominal elative markers *-adal*, with an *-i* of unclear origin and a prothetic *-j*-. On the other hand, while the elative appears in the corpus in the forms of *-ad*, *-adal* and *-adala*, the attested variants of the simulataneous converb include *-ijadal* and *-ijal*. In the examples below *-ijal* may be used without any change in meaning.


When/if formed from a perfective participle, the converb is semantically different from its imperfective counterpart in that it acquires a sense of immediacy:

(16) *predloženije* sentence *b-elč'-un-ijadal,* n-read:pfv-aor-smlt *perewod* translation *b-aq'-ib.* n-do:pfv-aor 'As soon as (s)he read the sentence, (s)he translated it.'

Note however that not all speakers accepted (16).

#### **3.1.5 Posterior converb**

The posterior converb either marks the final boundary of the event in the main clause or indicates that the event in the converb clause takes place after the event in the main clause. It could be translated into English as 'before'. The form can be used with both perfective and imperfective verb stems. This converb is formed by the affix *-če* added to the perfective stem followed by a vowel, either *a*, as in (19)

#### Maria Sheyanova

or *e*, as in (17), (18) and (20). The distribution of the vowel is phonologically conditioned and is exactly the same as the distribution of the vowel of the infinitive in the same verbs (see Daniel 2019 [this volume]), which suggests derivation of the converb from the infinitive stem. The suffix of the converb can be identified with the nominal marker *-če* (super).


'They saw a lot before they entered the town.'

(20) *luk'-eče,* write:ipfv-pstr *nuša-jni* we-erg *deč'* song *b-aq'-i-ra* n-do:pfv-aor-ego 'Before writing, we sang a song.'

### **3.2 Conditional and counterfactual converbs**

In this section I present a brief description of the morphosyntactic properties of the conditional and counterfactual converbs. For more information on the semantics of the conditional forms see Dobrushina (2019) [this volume], which is the source of most examples in this section.

#### **3.2.1 Hypothetical conditional converb**

The marker of the conditional converb *-k'a* attaches to the irrealis stem. The form is used with both perfective and imperfective verb stems.

8 Specialized converbs in Mehweb


In example (23), the main clause and the converb clause share the subject:

(23) *anwar* Anwar *w-ak'-i-le* m-come:pfv-aor-cvb *w-arg-a-k'a,* m-find:pfv-irr-cond *abaj-šu* mother-ad(lat) *uˤq'-es-i.* m.go:pfv-inf-atr 'If Anwar comes, he will go to his mother.' (Dobrushina 2019)

This conditional use of 'find' in (23) is discussed in Maisak & Daniel (2018). Example (24) shows the same converb formed from the imperfective stem:

(24) *d-aq-il* npl-much-atr *kung-ane* book-pl *luč'-a-k'a,* read:ipfv-irr-cond *d-aq-il* npl-much-atr *si-k'al* what-indef *nuša-ze* we-inter(lat) *d-alh-ul.* npl-know:ipfv-ptcp 'If we read many books, we will know many things**.'**

#### **3.2.2 Counterfactual converb**

The affix *-q'alle* forms the converb of counterfactual condition (that is, the event in the main clause could have taken place if the event in the converbal clause had taken place – but this did not happen). It attaches to the perfective stem (25, 26) or to the imperfective participle (27).

(25) *ħu* you.sg(nom) *anawaje* fast.advz *w-aq'-un-q'alle,* m-do:pfv-aor-ctrf *nuša-jni* we-erg *muħammad* Muhammad *ulc-a-re.* m.catch:ipfv-irr-pst 'If you had driven fast, we would have caught Muhammad.'

#### Maria Sheyanova

(26) *nu--ra* I(nom)--add *iχ* this *w-ebk'-ib-q'alle,* m-die:pfv-aor-ctrf *d-ubk'-a-re.* f1-die:ipfv-irr-pst 'If he had died, I would have also died.' (Dobrushina 2019)

Examples (27) shows this converb formed from the imperfective stem:

(27) *iχ* this *išbari* today *ʡaˤš-w-irk-ul-q'alle* pv-m-come.back:ipfv-ptcp-ctrf *ʡaˤχ-le* good-advz *b-uʔ-a-re.* n-be-irr-pst 'If he had come today, it would have been good.' (Dobrushina 2019)

#### **3.2.3 Concessive converb**

The concessive converb is formed by the complex suffix *-k'a-ra* (-cond-add) preceded by the irrealis suffix *-a*-, as exemplified in (28–30).


#### **3.2.4 Second concessive converb**

The marker *-leʡur* conveys a meaning close to concession and causality. It is used when the event described by the converbal clause was unlikely to happen and probably undesirable; but since it nevertheless did happen, the action in the main clause takes place. The structure of the form is unclear. In regular verbs, it may be analysed as the participle followed by the suffix *-leʡur*. For the auxiliary *-le-*cl, however, the suffix seems to attach directly to the finite form (*le-*cl*-leʡur*). This form is attested in Magometov's texts but is, at best, marginal. All my examples

8 Specialized converbs in Mehweb

are elicited. Not all speakers accept this form, and even those who find it acceptable with some verbs are unable to think of examples with other verbs.


### **3.3 Other converbs of logical relations**

#### **3.3.1 Causal converb**

The causal converb describes an event which is the cause of the situation described in the main clause. The causal converb affix *-na* is attached to the general converb. The converb is formed from both perfective (cf. 34 and 35) and imperfective (cf. 36) stems.


'We were waiting because Murad was due to come.'

#### Maria Sheyanova

#### **3.3.2 Purposive converb**

The purposive converb expresses an event conceptualized as the purpose of the action described in the main clause ('in order to'). It is formed by the affix *-alis* added to the bare verb stem and can be formed with both perfective and imperfective stems. The marker is likely to originate from *-a-li-s* (‑irr-obl-dat). Indeed, cross-linguistically, the dative often expresses a purposive meaning (see for example, Haspelmath 1995a), and purposive meaning is related to the irrealis domain (see for example Palmer 2001: 131).


The semantics of purpose can also be expressed by the infinitive, as in (40) and (41). Just as the construction with *-alis*, the infinitive construction may have the same subject as in the main clause or a different one. The difference in meaning between the two constructions, if it exists, has not been studied.


However, for the reasons discussed above in §2, I do not count the infinitival construction as converbal, though, in this case, it fits the definition from a functional point of view. For further detailed discussion, see Haspelmath (1995b: 28).

8 Specialized converbs in Mehweb

#### **3.3.3 Gradual converb**

The affix *-cad(i)* attaches to the participle to express graduality. Clauses with this converb can be translated into English using the expression 'the more…, the more …'. This affix also exists in standard Dargwa as a nominal marker expressing the meaning 'as much as, about' (called "equative" in van den Berg 2001: 25). It also occurs with verb forms and in this case shows nearly the same semantics as in Mehweb. Historically, the first part of this marker (*-ca-*) may derive from the spatial marker meaning 'from the speaker' (translocative), but the origins of the second part *-di* are not clear. A possible cognate of this affix is *-cat* in Tanti Dargwa, which conveys the meaning of approximation and similarity (Sumbatova & Lander 2014). The form can be derived from both the perfective and imperfective participles.


Apart from the semantics described above, this form may also have a temporal interpretation of simultaneity, as shown in (44):

(44) *dursi* girl *šaˤ-baˤʜ* village-dir *q'-uˤwe* go:ipfv-cvb.ipfv *le-r-cad,* aux-f-grad *iχija* this.gen *pikru-me* thought-pl *le-l-le* be-f-pst *ruzi-li-če-r.* sister-obl-super-f(ess) 'While going to the village, the girl was thinking about her sister.' (lit. her thoughts were about the sister)

#### **3.4 Locative converb**

The form that can be interpreted as locative converb was only identified in the corpus after the field phase of the research. Cf. the second wordform in (45):

Maria Sheyanova

(45) *nu* you.sg(nom) *q'oˤ-j-ħe* go:ipfv-ptcp-in(lat) *w-arg-a-k'a* m-find:pfv-irr-cond *uk-iša.* m.bring:pfv-fut.ego 'If it turns out that (you are going) where I am going, I will give you a ride.'

Morphologically, it is a participle followed by the suffix of localization inside a hollow container. For the reasons discussed above in §2.3, I consider this form together with converbs. The form only occurs in the corpus two more times (e.g. *dig-uj-ħe* 'to where one wants', 'wherever one wants (to go)'), also based on the imperfective participle (*-uj* < *-ul*). Under elicitation, the same form has also been produced for the perfective stem:

(46) *nu-ni* I-erg *uʡaˤ* cheese *b-ix-ib-i-ħe-w* m-put:pfv-aor-ptcp-in-m(ess) *it* this *kajʔ-ib.* m.sit:pfv-aor 'He sat down where I had put the cheese.'

### **4 Examples of the forms**

This section contains two tables. Table 1 shows the compatibility of converbal markers with different stems and the verb forms serving as the base for the cor-


Table 1: Compatibility of converbal markers with perfective or imperfective verb stems

#### 8 Specialized converbs in Mehweb

responding converbs, with references to the examples above. Table 2 provides an example of each of the specialized converb forms.


Table 2: Examples of specialized converbs

### **5 Conclusion**

Mehweb has a relatively rich inventory of specialized converbs, with five temporal converbs (anterior, immediate anterior, inceptive, simultaneous and posterior), seven converbs expressing logical relations (hypothetical conditional, counterfactual, concessive, converb expressing another meaning close to concessive, causal, purposive and gradual) and, probably, a locative converb. The anterior converb marker shows strong phonological variation. Other variants of converb markers include two variants for the simultaneous converb marker (*-jadal*/*-jal*) and two variants for the gradual converb (*-cad*/*-cadi*).

Specialized converbs are formed in several different ways, with the converb marker attached to either:


These types may be interpreted as reflecting an increasing degree of grammaticalization of the forms. The irrealis and the infinitive stems are bound (cannot appear without further marking). The third pattern includes converb suffixes that are selective in terms of the stem they attach to (aorist in the perfective, participle in the imperfective). The fourth type includes converbs derived from stems that also function as free forms (participles). However, as discussed in the introduction, participles do not typically function as nominal adjuncts in the clause, and their inflection is different from the converb markers, so a certain degree of grammaticalization is still present. Finally, in the fifth type, the forms to which the converb markers (causal and second concessive) are attached are not only free forms but are already marked for adverbial subordination. Their only function is to further specify the general converb in terms of its relation to the main clause, which makes these markers functionally similar to particles. On the other hand, they are bound and must be considered as derivational suffixes.

### **List of abbreviations**


#### Maria Sheyanova


### **References**


## **Chapter 9**

## **General converbs in Mehweb**

### Marina Kustova

National Research University Higher School of Economics

This paper deals with the morphological and syntactic properties of general converbs in Mehweb. I discuss the markers used to form general converbs, periphrastic converbs, independent uses of converbs, their behaviour in combination with verbs in the imperative, different strategies of argument sharing between the converb clause and the main clause, and coordination/subordination properties of the general converb. The description of the syntactic properties of the converbs is based on both elicited examples and corpus evidence.

### **1 Introduction**

Converbs are determined as nonfinite verb forms whose main function is to mark adverbial subordination (Haspelmath 1995: 3). Mehweb specialized converbs, i.e. converbs which specify the semantic relation between the main and the converb clause (e.g. causal, immediate precedence in time, other temporal relations and so on), are discussed in Sheyanova (2019) [this volume]. This paper is devoted to general converbs which do not specify this relation – or, at least, do it in a more subtle way, leaving some room for contextual interpretation (hence an alternative label for this category is *contextual* converbs).

In §2, the basic uses and morphology of perfective and imperfective converbs will be discussed, §3 describes periphrastic converbs, and §4 deals with independent uses of general converbs in Mehweb. §5 discusses different patterns of argument sharing between converb clauses and main clauses. Finally, in §6 I discuss the coordination and subordination properties of the Mehweb general converb.

Marina Kustova. 2019. General converbs in Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 255–270. Berlin: Language Science Press. DOI:10.5281/zenodo.3402070

#### Marina Kustova

### **2 Perfective and imperfective converbs: background information**

General converbs in Mehweb are derived from perfective and imperfective stems. Below I will refer to them as perfective and imperfective converbs, respectively. The perfective converb is formed by adding the converb marker *-le* to the verb in the aorist (Magometov 1982: 110); the affix undergoes a number of morphophonological alternations (see Moroz 2019; Daniel 2019). The formation of perfective converbs is presented in Table 1.


Table 1: The formation of the perfective converb

The imperfective converb is formed by adding *-uwe* to the imperfective stem. Here, the process is the same for all conjugation classes and could be interpreted as a combination of the participle suffix *-ul* and the converb suffix *-le* (Magometov 1982: 112). The formation of imperfective converbs is presented in Table 2.


The perfective converb is used to describe an event preceding the situation denoted in the main clause. Situations that take place simultaneously with the main event are described by the imperfective converb. Both imperfective and perfective converbs can be combined with finite verbs with present or past time reference, cf.:

9 General converbs in Mehweb


'Having sung a song, Musa is going home.'

(4) *deč'--ra* song--add *b-iq'-uwe* n-do:ipfv-cvb.ipfv *musa* Musa *ʡaˤr-q'-uwe* away-go:ipfv-cvb.ipfv *le-w* aux-m *quli.* house(lat)

'Singing a song, Musa is going home.'

In sentence (1), a perfective converb is combined with a finite verb in the aorist, in (2) an imperfective converb is combined with a verb in the aorist, in (3) a perfective converb is combined with a verb in the present tense, and in (4) an imperfective converb is combined with a verb in the present tense.

### **3 Periphrastic converbs**

Apart from the perfective and imperfective converbs described above, most speakers of Mehweb allow forms consisting of a converb and a copula in the converb form. Essentially, these are converbs formed from periphrastic verb forms. Below I refer to such forms as periphrastic converbs.

A periphrastic converb consisting of a perfective converb and a copula in the converb form corresponds to the resultative, a finite periphrastic form consisting of a perfective converb and a tensed copula.

(5) *jaˤbu* horse *b-ic-i-le* n-sell:pfv-aor-cvb *le-b-le* aux-n-cvb *maˤʜmud-ini* Mahmud-erg *χʷe* dog *as-ib.* take:pfv-aor 'Having sold a horse, Mahmud bought a dog.'

#### Marina Kustova

The same construction with an imperfective converb corresponds to the present progressive, which Magometov (1982: 87) terms definite imperfect.

(6) *jaˤbu* horse *b-ilc-uwe* n-sell:ipfv-cvb.ipfv *le-b-le* aux-n-cvb *maˤʜmud* Mahmud *le-w* aux-m *w-is-uwe.* m-weep:ipfv-cvb.ipfv 'While selling a horse, Mahmud is crying.'

Speakers also allow sentences like (7) and (8), where the copula in the converb form is preceded by a perfective or an imperfective infinitive. Morphologically, these forms correspond to the future resultative (composed of a perfective converb and a copula in the converb form) and the future progressive (an imperfective converb and a copula in the converb form). The semantic difference between the two periphrastic converbs remains unclear.


### **4 Independent use**

In most cases, converbs are used in complex clauses that also contain main finite clauses. However, some speakers allow sentences that contain only converbal predication.

When used independently, the perfective converb can have resultative semantics, as in (9).

(9) *urši-ni* boy-erg *diʔ* meat *b-erk-uwe.* n-eat:pfv-aor.cvb 'A boy has eaten the meat (he finished it, so there is none left for me).'

Imperfective converbs can have the same semantics as habitual forms, i.e. sentences (10) and (11) have the same meaning.

9 General converbs in Mehweb


Although examples with a converb as a sole predicate are allowed by some speakers, the corpus (about 900 sentences) does not contain any instances of such sentences.

### **5 Argument sharing**

In Mehweb, the S, A, P or other argument of the converb clause may – but need not – be referentially identical to an argument of the main clause. This shared argument can be expressed in either of the two clauses. Below I will refer to such situations as *argument sharing.* In this part I discuss sharing of core arguments, including S, A and P. Logically, a large list of different argument sharing configurations could be derived by alternating syntactic parameters, including the role of the shared argument in the main clause, the role of the shared argument in the converb clause and the locus of its expression (main or converb clause). However, not all of them are grammatical. Below I classify different argument sharing strategies in accordance with the consultants' ability to interpret them. Note that some of the sentences may be grammatical when interpreted in a different way, so I checked not just grammaticality but also the availability of the intended interpretation with shared arguments.

Generally, all configurations which include sharing of two S-arguments or an S-argument and an A-argument, regardless of the clause where it is expressed (the main or the converb clause), are interpretable, cf. (12) and (13).

(12) *musa* Musa *w-ak'-i-le* m-come:pfv-aor-cvb *rasuj-če* Rasul-super(lat) *b-aˤq-ib.* n-hit:pfv-aor

'When Musa came, (he) hit Rasul.' <sup>1</sup> (The S-argument of the intransitive converb clause is coreferential with the A-argument of the transitive main clause and is expressed in the converb clause.)

<sup>1</sup>The verb *--aqˤas* 'hit' takes the instrument as S, though it does not have to be expressed in the sentence. This is why the noun *Rasul* is not marked as S and the verb has a neutral gender agreement marker.

#### Marina Kustova

(13) *dag* yesterday *χʷe* dog *har-b-uq-uwe* pv-n-flee:pfv-aor.cvb *išbari* today *ʡaˤš-b-aˤq-ib.* back-n-come:pfv-aor 'Yesterday the dog ran away, today (it) returned.' (Two intransitive clauses sharing their S-argument, which is expressed in the converb clause)

In example (12), the fact that the shared argument is expressed in the converb clause is obvious from case marking. The verb *--ak'es* 'come' is intransitive and takes an S-argument, while *--aˤqas* 'hit' is transitive, with its A-argument in the ergative. Since the shared argument takes S-marking (nominative), it is dependent of the converb, not of the main verb. Therefore, it belongs to the converb clause.

As for (13), the same fact can be established on the basis of word order. The word *dag* 'yesterday' belongs to the converb clause, and the shared argument stands between this adverb and the converb. Therefore, I conclude that the shared argument belongs to the converb clause.

Sentences that include no argument sharing at all, like (14) and (15), are perfectly grammatical as well.


'The husband mowed the hay, the wife made the bed.'

Sharing that involves P-argument, like (16) and (17), is less straightforward. In (16), both clauses are transitive, the P-argument of the converb clause is coreferential with the A-argument of the main clause and the shared argument is expressed in the main clause (which can again be seen from the case marking of the shared argument):

(16) *maˤʜmud-ini* Mahmud-erg *as-i-le* take:pfv-aor-cvb *gatu-ini* cat-erg *waca* mouse *b-uc-ib.* n-catch:pfv-aor 'Mahmud bought a cat and it caught a mouse.'

9 General converbs in Mehweb

Note that in (16) the P-argument of the main clause cannot be coreferential with the P-argument of the converb clause, i.e. the example cannot mean that Mahmud bought a mouse who was then caught by a cat.

In (17), both clauses are transitive and share both their A- and P-arguments. The shared A-argument is expressed in the converb clause, and the shared Pargument belongs to the main clause (evidence based on word order, as in (13)):

(17) *dag* yesterday *ħamzat-ini* Hamzat-erg *as-i-le* take:pfv-aor-cvb *išbari* today *kʷiha* lamb *b-erh-un.* n-slaughter:pfv-aor 'Yesterday Hamzat bought a lamb, today he slaughtered it.'

Sentences where A- and P-arguments of one transitive clause were intended to be criss-cross coreferential with the P- and A-arguments of the other transitive clause were not interpreted in this way by any of the speakers. Cf. (18):

(18) *rasul* Rasul *uc-i-le* m.catch:pfv-aor-cvb *musa* Musa *w-aˤbʡ-ib.* m-kill:pfv-aor Intended \*'Musa caught Rasul, Rasul killed Musa.' Possible interpretation: 'Rasul was caught, Musa was killed.'

Table 3 below shows the distribution of different argument sharing strategies according to the native speakers' ability to interpret them in the intended way.


Table 3: The acceptability of different core argument sharing strategies

Note that not all theoretically possible configurations are included in the resulting table. It appears that configurations where the X-argument of the converb clause is coreferential with the Y-argument of the main clause behave in exactly the same way as those where the X-argument of the main clause is coreferential

Marina Kustova

with the Y-argument of the converb clause. The locus of expression did not seem to matter, either. The configurations in the table are thus only represented by the arguments which are shared.

### **6 Coordination and subordination properties**

An English translation equivalent for a converb construction is often coordination (Haspelmath 1995: 8). The syntactic status of this parallel is treated in Kibrik (2007). Below I will explore the syntactic properties of the Mehweb converb construction in terms of coordination vs. subordination.

### **6.1 Three syntactic tests**

To find out whether the converbal construction in Mehweb is subordinate to the main verb or not, three syntactic tests were applied, including changing the linear order (§6.1.1), embedding the converb clause in the main clause (§6.1.2), and relativization (§6.1.3) (the tests are described in Creissels 2012: 143–145).

To run the tests, I will use sentences (19) and (20). In sentence (19), the converb clause shares its A argument with the main clause, while sentence (20) has no argument sharing.


#### **6.1.1 Linear order of the clauses**

When two or more coordinate clauses describe a sequence of events, their order is iconic and cannot be changed without changing the sense of the entire sentence. In contrast, if one of the clauses is subordinate, the order can be changed with no influence on the general meaning. For instance, *I came, I saw, I conquered* is not semantically identical to *I came, I conquered, I saw*. However, the sentences *Having seen it, I conquered it* and *I conquered it, having seen it* are both possible

9 General converbs in Mehweb

and described the same sequence of events. In this respect, Mehweb general converbs seem to behave more like English subordinate clauses:


As can be seen from comparison of these examples with (19) and (20), in both cases the main and the converb clause can change places. It does not affect the interpretation of the order of the events. However, note that the translations provided by native speakers for both modified sentences changed so that their English translations now include the word 'because'. This fact will be discussed further in the paper.

#### **6.1.2 Embedding**

Further evidence for the subordination analysis is the possibility of embedding the converb clause in the main one.

In Mehweb, it is perfectly fine to place a converb clause that shares its Aargument with the main clause between the main verb and its dependents, cf. (23):

(23) *musa* Musa *qali* house *b-ic-i-le* n-sell:pfv-aor-cvb *iz-es* be.ill:ipfv-inf *w-aʔ-ib.* m-begin:pfv-aor 'Musa, as he sold the house, became ill.'

In this sentence, it is clear that the shared argument belongs to the main clause because of its case marking. The verb *izes --aʔes* 'become ill' is intransitive, which is why its only argument stands in the nominative. If the noun belonged to the converb clause, it would appear the ergative, cf. (24):

(24) *musa-ini* Musa-erg *qali* house *b-ic-ib.* n-sell:pfv-aor 'Musa sold the house.'

#### Marina Kustova

In the absence of argument sharing, however, embedding is severely degraded: speakers tend to either assign another interpretation or judge the sentence as unacceptable:

(25) *xunuj-ni,* wife.obl-erg *adami-li-ni* husband-obl-erg *q'ar* hay *b-iˤšq-i-le,* n-mow:pfv-aor-cvb *buruš* bed *b-aq'-ib.* n-make:pfv-aor 'The wife and the husband, having mowed the hay, made the bed.'

In (25), the converb clause with no argument sharing is embedded to the main clause. When the ergative arguments of the different clauses are placed next to each other as in (25), they are interpreted as belonging to one and the same clause (which can be either the converb clause or the main clause). As a result,

interpretation of the sentence becomes problematic.

#### **6.1.3 Relativization**

Generally, clause coordination tends to place more severe restrictions on the use of relativization strategies than clause subordination. For instance, the English sentence *The girl ran away when the boy punched her* can be relativized as *The girl who ran away when the boy punched her came back*, whereas no such construction is possible with a sentence like *The boy punched the girl, and she ran away* (*\*The boy, who punched the girl, and she ran away, felt sorry*). Thus, where the relative construction is allowed, I will consider this an argument for the subordinate status of the converb. Unavailability of relativization will be considered as evidence in favor of coordination.

In Mehweb, relativization is allowed if the converb clause shares its S- or Aargument with the main clause:

(26) *qali* house *b-ic-i-le* n-sell:pfv-aor-cvb *iz-es* be.ill:ipfv-inf *w-aʔ-ib-i* m-begin:pfv-aor-atr *musa* Musa *w-ebk'-ib.* m-die:pfv-aor

'Musa, who became ill because of selling the house, died.'

In (27), where no argument is shared, none of the speakers suggested the expected interpretation ('The wife, who made the bed after her husband had mowed the hay, came here'). They all suggested the paratactic reading, with the participle interpreted as the predicate of an independent main clause:

9 General converbs in Mehweb

(27) *adami-li-ni* husband-obl-erg *q'ar* hay *b-iˤšq-i-le* n-mow:pfv-aor-cvb *buruš* bed *b-aq'-ib-i,* n-make:pfv-aor-atr *xunul* wife *iše* herelat *r-ak'-ib.* f-come:pfv-aor 'The husband mowed the hay and made the bed (for his wife), the wife came here.' \* 'The wife, who made the bed after her husband mowed the hay, came here'

I conclude that, with respect to relativization, sentences with no argument sharing display more coordinate properties, while those with argument sharing tend to behave more like subordinate clauses. With respect to clause order, the constructions behave similarly, irrespective of the presence or absence of a shared argument: they both allow main clause – converb clause order, but the speakers then specify the causal relation between the two events.

#### **6.2 Semantic properties of the converb clause**

If two or more clauses are coordinated, each of them has a range of properties of their own, which means that features like tense, aspect and mood (and some others) are assigned to each predicate independently. A subordinate clause can, however, inherit some features from a main clause – or, in other words, fall under their scope. In this section, I will explore some of the converb clause properties which can potentially be inherited from the main clause. For each of the (non-)shared features, I will suppose that inheriting a feature implies that the construction behaves more like a subrodinate clause, and the absence of such inheritance will make an argument for the coordination analysis.

#### **6.2.1 Tense and taxis**

As was mentioned in §2, the perfective converb describes an event preceding the situation denoted in the main clause, whereas the imperfective converb describes an event which takes place simultaneously with the main event. In other words, the converb clause usually does not have a tense of its own, and its time reference fully depends on that of the main clause.

Sentences which imply the presence of independent time reference within the converb clause may nevertheless be accepted as fully grammatical, cf. (28):

#### Marina Kustova

(28) *išbari* today *duči-rk'-uwe* laugh-lv:ipfv-cvb.ipfv *dag* yesterday *pat'imat* Patimat *pašmaje* sad.advz *le-l-le.* be-f-cvb 'Today Patimat is smiling, yesterday she was sad.' ('Today smiling, yesterday Patimat was sad.')

Note that, however, such sentences are judged as ungrammatical if the converb clause is embedded to the main one, cf. (29):

(29) *\*dag* yesterday *pat'imat* Patimat *išbari* today *duči-rk'-uwe* laugh-lv:ipfv-cvb.ipfv *pašmaje* sad.advz *le-l-le.* be-f-cvb 'Today Patimat is smiling, yesterday she was sad.'

The same happens if the converb clause is placed after the main one: sentence (30) is ungrammatical as well.

(30) *\*dag* yesterday *pat'imat* Patimat *pašmaje* sad.advz *le-l-le* be-f-cvb *išbari* today *duči-rk'-uwe.* laugh-lv:ipfv-cvb.ipfv 'Today Patimat is smiling, yesterday she was sad.'

Overall, it seems that the Mehweb converb is capable of having a tense of its own, i.e. be tensed independently of the main clause. However, converbs inflected for a different tense than the main verb cannot be embedded to the main clause or placed after it. In other words, they fail the test on subordination. In this case, the converb clause is less clearly subordinate to the main clause.

#### **6.2.2 Illocutionary force**

When a subordinate predication depends on an imperative, it may or may not inherit the illocutionary force of the main clause. This means that the situation described in the subordinate predication can either be a part of the situation that the speaker wants to happen, or not. For instance, the English sentence *Having drunk wine, don't drive* does not mean that the speaker wants the addressee to drink the wine and then not to drive. This means that *Having drunk wine* does not inherit the illocutionary force of the main predication. On the contrary, the sentence *Having cut the tomatoes, add them to the salad*, which can easily be a part of a bigger instruction, does imply that the speaker wants the addressee both to cut the tomatoes and to add them to the salad. In this case, the subordinate clause inherits the main clause's illocutionary force.

In Mehweb, a converb depending on an imperative form may or may not inherit the illocutionary force of the main clause.

9 General converbs in Mehweb


In the contexts where the converb falls under the scope of the main verb's illocutionary force, using another imperative instead of the converb is possible. Thus, sentence (33) has almost the same reading as sentence (32).

(33) *kaltuška* potato *d-išq-aˤ* npl-peel:pfv-imp *ħarši* soup *d-aq'-a.* npl-cook:pfv-imp 'Peel the potatoes and cook the soup.'

The meaning of the two, however, is slightly different. Some speakers claim that (32) implies that potatoes should be peeled and then added to the soup, whereas (33) does not have this implication. Probably, using converbs with imperatives implies that there is a closer semantic link between the two events than there would be in a sentence with two imperatives. A similar phenomenon is described in Dobrushina (2008) for Archi.

#### **6.3 Coordination vs. subordination**

According to Creissels (2010), if it is difficult to determine whether a construction is a a case of coordination or subordination, there are a number of analytical possibilities. In particular, if one and the same construction within the same sentence can show both coordinate and subordinate properties, this would represent an instance of what he calls co-subordination. If a construction shows either coordinate or subordinate properties depending on the context, this is analysed as coordination in some of its uses and subordination in others.

After applying the tests to different sentences containing converbal predication, it seems that Mehweb converbal construction displays different coordination/subordination properties under different circumstances. I will take a closer look at the conditions that influence the syntactic properties of the constructions.

First, as can be seen from examples (21–23) and (26), in all the cases where the subordination tests worked, some sort of causal relation between the main clause

#### Marina Kustova

and the converb clause is implied. I suggest that the coordinate or subordinate characteristics of the construction mostly depend on the semantic relationship between the main clause and the converb clause. In other words, when a semantic link between the two appears, the converb construction is very likely to become subordinate.

Another important factor seems to be the presence or absence of argument sharing between the main and the converb clause. Examples (25) and (27) show that if the embedding test and the relativization test are applied to sentences with no argument sharing, the results may include the re-interpretation of the intended syntactic structure and lead to a different semantic interpretation. Relativisation and embedding of converb clauses without argument sharing is ungrammatical.

All in all, it seems that the behavior of the converb construction depends on (a) the semantic relation between the main and the converb clause and (b) the presence or absence of argument sharing between the clauses.

This seems very similar to the situation in Tsakhur as described by Kazenin & Testelets (2004). In this paper, the authors applied several tests for coordination vs. subordination to sentences containing general converbs. The tests turned out to give different results for one and the same sentence, depending on whether there was a causal relation between the converb clause and the main clause. If a Tsakhur sentence contains a converb construction and its semantics may imply some causal relation between the main clause and the converb clause, then embedding the converb clause into the main one is only possible with a causal interpretation. To put it differently, subordination tests produce positive results only if there exists a causal relation between the main clause and the converb clause. However, center embedding can also work without a causal relation between the clauses, if they both have the same subject.

### **7 Conclusion**

In this paper I have considered the properties of general converbs in Mehweb Dargwa. I have described the converb marker and its morphophonological features, the distribution of perfective and imperfective converbs, the use of periphrastic converbs, the independent use of converbs, the way they can combine with imperatives, and how they may share their S-, A- or P-arguments with the main clause. Coordination and subordination properties of the Mehweb general converb were discussed. The syntactic status of converb clauses is either coordinate or subordinate, depending on (a) whether there is a causal relation

between the main clause and the converb clause, and (b) whether the converb clause shares its main argument with the main clause or not. Which of the principles (a) and (b) is prior, however, is still a question to be discussed.

### **List of abbreviations**


### **References**

Creissels, Denis. 2010. Specialized converbs and adverbial subordination in Axaxdərə Akhvakh. In Isabelle Bril (ed.), *Clause linking and clause hierarchy: Syn-* *tax and pragmatics* (Studies in Language Companion Series 121), 105–142. Amsterdam Philadelphia: John Benjamins.


## **Chapter 10**

## **The self-pronoun in Mehweb**

### Aleksandra Kozhukhar

National Research University Higher School of Economics

This study deals with the phenomenon of the pronominal multifunctionality in Mehweb. The pronominal stem glossed as 'self' has four functions (reflexive, logophoric, intensifier, and resumptive) which are described in some detail.

*Keywords*: logophoricity, reflexivization, long-distance reflexives

### **1 Introduction**

In many typologically distinct languages, a formal relationship between reflexive pronouns, logophoric pronouns and intensifiers is attested (see König et al. 2013). In Mehweb these functions are fulfilled by the pronominal stem *sa*‹cl›*i*, 'self', which can also be used as a resumptive.

In this paper I will describe the formal and functional aspects of the pronoun *sa*‹cl›*i*, starting with a description of the structure of the relevant forms in §2. In §3 I will discuss their reflexive and logophoric usages, followed by a description of free logophors in §4. §5 is dedicated to the usage of *sa*‹cl›*i* as an intensifier and in §6 some examples of the resumptive function will be discussed.

### **2 Morphology**

The pronoun *sa*‹cl›*i* can appear in the form of what I refer to as a "bare pronoun", consisting of a pronominal stem inflected for number and case. A "complex pronoun" can be formed by adding the suffix *-al* to the bare pronoun. Both forms are described below.

Aleksandra Kozhukhar. 2019. The self-pronoun in Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 271–293. Berlin: Language Science Press. DOI:10.5281/zenodo.3402072

#### Aleksandra Kozhukhar

#### **2.1 Bare stem**

Mehweb employs the bare pronoun to refer to the antecedents in the longdistance domain (see §3.3) and possessive domain (see §3.2). The pronoun *sa*‹cl›*i* agrees in number, person and gender with the antecedent and can attach case suffixes (see Table 1).

Table 1: The paradigm of the bare pronoun


The bare pronoun has three suppletive allomorphs. The first, *sa*‹cl›*i*, is the nominative stem, which is the same in the singular and in the plural and carries a gender marker infix, agreeing with the antecedent of the pronoun. The second, *sune-*, is the oblique stem of the third person singular and can attach case suffixes. The third, *ču-*, is the oblique stem of the third person plural and can attach case suffixes.

#### **2.2 Complex pronouns**

The stem *sa*‹cl›*i* may attach the particle *-al*. The particle functions as emphatic when attached to nominal stems and demonstratives<sup>2</sup> :


<sup>1</sup> In Table 1 the genders are given as abbreviations as follows: m – masculine, f – feminine, n – neutral (i.e. all inanimate and animate non-human entities), hpl – human plural entities, npl – non-human plural entities.

<sup>2</sup> Suffix *-al* also marks cardinal numerals (Magometov 1982: 58).

#### 10 The self-pronoun in Mehweb

A partial paradigm of the complex pronoun is given in Table 2. For the sake of comparison, inflected forms of the first and second person pronouns are also presented.


Table 2: The paradigm of the complex pronoun

The suffix *-al* is preceded by an epenthetic consonant or deletion of the vowel in the suffix. If the last vowel of the stem is *-u*-, the epenthetic consonant is *-w*- (e.g. *nuwal*). If the last vowel of the stem is *-i*-, the epenthetic consonant is *-j*- (e.g. *sawijal*). If *-al* follows *-e-* or *-a*- then the vowel in the suffix is dropped (e.g. *ħušal* and *sunezel*). In the dative case, *-al* is simply attached to the case suffix. The distribution of these forms is discussed in the following sections.

### **3 Logophoric and reflexive contexts**

In this section, I will discuss the reflexive and logophoric functions of the pronominal stem.

Reflexives are typically used to show the coreference of the non-subject argument of the clause to another clause-mate argument (König et al. 2013). Testelets & Toldova (1998) argue that reflexives may be bound by a higher syntactic priority position (i.e. subject) which occurs in the same sentence. Logophoric pronouns are used to indicate "coreferenciality or conjoint reference with the argument of a higher predicate of communication or mental experience" (Sells 1987).

#### Aleksandra Kozhukhar

#### **3.1 Local domain**

The reflexive is bound within the local domain if it occurs within the same clause as its antecedent. Mehweb demonstrates no constraints on the syntactic position a reflexive can take in the clause. It can occupy the position of P as in (3) and (6), the indirect object position as in (4), or it can fulfill the role of adjunct (5). The antecedent, however, has to be the subject (cf. infelicitous (7)). This means it requires ergative marking with a transitive predicate, nominative for intransitive, and dative, inter-lative or inter-elative for experiential predicates (cf. examples (3), (4) and (6)). Within the local domain, the form of the pronoun is constrained: a bare pronoun with an antecedent in the local domain is considered ungrammatical and can only be interpreted as having logophoric meaning (compare (3) and (8)).

	- b. *\*sune-ze-l* self.obl-inter(lat) *rasul* Rasul.obl-erg *gu-b.* see:pfv-aor 'Rasul saw himself.' (lit. 'Himself<sup>i</sup> saw Rasul<sup>i</sup> .')

10 The self-pronoun in Mehweb

Because Mehweb is a pro-drop language, the reflexive can get a zero-antecedent, which is obligatorily in the subject position, as in (9).

(9) a. *it-ini* that-erg *sune-s-al* self.obl-dat-emph *jaˤbu* horse *as-ib.* take:pfv-aor b. *sune-s-al* self.obl-dat-emph *jaˤbu* horse *as-ib.* take:pfv-aor

> '(He<sup>i</sup> ) bought himself<sup>i</sup> a horse.'

The reflexive pronoun can be bound by a quantified NP.

(10) *har-il* each-atr *urši-li-ni* boy-obl-erg *sune-s-al* self.obl-dat-emph *jaˤbu* horse *as-ib.* take:pfv-aor 'Each boy<sup>i</sup> bought himself<sup>i</sup> a horse.'

Subordinate clauses work the same way. In a subordinate clause, the bare pronoun cannot be bound within the subordinate clause (11), while the complex pronoun has to be bound within it (12).


In example (12) the antecedent of the reflexive is within the local domain, whereas in (11) it is located in the distant domain (the latter will be discussed further in §3.3). The two domains differ as to which pronoun is used: the local domain employs the complex pronoun, whereas for an antecedent in the distant domain the bare pronoun is used.

#### **3.2 Possessive domain**

The possessive domain contains contexts where a genitive reflexive occurs in an NP within the same clause as its antecedent. In Mehweb, this domain is distin-

#### Aleksandra Kozhukhar

guished from the local domain in that both bare pronouns and complex pronouns can be employed<sup>3</sup> , as in (13).

	- b. *sune-la-l* self.obl-gen-emph *quli-w* house-m(ess) *ħa-jz-ur.* neg-live-aor '(He<sup>i</sup> ) did not live in his<sup>i</sup> house.'

Consider also the following examples where the complex and the bare pronoun are used in similar contexts by the same speaker:


Examples (14) to (17) prove that in natural texts the bare pronoun is available in possessive contexts. Consultants provide contradictory grammaticality judgements of constructed stimuli with the reflexive genitive. The majority consider

<sup>3</sup>This fact may serve as evidence for the idea that the possessive domain is a transition point between the local domain and the distant domain.

10 The self-pronoun in Mehweb

(13a) and (13b) to have the same meaning and to be fully grammatical. Some consultants suggest that *sunelal* adds emphatic meaning ('his own'), whereas *sunela* simply indicates possession. Other consultants suggest that the bare pronoun *sunela* is not bound within the sentence (for further discussion see §4), i.e. (13a) can be translated as 'He is living in his (someone else's) house'. Finally, some consultants consider *sunela* to be ungrammatical, apparently extending the constraints on the occurrence of bare pronouns in the same clause as their antecedents to possessive NPs.

#### **3.3 Distant domain**

Distant domain contexts are sentences in which the pronoun and its antecedent occur in different clauses. In Mehweb, the order of the antecedent and the pronoun is relevant within the local domain. The pronoun cannot precede its antecedent, otherwise it gets the free logophoric reading (more on free logophors in §4). The distant domain requires using the bare pronoun (see (18)).

(18) *sune-s* self.obl-dat *dig-uwe* want:ipfv-cvb.ipfv *le-w* aux-m *adaj-ze* father-inter(lat) *rasul* Rasul *daˤʜmic'aj-ħe-w* mirror-in-m(ess) *gʷ-es.* see:pfv-inf 'Rasul<sup>i</sup> wants his father<sup>y</sup> to see him<sup>i</sup> in the mirror.' lit. 'Himself<sup>i</sup> wants his father<sup>y</sup> to see Rasul<sup>i</sup> in the mirror.'

The bare stem can take subject and non-subject positions (P, IO, adjunct) in the subordinate or main clause and can be used in both finite and non-finite subordinate clauses, as shown in the following section.

#### **3.3.1 Finite subordinate clauses**

Mehweb employs finite subordinate clauses with predicates of speech and thought. Finite subordinate clauses in Mehweb may or may not be followed by the converb *ile* 'having said' and utilize either personal pronouns or a bare pronoun.

(19) *adaj-ni* father-erg *ib* say:pfv.aor *sune-ze* self.obl-inter(lat) *žanawar* wolf *gu-b* see:pfv-aor *(ile).* say:pfv.cvb 'Father<sup>i</sup> said he<sup>i</sup> saw a wolf.'

#### Aleksandra Kozhukhar

(20) *adaj-ni* father-erg *ib* say:pfv.aor *sune-ze* self.obl-inter(lat) *žanawar* wolf *gu-b-ra* see:pfv-aor-ego *(ile).* say:pfv.cvb 'Father<sup>i</sup> said he<sup>i</sup> saw a wolf.'

Considering Chechen and Ingush, Nichols (2000) refers to contexts such as (20) as semi-direct speech. In semi-direct speech "quoted matter is identical to the reported speech act except that coreferents to the speaker are reflexivized and the clause is marked with a quotative particle" (Nichols 2000). According to Nichols, Chechen uses reflexives to refer to the speaker, i.e. the subject of the main clause, only if subordinate finite clauses are marked by the quotation clitic *eanna*, while direct speech contexts use personal pronouns (1sg pronouns) and do not use the clitic.

In Mehweb, the quotative converb *ile* is optional with both types of reference. Compare the pronouns in (19) and (20) to those in (21) and (22); in all of these cases, the use of *ile* is optional.


Table 3 provides a summary of options for a pronoun used in a subordinate finite clause. It shows that *dize* behaves as a personal pronoun, since it can change

Table 3: Summary on the stimuli and antecedents


10 The self-pronoun in Mehweb

its antecedent between the actual and the reported speaker. The pronoun *suneze* on the other hand, behaves as a logophoric pronoun and always refers to the subject of the main clause; cf. the following examples:


Examples (26) and (27) additionally show subordinate clauses headed by different matrix predicates.


#### **3.3.2 Non-finite subordinate clauses**

Non-finite subordinate clauses in Mehweb can employ converbs, nominalizations or infinitives, depending on the predicate of the matrix clause. Non-finite

#### Aleksandra Kozhukhar

clauses can occur with a bare pronoun or with a zero pronoun in the subject position. Grammaticality of first person personal pronouns referring to the subject of the main clause in non-finite subordinate clauses is a matter of variation among the consultants (cf. 28 and 31). In non-finite subordinate clauses, the self-pronoun can occupy subject and non-subject positions (cf. 32).

Examples (28) and (29) demonstrate the use of the self-pronoun in subject and non-subject position in a subordinate clause headed by an infinitive.

(28) *it* this *uruχ* be.afraid *k'-uwe* lv:ipfv-cvb.ipfv *le-w* aux-m *sa‹w›i* ‹m›self ( *?nu*) ( ? I) *ʁaˤm-le* wrong-advz *w-ik-es* m-become:pfv-inf *(ile).* say:pfv.cvb 'He is afraid of making a mistake.'

(29) *rasuj-s* Rasul.obl-dat *dig-uwe* want:ipfv-cvb.ipfv *le-b* aux-n *adaj* father *sune-če-l* self.obl-super(lat)-emph *ħule* look *w-iz-es.* m-lv:pfv-inf 'Rasul<sup>i</sup> wants his father<sup>y</sup> to look at himself<sup>y</sup> .'

Subordinate clauses with an infinitive in Mehweb are employed as a strategy for marking sentential arguments, and can also express an aim (see (30–32)). In (31), the personal pronoun *nu* 'I' is grammatical.

(30) *ʡali-ni* Ali-erg *g-ib* give:pfv-aor *rasuj-ze* Rasul.obl-inter(lat) *arc* money *il* that *armi-li-ze* army-obl-inter(lat) *uˤq'-es.* m.go:pfv-inf 'Ali bribed Rasul so that he (Rasul or another person) go to the army.' (lit. 'Ali gave money to Rasul in order that Rasul (or another person) went to


'Ali bribed Rasul to go the army.' (lit. 'Ali gave money to Rasul in order Ali went to the army.')

10 The self-pronoun in Mehweb

(32) *ʡali-ni* Ali-erg *g-ib* give:pfv-aor *rasuj-ze* Rasul.obl-inter(lat) *arc* money *sa‹w›i* ‹m›self *armi-li-ze* army-obl-inter(lat) *uˤq'-es.* m.go:pfv-inf 'Ali bribed Rasul to go the army.' lit. 'Ali gave money to Rasul in order Ali went to the army.'

Examples (33) and (34) demonstrate the self-pronoun in a subordinate clause headed by a specialized converb.


Examples (35–37) show the use of the bare pronoun in a subordinate clause headed by an action nominal (masdar). In Mehweb there are two suffixes available for the derivation of action nominals: *-ri* and *-deš*. In most cases, these suffixes are interchangeable.


#### Aleksandra Kozhukhar

(37) *it-ini* that-erg *pikri* thought *b-aq-ib* n-do:pfv-aor *sa‹w›i* ‹m›self *q'am* late *uh-ub-le* become:pfv-aor-cvb *le-w-deš* aux-m-nmlz *(ile).* say:pfv.cvb 'He<sup>i</sup> thought that he<sup>i</sup> was late.'

The purpose of the examples above is to show that bare pronouns can be used in non-finite subordinate clauses. This fact blurs the distinction between the two functions the bare pronoun fulfills – that of the long-distant reflexive and the logophoric pronoun.

#### **3.3.3 Subject orientedness of the self-pronoun**

In a finite subordinate clause, the bare pronoun occurring in subject position is subject oriented. This means it is co-referent to the subject of the main clause, as in (25). Non-finite subordinate clauses on the other hand, show variation in what is interpreted to be the referent of the pronoun, depending on the presence of the suffix *-al*.

Most consultants interpret the self-pronoun with the suffix *-al* as subject oriented as well (see §3.3). In the case of two embedded predications, both the bare pronoun and the personal pronoun *nu* choose the subject of the embedded matrix clause; cf. (38–40).

(38) *ʡali-ni* Ali-erg *ib* say:pfv.aor *rasuj-ni* Rasul.obl-erg *ib* say:pfv.aor *sune-jni* self.obl-erg *eža* goat *as-i-ra.* take:pfv-aor-ego 'Ali<sup>y</sup> said that Rasul<sup>i</sup> said that he<sup>i</sup> bought a goat.' (39) *ʡali-ni* Ali-erg *ib* say:pfv.aor *rasuj-ni* Rasul.obl-erg *ib* say:pfv.aor *nu-ni* I-erg *eža* goat *as-i-ra.* take:pfv-aor-ego 'Ali<sup>y</sup> said that Rasul<sup>i</sup> said that he<sup>i</sup> bought a goat.' (40) *ʡali-ni* Ali-erg *ib* say:pfv.aor *rasuj-ni* Rasul.obl-erg *ib* say:pfv.aor *sune-jni-jal* self.obl-erg-emph *eža* goat *as-i-ra.* take:pfv-aor-ego

'Ali<sup>y</sup> said that Rasul<sup>i</sup> said that he<sup>i</sup> bought a goat.' 10 The self-pronoun in Mehweb

If a demonstrative is used instead of the self-pronoun or a personal pronoun, it does not take an antecedent in the same sentence:

(41) *ʡali-ni* Ali-erg *ib* say:pfv.aor *rasuj-ni* Rasul.obl-erg *ib* say:pfv.aor *il-ini--jal* this-erg--emph *eža* goat *as-i-ra.* take:pfv-aor-ego 'Ali<sup>i</sup> said that Rasul<sup>y</sup> said that he<sup>z</sup> bought a goat.'

The subject of the external embedded clause can be the antecedent of the logophoric pronoun if and only if the subject of the first embedded clause does not agree in person and/or number with the logophoric pronoun.


#### **3.3.4 Non-subject orientedness: a hypothesis**

A bare pronoun in subject position in a subordinate clause, whether it is finite or non-finite, is always 'subject oriented'. This means it is coreferent to the subject of the closest embedded clause (unless there is a mismatch in person or number properties).

In some speakers, the complex pronoun behaves in the same way. In other speakers, however, the complex pronoun has to be coreferent to the non-subject argument of the matrix clause (when present) (cf. 43–46).


#### Aleksandra Kozhukhar


In the four examples above, the self-pronoun takes the non-subject argument of the main clause as its antecedent. The referent of the embedded subject shifts from the subject to the non-subject argument of the embedding clause if the main clause contains more than one argument that can serve as an antecedent for the self-pronoun and matches it in person and number.

If all these conditions are satisfied, then the bare pronoun takes its reference from the subject of the main clause, whereas the complex pronoun takes its reference from another argument of the main clause. These rules apply to all complementation strategies and all predicates of the main clause that allow a second argument or adjunct as a potential antecedent. If the main clause lacks other arguments, or if the arguments of the main clause do not match the self-pronoun in person and number, the subject-to-non-subject shift does not occur.

The complex pronoun cannot take an argument outside the clause as its antecedent. The non-subject argument of the main clause thus may not be an immediate antecedent of the complex pronoun inside the subordinate clause. Examples (43) to (46) can be explained by introducing a zero pronoun in the subject position of the subordinate clause. This zero pronoun is non-subject-oriented (see Schema 1). On the other hand, the reference of the bare pronoun combined with an intensifier (*sunejni sunejnijal),* is always subject-oriented (that is, whenever the nearest subject matches the self-pronoun in person and/or number) – see (49).

Schema 1: Non-subject-oriented zero pronoun

[S intransitive predicate IO][self non-finite predicate] [S intransitive predicate IO][∅ self-emph non-finite predicate] [S intransitive predicate IO][self self-emph non-finite predicate] 10 The self-pronoun in Mehweb


An alternative explanation is that the complex pronoun in the subject position in the subordinate clause serves as the real subject of the clause and, unable to be bound within the local domain, takes the closest argument outside its clause as an antecedent. However, there is no evidence that an intensifier can serve as a subject of the clause.

### **4 Discourse usage**

In discourse the bare pronoun<sup>4</sup> can be used to refer to the narrator of a story. In the following contexts, the bare pronoun is used in various syntactic positions and does not have an antecedent within the sentence<sup>5</sup> .

(50) *sa‹r›i* ‹f›self *duc'* run *d-uq-un-na* f1-lv:pfv-aor-ego *k'ʷan* qot *ʡaj* perhaps *illi-šu.* that-ad(lat) 'Ii (the narrator) ran to her<sup>y</sup> .' (corpus, Poisoning: 1.20)

<sup>4</sup>There is evidence that the bare pronoun in its free logophoric function can be intensified with the suffix *-al* without changing the reference of the pronoun. The corpus, however, does not provide appropriate examples.

<sup>5</sup> It can also be hypothesized that the bare pronoun in its free logophoric function may refer to other participants of the narrative. The texts from the corpus do not provide any evidence in support of this, however, and the topic thus requires further investigation.

#### Aleksandra Kozhukhar

(51) *sune-jni* self.obl-erg *i-ra* say:pfv-ego *k'ʷan* qot *abaj-la* mother-gen *heš* this *dursi--ra* girl--and *d-aχ-uwe* f1-look.after:pfv-aor.cvb *d-uʔ-a-k'a* f1-be-irr-cond *ħu* you.sg *d-u-es* f1-be:pfv-inf *ʡaj.* perhars 'She<sup>i</sup> (the narrator) said that, my<sup>y</sup> daughter, you<sup>y</sup> better take care of her daughter.' (corpus, Poisoning: 1.8) (52) *sune-jni* self.obl-erg *i-ra* say:pfv-ego *k'ʷan* qot *marijan* marijan *ħad* you.sg.dat *d-ig-a-k'a* f1-want:pfv-irr-cond *d-uh-e* f1-become.pfv-imp *ʡaj* perhaps *ħad* you.sg.dat *ʡaˤχ-le* good-advz *b-uʔ-a-re.* n-be:pfv-irr-pst 'She<sup>i</sup> (the narrator) said: Marijam<sup>y</sup> , if you<sup>y</sup> want (to do this) marry him, maybe it would be good for you<sup>y</sup> .' (corpus, Poisoning: 1.17) (53) *sune-s* self.obl-dat *k'ʷan* qot *ʡaj* perhaps *urče* in.heart(lat) *c'a* fire *aq'-ur.* pour:pfv-aor 'She (the narrator) felt bad.' (corpus, Poisoning: 1.32) (54) *hanna* now *raχkʷar* man *r-uh-ub-le* f-become:pfv-aor-cvb *umma* kiss *r-uk'-uwe* f-lv:ipfv-cvb.ipfv *gʷa* ptcl *k'ʷan* qot *ʡaj* perhaps *sune-če* self.obl-super(lat) *hel* this.here *xunul.* woman

'Then the woman started to kiss him (the narrator).' (corpus, Speaking Lak: 1.14)

### **5 Intensifier**

The complex pronoun in Mehweb can be used as an intensifier. The intensifier is used in adposition to its head, which it emphasizes (cf. 55). This pronoun is formally identical to the reflexive pronoun<sup>6</sup> . The bare pronoun alone cannot be used as an intensifier (see 56).

<sup>6</sup>The functions of intensification and reflexivization are similarly combined in personal pronouns followed by the suffix *-al*; also cf. Table 2.

10 The self-pronoun in Mehweb

(55) *it-ini* this-erg *sune-jni-jal* self.obl-erg-emph *d-erk-un* n-eat:pfv-aor *χinč'-e.* khinkal-pl 'He<sup>i</sup> himself<sup>i</sup> ate all khinkals.'

the ram.'


The complex pronoun may intensify an overt NP (cf. 58), demonstratives (cf. 59), as well as pro-dropped pronouns in the subject position (cf. 60). The intensifier agrees in number, case and gender with its head. It can be used in all syntactic positions, including subject, P and other positions.


Some speakers are reluctant to accept intensification of NPs with low animacy:

#### Aleksandra Kozhukhar

(61) *? rasuj-ni* Rasul.obl-erg *muħammad-i-s* muhammad-obl-dat *sa‹b›i-jal* ‹n›self-emph *eža* goat *as-ib.* take:pfv-aor 'Rasul bought to Muhammad this the very goat.'

The intensifier may be preposed to its antecedent:

(62) *sa‹w›i-jal* ‹m›self-emph *wazil-li* chief-obl(erg) *b-arg-ib* n-find:pfv-aor *k'ʷan* qot *ʡilla--ra.* reason--and 'The chief<sup>i</sup> himself<sup>i</sup> found the reason.' (corpus, The story of Akula Ali: 1.7 (Magometov 1982))

The intensifier can co-occur with complex pronouns used as reflexives, as in (63) and (64). In such contexts, they seem to show a free relative order. However, (65) shows that the compound consisting of two complex pronouns cannot be split.


The intensifier can also be combined with a bare pronoun and can either precede or follow it, with no semantic contrast (cf. 66 and 67).

(66) *rasuj-s* Rasul.obl-dat *dig-uwe* want:ipfv-cvb.ipfv *le-b* aux-n *sawi* ‹m›self *sune-če-l* self.obl-super(lat)-emph *ħule* look *w-iz-es.* m-lv:pfv-inf 'Rasul<sup>i</sup> wants to look at himself<sup>i</sup> .'

10 The self-pronoun in Mehweb

(67) *rasuj-s* Rasul.obl-dat *dig-uwe* want:ipfv-cvb.ipfv *le-b* aux-n *sune-če-l* self.obl-super(lat)-emph *sa‹w›i* ‹m›self *ħule* look *w-iz-es.* m-lv:pfv-inf 'Rasul<sup>i</sup> wants to look at himself<sup>i</sup> .'

The intensifier can take the subject position in the subordinate clause since subject pro-drop is also acceptable in subordinate clauses (cf. 43–46 above). The reference of the intensifier in subject position is discussed in §3.3.4.

### **6 Resumptive**

The resumptive function of the self-pronoun is discussed in Lander & Kozhukhar (2015). Resumptive pronouns are optionally used in the position that is relativized (cf. 68, 69).


In resumptive contexts, the self-pronoun may also attach the suffix *-al*. As a result, the relativized argument is emphasized (cf. 70 and 71).

(70) *nu-ni* I-erg *sune-ze* self.obl-inter(lat) *arc* money *g-ib-i* give:pfv-aor-atr *insaj-ni* man.obl-erg *nab* I.dat *arc* money *ħa-lug-an.* neg-give:ipfv-hab

'The man<sup>i</sup> to whom<sup>i</sup> I gave the money doesn't give it back to me.'

#### Aleksandra Kozhukhar

(71) *nu-ni* I-erg *sune-ze-l* self.obl-inter(lat) *arc* money *g-ib-i* give:pfv-aor-atr *insaj-ni* man.obl-erg *nab* I.dat *arc* money *ħa-lug-an.* neg-give:ipfv-hab 'This very man<sup>i</sup> to whom<sup>i</sup> I gave money doesn't give me them back.'

Some consultants tend to use resumptives only with animate relative heads (72 and 73).

(72) *? sune-s* self.obl-dat *ʡadidi* behind *ħark'ʷ* river *b-aš-uwe* n-go:ipfv-cvb.ipfv *le-b-i* aux-n-atr *qali* house *le-b* be-n *rasuj-ja.* Rasul.obl-gen 'The house<sup>i</sup> behind which<sup>i</sup> the river flows belongs to Rasul.'

(73) *ʡadidi* behind *ħark'ʷ* river *b-aš-uwe* n-flow:ipfv-cvb.ipfv *le-b-i* aux-n-atr *qali* house *le-b* aux-n *rasuj-ja.* Rasul.obl-gen 'The house<sup>i</sup> behind which<sup>i</sup> there is a river belongs to Rasul.'

For further discussion on resumptives see Lander & Kozhukhar (2015).

### **7 Conclusion**

In this paper, I have considered the form and functions of the pronominal stem *sa*‹cl›*i* in Mehweb. This stem has at least the following functions: reflexive and long-distant reflexive, logophoric (including free logophoric), intensifier and resumptive. These functions, which are distinct from both syntactic and semantic perspectives, show different constraints on their antecedents.

The complex pronoun functions as a locally bound reflexive and may occupy any non-subject slots. The intensifier pronoun is homophonous to the reflexive and receives the same case, number and gender values as its head. The possible antecedents of an intensifier include locally bound reflexives, long-distance reflexives and logophoric pronouns; it can also be pro-dropped.

According to Reuland (2011) and Sells (1987), logophoric pronouns are pronouns used in finite subordinate clauses embedded under predicates of speech and mental experience. For Clements (1975) and Toldova (1999), the main function of the logophoric pronoun is to define the point of view. There are no typologically universal constraints on the syntactic position the logophoric pronoun, but

10 The self-pronoun in Mehweb

there is a strong tendency for the antecedent to be in the subject position of the embedded clause. Cole et al. (2000) however, discussing long-distance reflexives, argue that these take either subject or non-subject position within non-finite subordinate clauses. They also argue that long-distance reflexives manifest subject orientation: their antecedents have to be subjects of the main clause.

The pronoun *sa*‹cl›*i* covers both functions and fits both the description of the logophoric pronoun and that of the long-distance reflexive. Therefore, I suggest that in Mehweb, there is neither a morphological nor a (sharp) syntactic distinction between logophorics and long-distance reflexives.

### **List of abbreviations**


#### Aleksandra Kozhukhar


### **References**


## **Chapter 11**

## **Relative clause and resumptive pronouns in Mehweb**

### Yury Lander

National Research University Higher School of Economics

### Aleksandra Kozhukhar

National Research University Higher School of Economics

East Caucasian relative clause constructions (RCCs) are sometimes viewed as constructed mainly on the basis of semantic and pragmatic information. In this paper, we consider RCCs in Mehweb and argue that, despite the fact that the interpretation of some of them may rely exclusively on semantics, syntactic mechanisms may also come into play in their organization. We present evidence that Mehweb has grammaticalized the resumptive use of reflexive pronouns, which thus can be contrasted with other uses of reflexive pronouns due to the restrictions on animate antecedents observed only in RCCs.

*Keywords*: relative clause, relativization, resumptive pronoun.

### **1 Introduction**

Relativization is usually thought of as a mechanism which operates on an argument or an adjunct of a subordinate clause (see, for example, De Vries 2002). For example, in *the paper we are writing \_\_* the relativized argument is the direct object of the verb, while *the person that \_\_ wrote this sentence* presupposes that the relativized argument is the verb's subject.<sup>1</sup> Note that many scholars of relative clause constructions (RCCs) think of relativized arguments and adjuncts as syntactic positions and not as semantic roles. Indeed, studies of RCCs have revealed a number of restrictions on their formation which clearly have syntac-

<sup>1</sup> In both examples a gap is shown in the place of the relativized argument.

Yury Lander & Aleksandra Kozhukhar. 2019. Relative clause and resumptive pronouns in Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 295–314. Berlin: Language Science Press. DOI:10.5281/zenodo.3402074

#### Yury Lander & Aleksandra Kozhukhar

tic nature. These restrictions include, for instance, the continuous distribution of relative constructions along the Noun Phrase Accessibility Hierarchy (NPAH) Subject › Direct object › Indirect object › Obliqe object › Possessor › Object of comparison (Keenan & Comrie 1977) 2 and non-relativizability of nominals embedded in syntactic islands, like indirect questions and parts of coordinating constructions (Ross 1967).

The universality of this conception was called into question by Comrie (1996; 1998), who proposed, following Matsumoto (1988;1997), that some languages may construct what is, in their descriptions, usually considered an RCC on a semantic rather than on a syntactic basis. As was shown in the above-mentioned works and the subsequent literature (see especially the volume Matsumoto et al. 2017), such languages only need to establish a plausible semantic link between the head of the noun phrase and the subordinate clause which would be sufficient for the characterization of this head. This link sometimes involves an argument or an adjunct of the subordinate clause but it need not necessarily. Hence a new term was coined for this phenomenon, namely *generalized noun modifying clause constructions*. Naturally, such constructions do not display the syntactic restrictions proposed for languages with "canonical" relative clauses.

As we will see below, the contrast between RCCs proper and generalized nounmodifying clause constructions is not a clear-cut one. That is why in this paper we will use the terms *relative clause* and *relative clause construction* irrespectively of our stance as to the mechanisms behind the attributive patterns we discuss.<sup>3</sup> Nonetheless, we will distinguish between *syntactically-oriented RCCs* and *semantically-oriented RCCs* depending on whether or not, in a given case or set of cases, the syntactic information is relevant. If a construction contains a grammaticalized means intended for determining the relativized argument and displays clear syntactic constraints, it is considered a syntactically-oriented RCC. Otherwise, it may be considered semantically-oriented.

This paper presents a preliminary description of Mehweb RCCs in the perspective outlined above. At the clause level, Mehweb, as other Dargwa languages, is double-marking: it has case marking and verb agreement. Both kinds of marking display the ergative system, a remarkable exception being person marking, the

<sup>2</sup>This hierarchy was later extended and modified (for example, for ergative languages it was argued that the transitive undergoer has preference over the ergative argument); see Lehmann (1984: 211ff), Liao (2000), and specifically for Daghestanian languages, Lyutikova (1999; 2001).

<sup>3</sup>The terms *attributive clause* and *noun-modifying clause construction* used in literature are misleading, since cross-linguistically relatives do not always function as syntactic attributes/modifiers of nouns (cf. internally-headed RCCs or the amazingly wide use of RCCs without "head" nouns in some languages).

#### 11 Relative clause and resumptive pronouns in Mehweb

rules for which vary across Dargwa varieties (Sumbatova 2011; for discussion of the Mehweb system of personal agreement, seeGanenkov 2019 [this volume]). As for word order, Mehweb can be characterized as left-branching, although showing considerable freedom in independent clauses.

This paper is based on our fieldwork in Mehweb in 2013, 2015 and 2016. Most data were obtained through elicitation sessions. The structure of the paper is as follows: in §2 we describe the context in which we discuss Mehweb RCCs; in §3 we provide background information on relative clauses in this language; §4 is devoted to certain aspects of Mehweb RCCs that point to their syntactic nature; and §5 discusses these data from a theoretical point of view. The last section presents conclusions.

### **2 East Caucasian relative clauses**

As is typical for a left-branching language, the basic RCC in East Caucasian languages involves a relative clause preceding its head (if any).<sup>4</sup> In grammars, the form of the verbal predicate of the subordinate clause is traditionally described as a participle, although its real place in the verb paradigm varies. The difficulties in the attribution of these forms are related primarily to the fact that in many languages they coincide with some finite forms.

At first glance, East Caucasian RCCs seem like good candidates to be considered semantically-oriented. Kibrik (1980: 33) noticed that the syntactic characteristics of the relativized argument are not crucial for these constructions. Indeed, the role of the relativized argument cannot be deduced from the form of the predicate of the relative clause, neither can it be unambiguously recovered on the basis of any other grammatical property of the construction. There are no dedicated relative pronouns that mark the relativized argument, and the absence of a corresponding NP cannot serve as a reliable clue, since East Caucasian languages easily omit argument NPs even in independent clauses. Hence Comrie & Polinsky (1999), who analyzed RCCs in Tsez, argued that they may be constructed on the basis of semantic frames, and Comrie et al. (2017) continued this line of analysis for Hinuq and Bezhta, the languages of the same Tsezic branch of East Caucasian as Tsez. Daniel & Lander (2008; 2010) also proposed that RCCs in East Caucasian languages are not based on syntactic information. In this section we will illustrate the argumentation concerning these points with examples from Tanti Dargwa, a language belonging to the same branch of the family as Mehweb (see Sumbatova & Lander 2014 for details).

<sup>4</sup>A survey of the data available for East Caucasian relatives can be found in Barylnikova (2015).

#### Yury Lander & Aleksandra Kozhukhar

In general, Tanti Dargwa does not show any restrictions on what grammatical role is relativized. In this language, not only does the RCC relativize all roles in NPAH, but it is also not sensitive to syntactic islands. The following examples (both elicited) demonstrate what should presumably be described as relativization out of relative clauses and coordination constructions:<sup>5</sup>

(1) *dam* I.dat *č-ib-se* bring:pfv-aor-atr *kːata* cat *b-ibšː-ib* n-run.away:pfv-aor *хːunul* woman *simi* anger *r-ač'-ib.* f-enter:pfv-aor

'The woman such that the cat that she brought to me ran away got angry.'

(2) *aħmad-li--ra* Ahmad-erg--add *sun-ni--ra* self-erg--add *mura* hay *d-ertː-ib* npl-mow:pfv-aor *admi* man *dila* I.gen *χːutːu--sa-j.* father.in.law--cop-m

'The man with whom Ahmad mowed the hay (lit., Ahmad and who mowed the hay) is my father-in-law.'

Therefore, it seems that Tanti Dargwa lacks syntactic constraints on relativization. Moreover, a relative clause can appear even if there is no argument in the subordinate part that could be relativized. Cf. (3):<sup>6</sup>

(3) *ʕuˤ* you.sg *dam* I.dat *muher-li-cːe-r* dream-obl-inter-f(ess) *r-iž-ib-se* f1-sit:pfv-aor-atr *dila* I.gen *ʡamru* life *alžana--ʁuna--sa-tːe.* heaven--like--cop-npl+pst

'My life when I dreamt about you (lit., when you were sitting in my dream) was heaven-like.'

It is impossible to describe (3) as a result of any syntactic operation which deals with an argument of the relative clause. Hence, this RCC is likely to be semantically-oriented.

<sup>5</sup> For the reasons discussed in the paper, glossing occasionally follows rules that are different from other papers of the volume.

<sup>6</sup>The presence of the attributive suffix on the predicate of the relative clause in (3), which at first glance makes it different from the previous examples, is not related to any difference in the mechanisms of constructing the relation between the head and the relative clause. For a discussion of the distribution of the attributive suffix in Tanti Dargwa, see Lander (2014).

#### 11 Relative clause and resumptive pronouns in Mehweb

Still, it is doubtful that East Caucasian relatives never rely on syntactic information. As Daniel & Lander (2013) argued, the frequency of relativization of a syntactic position may depend on whether a language displays ergative features or not, even within this family. It may be that syntax is still engaged, even though, sometimes, these relatives only rely on semantics and pragmatics.

In addition, constraints on relativization have been reported for some East Caucasian languages. For example, according to Tatevosov (1996: 215), Godoberi does not relativize possessors, objects of comparison and objects of postpositions. Mutalov & Sumbatova (2003) note that in Itsari Dargwa "[r]elativization is impossible only for constituents of coordinate clauses and at least doubtful for constituents of adverbial clauses". Lyutikova (1999; 2001) reports that Tsakhur and Bagwalal prohibit relativization for the positions mentioned for Itsari as well. Moreover, although the syntactic limits of relativization are always quite loose, it is worth noting that informants do not always accept relativization of all syntactically peripheral participants without an appropriate context, even in languages whose RCCs are commonly believed to be semantically-oriented.

Another problem for a purely semantic treatment is posed by the fact that in many East Caucasian languages the relativized argument can be expressed within a relative clause by a reflexive pronoun, as in (4). Such pronouns look like resumptive pronouns, which directly point to the *syntactic* position that is relativized.

(4) *du* I (*sun-ni-šːu*) self-obl-ad(lat) *q'ʷ-aˤn-se* go:ipfv-prs-atr *qali* house 'the house where I am going'

Still, these pronouns differ from typical resumptives in various significant ways.

First, to refer to relativized arugments, East Caucasian languages use reflexive pronouns, while typical resumptives cited in the typological literature are nonreflexive.<sup>7</sup> Yet the appearance of reflexive pronouns in RCCs may be related to the fact that reflexive pronouns in this family have very wide distribution: for example, they are used as logophoric pronouns or in independent clauses both as intensifiers and as pronominals (Testelets & Toldova 1998). This suggests that reflexive pronouns in East Caucasian languages are much more neutral means

<sup>7</sup>Note, however, that reflexives used as resumptives are found outside the East Caucasian family as well. For example, Lee (2004) provides a detailed discussion of the resumptive use of a reflexive pronoun in Korean, Csató & Uchturpani (2010) describe reflexive resumptives in Uyghur, and Johanson & Csató (1998: 219) report the resumptive function of reflexives in Turkish.

#### Yury Lander & Aleksandra Kozhukhar

of pronominal reference than their counterparts in Standard Average European languages.

Second, East Caucasian languages sometimes allow resumptive reflexive pronouns in the most privileged syntactic positions occupying the top of NPAH, such as those of the intransitive subject (5), transitive actor (6) and transitive undergoer (7). Cf. the following Tanti Dargwa examples:


Typical resumptive pronouns in relative clauses prefer the positions that occur lower in syntactic hierarchies (Keenan & Comrie 1977; Maxwell 1979: 92). Hence, East Caucasian resumptives are different from typical resumptives.<sup>8</sup>

Daniel & Lander (2008) suggested that reflexives in relatives do not serve to mark the relativized position, i.e. they are only anaphoric devices, independent of relativization (cf. also Comrie et al. 2017: 133). If so, their existence does not contradict the idea that East Caucasian RCCs do not apply to syntactic information. The data from Mehweb we proceed to present make the issue of the use of reflexives more intriguing and return us to the idea that, after all, these can be treated as resumptives.

### **3 Relatives in Mehweb: first glance**

The basic RCC in Mehweb Dargwa involves a relative clause which precedes the head of the noun phrase, if there is one. The predicate of the relative clause is marked with an attributive suffix, which has allomorphs *-il*, *-i*, and *-l*. The same suffix is found with some other attributes, such as adjectival attributes. Some examples of RCCs are given in (8–9):

<sup>8</sup>Again, there do exist languages which allow resumptives in the subject position, but these uses are usually considered exceptional. We do not have information on the degree of markedness of such uses as (5–7) in East Caucasian languages.

11 Relative clause and resumptive pronouns in Mehweb

(8) *naˤʁ* hand *iz-u-l* hurt:ipfv-prs-atr *insan* person 'a person whose hand hurts'

(9) *nu* I *q'-oˤwe* go:ipfv-cvb.ipfv *d-uʔ-ub-i* f1-be:pfv-aor-atr *huni* road 'the road I was going with'

According to Magometov (1982: 112–115) and Khajdakov (1985: 105–107), Mehweb distinguishes between three types of participle with respect to the stem they are formed with and the variant of the attributive suffix they adjoin; cf. Table 1.

Table 1: Participles in Mehweb Dargwa


While the past and future participles are morphologically transparent and include just the corresponding base and the attributive suffix, the present participle contains the former marker of the present tense *-u*, which is found in present converbs.<sup>9</sup> While it is glossed as prs in this paper,<sup>10</sup> one should bear in mind that its distribution is limited to few non-finite forms and it can be used as a marker of a relative tense rather than as an absolute tense.<sup>11</sup>

We take the participles listed above as the canonical predicates of relative clauses. However, it should be noted that the predicates of relative clauses are not confined to these participles. For example, we have RCCs where the attributive suffix is added to the copula/existential verb, as in (10–11):

<sup>9</sup>Michael Daniel (pers. com.) noted that it is most likely that imperfective converbs are actually derived from imperfective participles.

<sup>10</sup>Note that in using this gloss for *-u*, our paper differs from other papers of this volume.

<sup>11</sup>The finite present tense is expressed periphrastically by a combination of the present converb with a copula.

Yury Lander & Aleksandra Kozhukhar


As shown by examples, the relativized argument need not be expressed overtly within the relative clause. As in Tanti Dargwa, it is not difficult to construct an example where the relation between the relative clause and the head must be established by the context:

(12) *nu-ni* I-erg *b-erk-un-na* n-eat:pfv-aor-ego *itti* that *b-urʁ-es* hpl-fight:ipfv-inf *b-aq-ib-i* hpl-let:pfv-aor-atr *t'ult'.* bread 'I ate the bread which served as the reason for them to fight.'

If the relativized argument can be reconstructed, it usually can be expressed with a pronoun *sa*‹cl›*i* (here cl is a gender marker), which has several suppletive forms and whose partial paradigm is given in Table 2. This pronoun also serves


Table 2: Case-number-gender forms of the pronoun *sa*‹cl›*i*

<sup>12</sup>The example is additionally interesting because it relativizes one of the arguments of the socalled biabsolutive construction. Cf. the original independent construction:

(i) adami man kʷiha ram b-erh-u-we n-slaughter:pfv-prs-cvb le-w aux-m 'The man slaughtered a ram.'

#### 11 Relative clause and resumptive pronouns in Mehweb

as a reflexive pronoun (both local and long-distance), as a logophoric pronoun, and as an intensifier (see Kozhukhar 2019 [this volume]).

Some examples of the use of *sa*‹cl›*i* as a resumptive are given below. In (13) it appears in the indirect object position, in (14) it serves as the possessor of the intransitive subject, and in (15) it refers to the experiencer with the experiential verb:

(13) *nu-ni* I-erg *ču-s* self.pl.obl-dat *kung* book *gib-i* give:pfv-atr *ule* child.pl *b-aˤq'-un* hpl-go:pfv-aor *uškuj-ħe.* school.obl-in(lat)

'The children to whom I gave a book went to school.'


'The shepherds who saw a demon were scared.'

### **4 Syntactic orientedness**

Even though Mehweb data show considerable resemblance to Tanti Dargwa, there are important differences between the two Dargwa varieties which suggest that relativization in Mehweb may be syntactically-oriented.

#### **4.1 Resumptives at the top of NPAH**

Unlike in Tanti Dargwa, the Mehweb pronoun *sa*‹cl›*i* is sometimes considered infelicitous at the top of NPAH. Cf. the following example where the position relativized into is the actor of a transitive clause:

(16) (*\*sune-jni*) self.obl-erg *kʷiha* ram *b-erh-un-i* n-slaughter:pfv-aor-atr *adami-li-ze* man-obl-inter(lat) *nu* I *g-ub.* see:pfv-aor 'The man who slaughtered the ram saw me.'

#### Yury Lander & Aleksandra Kozhukhar

When used as intensifier, *sa*‹cl›*i* is normally accompanied by the emphatic clitic --*al* (with an allomorph --*jal* after vowels). Crucially, the same speaker who found the use of the resumptive in (16) infelicitous allows the pronoun followed by --*al* in the same position:

(17) *sune-jni--jal* self.obl-erg--emph *kʷiha* ram *b-erh-un-i* n-slaughter:pfv-aor-atr *adami-li-ze* man-obl-inter(lat) *nu* I *g-ub.* see:pfv-aor 'The man who himself slaughtered the ram saw me.'

This example demonstrates that the impossibility of using *sa*‹cl›*i* in this position cannot be attributed to any morphological rule that prohibits this pronoun in this position in general: after all, it occurs there as an intensifier.

As noted by an anonymous reviewer, it could be that the emphatic clitic changes the distribution of the pronoun. Yet there are also speakers who have no problems with the use of the resumptive (lacking the emphatic particle) in all core syntactic positions, including the positions of the intransitive subject (18) and transitive actor (19):


Our data concerning the possibility of the use of a resumptive at the top of NPAH are not definitive. The fact that some speakers are more restrictive in the use of *sa*‹cl›*i* in the resumptive function suggests, however, that this function may be governed by syntactic rather than semantic rules.

#### **4.2 Coordinate structure constraint**

Mehweb does not allow relativization out of a conjunct in the coordination construction and hence follows one of the island constraints, namely the coordinate

11 Relative clause and resumptive pronouns in Mehweb

structure constraint. (20a) illustrates the coordination construction marked with the additive clitic *--ra*. (20b) demonstrates an unsuccessful attempt at relativizing one of the conjuncts.

	- b. *\*nu-ni--ra* I-erg--add *sune-jni--ra* self.obl-erg--add *heš* this *kung* book *b-elč'-un-i* n-read:pfv-aor-atr *adami* man *w-ak'-ib.* m-come:pfv-aor (Expected: 'The man who read this book together with me (lit., I and who read this book) came.')

This contrasts Mehweb with Tanti Dargwa, where the coordinate structure constraint does not apply (cf. (2) above), and again suggests that syntactic rules might be at work here.

### **4.3 An argument for resumptive function**

In general, reflexives in Dargwa languages and in Mehweb in particular are insensitive to the animacy or humanness of their antecedent. This is shown in (21–22), where in the first example *sunes* has an animate (human) antecedent and in the second example *sunela* has an inanimate antecedent:


However, some consultants claim that the appearance of *sa*‹cl›*i* in the resumptive function is only possible if the head of the relative clause is animate. Examples (23–24) show the possibility of the use of the pronoun in RCCs with human and non-human animate antecedents:

Yury Lander & Aleksandra Kozhukhar


On the contrary, (25) demonstrates that a resumptive reflexive with an inanimate antecedent is infelicitous:

(25) ( *⁇?sune-la*) self.obl-gen *baˤʜ* wall *ark-ib-i* turn.into.ruin:pfv-aor-atr *qali* house 'the house whose wall crashed down'

Interestingly, this restriction is independent from the gender system of Mehweb which contrasts humans and non-humans rather than animates and inanimates (see Footnote 13).

The restriction of *sa*‹cl›*i* to animates is crucial exactly because it is not observed in non-resumptive uses. As such, it separates the resumptive function from the other functions of the pronoun and goes against Daniel & Lander's (2008) hypothesis that reflexive pronouns in Daghestanian RCCs are not used as resumptives *per se*. If, according to some consultants' intuition, Mehweb has developed a dedicated resumptive use of pronouns characterized by specific restrictions, the RCCs involving such pronouns should be recognized as syntactically oriented. Again, no constraint of this kind is observed in Tanti Dargwa, where the reflexive pronoun easily occurs in the place of a relativized argument with an inanimate antecedent (4).

#### **4.4 Realizations of functions of** *sa***‹cl›***i*

In theory, when referring to a relativized argument within a relative clause, *sa*‹cl›*i* may fulfill not only the resumptive function but also the intensifier function and the reflexive proper function. These functions could in theory be distinguished on the basis of (i) the restriction to animates in the resumptive function, and (ii) the presence of the clitic *--al* in the intensifier function. In reality, however, the picture is more complex.

The intensifier function of *sa*‹cl›*i* is indeed observed, for example, in the following example:

11 Relative clause and resumptive pronouns in Mehweb

(26) *ʁarʁu-be* stone-pl *ar-d-ik-ib* pv-npl-fall:pfv-aor *sa‹r›i\*(--jal)* self‹cl›(\*--emph) *d-uʔ-ub-i* npl-be-aor-atr *merʔ-ani-če-la* place-pl-super-el

'The stones rolled from their own places.' (Lit., 'The stones rolled from the place they themselves occupied.')

In (26) *sari* refers to the intransitive subject and requires the emphatic clitic. Its inability to function as a resumptive (without the clitic) may be explained either by its high position in NPAH or by its inanimate reference. Importantly, the inanimate reference does not block its appearance in the intensifier function.

The realization of the reflexive function within a relative clause, on the other hand, turns out to be impossible, as (27) shows:

(27) *nu-ni* I-erg *(\*sune-la)* self.obl-gen *weˤʡi-ze* master-inter(lat) *g-ib-i* give:pfv-aor-atr *mažar* gun *b-oˤrʡ-oˤb* n-break:pfv-aor 'The gun that I returned to its owner broke.'

In this example, *sunela* could be expected to mark the coreference of the possessor with the undergoer argument (which is then relativized), yet it does not. Since the reflexive is possible in the same position in the independent clause (22), we suspect that the effect observed in (27) is due to the fact that the pronoun is interpreted as a resumptive, in which case it violates the animacy restriction.

Thus the resumptive function blocks the reflexive interpretation. This rule is not likely to be based on any semantic principle independent of the grammar, so we take it to be another piece of evidence for grammaticalization of the resumptive function in this language.

### **5 Towards an explanation of the Mehweb pattern**

To sum up, even though RCCs in Mehweb can be built on a semantic basis, in many cases their functioning relies upon strict syntactic mechanisms and constraints. At least when the relativized argument is animate, the construction resembles RCCs described for better known languages in a traditional fashion much more closely, since this argument can be expressed with a resumptive pronoun proper. These data support the conclusion made by Daniel & Lander (2013)

#### Yury Lander & Aleksandra Kozhukhar

that the borderline between RCCs involving syntactic mechanisms and RCCs which are based on the semantic information is not strict.

We have no obvious explanation for the Mehweb pattern we observed above. Nonetheless, below we present some speculations.

First, note that there are a number of languages where resumptive pronouns are found in RCCs mostly or even only when the relativized argument is animate; cf. Bošković (2009) on Serbo-Croatian and Bulgarian (Slavic), Csató & Uchturpani (2010) for Uyghur (Turkic), Erteschik-Shir (1992: 104–105) for Hebrew (Semitic), Kawachi (2007) for Sidaama (Cushitic). It may be that the Mehweb system results from grammaticalization of a similar tendency. Still, there are languages where at least in some contexts resumptives tend to be used for inanimates rather than animates, such as Arabic (Al-Zaghir 2014). Sometimes this can be grammaticalized. Lyutikova (1999: 474–475) reports that in another East Caucasian language, Tsakhur, the construction relativizing the object of a postposition only requires a resumptive pronoun if the relativized argument is inanimate.

Second, we may suspect that the most typical uses of relatives are associated with high accessibility of the relativized argument. This is partly reflected in NPAH but can also manifest itself in other parameters such as animacy, which is said to correlate with conceptual accessibility (see some discussion in van Nice & Dietrich 2003). Since more typical uses are more likely to be grammaticalized (see Lander 2015 for discussion), it is expected that relativization based on syntactic (i.e. grammatical) information is found for more accessible arguments. Note, however, that the construction with resumptives retains considerable semantic transparency (Keenan 1975) and therefore is in a sense less grammaticalized than constructions with the most accessible arguments. In other words, the absence of resumptives at the top of NPAH may be explained by the fact that this top is not primarily based on semantics, but the absence of resumptives for less accessible arguments may be explained by the fact that these constructions do not elaborate on syntactic information.

Still, this approach has a notable shortcoming. The evidence that relativization prefers animate arguments is somewhat scarce,<sup>13</sup> since most studies of the interaction between animacy and relativization are devoted to the way in which animacy affects the predictability of what is relativized. Moreover, things may be turned the other way round. The most accessible arguments are not normally described with a complex noun phrase with a modifier, since their accessibility

<sup>13</sup>For example, in Tsakhur, during elicitation the choice of what is relativized is sometimes influenced by animacy (Lyutikova 1999: 476–477), and for Turkish it is reported that headless RCCs by default have animate reference (Kerslake 1998). The latter, of course, may be just the property of headless relatives.

11 Relative clause and resumptive pronouns in Mehweb

allows them to be more economically expressed (such as by means of pronouns, proper names, simple noun phrases, etc.), cf. Ariel (1990). Since the inherent accessibility features of the antecedent and the relativized argument are (normally) identical, the very fact that the speaker has to use a highly complex phrase based on a RCC would imply that the target of relativization need not necessarily be accessible, at least as far as animacy is concerned. In any case, more research is needed on the issue of the interaction between animacy and relativization.

### **6 Conclusion**

In this paper, we provided a sketch of relativization in Mehweb against the background of the remarkable freedom of relativization in (at least some) other East Caucasian languages. In particular, we gave preliminary evidence for the idea that this language has grammaticalized resumptives and relies on syntactic information during relativization.

To be sure, these conclusions should not be taken for granted. In fact, even for resumptives, which we specifically addressed above, it is not clear whether all their uses should be considered alike; as argued by Erteschik-Shir (1992) among others, different types of resumptives may even occur in one language. A deeper investigation of the functioning of relatives in Mehweb and other East Caucasian languages, including both corpus analysis and psycholinguistic experiments, certainly may help to refine the conclusions presented here.

### **Acknowledgements**

We are grateful to our consultants for their help and their patience and to the editors and two anonymous reviewers for their comments on an earlier draft of the paper. All errors are ours.

### **List of abbreviations**


#### Yury Lander & Aleksandra Kozhukhar


### **References**

Al-Zaghir, Zainab Marzouk. 2014. *Relativisation and accessibility: A corpus analysis of relative clauses in Arabic written texts*. University of Otago. (Doctoral dissertation).

Ariel, Mira. 1990. *Accessing noun-phrase antecedents*. London: Routlege.


## **Chapter 12**

## **The Mehweb "assertive" copula gʷa: a sketch of a portrait**

### Yury Lander

National Research University Higher School of Economics

Mehweb Dargwa features a particle *gʷa*, a peculiar element which is basically used for emphasizing the assertion. The paper explores some grammatical characteristics of this particle. It is shown that, in both verbal and non-verbal clauses, *gʷa* serves as a predicative marker forming a complete predication and is an equivalent of a copula (even though, unlike the neutral copula in Mehweb, it lacks inflection). Similarly to typical East Caucasian predicative markers, *gʷa* may occur in different positions, though its place is syntactically constrained (e.g., it cannot be embedded within syntactic islands). Still, Mehweb speakers allow *gʷa* not to be adjoined to either the predicate or the focus. This makes the distribution of the particle surprising as compared with similar predicative markers in well-described East Caucasian languages, where they may either occur on the predicate or immediately follow the focused element.

### **1 Introduction**

This paper presents a preliminary description of the particle *gʷa* in Mehweb, a language of the Dargwa branch of the East Caucasian family. The following examples illustrate the use of this marker in a verbal clause (1) and in an equative clause (2):

(1) *ʔudidi-li* under.el-atr *ħark'ʷ-li* river-erg *ar-χ-uwe* away-bring:ipfv-cvb.ipfv *gʷa!* asrt

'The river carries away the one who is downstream!' (Molla Rasbaddin goes to the market place: 1.11)

Yury Lander. 2019. The Mehweb "assertive" copula gʷa: a sketch of a portrait. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 315–331. Berlin: Language Science Press. DOI:10.5281/zenodo.3402076

Yury Lander

(2) *hel* this *čudu* chudu *gʷa* asrt *di-la.* I.obl-gen 'This chudu (a kind of pie) is mine.'

The function of *gʷa* is not at all obvious. Etymologically, this particle is likely to originate from the imperative of the verb 'see' (which, as an imperative, is not fully felicitous – see Dobrushina 2019 [this volume]). Magometov (1982: 128) translated *gʷa* by the Russian particles *ved'* and *že*, whose semantics are by no means clear. The speakers often suggest that *gʷa* is frequent in disputes and emphasizes a claim. Given this, I will label it an *assertive* marker. Further research is needed for an exhaustive description of the rules that govern its use. What I will argue are the following two specific points:


The latter makes *gʷa* look quite peculiar against the background of what we know about copulas in many East Caucasian languages and in Dargwa languages in particular.

The issue of copula-ness is addressed in §2. In §3, I discuss the use of the marker in verbal predications and describe syntactic restrictions on its position. §4 describes the use of *gʷa* in non-verbal predications. The last section presents conclusions.

### **2 The assertive marker as a copula**

Many East Caucasian languages have elements which are described as copulas or predicative markers, i.e. as markers which are normally added to some lexical material in order to form complete predications (finite, unless these copulas themselves take a subordinate form).<sup>1</sup> Although their individual morphological and syntactic properties vary, these elements are clearly distinguishable from verbs. There are typically several predicative markers in a single language: for

<sup>1</sup> Some important studies addressing the behaviour of predicative markers in East Caucasian (especially with respect to their interaction with focus) include Harris (2000; 2002) on Udi, Kazenin (2002) on Lak, Sumbatova (2011) and Sumbatova & Lander (2014) on Tanti Dargwa. Forker (2013) discusses question particles which typically represent a kind of predicative markers in these languages. Testelets (1998), Kalinina & Sumbatova (2007) and Forker & Belyaev (2016) describe the influence of the position of some predicative markers on the overall clause structure.

#### 12 The Mehweb "assertive" copula gʷa: a sketch of a portrait

example, many languages have dedicated predicative markers used in questions in addition to those used in simple declaratives.

Predicative markers appear both in verbal and non-verbal predications. Below I will illustrate their use with a few examples from Udi, a language belonging to the Lezgic branch of the East Caucasian family, thus only distantly related to Mehweb.<sup>2</sup>

Predicative markers in Udi are highly grammaticalized and now commonly described as clitics (Harris 2000, 2002). They include personal markers which usually show agreement with the subject (S or A) and the question marker, which only appears in interrogative contexts and is not discussed here (but see Harris 1992). The following examples illustrate the use of the 1st person plural personal marker *--jan* in a non-verbal predication (3) and in verbal predications (4–5):<sup>3</sup>


Note that predicative markers attach not only to the lexical predicate (4) but also to the focused element (5). This can be viewed as a kind of competition for acquiring head properties between the semantic head (the predicate) and the most relevant element of the clause (i.e. focus).<sup>4</sup>

In Dargwa languages, predicative markers are less grammaticalized than in Udi. In particular, they show some properties of autonomous words. Many such markers readily constitute autonomous expressions (such as 'yes' or 'no'). Some of them take attributive and adverbial morphology and hence are akin to content words.

<sup>2</sup>Here I omit some important details of the Udi system, including the existence of a series of dative clitics and a more verb-like copula-like element used in existential, possessive predication, and identificational clauses, which also takes a predicative marker.

<sup>3</sup>The Udi examples are from the corpus of text in the Nizh dialect of Udi collected by Dmitry Ganenkov, Timur Maisak and the author.

<sup>4</sup> See Lander (2009) for some discussion of competition between semantically obligatory elements and the most relevant elements for the head properties.

#### Yury Lander

The primary Mehweb predicative marker is the copula *le-*cl (for morphology, see Daniel 2019 [this volume]), with a gender agreement slot controlled by the absolutive argument. Its use in non-verbal predications is shown in (6–7), while its use in verbal predications is illustrated in (8–9).


Like in Udi, the Mehweb predicative marker in verbal clauses can follow either the verb or the focused constituent. However, unlike in Udi, the Mehweb copula requires that a verb *be* in a non-finite (participial or converbal) form, while finite verb forms do not combine with the predicative marker. In fact, combinations of a copula and a lexical verb look like periphrastic forms, although the issue of monoclausality of these constructions is tricky.<sup>5</sup>

Turning to the assertive marker *gʷa*, it can be shown that it has the distribution of a copula. There are two pieces of evidence for this. First, similarly to *le*-cl, the assertive marker cannot appear in clauses that contain finite verb forms (10):

	- engine n-work:ipfv-hab asrt 'The engine works.'

<sup>5</sup> See Sumbatova & Lander (2014) for a detailed discussion of this issue in Tanti Dargwa, another Dargwa variety.

#### 12 The Mehweb "assertive" copula gʷa: a sketch of a portrait

Second, the assertive marker cannot combine with a copula (11a–b), unless the latter does not appear in a non-finite form, as in (11c). If we assume that *gʷa* is a copula, this is explained: a clause cannot contain two copulas.

(11) a. *dag* yesterday *it* that *derbenti-ze-la* Derbent-inter-el *w-ak'-i-le* m-come:pfv-aor-cvb *le-w* aux-m (*\*gʷa*)*.* asrt b. *dag* yesterday *it* that *derbenti-ze-la* Derbent-inter-el *w-ak'-i-le* m-come:pfv-aor-cvb *gʷa* asrt (*\*le-w*)*.* aux-m c. *dag* yesterday *it* that *derbenti-ze-la* Derbent-inter-el *w-ak'-i-le* m-come:pfv-aor-cvb *le-w-le* aux-m-cvb *gʷa.* asrt 'Yesterday he came from Derbent.'

It is worth mentioning, however, that *gʷa* differs from *le*-cl in that it does not take any morphology.

### **3 Verbal predications**

Just like the copula *le*-cl, the assertive marker need not follow the verb but can appear after focused elements:

	- b. *nuša-jni* we-erg *kulubi-s* club-dat *gʷa* asrt *remont* renovation *b-aq'-i-le* n-do:pfv-aor-cvb 'It was the club for which we made the renovation.'

I will distinguish between the wide scope use of *gʷa*, where it has a scope over the whole sentence or over the predicate and follows this predicate, and the narrow scope use of *gʷa*, where it should follow exactly the focused phrase. In verbal clauses, the wide scope *gʷa* is found with the neutral converb (13) and with

#### Yury Lander

the infinitive (14–15) but not with the participle (cf. the infelicitous (16) with (19) below):<sup>6</sup>

(13) *qʷe* vow *b-iq'-uwe* n-do:ipfv-cvb.ipfv:ipfv *gʷa,* asrt *ħu* you.sg *ħa-k-i-le* neg-bring:pfv-aor-cvb *ħa-wʔ-iša.* neg-m.be-fut.ego 'I swear I will take you (as a wife).' (Widow)

(14) *durʡa* lose *uh-ub-i-li* m.lv:pfv-aor-atr-erg *derqʷ* winning *uh-ub-i-s* m.become:pfv-aor-atr-dat *ca* one *dus-li* year-erg *quli-w* house.ess-m(ess) *w-at-ul-le* m-put:ipfv-ptcp-advz *uz-es* m.work:ipfv-inf *gʷa.* asrt

'The one who will lose will work as a servant for the one who will win, for one year.' (Widow)


If the assertive marker follows a constituent other than the predicate, the choice of the verb form is less restricted. In this construction not only a converbal form (17) and an infinitive (18) but also a participial form (19) is allowed:<sup>7</sup>


<sup>6</sup>Presumably, the assertive marker should combine with the participle where it functions as the head of the nominal predicate in a nominal clause. However, I lack relevant examples.

<sup>7</sup>These combinations stand in parallel with similar combinations of converbs, infinitives and participles with the standard copula (cf. Daniel 2019 [this volume]).

12 The Mehweb "assertive" copula gʷa: a sketch of a portrait

(19) *musa-ni* Musa-erg *gʷa* asrt *poˤroˤm* glass *b-oˤrʡ-aq-ib-i.* n-break:pfv-caus-aor-atr 'It was Musa who broke the glass.'

In examples (17–19) we observe the assertive copula following focused NPs. (20–22) demonstrate that *gʷa* can follow other kinds of constituents, such as adverbs and embedded clauses:


Still, we do find restrictions on what can be focused by means of *gʷa*. 8 For example, the assertive marker cannot immediately follow postpositional objects, but rather occurs after the whole postpositional phrase:

(23) a. *\*heč'* that *dubur-li-če* mountain-obl-super(lat) *gʷa* asrt *aqu-r* up-npl(ess) *dirigʷ* cloud *хaʔ* appear *d-uh-ub-le.* npl-become:pfv-aor-cvb b. *heč' dubur-li-če aqu-r gʷa dirigʷ хaʔ*

that mountain-obl-super upper-npl(ess) asrt cloud appear *d-uh-ub-le.* npl-become:pfv-aor-cvb

'It is over that mountain that the cloud appeared.'

Further, the assertive marker cannot be embedded in an NP. In particular, it cannot occur immediately after an adjective attribute (24), an attributive demonstrative (25) or a quantifier (26) when they precede the head noun:

<sup>8</sup> I hypothesize that these restrictions hold for the neutral copula as well, but I lack the necessary data.

#### Yury Lander

	- b. *ħunt'a-l* red-atr *burχa-li-če-r* roof-obl-super-npl(ess) *gʷa* asrt *ʁarʁ-ube.* stone-pl 'There are stones on the RED roof.'
	- b. *heš* that *ʁʷet'i-če-r* tree-super-npl(ess) *gʷa* asrt *d-aq-il* npl-much-atr *inc-be* apple-pl *d-urh-uwe.* npl-become:ipfv-cvb.ipfv 'There are many apples growing on THAT tree.'
	- b. *har-il* each-atr *urši-li-s* boy-obl-dat *gʷa* asrt *midal* medal *g-i-le.* give:pfv-aor-cvb 'He gave a medal to EACH boy.'

One natural way to focus an attribute is to place the assertive copula after the whole NP. Alternatively, one can split the description of a participant into two NPs with a semantic attribute being nominalized and taking its own case marker. Since the semantic attribute itself constitutes a complete NP in this construction, it becomes possible to place *gʷa* immediately after it (27). Notably, for absolutive NPs this results in the illusion of the embedding of the assertive marker in an NP (28), but this is likely to be a consequence of the fact that absolutive NPs do not receive overt case marking, so the two adjoined absolutive NPs look as a single phrase.


#### 12 The Mehweb "assertive" copula gʷa: a sketch of a portrait

Further, *gʷa* cannot occur within syntactic islands. For example, it cannot be embedded in a coordination construction (29) or in a converbal clause (30).

	- b. *b-urq'-il* n-old-atr *bartbisu* carpet *b-ic-i-le* n-sell:pfv-aor-cvb *gʷa,* asrt *d-aq-il* npl-much-atr *arc* money *d-aq'-i-le.* npl-do:pfv-aor-cvb 'After selling THE OLD CARPET, he got much money.'

Unlike most Dargwa varieties, Mehweb has developed a biabsolutive construction<sup>9</sup> (see also Daniel 2019 [this volume] and Ganenkov 2019 [this volume]). In this construction, a transitive verb appears as a converb and requires a copula but the actor appears in the absolutive, as does the undergoer. This construction is possible with *gʷa* (31a–b), yet the assertive copula cannot occur between the P-argument and the converb (31c).<sup>10</sup> This contrasts the biabsolutive construction with a simple combination of the converb with a copula and suggests that this pattern contains an embedded converbal clause which is an island, at least with respect to *gʷa*:

(31) a. *musa* Musa *kaš* kasha *d-uk-uwe* npl-eat:ipfv-cvb.ipfv *gʷa.* asrt 'Musa is eating kasha.'

<sup>9</sup>Biabsolutive (binominative) constructions are quite widespread in the East Caucasian family, but are not typical for the Dargwa branch, where they have been previously only reported for Itsari Dargwa (Mutalov & Sumbatova 2003). See Forker (2012) and Gagliardi et al. (2014) for surveys of some properties of this kind of constructions as well as for a discussion of their diversity and possible analyzes.

<sup>10</sup>The same set of facts is observed for the simple copula *le*-cl.

Yury Lander


With clausal complements, the situation is less obvious: some (but by no means all) speakers allow positioning *gʷa* within a clausal complement (32–33).<sup>11</sup>


While the placement of *gʷa* after a constituent other than the predicate usually indicates focus shift, even in this case it does not need to follow the constituents that are (likely to be) focused. Consider the following examples:

(34) a. *χadižat-ini--ra* Khadizhat-erg--add *heš* that *kung* book *gʷa* asrt *b-elč-u-we.* n-read:pfv-aor-cvb b. *χadižat-ini--ra* Khadizhat-erg--add *gʷa* asrt *heš* that *kung* book *b-elč-u-we.* n-read:pfv-aor-cvb 'Even Khadizhat has read that book.'

In (34) one can hypothesize that the focused constituent is the ergative NP, since it is marked with the additive clitic meaning 'even'. Yet as shown by these examples, the assertive copula may but need not be adjacent to the focused phrase: indeed, in (34a) it follows the absolutive argument. These examples suggest that focus is possibly not the only factor which determines the position of *gʷa*. More generally, we conclude that in verbal clauses the grammatical position of *gʷa* should be determined neither by the predicate nor by focus.

<sup>11</sup>The superscripted *%* in these examples refers to the fact that there is considerable variation among speakers in the acceptance of such examples.

12 The Mehweb "assertive" copula gʷa: a sketch of a portrait

### **4 Non-verbal predication**

Non-verbal predication is represented by two types, namely existential clauses and non-existential clauses with non-verbal predicates (nouns, adjectives, numerals, demonstratives, etc.). In Mehweb, the latter type allows the absence of a copula while the former normally does not.<sup>12</sup> The assertive copula can appear in both types.

(35–36) show examples of the use of *gʷa* in existential predication that assert the existence of entities or events described by an NP. Note that, in Mehweb, this type includes possessive predication (37).


The assertive copula is also found in clauses emphasizing the existence of the already known entities (sometimes in combination with the converbal form of the copula; cf. (38)) or describing the location of the already known entities (39):


(40–41) show examples of the use of *gʷa* in clearly non-existential predications.

(i) išbari today meħʷe-b in.Mehweb-n(ess) beʁ. wedding 'There is a wedding in Mehweb today.'

<sup>12</sup>An important exception is the use of NPs denoting events, which allow the absence of copula, as in (i):

Yury Lander


'This (cauldron) is the cauldron originating from (lit., made by) your cauldron.' (Molla Rasbaddin and the neighbour's cauldron, 1.5)

At least if the assertive marker follows the demonstrative, their combination can be embedded within the alleged subject phrase. In (42) the phrase *heš gʷa* 'that is' is embedded within the relative clause construction 'the house which Rasul built'.

(42) *rasuj-ni* Rasul.obl-erg *[heš* that *gʷa]* asrt *b-aq'-ib-i* n-do:pfv-aor-atr *qali.* house 'The house that Rasul built is that one.'

Negative non-verbal predication in Mehweb contains a dedicated negative copular verb. If *gʷa* is needed, this copula appears in a converbal form:

(43) *it* that *učitel* teacher *aħi-je* be:neg-cvb *gʷa.* asrt 'He is not a teacher!'

In equative clauses, determining what the predicate is presents a complex issue because of the formal similarity between the subject and the nominal predicate. Still, one can find indirect evidence for the predicate status of one of the noun phrases based on various semantic and syntactic tests. By using these tests, it is possible to show that the assertive marker does not have to immediately follow the predicate.

First, if a nominal phrase in an equative clause includes a reflexive pronoun bound by the other part of the clause, it is likely that it is a predicate and the reflexive is bound by the subject. However, *gʷa* need not follow such a nominal predicate:

(44) *šamil* Shamil *gʷa* asrt *sune-s-al* self.obl-dat-emph *weˤʡ.* master 'Shamil is a boss of himself.'

#### 12 The Mehweb "assertive" copula gʷa: a sketch of a portrait

Second, in an equative clause, an expression with a true distributive quantifier arguably should not function as a predicate (Partee 1987; but see Arkadiev & Lander 2013 for counterevidence). Yet, *gʷa* is possible with the quantified NP:

(45) *har* every *insan* person *gʷa* asrt *sune-s-al* self.obl-dat-emph *uħna-w* m.inside-m(ess) *rasul* Rasul *ħamzatow* Gamzatov. 'Everyone is Rasul Gamzatov (a famous Daghestanian writer) deep inside.'

Finally, if an equative clause contains an adjunct, the assertive copula may follow this adjunct:

(46) *anwar* Anwar *meħʷe-ja* in.Mehweb-gen *uškuj-ħe-w* school.obl-in-m(ess) *gʷa* asrt *učitel.* teacher 'Anwar is a teacher at the Mehweb school.'

Thus, the assertive marker need not follow the predicate. At the same time, it is not obvious that *gʷa* always follows the focus. For instance, in the elicited dialog (47), *gʷa* is attached to the first part of the clause 'Shamil is a singer', while its focus is constituted by its second part. In answers to content questions, *gʷa* is by default attached to the part of the utterance which does not contain new information, as in (48) and (49).


'Who is the head of Mehweb? – The head of Mehweb is Israpil.'

(49) *israpil* Israpil *či-ja?* who-q *– israpil* Israpil *gʷa* asrt *meħʷe-la* in.Mehweb-gen *χʷalajli.* chief 'Who is Israpil? – Israpil is the head of Mehweb.'

Thus, we find that, in non-verbal predications as well as in verbal predications, the assertive copula does not necessarily follow the predicate and the focused element.

Yury Lander

### **5 Conclusion**

To sum up, the assertive marker *gʷa* has the distribution of a copula (though lacking non-finite forms which are available for the copula), but its position does not fit into the picture that is usually documented in East Caucasian languages in that it does not need to be adjacent to the predicate or focus. At the same time, we observe some constraints on its distribution in complex constructions (in particular, its reluctance to syntactic islands). I conclude that more research is needed both to approach the functions of *gʷa* and to understand the principles that govern its syntactic position.

Further, it seems that our assumed knowledge of the principles regarding other kinds of predicative markers is overestimated. Indeed, while the idea of focusdetermined positions of copulas is important for East Caucasian, I am aware of no detailed corpus-based study of the position of predicative markers for any language of the family. Given the fact that during the last years the amount of corpora of East Caucasian languages has been increasing, one may hope that such studies will soon appear.

Moreover, as I emphasized in §2, predicative markers differ in their behavior, both within a single language and cross-linguistically. For East Caucasian, we need a more elaborated intragenetic typology of predicative markers. The present paper is to be considered a contribution to this line of investigation.

### **Acknowledgements**

I am grateful to all my consultants in Mehweb for their patience and to Michael Daniel, Nina Dobrushina and two anonymous reviewers for their useful comments on earlier versions of the paper.

### **List of abbreviations**



#### Yury Lander

### **References**


*toka, Jugo-Vostočnoj Azii i Zapadnoj Afriki (Moskva, 22–24 sentjabrja 2009): Tezisy i doklady*, 242–257. Moscow: Kluch-C.


## **Chapter 13**

## **Maps of Mehweb**

## Yuri Koryakov

Institute of Linguistics RAS

This section presents two maps of Mehweb, one showing the spread of the East Caucasian language family, together with Mehweb's closest relatives and its location with respect to the other languages of the Dargwa branch; the other showing the location of the linguistic family at a larger scale of the mountains of the Caucasus.

Figure 1: Mehweb on the map of East Caucasian languages

Yuri Koryakov. 2019. Maps of Mehweb. In Michael Daniel, Nina Dobrushina & Dmitry Ganenkov (eds.), *The Mehweb language: Essays on phonology, morphology and syntax*, 333–334. Berlin: Language Science Press. DOI:10.5281/zenodo.3402078

Figure 2: Mehweb on the map of North-East Caucasus

## **Name index**

Ageeva, Ekaterina, 100, 168, 169, 179, 183, 222 Al-Zaghir, Zainab Marzouk, 308 Anand, Pranav, 212 Ariel, Mira, 309 Arkadiev, Peter, 327

Bailey, Guy, 7 Barylnikova, Daria, 10, 97, 297 Belyaev, Oleg, 140, 145, 159, 160, 316 Bhatt, Rajesh, 211 Bošković, Željko, 308

Campbell, Lyle, 168 Charachidzé, Georges, 160 Chechuro, Ilya, 20, 22, 176, 190, 193 Clements, George N., 290 Cole, Peter, 291 Comrie, Bernard, 5, 90, 167, 296, 297, 300 Creissels, Denis, 76, 161, 202, 238, 262, 267 Csató, Éva Á., 299, 308 Cukor‐Avila, Patricia, 7 Daniel, Michael, 13, 39, 90, 103, 106,

120, 127, 128, 134, 153, 168, 169, 195, 198, 222, 242, 243, 256, 297, 299, 300, 307, 318, 320, 323 De Vries, Mark, 295 Dietrich, Rainer, 308

Dobrushina, Nina, iii, 7, 9, 87, 91, 92, 101, 103, 106, 125, 133, 138, 141, 145, 148, 159, 161, 242–244, 267, 316

Erteschik-Shir, Nomi, 308, 309 Evans, Nicholas, 145

Forker, Diana, 76, 140, 147, 148, 227, 228, 316, 323

Gagliardi, Annie, 227, 323 Ganenkov, Dmitry, 103–105, 297, 323 Goldsmith, John A., 34

Haiman, John, 147 Harris, Alice C., 168, 316, 317 Haspelmath, Martin, 41, 103, 235– 237, 246, 255, 262

Johanson, Lars, 299

Kalinina, Elena, 316 Kawachi, Kazuhiro, 308 Kazenin, Konstantin, 268, 316 Keenan, Edward L., 296, 300, 308 Kerslake, Celia J., 308 Khajdakov, Said, iii, 1, 2, 4, 117, 301 Khalilova, Zaira, 131 Kibrik, Aleksandr, v, 57, 106, 161, 262, 297 Kiparsky, Paul, 31 Kisriev, Enver, 2 Komarov, A. V., 4

#### Name index

König, Ekkehard, 271, 273 Koryakov, Yuri, iii, 1, 4 Kozhukhar, Aleksandra, 10, 156, 215, 216, 289, 290, 303 Kulikov, Leonid, 167, 169 Kustova, Marina, 127, 237 Ladefoged, Peter, 18 Lander, Yury, iii, 2, 55, 105, 123, 140, 148, 159, 160, 247, 289, 290, 297–300, 307, 308, 316–318, 327 Lee, Sun-Hee, 299 Lehmann, Christian, 296 Liao, Hsiu-Chuan, 296 Lyutikova, Ekaterina, 296, 308 Maddieson, Ian, 18 Magomedkhanov, Magomedkhan, 2 Magometov, Aleksandr, iii, iv,1, 2,18, 41, 42, 44, 47, 63, 73,105,108, 109, 117, 240, 244, 256, 258, 272, 276, 288, 301, 316 Maisak, Timur, 85, 106, 146, 147, 153, 243 Matsumoto, Yoshiko, 296 Mauri, Caterina, 146 Maxwell, Daniel N., 300 Merdanova, Solmaz, 106, 148 Moroz, George, 24, 41, 56, 69, 80, 108, 111, 127, 128, 256 Mutalov, Rasul,147,159,160, 299, 323 Nedjalkov, Vladimir, 167, 236 Nevins, Andrew, 212 Nichols, Johanna, 278 Nikitina, Tatiana, 212 Palmer, Frank R., 246 Partee, Barbara H., 327

Polinsky, Maria, 297 Reinhart, Tanya, 191 Reuland, Eric J, 290 Ross, John R., 296 Rubin, E., 57 Sansò, Andrea, 146 Schlenker, Philippe, 212 Sells, Peter, 273, 290 Sheyanova, Maria, 77, 87, 89, 145, 255 Shklovsky, Kirill, 212 Silnitsky, Georgij, 167 Sims, Andrea D., 41 Song, Jae Jung, 170 Sudo, Yasutada, 212 Sumbatova, Nina, iii, 1, 2, 73, 117, 123, 140, 147, 148, 159, 160, 247, 297, 299, 316, 318, 323 Tatevosov, Sergei, 299 Testelets, Yakov, 268, 273, 299, 316 Tishkov, Valentin, 2 Toldova, Svetlana, 273, 290, 299 Uchturpani, Muzappar A., 299, 308 Uslar, Petr, iii, 4 van den Berg, Helma E., 140, 159, 160, 247 van Nice, Kathy Y., 308 Wixman, Ronald, 5

## **Language index**

Agul, 106 Akhvakh, 76 Arabic, 308 Archi, 106, 267 Avar, 2, 3, 5, 6, 9–12, 76 Bagwalal, 299 Bezhta, 297 Bulgarian, 308 Chirag, 106 Dargwa, 13 Godoberi, 299 Hebrew, 308 Hinuq, 297 Itsari Dargwa, 299 Korean, 299<sup>7</sup> Lak, 2, 5, 6, 10–12, 316<sup>1</sup> Russian, 2, 3, 11 Serbo-Croatian, 308 Sidaama, 308 Tanti Dargwa, 123, 297, 298, 300, 303, 305, 306, 316<sup>1</sup> , 318<sup>5</sup> Tsakhur, 299, 308 Tsez, 297 Turkish, 299<sup>7</sup> , 308<sup>13</sup>

Udi, 316<sup>1</sup> , 317, 318 Uyghur, 299<sup>7</sup> , 308

## **Subject index**

abstract nouns, 74 accusativity, 75, 91 address, 123, 128, 143 agreement slot, 78, 106 andative, 95, 96 animacy, 305–307, 309 antipassive, 104 aorist, 87 apprehensive, 154–158 argument sharing, 259–262 aspectual stem, 77–79 atelic, 77 attributive suffix, 298<sup>6</sup> , 300, 301 auxiliary, 92–93 biabsolutive construction, 105, 302<sup>12</sup> 323 Bukhty, 6 causative, 97–99, 101, 109 causative, double, 100, 101, 183 causative, lexical, 105 causative, periphrastic, 169–177 causative, permissive, 174–175 caused motion verbs, *see* motion verbs causee, 177 census, 3 complex verbs, 106–111 concession, 137, 244–245 conditional, 146–150, 242–245 conjugation classes, 79–82

,

converb, 145, 151–152 converb, general, 87 converb, imperfective, 301<sup>9</sup> converb, neutral, 319, 320 converb, periphrastic, 257–258 converb, specialized, 89 coordinate structure constraint, 304– 305 coordination, 262–268 copula, 301, 318, 319, 320<sup>7</sup> , 321<sup>8</sup> , 323<sup>10</sup> counterfactual, 87, 146, 149–152 disjunct vs. conjunct forms, *see* egophoricity distributive quantifier, 327 ditransitive verbs, 180 durative, 92 egophoricity, 75–77 embedding, 263–264 endogamy, 5 ethnicity, 3 existential predication, 317<sup>2</sup> , 325 experiential verbs, 90, 104, 120, 179– 180 expression of wish, 137–146, 152 feminine, unmarried women, 74 focus, 317–319, 321, 322, 324, 327 future, 92 gender, 74–75, 306 grammatical case, 39

#### Subject index

habitual, 258 hortative, 130–133 hypothetical conditional, 146, 152– 153 identificational predication, 317<sup>2</sup> imperative, 89–92, 96, 102, 118–127, 266–267 imperative in clause chaining, 127, 133 imperative interjection, 91, 129 imperative particle, 125–127 imperative, plural, 89, 91, 122, 128, 131, 132 imperfective, 77–79, 87, 153 imperfective past, 87 infinitive, 89–92, 320 infixation, 79 insubordination, 145, 150 intensifier, 299, 303, 304, 306 irrealis, 87, 146–154 irregular verbs, 23, 78, 93–96 jussive, 121, 133–138 Kamakhal, 5 kindergarten, 6 Kumukh, 5, 10 labialization, 18<sup>1</sup> , 27, 28, 79–83 labile verbs, 90, 101, 104, 119 language loss, 2 lingua franca, 12 literacy, 2, 13 local domain, 274–275 logophoric pronoun, 273–285, 299, 303 market, 5, 10 mass nouns, 74

motion verbs, 91, 93, 107, 108, 110 motion verbs, caused motion, 96,104 Mugi, 4 Mukar, 5 multilingualism, 6, 9–12 multilingualism, asymmetrical, 11 negation, *see* polarity nominalization, 89, 93 Obokh, 5, 9, 11, 12 optative, 137–146 Palisma, 5 participle, 87, 301, 320 past, 144 perfective, 77–79, 87, 92, 153 periphrastic paradigm, 92 pharyngealization, 17, 18<sup>2</sup> , 19, 19<sup>3</sup> , 20, 21, 32–35, 80 plural, 41–55 plural imperative, *see* imperative, plural polarity, 82–86, 107, 132, 135, 181–182 polarity, negative volitional forms, 82–86 possessive domain, 275–277 possessive predication, 317<sup>2</sup> , 325 predicate topicalization, 147 predicative marker, 316–318 preverbs, 75 progressive, 92, 258 prohibitive, 127–129 Quran, 10 reason clauses, 153 recipient, 90 reduplication, *see* stem copy

Subject index

reflexive pronoun, 273–285, 299, 303, 307, 326 relativization, 264–265 reported speaker, 279 resultative, 92, 258 resumptive pronoun, 289–290, 299, 300, 303–308 school, 2, 6, 12 Shangoda, 5, 9, 11 Shitlib, 6 sound alternation, 27–32 stem copy, 85, 86, 146 stress, 17, 24–26, 35 subject, 278, 282–285 subordination, 262–268 syllable structure, 17, 20–22, 22<sup>4</sup> , 23, 24, 24<sup>5</sup> , 25, 28, 32, 34, 35 syntactic island, 298, 323–324 taxis, 265–266 tense, 265–266 topicalization, 146, 147 transitivity, 89, 97–106, 118, 122, 128, 131, 141, 180, 261 truncation, 140

#### Uri, 5

ventive, 95, 96 verificative, 105 vowel deletion, 29–32, 35

# 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.

## The Mehweb language

This book is an investigation into the grammar of Mehweb (Dargwa, East Caucasian also known as Nakh-Daghestanian) based on several years of team fieldwork. Mehweb is spoken in one village community in Daghestan, Russia, with a population of some 800 people, In many ways, Mehweb is a typical East Caucasian language: it has a rich inventory of consonants; an extensive system of spatial forms in nouns and converbs and volitional forms in verbs; pervasive gender-number agreement; and ergative alignment in case marking and in gender agreement. It is also a typical language of the Dargwa branch, with symmetrical verb inflection in the imperfective and perfective paradigm and extensive use of spatial encoding for experiencers. Although Mehweb is clearly close to the northern varieties of Dargwa, it has been long isolated from the main body of Dargwa varieties by speakers of Avar and Lak. As a result of both independent internal evolution and contact with its neighbours, Mehweb developed some deviant properties, including accusatively aligned egophoric agreement, a split in the feminine gender, and the typologically rare grammatical categories of verificative and apprehensive. But most importantly, Mehweb is where our friends live.