Explorations in Ethiopian Linguistics: Complex Predicates, Finiteness and Interrogativity [1 ed.] 3447102144, 9783447102148

Table of contents :
Contents
Figures, Maps And Tables
Preface
Question About Amharic Questions With
Interrogativity In Baskeet
Complex Predicates In Amharic Counterfactual
Complex Predicates In Zargulla
Grammaticalization Of Existential Auxiliaries
Benefactive Applicative Periphrases
Multiple Exponence
Transversal South Ethio-Semitic Languages
The Asymmetry Of Verbal Markedness
The Finite–Infinite Dichotomy In A
Finiteness In Gurage Languages
Case Marking In Amharic
Wandering Along The Border Of Finiteness:

Citation preview

Abhandlungen für die Kunde des Morgenlandes Band 91

Explorations in Ethiopian Linguistics: Complex Predicates, Finiteness and Interrogativity Edited by Ronny Meyer, Yvonne Treis and Azeb Amha

Deutsche Morgenländische Gesellschaft Harrassowitz Verlag

Abhandlungen für die Kunde des Morgenlandes Im Auftrag der Deutschen Morgenländischen Gesellschaft herausgegeben von Florian C. Reiter Band 91 Board of Advisers: Christian Bauer (Berlin) Desmond Durkin-Meisterernst (Berlin) Lutz Edzard (Oslo/Erlangen) Sebastian Günther (Göttingen) Jürgen Hanneder (Marburg) Herrmann Jungraithmayr (Marburg) Frank Kammerzell (Berlin) Karénina Kollmar-Paulenz (Bern) Jens Peter Laut (Göttingen) Michael Streck (Leipzig)

2014

Harrassowitz Verlag . Wiesbaden

Explorations in Ethiopian Linguistics: Complex Predicates, Finiteness and Interrogativity Edited by Ronny Meyer, Yvonne Treis and Azeb Amha

2014

Harrassowitz Verlag . Wiesbaden

Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.dnb.de abrufbar. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the internet at http://dnb.dnb.de.

For further information about our publishing program consult our website http://www.harrassowitz-verlag.de © Deutsche Morgenländische Gesellschaft 2014 This work, including all of its parts, is protected by copyright. Any use beyond the limits of copyright law without the permission of the publisher is forbidden and subject to penalty. This applies particularly to reproductions, translations, microfilms and storage and processing in electronic systems. Printed on permanent/durable paper. Printing and binding: Hubert & Co., Göttingen Printed in Germany ISSN 0567-4980 ISBN 978-3-447-10214-8 e-ISBN 978-3-447-19046-6

Contents List of Figures, Maps and Tables ................................................................

7

Preface Ronny Meyer & Yvonne Treis ................................................................

9

Interrogativity in Ethiopian Languages Question about Amharic Questions with yəhon: A Tentative Semantic Study Magdalena Krzyżanowska....................................................................... 17 Interrogativity in Baskeet Yvonne Treis ........................................................................................... 41

Complex Predicates in Ethiopian Languages Complex Predicates in Amharic Counterfactual Antecedent Clauses Abdu Ahmed............................................................................................ 79 Complex Predicates in Zargulla Azeb Amha .............................................................................................. 91 Grammaticalization of Existential Auxiliaries in Koorete Binyam Sisay Mendisu............................................................................ 121 Benefactive Applicative Periphrases with yɨw- ‘give’ in Xamtanga Chloé Darmon.......................................................................................... 137

Finiteness in Ethiopian Languages Multiple Exponence in the Long Prefix Conjugation of the Transversal South Ethio-Semitic Languages Maria Bulakh ........................................................................................... 149 The Asymmetry of Verbal Markedness in Libido Joachim Crass .......................................................................................... 179 The Finite–Infinite Dichotomy in a Comparative Semitic Perspective Lutz Edzard.............................................................................................. 205 Finiteness in Gurage Languages Ronny Meyer ........................................................................................... 225

6

Table of Contents

Case Marking in Amharic Copular Constructions Mulusew Asratie.......................................................................................259 Wandering along the Border of Finiteness: The Gəʿəz and Tigrinya Converb(s) in a Diachronic Perspective Stefan Weninger .......................................................................................283 Language Index ............................................................................................297

Figures, Maps and Tables Figures Figure 1: Types of Multiple Exponence ...................................................... 176
Figure 2: Genetic Classification of the Gurage Languages......................... 226


Maps Map 1: Approximate Location of Cited Ethiopian Languages.................... 9
Map 2: Location of the Basketo Special Woreda ........................................ 43
Map 3: Geographical Distribution of the Gurage Languages...................... 226
Map 4: Distribution of ICL Markers ........................................................... 252


Tables Table 1: Baskeet Consonant Phonemes....................................................... 44
Table 2: Declarative and Interrogative Perfective Paradigms (sool- ‘tell’). 47
Table 3: Declarative and Interrogative Jussive Paradigms (sool- ‘tell’) ..... 48
Table 4: Imperative and Hortative/Jussive .................................................. 183
Table 5: Declarative Verbs .......................................................................... 184
Table 6: APOD.HYP-Verb ............................................................................. 185
Table 7: Relative Verbs ............................................................................... 186
Table 8: The Three Narrative Converbs ...................................................... 187
Table 9: Specialized Converbs with Three Morphemes.............................. 189
Table 10: Specialized Converbs with Five Morphemes .............................. 190
Table 11: Main Verbs and Converbs with Five Morphemes....................... 191
Table 12: Main Verbs and Converbs with Three Morphemes .................... 192
Table 13: Declarative Verbs, Relative Verbs and la-Converb .................... 193
Table 14: The Three Types of Minus-Asymmetry between Declarative and Subordinate Verbs ............................................................... 194
Table 15: Affirmative and Negative PURP-Converb.................................... 197
Table 16: Affirmative and Negative HYP-Converb ..................................... 198
Table 17: Affirmative and Negative ANT-Converb ..................................... 199
Table 18: Symmetry vs. Asymmetry of Affirmative and Negative Main Verbs .................................................................. 200 Table 19: Asymmetry of Affirmative and Negative Subordinate Verbs..... 200
Table 20: Verbal Adjective/Stative in Akkadian......................................... 206
Table 21: Participle-Based Forms in Ṭuroyo .............................................. 207


8

List of Figures, Maps and Tables

Table 22: Converb/Gerund in Gəʿəz and Amharic ......................................208
Table 23: Converbs in Cushitic ....................................................................215
Table 24: Formation of Type A Verbs in Mesqan .......................................229
Table 25: Conjugational Templates of Type A Verbs .................................230
Table 26: Formation of Verbal Nouns .........................................................230
Table 27: Suffix Subject Indexes .................................................................231
Table 28: Pre-/Circumfix Subject Indexes ...................................................232
Table 29: Perfective Verbs in Main Clauses ................................................236
Table 30: Imperfective ICL Verbs in the Present Tense ..............................237
Table 31: Additional Morphemes with Imperfective Verbs in ICL .............241
Table 32: Relative Clause Verbs ..................................................................247
Table 33: Converbs in Gurage......................................................................249
Table 34: Compulsory Clause Marking in Gurage ......................................250
Table 35: Infinitives and Converbs ..............................................................288
Table 36: Suffixes of the Gəʿəz Converb Compared with Possessive Pronouns ............................................................289

Preface Ronny Meyer & Yvonne Treis The present volume brings together twelve contributions reflecting recent research trends in the description of Ethiopian languages. All contributions published in this volume deal with linguistic problems at the interface of morphosyntax and semantics/pragmatics; more precisely, they are analyses of interrogativity, complex predicates and finiteness in Ethiopian languages. Map 1: Approximate Location of Cited Ethiopian Languages

Most of the contributions are based on research presented in the linguistic panels at the 18th International Conference of Ethiopian Studies, held on 29 October – 02 November 2012 in Dire Dawa, Ethiopia. The multi-disciplinary conference was jointly organized by the French Center for Ethiopian Studies

10

Ronny Meyer & Yvonne Treis

(Addis Ababa) and the Institute of Ethiopian Studies (Addis Ababa University) under the motto Movements in Ethiopia, Ethiopia in Movement. Seven contributions are concerned with Ethio-Semitic languages, in particular with Amharic, Argobba, Gəʿəz, Harari, Tigrinya and different Gurage languages. Two contributions deal specifically with the Cushitic languages Xamtanga and Libido, while the remaining three articles focus on the Omotic languages Baskeet, Koorete and Zargulla. Furthermore, the Cushitic languages Bilin and Sidaama, the Omotic languages Haro, Maale, and Wolaitta, as well as various Semitic languages outside Ethiopia are cited in some contributions for comparative purposes. The approximate location of these languages is indicated on Map 1. The first section of this volume contains two articles from the panel on Interrogativity. Interrogativity has hitherto often only been handled in a very cursory manner in the grammatical descriptions of Ethiopian languages. Indepth analyses of this grammatical domain could, however, contribute interesting details to intonation research, language typology, and areal and historical linguistics. Omotic languages, many of which are hitherto little known, mark mood inflectionally and thus possess interrogative verbal morphology. Furthermore, the study of the morphology of interrogative pronouns and their diachronic origin can reveal interesting historical insights for the classification of individual languages or language groups. It also remains to be examined how widespread the use of interrogative pronouns plus an additive focus morpheme (similar to English ‘even’) in the function of indefinite pronouns is attested in Ethiopia. This phenomenon may well turn out to be an areal feature. The following two contributions take a first step towards a detailed examination of issues related to interrogativity in two selected Ethiopian languages. Yvonne Treis’ contribution on Interrogativity in Baskeet, a language of the Omotic family, is an in-depth study of morphological, syntactic and pragmatic aspects of interrogativity based on a corpus of recordings of natural speech events. Questions in Baskeet are either marked by intonation only, by an additional interrogative morpheme -a, or by a special interrogative verb form. The article discusses, among others, the form and function of the six simplex interrogative pronouns in Baskeet, namely PERSON, THING, TIME, QUANTITY, MANNER, PLACE, and SELECTION, and pronouns derived from them. Furthermore, these pronouns are compared with equivalents in related languages from the Ometo branch of Omotic. Finally, Treis demonstrates the use of interrogative pronouns in non-interrogative contexts in Baskeet. The second contribution on interrogativity, Question about Amharic Questions with yəhon: A Tentative Semantic Study by Magdalena Krzyża-

Preface

11

nowska, is concerned with the Ethio-Semitic language Amharic. It provides a semantic analysis of a special type of Amharic content and polar questions, namely those with the modal epistemic auxiliary yəhon ‘might’. Krzyżanowska’s analysis is based on a model developed by Andrzej Bogusławski and other Polish linguists. Accordingly, Amharic questions with yəhon are analyzed as consisting of three main components: the theme (i.e. a declarative sentence underlying the question), the rheme (i.e. the interrogative intention marked by prosody), and an additional predication on the theme conveyed by the auxiliary yəhon. It is shown that yəhon conveys epistemic modality de re in content questions, whereas it carries epistemic modality de dicto in polar questions. Polar questions with yəhon are then compared with declarative sentences marked by the auxiliary yəhonall ‘might’, and the additional pragmatic information conveyed by the particles mənalbat ‘perhaps’ and ənde ‘really’ in questions with yəhon is discussed. Finally, the use of yəhon as a marker of politeness is shown. The second section of the volume assembles contributions dealing with Complex Predicates, which attracted considerable theoretical interest over the past years. Here Ethiopian languages are of particular interest because of the diversity of complex predicate structures attested in different linguistic families in the country. The two most frequently mentioned features of complex predicates in the linguistic literature (see, for instance, Mengistu, Baker & Harvey 2010) are (i) that they comprise of a sequence of verbal predicates that are structurally and semantically related, and (ii) that verbs in complex predicates typically share an argument as well as tense, aspect, modality and/or polarity. Commonly, periphrastic causatives, converb-plus-verb constructions, particle-plus-verb constructions and ideophone-plus-verb constructions are subsumed under the concept of complex predicate. The four contributions of the second section discuss empirical issues related to complex predicates in individual Ethiopian languages. They point out the semantic/pragmatic differences between complex predicates and their monoverbal counterparts, the morphosyntactic properties of the component verbs in complex predicates, and the differences between complex predicates on the one hand and multi-clausal constructions or clause chains on the other. Of particular interest are syntactic and morphological mechanisms for combining verbs in a complex predicate, ordering restrictions regarding the component verbs, selection criteria for combining verbs in complex predicates as well as ways to subordinate or nominalize complex predicates and to mark modality and polarity in clauses headed by complex predicates. Thus, different types of complex predicates can be identified in individual languages, and

12

Ronny Meyer & Yvonne Treis

historical processes such as grammaticalization as well as contact phenomenon can be examined. Abdu Ahmed’s contribution Complex Predicates in Amharic Counterfactual Antecedent Clauses is concerned with syntactic and semantic aspects of complex predicates in a single clause type. There are two canonical types of counterfactual antecedent clauses in Amharic: one is based on a monoverbal construction, the other on a complex predicate consisting of a converb and the auxiliary verb honä ‘be(come)’. In addition to counterfactuality, the latter clause type encodes uncertainty of the speaker. It is shown that the converb encodes reference to the subject (and, optionally, to objects or adjuncts) of the verbal event, while the main verb, honä ‘be(come)’, is only marked for agreement with an expletive subject. The main verb is also the carrier of information on tense, aspect and modality. The author draws a distinction between converb-plus-verb and coverb-plus-verb complex predicates, which differ morphologically but share syntactic features. As subordinate conjunctions or adverbs can intervene between converb/coverb and main verb, complex predicates in Amharic cannot be considered serial verb constructions, which is in agreement with Azeb Amha’s analysis of complex predicates in Zargulla. Azeb Amha identifies in her contribution, Complex Predicates in Zargulla, three types of complex predicates. Based on the first component verb, she distinguishes between complex predicates based on converbs, intensive verb stems and ideophones. While basically every verb could occur as first component verb, the number of verbs occurring as second component in complex predicates is restricted to eight verbs – most of them are motion verbs. Semantically, complex predicates encode specific aspects of a verbal event as well as the effect, manner or path of the event. Complex predicates in Zargulla always consist of two component verbs, whereby the first component itself may be represented by a complex predicate. Binyam Sisay Mendisu’s contribution, Grammaticalization of Existential Auxiliaries in Koorete, explores the grammaticalization processes undergone by the verbs ba- ‘disappear, not exist’ and yes- ‘live, exist’ in Koorete, an Omotic language. It is shown that several grammaticalization stages (as full verb, existential verb, auxiliary, and grammatical morpheme) exist simultaneously in the language. In the analysis of complex predicates, in which the morphemes under investigation are used as auxiliaries, it is observed that inflectional morphemes occur on both the auxiliary and the main verb. Thus, auxiliary verb constructions in Koorete are categorized as split-headed. In her contribution Benefactive Applicative Periphrases with yɨw- ‘give’ in Xamtanga, Chloé Darmon shows that the applicative in the Central

Preface

13

Cushitic language Xamtanga is based on a biverbal periphrastic construction, which licenses the expression of an additional beneficiary participant. The construction involves the converb form of a lexical verb followed by the valency operator yɨw- ‘give’. Based on a comparison of the characteristics of the benefactive converb-plus-finite verb constructions with other formally identical V1–V2 patterns, it is argued that the ‘give’-periphrasis in Xamtanga is a complex predicate. While give-benefactives may be common in other parts of the world, Xamtanga is so far the only known language of the Ethiopian linguistic area that has developed an applicative construction with ‘give’. The third section of this volume concentrates on issues of Finiteness. The finite/non-finite distinction is frequently made in grammars but there is no general consensus on their cross-linguistic definition (cf. Nikolaeva 2007). Often finiteness is considered to be a phrasal or verbal category – mainly defined through specific values for TAM and person marking – but sometimes it is regarded as a clausal category that also encompasses, among others, illocutionary force, i.e. markers of assertion, questions, etc. With regard to Ethiopian languages, the finite/non-finite distinction displays a high degree of linguistic variability. Even within the same linguistic sub-family, independent affirmative clauses may contain a verb only inflected for aspect or modality, and subject agreement, or they may require additional information on tense, focus, or illocutionary force. Furthermore, there are several unsettled debates on certain grammatical phenomena in Ethiopian languages in which finiteness plays a crucial role: Regarding Ethio-Semitic languages, it is, for instance, contested that they have converbs in the strict sense. In Highland East Cushitic, the function of the so-called “additional morphemes” on main clause verbs is still unclear. Six contributions give an overview about the complexity of the questions relating to finiteness in Ethiopian languages. Their general tenor is that finiteness is a scalar or relational category rather than a binary one. One-to-many correspondence between a grammatical meaning and its morphological realization within one word is the topic of Maria Bulakh’s contribution Multiple Exponence in the Long Prefix Conjugation of the Transversal South Ethio-Semitic Languages. She focuses on the analysis of “multiple exponence” for subject referencing on imperfective verbs in main clauses of Amharic, Argobba, Harari, Wolane, and Zay. In these languages, this verb form consists of a main verb in the imperfective aspect followed by the auxiliary verbs *hallawa or *nabara. The original analytic construction of this verb form is said to have had double marking of person, once on the main verb and once on the auxiliary verb. The common trend in today’s lan-

14

Ronny Meyer & Yvonne Treis

guages, however, is to eliminate one of the redundant morphemes, either throughout the entire paradigm or in certain persons only. Bulakh shows that the individual languages use different patterns for dealing with multiple exponence, i.e. the discussed languages do not equally tolerate multiple exponence. There is a general tendency to avoid repetition of formally identical morphemes and to tolerate double marking only if the two markers are formally not too similar. Joachim Crass discusses clausal finiteness and the mismatch between morphological markedness of affirmative and negative main and subordinate verbs in his contribution The Asymmetry of Verbal Markedness in Libido. Libido – a Highland East Cushitic language – has a fairly complex verbal morphology. For the description of the verbal system, Bisang’s (1998; 2001; 2007) concept of asymmetry is applied, which is concerned with the occurrence of obligatory categories in main and subordinate verbs. Miestamo’s (2005) typology of standard negation is used to describe and categorize the structural asymmetries between affirmative and negative forms of main and subordinate verbs. Lutz Edzard is concerned with the notion of finiteness from a general comparative Semitic perspective in his contribution The Finite–Infinite Dichotomy in a Comparative Semitic Perspective, in which he discusses data from Akkadian, Arabic, Aramaic, Biblical Hebrew and Ethio-Semitic. It is shown that a clear-cut dichotomy between the categories finite and infinite is problematic in the verbal domain of Semitic – notably as far as the EthioSemitic converb/gerund, the Akkadian stative/verbal adjective, and the conjugated noun in predicative position are concerned. In many Semitic languages, verbal nouns or infinitives can adopt a finite function pragmatically, notably as imperatives. Furthermore, the first element in serial verb constructions tends to lose the morphological features associated with finiteness and to be grammaticalized into an adverb. In Amharic, certain converbs have become lexicalized in a frozen form with default 3SG.M subject reference; they function synchronically as adverbs. In his contribution Finiteness in Gurage Languages, Ronny Meyer outlines the remarkable variation in marking verbs for finiteness in independent main clauses vis-à-vis dependent clauses in genetically closely related and geographically adjacent languages. The detailed analysis of inflectional features of verbs including TAM, subject indexing, polarity and clausal status, and a comparison of these features in dependent and independent clauses shows that finiteness in Gurage languages is a relational category of the clause that can best be defined through a combination of morphological and syntactic features.

Preface

15

Mulusew Asratie’s contribution Case Marking in Amharic Copular Constructions discusses the nominative/accusative alternation of NP and AP predicates in Amharic copular constructions from a generative perspective. Based on differences between the accusative and nominative predicates regarding their semantic interpretation, agreement and word order, it is argued that accusative predicates must be assigned by a functional element that introduces “eventivity”. Stefan Weninger studies the Ethio-Semitic converb from a comparative and diachronic perspective in his contribution Wandering along the Border of Finiteness: The Gəʿəz and Tigrinya Converb(s) in a Diachronic Perspective. He compares the Gəʿəz converb with its counterpart in modern Tigrinya, the closest modern relative of Gəʿəz. In some respects, the Tigrinya converb has lost features of finiteness still present in Gəʿəz. In other respects, it behaves more like a finite verb. Moreover, the modern cognates of the Gəʿəz converb have neither developed into a fully finite nor into a fully infinite verb form in any modern Ethio-Semitic language. Instead, converbs have remained a fuzzy category, which makes the author develop a scalar definition of finiteness. We have made no attempt to unify the transliteration and transcription systems across the individual contributions. Apart from IPA, representations in Fidäl (for some Ethio-Semitic languages) and specialized transliteration/transcription systems are used. The reader should, therefore, be aware that a single sound might have various representations across the contributions, but is usually consistently reproduced by the same symbol in individual contributions. The ejective velar plosive, for instance, is represented by the IPA symbol k’, or the symbols ḳ and q according to the conventions of the Deutschen Morgenländischen Gesellschaft (DMG) and the Encyclopaedia Aethiopica (EAE), respectively. Palatal consonants are marked by a haček in some contributions, i.e. IPA ʤ is represented by ǰ or ǧ. Regarding the vowels, the mid-central vowel is either given as ä or ǝ. Note, however, that the symbol ǝ can also represent the mid-high vowel in some contributions on EthioSemitic languages, which appears as ɨ elsewhere. Vowel length is either marked by a double vowel, e.g. aa, or by a macron above the vowel symbol, e.g. ā. All contributions were peer-reviewed by at least two academic referees. The review process for this volume was separate from that of the ICES conference. We are grateful to all referees for their constructive comments. We would also like to thank the CNRS laboratory LLACAN for providing financial support for proofreading and cartography. Finally, we express our gratitude to Lutz Edzard and the Zeitschrift der Deutschen Morgenländischen Gesellschaft (ZDMG), who kindly accepted to

16

Ronny Meyer & Yvonne Treis

publish this volume in the Abhandlungen für die Kunde des Morgenlandes (AKM), and to Michael Fröhlich from the Harrassowitz Publishing House for his generous assistance with all our editorial enquiries. References Bisang, Walter. 1998. The view from the Far East. In Johan van der Auwera (ed.), Adverbial Constructions in the Languages of Europe, 641– 812. Berlin & New York: Mouton de Gruyter. Bisang, Walter. 2001. Finite vs. non finite languages. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible (eds.), Sprachtypologie und sprachliche Universalien – Ein internationales Handbuch, vol. II, 1400–1413. Berlin & New York: De Gruyter. Bisang, Walter. 2007. Categories that make finiteness: Discreteness from a functional perspective and some of its repercussions. In Irina Nikolaeva (ed.), Finiteness: Theoretical and Empirical Foundations, 115–137. Oxford: Oxford University Press. Mengistu Amberber, Brett Baker & Mark Harvey. 2010. Introduction: Complex predicates. In Amberber Mengistu, Brett Baker & Mark Harvey (eds.), Complex Predicates: Cross-Linguistic Perspectives on Event Structure, 1–12. Cambridge: Cambridge University Press. Miestamo, Matti. 2005. Standard Negation: The Negation of Declarative Verbal Main Clauses in a Typological Perspective. Berlin & New York: Mouton de Gruyter. Nikolaeva, Irina (ed.). 2007. Finiteness: Theoretical and Empirical Foundations. Oxford: Oxford University Press.

Question about Amharic Questions with yəhon: A Tentative Semantic Study Magdalena Krzyżanowska Poznan University and Hamburg University Abstract This paper presents a tentative semantic analysis of Amharic content and polar questions with the modal epistemic auxiliary ይሆን yəhon ‘might’, known as deliberative or meditative questions. It begins with a review of linguistic works where questions with yəhon have been most comprehensively treated. Next, it briefly discusses the concept of thematic-rhematic structure of sentences, which is then applied to investigate the meaning of questions with yəhon. In these questions the following main components are indicated and described in detail: the theme, being a declarative sentence contained in each question; an additional predication on the theme conveyed by yəhon (the thematic dictum); and the rheme, carried largely by prosodic features. Special attention is devoted to elucidating the meaning of yəhon. It is shown that yəhon conveys epistemic modality de re in content questions as it scopes over the variable, whereas it carries epistemic modality de dicto in polar questions as it scopes over the whole declarative sentence. The meaning of polar questions with yəhon is compared with the meaning of declarative constructions involving the distinct epistemic modal auxiliary ይሆናል yəhonall ‘might’. Then, the meaning of the particles ምናልባት mənalbat ‘perhaps’ and እንዴ ənde ‘really’ accompanying questions with yəhon is briefly discussed. Finally, the use of yəhon as a marker of politeness is presented. 1 Introduction The aim of this paper is to propose a tentative semantic analysis of Amharic questions with the modal epistemic auxiliary ይሆን yəhon.1 It will focus on questions like the following: 1

I like to express my heartfelt gratitude to Orin Gensler for providing insightful advice and criticisms and for supporting me before and during my conference presentation. I thank him also for English proofreading of the last version of this article. Equally, I am greatly indebted to my Ethiopian friends Tə’gəst Bərhanu, Abrəham Aduňňa, Sälomon Gäbräyäs and Gidäna Mäsfən for their invaluable help and for their patience.

18 (1)

Magdalena Krzyżanowska ሰላማዊት sälamawit

ነገ nägä

ትመጣ tə-məṭa

ይሆን? yəhon?

Sälamawit

tomorrow

3SG.F-come\IPFV

AUX.Q

‘Might Sälamawit be coming tomorrow?’ (2)

አሉላን alula-n

ማን man

ረድቶት rädət-o-t

ይሆን? yəhon?

Alula-ACC

who

help\CNV-3SG.M-OJ.3SG.M

AUX.Q

‘Who might have helped Alula?’ The basic conceptual apparatus which will be applied in the analysis is that elaborated by Bogusławski (1977). The main thrust of his approach is to investigate the meaning of questions by describing their thematic-rhematic structure. Since each type of sentence has its own particular thematic-rhematic structure, this structure can be used to describe questions of all types with all their characteristic features.2 Additionally, the present paper is to some extent inspired by Danielewiczowa (1996) where she applies Bogusławski’s concept to examine the thematic-rhematic structure of questions in Polish. Her work was especially instructive in pulling apart the various components of the structure, a task which is particularly difficult for questions. Our analysis will aim at tentatively indicating the main components of the thematic-rhematic structure of questions with yəhon and at establishing the meaning that each component contributes. Special attention will be devoted to elucidating the meaning of the auxiliary yəhon, which is crucial for distinguishing yəhon questions, known as deliberative or meditative, from other types of questions (in particular those with non-standard semantics such as rhetorical questions, riddles and exam questions) and from declarative constructions involving the epistemic modal auxiliary yəhonall. Also, it will be shown that the kind of epistemic modality that yəhon conveys depends on the type of question in which it is contained, i.e. de re in content questions and de dicto in polar questions. The semantic category of epistemic modality will be conceptualized in terms of a scale which involves different degrees of the speaker’s certainty in regard to the truth-value of the sentence.3 Since the 2 3

On the analysis of questions as kinds of statements and directives, see Danielewiczowa (1996:10 ff.). For epistemic modality treated in terms of the degree of the speaker’s certainty or similar notions (commitment, confidence), see e.g. Lyons (1977), Rytel (1982), Coates (1983), Givón (1994) and Nuyts (2001).

Question about Amharic Questions with yǝhon

19

scale is continuous the modal epistemic meanings may be delimitated only roughly as high, medium and low degree of certainty. The auxiliaries yəhon and yəhonall convey low degree of certainty. Finer differences in meanings between pairs of sentences will be discussed in terms of having greater, lower or equal degree of certainty. Morphologically, yəhon is the 3SG, masculine, simple imperfective form of the verb ሆነ honä ‘become’. The verb yəhon has undergone a grammaticalization process here and has come to serve as a frozen interrogative auxiliary which always follows the main verb (or the predicate nominal in a nominal sentence). It is completely invariant; there is no təhon ‘may you (SG.M)/she become’, etc. The next part of the article, Section 2, presents how the semantics of questions with yəhon has been dealt with in three previous works on Amharic grammar. It is followed by Section 3 containing a brief discussion of the thematic-rhematic structure of sentences, with a special focus on interrogatives. The next lengthy Section 4 treats the main individual components of the thematic-rhematic structure of questions with yəhon, i.e. theme, thematic dictum and rheme. These are described separately for content and polar questions. Section 5 discusses how the particles mənalbat ‘perhaps’ and ənde ‘really’ contribute to the meaning of questions. Section 6 presents the use of yəhon as a politeness marker. The last Section 7 contains the conclusion. The analysis will be illustrated with sentences representing both spoken and written Amharic.4 Some sentences were constructed by the author together with Amharic speakers.5 2 Review of literature Thus far the semantics of Amharic questions has been examined only briefly and not systematically. In linguistic descriptions of the language, questions with yəhon have been given relatively (though not absolutely) more attention in three works: Cohen’s Traité de langue amharique (1936), Goldenberg’s The Amharic Tense-System (1966) and Leslau’s Reference Grammar of Amharic (1995). Let us touch upon each work separately. 4

5

The examples are taken from real conversations, radio programs and novels. References are given in the footnotes only for sentences taken from five novels and one radio play. The abbreviated titles of novels are followed by the page number and the line number. Four Amharic speakers participated as the informants in writing this paper: Tə’gəst Bərhanu, Abrəham Aduňňa, Sälomon Gäbräyäs and Gidäna Mäsfən. All of them speak Amharic as their first language. Gidäna Mäsfən speaks also Tigrinya as his first language. Sälomon Gäbräyäs speaks Oromo as his second language while Tə’gəst Bərhanu speaks Tigrinya.

20

Magdalena Krzyżanowska

Cohen (1936:382 f.) treats questions with yəhon in a section on hypothetical interrogative sentences. He understands them as questions conveying a meaning of open or incomplete “éventualité”, that is, possibility or eventuality. Goldenberg (1966:77) first mentions yəhon as an independent form, i.e. a copula preceded by a nominal, expressing ጥርጣሬ ṭərəṭṭare “doubt” and መደንገጥ mädängäṭ “unpleasant surprise”. Further on, in the chapter devoted to compound tenses, he discusses the gerund with yəhon referring to an event which took place before the moment of speaking (Goldenberg 1966:206) and the imperfective with yəhon referring to an event which is taking place at the moment of speaking or will take place in the future (Goldenberg 1966:247 ff.); both of these are used in questions and exclamations of “surprise and fear”, especially in monologues and in inner speech. In his Reference grammar of Amharic, Leslau (1995:312, 328, 382, 776) takes into consideration similar aspects of questions with yəhon. He indicates that yəhon may be used alone as the main verb or it may be preceded by an imperfective or a gerund verb form to express probability or possibility in a question. Thus, questions with yəhon have been noted by linguists and their basic meaning has been described, but not in detail. To the best of my knowledge, there has been no study devoted to the prosodic features of Amharic questions, with the exception of Alemayehu Haile’s (1990) analysis of the intonation contour of Amharic polar questions. 3 Introduction to the thematic-rhematic structure of questions The analysis of questions with yəhon will consist in describing their thematicrhematic structure. The thematic-rhematic structure refers to the division of a sentence into two main semantic parts called theme and rheme.6 Theme is the thing or concept which is being talked about in the sentence while rheme is what is being said about this thing. Only simple sentences have such a simple binary structure. For complex sentences with several arguments and/or several predicates the structure may consist of several themes and rhemes (Karolak 1999:493 f., 596 f.). Furthermore, there can be various additional predications which are connected with the theme and which cannot be negated; these are called thematic dicta or presuppositions (Danielewiczowa 1996:42). As the axis of any question, that is, as the question’s rheme, Bogusławski (1977:242) suggests the formula: “I want to become confronted with a true sentence about...”; then the predication expressed by the question follows, which constitutes the question’s theme. In other words, by asking a question 6

In other scholarly traditions these terms correspond roughly to topic and comment, psychological subject and psychological predicate, datum and novum, respectively.

Question about Amharic Questions with yǝhon

21

the sender (i.e. speaker or writer) expresses her/his wish to be confronted with one or more true sentences about the theme. The formula takes this very general shape so as to embrace as many types of questions as possible, i.e. standard content and polar questions which do not contain any additional presuppositional predication about the theme, as well as questions with such a predication, e.g. questions posed during exams, riddles, rhetorical questions and deliberative questions, such as Amharic questions with yəhon. Now, we shall indicate and describe the components of the thematic-rhematic structure of Amharic questions with yəhon. 4 Thematic-rhematic structure of questions with yəhon Questions with yəhon appear as both content (wh-) and polar (yes/no) questions. Their thematic-rhematic structure in its essentials does not differ from that of standard questions. What is decisive for semantically discerning questions with yəhon is the presence of an additional predication (the thematic dictum) about the theme conveyed by yəhon. Because of this, relatively much space in this section will be devoted to elucidating its meaning. The first component to be dealt with is the theme, followed by the thematic dictum. Finally, the rheme will be discussed. 4.1 Theme of questions with yəhon The theme of any question, called the datum questionis, is a declarative sentence contained in the question. For instance, for question (3) the datum questionis is sentence (4): (3)

ይህንን yəh-ənnən

ቸኮሌት čäkolet

ሠናይት sännayt

አምጥታልኝ amṭət-a-ll-əňň

ይሆን? yəhon?

this-ACC

chocolate

Sännayt

bring\CNV-3SG.F-BEN-OJ.1SG

AUX.Q

‘Might Sännayt have brought me this chocolate?’ (4)

ይህንን yəh-ənnən

ቸኮሌት čäkolet

ሠናይት sännayt

አምጥታልኛለች amṭət-a-ll-əňň-all-äčč

this-ACC

chocolate

Sännayt

bring\CNV-3SG.F-BEN-OJ.1SGAUX.NPST-3SG.F

‘Sännayt has brought me this chocolate.’ The sender does not know whether sentence (4) is true or false and s/he wants to verify it. This is a polar (yes/no) question. Question (5) is a content question.

22 (5)

Magdalena Krzyżanowska ምን mən

ችግር čəggər

ገጥሟት gäṭm-o-at

ይሆን? yəhon?

what

problem

happen\CNV-3SG.M-OJ.3SG.F

AUX.Q

‘What problem might have happened to her?’ For this question the datum questionis is sentence (6): (6)

የሆነ yä-hon-ä

ችግር čəggər

ገጥሟታል gäṭm-o-at-all

REL-be\PFV-3SG.M

problem

happen\CNV-3SG.M-OJ.3SG.F-AUX.NPST

‘Some problem has happened to her.’ This sentence is considered to be true, that is, it is presupposed. This is a salient difference between content questions and polar questions: the datum questionis is presupposed in the former but not the latter. 4.2 Thematic dictum of questions with yəhon It should be noted that the datum questionis of (3) and (5) does not include yəhon. This is because yəhon conveys an additional predication about the theme. For expressing an event which may be taking place at the moment of speaking or may take place in the future, the main verb7 occurs in its simple imperfective form, e.g.: (7)

አልማዝ almaz

መቼ mäče

ትመጣ tə-mäṭa

ይሆን? yəhon?

Almaz

when

3SG.F-come\IPFV

AUX.Q

‘When might Almaz be coming?’ An event which might have taken place before the moment of speaking is indicated by the converb form, e.g.:

7

In my analysis of the meaning of yəhon, nominal (non-verbal) sentences have not been taken into account. Some examples of such sentences are given only in the part dealing with prosody (cf. 30 and 31). It appears that in nominal sentences yəhon may be both a copula and an interrogative auxiliary at the same time like in (30). Yet yəhon may also occur separately, following the copula as in (31). There seems to be a difference in meaning between the sentences: (30) conveys higher degree of certainty than (31).

Question about Amharic Questions with yǝhon (8)

ብርጭቆውን bərč̣əqqo-w-ən

ማን man

ሰብሮ säbr-o

ይሆን? yəhon?

glass-DEF-ACC

who

break\CNV-3SG.M

AUX.Q

23

‘Who might have broken the glass?’ Questions with yəhon most often occur in monologues and inner speech, a fact which is tightly connected with their semantics. Apart from the sender there might be a listener but s/he is essentially irrelevant. The listener is not really the intended addressee and therefore the sender does not expect an answer from her/him. The answer possibly exists, but at the moment of speaking it is not available. If there is an addressee at all, it is the sender her/himself. That is why, although the sender poses the question, s/he does not really ask the question,8 in the normal sense of seeking information in the answer and nothing more. Rather, s/he invites her/himself (and possibly her/his listener) to deliberate on the answer, which (s/he thinks) is somehow difficult to arrive at. In what follows we shall try to elucidate how questions with yəhon become deliberative questions. 4.2.1 Meaning of yəhon in content questions The datum questionis for question (9) is the declarative sentence (10) which contains a variable ‘someone’: (9)

አሉላን alula-n

ማን man

ረድቶት rädət-o-t

ይሆን? yəhon?

Alula-ACC

who

help\CNV-3SG.M-OJ.3SG.M

AUX.Q

‘Who might have helped Alula?’ (10)

(እገሌ)ይሆን (əgäle)yəhon

አሉላን alula-n

ረዳው rädda-w

(someone)AUX.Q

Alula-ACC

help\PFV.3SG.M-OJ.3SG.M

‘(Someone)yəhon has helped Alula.’ By asking the question, the sender is looking for a specific thing that s/he could substitute for the variable of the datum questionis, just as in standard questions without yəhon. The interrogative auxiliary yəhon scopes over the variable, bringing out a range of things which the sender could take into consideration as the true thing. Thus, in contrast to standard questions where 8

On the difference between asking and posing a question see Lyons (1977:753 ff.).

24

Magdalena Krzyżanowska

only one thing is taken into consideration as the true thing, in questions with yəhon there are more potential candidates which are taken into consideration as the true thing substitutable for the variable. In other words, the true thing appears against a background of other things; to reach the answer,9 the sender needs to sort these out.10 By way of illustration consider sentences (11) and (12) which are uttered in succession as the speaker’s proposed answer to question (9): (11)

ቶማስ tomas

ረድቶት rädət-o-t

ይሆናል yəhonall

Tomas

help\CNV-3SG.M-OJ.3SG.M

AUX.EMOD

‘Thomas might have helped him.’ (12)

አይ፤ ayy

እሱ əssu

አይደለም። aydällä-mm

ሰነፍ sänäf

ነው nä-w

no

he

NEG.COP.3SG.M-NEG

lazy

COP-3SG.M

‘No, it’s not him. He is lazy.’ (13)

አልማዝ almaz

ትሆናለች tə-hon-all-äčč

Almaz

3SG.F-be\IPFV-AUX.NPST-3SG.F

‘She might be Almaz.’ 4.2.2 Meaning of yəhon in polar questions In polar questions the sender presents the whole datum questionis as the object to be verified, i.e. s/he looks for the affirmation or for the negation of the truth of the datum questionis. However, polar questions with yəhon are enriched with an additional meaning which we shall present here. The datum questionis for (14) is the declarative sentence (15):

9 Sometimes the answer may not be the main aim as the sender knows that it does not exist or is unavailable at the moment of speaking. Even sorting out the things may only serve as a way to exclude those which appear as false. 10 I have adopted the basic concept for elucidating the meaning of Amharic content questions with yəhon from Bogusławski (1993), who examines Russian content questions with the verb možet ‘can’.

Question about Amharic Questions with yǝhon (14)

ሰላማዊት sälamawit

ነገ nägä

ትመጣ tə-mäṭa

ይሆን? yəhon?

Sälamawit

tomorrow

3SG.F-come\IPFV

AUX.Q

25

‘Might Sälamawit be coming tomorrow?’ (15)

(ሰላማዊት (sälamawit

ነገ nägä

ትመጣለች)yəhon tə-mäṭa-all-äčč)yəhon

(Sälamawit

tomorrow

3SG.F-come\IPFV-AUX.NPST-3SG.F)AUX.Q

‘(Sälamawit is coming tomorrow)yəhon’ The auxiliary yəhon scopes over the whole declarative sentence, indicating that it is only one of a range of sentences which are candidates for the true sentence. The sender does not know whether the datum questionis is verifiable at all. Yet, s/he poses the question because s/he does not exclude that the sentence might be true and thinks that it should be taken into consideration. To make this explanation clearer, let us consider other questions concerning Sälamawit that, in the same context, the sender might raise: (16)

ወይስ wäyss

ከነገ kä-nägä

ወዲያ wädiya

ትመጣ tə-mäṭa

ይሆን? yəhon?

or

from-tomorrow

beyond

3SG.F-come\IPFV

AUX.Q

‘Or might she be coming the day after tomorrow?’ (17)

አትመጣ a-t-mäṭa

ይሆን? yəhon?

NEG-3SG.F-come\IPFV

AUX.Q

‘Might she not be coming at all?’ One more example taken from a novel: (18)

ትርንጎ tərəngo

ትዝ təzz

አለችው። al-äčč-əw

በነፍስ bä-näfs

ትኖር tə-nor

ይሆን? yəhon?

Tərəngo

IDPH.

say\PFV-3SG.FOJ.3SG.M

in-soul

3SG.Fexist\IPFV

AUX.Q

remember

26

Magdalena Krzyżanowska ሞታ mot-a

ይሆን? yəhon?

አያውቅም a-y-awq-əmm

die\CNV-3SG.F

AUX.Q

NEG-3SG.M-know\IPFV-NEG

‘He remembered Tərəngo. Might she be alive? Might she have died? He does not know.’ (KB 13: 7–8) The deliberative or meditative character that questions with yəhon assume seems to result from the fact that the sender confronts her/himself with a range of things which s/he takes into consideration as possibly true and a range of sentences that are candidates for the true sentence. This is a sufficient condition for a question to become deliberative.11 Next may come the sender’s attempts to eliminate false candidates and select the true one. 4.2.3 Modal epistemic meanings of yəhon Whereas standard questions, as Bogusławski (1977:245) says, have “intellectual, non-egocentric and impartial […] character, especially, […] in comparison with other utterances expressing the will to learn something”, in questions with yəhon the sender highlights, basically, her/his lack of knowledge.12 In content questions with yəhon the sender does not know which one out of the range of candidate things s/he should substitute for the variable. In polar questions with yəhon the sender does not know whether the datum questionis is verifiable but s/he does not exclude that it might be true and thinks that it should be taken into consideration. Although both content and polar questions convey a modal epistemic meaning, as was already mentioned, there is a crucial difference in scope 11 Questions with yəhon cannot be identified with rhetorical questions, taken in the narrow sense as “questions which lead the addressee to understand the opposite” (Bussmann 1996:408). In fact, yəhon may occur in rhetorical questions as in the example below where a positive content question implies a negative assertion, i.e. ‘People will never get to know that life is holy.’ (i)

ሰዎች ሕይወት säw-očč həywät

ቅዱስ qəddus

መሆኑን mähon-u-n

የሚያውቁት yämm-iy-awq-u-t

መች mäč

man-PL

holy

be\VN-POSS. 3SG.M-ACC

REL-3PL-know\ IPFV-3PL-OJ.3SG.M

when AUX.Q

life

ይሆን? yəhon?

‘When might people get to know that life is holy?’ 12 One might argue that, in principle, we ask questions because we do not know something and we want to learn about it. However, ignorance is not a necessary component of questions. An examiner knows the answer but still poses a question which grammatically is in no way different from the same question asked by someone who does not know the answer. In light of this, questions with yəhon contain a clear and explicit signal of the sender’s relative ignorance.

Question about Amharic Questions with yǝhon

27

between them. In linguistics and logic this distinction can be conceptualized in terms of de re (Latin “about the thing”) modality and de dicto (Latin “about what is said”) modality.13 The auxiliary yəhon in content questions, whose scope ranges over variables within the datum questionis, conveys epistemic modality de re as it concerns things in the actual world, i.e. ‘Who/What is possibly X?’ In contrast, yəhon in polar questions, scoping over the whole datum questionis, carries epistemic modality de dicto as it concerns what is said about a state of affairs, i.e. ‘Is it possible that p?’ Thus, in this type of questions yəhon conveys a low degree of certainty in respect to the truth of the sentence. It is more problematic to assign the degree of certainty to yəhon in content questions since we cannot assign the truth-value to a non-sentential variable. To do so, we need to assume that even though yəhon scopes over the variable, it extends beyond it and embraces the whole sentence in which it is embedded. That means, the variable is taken together with what is predicated about it. Then, yəhon conveys a low degree of certainty in respect to the truth of the sentence containing a variable for which a range of candidate things may be substituted. 4.2.4 The auxiliary yəhon in comparison with yəhonall Beside yəhon there is a similar modal epistemic auxiliary ይሆናል yəhonall which occurs only in declarative sentences. From the morphological point of view, yəhonall is a standard compound imperfective form, composed (in the usual way) of the 3SG, masculine, simple imperfective form of the verb ሆነ honä ‘become’ and the 3SG, masculine, perfective form of the verb አለ allä ‘be present’. In the course of the grammaticalization process yəhonall has become a frozen form and, simultaneously, its function has shifted from lexical to purely grammatical. Thus, semantically, it conveys only the modal epistemic meaning whereas the main predication is expressed by the main verb that precedes it (and takes subject agreement). If we take into consideration its scope, yəhonall is more closely related to yəhon in polar questions than to yəhon in content questions. Consequently, it is more appropriate to compare its meaning with yəhon in polar questions. The essential difference between the two auxiliaries is that yəhon occurs exclusively in questions whereas yəhonall is used in statements. This implies that we may expect distinct intonation patterns for the two types of utter13 About de re and de dicto modalities in linguistics, see e.g. Lyons (1995:230 f.) and Chierchia & McConnell-Ginet (2000:308 f.). De re and de dicto modalities in logic are treated e.g. by Allwood et al. (1977:114 ff.) and Gamut (1991:45 ff.). Quine (1953) opposes this distinction.

28

Magdalena Krzyżanowska

ances. Besides, the meaning of yəhonall differs in at least two other points. First, the sender takes into consideration only one sentence at a time whose truth s/he is uncertain of and, therefore, the sentence with yəhonall does not assume a deliberative character. Secondly, and as a corollary of the previous point, the degree of certainty that it conveys is higher than in the case of polar questions with yəhon. Compare (19) vs. (20): (19)

ሰላማዊት sälamawit

ነገ nägä

ትመጣ tə-mäṭa

ይሆን? yəhon?

Sälamawit

tomorrow

3SG.F-come\IPFV

AUX.Q

‘Might Sälamawit be coming tomorrow?’ (20)

ሰላማዊት sälamawit

ነገ nägä

ትመጣ tə-mäṭa

ይሆናል yəhonall

Sälamawit

tomorrow

3SG.F-come\IPFV

AUX.EMOD

‘Sälamawit might be coming tomorrow.’ One may wonder whether both auxiliaries might perhaps be interpreted as conveying basically the same epistemic meaning of low degree of certainty. The difference in meaning between sentences in which they occur would be then ascribed to the type of utterance: question for yəhon and statement for yəhonall. 4.3 Rheme of questions with yəhon The rheme of questions, in general, comprises all those components which are responsible for interrogativity. For content questions interrogativity is conveyed in the first place by interrogative pronouns. They are aurally perceived as stressed14 (indicated here in bold), which is why they are the main bearers of the rheme.15 The final word or phrase may have a rising intonation, but this seems not to be obligatory.16

14 The author is not able to say which type of word stress is peculiar to Amharic. 15 To establish prosodic features of questions I used my own recordings of two Amharic speakers as well as recordings that accompany Obolenski et al. (1964). 16 In the speech of two Ethiopians rising intonation at the end of content questions did not occur.

Question about Amharic Questions with yǝhon (21)

ምን mən

ችግር čəggər

ገጠማት? gäṭṭäm-ä-at?

what

problem

happen\PFV-3SG.M-OJ.3SG.F

29

‘What problem happened to her?’ (22)

አዲስ addis

መኪና mäkina

መግዛት mägzat

ለ ምን lämən

ያስፈልጋታል? y-asfälləg-at-all?

new

car

buy\VN

why

3SG.M-need\IPFV-OJ.3SG.FAUX.NPST

‘Why does she need to buy a new car?’ (lit. ‘Why does it need her to buy a new car?’) (23)

እናቴ ənnat-e

አሁን ahun

ምን mən

እየሠራች əyyä-särra-čč

ነው? nä-w?

mother-POSS.1SG

now

what

PROG-work\PFV-3SG.F

COP-3SG.M

‘What is my mother doing now?’ The prosody of Amharic content questions with yəhon is essentially the same as in the above examples. The interrogative pronouns are always very clearly stressed, whereas yəhon, being the last word, does not have a rising intonation or, if it is there, it is not prominent. (24)

ምን mən

ችግር čəggər

ገጥሟት gäṭm-o-at

ይሆን? yəhon?

what

problem

happen\CNV-3SG.M-OJ.3SG.F

AUX.Q

‘What problem might have happened to her?’ (25)

አዲስ addis

መኪና mäkina

መግዛት mägzat

ለ ምን lämən

ያስፈልጋት y-asfälləg-at

ይሆን? yəhon?

new

car

buy\VN

why

3SG.M-need\IPFV-OJ.3SG.F

AUX.Q

‘Why might she need to buy a new car?’ (26)

እናቴ ənnat-e

አሁን ahun

ምን mən

እየሠራች əyyä-särra-čč

ይሆን? yəhon?

mother-POSS.1SG

now

what

PROG-work\PFV-3SG.F

AUX.Q

‘What might my mother be doing now?’

30

Magdalena Krzyżanowska

Hence, the rheme of Amharic content questions with yəhon is indicated by the interrogative pronoun, which is stressed, and, occasionally, by a rising intonation on yəhon. For Amharic polar questions, interrogativity is always conveyed by the intonation, such that the last word or phrase is uttered with rising pitch (indicated with ). Also, it might be, as in (28), that if the last word is a copula (ነው näw) then the last syllables of the penultimate word are stressed,17 because this is the essential part which is to be verified. (27)

ትላንት təlant

እዚህ əzzih

ዝናብ zənab

ዘነበ? zännäb-ä?

yesterday

here

rain

rain\PFV-3SG.M

‘Did it rain here yesterday?’ (28)

 ya


 säw-əyye

   ityop̣p̣yawi

? nä-w?

that

man-IND

Ethiopian

COP-3SG.M

‘Is that man an Ethiopian?’ Analogously, questions with yəhon, where the interrogative auxiliary occurs at the end, have a rising pitch on yəhon. (29)

ትላንት təlant

እዚህ əzzih

ዝናብ zənab

ዘንቦ zänb-o

ይሆን? yəhon?

yesterday

here

rain

rain\CNV-3SG.M

AUX.Q

‘Might it have rained here yesterday?’ (30)

ያ ya

ሰውዬ säw-əyye

ያዊ ኢትዮጵያ ityop̣p̣yawi

ይሆን? yəhon?

that

man-IND

Ethiopian

AUX.Q

‘Might that man be an Ethiopian?’ (31)

ያ ya

ሰውዬ säw-əyye

ያዊ ኢትዮጵያ ityop̣p̣yawi

ይሆን yə-hon

ይሆን? yəhon?

that

man-IND

Ethiopian

3SG.M-be\IPFV

AUX.Q

‘Might that man conceivably be an Ethiopian?’

17 The stressed syllables need to be identified precisely on the basis of acoustic analyses.

Question about Amharic Questions with yǝhon

31

To conclude, in Amharic polar questions with yəhon the bearer of the rheme is yəhon. This distinctive pattern of intonation is interpreted semantically as an expression of will or desire on the part of the sender to be confronted with a true sentence expressed by the theme. It should be added that especially in polar questions yəhon is syncretic, i.e. it conveys both the thematic dictum, expressing the modal epistemic meaning, and the rheme, expressing the meaning of sender’s desire. 5 Polar questions with yəhon accompanied by mənalbat and ənde Let us now move on to polar questions with yəhon which are accompanied by ምናልባት mənalbat (roughly ‘perhaps’). It is difficult to say what part of speech mənalbat represents, but it appears to be an epistemic modal particle. Moreover, mənalbat is somehow dependent on yəhon, since in a question which is meant to convey a modal epistemic meaning mənalbat cannot appear alone.18 Sentence (32) is unacceptable: (32)

*ምናልባት mənalbat

ከነገ kä-nägä

ወዲያ wädiya

ይመጣል? yə-mäṭa-all?

perhaps

from-tomorrow

beyond

3SG.M-come\IPFV-AUX.NPST

‘Perhaps he will be coming the day after tomorrow?’ One more hint to the interpretation of mənalbat is the intonation. After mənalbat the speaker makes a pause and only then continues the utterance. (33) a. ሣሩም sar-u-mm

ሁልጊዜ አረንጓዴ ነው። hulgize arängwade nä-w

grass-DEF-FOC always

green

COP-3SG.M

‘The grass is always green.’ b. ምናልባት# (ሰዓሊዎችን) ለዚህ የሚገፋፋቸው mənalbat # (säʾali-wočč-ən) lä-zzih yamm-i-gäfaffa-aččäw perhaps

painter-PL-ACC

for-that REL-3SG.M-inspire\IPFV-OJ.3PL

18 This conclusion is based on my corpus where mənalbat in polar questions always cooccurs with yəhon. Furthermore, it always occurs sentence-initially. The unacceptability of (32) was confirmed by native speakers.

32

Magdalena Krzyżanowska የኢትዮጵያ ልምላሜ ይሆን? yä-ityop̣p̣ya ləməllame yəhon? GEN-Ethiopia

greenness

AUX.Q

‘Might it perhaps be Ethiopia’s greenness of vegetation that inspires (painters) for that?’ (BBS 10: 29–30) Thus, the issue we shall raise is what, if any, additional meaning mənalbat contributes to the overall meaning of the question. Compare two pairs of questions, (34) with (35) and (36) with (37): the first item (34, 36) containing mənalbat and the second item (35, 37) lacking it: (34)

ምናልባት# mənalbat #

እስከ əskä

ዛሬ zare

እሱ əssu

ብቻ bəčča

perhaps

until

today

he

only

የሞከረውን yä-mokkär-ä-w-ən

ይሆን? yəhon?

REL-try\PFV-3SG.M-DEF-ACC

AUX.Q

‘Might it perhaps be that, until today, it was only he who has tried that?’ (AB 77: 16) (35)

እስከ əskä

ዛሬ zare

እሱ əssu

ብቻ bəčča

የሞከረውን yä-mokkär-ä-w-ən

ይሆን? yəhon?

until

today

he

only

REL-try\PFV-3SG.M-DEF-ACC

AUX.Q

‘Might it be that, until today, it was only he who has tried that?’ (36)

ምናልባት# mənalbat #

ነገ nägä

ይመጣ yə-mäṭa

ይሆን? yəhon?

perhaps

tomorrow

3SG.M-come\IPFV

AUX.Q

‘Might he perhaps be coming tomorrow?’ (37)

ነገ nägä

ይመጣ yə-mäṭa

ይሆን? yəhon?

tomorrow

3SG.M-come\IPFV

AUX.Q

‘Might he be coming tomorrow?’ It would seem that basically, there is no semantic difference between the two items in each pair. One possible nuance that mənalbat may contribute to questions with yəhon is that the sender indicates a change in her/his degree of

Question about Amharic Questions with yǝhon

33

certainty concerning a state of affairs.19 That is, a sender who was certain that a state of affairs takes place (or does not take place) starts doubting it, usually influenced by interlocutors or new circumstances. Unfortunately, at the moment the author is not able to provide a semantic test which would confirm that this meaning is truly encoded in mənalbat. Apart from mənalbat, polar questions with yəhon may be accompanied by the particle እንዴ ənde, which expresses surprise and disbelief. The particle always occurs sentence-finally. (38)

እያረጀሁ əyy-aräğğä-hu

ይሆን yəhon

እንዴ? ənde?

PROG-get_old\PFV-1SG

AUX.Q

really

‘Might I really be getting old?’ (TFT 13: 14) (39)

ሽቶውን šətto-w-ən

ሊሰጡኝ l-i-säṭ-u-ňň

ይሆን yəhon

እንዴ? ənde?

perfume-DEF-ACC

PURP-3PL-give\IPFV-3PL-OJ.1SG

AUX.Q

really

‘Might it be that they are really going to give me perfume?’ (FILEGA Ep. 18) The particle indicates the sender’s surprise and disbelief in what the datum questionis conveys. Since particles mənalbat ‘perhaps’ and ənde ‘really’ have different meanings they may co-occur in questions with yəhon: (40)

ምናልባት# mənalbat #

እያረጀሁ əyy-aräğğä-hu

ይሆን yəhon

እንዴ? ənde?

perhaps

PROG-get_old\PFV-1SG

AUX.Q

really

‘Might I perhaps really be getting old?’ 6 The use of yəhon as a politeness marker Finally, it remains to consider the use of yəhon in questions as a marker of negative politeness through which the sender attends to the addressee’s desire for freedom of action (Brown & Levinson 1987:129). In contrast to the usage of questions with yəhon described above, in such cases the presence of an addressee is obligatory; the concept of politeness intrinsically involves being polite to someone, not to the empty air. Consider the following conversation: 19 I thank Azeb Amha (p.c.) for sharing with me this suggestion.

34 (41)

Magdalena Krzyżanowska A

ሻምበል šambäl

ከእንግዲህ kä-əngədih

ይመጡ yə-mäṭ-u

ይሆን? yəhon?

Captain

from-hence

3SG.POL-come\IPFV-3SG.POL

AUX.Q

‘Might Captain come later?’ (42)

B

አይመስለኝም a-y-mäslä-ňň-əmm

ተሰናብተው täsänabtä-w

ነው nä-w

የሄዱ yä-hed-u

NEG-3SG.M-seem\IPFV-

say_good_bye\CNV3SG.POL

COP-

REL-go\PFV3SG.POL

OJ.1SG-NEG

3SG.M

‘I don’t think so, he left saying goodbye.’ (lit. ‘I don’t think so, it is [after] saying goodbye that he left.’) (43)

A

ነገስ nägä-ss

-

tomorrow TOP

ይመጡ yə-mäṭ-u

ይሆን? yəhon?

3SG.POL-come\IPFV-3SG.POL

AUX.Q

‘How about tomorrow? Might he come?’ (44)

B

መቼም mäčem

ከዚህ kä-zzih

ቤት bet

ጠፍተው ṭäftä-w

in_fact

from-this

house

disappear\CNV-3SG.POL

አያውቁም፤ a-y-awq-u-mm

ይመጡ yə-mäṭ-u

ይሆናል yəhonall

NEG-3SG.POL-know\IPFV-

3SG.POL-come\IPFV-3SG.POL

AUX.EMOD

3SG.POL-NEG

‘In fact, he has never disappeared from this house; he might come.’ (MAH 183: 5–10) In (41) and (43) the sender uses questions with yəhon to show deference to the addressee and, at the same time, to humble himself. The sender employs this strategy to express a non-imposing request which, as he believes, has more chances of bringing positive results. Since the addressee is of higher social rank (Captain) than the sender – which means also a certain social distance between them – by paying respect the sender clearly indicates that he recognizes the addressee’s status and power.20 If we replace yəmäṭu yəhon? ‘Might he come?’ by yəmäṭallu? ‘Will he come?’ in (41) and (43), the latter

20 For the politeness strategy of giving deference, see Brown & Levinson (1987:178 ff.).

Question about Amharic Questions with yǝhon

35

question could sound intrusive, too daring and as such be acceptable only in communication in familiar circles. 7 Conclusion This article treats the semantics of Amharic content and polar questions containing the epistemic modal auxiliary yəhon. The analysis of their meaning consists in indicating the main components of the thematic-rhematic structure, i.e. theme (datum questionis), thematic dictum and rheme and, subsequently, in establishing the meaning that each component contributes. The datum questionis of both types of questions with yəhon is a declarative sentence which is presupposed in content questions but not in polar questions. The thematic dictum, being an additional predication about the theme, is conveyed by the auxiliary yəhon. In content questions yəhon carries modality de re since the sender does not know which one out of the range of candidate things should be substituted for the variable. In polar questions yəhon carries modality de dicto since the sender does not know whether the datum questionis is verifiable but still s/he thinks that it should be taken into consideration as a true sentence. In Amharic polar questions with yəhon the bearer of the rheme is yəhon whereas in content questions it is the interrogative pronoun. Additional issues touched upon in the article are the semantic difference between yəhon and the distinct modal epistemic auxiliary yəhonall, the meaning of the particles mənalbat and ənde co-occurring in questions with yəhon and finally, the pragmatic use of yəhon as a marker of politeness. The analysis presented above should be treated as a proposal for how these challenging questions might be studied. Undoubtedly, it still needs refinement, especially in those parts that concern the meaning of yəhon and those describing the prosodic features of questions. The prosody of Amharic questions has not been carefully investigated yet; hence it is treated only tentatively and to a limited extent in this paper, as far as my knowledge and amateur recordings permitted. A thorough study of this aspect of questions may verify whether questions with yəhon indeed share the prosodic characteristics of standard questions. Also, the prosody of questions with yəhon and mənalbat is worth further research. In general, it seems that a more adequate analysis of these questions should be possible once we have at our disposal a good semantic description of standard questions. Two additional points concerning questions with yəhon, which may be studied in the future, are the syncretism of yəhon as a copula and an epistemic auxiliary (cf. footnote 7) and the semantic difference between questions where yəhon is preceded by the bare simple imperfective form of a verb

36

Magdalena Krzyżanowska

(simple imperfective + yəhon) and by the simple imperfective form of a verb with prefixed lə- (lə-simple imperfective + yəhon), e.g.: (45)

አዲሱ addis-u

ገንዘብ gänzäb

new-DEF

money

የመጠየቂያ yä-määyyäqiya

ስልት səlt

ምን mən

GEN-way_of_asking

way

what

ይሆን yəhon

ይሆን? yəhon?

be\IPFV.3SG.M

AUX.Q

‘What might be a new way of asking for money?’ (46)

አዲሱ addis-u

ገንዘብ gänzäb

የመጠየቂያ yä-määyyäqiya

ስልት səlt

ምን mən

new-DEF

money

GEN-way_of_asking

way

what

ሊሆን l-i-hon

ይሆን? yəhon?

PURP-3SG.M-be\IPFV

AUX.Q

‘What might become a new way of asking for money?’ (AB 36: 37–38) In (39) and (46), the prefix lə- is for the moment glossed as PURP (purposive) after Leslau (1995:674). Abbreviations # Pause 1, 2, 3 First, second, third person ACC Accusative AUX Auxiliary BEN Benefactive COP Copula CNV Converb DEF Definite EMOD Epistemic modality F Feminine FOC Focus GEN Genitive IDPH Ideophone IND Individuality IPFV Imperfective

M NPST OJ PFV PL POL POSS PROG PURP Q REL SG TOP VN

Masculine Non-past Object agreement Perfective Plural Polite Possessive Progressive Purposive Question Relative clause Singular Topic Verbal noun

Question about Amharic Questions with yǝhon

37

References Alemayehu Haile. 1990. An autosegmental approach to some Amharic declarative and interrogative contours. In Richard Pankhurst, Zekaria Ahmed & Beyene Tadesse (eds.), Proceedings of the First National Conference of Ethiopian Studies, 449–460. Addis Ababa: Institute of Ethiopian Studies, Addis Ababa University. Allwood, Jens, Lars-Gunnar Andersson & Östen Dahl. 1977. Logic in Linguistics. Cambridge: Cambridge University Press. Bogusławski, Andrzej. 1977. Problems of the Thematic-Rhematic Structure of Sentences. Warsaw: Państwowe Wydawnictwo Naukowe. Bogusławski, Andrzej. 1993. Ob “udivitiel’nom” možet v voprositel’nych predloženijach [On “puzzling” možet in interrogative sentences]. In Francesca Fici Giusti & Simonetta Signorini (eds.), Kategorija Skazujemogo v Slovjanskich Jazykach, 29–40. Munich: Sagner. Brown, Penelope & Stephen C. Levinson. 1987. Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press. Bussmann, Hadumod. 1996. Routledge Dictionary of Language and Linguistics. London & New York: Routledge. Chierchia, Gennaro & Sally McConnell-Ginet. 2000. Meaning and Grammar: An Introduction to Semantics. Cambridge, MA & London: MIT Press. Coates, Jennifer. 1983. The Semantics of the Modal Auxiliaries. London: Croom Helm. Cohen, Marcel. 1936. Traité de langue amharique (Abyssinie). Paris: Institut d’Éthnologie. Danielewiczowa, Magdalena. 1996. O Znaczeniu Zdań Pytajnych w Języku Polskim. Charakterystyka Struktury Tematyczno-Rematycznej Wypowiedzeń Interrogatywnych [On the Meaning of Polish Interrogative Sentences]. Warsaw: Wydawnictwa Uniwersytetu Warszawskiego. Gamut, L. T. F. 1991. Logic, Language and Meaning. Volume 2: Intensional Logic and Logical Grammar. Chicago & London: The University of Chicago Press. Givón, Talmy. 1994. Irrealis and the subjunctive. Studies in Language 18(2). 265–337. Goldenberg, Gideon. 1966. The Amharic Tense-System [in Hebrew]. Jerusalem: The Hebrew University (PhD Dissertation).

38

Magdalena Krzyżanowska

Karolak, Stanisław. 1999. Remat. Temat. In Kazimierz Polański (ed.), Encyklopedia Językoznawstwa Ogólnego [Encyclopedia of General Linguistics], 493–494, 596–597. Wrocław, Warsaw & Kraków: Zakład Narodowy Ossolińskich. Leslau, Wolf. 1995. Reference Grammar of Amharic. Wiesbaden: Harrassowitz. Lyons, John. 1977. Semantics, vol. II. Cambridge: Cambridge University Press. Lyons, John. 1995. Linguistic Semantics. Cambridge: Cambridge University Press. Nuyts, Jan. 2001. Epistemic Modality, Language and Conceptualization: A Cognitive-Pragmatic Perspective. Amsterdam: John Benjamins. Obolenski, Serge, Zelelie Debebow & Andualem Mulugeta. 1964. Amharic. Basic Course. Units 1–50. Washington, DC: Foreign Service Institute. Quine, Willard van Orman. 1953. Reference and modality. From a Logical Point of View: Nine Logico-Philosophical Essays, 139–159. Cambridge: Harvard University Press. Rytel, Danuta. 1982. Leksykalne Środki Wyrażania Modalności w Języku Czeskim i Polskim [Lexical Means of Expressing Modality in Czech and Polish]. Wrocław: Zakład Narodowy im. Ossolińskich.

Abbreviations for Amharic sources AB Abrəham Rätta Alämu. 1999 A.M. [2007 A.D.] Ababba säw yällämm: ač̣ač̣č̣ər ləbbwällädočč [Ababba nobody is there: Short stories]. Addis Ababa: Evangadi Productions. BBS

Mäsfən Habtämaryam. 1983/84 A.D. Yäbunna bet sə’əločč ənna leloččəmm wägočč [Pictures in a bar and other stories]. Addis Ababa: MI Printers.

FILEGA

Fǝlläga [The quest – radio play in Amharic]. 2006. Addis Ababa.

KB

Bä’alu Gərma. 1962 A.M. [1969/70 A.D.] Kadmas baššaggär [Beyond the horizon]. Addis Ababa: Ethiopian Book Centre.

MAH

Mamo Wəddənäh. 1982 A.M. [1989 A.D.] Mahəbärtäňňočču: tarik qässämä ləbbwälläd [Members of the association: A historical novel]. Addis Ababa: Commercial Printing Press.

Question about Amharic Questions with yǝhon

TFT

39

Asamnäw Barägga. 1994 A.M. [2001/02 A.D.] Yätroya färäs ənna leločč ač̣ač̣č̣ər tarikočč [Troyan Horse and other short stories]. Addis Ababa: Mega Printing Enterprise.

Interrogativity in Baskeet Yvonne Treis LLACAN (INALCO, CNRS, PRES Paris-Cité) Abstract The present article analyzes the grammatical means of encoding interrogativity in the Omotic language Baskeet. At first, an overview of interrogative marking on verbal and non-verbal predicates is given. Depending on the type of predicate, direct questions are marked by intonation only, or, additionally, by an interrogative morpheme -a, or special interrogative verb forms. Secondly, the forms and functions of the six simplex interrogative pronouns, i.e. PERSON, THING, TIME, QUANTITY, MANNER, PLACE, and SELECTION interrogatives, as well as pronouns derived from them, are discussed. Finally, the article considers the use of interrogatives in non-interrogative contexts. The analysis is primarily based on a corpus of recorded spontaneous speech events. 1 Introduction This article1 discusses the grammatical means of expressing interrogativity in Baskeet,2 an Omotic language spoken by about 80,000 speakers3 in the Basketo Special Woreda and the neighboring Melokoza Woreda of the GamoGofa Zone (SNNPR) in Ethiopia. Map 1 shows the approximate location of the Basketo Special Woreda and some of the places I have visited during my fieldtrips. The Baskeet-speaking area borders on Malo (Ometo) in the North, Dime (South Omotic) and the Bodi dialect of Me’en (Surmic) in the West, the Gelila dialect of Aari (South Omotic) in the South, and the Gofa dialect of 1

2 3

Data for this article has been collected during fieldtrips sponsored in part by a postdoctoral fellowship from La Trobe University (Melbourne) and by the Endangered Languages Documentation Programme (IPF-Project: Documentation of Baskeet song, verbal art and ceremonial language). This support is gratefully acknowledged. I am indebted to my Baskeet consultants; my special thanks go to Ambaye Tsedeke, Tamiru Admasu, Seid Ali, and Dutse Tamiru. I am grateful to Ronny Meyer, Dmitry Idiatov, and Azeb Amha for helpful comments on an earlier version of this paper. ISO 639-3 code: bst, glottocode: bask1236. According to the 2007 Ethiopian census, there are 78,284 members of the Baskeet ethnic group in Ethiopia, the large majority of which live in the Basketo Special Woreda (Federal Democratic Republic of Ethiopia Population Census Commission 2008:84).

42

Yvonne Treis

the North Ometo cluster in the East. In the Omotic language family, Baskeet belongs to the Ometo languages, which are a sub-group of the North Omotic branch. Baskeet and other Ometo languages spoken nearby (e.g. Malo, Maale, Oyda, and Wolaitta-Gamo-Gofa) are not mutually intelligible. Within Baskeet, no dialectal variation has been observed so far.4 Baskeet is the autonym of the Baskeet people. The term can refer to the language, the ethnic group, and the land (Treis 2011a). In the literature, the language and the people are either referred to by the Amharic term Basketo or by Mesketo of unknown origin (Treis 2011b; 2011c). Little has so far been published on Baskeet in English; see, however Sottile’s (2002) unpublished PhD thesis in Italian and Inui’s (2005; 2006; 2012 among others) work in Japanese. The analysis of many grammatical domains is still in its early stages. The earliest information on interrogative pronouns is found in Inui (2005:33 ff.) and Sottile (2002:105–117); interrogative marking on verbs is briefly addressed in Sottile (2002:178–181) and Inui (2006:38–43).5 Whereas earlier publications on Baskeet have been exclusively based on elicitation, my own work is mostly based on a corpus of recorded spontaneous speech events.6 The structure of this article is as follows: After an introduction to the Baskeet sound system in Section 2, Section 3 discusses interrogative marking on verbs. The core of this article, Section 4, analyses the use and the morphology of Baskeet interrogative pronouns. Section 5 considers the use of interrogative pronouns in non-interrogative contexts. Section 6 concludes the article. The present analysis of interrogativity is far from exhaustive. It especially lacks an analysis of interrogative intonation and a discussion of the grammar of subordinate questions, which will be attempted in the future.

4 5 6

The widespread claim that there are two Baskeet dialects, Dookka and Doolla (or the like), is unproven and probably wrong (Treis 2012). Inui’s description of interrogative verb forms contains many errors. In some of the paradigms, verb forms are assembled that actually belong to different paradigms. He also makes an unnecessary distinction between verbs in content and polar questions. The sound files of the examples presented in this article are available online under: http://llacan.vjf.cnrs.fr/fichiers/extraits/.

Map 2: Location of the Basketo Special Woreda

Interrogativity in Baskeet 43

44

Yvonne Treis

2 Phonological overview Baskeet has a five-vowel system (/a/, /e/, /i/, /o/, /u/) with a phonemic length contrast. Simplex vowels combine into five diphthongs: /ai/, /ei/, /oi/, /ui/, /au/, which are represented by , , , , and in the orthography. The Baskeet consonant chart with 25 consonant phonemes is provided below.

STOPS

voiceless voiced glottalized

AFFRICATES FRICATIVES

LIQUIDS GLIDES

t͡ ʃ t͡ ʃ ’

voiceless voiced

s z

ʃ ʒ

m tap lateral

k g k’

GLOTTAL

VELAR

PALATO-

ALVEOLAR

voiceless glottalized

t d ɗ t͡ s t͡ s’

NASALS

p b ɓ

ALVEOLAR

PLACE OF ARTICULATION → MODE OF ARTICULATION ↓

LABIAL

Table 1: Baskeet Consonant Phonemes

ʔ

h

n ɾ l

w

j

The orthographic representation, as used in this article, is given in angle brackets if it is different from the IPA symbol. Length is phonemic in the consonant system, where all consonants (except h and r) distinguish between a simplex and a geminate realization. Baskeet is a tonal language with two tonemes, high (H) and low (L). Tone is used to differentiate lexemes and tonal minimal pairs are easily found in the nominal and verbal domain, e.g. zarzará ‘lizard’ (LLH) vs. zarzára ‘sieve’ (LHL), k’áb’a ‘tongs’ (HL) vs. k’áb’á (HH) ‘remaining unfinished work’, mícc- (H) ‘spread (something) in the sun’ vs. micc- (L) ‘burn (something)’. As the tonal analysis (especially of tonal sandhi rules) is still ongoing, tone is not marked in this article.

Interrogativity in Baskeet

45

3 Interrogative marking on predicates 3.1 General interrogative marking In Baskeet, interrogativity is marked: (i) By interrogative intonation only (STRATEGY 1), (ii) By an interrogative suffix -a (or its variant -(i)ya) on the predicate (STRATEGY 2) plus interrogative intonation, or (iii) By the use of an interrogative verb form (STRATEGY 3) plus interrogative intonation. The choice of the interrogative strategy is dependent on the type of sentence predicate. It is not dependent on the type of question: polar questions and content questions both require interrogative marking. The Baskeet verbal paradigms can be separated into (i) those that can be used in interrogative and non-interrogative contexts alike, without any additional morphology, (ii) those that receive an additional suffix -a when used in questions, and (iii) those that have a non-interrogative and an interrogative sub-paradigm that cannot be derived from each other. Affirmative and negative intentional verbs, for instance, belong to verbal paradigm type (i). Thus only the associated intonational curve marks the utterance as a question.1 As (1) shows, the intentional verb form in -anda is not morphologically marked for interrogativity; this is proved by the use of the same verb form in a declarative sentence in (2).2 (1) nuuni

zuggi

miiz yel-in-e

ga-ar

pird-ar

1PL.LG

Zugga

cow

say-CNV2

judge-CNV2

nu

giyaab

wonda

ak’-anda?

1PL

tomorrow

where

pass_the_night-INT

give_birth-PRF-DECL

‘(If) we conclude that Zugga’s3 cow has given birth (to the calf), where do we want to pass the night (i.e. to live) tomorrow?’ (LS100021_000313013-000323530)

1

2 3

An in-depth intonational analysis of questions is still to be carried out in the future. Polar questions are characterized by extra-high pitch rise on the penultimate syllable of the utterance and a long falling tone movement on the last syllable. In content questions, the pitch peaks on the penultimate syllable of the interrogative pronoun. Jussives of the first person (see below) are another example of verb forms that are only marked by intonation as being interrogative. Zugga is a fairy-tale character. The small fury animal does not seem to exist in real life.

46

Yvonne Treis

(2) ee

nu

lukk-anda

yes

1PL

go-INT

‘Yes, we are decided to go.’ (LS100714_000008406-000010508, elicited)

Some verbal paradigms, such as affirmative and negative imperfective verbs, affirmative perfect verbs, and negative perfective verbs,4 require STRATEGY 2, i.e. the suffixation of an interrogative morpheme -a as well as associating the utterance with the appropriate interrogative intonation. In (3) the morpheme is attached to the negative imperfective, in (4) to the perfect form. (3) “te 1SG

yinti

dabbi-shoa

ha

te

acc-aa

2PL

relative-AFF

DEM1.M

1SG

tooth-M

bek’baas-a”

ga-ar

see\M.NEG.IPFV-Q

say-CNV2

[…]

‘(The bat) said: “I am your relative! Don’t (you) see these, my teeth?” […]’ (LS100441_000034229-000037411) (4) tuus-in

tol-in-a?

center_pole-F

erect-PRF-Q

‘Have (they) erected the center pole?’ (LS100710_000050027-000052119, elicited)

The interrogative STRATEGY 2 is also applied to non-verbal predicates. In (5) the interrogative suffix -a is added to the nominal predicate baz ‘thing’. (5) nuub 1PL.DAT

ol-i

awt-iz

baz-a?

fight-NOM

be_what-M.IPFV.REL

thing-Q

‘What is the benefit of a fight for us?’ (lit. ‘A fight is a thing that does what for us?’) (LS100441_000100355-000101895) A third type of verbal paradigms – including, among others, affirmative perfective forms – requires interrogative STRATEGY 3. The paradigms of this verb type subdivide into interrogative and non-interrogative (sub-)paradigms. If the interrogative and declarative forms are compared, no interrogative morpheme can be isolated. Instead, subject agreement, aspect/modality and

4

The perfective and perfect paradigms share one negative paradigm.

Interrogativity in Baskeet

47

mood are marked by portmanteau-morphemes; see the perfective forms in Table 2.5 Table 2: Declarative and Interrogative Perfective Paradigms (sool- ‘tell’)

6

AFFIRMATIVE PERFECTIVE

NEGATIVE PERFECTIVE

DECLARATIVE

INTERROGATIVE

DECLARATIVE

INTERROGATIVE

sool-akkay(-e)

sool-akkay-a

1 form

1 form

SG

1 2 3M 3F

sool-ade sool-ade sool-ide sool-ade

sool-e sool-i ~ -iye sool-e sool-a

PL

1 2 3

sool-ide sool-ide sool-ide

sool-e sool-it(e) sool-e

2 forms

4 forms

In the affirmative, the interrogative perfective paradigm distinguishes more persons than the corresponding declarative paradigm. Whereas there is only a two-way opposition between masculine and feminine verb forms (3M/1PL/2PL/3PL vs. 1SG/2SG/3F) in the declarative paradigm, four forms (1SG/3M/1PL/3PL vs. 2SG vs. 3F vs. 2PL) are distinguished in the corresponding interrogative paradigm. Example (6) illustrates the use of an interrogative perfective verb form.7 (6) ne 2SG

woy-di

d’ay-i?

where-LOC

disappear-2SG.PFV.Q

‘Where did you disappear to?’ (LS100429_000442832-000443892) In the negative perfective, the person distinctions are neutralized; all verb forms end invariably in -akkay. In questions, the interrogative suffix -a (see STRATEGY 2 above) is added to this verb form (7).

5 6 7

The bracketed vowels in the tables are optional. The rules that determine their occurrence are so far unknown. The final vowels of the interrogative perfectives tend to be lengthened. Other examples: (48), (49), (51), (52), (56), (57).

48

Yvonne Treis

(7) pit-in sweepings-F

keetts-app

tong-int-akkay-a?

house-ABL

throw_out-PASS-NEG.PFV-Q

‘Haven’t the sweepings been thrown out of the house?’ (LS100288_000020645-000023594, elicited)

The jussive is a verb form which expresses, generally speaking, obligations, permissions, commands, and wishes. As shown in Table 3, the jussive distinguishes between a declarative and an interrogative (sub-)paradigm inflectionally. The negative jussive forms are marked by the morpheme -app, both in the declarative and interrogative mood.8 The interrogative forms, like the declarative forms, consist of a verb stem plus a portmanteau-morpheme for subject agreement and mood. No interrogative morpheme can be isolated, if the declarative and interrogative forms are compared. However, the endings of the interrogative jussive partly overlap with the endings of the interrogative perfective (compare Table 2 and Table 3). This means that the interrogative jussives could be further segmented into an invariable jussive interrogative morpheme -an plus the inflectable interrogative endings -i in 2SG, -e/-o in 1SG9 and -e in 3M, 1PL, 3PL, -a in 3F, and -t(e) in 2PL. Table 3: Declarative and Interrogative Jussive Paradigms (sool- ‘tell’) DECLARATIVE JUSSIVE

INTERROGATIVE JUSSIVE

AFFIRMATIVE

NEGATIVE

AFFIRMATIVE

NEGATIVE

sool-app-ano 10 ⎯ sool-app-ane sool-app-ana

SG

1 2 3M 3F

sool-o sool-ab(e) sool-o sool-o

sool-app-o sool-app(-e) sool-app-o sool-app-o

sool-ano sool-ani sool-ane sool-ana

PL

1 2 3

sool-o sool-ibt(e) sool-o

sool-app-o sool-app-it(e) sool-app-o

sool-ane sool-app-ane sool-ant(e) ⎯ sool-ane sool-app-ane

3 forms

3 forms

5 forms

3 forms

8 Note, however, that the 2SG and 2PL negative declarative forms are not directly formed on the basis of the affirmative forms. 9 In the interrogative perfective, the 1SG forms ends in -e, whereas it ends in -o in the interrogative jussive (compare Table 2 and 3). 10 The missing forms are neither attested in my corpus nor could they be elicited. The hypothetical forms *sool-app-ani and *sool-app-ante (or corresponding forms of other verbs) were considered non-existing or ungrammatical by two native speakers consulted.

Interrogativity in Baskeet

49

No translations of the verb forms are given in Table 3 because the meaning of the jussive forms interacts with person (of the subject) and mood: • In the first person, the declarative and interrogative jussives overlap functionally. The interrogative jussive forms are used to ask for permission or about an obligation (‘can I/we tell?’, ‘should I/we tell?’; cf. (8)). Requests for permission or inquiries about an obligation can also be expressed by the declarative jussive in combination with the interrogative intonation. In fact, this is the normal way of using declarative jussives in the first (!) person. Thus, there are only a few attested cases in the corpus in which the first person jussive forms are realized without interrogative intonation, in which case they express directives (hortatives) (‘let me/let us tell’).11 • In the declarative mood of the second person, the jussive expresses directives and is best translated by English imperatives (‘tell!’). In the interrogative mood of the second person, the jussive expresses polite directives (‘could you (SG/PL) please tell?’) or inquires about a wish (‘do you (SG/PL) wish to tell?’). • In the third person, the declarative jussive forms express permission (‘he/she/they is/are allowed to tell’), directives to grant permission (‘let him/her/them tell’), obligation (‘he/she/they should tell’), wishes and curses (‘may he/she/they tell’). The interrogative jussive forms express requests for permission (‘is/are he/she/they allowed to tell?’) or questions about obligations (‘should he/she/they tell?’). Example (8) illustrates the use of an interrogative jussive verb form. (8) nu

wol’

1PL together

a

galass shiikk’-ar

igint-ane?

what

day

be_reconciled-1PL.JUS.Q

gather-CNV2

‘When (lit. ‘on what day?’) should we gather and be reconciled?’ (LS100441_000114240-000116439)

As the morphological analysis of the Baskeet verb is not yet completed, I am still unable to give a comprehensive account of interrogative marking. For some verb forms that are not discussed above it is still unclear how they are marked in questions. Further research on interrogative marking is definitely required in the future.

11 For hortatives, there are also forms that are not part of the jussive paradigm: sool-ast(o) ‘let me tell’, sool-ist(o) ‘let us tell’. So far, no functional difference between the hortatives in -o or in -ast(o)/-ist(o) could be determined.

50

Yvonne Treis

If the morphological complexity of verbal interrogative marking of Baskeet is compared to that of related languages, Baskeet can be said to occupy an intermediate position. On the low end of the scale of complexity, Maale (South Ometo) marks interrogativity on verbs by the affixes -y in content questions and -íya (or only interrogative intonation) in polar questions (Azeb 2001:155 f.). On the high end of the scale, Wolaitta (North Ometo) possesses complex interrogative paradigms distinguishing up to seven forms per paradigm (Azeb 2012:465 f.; Wakasa 2008:753–760, 774–780, 795– 802).12 The minimal answer to polar questions in Baskeet is aa ‘yes’ (9), ee(’ee) ‘yes, ok’ (2), or wo’’ak(e) ‘no, not at all’. There is no difference between answers to affirmative and negative polar questions: both are confirmed with ‘yes’ and contradicted with ‘no’. The interjection aa ‘yes’ is often pronounced with pulmonic ingressive airstream (10). (9) S1: zimb-in=garta hat-hat lyre-F=inside

stumble-stumble

wot-az-in

wod’

make-F.IPFV.REL-F

exist

–

‘Among the lyre songs there are (some) that make me stumble (i.e. which I don’t master well).’ S2: wod’-a?

S1: aa

exist-Q

yes

‘Are there (such songs)?’

‘Yes.’

(ML2012-01-10_001_001344057-001346866)

(10) S1: Gayl-ants-i Gelila_Aari-PL-NOM

izi yetts’-ir-a? 3F

sing-M.IPFV-Q

‘Do the Gelila Aari sing it (= the song)?’

S2: ↓aa yes

‘Yes.’

(ML2012-01-10_001_002221154-002222441)

Polar questions are also commonly answered by echoing the verbal or nonverbal predicate in the declarative form for confirmation (11). If speakers want to contradict the question, they change the polarity value of the predicate that they echo, e.g. from er-ir to er-baas in (12).

12 Furthermore, Baskeet – unlike various other Omotic languages – is not a language with “interrogative zero-marking” in the sense of Köhler (2013).

Interrogativity in Baskeet (11) S1: meli ootsi other

work

intab

bayy-a?

3PL.DAT

not_exist-Q

51

S2: bayy not_exist

‘They don’t have any other work?’

‘(No, they) don’t have.’

(ML2012-01-10_001_000546665-000548124)

(12) S1: an 2M.VOC

ne

baab-aa tiin-o

yinni

zay-in

2SG

father-M

DEM2.F

bamboo_trumpet-F

pugg-i

er-ir-a?

blow-CNV1

know-M.IPFV-Q

first-M

‘Did your father, the first one (i.e. the biological father), know how to play this bamboo trumpet?’ S2: te 1SG

baabi

er-baas

father

know-M.NEG.IPFV

‘(No,) my father didn’t know.’ (BI2012-01-21_002_000422232-000425581)

3.2 Specialized interrogative morphemes Apart from the general interrogative morpheme -a, Baskeet possesses three other functionally specialized interrogative markers, -ish for follow-up questions, -be for alternative and confirmation questions, and -baya for strong confirmation questions. The morpheme -ish is attached to nouns, pronouns or nominalized phrases to mark a follow-up question ‘and what about X’.13 In (13) the interviewer first asks about the number of musical pieces played at mourning ceremonies. After having received the answer, he asks a follow-up question marked by -ish. A question with -ish can follow up on a content question or a polar question (see also (33)). (13) S1: yeepi=gall

aapin

mourning_ceremony=top how_many

pugg-ir-a? blow-M.IPFV-Q

‘How many (different musical pieces) do you play at mourning ceremonies?’

S2: tabza seven

‘Seven.’

13 Functionally similar – but non-cognate – interrogative morphemes are also found in other unrelated Ethiopian languages; see, for instance, -’nnu in Kambaata (Cushitic) (Treis 2008:228 f.) and -ss(a) in Amharic (Semitic) (Leslau 1995:775, 889–892).

52

Yvonne Treis S1: issi=gall-ish?

S2/3: isshin

wedding=top-SUPPL.Q

five

‘And (what about) at weddings?’

‘Five.’

(BI2012-01-21_002_002941118-002946070)

The morpheme -be is, first of all, a disjunctive interrogative morpheme. It is attached to the first phrasal or clausal coordinand of a disjunction in an alternative question (‘Is it X or Y?’); in (14) two clauses are coordinated, in (15) two postpositional phrases.14 The second coordinand is marked like a regular polar question; see the sentence-final interrogative morpheme -a. (14) ne 2SG

baabi bess-in=gall

ini

wod’i-be

bet-i

wod’-a?

father

there

exist-DISJ.Q

migration-NOM

exist-Q

place-F=top

‘Are (you) (still) there at your father’s place or have (you) migrated (lit. ‘is there migration’)?’ (BI2012-01-21_002_000112843-000114693) (15) S1: maall-ants=bar-be non_Mani-PL=ICP-DISJ.Q

man-ants=bar

kotts’-aar-a?

Mani-PL=ICP

share-F.IPFV-Q

‘Do (you) share (the farm work) with non-Mani or with Mani (= members of the potter and smith clans)?’ S2: maall-ants=bar-ar non_Mani-PL=ICP-ADD

kotts’-aar share-F.IPFV

‘(I) also share with non-Mani.’ (BI2012-01-21_002_001248544-001251068)

The morpheme -be is also attested outside alternative questions, namely as the final morpheme in affirmative polar questions. In this position, its exact function is not yet clear. It can be tentatively described as marking confirmation questions to which the speaker expects a positive answer.15 In (16) the interviewed tells the interviewer that he learnt to play (lit. ‘blow’) the lyre from his father. Since he had mentioned earlier in the interview that he grew up with his stepfather, the interviewer asks the interviewed to confirm that his stepfather was his lyre teacher.

14 Another example with the -be-morpheme can be found in (63). 15 Confirmation questions with final -be could be analyzed as alternative questions that lack the second coordinand.

Interrogativity in Baskeet (16) S1: te 1SG

53

baab-aa

min

pugg-ir=don

te

father-M

once

blow-M.IPFV=SIMUL.DS

1SG

baab-aadi

te

g[illi=don] ...

father-M.DEF.NOM

1SG

small=SIMUL.DS

‘When my father played in the old days, my father, when I was [small] … [interrupted].’ S2: injeri baab-aadi-be? injera

father-M.DEF.NOM-DISJ.Q

‘Your stepfather, wasn’t it?’ S1: injeri injera

baab-aadi

pugg-ir

father-M.DEF.NOM

blow-M.IPFV

‘(Yes,) my stepfather played.’ (BI2012-01-21_002_001705191-001710141)

The morpheme -baya marks strong confirmation questions for which the speaker firmly expects a positive answer. The morpheme probably goes back to negative existential copula bayy plus the general interrogative morpheme -a, which came to be attached to the last constituent of a polar question. Like the morpheme -be, -baya is only attested on affirmative (verbal and nonverbal) clauses. (17) S1: tiin zay-ants

yinti

first bamboo_trumpet-PL 2PL

pugg-iz-ants

hayzz-baya?

blow-M.IPFV.REL-PL

three-CONF.Q

(Speaker 1 mentions what he has learned earlier in the conversation:) ‘First of all, the bamboo trumpets that you play are three (in number), aren’t they?’ S2: hayzzi three

‘(Yes,) three.’ (BI2012-01-21_002_002002759-002005391) 4 Interrogative pronouns The following six interrogative categories are expressed by unanalyzable morphemes in Baskeet.16 The “>” symbol introduces derived forms or functional extensions; the tilde separates variants.

16 The categories are named after Cysouw (2004).

54

Yvonne Treis PERSON (‘who’): THING (‘what’): > REASON (‘why’): > QUALITY (‘what kind of’): TIME (‘when’): QUANTITY (‘how much/how many’): > RANK (‘how manieth’): MANNER (‘how’): PLACE (‘where’): > SELECTION (‘which’): SELECTION (‘which’): > PLACE (‘where’):

oona ~ ooni a ~ abz abz-ab a ~ abz aanta ~ aanti aapin aapintsa wozar woy ~ woy-da ~ woy-di woy-di woodi (M), wonni (F), wonts (PL) won-da ~ won-di

Baskeet interrogative pronouns are freestanding morphemes, some of which inflect for case like other nominals of the language (e.g. the PERSON interrogative), while others are invariant and adverb-like (e.g. the MANNER interrogative). Number marking is not attested on interrogative pronouns. Plural number is instead indicated by reduplication, e.g. abz abz ‘what (and) what’, meaning: ‘what multiple things’; see (34) and (64). Most interrogatives (PERSON, THING, QUANTITY, PLACE, SELECTION) can be used as heads of noun phrases or as modifiers. Synchronically, the interrogative stems are not analyzable but it is noteworthy that the PLACE, SELECTION and MANNER interrogatives start with the formative wo_ and the THING, QUANTITY, and TIME interrogatives with the formative a(a)_. This might point to a common origin in their history.17 However, the putative historical links between the interrogative pronouns cannot be discussed in detail in this predominately synchronically oriented study. In content questions, interrogativity is marked by interrogative pronouns and – depending on the verb (see Section 3) – by interrogative morphology on the verb. The following sections discuss the individual interrogative pronouns in detail. 4.1 Person The PERSON interrogative pronoun oona ~ ooni ‘who’ inflects for case, cf. (18). The final vowels of the dative, ablative, and allative forms, which are

17 Recurrent formatives which have no morpheme status are marked by an underscore in the place of a hyphen.

Interrogativity in Baskeet

55

given here and in other (pro-)nominal paradigms, are only pronounced in prepausal position. (18) ACC: NOM: GEN: DAT: ABL: ALL: Q.PRED:

oona ~ ooni o ~ ooni o oob(o) < o + ab(o) oopp(o) < o + app(o) ook(o) < o + ak(o) oono

In the nominative and accusative, two case variants are attested in the data. The short nominative form is formally identical to the genitive pronoun ‘whose’. Nominative and accusative forms overlap partly: The long nominative form is identical to one of the accusative variants. The dative (‘for whom’), ablative (‘from whom’), and allative forms (‘to whom’) can be analyzed as mergers of the short interrogative o with the dative, ablative, and allative morpheme, respectively. The PERSON interrogative paradigm has a unique predicate form oono. The PERSON interrogative pronoun is morphologically similar to some emphatic personal pronouns. 1SG, 2SG, and 1PL personal pronouns also contain a formative _nV in the nominative, accusative and predicative forms; see taana ‘I’, neena ‘you (SG)’, and nuuna ‘we’. The genitive case form is used as an interrogative modifier (‘whose’). It also combines with relational enclitics (19), which can carry their own case markers. (19) o=bar(a) ‘with whom’ o=gal(la) ‘on whom, to whose detriment’ o=gart(a) ‘in whom’ Baskeet is a verb-final language. Examples from recorded texts (20)–(23) show that the typical position of a non-clefted interrogative pronoun is preverbal, i.e. in the position where new information is found; cf. the answer to the question in (21). (20) yintana 2PL.ACC

o who

er-is-in-a?

know-CAUS-PRF-Q ‘Who has taught you?’ (BI2012-01-21_002_000039697-000001637)

56

Yvonne Treis

(21) S1: inti ooni dokkis-ir-a? 3PL who

give_leftovers\CAUS-M.IPFV-Q

‘Who do they provide with leftovers?’ S2: inti 3PL

c’inc’al ant

‘They … to the ants.’ S3: inti 3PL

c’inc’al

dokkis-ir

ant

give_leftovers\CAUS-M.IPFV

‘They give the leftovers to the ants.’ (BI2012-01-21_002_000722501-00725936)

(22) hatt giir-in oob now tax-F

giir-aar-a?

who.DAT pay_tax-F.IPFV-Q

‘Whom do you now pay the tax?’ (BI2012-01-21_002_000151522-000153459)

Content questions are often cleft-sentences with the interrogative pronoun serving as the predicate. The rest of the clause is nominalized and functions as the subject of the sentence (23). (23) aal home

ehe-in

yintab

k’atts’-iz-a

oono?

bring-CNV.DS

2PL.DAT

cut-M.IPFV.REL-M

who.Q.PRED

‘Having taken (a goat/sheep) home, who slaughters (it) for you?’ (lit. ‘… who is the one who slaughters (it) for you?’) (BI2012-01-21_002_002837328-002839013)

Apart from inquiring about a person, the PERSON interrogative can also inquire about a person’s name. The same is observed in other Ethiopian languages, e.g. Wolaitta (Wakasa 2008:356), Koorete (Hayward 1982:231), Kambaata (Treis 2008:263), and Amharic (Leslau 1995:70).18 (24) ne 2SG

sumts-a

oono?

name-M

who.Q.PRED

‘What is your name?’ (lit. ‘Who is your name?’) (LS100004_000001342-000002562, elicited)

18 Idiatov (2007:61–94) gives a cross-linguistic overview of the use of ‘who’ in questions about names, showing that the Ethiopian case is also frequent in other languages.

Interrogativity in Baskeet

57

In all Ometo languages on which literature is available to me, i.e. Maale, Malo, Wolaitta, Koorete, Zargulla, Zayse, and Haro, the PERSON interrogative starts with a formative o(o)_.19 This formative is followed by a case marker (e.g. dative, ablative), either directly or after a _nV or _dV formative. It is obvious that all PERSON interrogatives in Ometo are cognate. 4.2 Thing The THING interrogative pronoun ‘what’ inflects for case as shown in (25).20 For most forms, a short and a long variant are attested in my database. The short form is realized as a in isolation and as ab if followed by another morpheme. With the exception of the predicate form (discussed below), the short form a(b) and the long form abz are in free variation. The short ablative form (*abappo) is not attested in the corpus and has been rejected by a native speaker. (25) ACC: NOM: GEN: DAT: ABL: ALL:

a ~ abz a ~ abz-i a ~ abz ab-ab(o) ~ abz-ab(o) abz-app(o) ab-ak(o) ~ abz-ak(o)

The genitive form a or abz combines with relational enclitics (26). (26) a=bar(a) ~ abz=bar(a) ‘with what; through what’ a=gal(la) ~ abz=gal(la) ‘on what’ a=gart(a) ~ abz=gart(a) ‘in what’ Baskeet examples from recorded texts are provided below. In (27) the interrogative pronoun is used in object function. In (28) it is the predicate of the question. In contrast to the PERSON interrogative (18), there is no separate predicate form for the THING interrogative.

19 The sources consulted are Azeb (2001; 2007), Siebert & Caudwell (2002), Wakasa (2008), Binyam (2010), Hayward (1982; 1990), and Hirut (2003). Note that most sources transcribe the pronoun with an initial prevocalic glottal stop. 20 Sottile (2002:109) mentions that the THING interrogative can be marked for definiteness but this is not confirmed by my data – neither for this nor for other interrogatives.

58

Yvonne Treis

(27) abz

ding-ir-a?

what

find-M.IPFV-Q

‘What do (you (PL)) earn?’ (BI2012-01-21_002_001428402-001429561)

(28) nee-ko-n

k’omm-in abz-a?

2SG-NMLZ-F clan-F

what-Q

‘Yours, the clan, what is (it)?’ (BI2012-01-21_002_000844146-00845485)

While the short and the long variants of ‘what’ in (25) can usually be used interchangeably, there is a significant meaning difference between the two forms in predicate function. While the long form abz-a is used in neutral information questions as in (28), the short form ab-a conveys that the speaker is irritated, annoyed or worried about the situation s/he inquires about. In utterances like (29) and (30), native speakers considered the use of the long form inappropriate. (29) neek

ab-a (*abz-a)?

2SG.ALL

what-Q

(Speaker worried about the state of health of the addressee:) ‘What happened to you!?/What’s wrong with you!?’ (Tam2013-06-17_001_000044791-000046747, elicited)

(30) zinaab yesterday

yinti

aal

yii

ab-a (*abz-a)?

2PL

home

DEM2.M

what-Q

(Speaker irritated about the noise from the neighboring compound:) ‘What was (happening) at your house yesterday?!’ (Tam2013-06-17_002_0009057-000011091, elicited)

The THING interrogative usually asks about something non-human. In one context, however, it can be also used to inquire about persons, namely in questions about a person’s classification, e.g. kinship category (31). (31) ne 2SG

aakki

mish-in

neeb

abz-a?

grandmother

older_sister-F

2SG.DAT

what-Q

‘What is your grandmother’s older sister for you?’ (no recording)

Interrogativity in Baskeet

59

The dative-marked THING interrogative is used to inquire about a REASON (32). Baskeet thus follows a very common cross-linguistic strategy in the formation of reason interrogatives (Cysouw 2004:18). (32) ne

wod’-appe!

abz-ab

wod’-aar-a?

2SG kill-2SG.IMP.NEG what-DAT kill-F.IPFV-Q

‘Don’t kill (her)! Why do you (SG) (want to) kill her?’ (BI2012-01-21_002_010554066-010555555)

As a modifier, the THING interrogative inquires about a QUALITY (‘what (kind)?’), as in (33). (33) ne 2SG

macc’-ind-ish?

a

woman-F.DEF.NOM-SUPPL.Q what

ootsa

ak’-aar-a?

work

pass_the_night-F.IPFV-Q

‘And what about your wife? What work does she live (lit. ‘pass the night’) on?’ (BI2012-01-21_002_001317760-001319610) If the modifier is reduplicated, as in (34), one inquires about things of different qualities or types. (34) hatt … now

izana …

zay-in

iz=galla …

abz

abz …

3F.ACC

trumpet-F

3F=top

what

what

yetts-iya

yinti

er-az-inda? 21

song-Q

2PL

know-F.IPFV.REL-F.DEF.NOM

‘Now, it (lit. ‘her’), the trumpet, what different kinds of (lit. ‘what and what’) songs can you play on it (lit. ‘her’)?’ (BI2012-01-21_002_002846155-002855354)

Due to frequent collocations, the short THING interrogative a and the verb wot- ‘do’ have merged into an interrogative verb awt- ‘do what’ in Baskeet, cf. (35)–(36). (35) neeb 2SG.DAT

santi

awt-ir-a?

cent

do_what-M.IPFV-Q

‘What does money mean to you?/What do you need money for?’ (lit. ‘What do cents do for you?’) (BI2012-01-21_002_010449247-010450399) 21 The symbol marks here speech errors and hesitations.

60

Yvonne Treis

(36) ne 2SG

awt-aar-a? do_what-F.IPFV-Q

‘What are you doing?’ (LS100050_000444844-000446028) Interrogative verbs are not only used as main verbs in questions but also in subordinate clauses. The intentional verb form of awt- ‘do what’ serves, for instance, to inquire about a PURPOSE (37). In (38) the relative form of awt- ‘do what’ modifies the predicate noun. (37) an

zugg-o

saatts awt-anda

koy-aar-a?

2M.VOC Zugga-VOC pebble do_what-INT search-F.IPFV-Q

‘Hey, Zugga, why are you looking for a pebble?’ (LS100259_000549646-000551482)

(38) nuub 1PL.DAT

ol-i

awt-iz

baz-a?

fight-NOM

do_what-M.IPFV.REL

thing-Q

‘What is the benefit of a fight for us?’ (lit. ‘A fight is a thing that does what for us?’) (LS100441_000100355-000101895) For other interrogative verbs in Baskeet and related Omotic languages see Section 4.5. In all Ometo languages, the THING interrogative starts with a formative a(V)_ like in Baskeet; see, for instance, ʔáa in Zayse (Azeb 2007:208), ʔal-á in Haro (Hirut 2003:181), ʔáí- in Wolaitta (Wakasa 2008:530).22 It can thus be assumed that the THING interrogative of Baskeet is cognate with that of other Ometo languages. 4.3 Time The TIME interrogative ‘when’ is so far attested in the following three case forms. (39) ACC: aanta ~ aanti DAT: aant-ab(o) ABL: aant-app(o) The use of the accusative form is exemplified in (40).

22 Note that Wakasa’s symbol for the glottal stop, , is here replaced by .

Interrogativity in Baskeet (40) aanti when

61

er-in-a? know-PRF-Q

‘When have (you) learnt it?’ (BI2012-01-21_002_000012332-000014456)

In questions about the starting and end point of a period of time, the ablative marks the former and the accusative the latter (41). (41) te 1SG

geents-a

aantapp

aanti

yel-ir-a?

bull-M

when.ABL

when

give_birth-M.IPFV-Q

(From a fairy tale:) ‘From when to when does my bull (finally) give birth?’ (LS100021_000032427-000034308) Questions about a precise point in time are combinations of the THING interrogative a (Section 4.2) and a temporal noun, e.g. saata ‘hour’ (42) or galass ‘day’ (43). (42) yinti 2PL

a

saati

lukk-i

gel-ir-a?

what

time

go.CNV1

enter-M.IPFV-Q

‘At what time do you (PL) leave (and) enter (into the compound of the bridegroom)?’ (BI2012-01-21_002_004350995-004352374) (43) nu

wol’[i]

a

galass

shiikk’-ar

igint-ane?

1PL

together

what

day

gather-CNV2

be_reconciled-1PL.JUS.Q

‘When (lit. ‘on what day?’) should we gather and be reconciled?’ (LS100441_000114240-000116439)

The TIME interrogative of Baskeet is cognate to that of other Ometo languages; see, for instance, aide in Koorete (Binyam 2010:47) and ʔandé in Haro (Hirut 2003:181). The general structure of a TIME interrogative in Ometo is as follows: a(V)(n)DV. The first syllable consists of a vowel a(a) or ainitial diphthong plus, in some languages, a phoneme n; the second syllable starts with an alveolar stop d or t and is completed with a vowel. Azeb (2001:215) and Wakasa (2008:544) assume that the second syllable of the TIME interrogative in Maale and Wolaitta originates from the word ‘time’, i.e. wode (Maale) and d-é (Wolaitta), whose cognate is woda in Baskeet. ‘When’ could thus historically go back to ‘what time’.

62

Yvonne Treis

4.4 Quantity The QUANTITY interrogative does not inflect for case. The non-predicate form is aapin (44). When used as the predicate of a question, the interrogative morpheme -a is added to it (45). (44) aapin how_many

birr=bar

woong-ir-a,

pettan-indo?

birr=ICP

trade-M.IPFV-Q

one-F.DEF.ACC

‘For/with how many birr (= Ethiopian currency) does one sell it, one (of them)? (BI2012-01-21_002_002421706-002423231) (45) haykk’-in-onts die-PFV.REL-PL

aapin-a? how_many-Q

‘How many have died?’ (lit. ‘The ones who have died are how many?’) (BI2012-01-21_002_000541114-000542375)

Like many other languages of the area, Baskeet does not differentiate between a COUNT QUANTITY interrogative (‘how many’) and a MASS QUANTITY interrogative (‘how much’). The interrogative aapin belongs to the word class of numerals; as such, it can be the input of the ordinal derivation – marked with the morpheme -(it)tsa – and thus inquire about a RANK (46). (46) ne 2SG

aapin-tsa-a?

–

how_many-ORD-Q

te

hayzittsa

1SG

three\ORD

‘On what rank are you?’ (lit. ‘How maniest are you?’) – ‘I am third.’ (LS100373_000107016-000110387, elicited)

Historically, the QUANTITY interrogative aapin can be assumed to consist of the THING/QUALITY interrogative a ‘what’ (Section 4.2) plus a formative ‘much, many’. In synchronic Baskeet, _pin is not a morpheme. However, it can be traced back to the etymon *(p)puna ‘many, much’, which is part of the numerals from ‘six’ to ‘nine’ in all Ometo languages except Baskeet and Maale (Hirut 2005). The Baskeet QUANTITY interrogative is very similar to that of Wolaitta, ʔááppun-a (Wakasa 2008:362), and Malo, apunu (Hirut 2005:136). The QUANTITY interrogatives of East Ometo (Koorete, Zargulla, Zayse, Haro) can also be analyzed as consisting of two formatives, an element a(V)(n)_ and an element _s’(V)(na); they are thus structurally similar to that of Baskeet. Non-cognate QUANTITY interrogatives are attested in Maale,

Interrogativity in Baskeet

63

wááʔʔí (Azeb 2001:214), and Wolaitta, wókk’- ‘how much, how many’ and wóís-á ‘how much’ (Wakasa 2008:369).23 4.5 Manner The MANNER interrogative in Baskeet is invariable: wozar is the nonpredicate form (47); as the predicate of a question, the interrogative morpheme -a is attached (48).24 (47) wozar how

pirk’-aar-a? open-F.IPFV-Q

‘How do (you) open it?’ (LS100373_000030246-000031426, elicited) (48) S1: ne

wozar peeshk-i?

2SG how

pass_the_day-2SG.PFV.Q

(Greeting in the afternoon or evening:) ‘Have you passed the day well?’ S2: te 1SG

kossh-e,

ne

good-DECL

2SG how-Q

wozar-a?

‘I (passed it) well. How are you?’ (LS100009_000010617-000013601, elicited)

The MANNER interrogative is also frequently applied to inquire about an UTIn Baskeet, one does not ask ‘What did he say?’ but ‘How did he say?’ The latter question is not a question about the subject’s way or manner of speaking but about the content of his/her speech. TERANCE (49).

(49) aash 2F.VOC

wozar

gey-e?

how

say-3M.PFV.Q

‘Hey (lit. ‘you (F)’), what (lit. ‘how?’) did he say?’ (LS100061_000233241-000234635)

23 The w-initial QUANTITY interrogatives of Maale and Wolaitta could, however, be linked to the w-initial MANNER, PLACE, and SELECTION interrogatives in Baskeet and beyond (see next sections). 24 Given that some other interrogative pronouns also start with a formative wo_ (see below), a reviewer has suggested to segment wozar into wo- and -zar. However, so far, there is no synchronic or comparative evidence for the existence and for a possible source of a morpheme -zar.

64

Yvonne Treis

The Baskeet MANNER interrogative is closely related to that of Maale, wozí (Azeb 2001:215), while the East Ometo languages express ‘how’ by a form waydi or similar; see e.g. wáydí in Zayse (Hayward 1990:277). The East Ometo MANNER interrogative is formally close to the Baskeet PLACE interrogative discussed in the next section. The Baskeet MANNER interrogative wozar and the verb wot- ‘do’ have merged to form an interrogative verb wozaart- ~ wozoort- ‘do how, happen how’ (50); recall the merger of a ‘what’ and wot- ‘do’ in Section 4.2 above. The manner interrogative verb inflects like any other verb of the language. It can be used as the main verb in questions (50) or as a subordinate verb, e.g. as an intentional verb or as a converb (51). The interrogative verb wozaart- ~ wozoort- ‘do how’ can govern a direct object (52) in the same way as the verb wot- ‘do’. (50) te 1SG

wozaart-ano? do_how-1SG.JUS.Q

‘What (lit. ‘how’) should I do?/How should I go about?’ (LS100259_000420181-000420905)

(51) aap-in eye-F

wozaart-i

wobb-a?

do_how-CNV1

become_disabled-3F.PFV.Q

‘How did she become blind?’ (lit. ‘Doing how, did she become disabled with respect to the eyes?’) (BI2012-01-21_002_000526827-000527953)

(52) ne 2SG

okar-in

wozaart-iye?

leather_sack-F

do_how-2SG.PFV.Q

‘How did you handle the sack?/What did you do with the sack?’ (LS100362_000158128-000159503)

Apart from wozaart- ~ wozoort- ‘do how’, Baskeet possesses another MANNER interrogative verb woytt- ‘do how’, which seems to be based historically on woy ‘where’ (see Section 4.6) and the verb wot- ‘do’. Synchronically, woytt- is not (or no longer) attested in questions about a PLACE. The etymology may, however, explain the allative marking on the objects of woytt- in (53). The semantic difference between the two MANNER interrogatives wozaart- ~ wozoort- and woytt- is difficult to grasp; it seems, however, that woytt- cannot be used in neutral information questions but only in questions expressing the speaker’s worry, indignation, or irritation (53).

Interrogativity in Baskeet (53) S1: an 2M.VOC

65

neek

woytt-e?

2SG.ALL

do_how-3M.PFV.Q

‘Hey, what’s wrong with you (lit. ‘it does how to you’)?’ S2: taak 1SG.ALL

woytakkay,

taana

assh-ab

do_how\NEG.PFV

1SG.LG

leave-2SG.IMP

‘Nothing’s wrong with me (lit. ‘it does not do how to me’).25 Leave (SG) me alone!’

(Tam2013-06-17_003_000005473-000010859, elicited)

A cognate interrogative verb wóít- ‘do how, do what’ is also found in the related Ometo language Maale.26 Azeb (2001:218) hypothesizes that wóítconsists of the interrogative formative wo_ and the “predicative verb” t_, which have merged into one stem. Wakasa (2008:550) also reports about interrogative verbs in the Ometo language Wolaitta. The verb wóíg- ‘say what’, wáát- ‘do what’, and wáán- ‘become what’ are analogous to the demonstrative verbs yáág- ‘say so’, yáát- ‘do so’, and yáán- ‘become so’. The verb wáát- ‘do what’ – sometimes also translated as ‘do how’ (Wakasa 2008:553) – is cognate to woytt- in Baskeet and wóít- in Maale. 4.6 Place The PLACE interrogative woy ‘where’ is used in its bare form or in combination with the locative derivational morpheme -da (~ -di) > woyda (~ woydi) ‘where’. The following forms are attested: (54) ACC: woy ~ woy-da ~ woy-di ‘where (to)’ GEN: woy-di ‘of where’ ‘from where’ ABL: woy-d-app(o) Irrespective of the presence of the locative derivational morpheme, the PLACE interrogative can be used to ask for a LOCATION (55) or a GOAL (56) and (6).

25 For the use of interrogative pronouns and verbs in non-interrogative contexts see Section 5. 26 The verb is sometimes written wóít-, sometimes wóyt-. Note that Azeb (2001:218) does not speak of an interrogative verb but of a sentential interrogative.

66

Yvonne Treis

(55) ne 2SG

tiin

yel-int-i

dicc’-in-onda

woy-di-ya?

first

bear-PASS-CNV1

grow_up-PFV.REL-F.DEF.NOM

where-LOC-Q

‘Where were you born (and) (where) did you grow up (in the) first (years)?’ (lit. ‘It is where that …?’) (BI2012-01-21_002_000101927-000104531)

(56) aash

ullo

ne

asini

woy

lukk-e?

2F.VOC my_dear 2SG husband where go-3M.PFV.Q

‘Hey, my dear, where has your husband gone to?’ (LS100912a_000001009-000003231, elicited)

The ablative form inquires about the SOURCE or ORIGIN (57). (57) woy-d-app where-LOC-ABL

yey-e? come-3M.PFV.Q

‘Where did he come from?’ (LS100371_000043703-000046757, elicited)

There are formal and functional overlaps between PLACE and SELECTION interrogatives in Baskeet. Historically, the PLACE interrogative woy may go back to a combination of the SELECTION interrogative woo ‘which (M)’ (see next section) and an obsolete locative morpheme -(a)y, which is also attested in hay ‘here’, geday ‘over there’, sekay ‘over there (far)’, yetay ‘down there’, and lokay ‘up there’. Apart from this diachronic link, there are two synchronic links between PLACE and SELECTION interrogatives. Firstly, the PLACE interrogative pronoun functions as a SELECTION interrogative, if it is used as a modifier (‘of where’ = ‘which’) (58). (58) ne 2SG

woy-di

gab-ipp

yeer-a?

where-LOC

market-ABL

come.F.IPFV-Q

‘From which (lit. ‘of where’) market do you come?’ (LS100248_00013056-000014690)

Secondly, Baskeet can also derive a PLACE interrogative pronoun wonda ~ wondi ‘where’ from the SELECTION interrogative wonni ‘which (F)’ (see next section) by attaching the – synchronically fairly productive – locative derivational morpheme -da (~ -di). The derived PLACE pronoun is attested in the following case forms:

Interrogativity in Baskeet

67

(59) ACC: won-da ~ won-di ‘where (to)’ ‘from where’ ABL: won-d-app(o) Formal and functional overlaps between PLACE and SELECTION interrogatives are also observed in other Ometo languages (and beyond that in other Ethiopian languages). Wakasa (2008:543) reports about the use of the PLACE interrogative in modifying function to inquire about a SELECTION (‘of where’ for ‘which’) in Wolaitta. Examples in Azeb (2007:207) and Hirut (2003:188 f.) show that SELECTION interrogatives are based on PLACE interrogatives in Zargulla and Haro. In Maale, the PLACE interrogative wo-ka ‘where’ and the SELECTION interrogatives, e.g. wo-yí ‘which (M.NOM)’, are based on the same pronominal stem (Azeb 2001:215). Functional differences between woy(da) ‘where’ (54) and wonda ‘where’ (59) could so far not be determined. Compare (55) and (60), in which the speaker inquires about someone’s place of birth with woyda and wonda, respectively. Wonda can equally be used to ask for a GOAL (61). (60) wondi where

yel-int-in-a,

wondi

dicc’-in-a?

bear-PASS-PRF-Q

where

grow_up-PRF-Q

‘Where have (you) been born? Where have you grown up?’ (BI2012-01-21_002_000034971-000036474)

(61) deyshi

na’-i

goat_sheep child-NOM

taani

dakk-in

miiz-i

na’-i

1SG.LG

chase-CNV.DS

cow-GEN

child-NOM

taani

dakk-in

te

wonda-a?

1SG.LG

chase-CNV.DS

1SG

where-Q

‘When the goat/sheep children (i.e. shepherds) chase me and when the cow children (i.e. shepherds) chase me, where (do) I (escape to)?’ (LS100063_000534114-00536269)

It is interesting to note that the Baskeet PLACE interrogatives in (54) are not cognate with the PLACE interrogatives of most other Ometo languages, which possess PLACE interrogatives with an initial formative a(V)(n)_; see Wolaitta ʔáu ‘where (to)’ (Wakasa 2008:542), Koorete aya ‘where’ (Binyam 2010:46), Zayse ʔáná ‘where’ (Hayward 1990:271), and Maale ʔánko (Azeb 2001:215). Only Baskeet woy and Maale wo-ka ‘where’ (-ka LOC) (Azeb 2001:215) are cognate.

68

Yvonne Treis

4.7 Selection Like demonstratives, the SELECTION interrogatives (‘which’) inflect for gender, number, and case in Baskeet. The following forms are attested in my corpus; the gaps are accidental: (62) ACC: NOM: GEN: DAT: ABL: ALL:

MASCULINE woo ~ wooda ~ woodi woo ~ woodi n.a. woob(o) woopp(o) n.a.

FEMININE wonni n.a. wonni wonn-ab(o) wonn-app(o) n.a.

PLURAL wonts n.a. n.a. wonts-ab(o) wonts-app(o) n.a.

The use of the ablative form is exemplified below: (63) S1: te 1SG

yinni

naar-aa sekii=gall-appo

ekk-ash-e

DEM2.F

boy-M

take-RCT-DECL

DEM4.M=top-ABL

‘I have just taken this from the boy over there.’ S2: naar-aa boy-M

yii

wooppo?

karts-a-appi-be

DEM2.M

which.M.ABL

black-M-ABL-DISJ.Q red-M-ABL

aars-a-appo?

‘From which boy? From the dark or the fair-skinned one?’ (Tam2013-06-17_004-000101414-000109061, elicited)

Additional examples from the text corpus illustrate the use of the plural form in direct object function (64) and of the masculine form in predicate function (65). (64) […]

ne

yints=gart-app

wonts

wonts

er-aar-a?

2SG

DEM2.PL=inside-ABL

which.PL

which.PL

know-F.IPFV-Q

‘[People say that there are different styles,] which ones (lit. ‘which ones and which ones’) of them do you know?’ (ML2012-01-10_001_001403898-001405485)

(65) zay-in

iit-aa-pen wot-egeyttsi c’add-i

trumpet-F bad-M-SIM do-COND

mah-iz-a

stab-CNV1 return-M.IPFV.REL-M

woodi-ya? which.M-Q

‘If the trumpet plays a mistake, which one (of the trumpet players) rearranges (the song)?’ (BI2012-01-21_002_002143706-002145778)

Interrogativity in Baskeet

69

If the SELECTION interrogative is combined with the locative derivation -da, it turns into a PLACE interrogative won-da ‘where (to)’ (cf. Section 4.6). Not much data is available on SELECTION interrogatives in other Ometo languages. If SELECTION interrogatives are mentioned in the literature, they can clearly be shown to be based on PLACE interrogatives, as is the case in Wolaitta (Wakasa 2008:543), Zargulla (Azeb 2007:207), and Haro (Hirut 2003:188 f.). 5 Non-interrogative uses of interrogative pronouns Baskeet uses interrogative pronouns not only in questions but also in declarative sentences. Like in many languages of the world, interrogatives are used as indefinite pronouns (Section 5.1). In addition, interrogatives can be used in one type of assertive construction (Section 5.2). 5.1 Interrogative pronouns as indefinite pronouns In Baskeet, interrogative pronouns form the basis of (i) realis indefinites (cf. some-indefinites in English), (ii) free-choice indefinites (cf. any-indefinites in English), and (iii) indefinites in negative sentences (cf. no-indefinites in English). In these functions, the interrogative pronouns usually combine with additional morphemes.27 As realis indefinites (66)–(69), the interrogative pronouns receive an additional morpheme -(V)k(o), which is (very likely to be) identical to and functionally derived from the conditional morpheme – even though the grammaticalization path is not yet entirely clear.28 Example (67) illustrates the use of -(V)k(o) in its function as a conditional clause marker (first occurrence) and in its function as a marker of indefinite pronouns (second occurrence). (66) ab-ak

taani

iishicc-ash

what-COND 1SG.LG frighten-RCT

‘Something has scared me.’ (LS100533_000005101-000009902, elicited) 27 There are a few exceptions: In the periphrastic expression a ga-akkay /what sayNEG.PFV/ ‘without being hungry’ (lit. ‘without saying what(ever)’), the indefinite pronoun a ‘what(ever)’ carries no additional morpheme. 28 For cross-linguistic (near) parallels, i.e. languages in which free-choice indefinite pronouns include concessive conditional marking, see Haspelmath (1997:135–140). In contrast to the languages discussed there, Baskeet uses a conditional morpheme on all types of interrogative-based indefinite pronouns. Concessive conditional clauses as well as negative and free-choice indefinite pronouns receive, in addition to the conditional morpheme, an additive morpheme (‘also, even’); see below.

70

Yvonne Treis

(67) pil’is-in gub’i=bar-ke toh=bar koc-int-in-o-ko chalice-F lid=ICP-and

foot=ICP

make-PASS-PFV.REL-M-COND

iz=garta

pila

disti

ab-ak

gadd-i

mu-ir-e

3F=inside

cheese

sauce

what-COND

put-CNV1

eat-M.IPFV-DECL

‘If the chalice is made with a lid and a foot, one keeps cheese, sauce, and anything (else) in it and eats (from it).’ (LS100865_000118273-000124998)

(68) oon-ak

te

apil-in

eh-akka!

who-COND 1SG clothes-F take-SURP

‘Someone has taken my clothes!’ (LS100533_000137806-000143067, elicited)

The final vowel of the conditional morpheme -(V)k(o) is only realized before pauses, e.g. when the indefinite pronoun is used sentence-finally in predicate function (69). The conditions for the occurrence and the quality of the initial vowel (i or a) – symbolized here by V – are still unknown. (69) S1: yii DEM2.M

oono? who.Q.PRED

(A voice is heard from outside:) ‘Who is this?’

S2: i 3M

oon-iko who-COND

‘It’s someone.’ (i.e. I don’t know this voice.)

(Tam2013-06-17_001_000000000-000004093, elicited)

If the interrogative is followed by a relational enclitic, the conditional morpheme moves to the right of the enclitic’s case marker; see (70) and (72).29 (70) o=gall-ap-k

ekk-ini-shoa

who=top-ABL-COND take-PRF-ASS

‘He has taken it from someone!’ (LS100533_000228600-000233681, elicited)

To generate free-choice (71) or negative indefinites (72), the additive morpheme -ar ‘also, even’ is suffixed to the conditional morpheme. Note that the

29 Note that one example in Sottile (2002:122) contains the form ab-ak=bar ‘with something’, which contradicts the morpheme order rule formulated here. The form abak=bar is, however, not attested in my corpus.

Interrogativity in Baskeet

71

additive morpheme also marks conditional clauses as concessive, i.e. freechoice, in Baskeet (73). (71) ne 2SG

iib

ab-ak-ar

imm-in

ts’oossi

3M.DAT

what-COND-ADD

give-CNV.DS

God

imm-o

ga-abas

give-3.JUS

say-M.NEG.IPFV

‘Whatever you give to him, he won’t say “thank you (lit. ‘may God give’)”.’ (LS100533_000041040-000044052, elicited) (72) o=bar-k-ar

lukapp-e

who=ICP-COND-ADD

go\2SG.NEG.IMP-DECL ‘Don’t go with anybody!’ (LS100533_00115007-000118890, elicited)

(73) i

ye-ibaas-k-ar

nu iy

ts’eyg-ibaas-e

3M come-M.NEG.IPFV-COND-ADD 1PL 3M call-M.NEG.IPFV-DECL

‘Even if he doesn’t come, we won’t call him.’ (LS100370_000004075-000007158, elicited)

Apart from interrogatives, generic nouns such as asi ‘person’, baz ‘thing’, and bessi ‘place’ can function as indefinites. In their function as free-choice or negative indefinite pronouns, generic nouns do not combine with the conditional morpheme, but – like interrogatives – they receive the additive morpheme -ar. An utterance like ‘I don’t have anybody’ could thus be rendered in Baskeet either as ‘to_me who-COND-ADD be_not’ (interrogative as indefinite pronoun) or as ‘to_me person-ADD be_not’ (generic noun as indefinite pronoun) (74).30 (74) taab 1SG.DAT

as-ar

bayy-e

person-ADD

be_not-DECL

‘I don’t have anybody.’ (lit. ‘Even a person is not to me.’) (LS100048_000044975-000045866)

The use of interrogatives as indefinite pronouns is also attested in related Ometo languages, among them Maale (Azeb 2001:230), Koorete (Binyam 2010:154), Haro (Hirut 2003:201), and Wolaitta. Most examples given in the 30 Similarly, ab-ak ‘something’ (what-COND) is interchangeable with petti baz ‘something’ (lit. ‘one thing’).

72

Yvonne Treis

respective grammatical descriptions illustrate the use of interrogatives in negative sentences. There the interrogatives combine – like in Baskeet – with an additive morpheme.31 Realis indefinite pronouns in Wolaitta consist of an interrogative pronoun plus the conditional morpheme -kko ‘if’ (75). The multifunctionality of -kko in Wolaitta rules out that the identity of the conditional morpheme and the final morpheme of realis indefinites in Baskeet could be coincidental. Free-choice indefinites in Wolaitta are also based on interrogatives to which the additive morpheme -kká is attached, either directly (76) or to the clause in which they occur (Wakasa 2008:532). The same morpheme is found on free-choice conditional clauses.32 (75) ʔóóna-kko

y-íis

who.ABS-COND

‘Someone came.’ (76) ʔóóní-kká who.NOM-ADD

WOLAITTA

come-3M.SG.PFV

(Wakasa 2008:435; glosses adjusted)

he

wott-áa

danday-ées

that

running-M.SG.ABS

can-3M.SG.IPFV

WOLAITTA

‘Anyone can run that running.’ (Wakasa 2008:435; glosses adjusted) Indefinites in negative contexts are either marked by -kká (Wakasa 2008:163; example 32) or another additive morpheme -nné (Wakasa 2008:432; footnote 278) in Wolaitta. 5.2 Interrogative pronouns in assertive constructions Baskeet possesses a periphrastic construction which expresses that something is blatantly obvious, incontestable, or irrefutable. The assertive construction is often used by speakers in response to a question which they consider unnecessarily asked – either because the person who has asked the question could have easily found out the answer by himself/herself or because the answer to the question is general world knowledge. The assertive construction consists of a clause33 including a dubitative morpheme -dok attached to the predicate and a postposed interrogative pronoun. In (77)–(79) the predicate is a THING, a PLACE, and a TIME interrogative, respectively. 31 Note that this morpheme is sometimes called “inclusive morpheme” in the sources consulted. 32 It is not clear whether Wolaitta -kká could be further segmented into a conditional morpheme -kko plus an additive morpheme proper as in Baskeet. 33 Usually, the assertive construction is based on a nominal clause. So far, only one verbal clause is attested in such a construction in my corpus.

Interrogativity in Baskeet (77) S1: yii

abz-a?

DEM2.M

S2: yii

what-Q

DEM2.M

‘What is this?’

73

buy-dok

abz-a

yam-DUB

what-Q

‘This is a yam!’ (You should know that! This is so obvious.)

(LS100533_000008646-000012113, elicited)

(78) S1: ne

woy-a?

2SG

S2: gabi-dok

where-Q

market-DUB

‘Where are you (going) to?’

woy-a where-Q

‘To the market, of course!’

(LS100498_000016571-000019093, elicited)

(79) S1: sibsaabi aant-iya? meeting

S2: sibsaabi k’eeri-dok

when-Q

meeting

‘When is the meeting?’

Saturday-DUB

aant-a when-Q

‘The meeting is on Saturday!’ (You should know that! It was announced everywhere.)

(LS100498_000040439-000044582, elicited)

Although the assertive construction contains an interrogative pronoun, its intonational pattern is clearly different from that of questions. Questions with the interrogative pronoun in clause-final position are characterized by a pitch rise that peaks on the stem of the interrogative pronoun, e.g. on aant- in (79, S1), and a subsequent pitch fall on the last syllable of the utterance. In the assertive construction (79, S2), there is no question-like pitch rise. Instead the pitch falls on the morpheme -dok and the final interrogative pronoun has the lowest pitch value of all the elements in the utterance. The interrogative pronoun of the assertive construction does not constitute an intonational phrase of its own; there is no pause between the morpheme -dok and the interrogative. The assertive construction is not only common as an answer to unnecessary questions; it is also attested elsewhere as a signal of incontestable statements. The statements in (80), to which the singer expects no objection, are taken from a praise song. (80) baskeet yizi Baskeet

thus

maaki-dok ab-a

baskeet

yizi zobb-dok

ab-a

leopard-DUB

Baskeet

thus lion-DUB

what-Q

what-Q

‘The Baskeet are leopards! The Baskeet are lions!’ (LS100595_000216005-000220950)

74

Yvonne Treis

The examples in (81) are used by the protagonist of a song to remind the listeners of features that are well-known to all Baskeet speakers. (81) gaari kaat yizi Gaara

king

thus

gembi

kaat-dok

ab-a

type_of_sorghum

king-DUB

what-Q

gaari

kaat

yizi

gaashi

kaat-dok

ab-a

Gaara

king

thus

tef

king-DUB

what-Q

‘(It is obvious that) the king of Gaara is the king of gemba-sorghum. (There is no doubt that) the king of Gaara is the king of tef.’ (LS100593_000234052-00239205)

6 Summary and outlook Baskeet has been shown to be one of the Omotic languages with inflectional interrogative marking. Interrogative inflectional morphology is, however, only found in some verbal paradigms (the perfective and jussive paradigms, in particular). Verb forms of other paradigms can simply be marked by the appropriate question intonation or by the addition of an interrogative suffix -a (or its variant -(i)ya) when they are used in questions. Apart from the presence of a question word, no formal distinction is made between content and polar questions in Baskeet. In addition to the morphology used to mark canonical questions, Baskeet has three other interrogative morphemes at its disposal: The morpheme -ish marks follow-up questions. The morpheme -be occurs as a disjunctive morpheme (‘or’) in alternative questions and as a marker of confirmation questions. The morpheme -baya requests a confirmation as well. In the interrogative pronoun system of Baskeet a distinction is made between the semantic categories of PERSON, THING, TIME, QUANTITY, MANNER, PLACE, and SELECTION. Most interrogative pronouns are case-marked like other nouns and pronouns in the language. As in many languages of the world, the semantic category REASON is expressed by a morphologically analyzable interrogative, namely the dative form of THING, i.e. ‘for what’ = ‘why’. QUALITY is inquired about by a THING interrogative in modifying function, i.e. ‘what N’ = ‘what kind/type of N’. The RANK interrogative is derived from the QUANTITY interrogative by an ordinal derivational suffix. Apart from having a dedicated PLACE interrogative, a locative-derived SELECTION pronoun can also be used to ask for a location. Apart from having a dedicated SELECTION interrogative, a PLACE interrogative in modifying function can express SELECTION, i.e. ‘of_where N’ = ‘which N’. Formal over-

Interrogativity in Baskeet

75

laps between SELECTION and PLACE interrogatives are also common in other Ometo languages. Although the abovementioned six simplex interrogatives of Baskeet cannot be analyzed further from a synchronic point of view, certain phonological resemblances are apparent between them. These resemblances may point to a common historical origin. The Baskeet interrogatives can be categorized into three groups: (i) the PERSON interrogative oona ~ ooni, which shows no phonological resemblance to any other interrogative, (ii) the THING, TIME, and QUANTITY interrogatives, which start with a formative a(a)_ and (iii) the MANNER, PLACE, and SELECTION interrogatives, which start with a formative wo_. Note that these formatives are not morphemes synchronically. The same formatives are, however, found in functionally equivalent interrogative pronouns in other Ometo languages. The only major difference between Baskeet and the other Ometo languages concerns the initial formative of PLACE interrogatives, which is wo_ in Baskeet and Maale but a(V)(n)_ elsewhere. The analysis of interrogativity in Baskeet is still on-going. The present article could only give a preliminary overview of interrogative marking on predicates in Section 3 and of the morphology and use of interrogative pronouns in Section 4 and 5. Gaps in the description remain, even for these domains, because some grammatical forms are not (yet) attested in my corpus of recorded texts, which are in the process of being transcribed, glossed and translated. Other important aspects of interrogativity, especially the analyses of how questions are marked prosodically and of how subordinate (indirect) question are marked, have to be dealt with in future publications. Abbreviations ABL Ablative ACC Accusative ADD Additive AFF Affirmative ALL Allative CAUS Causative COND Conditional CONF.Q Confirmation question CNV1 Converb in -i CNV2 Converb in -ar CNV.DS Different subject converb in -in DAT Dative

DECL DEF DEM1 DEM2 DEM4 DISJ.Q DS DUB F GEN ICP IMP

Declarative Definite Demonstrative, proximal Demonstrative, invisible Demonstrative, distal Disjunction in questions Different subject Dubitative Feminine Genitive Instrumental, comitative, perlative Imperative

76 INT IPFV JUS LOC LG M

n.a. NEG NMLZ NOM ORD PASS PFV

Yvonne Treis Intentional Imperfective Jussive Locative Long (emphatic) pronoun Masculine Not attested Negation Nominalizer Nominative Ordinal Passive Perfective

PL PRED PRF Q RCT REL SG SIM SIMUL SUPPL.Q SURP VOC

Plural Predicative Perfect Interrogative, question Recent Relative Singular Similative Simultaneity Supplementary question Surprise Vocative

References Azeb Amha. 2001. The Maale Language. Leiden: LOT. Azeb Amha. 2007. Questioning forms in Zargulla. In Rainer M. Voigt (ed.), From Beyond the Mediterranean: Akten des 7. Internationalen Semitohamitistenkongresses, Berlin 13-15 September 2004, 197– 210. Aachen: Shaker. Azeb Amha. 2012. Omotic languages. In Zygmunt Frajzyngier & Erin Shay (eds.), The Afroasiatic Languages, 423–504. Oxford: Oxford University Press. Binyam Sisay Mendisu. 2010. Aspects of Koorete Verb Morphology. Cologne: Köppe. Cysouw, Michael. 2004. Interrogative words: An exercise in lexical typology. Handout presented at the seminar “Bantu Grammar: Description and Theory 3”, session on question formation in Bantu, ZAS Berlin, 13 February 2004. Federal Democratic Republic of Ethiopia Population Census Commission. 2008. Summary and Statistical Report of the 2007 Population and Housing Census: Population Size by Age and Sex. Addis Ababa. http://www.csa.gov.et/pdf/Cen2007_firstdraft.pdf (22 August 2012). Haspelmath, Martin. 1997. Indefinite Pronouns. Oxford: Clarendon. Hayward, Richard J. 1982. Notes on the Koyra language. Afrika und Übersee 65. 211–268.

Interrogativity in Baskeet

77

Hayward, Richard J. 1990. Notes on the Zayse language. In Richard J. Hayward (ed.), Omotic Language Studies, 210–355. London: SOAS. Hirut Woldemariam. 2003. A Grammar of Haro with Comparative Notes on the Ometo Linguistic Group. Addis Ababa: Addis Ababa University (PhD Dissertation). Hirut Woldemariam. 2005. Historical notes on numerals in Ometo: The obsolete quinary system. Journal of Ethiopian Studies 38(1/2). 131–150. Idiatov, Dmitry. 2007. A Typology of Non-Selective Interrogative Pronominals. Antwerpen: Universiteit Antwerpen (PhD Dissertation). Inui, Hideyuki. 2005. A grammatical survey of the Basketo language [in Japanese]. Cushitic–Omotic Studies 2004. 1–40. Inui, Hideyuki. 2006. A grammatical survey of the Basketo language II [in Japanese]. Cushitic–Omotic Studies 2006. 15–60. Inui, Hideyuki. 2012. Basic verb sentences for Ethiopian language research [in Japanese]. Cushitic–Omotic Studies 2012. 48–211. Köhler, Bernhard. 2013. Interrogative zero-marking in some Ometo languages. In Marie-Claude Simeone-Senelle & Martine Vanhove (eds.), Proceedings of the 5th International Conference on Cushitic and Omotic Languages, Paris, 16–18 April 2008, 135–155. Cologne: Köppe. Leslau, Wolf. 1995. Reference Grammar of Amharic. Wiesbaden: Harrassowitz. Siebert, Ralph & Simon Caudwell. 2002. Sociolinguistic Survey Report of the Melo (Malo) and Mursi Languages of Ethiopia. Dallas, TX: SIL International. Sottile, Roberto. 2002. Schizzo grammaticale del Basketo (Etiopia sudoccidentale). Naples: Università degli Studi di Napoli “L’Orientale” (PhD Dissertation). Treis, Yvonne. 2008. A Grammar of Kambaata. Part 1: Phonology, Nominal Morphology, and Non-verbal Predication. Cologne: Köppe. Treis, Yvonne. 2011a. Names and naming: Baskeet. Basket to Ethiopia. http://baskettoethiopia.wordpress.com/2011/10/31/names-andnaming-1 (31 October 2011). Treis, Yvonne. 2011b. Names and naming: Basketo (and other variants). Basket to Ethiopia. http://baskettoethiopia.wordpress.com/2011/ 10/31/names-and-naming-2 (31 October 2011).

78

Yvonne Treis

Treis, Yvonne. 2011c. Names and naming: Mesketo (and other variants). Basket to Ethiopia. http://baskettoethiopia.wordpress.com/2011/ 11/05/names-and-naming-3 (5 November 2011). Treis, Yvonne. 2012. Names and naming: Dookka and Doolla are not Baskeet “dialects”. Basket to Ethiopia. http://baskettoethiopia. wordpress.com/2012/09/05/names-and-naming-dookka-and-doollaare-not-baskeet-dialects/ (5 September 2012). Wakasa, Motomichi. 2008. A Descriptive Study of the Modern Wolaytta Language. Tokyo: University of Tokyo (PhD Dissertation).

Complex Predicates in Amharic Counterfactual Antecedent Clauses Abdu Ahmed Addis Ababa University Abstract Amharic distinguishes between two canonical counterfactual antecedent clauses in conditional constructions, namely clauses based on an imperfective verb and clauses based on a converb. This paper aims to describe the morphosyntactic features and the semantic composition of complex predicates in these two types of antecedent clauses. Complex predicates either consist of a converb plus a superordinate verb or a coverb plus a light verb. 1 Introduction Prototypically, conditional constructions consist of an adverbial clause, often referred to as the conditional clause, antecedent or protasis, and a main clause, called consequent or apodosis (Bhatt & Pancheva 2005:639; Kaufmann 2006:6; Payne 1997:316). The precedence of the antecedent over the consequent and the fact that the former represents the morphosyntactically marked part of a conditional construction are considered a defining feature of conditional constructions cross-linguistically (Bhatt & Pancheva 2005:642). In conditional constructions, markers of conditionality can either be affixes attached to the subordinate verb or free morphemes in the subordinate clause. Regarding their meaning, Bhatt & Pancheva (2005:639) state that conditional clauses are interpreted “with the proposition expressed by the antecedent clause specifying the (modal) circumstances in which the proposition expressed by the main clause is true.” Although it is not a sharp distinction, most languages basically distinguish between real and unreal conditionals (Thompson, Longacre & Hwang 2007:255). Contrary to the fact that a large number of linguistic works have been dedicated to various aspects of Amharic grammar, little has so far been written on Amharic conditionals. Apart from Abdu (2011), Hartmann (1980) and Leslau (1995) are the only linguistic works that have addressed certain aspects of Amharic conditionals. However, it is necessary to note the descriptive inadequacy of these two studies: They do not clearly define important concepts such as conditionals, real conditionals, unreal conditionals, and

80

Abdu Ahmed

hypothetical conditionals, and they gloss and analyze the data in such a way that the structural components of conditional sentences cannot be easily identified. The present contribution is based on Abdu (2011) but concentrates only on counterfactual conditionals.1 Since I am a native speaker of Amharic, the data was mainly collected through introspection. Prototypical conditional constructions in Amharic have a conditional marker in the antecedent clause: kǝ- occurs with verbs in the perfective aspect and marks a real condition, whereas b- is prefixed to verbs in the imperfective aspect and marks an unreal condition (cf. Abdu 2011:32–35, 38–39). Within the realm of unreal conditions, counterfactuals differ from hypothetical conditionals in that they mostly contain the invariant noro as a marker of counterfactuality in the antecedent clause (cf. Abdu 2011:40 f.). In addition, the consequent clause is marked by the past-tense auxiliary nǝbbǝr (cf. Abdu 2011:64 ff.). (1) a.

k’ǝss bɨlo binǝda noro adǝgaw ajdǝrsɨmm nǝbbǝr [k’ǝss bɨlo IDPH.slow say\CNV.3SG.M [adǝga-w accident-DEF

b-j-nǝda

noro]antecedent

COND-3SG.M-drive\IPFV

AUX.CF

a-j-dǝrs-mm NEG-3SG.M-reach\IPFV-NEG

nǝbbǝr]consequent AUX.PST

‘If he had driven slowly, the accident would not have occurred.’ b.

k’ǝss bɨlo nǝdto bihon noro adǝgaw ajdǝrsɨmm nǝbbǝr [k’ǝss bɨlo IDPH.slow say\CNV.3SG.M [adǝga-w accident-DEF

nǝdto drive\ CNV.3SG.M

a-j-dǝrs-mm NEG-3SG.M-reach\IPFV-NEG

b-j-hon

noro]antecedent

COND-3SG.M-

AUX.CF

be(come)\IPFV nǝbbǝr]consequent AUX.PST

‘If he had driven slowly, the accident would not have occurred.’ Example (1) shows the two canonical types of counterfactual constructions in Amharic. These types are similar in meaning but differ in the composition of their antecedent clauses (see Section 2). There is another, perhaps more archaic, way of constructing counterfactual conditionals, in which the verbs in the antecedent as well as in the conse-

1

Cf. Abdu (2011:24 ff.) for an overview of the various conditional clause types in Amharic.

Complex Predicates in Amharic Counterfactual Antecedent Clauses

81

quent clause occur in the perfective aspect and are both preceded by the locative prefix bǝ-, as in (2). (2) k’ǝss bɨlo bǝnǝdda adǝgaw baldǝrrǝsǝ nǝbbǝr k’ǝss IDPH.slow

bɨlo say\

LOC-drive\

bǝ-nǝdda

CNV.3SG.M

PFV.3SG.M

adǝga-w accident-DEF

bǝ-al-dǝrrǝs-ǝ LOC-NEG-

nǝbbǝr AUX.PST

reach\PFV-3SG.M

‘If he had driven slowly, the accident would not have occurred.’ Due to its pragmatic markedness and overall rarity, this type of counterfactual construction will not be considered here.2 Constructions combining a converb or a coverb3 with another verb are usually considered complex predicates (Baker & Harvey 2010:13). However, there is no single, commonly accepted, all-inclusive characterization of complex predicates in the linguistic literature. The various definitions (cf. Alsina, Bresnan & Sells 1997; Mengistu, Baker & Harvey 2010) focus on monoclausality and the involvement of two or more predicating morphemes. With regard to Amharic, the combination of a converb plus another verb is a widely employed mechanism to form complex predicates (cf. Azeb & Dimmendaal 2006:429), as in k’ort’o tǝnǝssa [cut\CNV.3SG.M rise\PFV.3SG.M] ‘make a decision’, dǝrso tǝmǝllǝsǝ [reach\CNV.3SG.M return\PFV.3SG.M] ‘come back’ (see Desalegn 2012:2 for further examples of verbs that frequently occur in complex predicates). Note, however, that not all sequences of a converb plus another verb are complex predicates.4 Semantically, the converb in a converb-plus-verb combination in Amharic can have two meanings: it can express a simultaneous event or a consecutive event (cf. Meyer 2012:172 ff.; but also Azeb & Dimmendaal 2006:412 f.). In the constructions that express simultaneous events, the converb serves as adverbial modifier, which is the case in counterfactual antecedent clauses, whereas in constructions that encode consecutive events, the converb is used as a clause-

2 3 4

See Abdu (2011:65) for further details. In the literature, the first invariant component in complex predicates is also referred to as particle, preverb, ideophone, etc. As a matter of personal choice, the term coverb is used in this study. Generally, there is no clear-cut boundary between converb-plus-verb constructions and complex predicates in Amharic (cf. Meyer 2012:175 ff. for more details). Complex predicates based on converb constructions are often lexicalized in such a way that their meaning cannot be derived from the two component verbs. Syntactically, the two component verbs usually occur as a fixed unit, but see Section 3.1 for exceptions.

82

Abdu Ahmed

chaining device. Converbs in Amharic occur in the completive aspect, which is semantically related to the perfective aspect.5 The combination of a coverb plus a light verb is another common way to form complex predicates in Amharic. The light verbs in such constructions are inflected forms of alǝ ‘say’ or adǝrrǝgǝ ‘do, make’. Baye (2008:199 ff.) makes a distinction between “simple” vs. “complex” coverb constructions (cf. also Baye 1999:23; Tadesse 1972), which I use here. In simple constructions, the coverb is a synaesthetic or conventional ideophone, which is derived from a regular lexical root (cf. also Wetter 2003:263 ff.). These synaesthetic ideophones convey either an intensive or an attenuative meaning, as in sɨbbɨrr alǝ ‘break with force’ (intensive) vs. sǝbǝrr alǝ ‘break lightly’ (attenuative); they are both derived from the root √sbr ‘break’, which also occurs in the (semantically neutral) verb sǝbbǝrǝ ‘break’. Complex coverb constructions contain any kind of ideophone (imitative, synaesthetic or conventional) that are not derived from lexical verbs, as in k’utʧ’ alǝ ‘sit down’ or zɨmm alǝ ‘be quiet’. Furthermore, coverbs can be totally or partially reduplicated to encode iterative events, like fǝt’ǝnn fǝt’ǝnn alǝ ‘hurry up’ or ʧ’ɨnk’ɨnk’ɨnk’ alǝ ‘get overcrowded’. This paper is only concerned with complex predicates in canonical counterfactual antecedent clauses in Amharic, as shown in (1). It aims to analyze the structural composition of converb- and coverb-based complex predicates in these clauses. In Section 2, the structure of counterfactual antecedent clauses is described. Then converb-plus-verb and coverb-plus-verb predicates of this clause type are discussed in detail in Section 3. Section 4 summarizes the main points of the discussion. 2 Counterfactual antecedent clauses The Amharic counterfactual antecedent clause often contains a series of different verb forms that constitute a complex predicate. This makes counterfactual antecedents an interesting study object regarding the syntactic composition of their complex predicates. As mentioned in Section 1, the canonical counterfactual antecedent clause in Amharic occurs in two syntactic types with identical meaning. In the first type, the semantic core of the predicate is encoded by a verb in the imperfec5

The term completive aspect was introduced by Baye (2006:197) to refer to a dependent verb form that is commonly known as gerund (cf. Leslau 1995:355 ff.) or converb (cf. Hetzron 1972:99 f.) in Amharic scholarship. In contrast to other studies on this verb form, which commonly assume that it is devoid of TAM features (cf., for instance, Meyer 2012), Baye (2006:197) considers it a (semantic) subcategory of the perfective aspect. I follow Baye’s (2006) analysis in this study.

Complex Predicates in Amharic Counterfactual Antecedent Clauses

83

tive aspect, in the second type by a converb that is followed by an auxiliary in the imperfective aspect. The two types have the following syntactic structure: (3) Structure of Canonical Counterfactual Antecedent Clauses in Amharic a.

Imperfective-based counterfactual antecedent clause: (NPSJ) (NPOJ) COND-AGR.SJ-VERB\IPFV(-AGR.OJ) AUX.CF

b. Converb-based counterfactual antecedent clause: (NPSJ) (NPOJ) VERB\CNV-AGR.SJ(-AGR.OJ) COND-AGR.3SG.M-be(come)\IPFV AUX.CF

Both canonical counterfactual antecedent clauses contain the conditional conjunction b- (COND) and the counterfactuality auxiliary noro (AUX.CF) as the final element. Note that the counterfactuality auxiliary noro itself represents the converb form of the verb norǝ ‘live’ marked for an invariable 3SG.M subject. This counterfactuality auxiliary must always be preceded by a verb in the imperfective aspect to which the conditional conjunction b- is attached. Formally, the composition of the verb in counterfactual antecedent clauses mirrors that of counterfactual consequent clauses: an inflected verb in the imperfective aspect is followed by an invariable auxiliary verb, i.e. the pasttense auxiliary nǝbbǝr in the consequent clause and the counterfactual auxiliary noro in the antecedent clause. The auxiliary nǝbbǝr is diachronically also related to the main verb norǝ ‘live’ (cf. Podolsky 1991:48). This is still reflected in certain types of subordinate clauses (including converb clauses), in which the auxiliary nǝbbǝr is replaced by inflected forms of the verb norǝ. The verb bihon, i.e. the invariable 3SG.M imperfective form of the auxiliary honǝ ‘be(come)’ preceded by the conjunction b- in converb-based counterfactual clauses, cf. (1b), (3b),6 encodes – in addition to a counterfactual condition – uncertainty of the speaker with regard to the verbal event denoted by the converb (for more examples, cf. Abdu 2011:40 f.). The two canonical types k’ǝss bɨlo binǝda noro and k’ǝss bɨlo nǝdto bihon noro in (1) may be compared with two other conditional antecedent types that are closely related. These are the imperfective-based k’ǝss bɨlo binǝda and the converb-based k’ǝss bɨlo nǝdto bihon in (5), which lack the auxiliary noro. 6

Note that in the converb-based counterfactual antecedent construction, the subject agreement is twofold: variable agreement on the converb in accordance with the subject of the clause but invariable 3SG.M subject agreement on the auxiliary honǝ referring to an expletive subject (cf. Abdu 2011:42 ff.).

84

Abdu Ahmed

The latter types of conditional clauses tend not to predict or evaluate what is to happen but rather to express a wish, especially if they contain temporal adverbials of future reference, as in (4). Thus, they constitute hypothetical – not counterfactual – conditional antecedent clauses. (4) a. bɨzu gǝnzǝb bagǝññ mǝkina ɨgǝza nǝbbǝr bɨzu gǝnzǝb many money

b-j-agǝññ COND-1SG-find\IPFV

mǝkina car

ɨ-gǝza 1SG-buy\IPFV

nǝbbǝr AUX.PST

‘If I got a lot of money, I would buy a car.’ b. nǝgǝ bizǝnb wǝdǝ tɨmhɨrt bet alhedɨmm nǝbbǝr nǝgǝ b-j-zǝnb tomorrow COND-3SG.Mrain\IPFV

wǝdǝ to

tɨmhɨrt_bet school

al-ɨ-hed-mm NEG-1SG-go\IPFV-

nǝbbǝr AUX.PST

NEG

‘If it rained tomorrow, I would not go to school.’ The past tense auxiliary nǝbbǝr in the consequent clauses of such hypothetical conditionals does not refer to a state of affairs prior to the moment of utterance but serves the modal function of indicating irrealis (cf. Abdu 2011:60 f.). Antecedent forms lacking the auxiliary noro, as those in (4), necessarily occur with temporal adverbials of past reference in order to render counterfactual readings, as in (5) below. Past-tense reference is indicated either by the auxiliary noro or a temporal adverbial. However, in the absence of either of the two, the converb-based forms, like k’ǝss bɨlo nǝdto bihon in (5b), seem to have a counterfactual meaning more commonly than the imperfectivebased forms, like k’ǝss bɨlo binǝda in (5a). (5) a.

k’ɨddɨm k’ǝss bɨlo binǝda adǝgaw ajdǝrsɨmm nǝbbǝr [k’ɨddɨm earlier

k’ǝss IDPH.slow

[adǝga-w accident-DEF

bɨlo say\CNV.3SG.M

a-j-dǝrs-mm NEG-3SG.M-reach\IPFV-NEG

b-j-nǝda]antecedent COND-3SG.M-drive\IPFV

nǝbbǝr]consequent AUX.PST

‘If he had driven slowly earlier, the accident would not have occurred.’ b.

k’ɨddɨm k’ǝss bɨlo nǝdto bihon adǝgaw ajdǝrsɨmm nǝbbǝr [k’ɨddɨm earlier

k’ǝss IDPH.slow

bɨlo say\CNV.3SG.M

nǝdto drive\CNV.3SG.M

Complex Predicates in Amharic Counterfactual Antecedent Clauses b-j-hon]antecedent COND-3SG.Mbe(come)\IPFV

[adǝga-w accident-DEF

a-j-dǝrs-mm NEG-3SG.Mreach\IPFV-NEG

85

nǝbbǝr]consequent AUX.PST

‘If he had driven slowly earlier, the accident would not have occurred.’ In general, the use of either the imperfective-based antecedents or their converb-based counterparts displays a clear tendency: In combination with the auxiliary noro, the imperfective-based antecedents tend to be preferred for their economy (i.e. shortness). In contexts where noro is left out, however, the converb-based antecedents are preferred because they are usually interpreted as encoding counterfactuality. Imperfective-based antecedents without noro, in contrast, do not show such a preference, i.e. they can either be interpreted as expressing a hypothetical or as expressing a counterfactual reading (cf. Abdu 2011:61 f.).7 3 Complex predicates in Amharic counterfactual antecedent clauses 3.1 Converb constructions When a complex predicate consisting of converb plus superordinate verb occurs in an imperfective-based counterfactual antecedent clause, like (6a), the conditional conjunction b- is prefixed to the superordinate verb but not to the converb. When such a complex predicate occurs in the converb-based counterfactual antecedent clause, the auxiliary honǝ carries the conditional marker, as in (6b). (6) a. aster jǝmɨssɨraʧ jɨza bɨtɨmǝt’a noro aster Aster

jǝmɨssɨraʧ good_news

jɨza hold\CNV. 3SG.F

b-t-mǝt’a

noro

COND-3SG.F-

AUX.CF

come\IPFV

‘If Aster had brought (lit. had come holding) good news …’ b. aster jǝmɨssɨraʧ jɨza mǝtta bihon noro aster Aster

jǝmɨssɨraʧ good_news

jɨza hold\ CNV.3SG.F

mǝt’ta come\ CNV.3SG.F

b-j-hon

noro

COND-3SG.M-

AUX.CF

be(come)\IPFV

‘If Aster had brought (lit. had come holding) good news …’

7

However, in addition to the presence of noro, the contexts that trigger imperfectivebased vs. converb-based antecedents are still to be examined.

86

Abdu Ahmed

Converb-based complex predicates allow for the insertion of adverbs, cf. (7), or subordinating conjunctions attached to the second component of the converb-plus-verb construction, cf. (8). (7) aster jǝmɨssɨraʧ jɨza tolo bɨtɨmǝt’a noro … aster jǝmɨssɨraʧ jɨza tolo b-t-mǝt’a noro Aster good_news hold\CNV.3SG.F quickly COND-3SG.F-come\IPFV AUX.CF

‘If Aster had quickly brought (lit. had come holding) good news …’ (8) aster jǝmɨssɨraʧun jɨza ɨndəmətt’atʧ bɨtɨnnaggər noro … aster jǝmɨssɨraʧ-u-n Aster good_news-DEFACC

jɨza hold\CNV. 3SG.F

ɨndə-mətt’a-ǝčč as-come\PFV3SG.F

b-t-tnaggər

noro

COND-3SG.F-

AUX.CF

tell\IPFV

‘If Aster had told the good news as soon as she had come …’ As shown in (7) and (8), the adverbs tolo ‘quickly’ and the subordinating conjunction ɨndə- intervene between the converb jɨza and its superordinate verb. Agreeing with Desalegn (2012), these examples prove that converb constructions in Amharic cannot be considered serial verb constructions. 3.2 Coverb constructions When a coverb-based complex predicate occurs in an imperfective-based counterfactual antecedent clause, as in (9a)– (9d), the conditional conjunction b- is prefixed to the light verbs alǝ ‘say’ or adǝrrǝgǝ ‘do, make’ but not to the coverbs. (9)

a. bǝrru kǝfǝtt bil noro … bǝrr-u door-DEF

kǝfǝtt

b-j-l

noro

IDPH.open\ATTEN

COND-3SG.M-say\IPFV

AUX.CF

‘If the door had opened (slightly) …’ b. aster fǝt’ǝnn fǝt’ǝnn bɨtɨl noro … aster Aster

fǝt’ǝnn_fǝt’ǝnn IDPH.quick\ITER

b-t-l

noro

COND-3SG.F-say\IPFV

AUX.CF

‘If Aster had hurried up (slightly) …’ c. botaw c’ɨnk’ɨnk’ɨnk’ bil noro … bota-w place-DEF

c’ɨnk’ɨnk’ɨnk’

b-j-l

noro

IDPH.crowded\ITER

COND-3SG.M-say\IPFV

AUX.CF

‘If the place had become overcrowded …’

Complex Predicates in Amharic Counterfactual Antecedent Clauses

87

d. kassa tɨmhɨrtun jazz lǝk’ǝkk’ bijadǝrgəw noro … kassa Kassa

tɨmhɨrt-u-n education-POSS.3SG.M-ACC

b-j-adǝrg-əw COND-3SG.M-do\IPFV-OJ.3SG.M

jazz

lǝk’ǝkk’

IDPH.hold

IDPH.release

noro AUX.CF

‘If Kassa had not taken his education seriously (lit. had once held it and another time released it) …’ In converb-based counterfactual antecedent clauses, however, the conditional conjunction b- is not attached to the coverb-based complex predicate but to the additional auxiliary honǝ, cf. (10). (10) a. bǝrru kǝfǝtt bɨlo bihon noro bǝrr-u door-DEF

kǝfǝtt IDPH.open\ATTEN

blo say\CNV.3SG.M

b-j-hon

noro

COND-3SG.M-

AUX.CF

be(come)\IPFV

‘If the door had opened (slightly) …’ b. aster fǝt’ǝnn fǝt’ǝnn bɨla bihon noro aster Aster

fǝt’ǝnn_fǝt’ǝnn IDPH.quick\ITER

bla say\CNV.3SG.F

b-j-hon

noro

COND-3SG.M-

AUX.CF

be(come)\IPFV

‘If Aster had hurried up (slightly) …’ Negated coverb constructions tend to denote actions or states that are inconsistent, indecisive or undesirable (Baye 2008:228). For the two affirmative coverb-based counterfactual antecedent clauses (9a) and (9c) above, two different negation possibilities exist: The negative clauses in (11) are imperfective-based counterfactual antecedent clauses, while the negative clauses in (12) are converb-based. (11) a.

bǝrru kǝfǝtt bajl noro … bǝrr-u door-DEF

kǝfǝtt

b-a-j-l

noro

IDPH.open\ATTEN

COND-NEG-3SG.M-say\IPFV

AUX.CF

‘If the door had not opened (slightly) …’ b. botaw ʧ’ɨnk’ɨnk’ɨnk’ bajl noro … bota-w ʧ’ɨnk’ɨnk’ɨnk’ b-a-j-l place-DEF IDPH.crowded\ITER COND-NEG-3SG.M-say\IPFV

‘If the place had not got overcrowded …’

noro AUX.CF

88

Abdu Ahmed

(12) a.

bǝrru kǝfǝtt bɨlo bajhon noro … bǝrr-u door-DEF

kǝfǝtt IDPH.open\ ATTEN

blo say\CNV. 3SG.M

b-a-j-hon

noro

COND-NEG-3SG.M-

AUX.CF

be(come)\IPFV

‘If the door had not become open (slightly) …’ b. botaw ʧ’ɨnk’ɨnk’ɨnk’ bɨlo bajhon noro … bota-w place-DEF

ʧ’ɨnk’ɨnk’ɨnk’ IDPH.crowded\ ITER

blo say\CNV. 3SG.M

b-a-j-hon COND-NEG-3SG.Mbe(come)\IPFV

noro AUX.CF

‘If the place had not got overcrowded …’ In the imperfective-based counterfactual antecedents in (11) the negation marker a- is directly attached to the light verbs. However, in the converbbased antecedent clauses in (12), it is attached to the auxiliary verb honǝ. It is not clear whether there exists a pragmatic difference between the two negative constructions. 4 Summary In summary, the analysis of converb- and coverb-based complex predicates in canonical Amharic counterfactual antecedent clauses brings forth some interesting structural phenomena. The use of either imperfective-based antecedents or converb-based counterparts follows a specific tendency. The imperfective-based antecedent clauses are preferred if counterfactuality is marked by the auxiliary noro. In contexts where this auxiliary is absent, the converbbased antecedent clauses are preferred. As in the affirmative, two different constructions are possible in the negative of converb- and coverb-based complex predicates in the two canonical counterfactual antecedent clauses. Abbreviations 1, 2, 3 First, second, third person ACC Accusative case AGR Subject/object agreement ATTEN Attenuative AUX Auxiliary verb CF Counterfactuality CNV Converb DEF Definiteness F Feminine

IDPH ITER IPFV M NEG NP OJ PFV PL

Ideophone Iterative Imperfective Masculine Negation Noun phrase Object Perfective Plural

Complex Predicates in Amharic Counterfactual Antecedent Clauses PST SG

Past tense Singular

SJ

89

Subject

References Abdu Ahmed. 2011. Conditional Constructions in Amharic: A Description. Saarbrücken: VDM. Alsina, Alex, Joan Bresnan & Peter Sells (eds.). 1997. Complex Predicates. Stanford: CSLI Publications. Azeb Amha & Gerrit Dimmendaal. 2006. Converbs in an African perspective. In Felix K. Ameka, Alan Dench & Nicholas Evans (eds.), Catching Language: The Standing Challenge of Grammar Writing, 393–440. Berlin & New York: Mouton de Gruyter. Baker, Brett & Mark Harvey. 2010. Complex predicate formation. In Mengistu Amberber, Brett Baker & Mark Harvey (eds.), Complex Predicates: Cross-Linguistic Perspectives on Event Structure, 13– 47. Cambridge: Cambridge University Press. Baye Yimam. 1999. The verb “to say” in Amharic. Journal of Ethiopian Studies 32(1). 1–50. Baye Yimam. 2006. The interaction of tense aspect and agreement in Amharic syntax. In John Mugane, John P. Hutchison & Dee A. Worman (eds.), Selected Proceedings of the 35th Annual Conference on African Linguistics, 193–202. Somerville, MA: Cascadilla Proceedings Project. Baye Yimam. 2008. Jǝʔamarɨɲɲa sǝwasɨw. [Amharic Grammar] Addis Ababa: Elleni Publishing. Bhatt, Rajesh & Roumyana Pancheva. 2005. Conditionals. In Martin B. H. Everaert & Hendrik C. van Riemsdijk (eds.), The Blackwell Companion to Syntax, vol. I, 638–687. Oxford: Blackwell. Desalegn H. Asfawwesen. 2012. Amharic as a converbal language – as opposed to a serial one. Proceedings of the Doctoral Festival 2012, Stockholm University. http://www.ling.su.se/polopoly_fs/1.99115. 1346331385!/menu/standard/file/Hagos_Desalegn.pdf (10 January 2014). Hartmann, Josef. 1980. Amharische Grammatik. Wiesbaden: Steiner. Hetzron, Robert. 1972. Ethiopian Semitic: Studies in Classification. Manchester: Manchester University Press.

90

Abdu Ahmed

Kaufmann, S. 2006. Conditionals. In Keith Brown (ed.), Encyclopedia of Language and Linguistics, vol. III, 6–9. Oxford: Elsevier. Leslau, Wolf. 1995. Reference Grammar of Amharic. Wiesbaden: Harrassowitz. Mengistu Amberber, Brett Baker & Mark Harvey (eds.). 2010. Complex Predicates: Cross-Linguistic Perspectives on Event Structure. Cambridge: Cambridge University Press. Meyer, Ronny. 2012. The converb in Amharic. In Domenyk Eades (ed.), Grammaticalization in Semitic, 165–192. Oxford: Oxford University Press. Payne, Thomas E. 1997. Describing Morphosyntax: A Guide for Field Linguists. Cambridge: Cambridge University Press. Podolsky, Baruch. 1991. Historical Phonetics of Amharic. Tel Aviv: Baruch Podolsky. Tadesse Beyene. 1972. Aspects of the Verb in Amharic. Georgetown: Georgetown University (PhD Dissertation). Thompson, Sandra A., Robert E. Longacre & Shin Ja J. Hwang. 2007. Adverbial clauses. In Timothy Shopen (ed.), Language Typology and Syntactic Description. Volume II: Complex Constructions, 236–299. Cambridge: Cambridge University Press. Wetter, Andreas. 2003. Ideophones in Amharic. In Kézié K. Lébikaza (ed.), Actes du 3e Congrès Mondial de Linguistique Africaine, Lomé 2000, 257–267. Cologne: Köppe.

Complex Predicates in Zargulla Azeb Amha Leiden University Abstract The paper presents an analysis of complex predicates in Zargulla, a member of the East Ometo branch of Omotic. Complex predicates are constructions in which a single clause is headed by two verbal constituents. The V1 constituent of complex predicates may be any lexical or ideophonic verb whereas the V2 constituent is restricted to about ten lexical verbs, which comprise several motion verbs. Ideophonic verbs are used as V1 constituents in their basic (lexical) form without any further morphological marking. Other V1 constituents must be either in the converb or intensive verb form. Thus on the basis of the structure of V1, the language has three types of complex predicates: [V1[CNV]–V2], [V1[INT]–V2], and [V1[IDPH]–V2]. The basic, minimal form of a complex predicate in Zargulla consists of two verbs: V1 plus V2. Surface realizations of complex predicates with three verbs are thus analyzed hierarchically: V1–V2 being used as a V1 input for a higher-level complex predicate. Schematically represented, the two hierarchical structures are: [V1–V2] and [V1[V1+V2]–V2]. Support for the two-level analysis stems from the fact that in complex predicates of the latter type the final V2 may be realized once after a series of V1[V1+V2] constituents. Semantically, complex predicates are characterized by expressiveness, i.e. unlike simple verbs, they encode events together with manner (e.g. speed), context (e.g. unexpectedness), direction, etc. as integral part of their meaning. Complex predicates are compact structural parallels to syntactic constructions involving a simple verb and one or more adverbial phrases. 1 Introduction Zargulla is an endangered Omotic language, classified as a member of the East Ometo branch of Omotic. Administratively, the Zargulla are part of the Bonke District Administration (or Bonke Woreda) within the Gamo-Gofa Zone. Although known by the name Zargulla in various scientific publications and in official documents including the national censuses prior to 2007, the speakers identify themselves as Gamo and they refer to their language as Gamotso. In the 1994 national census it is reported that there are 7,625 mother tongue speakers of whom 390 claimed to be ethnic Zargulla (cf. CSA

92

Azeb Amha

1998:67, 80). However, the 2007 census does not mention the Zargulla at all (cf. CSA 2008). It seems that because of their ethnic identity name Gamo, the Zargulla are included in the linguistic group of the neighboring Gamo, who speak a North Ometo language. However, these two languages are not mutually intelligible and belong to different branches of the Ometo cluster: both the four-way division of Ometo in Fleming (1976) as well as the broader, three-way grouping proposed in Bender (2003) keep Zargulla and Gamo in different branches: East Ometo vs. North Ometo or Southeast Ometo vs. Northwest Ometo, respectively. Before the discussion of complex predicates, I mention here a few phonological and morpho-syntactic properties of the language that are relevant in the examination of the data used in the present contribution. The reader will notice that words may have different realizations of tone-accent. The realization depends on the tone of the lexeme and its affixes as well as on word-level rules. Zargulla is a tone-accent language in which high-tone (marked by the acute accent) and low-tone (unmarked) are distinguished. Multi-syllabic words with all of the syllables bearing low-tone are avoided. Verb roots, most of which have a CVC structure, may be inherently high- or low-toned. When a low-tone affix is added to a low-tone verb root, a default high-tone is realized on the ultimate or penultimate syllable, so that there is at least one high tone-accent in an independent word. For example, the imperative marker is either -a or -á depending on the tone of the verb. Thus, the imperative of dáp- ‘release’ is dáppa ‘release!’ whereas that of dap- ‘pull’ is dapá ‘pull!’. In addition, certain verb paradigms have a fixed tone-accent pattern. Thus both the high-tone verb root dáp- ‘pull’ and the low-tone verb root dap- ‘release’ are marked by a high tone in the past, present and future declarative: dápóttesínne ‘he released’ and dápáttesínne ‘he pulled’. However, they maintain their lexical tone-accent in the progressive declarative as well as in the imperative and optative (cf. also Azeb 2009; 2013). Basic, unmarked nouns may have different syntactic functions. The indefinite object form is identical to the basic noun; see, for example, budó ‘fire’ in (9a) or Ɂoótsó ‘work’ in (10). Goal nouns may also be morphologically unmarked basic forms. In (6) for instance, the goal noun gidoóle is without morphological case marking; see also deré ‘a big village/country’ in (17b) in the same role. Moreover, basic, unmarked nouns indicate the possessor in (indefinite) possessive constructions, e.g. haré ‘donkey’ in haré Ɂáde ‘donkey’s father’ in (8a). In analogy to the morphologically marked accusative and possessive, I gloss unmarked object and goal nouns as ACC and possessive nouns as POSS. Language-internal and comparative data motivate

Complex Predicates in Zargulla

93

grouping goal and object nouns together, the details of which cannot be discussed here. Pronouns have long and short forms, e.g. subject ʔé or ʔésí (3SG.M); tá or tání (1SG); né or néní (2SG); ʔú or ʔúsúní (3PL), etc. Some of the object pronouns exhibit the same alternation, e.g. táná or taa (1SG.OJ), néná or nee (2SG.OJ). It remains to be investigated whether the choice of the long or short form is motivated by pragmatic considerations. The long pronouns are morphologically marked for nominative and accusative case – but note that accusative case marking in pronouns and nouns is not identical. The short pronouns do not distinguish between subject and object by case suffixes but by different forms, e.g. tá (1SG.SJ) vs. taa (1SG.OJ). For this reason, I decided not to gloss nominative and accusative case in pronouns but rather indicate the syntactic function of short and long pronouns in the examples as subject and object. In nouns, however, I gloss case explicitly both for morphologically marked and unmarked accusative and nominative. 1.1 Overview of complex predicates in Zargulla In the present contribution, the structural and semantic properties of complex predicates in Zargulla are discussed. An example for a complex predicate is yeédí ʔolí ‘having come’ in (1):1 (1) haré

ʔáde

donkey.POSS father.POSS

yeepá

yeéd-í

ʔol-í

funeral.ACC

come-SS.CNV

leave-SS.CNV

yeép-á-tt-ús-ínne weep-INT-FOC-3PL-PST

‘Having come to the funeral of the donkey’s father, they [i.e. the hyenas] WEPT.’ I use Butt’s (2010:49) definition of complex predicates:2 1

2

The present contribution includes data from narratives and conversation as well as elicited material which I collected during four fieldwork trips between 2003 and 2009. I would like to thank Ato Tariku Chifaw, Ato Ayyele Bola, Ato Arche Apa, Ato Andreas Worera, Ato Nega Abdisa, Ato Mulatu Aberra, and my youngest assistants Temesgen Tariku and Mulu Tsegaye – all native speakers of Zargulla – for their unreserved help during my fieldwork and for their friendship. Research on Zargulla was made possible through the financial support of the Dutch National Science Foundation (NWO) through its Endangered Languages Programme. I am very grateful to Sasha Aikhenvald, Felix Ameka, Ronny Meyer and Yvonne Treis for helpful comments and suggestions. Other terms that are sometimes interchangeably used with complex predicate are light

94

Azeb Amha “[A] construction that involves two or more predicational elements (such as nouns, verbs and adjectives) which predicate as a single unit, i.e. their arguments map onto a monoclausal syntactic structure.”

Mengistu, Baker & Harvey (2010:2) further specify that “the term ‘complex predicate’ properly construed is shorthand for complex predicate head.” The components of complex predicates in Zargulla are only verbal elements. Each component contributes to the semantic interpretation of the predicate head. This prompts the question how complex predicates are distinguished from comparable V1–V2 sequences, such as serial verb constructions. There are indeed a number of similarities between complex predicates, such as that illustrated in (1), and serial verb constructions. Both are monoclausal; they have just one tense, aspect, and polarity value; they share core and other arguments. However, as Aikhenvald (2006:1) emphasizes, the constituent verbs in a serial verb construction (SVC) are “without any overt marker of coordination, subordination, or syntactic dependency of any other sort … Each component of an SVC must be able to occur on its own.” V1 components of Zargulla complex predicates, however, are morphologically marked as dependent verbs. Only in occasional cases of ellipsis, they may occur on their own in the head position of main clauses (see Sections 2.4 and 3). Based on the morphological properties of V1, three basic types of complex predicates are distinguished in Zargulla: i. Complex predicates in which V1 is a converb and V2 is from a restricted class of verbs including ʔol- ‘give up, abandon’, ʔer- ‘know’, or motion verbs; ii. Complex predicates in which V1 is an intensive verb stem and V2 is hi(yy)- ‘say’ or his- ‘say.CAUS’; iii. Complex predicates in which V1 is an ideophonic verb and V2 is hi(yy)‘say’ or his- ‘say.CAUS’ Besides structural similarities, the three types of complex predicates share in common semantic expressivity. That is, such verbs denote an event together with the manner, direction and/or effect of the event, which would otherwise be expressed by adverbs or modifying clauses. This is discussed in Section 2.1 for V1[CNV]–V2 constructions. In Section 2.2, I examine the morphosyntax and semantics of V1[INT]–V2 constructions. Ideophone-based complex predicates are discussed in Section 2.3. verb construction (Mengistu 2010; Darmon 2012; among others) or multi-verb construction (but cf. Ameka 2005 for the use of the latter term to refer to constructions that involve more than one clause). For cross-linguistic studies on complex predicates, cf. also Alsina, Bresnan & Sells (1997) and Butt (1997; 2003; 2010), among others.

Complex Predicates in Zargulla

95

In Zargulla a V1–V2 complex predicate may be the head of a main clause or a dependent clause, e.g. a conditional or temporal clause. In addition, a complex predicate as a single unit may fit into a V1 slot of a higher-level complex predicate, resulting in a sequence of three verbs. In this case, the V2 of the lower-level complex predicate is marked as a converb or an intensive verb, as the discussion in Section 2.4 shows. Preceding the discussion on complex predicates, I briefly present the structure of simple-verb clauses in Zargulla in Section 1.2 in order to compare them with complex predicates and to illustrate how verbal categories such as tense-aspect, verbal agreement, and polarity are marked. 1.2 Simple verbal predicates and auxiliary verb constructions By simple verbal predicate I refer to a single verb heading a clause. As a main verb, a simple predicate may be represented by the minimally inflected affirmative or negative existential verbs yéne ‘exist’ or báaʔa ‘not exist’, as shown in (2a–b). (2) a. taára-tt-iʃ

yéne

1SG.COMT-FOC-3SG.F

exist.PRES

‘She is/lives WITH ME.’ [as a reply to the question: ‘Does the mother of the children live with you?’] b. zéré descent.NOM

lóʔo

maákk’-e

metó-y

baáʔa

good

happen-REL

problem-NOM

not_exist.PRES/PST

‘There is no problem [to marry someone if the person’s] descent is a good one.’ A simple verbal predicate can also be a partially or fully inflected lexical verb, as in (3a) and (3b), respectively. In (3a) the verb zeérínne consists of the verb stem and a tense marker, i.e. it is only partially inflected. Such an inflection is used when the verbal predicate is not in focus. In contrast, when the verb is in focus, it must include a further morphological category, namely the intensive verb-stem marker -á or -ó, which is often followed by a subjectagreement marker, as in (3b–c).3 The focus marker may also be affixed to a converb verb; see haríttet ‘I SIMPLY’ in (16a), or Ɂórdíttes ‘HE BECOMING STOUT’ in (18).

3

In a few cases a verb in focus may be realized without a subject-agreement marker. The motivation for such an omission is not clear and needs further investigation.

96

Azeb Amha

(3) a. lóʔo-tt-it

zeér-ínne

good-FOC-2PL

speak-PST

‘You (PL/honorific) spoke/told WELL.’ b. ʔizípe-ga August-LOC

kiíló-tt-un-íne take_out_weed.INT-FOC-1PL-PRES

‘In August, we REMOVE THE UPROOTED WEEDS.’ c. kettéle

hátte haíʔ-aá-tt-un-éne

Kettele.NOM DM

híd-í

taákko yeéd-í

die-PROG-FOC-1PL-FUT say-SS.CNV 1SG.DIR come-SS.CNV

‘Well, the Kettele [villagers], having come to me and saying “We ARE DYING” …’ The progressive in (3c) can alternatively be expressed by an auxiliary verb construction, as in (4a), in which the present-tense existential verb yéne is combined with the fully inflected progressive verb. Note that subjectagreement and tense-aspect morphemes are not realized on V1 when the auxiliary is used as V2. It is not clear whether there is a pragmatic difference between the inflected present progressive, as in (3c), or the auxiliary-based one, as in (4a). As (4b) shows, a past progressive is formed by replacing the present-tense existential verb yéne with the past-tense existential verb yéʃʃe:4 (4) a. budó-y fire-NOM

Ɂeéss’-aá-tte

yéne

burn-PROG-FOC

exist.PRES

‘The fire is BURNING.’ b. Ɂé 3SG.M.SJ

yeés-eéska budó-y

Ɂeéss’-aá-tte

yéʃʃe

come-TEMP

burn-PROG-FOC

exist.PST

fire-NOM

‘When he came, the fire was BURNING.’ Although two predicative elements are involved in the constructions in (4), I do not analyze them on a par with complex predicates. In (4) the V2 component contributes a grammatical meaning to the construction, i.e. tense, which is otherwise expressed by the bound morphemes -íne (present/habitual), -ínne (past), and -éne/-ené (future). In contrast, in the complex predicate yeéd-í 4

The existential verbs yéne and yéʃʃe have the base yé-. The morpheme -ʃʃe may also be affixed to lexical verbs to indicate the past progressive. We find -ʃʃe, for instance, on V2 of the complex predicate Ɂárdáttus hí-ʃʃe ‘was living’ in (17b). However, as the contrast between present and past tense in lexical verbs is expressed by other morphemes, the tense morpheme is not separated on the existential verbs yéne and yéʃʃe. See also the contribution by Binyam Sisay Mendisu (this volume) on a similar use of existential verbs in Koorete.

Complex Predicates in Zargulla

97

ʔol-í in (1), both verbs contribute to the lexical meaning of the predicate head while tense-aspect – if marked – is expressed on V2 by the bound morphemes just mentioned. For instance, if the complex predicate in (1) were a nonfocused main verb, it would be inflected as yeéd-í ʔól-ínne (past), yeéd-í ʔól-íne (present/habitual), etc. Secondly, the V1–V2 construction in (1) – further illustrated and discussed in Section 2 – is comparable to simple verbal predicate in that it can be (co)subordinated as a single unit, which is not possible for a V1–V2 auxiliary verb construction. Finally, as I show in Section 2, the V2 element of complex predicates contributes to the argument structure of the clause, e.g. by determining transitivity, but this is not the case with the existential verbs yéne and yéʃʃe.5 In typological-comparative studies, contribution to argument structure is mentioned as one of the main reasons for distinguishing between auxiliary verb constructions and complex predicates (cf. Baker & Harvey (2010:2, 31); see also Aikhenvald (2006:5) on the differences between auxiliary verb constructions in English and serial verb constructions in serializing languages). With regard to the lexical, syntactic or inflectional information that the components of a complex predicate contribute to the clause and the way these are analyzed, Butt (2010:49) writes: “[L]ight verbs [i.e. forms corresponding to V2 in Zargulla, AA] are always part of a joint predication within a complex predicate. Indeed, it is this central characteristic that has rendered complex predication and the representation of light verbs a tough nut for syntactic theories. This is because a very fundamental assumption underlying all syntactic theories has been that the main verb is the predicational lynch-pin of the clause and that all other elements in the clause are either arguments or modifying elements of some sort. However, there was no sense that two or more predicational elements could come together to form a joint predication, with a jointly determined argument structure.” Anderson (2006:9–18) points out that serial verb constructions, verbcomplement structures and coordinate or conjoined constructions are diachronic sources of auxiliary verb constructions and treats all of them including auxiliary verb constructions as complex predicates. In the present synchronic analysis of Zargulla, auxiliary verb constructions are not discussed further. As outlined above, they have different morpho-syntactic and seman-

5

Zargulla has also other auxiliary verbs, which are not considered in the present contribution (cf. Azeb 2009).

98

Azeb Amha

tic properties compared to complex predicates that are discussed in the remaining sections. 2 Complex predicates in Zargulla There are three basic types of complex predicates in Zargulla. In the present section, I discuss each of them in turn. 2.1 Converb-based complex predicates In an excellent summary of the diverging definitions, morphosyntactic properties and functions of the converb, Rapold (2007:9) writes: “Most linguists agree that a converb is a verb form specialized for combining clauses and, more specifically, marking a dependent verb form that is neither argumental nor adnominal.” Among the “more commonly attested functions” of the converb Rapold (2007:13 f.) includes: • Part of compound TAM or aktionsart formulation, e.g. come-CNV stay = ‘keep coming’ • Part of a construction involving a directional, e.g. fly-CNV go = ‘fly away’ • Part of lexicalized combinations of verbs, e.g. know-CNV hold = ‘understand’ We find parallel constructions in Zargulla. Examples (5) and (6) illustrate the clause-linking function of the converb. The examples also show that there are two types of converbs in Zargulla: the same-subject converb marked by -í vs. the different subject converb, which is marked by -um/-úm. (5) hang-í go-SS.CNV

kaat-ús

ʃída

kes-us-í

king-DAT

honey.ACC

go_out-CAUS-SS.CNV

yéwa

híd-í

káʃʃ-ínne

come.IMP

say-SS.CNV

tell-PST

‘[The king’s messenger] told [the monkey]: “Go and have the people collect honey for the king and come!”’

Complex Predicates in Zargulla (6) gidoóle Gidole.ACC

99

tá

ham-úm

ʔaálá-kko-y

taa

1SG.SJ

go-DS.CNV

what-ADD-NOM

1SG.OJ

harg-úss-úm

híkke

hií-nna

yeéd-í …

be_sick-CAUS-DS.CNV

DM

DIST-INS

come-SS.CNV

‘I having gone to Gidole, something made me ill and having returned because of that …’ In (5), the subject of the imperative sentence (i.e. ‘monkey’) is not overtly realized but is mentioned earlier in the story. In this sentence, the subject of the verb hang- ‘go’, the causative verb kesus- ‘make go out’ and the imperative main verb yewá ‘come’ is the same (i.e. ‘monkey’). Accordingly, the same-subject converb marker -í is suffixed to each dependent verb. Similarly, the subject of the two verbs in the matrix clause hídí káʃʃínne is the same (i.e. ‘the king’s messenger’). Therefore, V1 of this quotative construction (hídí ‘having said’) bears the same-subject converb marker -í. In contrast, in (6), the subject of the first clause is the 1SG pronoun tá, whereas that of the immediately following clause is the derived nominal ʔaálákko ‘something’. Anticipating the change of subject in the second clause, the verb ham- ‘go’ is marked by the different-subject marker -um/-úm. The subject of the verb yeéd- ‘come’ in the third dependent clause is again 1SG. To indicate the switch-back to the 1SG, the different-subject marker -um/-úm is also suffixed to the causative verb hargúss- ‘make sick’. Of the two converb types in Zargulla, the same-subject converb is combined with other verbs to form a complex predicate. The different-subject converb is not used for this purpose. This is a common tendency crosslinguistically both in serial verb constructions and in complex predicates (cf. Aikhenvald 2006:14; 2008: Chapter 17). The same-subject converb is used in different types of complex predicates in Zargulla, including those in (7a) and (7b). The relevant verbs are given in boldface: (7) a. hátte-tte-t now-FOC-1SG

lább-í

yeénne

be_tired-SS.CNV

come.PST

‘NOW I came tired.’

100

Azeb Amha b. ʃaat-t-ó

ʔeéd-í

girl-F.DEF- F.ACC bring-SS.CNV

maák’-í

hang-ȇn

return-SS.CNV

go-EMPH.Q

ʔol-í

ʔul-í-us

leave-SS.CNV

return-SS.CNV-3PL

6

‘Having brought the bride, they go back directly to their home?’ In (7a), V1 lábbí and V2 yeénne form a complex predicate. They jointly copredicate the subject similar to V1 ʔeédí and V2 ʔolí in the highlighted complex predicate in (7b). However the two complex predicates differ in a significant way. In (7a), V1 and V2 each represent a co-temporal event (state and motion). V1 does not form a distinct clause on its own and it is not an adverbial modifier of V2. Rather V1 lábbí in (7a) co-predicates the subject together with the main predicate V2, which carries the tense/aspect and polarity value that is shared by V1 and V2. V1 lábbí ‘be tired’ expresses the state the subject is in when the event expressed by the main verb (yeénne ‘came’) is realized. Further, the lexical meanings of both V1 and V2 are still explicit in the semantic interpretation of the sentence. In (7b) on the other hand, V1 ʔeéd- ‘bring’ is central to the interpretation of the sentence, whereas the lexical meaning of V2 ʔol- ‘leave, abandon, give up’ (hereafter glossed as ‘leave’) is not directly apparent. That is, although both ʔeéd- ‘bring’ and ʔol- ‘leave’ are in the predicate head position, V1 ʔeéd- represents the semantic content of the sentence more directly. The semantic contribution of V2 ʔol- ‘leave’ in (7b) is not identical to its contribution in clauses in which it is used alone as the head or in constructions in which it is used as V1 (see next section for details). In the present contribution, I do not deal with the details of the complex predicate type illustrated in (7a).7 Rather, I focus on the complex predicate type illustrated in (7b). 2.1.1 Converb plus ʔol- ‘leave’ The converb plus ʔol- construction is very frequently used. In the interpretation of (1), repeated below as (8a), nothing is included in the translation that reflects the lexical meaning of ʔol- ‘leave’. In addition, the utterance is per6 7

The verb ʔul- is translated here as ‘return home’. However, this verb has a broader meaning: ‘return to the location one came from’. In (16c), for instance, the location the addressee must return to is a riverside. Himmelmann & Schultze-Berndt (2005) is a typological study of such constructions, known as depictive secondary predicates. A similar use of the converb as a secondary predicate is attested in a number of Ethiopian languages including Amharic and Wolaitta (cf. Azeb & Dimmendaal 2005).

Complex Predicates in Zargulla

101

fectly grammatical without V2 as shown in (8b). Off hand, the two sentences have the same translation: (8) a. haré donkey.POSS

ʔáde

yeepá

yeéd-í

ʔol-í

father.POSS

funeral.ACC

come-SS.CNV

leave-SS.CNV

yeép-á-tt-ús-ínne weep-INT-FOC-3PL-PST

‘Having come to the funeral of the donkey’s father, they (i.e. the hyenas) WEPT.’ b. haré

ʔáde

yeepá

yeéd-í

yeép-á-tt-ús-ínne

donkey.POSS father.POSS funeral.ACC come-SS.CNV weep-INT-FOC-3PL-PST

‘Having come to the funeral of the donkey’s father, they (i.e. the hyenas) WEPT.’ However, from cross-checks with consultants and an examination of the contexts in which V1[CNV] plus ʔol- is used, it is clear that this construction is semantically distinct from the corresponding sentence with just V1 yeéd‘come’. Comparing the two examples, the addition of the V2 ʔol- ‘leave’ in (8a) contributes a semantic nuance to the lexical meaning of V1 which can be characterized as accomplishment of the action expressed by V1 and a turning point in the story line, e.g. in the form of surprise, suspense or unexpectedness of what is expressed by the phrase or clause that immediately follows V1[CNV] plus ʔol-. In (8a), the fact that the hyenas attend the funeral of a donkey and weep is presented as a surprising and unexpected action and this is indicated by the addition of ʔol-. In (8b) on the other hand, the event is expressed without such an implication; the action is what most attendants would ordinarily do during a funeral (disregarding the strange friendship between hyenas and donkeys that the story relates). Consider (9a) and (9b), which occur in a story as they are presented below: (9) a. budó fire.ACC

Ɂekk-í

Ɂol-í

take-SS.CNV leave-SS.CNV

Ɂudúla-gá-tte

ɗáy-ínne

mortar-LOC-FOC

throw-PST

‘[Tsuho] took fire and put it IN A MORTAR.’

102

Azeb Amha b. Ɂudúla-ga ɗay-í mortar-LOC

throw-SS.CNV

waatsé-tte

huúɁɁ-ínne

water-FOC

spill-PST

Ɂol-í

budó-ga

leave-SS.CNV

fire-LOC

‘After putting the fire in a mortar he spilled WATER on the hearth.’ In the story, Lion and his best friend Tsuho, a trickster, arrive at home to prepare a meal from the meat of their cow, which has died in suspicious circumstances. The clauses that immediately follow the V1[CNV] plus ʔol- construction express two unexpected events: instead of kindling the fire to cook the meat, Tsuho hides part of the hot embers in the mortar and extinguishes the remaining fire in the hearth. It becomes clear later that Tsuho carries out these actions so that Lion would go away to fetch fire and enable Tsuho to carry out his secret plan to take revenge on Lion. If ʔol- were left out and the dependent clauses in (9a) or (9b) ended with just the converb Ɂekk-í or ɗay-í, the two clauses would be interpreted as mere sequential events without the additional meaning just outlined. That is, the sequential/anteriority meaning is expressed by the converb marker -í, not by ʔol-. Complex predicates with ʔol- often occur at a transition point in narratives, where speakers repeat the main verb of the preceding sentence as a head-tail construction and raise the rhetoric question What next?, as shown in (10). This is in line with the function of ʔol- in V1 plus Ɂol- constructions to indicate important, unexpected or surprising events or actions. (10) ʔoótsó

ʔálts-á-tte-s-ínne.

work.ACC finish.CAUS-INT-FOC3SG.M-PST

ʔalts-í

Ɂol-í

waínne-s-ínne

finishSS.CNV

leaveSS.CNV

what_do-3SG.M-PST

‘He FINISHED the work. Having finished, what did he do?’ The verb ʔol- in (8a), (9) and (10) is not a grammaticalized verb that expresses an obligatory tense/aspect meaning. As might be expected of constituents of complex predicates (cf. Butt 2003:2; 2010:53–54), it is “formidentical” to a lexical verb that can stand on its own and head an independent clause (11a). In addition, like any other lexical verb, ʔol- can be used as V1 in a complex predicate, as illustrated in (11b). Note that the latter option is not available for the existential auxiliary verbs yéne and yéʃʃe even though, like ʔol- in (11a), each of the two verbs can head an independent clause (cf. Section 1.2 above).

Complex Predicates in Zargulla (11) a. ʔésá 3SG.M.OJ

103

ʔol-átt-iʃ-ínne leave-FOC-3SG.F-PST

‘It STOPPED.’ [talking of somebody troubled by diarrhea; lit. ‘It left him.’] b. ʔésí 3SG.M.SJ

bé

gutá

ʔol-í

3.LOG village.ACC leave-SS.CNV

ham-ínne go-PST

‘He abandoned his village.’ The verb ʔol- can also follow a lexical verb in the negative converb form, as in (12a) and (12b), and be either in the affirmative or negative polarity. (12) a. kiít-t-ó

ʔaálassun gats-átʧe

message-F.DEF-F.ACC why

ʔol-ínne

make_reach-NEG.CNV leave-PST

‘Why didn’t you deliver the message?’ b. wuts-átʧe

ʔol-aáʔa

bring_down-NEG.CNV

leave-PRES.NEG

‘I must bring it down.’ [i.e. a ball from a tree] In the two cases in (12), it is not clear whether ʔol- forms a complex predicate with V1 or whether it should be interpreted as an independent clause. Such sentences have not yet been sufficiently examined in context. 2.1.2 Converb plus ʔer- ‘know’ A converb-based V1 plus ʔer- construction is another frequently used complex predicate in Zargulla. In this case V2 indicates whether the state or event expressed by V1 has been experienced before. Typically, it is realized in the present tense form: (13) a. ʧ’ilóoʃe

ʔekk-í

ʔer-aáʔa

dowry.ACC take-SS.CNV know-PRES.NEG

‘(As a custom) we do not take dowry.’ b. tá 1SG.SJ

s’aap-í

ʔer-íne

write-SS.CNV

know-PRES

‘I can write.’ [lit. ‘Having written I know’] In the negative, V2 ʔer- is often realized twice as shown in (14b). There is a slight semantic difference between sentences with single ʔer- as in (14a), and those in which ʔer- occurs twice (14b). Example (14a) can have either of the

104

Azeb Amha

interpretations given, whereas (14b) can only be interpreted as a negative experiential. (14) a. gútó-y

kats-í

ʔer-aáʔa

Gute-NOM cook-SS.CNV

know-PRES.NEG

i. ‘Gute has never cooked.’ [within a certain period or at a certain place] ii. ‘Gute cannot/does not know how to cook.’ [i.e. she never tried] b. yébi

né

ʔaákko-kka

2SG.F.VOC 2SG.SJ

what-ADD.ACC

ʔoots-í

ʔer-í

ʔer-aáʔa

work-SS.CNV

know-SS.CNV

know-PRES.NEG

‘Hey, haven’t you ever done any (domestic) work before?’ The following examples illustrate the use of the verb ʔer- ‘know’ as a main predicate in a simple sentence. (15) a. tá

ʔaákko-nne

1SG.SJ what-COORD

ʔer-aáʔa know-PRES.NEG

‘I do not know anything.’ b. ʔaráda Arada.POSS

ʔíndo

ʔer-áye

mother.ACC know-2SG.NEG.PRES.Q

‘Don’t you know Arada’s mother?’ 2.1.3 Converb plus motion verbs Motion and direction are often expressed by complex predicates in Zargulla. See the complex predicates marked in boldface in (16): (16) a. hí-gá-tte-t DIST-LOC-FOC-1SG

basaaʧ’aútt-í 8

waatsé

kís’o-ga

be_upset-SS.CNV

water.POSS

abscess-LOC

har-í-tte-t

wúɗɗ-í

pínne

be_nothing/empty-SS.CNV-FOC-1SG

descend-SS.CNV

cross.PST

‘Being upset by THAT [i.e. a land dispute, on account of which the speaker had to travel even in the rainy season to bring his case to the court in the regional capital] I SIMPLY crossed high rivers.’

8

The verb basaaʧ’aútt- is a loan from Amharic täbäsatʧ’ä ‘be upset’.

Complex Predicates in Zargulla b. ʔing-í

ʔol-í

ʔul-í

105

hamm-a-íte

give-SS.CNV leave-SS.CNV return-SS.CNV

go-IMP-2PL

ʔísó-y

haná

yéss-áya

3SG.F-NOM

PROX.LOC

exist-3SG.F.OPT

‘[The cat] gave [the verdict that sentenced the rat and told the people]: “You go back home! Let her [i.e. the rat] stay here!”’ c. Ɂul-í return-SS.CNV

Ɂekk-í

yeéd-í

s’éga

neéra

come-SS.CNV

goat.ACC

2SG.COMT

hamm-á

take-SS.CNV go-IMP

‘Come back and take the goat with you!’ Syntactically, component verbs of complex predicates with motion verbs as V2 are loosely bound. In (17), V1 and V2 may be separated by the subject (17a) or by a goal noun (17b). (17) a. míɁɁó 9

tookk-í

cow.F.DEF.ACC

Ɂekk-í

gaámmé yeénne

carry-SS.CNV take-SS.CNV lion.NOM

come.PST

‘The lion brought the cow carrying.’ b. moog-í

Ɂol-í-tt-us

tolkó-y

bury-SS.CNV leave-SS.CNV-FOC-3PL

Ɂul-í

hyena-NOM return-SS.CNV

bé

deré

hang-í

Ɂárdá-tt-us

hí-ʃʃe

3.LOG

village.ACC

go-SS.CNV

live.INT-FOC-3PL

say-AUX.PST

‘Having BURIED [the eldest donkey], the hyenas returned to their village and were LIVING [there].’ [i.e. not showing any signs of the adverse plan they had for the donkeys] V2 yeénne ‘came’ in (17a) indicates direction of movement towards the deictic center (i.e. the speaker), whereas V2 hangí ‘go’ in (17b) indicates movement away from the deictic center.

9

The basic, indefinite form of the noun is míis ‘cow’ which can also be used as indefinite object noun. The definite object form of the noun can be míɁɁó, as in this example, or míɁáttó, as in (26). Note also that we find gaámmé ‘the lion (NOM)’ in (17), instead of the expected nominative form gaammóy. The basic, indefinite form of the noun is gaammó ‘lion’; the nominative case marker is -y/-í. It is not clear why such alternations occur with a small number of nouns.

106

Azeb Amha

Furthermore, V1 can be reduplicated on its own to express an augmentative, as in the following (pseudo-)cleft construction, which involves V1 plus the relative form of the verb ham- ‘go’ as V2. (18) tá

ʔiʧé-y

1SG.POSS

ʔórd-í

ʔórd-í-tte-s

hám-e

brother-NOM be_stout-SS.CNV be_stout-SS.CNV-FOC-3SG.M go-REL

‘My brother is increasingly becoming FAT/STOUT.’ [lit. ‘It is having become FAT, having become FAT that my brother goes’] Duration may also be expressed by reduplicating the V1 component with just one final V2. This is attested in complex predicates involving intensive verb stems, which are discussed in the next section. Zargulla has another set of converb-based complex predicates of which the V1 component is always hí(yy)- ‘say’ and the V2 component is one of the utterance verbs káʃʃ- ‘tell’, ʔoótʧ’-/ʔoitʧ- ‘ask’, s’eég- ‘call’. Such complex predicates express the quotative. In (5) above, the final verb hídí káʃʃínne [sayCNV tellPAST] ‘told’ is an instance of the quotative. This construction is not further discussed. 2.2 Intensive verb plus hí(yy)- ‘say’ and hís- ‘say.CAUS’ The V1[INT] plus hí(yy)-/hís- complex predicate expresses that the action or event denoted by V1 is effected with great rigor or intensity. There are two variants of the intensive verb morpheme: -a and -o, whose distribution is lexically determined (cf. Azeb 2007). Thus the verb ʔard- ‘live’ in (17b), repeated below as (19a) for convenience, takes -a while ɗoll- ‘jump’ and kes‘go out’ in (19b) take -o. (19) a. moog-í

Ɂol-í-tt-us

tolkó-y

Ɂul-í

bury-SS.CNV leave-SS.CNV-FOC-3PL hyena-NOM return-SS.CNV

bé

deré

hang-í

Ɂárdá-tt-us

hí-ʃʃe

3.LOG

village.ACC

go-SS.CNV

live.INT-FOC-3PL

say-AUX.PST

‘Having BURIED [the eldest donkey], the hyenas returned to their village and were LIVING [there].’ [i.e. not showing any signs of the adverse plan they had for the donkeys]

Complex Predicates in Zargulla b. beé-s

tooruts-í

ɗóllo

3.LOG-DAT fight-SS.CNV jump.INT

késo

his-eés

go_out.INT say.CAUSIPFV.REL

107 hiídáas’ DIST.ASS.PL

‘Having fought themselves, it is those who make [young people] come out jumping.’ [i.e. having fought for themselves, being familiar with it, it is those who make the youth come out jumping, eager to fight whenever a conflict arises] The selection of V2 hí(yy)- ‘say’ or hís- ‘say.CAUS’ depends on transitivity: hí(yy)- ‘say’ is used in intransitive clauses, the single participant of which has the semantic function of actor or undergoer. The V2 hís- ‘say.CAUS’ is used in clauses with at least two participants, of which the subject participant is the actor, initiator or controller of the event. In (19a), for example, the verb V2 hí(yy)- ‘say’ is used since V1 Ɂárd- ‘live’ is intransitive. In (19b) on the other hand, the V2 hís- ‘say.CAUS’ is used to indicate that the (covert) 3PL subject of the clause (i.e. those referred to as being familiar with fighting) are not the actors of the intransitive verbs ɗóll- ‘jump’ and kés- ‘go out’ but are the causers of these actions. See also example (21) below for the use of hís- with a simple (non-causative) transitive verb. A sequence of dependent clauses, each headed by a distinct lexical V1 constituent, may be followed by a single V2 híyy-/hís- at the end of the series of clauses. In (20), there are three dependent clauses, ending in the intensive verbs gaádda ‘dance.INT’, kírba ‘dance.INT’ and ɗóllo ‘jump.INT’ (see brackets indicating the clause-boundaries). In principle, each of the three verbs should be followed by V2 hí(yy)- ‘say’ or hís- ‘say.CAUS’. But the speaker chose to omit V2 in the first two dependent clauses. The complex predicate in the final clause is complete, consisting of V1 ɗóllo and V2 híyy-áwo. Thus, in (20), there is a case of V2 ellipsis in the first two dependent clauses; the V2 can be recovered from the final clause (see also Section 3). It may be argued that cases like (20) are chaining constructions comparable to series of converbal clauses. However, converb forms have two functions: they are used to link two or more clauses in sequential or anterior relationship, and they contribute the V1 of a complex predicate head in a monoclausal structure. This is not true for intensive verbs. The intensive verb is not used in functions other than as V1 of a complex predicate. Cases like (20) are exceptions in which the V2 constituent that should indicate the actual clause-level relation is omitted. In (20), there is no overt anterior, simultaneous, causal or other clause-level relation among the clauses headed by gaádda, kírba and the final ɗóllo híyyáwo. Unlike the converb marker -í, the intensive verb marker -a/-o does

108

Azeb Amha

not contribute such clause-level meaning. The construction in (20) is thus different from chaining constructions with a series of converbal clauses. (20) [[be

gaadd-eɗé

gaádda] e1

3.LOG kind_of_dance-PL.ACC dance.INT

[ʔogé-ga

ʔaáɗɗ-í

road-LOC

pass-SS.CNV

kirbé

kírba] e1

[ʔapáa

ɗóllo

híyy-áwo1]]

kind_of_dance.ACC

dance.INT

sky.ACC

jump.INT

say-3PL.NEG.Q

‘[The young people] dance their gaaddé dances, they go to the road and dance kirbé [and] jump up high [for the ɗóllo dance], don’t they?’ Similarly, the V1[INT] constituent may be reduplicated to express distributive or durative actions. Such reduplicated verbs are then followed by a single final V2 as in (21). (21) híkke ʔóla DM

ɗuuss-í

yeétta

war.ACC explode.CAUS-SS.CNV

k’eés’o

laarká

come.INT clear_forest.INT

Laarka.ACC

k’eés’o

k’eés’o

k’eés’o

ʔú

hís-eéska

clear_forest.INT

clear_forest.INT

clear_forest.INT

3PL.SJ

say.CAUS-TEMP

‘Now, they started war and came and cleared Laarka [i.e. damaged the village badly], when they cleared, cleared and cleared …’ Intensive and durative expressions such as those in (21) are often accompanied by gestures; when uttering (21), the speaker raised his two arms and snapped his middle finger and the thumb to signal shooting. The V2 verbs híyy- ‘say’ and hís- ‘say.CAUS’ may also combine with an ideophonic verb, as I discuss in the next section. The intensive verb and the ideophonic verb share not only a constructional similarity but also typical expressive semantics. 2.3 Ideophonic verb plus híyy- ‘say’ and hís- ‘say.CAUS’ Dingemanse (2011:25) proposes the following cross-linguistic definition of ideophones: “Ideophones are marked words that depict sensory imagery.” They are “marked” because they tend to have special phonotactic and morphological properties. They represent “sensory imagery” in that “they evoke not just perceptions of the external world, but also kinaesthetic sensations, interoceptive experience and balance” (Dingemanse 2011:29). With this

Complex Predicates in Zargulla

109

definition as a background, I provide an overview of ideophones in Zargulla in the present section. The emphasis is on showing the structural and, to some extent, functional similarity between ideophonic constructions and complex predicates based on converbs and intensive verbs. Zargulla has a large repertoire of verbal and nominal ideophones. The two have different shapes: verbal ideophones often end in a consonant or in a vowel that is identical to the vowel in the stem, e.g. u in dúnku ‘hold someone’s waist from behind using two hands’. Most nominal ideophones end in a, e.g. ʃonnáʃonná ‘darkish or fading color’, and some in e, e.g. bólje ‘big and unattractive eye’. In addition, nominal ideophones take the suffix -(á)ma, which appears to be a derivational morpheme, although I could not establish the non-derived base form for most forms. Compare ideophonic tembáma ‘overly ripe fruit (e.g. banana that will spoil if not eaten soon)’ and tembó ‘inedible fruit of the wild fig’; the former corresponds to ideophonic nominals such as teɗɗáma ‘physically inactive person, one who works seated (especially a woman)’ and tengáma ‘one who sleeps with legs spread’. Nominal ideophones are used as attributive or referring forms and are not considered further in the present discussion. Morpho-phonologically, most ideophonic verbs in Zargulla belong to the following two groups: (i) multisyllabic words involving a geminate consonant or (ii) fully reduplicated ideophonic verbs.10 An example of ideophones from group (ii) is líɓlíɓ in (22a). In group (i), mono- or two-syllable ideophones which typically involve a geminate C2 consonant are found, cf. díppu in (22b). (22) a. wáʔ ʔáa-s INTJ

what-DAT

líɓlíɓ

hiíne

be_hasty.IDPH.V

say.PRES

‘Why are you so hasty to speak?’ b. worobbá Worobba.LOC

dok’aás-um

dok’aás-um

kás’ó

pray-DS.CNV

pray-DS.CNV

grain/food.F.ACC

díppu

yáa

híyy-um

eat.IDPH.V

DIST.DEM

say-DS.CNV

‘While (the people) were praying at Worobba [that the locusts leave the area] (the locusts) ate all the grain [i.e. young plants].’ Similar to the intensive verbs, the choice of the co-predicate with ideophonic verbs depends on transitivity. However, unlike lexical verbs, ideophonic 10 Note that reduplication is not common in other word classes in Zargulla.

110

Azeb Amha

verbs themselves are not defined for argument structure. The same ideophone may combine with either hí(yy)- or hís-; accordingly it is interpreted as intransitive or transitive. The examples in (22) illustrate intransitive ideophonic clauses where the ideophone is followed by V2 hí(yy)- ‘say’. In (23), an example of a transitive ideophonic complex predicate (i.e. ‘eat completely’) with V2 hís- ‘say.CAUS’ is given. (23) ʧ’ínʧa maák’-úkko sírs’ smart

happen-CND

do_well.IDPH

hís-í

míy-a

say.CAUS-SS.CNV

eat-IMP

néná-tt-iʃ

nay-ené

2SG.OJ-FOC-3SG.F

hunger-FUT

‘If you are smart, eat well! YOU will be hungry.’ [lit. ‘If you are smart, eat well! (It) will hunger YOU.’] Example (24) is an extract illustrating that speakers may use ideophones to define or explain concepts. The speaker explains the term luutsé, which refers to a kind of dance performed by men during funerals, and comments on how a good performance can draw attention to a dancer, particularly from young women. The complex predicates in boldface and their translations indicate that the ideophones designate motion events together with the manner and source of the movement. (24) luutsé

híkke s’ónón

Luutsé.NOM DM

toorá

hítta

híkke wálál

run.IDPH.V

say.INT

DM

k’aíso

hítta

run.IDPH.V

say.INT

ɗáwo

híkke

haʧé-gaá-nna

spear.ACC break.INT

drop.INT

DM

shoulder-LOC-INS

luús’s’a

ʔóla

híso11

híkke

góde ʧ’aláʔo

shake_vigorously.INT

leave.INT

say.CAUS.INT

DM

aside

12

maáʔa

walalíye

gélaiso

wear.INT

run.IDPH.N

enter.INT.say.CAUS.INT

híkke s’eéla DM

see.INT

hide.ACC

ʔésí 3SG.M.SJ

naʔá-z-í-nne

ʔepp-ené

hís-eésa-ttó-y

child-M.DEF-NOM-COORD

take-FUT

say.CAUS-IPFV.REL-F.DEF-NOM

11 The complex predicate ʔóla híso ‘leave’ is structurally similar to gélaiso ‘enter’ in the next line. The verbs hí(yy)- ‘say’ and hís- ‘say.CAUS’ are often encliticized to a preceding element. Thus complex predicates based on these verbs may be pronounced as a single phonological word or as two words. 12 The verb is a fused form of V1 géla ‘enter.INT’ and V2 híso ‘say.CAUS.INT’.

Complex Predicates in Zargulla nee

ʔoitʧ-eés-í

2SG.OJ

ask-IPFV.REL-NOM 3SG.M.SJ-FOC

ʔésu-tte

111

hítta

híkke

miʧ-eɗé-y

say.INT

DM

sister-PL-NOM

káʃa

ʔí-ga

s’eell-áya

híkke

tell.INT

3SG.F-LOC

see-3SG.F.OPT

DM

‘Luutse (is/means) to suddenly appear from among a large group of people and run fast (s’ónón); it is moving fast with grace (wálál), (almost) break a spear by forcefully shaking the shaft on one’s shoulder, then leave and put on the over-coat made from hide (and) start the beautiful and smooth running (walalíye). Now, he, the boy and the one to whom he proposed [lit. ‘make say: “I will take her”’] see each other. The sisters [i.e. female relatives of the boy] tell [the girl]: “It is him who is asking [to marry] you.” Let her see [him] there then!’ I do not claim that all the English words/phrases that are used to specify the meaning of the ideophonic verbs (s’ónón and wálál ) are strictly speaking part of the lexical meaning of the ideophones (the English translations are based on word-by-word translations from Zargulla to Amharic). The examples are meant to illustrate that ideophonic complex predicates express an elaborate and specialized lexical meaning, depicting actions together with the manner in which they are carried out (e.g. ‘with grace’) and/or affective, emotive meanings related to the actions (e.g. ‘beautiful’). They are in this regard very different from the semantically more general corresponding simple lexical motion verbs such as wos’- ‘run’. 2.4 Are there three-component complex predicates? I assume that the minimal form of a complex predicate involves two verbs (V1 and V2). However, complex predicates expressing motion events can occur with the verb ʔol- ‘leave’ as in (25), resulting in one of the few instances where three predicative elements constitute the verbal head. (25) s’úho-y

Ɂiʃ-í

Ɂekk-í

hám-ínne

Tsuho-NOM drive-SS.CNV take-SS.CNV go-PST

[[Ɂekk-í

hang-í ]

Ɂol-í ]]

waínne-s-ínne

take-SS.CNV

go-SS.CNV

leave-SS.CNV

what_do-3SG.M-PST

‘Tsuho drove [the cattle] and took them away. Having taken [the cattle] away, what did he do?’

112

Azeb Amha

The above example illustrates that the complex predicate comprising a V1 plus V2 unit can also function as V1 in a higher-level structure, i.e. V1[V1+V2] plus V2. This shows that the complex predicate is treated on a par with a simple verb in forming a converb, and either of these converb forms can be the V1 constituent of a complex predicate. Likewise, the complex predicate as a whole may be marked by the intensive verb formatives -a or -o, as illustrated by ʔekkí háma in (26). It is in this kind of configuration that the V1–V2 unit is said to be a dependent verb. (26) míʔ-átt-ó

ʔúsúní ʔekk-í

cow-F.DEF-ACC 3PL.SJ

hám-a diitá

take-SS.CNV go-INT

ʔeém-a

Diita.ACC bring_away-INT

‘They took the cow away and brought it to Diita.’ In (26) the expected V2 verbs hi(yy)- ‘say’ or his- ‘say.CAUS’ are not realized, thereby allowing the dependent V1[CNV] plus V2 complex predicate to occur without a fully inflected head. Why the speaker chose to omit the V2 hi(yy)‘say’ or his- ‘say.CAUS’ is not clear. In (20), we have seen a sequence of two dependent clauses each of which is realized with just the V1 constituent of the complex predicate, eliding the V2 constituent. These clauses are followed by a final main clause in which a complete complex predicate with a fully inflected V2 is attested. Somewhat similar, in (21), a series of V1 elements are realized without the V2 element and only the final V1 is followed by V2. These examples show that the independent occurrence of V1 is tolerated but there is always a V2 element in the sentence which can be linked to these incomplete complex predicates. The complex predicate illustrated in (26) goes a step further in that a V2 element (hi(yy)- ‘say’ or his- ‘say.CAUS’) is absent from the whole sentence. This makes V1[V1[CNV]+[V2[INT]] ʔekkí háma and V1[INT] ʔeéma the only verbal elements in the head position of their respective clauses, similar (at least in surface) to independent main clause verbs. A similar situation holds for converbs (see next section). However, because of the relatively low frequency of such usages, it is reasonable to analyze the omission of V2 in Zargulla as ellipsis and V1 as a structurally dependent verb. In some Ethio-Semitic languages such omissions are apparently common in complex predicates involving converbs so that the converb is treated as a main verb in these sentences. Consequently, V1[CNV] plus V2 complex predicates are analyzed as serial verbs because each of the verbs has the potential to head an independent sentence on its own. This is briefly discussed in the next section to view the Zargulla ellipsis cases in a comparative perspective.

Complex Predicates in Zargulla

113

3 Complex predicates in related languages In Omotic and other Afroasiatic languages, constructions that are similar to Zargulla complex predicates have been reported. However, at least two significant differences can be observed. First, in some languages the V2 component of a V1[CNV] plus V2 construction may be dropped. This is the case in Amharic as observed in Leslau (1995:363): “At times the gerund [i.e. converb, AA] stands alone at the end of the sentence without a principal verb … This usage of the gerund occurs when it refers to, or is a continuation of, a thought expressed in the preceding statement, or in an answer to a question.” In Tigrinya, it appears that such usage is so common that the “gerund” is treated as one of the “three basic tenses” together with the perfect and imperfect (cf. Kogan 1997:437). Kefyalew (2007), Weldu (2008), and Nazareth (2011) argue that since sentences with the converb as the predicate head are commonly attested, the V1[CNV] and V2 components of complex predicates are symmetrical, and analyze V1[CNV] plus V2 combinations as serial verb constructions. In Zayse, a closely related language or dialect of Zargulla, a converb with the cognate morpheme -í occurs as the final verb in main clauses. Because of this, Hayward (1990) analyzed the morpheme as a marker for a “converb” and a “short perfect” main predicate. Similarly, in Zargulla the converb is sometimes used as the only head of an independent clause. In a few cases it is observed that the predicate head of a sentence can also be the intensive verb (see (26)), without its V2 hí(yy)- ‘say’ or hís- ‘say.CAUS’. However, when probed, Zargulla speakers readily repeat the complex predicate in its full form. Despite its occasional uses as an independent form, I therefore analyze the V1 of complex predicates in Zargulla as a morphologically dependent form and the composite V1–V2 construction as a complex predicate, not as a serial verb construction. Second, in terms of compositionality, the Zargulla complex predicates differ slightly from parallel structures reported for both the closely related Ometo language Wolaitta as well as for Cushitic and Semitic languages, e.g. Afar and Tigrinya (Palmer 1974; Hayward 1996). In the latter languages the constructions have been analyzed as compound verbs. Güldemann (2005) reanalyzes data from Afar, Tigrinya, and a number of other East African languages as an areal phenomenon of complex predicate formation. He cautions against the use of the term compound when the component verbs are not required to be adjacent (Güldemann 2005:133). As shown in Section 2, Zargulla complex predicates are not necessarily adjacent either. In contrast,

114

Azeb Amha

V1–V2 complex predicates in Wolaitta are contiguous, and the converb morpheme on V1[CNV] undergoes phonological reduction, suggesting stronger unification of the constituent parts than what has been observed in Zargulla (cf. Azeb & Dimmendaal 2006; Azeb 2010). Moreover, Wolaitta has numerous fixed V1[CNV] plus V2 combinations, e.g. bakk’-í ʔoikk’- [slap1-CNV hold2] ‘be tight (of clothing), stick to (of cooking pans)’, the V2 of which is rarely attested in combination with other verbs. In my data on Zargulla such highly lexicalized complex predicates are not attested. On the other hand, Zargulla has V1[INT] plus V2[say/say.CAUS] constructions, the V1 of which can be derived from any lexical verb. The construction is also attested in Zayse (Hayward 1990:314). Wolaitta does not have this intensive verb construction and, as far as I know, it is not reported for other Omotic languages either. Appleyard (2001) is a typological study of the V1–V2[say/do] construction in a number of Cushitic, Omotic and Ethio-Semitic languages. He distinguishes three subtypes of the construction, which he labels “composite verbs”: (i) ideophone plus say-verbs, (ii) quotative particle plus say-verbs, and (iii) derivational particle plus say-verbs. Appleyard (2001:5) states that the third type, which is derived from existing verb roots through a regular, open-ended morphological process, is “recorded to date as a regular device only in Amharic, Tigrinya and Qafar [i.e. Afar in earlier works, AA].” An Amharic example of this is what Leslau (1995:582) labeled the “intensive”: kïffïtt alä ‘open completely’, k’ïllïtt’ alä ‘melt completely’. The V1 of these constructions is derived from the regular roots kft ‘open’ and k’lt’ ‘melt’, respectively. The Amharic intensive construction corresponds semantically and structurally to the V1[INT] plus V2[say] construction in Zargulla. As reported in Azeb (2007:17), speakers are aware of the semantic correspondences and point this out in the transcription and translation process. Amharic also makes use of an “attenuative” – the semantic opposite of the intensive – in complex predicates. Thus, from the above-mentioned roots kft ‘open’ and k’lt’ one can also derive käfätt alä ‘open slightly’ and k’älätt’ alä ‘melt slightly’ (cf. Mengistu 2002; 2010 for a detailed discussion). Such a form is not found in Tigrinya, Afar and Zargulla. Thus, while there are structural similarities between Zargulla complex predicates and so-called compound verbs in Ethiopian Afroasiatic languages, there are also morpho-phonological and lexical differences. 4 Summary and conclusion Three major types of complex predicates in Zargulla have been discussed: complex predicates based on converbs, intensive verbs, and ideophones. Structurally, the V1 component is always a dependent verb form whereas V2 is often independent. Furthermore, V1 plus V2 as a unit may function as V1 in

Complex Predicates in Zargulla

115

a hierarchically higher structure. In this case, the V2 component of V1 may be marked as a dependent (converb or intensive) verb as demonstrated in Section 3. With the exception of quotative complex predicates in which the V1 component can only be the converb form of hí(yy)- ‘say’, there are no restrictions on the lexical verb that is used as V1. In contrast, the V2 component in Zargulla is limited to the following verbs: ʔol- ‘leave, give up’, ʔer- ‘know’, the motion verbs pínn- ‘cross’, ham(m)-/hang- ‘go’, yew-/yeéd- ‘come’, ʔeem- ‘take away’, ʔeew- ‘bring’, and the utterance verbs hi(yy)- ‘say’ and his- ‘say.CAUS’, kaʃʃ- ‘tell’, and ʔoitʧ-/ʔoótʧ’- ‘ask’. Considering the semantic input, the three types of complex predicate in Zargulla are asymmetrical. V1 expresses a state or event whereas V2 specifies fixed, constructional meanings such as attitude (e.g. surprise) or experience in effecting the state/event denoted by V1 (cf. Sections 2.1.1 and 2.1.2), or the direction or orientation of a motion event expressed by V1 (cf. Section 2.1.3). Finally, V2 may determine the transitivity value as shown in Sections 2.2 and 2.3. As documented in Aikhenvald (2006:21–29) some of these functions are achieved by verb serialization in other languages, as is also claimed for Tigrinya. The constructions thus serve to express complex events that can be construed to involve various sub-events or indicate different predications on the subject. However, even though the functions look similar, even among the closely related Ometo languages, there is variation in the complex predicate constructions. Abbreviations 1, 2, 3 First, second, third person ACC Accusative ADD Additive ASS Associative AUX Auxiliary CAUS Causative CND Conditional CNV Converb COMT Comitative COORD Coordinative DAT Dative DEF Definite DIR Directional DIST Distal demonstrative DM Discourse marker DS Different subject

EMPH F FOC FUT IDPH IMP INS INT INTJ IPFV LOC LOG M N NEG NOM

Emphasis Feminine Focus Future Ideophone Imperative Instrumental Intensive Interjection Imperfective Locative Logophoric pronoun Masculine Noun, nominal Negative Nominative

116

Azeb Amha

Object Optative PL Plural POSS Possessive PRES Present tense PRES/HAB Present tense or habitual PROG Progressive aspect PROX Proximal demonstrative PST Past OJ

Q

OPT

REL SG SJ SS TEMP V VOC

Question Relative Singular Subject Same subject Temporal Verb(al) Vocative

References Aikhenvald, Alexandra Y. 2006. Serial verb constructions in typological perspective. In Alexandra Y. Aikhenvald & R. M. W. Dixon (eds.), Serial Verb Constructions: A Cross-Linguistic Typology, 1–68. Oxford: Oxford University Press. Aikhenvald, Alexandra Y. 2008. The Manambu Language from East Sepik, Papua New Guinea. Oxford: Oxford University Press. Alsina, Alex, Joan Bresnan & Peter Sells (eds.). 1997. Complex Predicates. Stanford: CSLI Publications. Ameka, Felix K. 2005. Multiverb constructions on the West African littoral: Microvariation and areal typology. In Mila Vulchanova & Tor A. Åfarli (eds.), Grammar and Beyond: Essays in Honour of Lars Hellan, 15–42. Oslo: Novus Press. Anderson, Gregory D. S. 2006. Auxiliary Verb Constructions. Oxford: Oxford University Press. Appleyard, David. 2001. The verb “to say” as a verb “recycling-device” in Ethiopian languages. In Andrzej Zaborski (ed.), New Data and New Methods in Afroasiatic Linguistics: Robert Hetzron in Memoriam, 1–11. Wiesbaden: Harrassowitz. Azeb Amha. 2007. Are -a- and -o- in the indicative verb paradigms of Zargulla nominalizers? In Azeb Amha, Graziano Savà & Maarten Mous (eds.), Omotic and Cushitic Languages Studies: Papers from the Fourth Cushitic Omotic Conference, Leiden, 10–12 April 2003, 1– 22. Cologne: Köppe. Azeb Amha. 2009. The morpho-syntax of negation in Zargulla. In W. Leo Wetzels (ed.), The Linguistics of Endangered Languages: Contributions to Morphology and Morpho-Syntax, 199–220. Utrecht: LOT.

Complex Predicates in Zargulla

117

Azeb Amha. 2010. Ideophones and compound verbs in Wolaitta revisited. In Mengistu Amberber, Brett Baker & Mark Harvey (eds.), Complex Predicates: Cross-Linguistic Perspectives on Event Structure, 259– 290. Cambridge: Cambridge University Press. Azeb Amha. 2013. Directives to humans and to domestic animals – the imperative and some interjections in Zargulla. In Marie-Claude Simeone-Senelle & Martine Vanhove (eds.), Proceedings of the 5th International Conference on Cushitic and Omotic Languages, Paris, 16–18 April 2008, 211–229. Cologne: Köppe. Azeb Amha & Gerrit Dimmendaal. 2005. Secondary predicates and adverbials in Nilotic and Omotic: A typological comparison. In Nikolaus P. Himmelmann & Eva Schultze-Berndt (eds.), Secondary Predication and Adverbial Modification: The Typology of Depictives, 299–321. Oxford: Oxford University Press. Azeb Amha & Gerrit Dimmendaal. 2006. Verbal compounding in Wolaitta. In Alexandra Y. Aikhenvald & R. M. W. Dixon (eds.), Serial Verb Constructions: A Cross-Linguistic Typology, 319–337. Oxford: Oxford University Press. Baker, Brett & Mark Harvey. 2010. Complex predicate formation. In Mengistu Amberber, Brett Baker & Mark Harvey (eds.), Complex Predicates: Cross-Linguistic Perspectives on Event Structure, 13–47. Cambridge & New York: Cambridge University Press. Bender, M. Lionel. 2003. Omotic Lexicon and Phonology. Southern Illinois: SIU Printing/Duplicating. Butt, Miriam. 1997. Complex predicates in Urdu. In Alex Alsina, Joan Bresnan & Peter Sells (eds.), Complex Predicates, 107–149. Stanford: CSLI Publications. Butt, Miriam. 2003. The light verb jungle. Harvard Working Papers in Linguistics 9. 1–49. Butt, Miriam. 2010. The light verb jungle: Still hacking away. In Mengistu Amberber, Brett Baker & Mark Harvey (eds.), Complex Predicates: Cross-Linguistic Perspectives on Event Structure, 48–78. Cambridge: Cambridge University Press. CSA = Central Statistical Authority. 1998. The 1994 Population and Housing Census of Ethiopia. Results at Country Level. Volume I: Statistical Report. Addis Ababa: Population Census Commission.

118

Azeb Amha

CSA = Central Statistical Authority. 2008. Summary and Statistical Report of the 2007 Population and Housing Census. Population Size by Age and Sex. Addis Ababa: Population Census Commission. Darmon, Chloé. 2012. Light verb constructions in Xamtanga and in the Ethiopian linguistic area. In Michael L. Marlo (ed.), Selected Proceedings of the 42nd Annual Conference on African Linguistics, 183–194. Somerville, MA: Cascadilla Proceedings Project. Dingemanse. 2011. The Meaning and Use of Ideophones in Siwu. Nijmegen: Radboud University (PhD Dissertation). Fleming, Harold C. 1976. Cushitic and Omotic. In M. L. Bender, J. Donald Bowen, Robert L. Cooper & Charles A. Ferguson (eds.), Language in Ethiopia, 34–53. London: Oxford University Press. Güldemann, Tom. 2005. Complex predicates based on generic auxiliaries as an areal feature in Northeast Africa. In Erhard F. K. Voeltz (ed.), Studies in African Linguistic Typology, 131–154. Amsterdam & Philadelphia: John Benjamins. Hayward, Richard J. 1990. Notes on the Zayse language. In Richard J. Hayward (ed.), Omotic Language Studies, 210–355. London: School of Oriental and African Studies. Hayward, Richard J. 1996. Compounding in Qafar. Bulletin of the School of Oriental and African Studies 59(3). 525–545. Himmelmann, Nikolaus P. & Eva Schultze-Berndt. 2005. Issues in the syntax and semantics of participant-oriented adjuncts: An introduction. In Nikolaus P. Himmelmann & Eva Schultze-Berndt (eds.), Secondary Predication and Adverbial Modification: The Typology of Depictives, 1–67. Oxford: Oxford University Press. Kefyalew Gebregziabher. 2007. Are there serial verbs in Tigrinya? Paper presented at the 16th International Conference on Ethiopian Studies, 2-7 July 2007, Trondheim, Norway. Kogan, Leonid. 1997. Tigrinya. In Robert Hetzron (ed.), The Semitic Languages, 424–445. London & New York: Routledge. Leslau, Wolf. 1995. Reference Grammar of Amharic. Wiesbaden: Harrassowitz. Mengistu Amberber. 2002. Quirky alternations of transitivity: The case of ingestive predicates. In Mengistu Amberber & Peter Collins (eds.), Language Universals and Variation, 1–20. Westport, CT: Praeger.

Complex Predicates in Zargulla

119

Mengistu Amberber. 2010. The structure of the light verb construction in Amharic. In Mengistu Amberber, Brett Baker & Mark Harvey (eds.), Complex Predicates: Cross-Linguistic Perspectives on Event Structure, 291–318. Cambridge: Cambridge University Press. Mengistu Amberber, Brett Baker & Mark Harvey. 2010. Introduction: Complex predicates. In Mengistu Amberber, Brett Baker & Mark Harvey (eds.), Complex Predicates: Cross-Linguistic Perspectives on Event Structure, 1–12. Cambridge: Cambridge University Press. Nazareth Amlesom Kifle. 2011. Tigrinya Applicatives in Lexical-Functional Grammar. Bergen: University of Bergen (PhD Dissertation). Palmer, F. R. 1974. Some remarks on the grammar and phonology of the “compound verbs” in Cushitic and Ethiopian Semitic. IV Congresso Internazionale di Studi Etiopici, Roma 10-15 Aprile 1972, Tomo II (Sezione Linguistica), 71–77. Rome: Accademia Nazionale dei Lincei. Rapold, Christian J. 2007. Defining converbs ten years on – a hitchhiker’s guide. Frankfurter Afrikanistische Blätter 19. 7–30. Weldu Michael Weldeyesus. 2008. A diachronic change of the gerundive in Tigrinya. Paper presented at the 39th Annual Conference of African Linguistics Athens, Georgia, April 18-20, 2008, Athens, Georgia.

Grammaticalization of Existential Auxiliaries in Koorete Binyam Sisay Mendisu Addis Ababa University Abstract This paper investigates existential auxiliaries in the Omotic language Koorete in the light of grammaticalization theory. The negative existential auxiliary ba- ‘disappear, not exist’ and the affirmative existential auxiliary yes- ‘live, exist’ are discussed in detail. Both auxiliaries exhibit three stages of grammaticalization synchronically. Regarding the negative existential auxiliary ba-, it occurs as a full verb meaning ‘disappear’ from which it grammaticalized into the negative existential verb with the more abstract meaning ‘not exist’, and then into a verb marking negation, and finally into a negation suffix. A similar path of grammaticalization was followed by the affirmative existential auxiliary yes-. The auxiliary is synchronically used as a full verb meaning ‘live’, as an existential verb meaning ‘exist’, as an auxiliary occurring with a main verb in periphrastic constructions, and as a phonologically reduced form. The description of these auxiliaries is in line with the crosslinguistically attested grammaticalization path Lexical Verb > Auxiliary Verb > Affix. Moreover, the study discusses the structure of auxiliary verb constructions in Koorete taking into account typological generalizations provided in Anderson (2006). Accordingly, three kinds of auxiliary verb constructions occur in Koorete, which are classified as split-headed. 1 Introduction This article examines the grammaticalization processes involving two existential auxiliaries in Koorete: ba- ‘not exist’ and yes- ‘exist’.1 Koorete is an Omotic language. It is one of the five languages classified under the Eastern branch of the Ometo cluster (Fleming 1976). The language is spoken by more 1

Binyam (2010) also provides some fragmentary information on the grammaticalization of existential auxiliaries in Koorete. Yet, the present study investigates the grammaticalization of affirmative and negative existential auxiliaries as a topic in its own right and in a more organized and systematic way with additional data. Furthermore, this contribution includes a discussion of the Koorete data and analysis in the light of typol i l li ti

122

Binyam Sisay Mendisu

than 150,000 speakers in the southwestern part of Ethiopia (CSA 2008). First, the paper analyzes the grammaticalization process affecting existential verbs in Koorete and then discusses the structure of auxiliary verb constructions in the language. The discussion of auxiliary verb constructions relies heavily on insights from grammaticalization theory and cross-linguistic observations made in Anderson (2006). In Section 2, I provide preliminary information about grammaticalization theory and define auxiliary verb constructions. Next, an account of the different stages of grammaticalization of the two existential auxiliaries of Koorete is provided in Section 3. This is followed by a discussion of auxiliary verb constructions in Section 4. Finally, a short summary of the findings of the paper is given in Section 5. 2 Theoretical preliminaries Grammaticalization is commonly defined as a process in which “a lexical unit or structure assumes a grammatical function, or where a grammatical unit assumes a more grammatical function” (Heine, Claudi & Hünnemeyer 1991:2). The main concern of grammaticalization, therefore, is the identification, description and explanation of linguistic items that show signs of development from lexical sources into grammatical elements and from grammatical items into even more grammatical ones. In the literature, there are two ways of understanding grammaticalization as a process. Traugott (2010:97) states that research and explanations in grammaticalization have focused in two main areas which are in contention and complementary to each other, “one invoking reduction, the other extension”. The main assumption behind those who regard grammaticalization as reduction is the change of lexical items to grammatical ones, and the phonological reduction in form. Those scholars belonging to the extension camp, on the other hand, draw attention to the expansion of the semantic and syntactic context in which a form can be used following the change of a lexical item to a grammatical one. Although some scholars label grammaticalization as a diachronic phenomenon, the literature on grammaticalization theory emphasizes that the process of grammaticalization has both a synchronic and a diachronic dimension (Heine, Claudi & Hünnemeyer 1991; Heine 2003; Lehmann 2004). Lehmann (2004:2) succinctly explains that grammaticalization “could never be a kind of diachronic change if it were not a kind of synchronic variation, too.” Having said this, the primary focus of this study is to describe the functional variation of the existential auxiliaries in Koorete as observed in the synchronic data

Grammaticalization of Existential Auxiliaries in Koorete

123

Grammaticalization as a process displays different stages which tend to overlap with one another. Heine (2003:579) identifies four interconnected stages of grammaticalization: (a) Desemanticization (or bleaching, semantic reduction), i.e. loss in meaning, content; (b) Extension (or context generalization), i.e. use in new contexts; (c) Decategorization, i.e. loss in morphosyntactic properties characteristic of the source form, including the loss of independent word status (cliticization, affixation); (d) Erosion (or phonetic reduction), i.e. loss in phonetic substance. The boundary between the different stages of grammaticalization is blurred and the various interrelated mechanisms form a chain in the process. This chain-like process from Lexical Verb > Auxiliary Verb > Affix is frequently observed in the development of auxiliaries from full verbs cross-linguistically (Heine 1993; Kuteva 2001; Anderson 2006). In relation to this, Heine (1993:66) mentions that “chains are by definition continuous structures, setting up stages along these structures must remain an arbitrary and/or artificial endeavor to some extent.” Many scholars agree that it is difficult to come up with an incontestable definition of auxiliary that accounts for the nature of auxiliaries in all languages (Heine 1993; Anderson 2006). Be that as it may, Anderson (2006:4) defines an auxiliary verb from the perspective of a chain-like grammaticalization process as: “[A]n item on the lexical verb-functional affix continuum, which tends to be at least somewhat semantically bleached, and grammaticalized to express one or more of a range of salient verbal categories, most typically aspectual and modal categories, but also not infrequently temporal, negative polarity, or voice categories.” Auxiliary verb constructions are formed by joining an auxiliary verb and a lexical verb. In line with this, Anderson (2006:7) defines the auxiliary verb construction as follows: “[A] mono-clausal structure minimally consisting of a lexical verb element that contributes lexical content to the construction and an auxiliary verb element that contributes some grammatical or functional content to the construction.” In Section 4, the auxiliary verb constructions in Koorete are discussed in detail. Prior to that, however, Section 3 deals with the grammaticalization chains of existential auxiliaries in the language

124

Binyam Sisay Mendisu

3 Grammaticalization of existential auxiliaries in Koorete There are two existential auxiliaries in Koorete: ba- ‘not exist’ and yes- ‘exist’.2 Synchronically, the two verbs exhibit different stages in grammaticalization, which will be described in detail in this section. For this purpose, basic information on the structure of the verb in Koorete is in order. Binyam (2010:71) identifies three types of verb forms: (i) simple verbs which only consist of the verb stem (i.e. VERB), (ii) compound verbs that reduplicate the verb stem and insert a clitic pronoun between the reduplicated stems (i.e. VERB1-PRO-VERB1), and (iii) complex verbs that consist of a main verb and an auxiliary verb between which a clitic pronoun is inserted (i.e. VERB-PRO-AUX). Furthermore, these three verb forms may contain additional markers for person, aspect/mood and tense (Binyam 2010:81). Typologically, Koorete is an aspect-prominent language, but it marks both aspect and tense morphologically. According to Binyam (2010:82 ff.), the primary aspect distinction (also called ASPECT 1) is between perfective, marked by the suffix -d, and imperfective, unmarked or marked by the suffix -g. Two additional secondary aspects are distinguished in Koorete (called ASPECT 2), namely the perfect, marked by the suffix -i, which always cooccurs with the perfective aspect marker, and the progressive, marked by the suffix -iya and which always co-occurs with the imperfective aspect marker. Regarding tense, Koorete distinguishes between present, marked by the suffix -e, and past, marked either by the suffix -o or the suffix -ʧa (for further details, cf. Binyam 2010:89 f.). These suffixes can be combined in various ways yielding specific aspect/tense categories (cf. Binyam 2010:108 for an overview). With regard to person marking, Koorete employs agreement suffixes, which are restricted to simple verb forms,3 as well as pronoun clitics, which only occur with compound or complex verb forms (cf. Binyam 2010:114 f.). The order of the aspect, tense and person morphemes depends on the verb form. In simple verbs, the primary aspect morphemes are immediately suffixed to the verb stem, and then followed by secondary aspect and tense morphemes or only tense morphemes (Binyam 2010:108 f.). The agreement suf2

3

In this contribution, the meanings ‘disappear’ and ‘not exist’ for ba- and ‘live’ and ‘exist’ for yes- are considered as two different stages in the grammaticalization process. This is based on the assumption that the meaning ‘not exist’/‘exist’ is more abstract and semantically bleached than ‘disappear’/’live’ for lacking physical presence and immediacy. Yet, it is plausible to claim that the two meanings of the auxiliary verbs are multiple related meanings (i.e. polysemy) of the verbs. There are two distinct paradigms of agreement suffixes for verbs in declarative vs. i t ti l (Bi 2010 115)

Grammaticalization of Existential Auxiliaries in Koorete

125

fixes, whose occurrence depends on various factors (cf. Binyam 2010:115 ff. for further details), are attached to the aspect/tense morphemes. In compound verb forms, the first occurrence of the verb stem does not bear aspect or tense morphemes but is only marked as a dependent element. Only the reduplicated second verb stem bears information on aspect and tense (Binyam 2010:87). In complex verb forms, the primary and secondary aspect morphemes are suffixed to the initial verb while the tense morphemes are suffixed to the reduplicated verb stem or the auxiliary verb (Binyam 2010:108 f.). In addition to aspect, tense and person morphemes, focus and clause markers also can be attached to verbs (cf. Binyam 2010:157 ff.). 3.1 Verb ba- ‘disappear, not exist’ Synchronically, the verb ba- ‘disappear, not exist’ shows four stages of use in Koorete: full verb, existential verb, negation marker, and negation suffix. 3.1.1 Full verb baThe full verb ba- has the lexical meaning ‘disappear’. As a full verb, it can be marked for different aspect/tense categories like any other full verb of the language, as shown in (1). (1) a. ʔis-i she-NOM

ba-d-o disappear-PFV-PST1

‘She disappeared.’ b. kana-i dog-NOM

ba-d-i-ko

yes-e

disappear-PFV-PRF-FOC.DECL

exist-PRS

‘A dog has DISAPPEARED.’4 c. kana-i dog-NOM

ba-d-i-y-a

yes-e?

disappear-PFV-PRF-EPN-FOC.Q

exist-PRS

‘Has a dog DISAPPEARED?’ d. ʔes-i he-NOM

ba-tta-ko

baa-ne

disappear-DEP-FOC.DECL

not_exist-PRS

‘He DISAPPEARS.’ As can be seen in the examples in (1), the full verb ba- ‘disappear’ appears in the past perfective in (1a), in the present perfect in (1b) and (1c), and in the present habitual in (1d).

4

Th

f

i th t

l ti

i di t

f

126

Binyam Sisay Mendisu

3.1.2 Existential verb baIn the first grammaticalization stage, ba- still occurs as a lexical verb, but in the more abstract meaning ‘not exist’, as compared to the lexical verb meaning ‘disappear’ discussed in Section 3.1.1, i.e. ba- developed into an existential verb through semantic bleaching. Unlike the examples in (1), the use of ba- as an existential verb in main clauses is limited to the present and past tenses, i.e. it is not marked for aspect. In other words, while the occurrence of the existential verb ba- ‘not exist’ in (2) is mainly limited to simple verb forms (see (3) for an exception), the lexical verb ba- ‘disappear’ in (1) may occur in simple, compound and complex verb forms. (2) a. nen-i

ba-na-ko

you-NOM

not_exist-2.SG-FOC.DECL

‘You (SG) do NOT EXIST.’ (i.e. the addressee is not in the vicinity of the speaker, for instance, during an intended activity) b. ʔes-i he-NOM

baa-s-so not_exist-3SG.M-FOC.DECL

‘He does NOT EXIST.’ (i.e. he is not in the vicinity of the speaker) In the present tense, the subject agreement suffix and the focus marker are added to the existential verb ba- ‘not exist’. This is quite different from the construction with the full verb ba- ‘disappear’ in (1d) above. First, whereas (1d) is a compound verb form in which the verb ba- occurs twice, (2b) employs the simple verb form. Second, in (2b), the subject agreement marker is suffixed to the verb, while person marking is absent in (1d). In the past perfect, the verb ba- ‘not exist’ occurs as shown in (3). (3) zawa-ka

mitsi ba-d-i-ko

ye-ʧa

house-LOC wood disappear/not_exist-PFV-PRF-FOC.DECL exist-PST2

i. ‘There HAD NOT BEEN wood in the house.’ ii. ‘Wood HAD DISAPPEARED (lit. had been lost) in the house.’ Interestingly, the sentence is ambiguous between the two readings of ba- as negative existential auxiliary ‘not exist’ and full verb ‘disappear’. It is common in grammaticalization to come across ambiguous structures when the development from one stage to the other is not completed (cf. Heine 1993; Anderson 2006).

Grammaticalization of Existential Auxiliaries in Koorete

127

3.1.3 Negation marker baThe next stage of grammaticalization noted in Koorete is the occurrence of the existential auxiliary ba- as a marker of verbal negation. Anderson (2006) states that the development of an auxiliary verb into a negation-marking element is attested in many languages. As to Koorete, the existential auxiliary ba- ‘not exist’ always occurs together with a main verb forming an auxiliary verb construction denoting verbal negation. The existential auxiliary follows the main verb. No morphological element intervenes between the main verb and the auxiliary. (4) a. ʔis-i she-NOM

dana ʔush-iya

ba-nni-ko

beer

not_exist-3SG.F-FOC.DECL

drink-PROG

‘She is NOT drinking beer.’ b. dana ʔush-iya beer

drink-PROG

ba-d-i-ko

ʔi

ye-ʧa

not_exist-PFV-PRF-FOC.DECL

she

exist-PST2

‘She was NOT drinking beer.’ As (4) illustrates, the verb ʔush- ‘drink’ is negated by the negative existential auxiliary ba- which immediately follows it. The full verb ba- ‘disappear’ is also negated by the negative existential auxiliary like any other verb. (5) ʔis-i

ba-d-o

ba-nni-ko

she-NOM disappear-PFV-PST1 not_exist-3SG.F-FOC.DECL

‘She did NOT disappear.’ In the grammaticalization continuum, two mechanisms overlap in the development of the existential auxiliary ba-. That means the auxiliary extends the contexts in which it can occur, since it can be found with any lexical verb as a negative marker, and it shows decategorization as it distinctly occurs as negative existential auxiliary verb. 3.1.4 Negative suffix baIn the present habitual and the past habitual, the negative existential auxiliary is suffixed to the verb, thereby undergoing a morphophonemic process in which the initial plosive b is weakened to w. Here, the auxiliary shows phonetic erosion, i.e. it has lost its status as an independent word.

128

Binyam Sisay Mendisu

(6) a. ʔis-i

dana ʔush-u-wa-nni-ko

she-NOM

beer

drink-PRS-not_exist-3SG.F-FOC.DECL

‘She does (will) NOT drink beer.’ b. dana ʔush-u-wa-d-i-ko beer

drink-PRS-not_exist-PFV-PRF-FOC.DECL

ʔi

ye-ʧa

she

exist-PST2

‘She used NOT to drink beer.’ In related Ometo languages, similar morphemes are employed to mark negation. In Haro, the negation is expressed by the negative suffix -(ww)á, which is always attached to verbs (Hirut 2003:190 ff.). Also in Maale the suffixes -iba and -uwa mark negation (Azeb 2001:119). In fact, Azeb (2001:119) also argues that the negative marker in Maale has gradually developed from the negative existential verb ba-. Gelderen (2008) notes that cross-linguistically there are two grammaticalization paths in the development of negative markers: one departs from indefinite pronouns, the other from verbs. The Koorete data given in (1) to (6) clearly shows that the verb ba- is undergoing grammaticalization: from full verb to existential verb to auxiliary verb for verbal negation to negative suffix. Thus, synchronically, the language exhibits the grammaticalization chain from lexical verb to negative suffix, i.e. the different interrelated grammaticalization mechanisms given by Heine (2003) as semantic reduction, context generalization, decategorization and phonetic reduction can all be observed to different extents. 3.2 Verb yes- ‘live, exist’ In this section, the different grammaticalization stages in which the verb yes‘live, exist’ is attested are examined. 3.2.1 Full verb yesIn Koorete, yes- appears as a full verb carrying the lexical meaning ‘live’. In this function, yes- ‘live’ appears in all the different aspect/tense categories akin to other full verbs. This use of yes- ‘live’ corresponds to that of ba- as a full verb meaning ‘disappear’ in Section 3.1.1. (7)

a. ʔes-i he-NOM

yes-s-o5 live-PFV-PST1

‘He lived.’

5

Note that the perfective marker -d totally assimilates to immediately preceding coronal bt t i * d ( f Bi 2010 59)

Grammaticalization of Existential Auxiliaries in Koorete b. yes-s-i-ko live-PFV-PRF-FOC.DECL

ʔu

ye-ʧa

they

exist-PST2

129

‘They had LIVED.’ c. yes-s-i-ko live-PFV-PRF-FOC.DECL

ʔu

yes-e

they

exist-PRS

‘They have LIVED.’ d. yes-iya-ko live-PROG-FOC.DECL

ta

yes-e

I

exist-PRS

‘I am LIVING.’ The full verb yes- ‘live’ occurs in the past perfective in (7a), in the past perfect in (7b) in the present perfect in (7c) and present progressive in (7d). 3.2.2 Existential verb yesIn the first grammaticalization stage, yes- appears as an existential verb with the more abstract meaning ‘exist’ as compared to ‘live’, which indicates physical presence and immediacy, as described in Section 3.2.1. This grammaticalization stage of yes- corresponds to that of the negative existential verb ba- ‘not exist’ discussed in 3.1.2. While yes- with the meaning ‘live’ can be used in all the tense/aspect categories (cf. (7)), only the present and past forms can be used with the existential verb yes- ‘exist’. (8)

a. ʔes-i he-NOM

yes-a-ko

yes-e

exist-DEP-FOC.DECL

exist-PRS

i. ‘He EXISTS.’ (i.e. he is currently in the vicinity of the speaker) ii. ‘He LIVES.’ b. ʔes-i he-NOM

yes-a-ko

ye-ʧa

exist-DEP-FOC.DECL

exist-PST2

i. ‘He EXISTED.’ ii. ‘He used to LIVE.’ Example (8a) is in the present and (8b) is in the past. Interestingly, the two examples in (8) are ambiguous, allowing the readings ‘exist’ and ‘live’. The same ambiguity was observed with the verb ba- ‘disappear, not exist’ in 3.1.2. Since grammaticalization forms a continuum, it is typical to have meaning overlap and ambiguity between the imprecisely defined grammaticalization stages (Heine 1993; Anderson 2006).

130

Binyam Sisay Mendisu

3.2.3 Auxiliary yes- in complex predicates When used as an existential auxiliary verb, yes- ‘exist’ has to co-occur with a main verb in an auxiliary verb construction. The tense marker in such a construction is always suffixed to the auxiliary. By combining the main verb and the existential auxiliary, four aspect/tense categories can be formed: present progressive, past progressive, present perfect and past perfect. (9)

a. zawa woon-g-iya-ko house buy-IPFV-PROG-FOC.DECL

ta

yes-e

I

exist-PRS

‘I am BUYING a house.’ b. zawa woon-g-iya-ko house buy-IPFV-PROG-FOC.DECL

ta

ye-ʧa

I

exist-PST2

‘I was BUYING a house.’ (10) a. yoo-d-i-ko eat-PFV-PRF-FOC.DECL

ʔu

yes-e

they

exist-PRS

‘They have EATEN.’ b. yoo-d-i-ko

ʔu

ye-ʧa

eat-PFV-PRF-FOC.DECL they

exist-PST2

‘They had EATEN.’ Examples (9a–b) are in the present progressive and past progressive respectively, whereas (10a–b) are in the present perfect and past perfect. As can be seen, clitic pronouns usually appear between the main verb and the auxiliary yes- ‘exist’. In contrast, no elements were allowed to intercede between the main verb and the negative auxiliary ba- ‘not exist’ as shown in (4) in Section 3.1.3. 3.2.4 Reduced form of yesKoorete finally exhibits a case in which the existential auxiliary yes- ‘exist’ optionally merges with the main verb. The structures given in (a) and (b) in (11) and (12) below may be used alternatively: (11) a. tan-i I-NOM

dana-ko

ʔush-iya

yes-e

beer-FOC.DECL

drink-PROG

exist-PRS

‘I am drinking BEER.’

Grammaticalization of Existential Auxiliaries in Koorete b. tan-i I-NOM

dana-ko

ʔush-iya-s-e

beer-FOC.DECL

drink-PROG-exist-PRS

131

‘I am drinking BEER.’ (12) a. tan-i I-NOM

doro-ko

woon-g-iya

ye-ʧa

sheep-FOC.DECL

buy-IPFV-PROG

exist-PST2

‘I was buying SHEEP.’ b. tan-i I-NOM

doro-ko

woon-g-iya-ʧa

sheep-FOC.DECL

buy-IPFV-PROG-PST2

‘I was buying SHEEP.’ As can be noted from (11b) and (12b), the first open syllable of the existential auxiliary yes- is deleted while the remaining syllable merges with the preceding main verb. In summary, three overlapping, chain-like grammaticalization stages are observed with respect to yes- ‘live, exist’ and the negative existential auxiliary verb ba- ‘disappear, not exist’. Koorete is not unique in exhibiting this phenomenon. For example, as reported in Anderson (2006:6 f.), Xakas, a Turkic language of Siberia, synchronically shows the use of the same form as a lexical verb, as an auxiliary verb within an auxiliary verb construction and as a suffix. 4 Auxiliary verb constructions Auxiliary verb constructions generally consist of a lexical and an auxiliary verb. While the lexical verb contributes the lexical content, the auxiliary verb carries grammatical information (Anderson 2006:7). As discussed in Section 3, the verbs ba- ‘not exist’ and yes- ‘exist’, which are used in auxiliary verb constructions in Koorete, are in an intermediate stage of grammaticalization: between a full verb and an affix. In Koorete, three different auxiliary verb constructions have been identified involving these existential auxiliaries: (a) Main verb plus negative existential auxiliary: VERB-NEG.AUX (b) Main verb with optional clitic subject pronoun plus affirmative existential auxiliary: VERB-(PRO)-AFF.AUX (c) Main verb followed by the negative existential auxiliary plus optional clitic subject pronouns followed by the affirmative existential auxiliary: VERB-NEG.AUX-(PRO)-AFF.AUX The way inflection is marked within an auxiliary verb construction varies from language to language. The component within the auxiliary verb construction that carries the inflectional suffixes is considered as a head. For

132

Binyam Sisay Mendisu

example, if inflectional markers appear on the auxiliary, then the auxiliary is taken as the head of the construction. In this regard, Anderson (2006:25) identifies the following four main head types, and a fifth one which is a combination of the last two: (a) Auxiliary-headed: the auxiliary is marked for inflectional categories (b) Lexical-headed: the lexical verb is inflected for inflectional categories (c) Double-headed: the same inflectional suffixes occur on both the auxiliary and the lexical verb (d) Split-headed: some inflectional suffixes occur on the auxiliary and others in the lexical verb In addition to these, Anderson (2006) states that some languages exhibit the unusual pattern which he labeled “split/doubled”, which involves a combination of the patterns in (c) and (d) above. In what follows, the three auxiliary verb constructions in Koorete will be examined as to their internal structure and inflectional headedness. 4.1 VERB-NEG.AUX The VERB-NEG.AUX construction consists of a lexical verb and the negative existential auxiliary. While the main verb contributes the lexical content, the negative existential auxiliary marks negation. The lexical verb and the auxiliary are syntactically adjacent and no morphological element can occur between these two constituents. This construction is attested in the past perfective and present progressive aspect/tense categories. (13) ʔis-i

[ yoo-d-o

ba-nni-ko]

she-NOM come-PFV-PST1 not_exist-3SG.F-FOC.DECL

‘She did NOT come.’ (14) ʔis-i

[ geh-iya

ba-nni-ko]

she-NOM sleep-PROG not_exist-3SG.F-FOC.DECL

‘She is NOT sleeping.’ Examples (13) and (14) are negative sentences in the past perfective and present progressive, respectively. In both cases, the main verb is followed by the negative existential auxiliary. In (13), the main verb carries the perfective aspect and the past tense markers. In (14), the main verb takes the progressive aspect marker. The negative auxiliary in both cases carries the person agreement and focus/sentence type markers. Since the main verb and the auxiliary carry inflectional markers, neither of them can be considered to be the head

Grammaticalization of Existential Auxiliaries in Koorete

133

of the construction. According to Anderson (2006), this type of auxiliary verb construction is split-headed. 4.2 VERB-(PRO)-AFF.AUX Koorete also has an auxiliary verb construction with three constituents: main verb plus clitic subject pronoun plus affirmative existential auxiliary. The main verb and the affirmative auxiliary are obligatory, whereas the clitic pronoun between them is optional. Affirmative sentences in the present perfect, past perfect, present progressive and past progressive employ this auxiliary verb construction. Unlike the construction discussed in Section 4.1, the link between the main verb and the auxiliary verb in this auxiliary verb construction is not tight due to the optional clitic pronoun between the two verbal constituents. Examples of the present perfect, past perfect, present progressive and past progressive forms are given below: (15) ʔera

zawa-ko

[han-d-i

ʔe yes-e]

knowledge house-FOC.DECL go-PFV-PRF he exist-PRS

‘He has gone to SCHOOL.’ (16) ʔera

zawa-ko

[han-d-i

ʔe ye-ʧa]

knowledge house-FOC.DECL go-PFV-PRF he exist-PST2

‘He had gone to SCHOOL.’ (17) ʔera

zawa-ko

[han-g-iya

ʔe yes-e]

knowledge house-FOC.DECL go-IPFV-PROG he exist-PRS

‘He is going to SCHOOL.’ (18) ʔera

zawa-ko

[han-g-iya

ʔe ye-ʧa]

knowledge house-FOC.DECL go-IPFV-PROG he exist-PST2

‘He was going to SCHOOL.’ In all four examples, aspect is marked on the main verb while tense is realized on the auxiliary verb. In addition, the clitic pronouns indicate the person of the subject. Akin to the constructions described in 4.1, inflectional suffixes appear on both the main verb and the auxiliary. The auxiliary verb constructions exemplified in (15)–(18) are thus cases of the split-headed type according to Anderson (2006). 4.3 VERB-NEG.AUX-(PRO)-AFF.AUX The third type of auxiliary verb construction that is used in Koorete consists of a main verb a negative existential auxiliary a clitic pronoun and an af

134

Binyam Sisay Mendisu

firmative existential auxiliary. The construction involves the negative polarity forms of the present perfect, past perfect and past progressive aspect/tense clauses. Although the negative existential verb carries perfective and perfect aspect morphemes, these morphemes are neutralized in this construction in the sense that they do not mark aspect anymore (Binyam 2010:138). The clitic pronoun is optional. When it is used, its position is between the two existential auxiliaries. The following are examples from the present perfect, past perfect and past progressive: (19) ʔogas’o-i

katsa [kas-s-o

Ogatso-NOM food

ba-d-i-ko

ʔe yes-e]

cook-PFV-PST1 not_exist-PFV-PRF-FOC.DECL he exist-PRS

‘Ogatso has NOT cooked food.’ (20) ʔogas’o-i

katsa [kas-s-o

Ogatso-NOM food

ba-d-i-ko

ʔe ye-ʧa]

cook-PFV-PST1 not_exist-PFV-PRF-FOC.DECL he exist-PST2

‘Ogatso had NOT cooked food.’ (21) ʔogas’o-i

katsa [kas-iya

Ogatso-NOM food

ba-d-i-ko

ʔe ye-ʧa]

cook-PROG not_exist-PFV-PRF-FOC.DECL he exist-PST2

‘Ogatso was NOT cooking food.’ The constructions in (19) to (21) contain the negative and the affirmative existential auxiliaries. The negative existential auxiliary carries three suffixes: the perfective marker -d, the perfect maker -i and the focus/sentence type marker -ko. The aspect markers are neutralized morphemes (cf. Binyam 2010:138 for more discussion) while the suffix -ko still functions as focus/sentence type marker. In all the tense/aspect categories, the affirmative existential auxiliary takes the tense marker. In (19) and (20), the past tense and perfective aspect suffixes are added to the main verb. Likewise, in (21), the progressive aspect marker -iya is suffixed to the main verb. Typologically, the auxiliary verb constructions exemplified in (19) to (21) fall into the split-head type,6 just like the auxiliary verb constructions discussed in Section 4.1 and 4.2. In all the three types of auxiliary verb constructions, inflectional suffixes occur on both the main verb and the auxiliary.

6

If the aspect markers were not analyzed as neutralized morphemes, the auxiliary verb constructions in (19) and (20) would be of the split/doubled-head type because the neut li d h ld t i

Grammaticalization of Existential Auxiliaries in Koorete

135

5 Summary In the present contribution, I discussed the different grammaticalization stages of the two existential auxiliaries in Koorete: yes- ‘exist’ and ba- ‘not exist’. Both of them exhibit synchronically three stages of grammaticalization. They function as full verbs on the one end of the continuum and as a phonetically eroded grammatical markers at the other end. This is an instance of the cross-linguistically attested grammaticalization path Lexical Verb > Auxiliary Verb > Affix. Note, however, that this phenomenon is not unique to Koorete; it is also attested in related Ometo languages. This might imply that the discussed grammaticalization processes in Koorete probably took place in an earlier Ometo stage. Further, three types of auxiliary constructions containing yes- ‘exist’ or ba- ‘not exist’ were identified for Koorete: (i) a lexical verb plus the negative existential auxiliary, (ii) a lexical verb plus the affirmative existential auxiliary, and (iii) a lexical verb, the negative auxiliary plus the affirmative auxiliary. I demonstrated that auxiliary verb constructions in Koorete are of the split-headed type, since inflectional suffixes occur in both the auxiliary and the main verb. Abbreviations 2, 3 Second, third person AFF.AUX Affirmative existential auxiliary DAT Dative DECL Declarative DEF Definite DEP Dependant EPN Epenthetic F Feminine FOC Focus IPFV Imperfective LOC Locative M Masculine

NEG NEG.AUX NOM PFV PRF PRO PROG PRS PST Q SG

Negation, negative Negative existential auxiliary Nominative Perfective Perfect Pronoun (clitic) Progressive Present Past Interrogative Singular

136

Binyam Sisay Mendisu

References Anderson, Gregory D. S. 2006. Auxiliary Verb Constructions. Oxford: Oxford University Press. Azeb Amha. 2001. The Maale Language. Leiden: LOT. Binyam Sisay Mendisu. 2010. Aspects of Koorete Verb Morphology. Cologne: Köppe. CSA. 2008. Summary and Statistical Report of the 2007 Population and Housing Census. Population Size by Age and Sex. Addis Ababa: Population Census Commission. Fleming, Harold C. 1976. Omotic overview. In Lionel M. Bender (ed.), The Non-Semitic Languages of Ethiopia, 299–323. East Lansing, MI: African Studies Center, Michigan State University. Gelderen, Elly van. 2008. Negative cycles. Linguistic Typology 12. 195–243. Heine, Bernd. 1993. Auxiliaries: Cognitive Forces and Grammaticalization. New York: Oxford University Press. Heine, Bernd. 2003. Grammaticalization. In Brian D. Joseph & Richard D. Janda (eds.), The Handbook of Historical Linguistics, 575–601. Oxford: Blackwell. Heine, Bernd, Ulrike Claudi & Frederike Hünnemeyer. 1991. Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Hirut Woldemariam. 2003. A Grammar of Haro with Comparative Notes on the Ometo Linguistic Group. Addis Ababa: Addis Ababa University (PhD Dissertation). Kuteva, Tania A. 2001. Auxiliation. Oxford: Oxford University Press. Lehmann, Christian. 2004. Theory and method in grammaticalization. Zeitschrift für Germanistische Linguistik 32. 152–187. Traugott, Elizabeth Closs. 2010. Grammaticalization. In Andreas H. Jucker & Irma Taavitsainen (eds.), Historical Pragmatics, 97–126. Berlin & New York: De Gruyter Mouton.

Benefactive Applicative Periphrases with yɨw- ‘give’ in Xamtanga Chloé Darmon Dynamique Du Langage, Université Lumière Lyon 2 Abstract The Central Cushitic (Agaw) language Xamtanga marks applicative operations by mean of periphrases. This paper deals with a type of biverbal construction employed to license the expression of an additional beneficiary participant. The construction involves the converb form of a lexical verb followed by the valency operator yɨw- ‘give’. The characteristics of such converb-finite verb sequences are presented in comparison with other formally identical V1-V2 patterns attested in the language, arguing that giveperiphrases function like complex predicates. The semantic role assigned by the verb operator to its new argument is also discussed. 1 Introduction Xamtanga is a Central Cushitic (Agaw) language of the northern Ethiopian Highlands numbering approximately 200 000 speakers. Based on data from the Abɨrgälle dialect collected during fieldworks carried out between 2008 and 2011, this paper discusses the semantic and morphosyntactic properties of Xamtanga benefactive applicative periphrases with yɨw- ‘give’.1 According to Peterson (2007:1), “applicative constructions are a means some languages have for structuring clauses which allow the coding of a thematically peripheral argument or adjunct as a core-object argument.” Like other valencychanging operations, applicatives might be marked by a morphological modification of the verb, or by an analytic/periphrastic construction, as in Xamtanga. Creissels (2010:30) defines periphrastic applicatives as “biverbal constructions functionally comparable to monoverbal constructions headed by applicative verb forms.” They are made of two elements: a lexical verb that 1

I would like to thank the French Center of Ethiopian Studies for their financial support as well as the participants of the working group atelier morphosyntaxe in Dynamique Du Language for their valuable comments and suggestions when I first presented these data in 2011. I am also grateful to Workayehu Chekole, my main consultant on Xamtanga, and Tilahun Ayele, who considerably helped me with Amharic translations, for their patience and precision as we were working on this topic.

138

Chloé Darmon

determines the type of event encoded, and a verb operator that acts as a valency operator licensing an additional participant. Xamtanga displays three distinct applicative periphrases depending on the verb operator employed: yɨw- ‘give’, y- ‘say’, or bär- ‘leave’, for it is the verb operator that determines the semantic role assigned to the additional argument. The verbs yɨw- ‘give’ and y- ‘say’ are specialized in the expression of benefaction, while bär- ‘leave’ is used either in malefactive constructions or in generic periphrases that refer to a non-specified affected participant. For reasons of space limitation, I shall only focus on yɨw- ‘give’ periphrases here. It is worth mentioning that ‘give’ is the commonest verb operator among languages of the world that have benefactive applicative periphrases (hence abbreviated BAPs). In his cross-linguistic study of BAPs, which greatly contributed to my analysis of the phenomenon in Xamtanga, Creissels (2010:34) ascribes the widespread use of give-periphrasis to the fact that the verb ‘give’ can encode recipient-like participants as in ‘I baked a cake and gave it to the children’, from which follows the benefactive reading: ‘I baked a cake for the children’. 2 General structure of Xamtanga give-BAPs In its lexical use, yɨw- is a ditransitive verb that licenses a direct object (O) and a dative argument (X) marked by the case suffix -z or -s. The two wordorders SOXV and SXOV are possible with verbs of giving: (1)

a. k’uršan-dyän-t-a money-DEF-ACC-ENP

mɨkän-i-z

yɨw-ŋ-äkʷ

church-DEF-DAT

give-3PL-IPFV

‘They give the money to the church.’ b. guläšu Guläšu

šäggä

ɨfärä-s

abäbä

yɨw -u

good

child-DAT

flower

give-PFV[3M.SG]

‘Guläšu gave flowers to a beautiful girl.’ The structure of the BAP is illustrated in (2): the lexical component is in a converbal form in V1 position, whereas the operator in V2 is in a finite form, inflected like in an independent clause. The dative-marked beneficiary introduced by the periphrasis may be placed either between V1 and V2 (2a) or before the direct object (2b) without any apparent change in meaning.

Benefactive Applicative Periphrases with yɨw- ‘give’ in Xamtanga 139 (2)

a. abbäbä dɨbdabe-d s’af-ø Abbäbä letter-DEF

bɨrtukʷan-sɨ

write-3M.SG[CNV] Bɨrtukʷan-DAT

yɨw-u give-PFV[3M.SG]

‘Abbäbä wrote a letter for Bɨrtukʷan (in her favor, because she cannot write).’ (lit. ‘Abbäbä wrote the letter and gave (it) to Bɨrtukʷan.’) b. abbäbä Abbäbä

bɨrtukʷan-sɨ dɨbdabe-d s’af-ø Bɨrtukʷan-DAT letter-DEF

yɨw-u

write-3M.SG[CNV] give-PFV [3M.SG]

‘Abbäbä wrote the letter for Bɨrtukʷan (in her favor, because she cannot write).’ (lit. ‘Abbäbä wrote the letter and gave (it) to Bɨrtukʷan.’) Looking at the sentences above, the main issue that arises concerns the possibility to differentiate between verb operator and lexical instances of ‘give’, i.e. between the benefactive reading suggested in the translation (‘do something for someone’) and the literal reading consisting of a succession of actions (‘do something and give it to someone’). In Section 3, I argue that Xamtanga BAPs exhibit both morphosyntactic and semantic characteristics that distinguish them from biclausal constructions and rather tend to relate them to complex predicates. Another subject that needs to be addressed is the function of the BAP: examples (1a) and (1b) show that the dative case itself prototypically encodes recipient-like beneficiaries, one might thus question the applicative nature of Xamtanga give-periphrases on the grounds that their dative NP could be licensed by monoverbal ditransitive predicates as well. The semantic role of the dative argument assigned by the valency operator yɨw- is discussed in Section 4. 3 Valency operator vs. lexical uses of ‘give’ 3.1 Survey of converb-verb constructions in Xamtanga As previously mentioned, the lexical component of the BAP in V1 position is in a converbal form. The term converb refers to a paradigm of non-finite verb forms “morphologically distinct from main verbs as well as dependent verb forms occurring in conditional, purposive, or reason clauses” (Azeb & Dimmendaal 2006:394). In Xamtanga, converbs are not marked for tense or aspect, but they are inflected for person with the realis mood set of subject agreement morphemes which is used with most adverbial subordinate clauses:2 1SG -ø (+ ablaut), 2SG and 3F.SG -r, 3M.SG -ø, 1PL -n, 2PL -rn, 3PL -ŋ. The 2

As against the irrealis mood set of subject agreement markers that is employed with part of the jussive paradigm and purpose clauses.

140

Chloé Darmon

-ŋ. The following examples illustrate the three major types of constructions the converb is involved in: conjoining of a succession of events (3a), expressing manner modification of a finite verb (3b), and joint predication (3c). (3) a. ŋɨ=xɨrtä-s 3M.SG.POSS=sleep-ABL

gʷɨ-ø

ŋɨ=gas’-ɨd

get_up-3M.SG[CNV]

3M.SG.POSS=face-DEF

aqa-r-ø

ŋɨ=t’amänä-d

xʷɨ-ø

wash-MEDP-3M.SG[CNV]

3M.SG.POSS=breakfast-DEF

eat-3M.SG[CNV]

kɨndɨŋ

ŋɨn-i-z

gʷä

fir-u

study

house-DEF-GEN

to

go-PFV[3M.SG]

‘He woke up, washed his face, ate his breakfast, and went to school.’ b. albb-ɨd cloth\PL-DEF

gʷäyy-ɨr

aqa-s-ɨ-č

sit-3F.SG[CNV]

wash-CAUS-PFV-3F.SG

‘She washed the cloth sitting.’ c. ŋɨn-i-s house-DEF-ABL

fɨ-ø

fir-u

go_out-3M.SG[CNV]

go-PFV[3M.SG]

‘He left the house.’ The number of converbs occurring in a sequential construction like (3a) is theoretically unlimited, as each of them represents a distinct action. On the other hand, (3b) and (3c) do not allow for more than one converb before the main verb. In (3b) the position of the direct object of ‘wash’ plays a role in the understanding of ‘sit’ as an adverbial modifier describing a single situation, because the same sentence with the object preceding the final verb (gʷäyyɨr albbɨd aqasɨč) would preferably be interpreted ‘She sat down and washed the cloth’. The structure given in (3c), particularly frequent with motion and speech verbs in V2, straightforwardly corresponds to a complex predicate. Two characteristics set converbs that participate in joint predication apart from sequential and adverbial ones: no element may be inserted between V1 and V2, and V1 cannot bear negation markers. While negation of other Xamtanga verb forms is primarily achieved through infixation of a negative morpheme between the verb stem and subject agreement (see Appleyard 1984 for more details), converbs, like subordinates expressing simultaneity, anteriority, and subsequence, are negated with the suffix -ink’ä added after the bare lexical base. Consider (4a) and (4b).

Benefactive Applicative Periphrases with yɨw- ‘give’ in Xamtanga 141 (4) a. ŋɨ=t’amänä-d 3M.SG.POSS=breakfast-DEF

gʷä

fir-u

to

go-PFV [3M.SG]

xʷ-ink’ä

kɨndɨŋ

ŋɨn-i-z

eat-NEG

study

house-DEF-GEN

‘He went to school without (even) eating his breakfast.’ b. albb-ɨd cloth\PL-DEF

gʷäyy-ink’ä

aqa-s-ɨ-č

sit-NEG

wash-CAUS-PFV-3F.SG

‘She washed the cloth standing (lit. ‘not sitting’).’ The two components of a complex predicate, on the contrary, are under the same scope of negation, which is necessarily marked on the main verb: (5) ŋɨn-i-s

fɨ-ø

fir-ɨ-y-äw=ɨm

house-DEF-ABL go_out-3M.SG[CNV] go-PFV-NEG-3M.SG=PFC

‘He did not leave the house.’ (*‘He went but did not go out from the house.’) 3.2 BAPs as complex predicates: Morphosyntactic and semantic criteria The properties of V1–V2 constructions surveyed in 3.1 shall serve as a starting point for our analysis of give-periphrases as complex predicates. The possibility to insert an argument (the beneficiary) between the lexical verb and the verb operator of BAPs has already been underlined in Section 2, where (2a) and (2b) were in free variation, though it is worth mentioning that I found more occurrences of the SXOV1–V2 pattern in my corpus. Our second criterion, however, clearly distinguishes BAPs from sequential constructions. The comparison of (6a) with (6b) shows that only the final verb may be formally negated, whether the scope of negation is limited to the beneficiary participant or not, as both interpretations are available. (6)

a. ŋɨ=xʷɨrä-sɨ 3M.SG.POSS=child-DAT

ŋɨn

täss-ø

yɨw-u

house

build-3M.SG[CNV]

give-PFV[3M.SG]

‘He built a house for his son (because he did not have enough money to do it himself).’ (lit. ‘He built a house and gave (it) to his son.’) b. ŋɨ=xʷɨrä-sɨ

ŋɨn

3M.SG.POSS=child-DAT house

täss-ø

yɨw-ɨ-y-äw=ɨm

build-3M.SG[CNV]

give-PFV-NEG-3M.SG=PFC

‘He did not build a house for his son.’ (either ‘He did not build any house.’ or ‘He did build a house, but not for his son.’)

142

Chloé Darmon

In order to introduce an alternative beneficiary, the lexical verb needs to be repeated along with the valency operator: (7)

ŋɨ=xʷɨrä-sɨ

ŋɨn

täss-ø

yɨw-ɨ-y-äw=ɨm

3M.SG.POSS=child-DAT

house

build-3M.SG[CNV]

give-PFV-NEG-3M.SG=PFC

gɨn

ŋ=ɨwnä-s

täss-ø

yɨw-u

but

3M.SG.POSS=woman-DAT

build-3M.SG[CNV]

give-PFV[3M.SG]

‘He did not build a house for his son, but he did build one for his wife.’ Apart from these morphosyntactic considerations, all of the instances of BAPs seen so far are ambiguous when extracted from their enunciative context, allowing for the benefactive reading as well as the sequential one provided in the literal translation. However this is not always the case. Examples (8a–c) unmistakably establish the valency operator function of ‘give’ in BAPs. They additionally demonstrate that the V1–V2 pattern of applicative constructions must be analyzed as complex predication, that is “monoclausal structures involving two or more predicating morphemes” (Baker & Harvey 2010:13). (8)

a. mi-dyän-t injera-DEF-ACC

abɨz-ø

dawit-ɨs

yɨw-u-n

finish-1SG[CNV]

Dawit-DAT

give-PFV-1SG

‘I finished the injera for the sake of Dawit (otherwise he would have been punished).’ b. bir-ɨd door-DEF

bɨz-ɨr

šɨmir-yän-s

yɨw-ɨ-č

open-3F.SG[CNV]

old_woman-DEF-DAT

give-PFV-3F.SG

‘She helped the old woman to open the door (because she did not manage to).’ c. yɨna=ak’ɨl-dyän-t 1PL.POSS=relatives-DEF-ACC

käb-n-ansɨ

sab

säraš-ɨn

help-1PL-PURP

work

work-1PL[CNV]

yɨw-ɨn-u-n give-1PL-PFV-1PL

‘In order to help our relatives, we work for/with them.’ One piece of evidence, however, suggests that ‘give’ is not fully grammaticalized as a valency operator: it cannot combine with ‘give’ in V1.

Benefactive Applicative Periphrases with yɨw- ‘give’ in Xamtanga 143 (9) *guläšu šäggä ɨfärä-y-sɨ Guläšu

good

abäbä yɨw-ø

child-DEF-DAT flower

yɨw-u

give-3M.SG[CNV] give-PFV[3M.SG]

Intended: ‘Guläšu gave flowers to/for/on behalf of the beautiful girl.’ Note that when the beneficiary licensed by the periphrasis is the speaker, the verb operator näy-, translated ‘give [to the speaker], give here’ by Appleyard (2006), is employed instead of yɨw-. ‘Give’ verbs with a deictic component of this type are attested cross-linguistically (Creissels 2010:50). (10) ŋɨn

täss-ø

näy-u

house build-3M.SG[CNV] give_to_the_speaker-PFV[3M.SG]

‘He built a house for me.’ 4 The semantic role of the dative NP 4.1 Semantic roles encoded by the dative case In this section I examine the semantic role assigned by yɨw- to its dativemarked participant in BAPs. In order to do so, I shall begin by introducing the three main semantic subtypes of beneficiaries recognized by Van Valin & LaPolla (1997:383 f.): recipient, deputative, and plain. A recipient beneficiary receives a concrete entity (‘bake a cake for someone’); a deputative beneficiary profits from a situation because the agent performs the action in place of him, so that he does not have to do it himself (‘stand in line for someone’); and a plain beneficiary benefits from a state of affairs that the agent created to make him happy, for his enjoyment (‘sing for someone’). It is important to emphasize that the semantic role of the beneficiary is usually highly contextual. The action of ‘baking a cake’, for instance, can be performed with the aim of giving the cake to someone (recipient), instead of someone (deputative), or to amuse someone (plain). Like the English postposition ‘for’, the dative case of Xamtanga supposedly leaves room for several interpretations, namely recipient-like (11a) or plain ones (11b), the deputative reading being judged ungrammatical in (11c). (11) a. abbäbä bɨrtukʷan-sɨ Abbäbä

dɨbdabe s’af-u

Bɨrtukʷan-DAT letter

write-PFV[3M.SG]

‘Abbäbä wrote a letter to Bɨrtukʷan.’ b. y=ɨnyä

yɨggɨs

tär-ɨ-č

1SG.POSS=mother 1SG.DAT come-PFV-3F.SG

‘My mother came for me.’

144

Chloé Darmon c. *yɨggɨs 1SG.DAT

aryä

fir-u

market

go-PFV[3M.SG]

Intended: ‘He went to the market instead of me.’ The choice to employ a linguistically more marked structure, an applicative periphrasis, is thus primarily motivated by the speaker’s need to specify a semantic subtype of beneficiary. Although the verb operator yɨw- ‘give’ may license recipient-like participants (6a), we already observed that several occurrences of BAPs cannot be accounted for in this way. In fact, comparison between the monoverbal sentence (11a) ‘Abbäbä wrote a letter to Bɨrtukʷan’ and the biverbal construction (2) ‘Abbäbä wrote the letter for Bɨrtukʷan (in her favor, because she cannot write)’ reveals that one of the purposes of BAPs is to exclude a recipient/addressee reading. BAPs also encode deputative beneficiaries such as (8a), (8b) and (8c). Hence there is no one-to-one correspondence between the semantic roles of Van Valin & LaPolla’s (1997) typology and the dative arguments licensed by give-periphrases in our case. 4.2 BAPs and the notion of engagement The notion of engagement put forward by Song (2010) to account for Korean data offers a more appropriate starting point for our analysis of giveperiphrases. The author provides the following definition of situations of engagement (Song 2010:408 f.): “What the agent (i.e. X) […] does it to create a situation in which the goal (i.e. Y) engages with the theme (i.e. Z) to the goal’s benefit, whether Y consumes Z (e.g. eating a pie), watches Z (e.g. watching a DVD) or makes use of Z (e.g. walking through an opened door).” He adds that Y has to create a “humanly relevant” contact with Z, that is to engage with it in an expected and plausible manner. This last requirement rules out utterances such as ‘X closed the door for Y’, since it is difficult to imagine the kind of daily-life activity Y could experience by connecting with a closed door. To this respect, Xamtanga differs from Korean: in (8a) ‘I finished the injera for the sake of Dawit (otherwise he would have been punished)’, it is obviously not possible for Dawit to engage with an already eaten injera in practical terms. The verb operator yɨw- therefore appears to be employed in a broader range of contexts than what Song describes. For this reason, I suggest adopting a wide definition of ‘engagement’ that encompasses situations where the beneficiary engages either directly with the theme object, or with the overall state of affairs resulting from the agent’s

Benefactive Applicative Periphrases with yɨw- ‘give’ in Xamtanga 145 action. The participant licensed by a give-periphrasis additionally needs to take part in the activity encoded by the lexical verb, the agent being perceived has someone who assists the beneficiary. Contrary to (11a) ‘Abbäbä wrote a letter to Bɨrtukʷan’, which does not tell us anything about the interest of Bɨrtukʷan towards the letter, the applicative construction (2) ‘Abbäbä wrote a letter for Bɨrtukʷan’ thus indicates both that Abbäbä writes the letter so that Bɨrtukʷan can do something with it (e.g. send it to someone else), and that Bɨrtukʷan herself is involved in the writing process (e.g. by dictating the text). The same explanation is also accurate for other instances of giveconstructions such as (8b) ‘open a door for someone’, as opposed to ‘open a door to someone’ below. In (12) the goal is likely to engage with the theme as well (i.e. walk through the opened door), but the structure itself neither implies it, nor conveys the idea that ‘the guest’ took part in the event. (12) abɨn-i-z

bir-ɨd

bɨz-ɨ-č

guest-DEF-DAT door-DEF open-PFV-3F.SG

‘She opened the door to the guest.’ Cases where the agent’s action results in the ownership of an entity (‘house’, ‘letter’) are clearly connected to the initial lexical meaning of yɨw-. Elsewhere, as in (8a) and (8b), the direct object of the complex predicate cannot be understood as the patient of the lexical verb ‘give’, demonstrating that the grammaticalization of yɨw- into a verb operator is well advanced. 5 Conclusion We have seen that Xamtanga BAPs involve a lexical verb in a converbal form in V1 and the verb operator yɨw- in V2 that assigns the semantic role of engager beneficiary to its dative argument. They thus belong to the markedlexical verb syntactic type of BAPs, where only the lexical verb is nonautonomous. This is in contrast with the marked verb operator (in which only the verb operator is in a non-autonomous form) and the serializing (in which the lexical verb and the verb operator are identically marked, none of them is in a non-finite form and there is no linking element between them) types identified by Creissels (2010). Despite the fact that converbs are widely attested in verb-final languages of north-east Africa, Creissels (2010) observes that he found only one instance of such constructions with ‘give’ in this area (in the Saharan language Beria). When it comes to Afroasiatic languages spoken in Ethiopia and Eritrea, give-benefactive periphrases in general seem to be rare, or at least rarely described, as none of the descriptions I was able

146

Chloé Darmon

to check mentioned them. This study shall thus contribute to drawing attention to the existence of BAPs in Ethiopian languages, and perhaps specifically in other members of the Agaw languages cluster. Abbreviations ABL Ablative ACC Accusative BAP Benefactive applicative periphrases CAUS Causative CNV Converb DAT Dative DEF Definite article ENP Enunciative particle F Feminine GEN Genitive IPFV Imperfective M Masculine

PFC MEDP NEG PFV PL POSS PURP SG V1 V2

Polyfunctional clitic =(ɨ)m Medio-passive Negation Perfective Plural Possessive Purpose Singular First verb (in complex predicates) Second verb (in complex predicates)

References Appleyard, David. 1984. The morphology of the negative verb in Agaw. Transactions of the Philological Society 82(1). 202–219. Appleyard, David. 2006. A Comparative Dictionary of the Agaw Languages. Cologne: Köppe. Azeb, Amha & Gerrit Dimmendaal. 2006. Converbs from an African perspective. In Felix K. Ameka, Alan Dench & Nicholas Evans (eds.), Catching Language: The Standing Challenge of Grammar Writing, 393–440. Berlin: Mouton de Gruyter. Baker, Brett & Mark Harvey. 2010. Complex predicate formation. In Amberber Mengistu, Brett Baker & Mark Harvey (eds.), Complex Predicates: Cross-Linguistic Perspectives on Event Structure, 13–47. Cambridge & New York: Cambridge University Press. Creissels, Denis. 2010. Benefactive applicative periphrases: A typological approach. In Fernando Zúñiga & Seppo Kittilä (eds.), Benefactives and Malefactives: Typological Perspectives and Case Studies, 29– 70. Amsterdam & Philadelphia: John Benjamins.

Benefactive Applicative Periphrases with yɨw- ‘give’ in Xamtanga 147 Peterson, David A. 2007. Applicative Constructions. Oxford: Oxford University Press. Song, Jae Jung. 2010. Korean benefactive particles and their meanings. In Fernando Zúñiga & Seppo Kittilä (eds.), Benefactives and Malefactives: Typological Perspectives and Case Studies, 393–418. Amsterdam & Philadelphia: John Benjamins. Valin, Robert D. Van & Randy J. LaPolla. 1997. Syntax: Structure, Meaning and Function. Cambridge & New York: Cambridge University Press.

Multiple Exponence in the Long Prefix Conjugation of the Transversal South Ethio-Semitic Languages Maria Bulakh Russian State University for the Humanities Abstract The phenomenon of one-to-many correspondence between a grammatical meaning and its morphological realization within one word, referred to as multiple exponence, is well-known cross-linguistically. This article deals with this phenomenon in the Transversal South Ethio-Semitic (TSES) languages and focuses on the evolution of the analytic constructions involving the long prefix conjugation (LPC) form of the main verb and the auxiliary verbs *hallawa and *nabara. The original analytic constructions had double marking of person, once on the main verb and once on the auxiliary verb: *yəsabr-u hallaw-u, with the 3PL agreement markers yə-...-u on the main verb and -u on the auxiliary verb. The common trend is to eliminate one of the redundant morphemes, either throughout the paradigm or in certain slots only. The exhaustive survey of the TSES paradigms going back to combinations of the LPC with *hallawa and *nabara enables examination of various patterns of dealing with multiple exponence and various degrees of toleration of multiple exponence. The general tendency to avoid repetition of formally similar morphemes and to tolerate double marking by way of disparate exponents can be observed.1 1 Introduction The languages discussed in this article – Amharic, Argobba, Harari and the three East Gurage languages Zay, Wolane, and Silt’e – are Transversal South Ethio-Semitic (TSES) languages. This term was introduced by Hetzron (1972:6 f.) as a name of a genealogical sub-branch within South Ethio-Semi-

1

I am most grateful to the Russian Foundation for Humanities (RFH/РГНФ) for supporting this investigation (grant #12-04-00092a). Warm thanks go to the anonymous reviewers of the draft version of the paper for their careful reading, corrections and ini htf l t N dl t ll th i

150

Maria Bulakh

tic (SES) which, in turn, is a genealogical unit within Ethio-Semitic (ES) (cf. also fn. 4). The present article focuses on the paradigm of the long prefix conjugation (LPC), which is called imperfect in traditional grammars of Ethio-Semitic languages, and imperfective in modern descriptive works. The term long prefix conjugation (opposed to short prefix conjugation = traditional jussive, and to suffix conjugation = traditional perfect) is preferred here as unambiguous reference to the form rather than to the meaning.2 The paradigm of the simple LPC is marked by a special verb base and a set of pre- and circumfixes to mark person, gender and number:3 (1) a. yə-säbər

AMHARIC

3SG.M-break\IPFV

‘he breaks’ b. tə-säbr-i

c. yə-säbr-u

2SG.F-break\IPFV-CIRC

3PL-break\IPFV-CIRC

‘you (SG.F) break’

‘they break’

In all TSES languages, the LPC has two subtypes, traditionally called simple (e.g. Amharic yəsäbər ‘he breaks’) and compound (e.g. Amharic yəsäbrall ‘he breaks’).4 The semantic content of the compound LPC in modern TSES languages can be tentatively described as non-past. Usually, the aspectual meaning of the imperfective is also attributed to the compound (as well as simple) LPC in modern descriptive works (cf. Meyer 2005:183 ff., 2006:117 for Zay and 2 3 4

Note that in the language examples, however, the long prefix conjugation is glossed as IPFV for the sake of consistency within the volume. Throughout the article, the terms prefix and suffix will be applied to both independent morphemes and parts of circumfixes. In the glosses, the meaning of a circumfix is indicated on the prefix, whereas the suffix is marked as CIRC. The development of the compound LPC, which has replaced the simple LPC in the function of the basic exponent of non-past tense events, is one of the most significant arguments for positing the genealogical unity of TSES (Hetzron 1972:38 ff.). However, this feature is apparently a relatively recent phenomenon that developed independently in individual languages or groups of languages. Notably, it is absent in Ancient Harari, and thus, its presence in Modern Harari is to be attributed to areal diffusion (Wagner 2011:1259). Other common innovations postulated by Hetzron for TSES can be explained by a period of areal affinity rather than by inheritance from a common ancestor language. This is why in the present article the term TSES will be applied as a cover label for Amharic, Argobba, Harari and East Gurage without iml i th t th tit t l i l it

Multiple Exponence in the Long Prefix Conjugation

151

Wolane; Wetter 2010:151 ff. for Argobba of Ṭollaha). Indeed, a verb in the form of the compound LPC usually refers to an unbounded situation in the present. However, if the verb refers to a situation in the future, it may well designate a bounded action: (2) a. zuhur əsseggädäll ‘one says the midday prayer’ (present time reference, unbounded)

ARGOBBA OF ṬOLLAHA

b. lägan kämise ənnahedäšənna bäyya ‘Tell her: tomorrow we shall take you to Kämise’ (future time reference, bounded) (Wetter 2010:153) The simple LPC forms occur either in subordinate clauses or in analytic constructions involving various auxiliaries, of which the combination with the past-tense auxiliary *nabara is the most important and most integrated into the tense-aspect system of TSES. In the modern TSES languages, this construction is the basic exponent of past imperfective. It is opposed to the suffix conjugation (SC) paradigm, which is the exponent of the perfective. The compound LPC goes back to an analytic construction “simple LPC + non-past tense auxiliary *hallawa” (cf. Hetzron 1972:38 ff. and Voigt 1977:336 ff. for a detailed overview of the compound LPC paradigms and comparable analytic constructions in Gəʿəz, Tigre and Tigrinya). The use of the verb *hallawa as an auxiliary is restricted to the non-past tense. Its past tense counterpart is the auxiliary *nabara. The construction “simple LPC + auxiliary *hallawa” was thus once structurally parallel to “simple LPC + auxiliary *nabara”. In the course of time, it has achieved a higher degree of grammaticalization, which includes both the degree of morphological fusion between the main verb and the auxiliary, and the semantic generalization. Originally this construction expressed progressive or durative aspectual nuances in the present tense and was opposed to a more neutral simple LPC form. Eventually it ousted the simple LPC form in the main clause and took over its functions and semantics. Still, some parallels can be traced in the development of both *hallawaand *nabara-constructions in TSES. One of their common features is the phenomenon of multiple exponence and the tendency to avoid it, which is revealed in the evolution of both constructions in most of the TSES languages. It is this phenomenon that is in the focus of the present investigation. The language specific reflexes of both *nabara and *hallawa are referred to as auxiliary elements (and, accordingly, glossed as AUX) – a term that, at

152

Maria Bulakh

least when applied to *hallawa, is used in a diachronic sense. Similarly, the distinction between the subject agreement markers of main verbs and those of auxiliary elements is to be understood in a diachronic sense. The two constructions, at any stage of their evolution, will be occasionally referred to as *hallawa-construction and *nabara-construction. 2 Multiple exponence The phenomenon of one-to-many correspondence between a grammatical meaning and its morphological realization within one word or one syntactic construction, referred to as multiple exponence or extended exponence (cf. Matthews 1974:149 f.), has enjoyed a great deal of attention from the adherents of various linguistic theories in the last decades (cf., e.g. Caballero & Harris 2012:163 f. with further references). It has been observed (e.g. by Anderson 2001) that multiple exponence is frequently a result of grammaticalization of analytic (periphrastic) constructions, and that in the course of time it tends to be eliminated, in full accordance with the principle of economy (cf., e.g. Kiparsky 2005; Kibrik 1992:188; for an English summary cf. Testelets 2008:318). In earlier works, the term multiple exponence was usually applied to synthetic forms. The more recent trend is to make this term applicable to the domain of syntax as well (as in Caballero & Harris 2012:166). The theoretical aspects of this expansion – in particular, a precise definition that would draw a borderline between multiple exponence and morphological agreement – remain to be elaborated. Still, this latter approach is fruitful for the present investigation because it legitimizes the inclusion of *nabaraconstructions among the forms to be analyzed. These collocations, usually treated as analytic forms in the modern TSES languages, exhibit the same tendency to avoid multiple exponence which is observed in the more grammaticalized *hallawa-constructions. Of course, the deletion of redundant morphemes in an analytic construction with *nabara should be regarded as the first stage of its grammaticalization into a synthetic form. Multiple exponence is widespread in the languages of the world. However, there is at least one subtype of multiple exponence that occurs only rarely. This subtype, labeled exuberant exponence (cf., e.g. Harris 2008), is double marking of the same grammatical meaning by means of two phonetically identical morphemes within one word. The data from the TSES languages confirm the existence of special constraints on exuberant exponence: in some of the paradigms considered in the present investigation, only exuberant exponence is eliminated, whereas other cases of multiple exponence,

Multiple Exponence in the Long Prefix Conjugation

153

in which formally different morphemes with the same grammatical meaning are employed, remain intact (cf. Section 6.3). 3 Multiple exponence in *hallawa- and *nabara-constructions Obviously, the original analytic *hallawa- and *nabara-constructions had double marking of subject agreement: once on the main verb and once on the auxiliary verb. Thus, the form *tə-sabr-i hallo-ki, a presumable predecessor of the Amharic form in (3), contains the 2SG.F AGRS marker *tə-...-i on the main verb and in the same function *-ki on the auxiliary verb. In this particular case, the double marking is preserved in modern Amharic: (3) tə-säbr-iy=allä-š

AMHARIC

2SG.F-break\IPFV-CIRC=AUX.NPST-2SG.F

‘you (SG.F) break’ The TSES languages exhibit a strong tendency to get rid of multiple exponence in the *hallawa- and *nabara-constructions. The evolution of the analytic constructions with *hallawa and *nabara is frequently accompanied by the loss of some of the AGRS morphemes, either on the auxiliary or on the main verb. However, a lot of variation can be observed as far as the realization of this general tendency is concerned. Even within one language, multiple exponence may be dealt with in different ways in each of the two constructions or even in different slots of one paradigm. Thus, in the Amharic compound LPC, the suffixal element of the AGRS marker of the main verb is deleted in the 3PL (4) but is preserved in the 2SG.F, shown in (3) above (cf. Leslau 1995:341; for a detailed discussion cf. Diertani & Eilam 2010). (4) yə-säbr=all-u

AMHARIC

3PL-break\IPFV=AUX.NPST-3PL

‘they break’ The 3PL forms with double marking are sometimes found in modern texts as a dialectal feature (cf. Goldenberg 1977:494). It is the purpose of the present investigation to survey the paradigms of *hallawa- and *nabara-constructions in all TSES languages, to analyze various strategies of eliminating multiple exponence and to identify the factors that may favor the deletion or preservation of the redundant morphemes. Of course, multiple exponence is not confined to these two constructions in TSES. Other verbal paradigms exhibiting the same phenomenon were not

154

Maria Bulakh

included in the present investigation. Most of them are not omnipresent in TSES and/or are still non-grammaticalized periphrastic constructions. The compound gerund in Amharic and Argobba is the most attractive object of analysis; a future investigation of multiple exponence in this domain will undoubtedly yield interesting results. A cursory glance at these paradigms suggests that their development follows the same principles as those detected in the course of the present investigation. 4 Multiple exponence and syncretism Syncretism, i.e. lack of distinction between certain forms within a paradigm (cf. Baerman, Brown & Corbett 2005), can be viewed as a phenomenon opposite to multiple exponence. Whereas multiple exponence involves morphological superfluity, syncretism appears as a shortage of morphological means. If multiple exponence emerges as a result of morphological fusion between two paradigms (as in the case of the constructions under scrutiny, between paradigms of the main verb and its auxiliary) and one of these paradigms is syncretic, one could expect multiple exponence to be preserved as a means of distinguishing the originally syncretic forms. Indeed, in TSES, when an auxiliary is grammaticalized and merges with a preceding main verb, the AGRS morphemes on the auxiliary are not always redundant. In certain cases, these additional morphemes can make the paradigm more distinctive: they can serve to disambiguate some slots of the paradigm that had been syncretic. Consider the Amharic 2SG.M and 3SG.F forms in (5), homophonous in the simple LPC but distinctive in the compound LPC due to the auxiliary AGRS suffixes. (5)

Simple LPC a. tə-säbr

AMHARIC

2SG.M;3SG.F-break\IPFV

i. ‘you (SG.M) break’ ii. ‘she breaks’ Compound LPC b. tə-säbr=allä-h 2SG.M;3SG.F-break\IPFV=AUX.NPST-2SG.M

‘you (SG.M) break’

vs.

c. tə-säbr=allä-äčč 2SG.M;3SG.F-break\IPFV=AUX.NPST-3SG.F

‘she breaks’

Multiple Exponence in the Long Prefix Conjugation

155

The omission of the AGRS suffixes -h and -äčč would lead to a loss of grammatical distinctiveness. Interestingly, the data from TSES testify that such cases of newly introduced disambiguation are still to be considered as instances of multiple exponence: the AGRS suffixes in such positions are eliminated as easily as in the rest of the paradigm. This is exemplified by the Amharic *nabara-construction, where the AGRS suffixes of the auxiliary are consistently dropped: (6) tə-säbr 2SG.M;3SG.F-break\IPFV

näbbär

AMHARIC

AUX.PST

i. ‘you (SG.M) were breaking’ ii. ‘she was breaking’ Compare the paradigm with preservation of multiple exponence (see also Section 6.1): (7)

tə-säbr

näbbär-k

2SG.M;3SG.F-break\IPFV

AUX.PST-2SG.M

AMHARIC

‘you (SG.M) were breaking’ vs. tə-säbr

näbbär-š

2SG.M;3SG.F-break\IPFV

AUX.PST-3SG.F

‘she was breaking’ In order to account for this phenomenon one has to remember that syncretism between 2SG.M and 3SG.F is an archaic feature of the prefix conjugation in Semitic, and that it is easily tolerated in the daughter languages. Notably, it is omnipresent in the simple LPC of TSES.5 Even if, in the course of evolution, disambiguation is by chance achieved, as in the TSES *hallawa- and *nabara-constructions, the information on person-gender-number of the subject is rather perceived as superfluous and can be easily omitted again. On the contrary, if any of the old paradigmatic oppositions is endangered, the additional AGRS morphemes on auxiliaries are readily employed as 5

The question why the homophony between these two verbal forms was allowed in the protolanguage lies beyond the scope of the present paper. In their fundamental work, Baerman, Brown & Corbett (2005:104 f.), who classify this type of syncretism as partial polarity, make no claims as to whether it is ever preferred to other types of syncretism. In fact, they admit that polarity effects in various languages of the world may be l id t l

156

Maria Bulakh

means to reinforce the archaic ones. Thus, in some of the TSES languages the prefixal subject markers *ʔə- (1SG) and *yə- (3SG.M;3PL) are no more distinguished in *hallawa- and *nabara-constructions: in East Gurage they merge into *yə-, in Argobba of Ṭollaha they are dropped altogether and thus shift to a single zero-marker opposed to overt prefixal elements in the other slots of the paradigm. Both in East Gurage and in Argobba of Ṭollaha this merger leads to block syncretism (Stump 2001:217 f.) between 1SG and 3SG.M prefixes but does not result in the coincidence of the 1SG and 3SG.M forms. The AGRS markers on the auxiliary are effectively used to avoid the obviously undesirable whole-word syncretism:6

(8)

Compound LPC a. 1SG yə-näḳəl=ā-hʷ

ZAY

1SG;3SG.M-take\IPFV=AUX.NPST-1SG

‘I take’ b.

3SG.M

yə-näḳəl=äl-ø 1SG;3SG.M-take\IPFV =AUX.NPST-3SG.M

‘he takes’ Simple LPC + *nabara c. 1SG yə-näḳəl 1SG;3SG.M-take\IPFV

(Meyer 2005:159) nār-ux AUX.PST-1SG

‘I used to take/I was taking’ d.

3SG.M

yə-näḳəl

nār-ø

1SG;3SG.M-take\IPFV

AUX.PST-3SG.M

‘he used to take/he was taking’ (9)

Compound LPC a. 1SG i-nakt=ā-hu

(Meyer 2005:160) SILT’E

1SG;3SG.M-beat\IPFV=AUX.NPST-1SG

‘I beat’ b.

3SG.M

i-nakt=ān-ø 1SG;3SG.M-beat\IPFV=AUX.NPST-3SG.M

‘he beats’

6

(Gutt 1997:922)

For convenience’s sake, the term whole-word syncretism, i.e. homophony of wordforms, as opposed to block syncretism, i.e. syncretism of morphemes (cf. Stump 2001:217 f.), will also be applied to *hallawa- and *nabara-constructions (albeit the i t t f th l tt ld b h l i i )

Multiple Exponence in the Long Prefix Conjugation Simple LPC + *nabara c. 1SG i-ḳaba 1SG;3SG.M-paint\IPFV

157

nār-ku AUX.PST-1SG

‘I was painting’ d.

3SG.M

i-ḳaba

nār-ø

1SG;3SG.M-paint\IPFV

AUX.PST-3SG.M

‘he was painting’

(Gutt 1997:922) ARGOBBA OF ṬOLLAHA

Compound LPC (10) a. 1SG ø-säbr=əll-äw 1SG;3SG.M-break\IPFV=AUX.NPST-1SG

‘I break’ b.

3SG.M

ø-säbr=äll-ø 1SG;3SG.M-break\IPFV=AUX.NPST-3SG.M

‘he breaks’ Simple LPC + *nabara c. 1SG ø-zeyyər 1SG;3SG.M-greet\IPFV

(Wetter 2010:400) əmbär-ew AUX.PST-1SG

‘I used to greet’ d.

3SG.M

ø-zeyyər

əmbär-ø

1SG;3SG.M-greet\IPFV

AUX.PST-3SG.M

‘he used to greet’

(Wetter 2010:403)

In Zay (8) and Argobba (10) as well as in the *hallawa-construction of Silt’e (9a–b), the 1SG and 3SG.M slots are not the only positions where the AGRS markers of the auxiliary are preserved. Still, a highly illustrative example of the direct connection between the distinctiveness of AGRS markers of the main verb and the auxiliary is provided by Silt’e, where the AGRS markers of the auxiliary *nabara are dropped throughout the paradigm except for 1SG (cf. (9c) and Section 6.1). However, a paradigm in which the auxiliary is only marked for subject agreement in one single person (1SG) is not expected to be stable. Indeed, in Wolane, the closest relative of Silt’e, the AGRS markers of *nabara are dropped even in the 1SG, so the *nabara-construction, unlike the *hallawa-construction, is characterized by whole-word syncretism between 1SG and 3SG.M:

158

Maria Bulakh

Compound LPC (11) a. 1SG

WOLANE

yə-säbr=ā-ʷh7 1SG;3SG.M-break\IPFV=AUX.NPST-1SG

‘I break’ b.

3SG.M

yə-säbr=ān-ø 1SG;3SG.M-break\IPFV=AUX.NPST-3SG.M

‘he breaks’ Simple LPC + *nabara c. 1SG/3SG.M yə-näksə 1SG;3SG.M-bite\IPFV

(Meyer 2006:97) när AUX.PST

i. ‘I used to bite’ ii. ‘he used to bite’ Obviously, the tendency to preserve the morphological distinction was outweighed in Wolane by the urge towards leveling the paradigm. It should be observed that the merger between the prefixal elements *ʔəand *yə- in itself is likely a direct result of the integration of the auxiliaries with their AGRS morphemes into the verbal paradigms. The tendency of merging these two morphemes, due to their phonological similarity, is observed in many languages throughout Semitic, and various morphological replacements can be employed to avoid it and thus to preserve the paradigmatic distinction (cf. Rubin 2007). In the *hallawa- and *nabara-constructions of East Gurage, as soon as the distinction was secured by the AGRS markers of the auxiliaries, the natural tendency to merge *ʔə- and *yə- was no longer constrained. Thus, the two synchronically very similar occurrences of block syncretism – yə- for 1SG/3SG.M and tə- for 2SG.M/3SG.F – exhibit striking dissimilarity in the pattern of interaction with the AGRS suffix of the auxiliary. This dissimilarity can only be explained in diachronic terms. The archaic 2SG.M/3SG.F whole-word syncretism, an inherent feature of the Semitic prefix conjugation, is easily tolerated, thus newly introduced disambiguation is not welcome and can be abandoned by means of deletion of the redundant AGRS morphemes. On the contrary, the 1SG/3SG.M whole-word syncretism is usually avoided: the innovative block syncretism of the 1SG/3SG.M prefixes is counterbalanced by the innovative distinctive AGRS morphemes of the auxiliaries, which are not perceived as redundant, and hence, are not deleted.8 7 8

This verb also occurs in the variant yə-säbr=ā-w. I Z lt ti d LPC i f d ith th

ffi

ā (M

2005 148)

Multiple Exponence in the Long Prefix Conjugation

159

For a similar case of preservation of multiple exponence in order to avoid innovative whole-word syncretism, cf. Section 6.2. 5 Preservation of multiple exponence throughout the paradigm Among the *hallawa- and *nabara-constructions in the TSES languages, consistent preservation of multiple exponence is found only in Argobba of Ṭollaha. In this dialect, multiple exponence is preserved in both constructions.9 (12) Compound LPC a. ø-säbr=əll-äw

ARGOBBA OF ṬOLLAHA

1SG-break\IPFV=AUX.NPST-1SG

‘I break’ b.

ə-säbr=əll-əx 2SG.M-break\IPFV=AUX.NPST-2SG.M

‘you (SG.M) break’ c.

ə-säbr-u=ll-uxum 2PL-break\IPFV-CIRC=AUX.NPST-2PL

‘you (PL) break’ d.

ø-säbr-u=ll-äy 3PL-break\IPFV-CIRC=AUX.NPST-3PL

‘they break’ Simple LPC + *nabara e. ø-zeyyər 1SG-greet\IPFV

(Wetter 2010:400) əmbär-ew AUX.PST-1SG

‘I used to greet’ f.

ə-zeyyər

əmbär-ex

2SG.M-greet\IPFV

AUX.PST-2SG.M

‘you (SG.M) used to greet’

9

which is likely to be a further simplification of *hallawa. In this paradigm, the auxiliary has no personal markers of its own. Remarkably, it has a lacuna in the 1SG form – a fact that can be easily accounted for by the lack of means to sustain the distinction between 1SG and 3SG.M. In Argobba of Ṭollaha, the AGRS prefixes *ʔə- and *yə- are replaced by zero. This development cannot be explained as an elimination of redundant AGRS morphemes, since the zero morphemes are meaningful elements structurally opposed to the overt prefixes in the rest of the paradigm. The introduction of zero morphemes only leads to l k f di ti ti b t 1 d3 bj t k ( f S ti 4)

160

Maria Bulakh g.

ə-zeyyər-u

əmbär-exum

2PL-greet\IPFV-CIRC

AUX.PST-2PL

‘you (PL) used to greet’ h.

ø-zeyyər-u

əmbär-äy

3PL-greet\IPFV-CIRC

AUX.PST-3PL

‘they used to greet’

(Wetter 2010:403)

6 Elimination of multiple exponence Apart from Argobba of Ṭollaha, all TSES languages demonstrate some degree of constraint on multiple exponence which is manifested by the deletion of certain AGRS morphemes. By what means is the elimination of multiple exponence achieved? The most straightforward way is to drop one of the redundant morphemes. The considered data shows that the morphemes to be deleted are not chosen randomly. Rather, a certain hierarchy of elements subject to deletion can be established. In the first place, the choice of the morpheme to be eliminated depends on the type of the morpheme itself. In the constructions under scrutiny, the AGRS morphemes of the auxiliary are (at least diachronically, but in most cases also synchronically) suffixes, whereas the proto-forms of the AGRS morphemes of the main verb are prefixes or circumfixes. The investigation shows that the elimination of multiple exponence in TSES never involves deletion of the prefixes. In contrast, the omission of suffixes is not infrequent, either on the main verb, or on the auxiliary. I will use the term suffix multiple exponence to refer to the preservation of AGRS suffixes both on the main verb and on the auxiliary. Since ARGS prefixes are never deleted, one can establish the following implicational hierarchy: if the LPC contains suffix multiple exponence, it necessarily contains other types of multiple exponence, that is, involving only prefixes of the main verb and suffixes of the auxiliary. The reverse does not hold: in some paradigms, multiple exponence is restricted to verbal forms with AGRS prefixes on the main verb and AGRS suffixes on the auxiliary. Thus, suffix multiple exponence is less tolerated in TSES than multiple exponence involving a prefix and a suffix. For the sake of brevity, I will refer to the slots of the paradigm where the main verb can attach an AGRS suffix as positions of suffix multiple exponence, that is, positions where such multiple exponence is possible. There is a certain variation among TSES languages concerning the number of such positions. The main verb AGRS suffixes going back to Proto-ES are 2SG.F *-i

Multiple Exponence in the Long Prefix Conjugation

161

and 2/3PL *-u. In Argobba and East Gurage, the innovative 1PL suffix *-na is introduced in the LPC. Thus, in all TSES languages positions of suffix multiple exponence include 2SG.F, 2PL and 3PL, whereas in Argobba and East Gurage this list is expanded by 1PL. 6.1 Total elimination of multiple exponence by full deletion of the AGRS morpheme of the auxiliary Total elimination of multiple exponence by means of full deletion of the AGRS morpheme of the auxiliary is registered in *nabara-constructions only. In Amharic, the analytic forms “simple LPC + auxiliary näbbär” usually mark subject agreement only once, on the main verb, whereas the auxiliary remains indeclinable: (13) a. yə-sära

näbbär

AMHARIC

3SG.M-work\IPFV AUX.PST

‘he was working’ b. yə-sär-u

näbbär

3PL-work\IPFV-CIRC

AUX.PST

‘they were working’ According to Leslau (1995:316), the AGRS marking of the auxiliary is not altogether dropped but preserved as optional. The same phenomenon can be observed in a number of other analytic constructions involving the simple LPC (cf. Leslau 1995:321, 323, 326, 332, etc.).10 The *nabara-construction in Wolane (cf. Section 4) is likewise characterized by the full deletion of the AGRS markers on the auxiliary. The same strategy may be suspected for Argobba of Aliyu Amba. Unfortunately, in Leslau’s (1997:49) description only 3SG.M forms with *nabara are given, as in (14), so it remains uncertain whether the auxiliary is conjugated or not. (14) yə-hed 3SG.M-go\IPFV

‘he used to go’

əmbär

ARGOBBA OF ALIYU AMBA

AUX.PST

(Leslau 1997:49)

10 According to Baye Yimam (p.c.), the forms with multiple exponence are more usual f l A h i t t df lA h i

162

Maria Bulakh

However, the fact that the auxiliary is given in its shortened form, rather than əmbär-a, with the overt 3SG.M morpheme -a (cf. Leslau 1997:36), seems to imply the deletion (obligatory or optional, as in Amharic) of the person markers on the auxiliary. Incomplete elimination or selective preservation of multiple exponence is a means to maintain paradigmatic distinctiveness. In Silt’e, the auxiliary verb nār (< *nabara) has no AGRS marking except for the 1SG (Gutt 1997:922). The principle of this distribution is quite transparent. The 1SG marker of the auxiliary is not a redundant morpheme since it serves to maintain the opposition between 1SG and 3SG.M, as was shown in (9c) vs. (9d) above. Such a strategy is apparently very rare and one can expect that the general principles of paradigmatic leveling will sooner or later overrule the tendency towards preservation of the morphological distinction. Notably, in Wolane, the auxiliary verb of the *nabara-construction is indeclinable (Meyer 2006:118), despite the same syncretism in the prefixes of the main verbs (cf. also Section 4). 6.2 Elimination of suffix multiple exponence by deleting the AGRS suffixes of the main verb The full deletion of the suffixal elements of the main verb is recorded in Harari before the reflex of the non-past auxiliary *hallawa (on the variation in 2/3PL forms cf. Section 6.4): (15) a. ti-säbr-ā-š

HARARI

2SG.F-break\IPFV-AUX.NPST-2SG.F

‘you (SG.F) break’ b. ti-säbr-ā-ḵu 2PL-break\IPFV-AUX.NPST-2PL

‘you (PL) break’ c. yi-säbr-āl-u 3PL-break\IPFV-AUX.NPST-3PL

‘they break’

(Leslau 1958:19)

Among the positions of suffix multiple exponence in Harari, the 2SG.F deserves special attention. The suffixal element of the 2SG.F AGRS marker *tə-…-i has phonetically coincided with the epenthetic vowel -i serving to avoid a consonant cluster. Compare the two forms of the simple LPC in (16):

Multiple Exponence in the Long Prefix Conjugation (16) a. tič̣ämḳi

163 HARARI

ti-č̣ämḳ-ʲ 2SG.F-squeeze\IPFV-CIRC

‘you (SG.F) squeeze’ b. tič̣ämḳi ti-č̣ämḳ[-i] 2SG.M-squeeze\IPFV[-EPV]

‘you (SG.M) squeeze’ As a result, in the simple LPC the only trace of the 2SG.F suffix -i is palatalization, which affects final or even non-final alveopalatal consonants (cf. Leslau 1958:17 f.): (17) a. tilämǧi

HARARI

ti-lämd-ʲ 2SG.F-learn\IPFV-CIRC

‘you (SG.F) learn’ b. tilämdi ti-lämd[-i] 2SG.M-learn\IPFV[-EPV]

‘you (SG.M) learn’ At the morphemic level, one can postulate a complex 2SG.F morpheme that combines the prefix ti- and the process of palatalization -ʲ, which obtains a surface realization only if the verb contains at least one palatalizable consonant, as in (17a) vs. (16a). Absence of palatalization in the compound forms of the LPC is therefore a proof that the suffixal element - ʲ was really omitted: Simple LPC (18) a. tisägǧile ti-sägd-ʲ=le 2SG.F-prostrate\IPFV-CIRC=PURP

‘in order that you (SG.F) prostrate’ vs.

HARARI

164

Maria Bulakh Compound LPC b. tisägdāš ti-sägd=ā-š 2SG.F-prostrate\IPFV=AUX.NPST-2SG.F

‘you (SG.F) prostrate’

(Cohen 1931:278, 280)11

According to Cohen (1931:278), the suffix of the main verb is likewise omitted in the *nabara-construction of Harari:12 (19) a. tisägdi nārši ti-sägd[-i] 2SG.F-prostrate\IPFV[-EPV]

HARARI nār-š AUX.PST-2SG.F

‘you (SG.F) were prostrating’ b. tisägdi nārḫu ti-sägd[-i] 2PL-prostrate\IPFV[-EPV]

nār-ḫu AUX.PST-2PL

‘you (PL) were prostrating’ c. yisägdi nāru yi-sägd[-i] 3PL-prostrate\IPFV[-EPV]

nār-u AUX.PST-3PL

‘they were prostrating’ However, Abdurahman & Wagner (1998:21 f.) have recorded alternative forms for 2/3PL (cf. Section 6.4). The compound LPC given by Leslau (1997:50) for Argobba of Aliyu Amba is also characterized by the consistent deletion of the AGRS suffixes of the main verb and the preservation of the suffixes of the auxiliary *hallawa; for the variation in 1PL cf. Section 6.4. Incomplete elimination of suffix multiple exponence through selective preservation of suffix multiple exponence is a means to maintain paradigmatic distinctiveness. In the compound LPC of Amharic, the redundant suf11 Cohen’s transcription was standardized in accordance with the one adopted by Leslau (1958). 12 During a short stay in Harar in November 2012 I had the opportunity to check this data. Bahar Ali (a 28-year-old teacher of Harari in the Elementary School of Harar), to whom I am infinitely grateful for his cooperation and hospitality, provided me with the full paradigm of the verb lämädä ‘teach’, in which palatalization was absent both in *hallawa- and *nabara-constructions: tilämdāš ‘you (SG.F) teach’, tilämdi nārši ‘you ( ) t hi ’

Multiple Exponence in the Long Prefix Conjugation

165

fixes of the main verb rather than those of the auxiliary are dropped. Only some slots of the paradigm are affected by this process. The suffix -u of the main verb is omitted both in 2PL (the AGRS suffix on the auxiliary is -aččəhʷ) and in 3PL (the AGRS suffix on the auxiliary is -u), whereas the 2SG.F suffix -i is preserved, being doubled by the auxiliary suffix -š. Simple LPC (20) a. tə-säbr-i

AMHARIC

2SG.F-break\IPFV-CIRC

‘you (SG.F) break’ b.

tə-säbr-u 2PL-break\IPFV-CIRC

‘you (PL) break’ c.

yə-säbr-u 3PL-break\IPFV-CIRC

‘they break’ Compound LPC d. tə-säbr-iy=all-äš 2SG.F-break\IPFV-CIRC=AUX.NPST-2SG.F

‘you (SG.F) break’ e.

tə-säbr=all-aččəhʷ 2PL-break\IPFV=AUX.NPST-2PL

‘you (PL) break’ f.

yə-säbr=all-u 3PL-break\IPFV=AUX.NPST-3PL

‘they break’

(cf. Leslau 1995:301, 341)

A detailed description of multiple exponence in Amharic was presented by Diertani & Eilam (2010). In their opinion, the preservation of the 2SG.F suffix of the main verb is dictated by the weakness of distinctive force of the 2SG.F suffix of the auxiliary.13 Indeed, the 2SG.F suffix -š is phonetically close to 13 There is another possible explanation offered by Orin Gensler (p.c.): the 2SG.F suffix is less subject to elimination because of the palatalization. In verbs with a final palatalizable consonant, the 2SG.F suffix fuses with the root: təläbšəyalläš/təläbšalläš < *təläbs-i=allä-š 2SG.F-wear\IPFV-CIRC=AUX.NPST-2SG.F ‘you (SG.F) wear’. Due to the lack of a clear-cut boundary the suffix of the main verb resists deletion. However, in Harari (see above) the same conditions do not provide any obstacle to the elimination f th 2 k

166

Maria Bulakh

the 3SG.F suffix -äčč which, moreover, can shift to -äšš in the speech of some informants. In Diertani & Eilam (2010) this process of deaffricatization is defined as recent and/or dialectal. If their explanation is correct, the preservation of multiple exponence is triggered by the same factor as in East Gurage (cf. Sections 4 and 6.1): phonetic similarity of two personal markers favors preservation of the additional morphemes which help to keep the distinction between two verbal forms. The only difference is that in Amharic we deal with the (tendency towards) block syncretism of the AGRS suffixes of the auxiliary, while in East Gurage multiple exponence is preserved because of the block syncretism of the AGRS prefixes of the main verb.14 6.3 Elimination of exuberant exponence by selective deletion of the AGRS suffixes of the auxiliary The East Gurage languages have developed a common strategy of dealing with multiple exponence in the compound LPC. The AGRS suffixes are deleted on the auxiliary rather than on the main verb. Their pattern is thus different from the one adopted in Amharic compound LPC: (21) a. yə-säbr-w=ān

WOLANE

3PL-break\IPFV-CIRC=AUX.NPST

b. yə-säbr=all-u

AMHARIC

3PL-break\IPFV=AUX.NPST-3PL

‘they break’ In all East Gurage languages, the redundant morphemes are dropped in some positions of the paradigm, but preserved in others. There are four positions of suffix multiple exponence in East Gurage: 1PL, 2SG.F, 2PL, 3PL. The deletion takes place in the 1PL and 3PL only. It can be argued that in East Gurage the redundant AGRS suffix of the auxiliary (which is the regular SC AGRS marker) is preserved unless it is phonologically identical with the suffixal element of the AGRS marker of the main verb. Indeed, the AGRS suffixes of the main verb differ from the AGRS suffixes of the auxiliary in 2SG.F (*-i vs. *-aš) and 2SG.PL (*-u vs. *-kum) but are identical in 1PL (*-na) and 3PL (*-u).

14 For a different interpretation cf. Goldenberg (1977:494): the AGRS suffix -u of 2/3PL main verbs is deleted because it is phonologically identical with “the end of the auxiliary”. This rule is less convincing since it goes across the morphological borders. As the cumulative data of TSES show, the deletion of the redundant markers is usually h l i ll diti d

Multiple Exponence in the Long Prefix Conjugation

167

Thus, in Wolane, double marking is tolerated in 2PL, but the auxiliary verb lacks the AGRS marker in 1PL and 3PL: (22) a. tə-säbr-w=ā-hum

WOLANE

2PL-break\IPFV-CIRC=AUX.NPST-2PL

‘you (PL) break’ but: b. yə-säbr-ən=ān 1PL-break\IPFV-CIRC=AUX.NPST

‘we break’ c. yə-säbr-w=ān 3PL-break\IPFV-CIRC=AUX.NPST

‘they break’

(Meyer 2006:97, 108 f.)

The subtype of multiple exponence which employs identical morphemes, labeled exuberant exponence in contemporary linguistics, appears to be indeed extremely rare cross-linguistically (cf., e.g. Harris 2008). The constraint on exuberant exponence (but not on other types of multiple exponence) is thus a typologically predictable feature. Indeed, an instance of the same strategy (avoidance of exuberant exponence, but toleration of other types of multiple exponence) can be found in Amharic in another segment of verbal morphology, namely in the compound gerund (Diertani & Eilam 2010). In Zay, the constraint on exuberant exponence is observed in the *nabaraconstruction. The auxiliary verb preserves the AGRS morphemes throughout the paradigm, except for the 1PL and 3PL, i.e. the forms in which the AGRS markers of the main verb coincide with those of the auxiliary: (23) a. tineḳil nāriš tə-näḳl-ʲ 2SG.F-take\IPFV-CIRC

ZAY nār-š AUX.PST-2SG.F

‘you (SG.F) used to take/took’ – but b. yunoḳul nār yə-näḳl-ʷ 3PL-take\IPFV-CIRC

nār AUX.PST

‘they used to take/took’

(Meyer 2005:160 f.)

In Zay, the vocalic suffixes *-i and *-u came to be realized by palatalization and labialization, respectively (marked as ʲ and w), applied to all the vowels of

168

Maria Bulakh

the word. As a result, the deletion of the reflex of *-u in 3PL cannot be proved. The labialization in the auxiliary in (24) can be triggered by one or two AGRS markers. The morphophonemic structure given by Meyer (2005:160) clearly implies exuberant exponence: (24) yunoḳulol

ZAY

yə-näḳl-ʷ=äl-ʷ 3PL-break\IPFV-CIRC=AUX.NPST-3PL

‘they take’

(Meyer 2005:160)

However, the overt constraint on exuberant exponence in the *nabara-construction in Zay as well as in Silt’e and Wolane favors an alternative interpretation, according to which the compound LPC paradigm in Zay is subject to the same constraint. Indeed, the change ä > o in the auxiliary could have been triggered by the AGRS morpheme of the main verb, after the redundant marker of the auxiliary was omitted: (25) a. yunoḳulol

ZAY

yə-näḳl-ʷ=äl 3PL-take\IPFV-CIRC=AUX.NPST

‘they take’ b. tunoḳulōhum tə-näḳl-ʷ=ā-hum 2PL-take\IPFV-CIRC=AUX.NPST-2PL

‘you (PL) take’

(Meyer 2005:160)

The labialization as part of the AGRS circumfix of the main verb in (25b) affects the vowel of the auxiliary, while the absence of a second labialization morpheme is evident. There is only one TSES language which apparently tolerates exuberant exponence in one position of the paradigm while it avoids any kind of suffix multiple exponence in the rest of the paradigm. In Argobba of Aliyu Amba, a form with exuberant exponence is recorded in the 1PL: (26) ə-sädb-ən=äll-ən

ARGOBBA OF ALIYU AMBA

1PL-insult\IPFV-CIRC=AUX.NPST-1PL

‘we insult’

(Leslau 1997:50, fn. 2)

Multiple Exponence in the Long Prefix Conjugation

169

In the rest of the paradigm, the AGRS suffixes of the main verb are consistently deleted even if they do not formally coincide with the AGRS suffixes of the auxiliary: (27) tə-sädb=äll-əhum

ARGOBBA OF ALIYU AMBA

2PL-insult\IPFV=AUX.NPST-2PL

‘you (PL) insult’

(Leslau 1997:50)

This asymmetry may be due to the fact that the 1PL suffix of the main verb is younger than other AGRS suffixes. Unlike 2PL and 3PL suffixes, the 1PL suffix -ən does not go back to the Proto-SES stage. Rather, its introduction is a recent process, obviously connected with the merging of 1SG and 1PL prefixes.15 The introduction of the 1PL suffix into the system of AGRS markers of Argobba could sporadically extend into the compound LPC forms, thus ruining the original principle of constraint on suffix multiple exponence in this paradigm. Notably, the form with exuberant exponence is recorded only as a variant of more usual forms without AGRS suffix of the main verb (cf. Section 6.4). On the elimination of exuberant exponence in object-marked paradigms cf. Section 7 (and cf. also Section 7 on the relative paradigm in Harari, with potential preservation of exuberant exponence). 6.4 Variation within paradigms According to Abdurahman & Wagner (1998:21 f.), the AGRS morphemes are consistently preserved on the auxiliary of the Harari *nabara-construction except for the 2/3PL forms, where optional deletion of the suffixes takes place. Notably, these are the only forms where overt AGRS suffixes of the main verb appear (on the reflex of 2SG.F *-i in Harari cf. Section 6.2), in other words, the only positions of synchronically valid suffix multiple exponence in this language. In 2/3PL, all three possibilities of dealing with multiple exponence apparently coexist as free variants: the double marking can be preserved (28a), or the AGRS suffix can be omitted on the main verb (28b) – as in Cohen’s paradigm16 – or on the auxiliary (28c). 15 In the simple LPC, the opposition *əl-säd(ə)b/*ən-säd(ə)b ‘we insult’ vs. *ə-säd(ə)b ‘I insult’ is replaced by əl-sädb-ən ‘we insult’ vs. əl-säd(ə)b ‘I insult’. In the compound LPC, the 1SG prefix ə-, when taken over into the 1PL form, apparently causes the introduction of the same innovative suffix -ən. 16 Cf S ti 6 2 Th t t i b d i th di id d t b

170

Maria Bulakh

(28) a. yi-ṭṭaraḥ-u 3PL-be_called\IPFV-CIRC

nār-u

HARARI

AUX.PST-3PL

‘they used to be called’ b. yi-tlamad-i 3PL-be_educated\IPFV-EPV

nār-u AUX.PST-3PL

‘they used to be educated’ c. y-adgrābg-u 3PL-take\IPFV-CIRC

nāra AUX.PST

‘they used to take it here and there’ Thus, the paradigm in question varies between fully preserving multiple exponence and constraining suffix multiple exponence. In the compound LPC paradigm of Argobba of Aliyu Amba, suffix multiple exponence is avoided by means of deletion of AGRS suffixes of the main verb. The only position where alternative forms have been recorded is 1PL, where, alongside the form in (29a), the form with suffix multiple exponence is available, as in (29b): (29) a. əl-17sädb=äll-ən

ARGOBBA OF ALIYU AMBA

1PL-insult\IPFV=AUX.NPST-1PL

b. ə-sädb-ən=äll-ən 1PL-insult\IPFV-CIRC=AUX.NPST-1PL

‘we insult’

(Leslau 1997:50 with fn. 2)

The form in (29b) is all the more interesting since it involves formally identical morphemes (cf. Section 6.3). 7 Multiple exponence in the compound LPC with object pronouns and other inserted elements In Amharic and Harari, insertion of object suffixes (or AGRO morphemes, as they are sometimes called) into compound LPC forms changes the system of subject agreement markers drastically, so that one has to speak of a special subtype of the compound LPC paradigm. I will call it object-marked compound LPC,18 opposed to neutral compound LPC. Bahar Ali: tilämdi nārḫu ‘you (PL) were learning’; yilämdi nāru ‘they were learning’. 17 This prefix has the optional variant ən-. 18 This term will be applied to paradigms in which the object suffix is the only element which divides the main verb and the auxiliary element. The Harari paradigms where l l t i t d ill b t t d t l

Multiple Exponence in the Long Prefix Conjugation

171

The object-marked compound LPC paradigms also display a tendency toward elimination of multiple exponence. However, the strategy of dealing with multiple exponence implemented in an object-marked paradigm usually differs from the one of the neutral paradigm. Thus, the object-marked compound LPC paradigms in Amharic and Harari offer additional material relevant for the present investigation.19 As outlined in Section 6.2, in Amharic the suffixes of the main verb are deleted in the neutral compound LPC, with the exception of 2SG.F. In the object-marked compound LPC, all the AGRS suffixes of the main verb are preserved. As for the AGRS suffix of the auxiliary, it is preserved throughout the paradigm except for 3PL: (30) a.

tə-nägr-i=ňň=allä-š

AMHARIC

2SG.F-tell\IPFV-CIRC=OJ.1SG=AUX.NPST-2SG.F

‘you (SG.F) tell me’ b.

tə-nägr-u=ňň=all-aččǝhu 2PL-tell\IPFV-CIRC=OJ.1SG=AUX.NPST-2PL

‘you (PL) tell me’ – but c.

yə-nägr-u=ňň=all 3PL-tell\IPFV-CIRC=OJ.1SG=AUX.NPST

‘they tell me’ Thus, the object-marked compound LPC paradigm of Amharic manifests a constraint on exuberant exponence, which is avoided by means of deleting the AGRS suffixes of the auxiliary (cf. Section 6.3). The same strategy is adopted in the object-marked paradigm of Harari: (31)

a.

ti-gädl-i=ň=ā-š

HARARI

2SG.F-kill\IPFV-CIRC=OJ.1SG=AUX.NPST-2SG.F

‘you (SG.F) kill me’

19 I was unable to find full paradigm of the object-marked compound LPC for Argobba of Aliyu Amba, Zay, Silt’e and Wolane. In the East Gurage languages, however, no difference in the AGRS marking of neutral and object-marked paradigm is to be expected. These languages preserve the AGRS suffixes of the main verb in the neutral paradigms, and the same strategy is in all probability implemented in the object-marked paradigms as well. In Argobba of Ṭollaha, the differences in the AGRS marking of neutral and object-marked paradigms are irrelevant to the present investigation: both paradigms coni t tl lti l ( f W tt 2010 180 188 387 394)

172

Maria Bulakh b. ti-gädl-u=ň=ā-ḵu 2PL-kill\IPFV-CIRC=OJ.1SG=AUX.NPST-2PL

‘you (PL) kill me’ but: c.

yi-gädl-u=ň=āl 3PL-kill\IPFV-CIRC=OJ.1SG=AUX.NPST

‘they kill me’

(Leslau 1958:78)

Unlike in the neutral compound LPC, the 2SG.F AGRS marker ti-...-i in the object-marked paradigm is distinct from the 2SG.M marker ti-: (32) ti-gädl=äň=ā-ḵ

HARARI

2SG.M-kill\IPFV=OJ.1SG=AUX.NPST-2SG.M

‘you (SG.M) kill me’ Interestingly, according to Leslau’s paradigm, no palatalization occurs in the object-marked compound LPC, whereas it is recorded in the object-marked simple LPC paradigm. Compare (33) vs. (31a): (33) tigädyiň

HARARI

ti-gädl-ʲ=ň 2SG.F-kill\IPFV-CIRC=OJ.1SG

‘you (SG.F) kill me’

(Leslau 1958:77)

Conversely, a paradigm with an inserted AGRO morpheme and a question marker, offered in Abdurahman & Wagner (1998:171), suggests the absence of the overt 2SG.F AGRS suffix marker, but its preservation as a palatalization trigger: (34)

čigäǧlānaš ~ tigäǧlānaš

HARARI

ti-gädl-ʲ=ā=n=a-š 2SG.F-kill\IPFV-CIRC=OJ.3SG.M=Q=AUX.NPST-2SG.F

‘Do/will you (SG.F) kill him?’ Both in Amharic and Harari, besides the object suffixes, other elements can be inserted between the main verb and auxiliary. In some cases the insertion is accompanied by the same rearrangement of the AGRS morphemes as in the object-marked paradigm. This is apparently the case in Amharic, where the

Multiple Exponence in the Long Prefix Conjugation

173

deletion of the 3PL suffix of the auxiliary described above occurs also if a particle, rather than an object suffix, is inserted, as shown in (35a) vs. (35b): (35) a. yə-nägr-u=mm=all

AMHARIC

3PL-tell\IPFV-CIRC=and=AUX.NPST

‘and they will tell’ b. yə-nägr=all-u 3PL-tell\IPFV=AUX.NPST-3PL

‘they will tell’

(Leslau 1995:887)

In Harari, the constructions with inserted elements are more widespread than in Amharic and exhibit considerable variation as far as the elimination of multiple exponence is concerned. Some elements, like -n (question marker) and -t ‘only’, do not change the neutral AGRS marking at all: (36) a. ti-katb=īt=a-š

HARARI

2SG.F-write\IPFV=only=AUX.NPST-2SG.F

‘you (SG.F) only write’ b. ti-katb=īt=a-ḫu 2PL-write\IPFV=only=AUX.NPST-2PL

‘you (PL) only write’ c. yi-katb=īt=al-u 3PL-write\IPFV=only=AUX.NPST-3PL

‘they only write’

(Abdurahman & Wagner 1998:170, 173 f.)

Certain combinations of inserted elements follow the object-marked paradigm, where multiple exuberance is constrained. In addition to (34), consider the following examples: (37) a.

tigadlōnaḫu

HARARI

ti-gadl-u=ā=n=a-ḫu 2PL-kill\IPFV-CIRC=OJ.3SG.M=Q=AUX.NPST-2PL

‘Do you (PL) kill him?’ b.

yigadlōnal yi-gadl-u=ā=n=al 3PL-kill\IPFV-CIRC=OJ.3SG.M=Q=AUX.NPST

‘Do they kill him?’

(Abdurahman & Wagner 1998:171)

174

Maria Bulakh

Still other combinations exhibit variation in the 3PL. Thus, they vacillate between the total preservation of multiple exponence and the constraint on exuberant exponence: (38) a.

tikībbāmēš

HARARI

ti-kīb-i=b-ā=m=ē-š 2SG.F-testify\IPFV-CIRC=MAL-OJ.3SG.M=NEG=AUX.NPST.NEG-2SG.F

‘you (SG.F) do not testify against him’ b.

tikībbōmēḫu ti-kīb-u=b-ā=m=ē-ḫu 2PL-testify\IPFV-CIRC=MAL-OJ.3SG.M=NEG=AUX.NPST.NEG-2PL

‘you (PL) do not testify against him’ – but c.

yikībbōmēl(-u) yi-kīb-u=b-ā=m=ēl(-u) 3PL-testify\IPFV-CIRC=MAL-OJ.3SG.M=NEG=AUX.NPST.NEG(-3PL)

‘they do not testify against him’ (Abdurahman & Wagner 1998:184 f.) Note that the AGRS suffix of the auxiliary in (38c) is optional. A great deal of variation is also observed in the relative paradigm, marked by the inserted relative marker z-. According to Leslau (1958:19 f.), it consistently preserves multiple exponence. However, the paradigm given by Cohen (1931:281) is formed after the neutral AGRS paradigm with total elimination of multiple exponence.20 Cohen (1931:281) and Abdurahman & Wagner (1998:187) note variation in the 3PL position, where the redundant morphemes can be preserved or dropped on the main verb. If the variation is indeed restricted to the 3PL position, we deal with still another remarkable case of preservation of exuberant exponence in the absence of other types of multiple exponence (cf. 6.3). 8 Conclusions The reflexes of *hallawa- and *nabara-constructions in the TSES languages exhibit various degrees of toleration of multiple exponence, which can be arranged in the following way:

20 The same strategy is observed in the relative paradigm of the verb lämädä ‘learn’ provided to me by Bahar Ali: ʔilämdi-z-āḫ (1SG), tilämdi-z-āḫ (2SG.M), tilämdi-z-āš (2SG.F), yilämdi-z-āl (3SG.M), tilämdi-z-āt (3SG.F), nilämdi-z-āna (1PL) tilämdi-z-āḫu (2 ) ilä di āl (3 )

Multiple Exponence in the Long Prefix Conjugation

175

(1) Full toleration of multiple exponence (2) Constraint on exuberant exponence (3) Constraint on suffix multiple exponence (4) Total constraint on multiple exponence Multiple exponence, which undoubtedly was present in the source constructions, is fully preserved only in one variety, namely in Argobba of Ṭollaha. This may be considered an additional indication of the general conservatism of this variety in comparison to the other TSES languages. Total avoidance of multiple exponence is only achieved by means of deleting AGRS morphemes on the auxiliary, as in the Amharic *nabara-construction. The elimination of multiple exponence through the deletion of prefixed AGRS morphemes of the main verb is not attested in TSES. Two alternative principles that can govern the deletion of the suffixes of the main verb can be observed. One of them is the constraint on multiple exponence of suffixes, which leads to the omission of all main verb AGRS suffixes. The second principle – the constraint on exuberant exponence – leads to the deletion of only those main verb AGRS suffixes that are formally identical to the corresponding AGRS suffixes of the auxiliary. Both strategies apparently realize the same general tendency to avoid the repetition of formally similar morphemes and to tolerate double marking by way of dissimilar exponents. It remains to be stated that the hierarchy sketched above should not be understood as a successive chain of states accompanying the process of grammaticalization. Rather, any of the types (2)–(4) can develop directly from type (1). This is supported by the patterns of variation: in Harari *nabara-constructions, oscillation between (1) and (3) can be observed, whereas some Harari paradigms of the compound LPC with inserted elements offer instances of variation between (1) and (2) (cf. Sections 6.4 and 7). Moreover, as long as the constraint on the deletion of the AGRS prefixes remains valid, the systems with deleted main verb suffixes (that is, belonging to the type (3)) cannot further develop into (4). Thus, the diachronic arrangement of these four types within TSES can be represented as three alternative paths of evolution from type (1):

176

Maria Bulakh

Figure 1: Types of Multiple Exponence (1) Full toleration of multiple exponence ↓

↓

↓

(2) Constraint on exuberant exponence

(3) Constraint on suffix multiple exponence

(4) Total constraint on multiple exponence

Abbreviations * Reconstructed form (mostly proto-ES, but also forms of recent stages) ʲ Palatalization ʷ Labialization 1, 2, 3 First, second, third person AGRO Object agreement AGRS Subject agreement AUX Auxiliary verb/element CIRC Suffixal part of a circumfix EPV Epenthetic vowel ES Ethio-Semitic F Feminine IPFV Imperfective aspect

LPC M MAL NEG NPST OJ PL PST PURP Q SC SES SG TSES

Long prefix conjugation Masculine Malefactive Negation Non-past tense Object Plural Past tense Purposive Question Suffix conjugation South Ethio-Semitic Singular Transversal South EthioSemitic

References Abdurahman Garad & Ewald Wagner. 1998. Harari-Studien: Texte mit Übersetzung, grammatischen Skizzen und Glossar. Wiesbaden: Harrassowitz. Anderson, Stephen R. 2001. On some issues of morphological exponence. In Geert Booji & Jaap van Maarle (eds.), Yearbook of Morphology 2000, 1–18. New York, Boston, Dordrecht, London & Moscow: Kluwer Academic Publishers. Baerman, Mathew, Dunstan Brown & Greville G. Corbett. 2005. The SyntaxMorphology Interface: A Study of Syncretism. Cambridge: Cambridge University Press.

Multiple Exponence in the Long Prefix Conjugation

177

Caballero, Gabriela & Alice C. Harris. 2012. A working typology of Multiple Exponence. In Ferenc Kiefer, Mária Ladányi & Péter Siptár (eds.), Current Issues in Morphological Theory: (Ir)Regularity, Analogy and Frequency. Selected Papers from the 14th International Morphology Meeting, Budapest, 13–16 May 2010, 163–188. Amsterdam & Philadelphia: John Benjamins. Cohen, Marcel. 1931. Études d’éthiopien méridional. Paris: Librairie Orientaliste Paul Geuthner. Diertani, Ariel & Aviad Eilam. 2010. How Amharic Deals with Multiple Exponence. Paper presented at the 84th Annual Meeting of the Linguistic Society of America (LSA), 7–10 January 2010. Goldenberg, Gideon. 1977. The Semitic languages of Ethiopia and their classification. Bulletin of the School of Oriental and African Studies 40. 461–507. Gutt, Ernst-August. 1997. Concise grammar of Silt’e. In Eeva H. M. Gutt and Hussein Mohammed, Silt’e-Amharic-English Dictionary (with Concise Grammar by Ernst-August Gutt), 895–957. Addis Ababa: Addis Ababa University Press. Harris, Alice C. 2008. Explaining exuberant agreement. In Thórhallur Eythórsson (ed.), Grammatical Change and Linguistic Theory: The Rosendal Papers, 265–284. Amsterdam & Philadelphia: John Benjamins. Hetzron, Robert. 1972. Ethiopian Semitic: Studies in Classification. Manchester: Manchester University Press. Kibrik, Aleksandr. 1992. Podlezhashchee i problema universal’noy modeli yazyka. Ocherki po obshchim i prikladnym voprosam yazykoznaniya, 179–197. Moskva: MGU. Kiparsky, Paul. 2005. Blocking and periphrasis in inflectional paradigms. In Geert Booji & Jaap van Maarle (eds.), Yearbook of Morphology 2004, 113–135. Dordrecht: Kluwer Academic Publishers. Leslau, Wolf. 1958. The Verb in Harari (South Ethiopic). Berkeley, Los Angeles: University of California Press. Leslau, Wolf. 1995. Reference Grammar of Amharic. Wiesbaden: Harrassowitz. Leslau, Wolf. 1997. Ethiopic Documents: Argobba. Grammar and Dictionary. Wiesbaden: Harrassowitz.

178

Maria Bulakh

Matthews, Peter H. 1974. Morphology. An Introduction to the Theory of Word-Structure. London: Cambridge University Press. Meyer, Ronny. 2005. Das Zay: Deskriptive Grammatik einer Ostguragesprache (Äthiosemitisch). Cologne: Köppe. Meyer, Ronny. 2006. Wolane: Descriptive Grammar of an East Gurage Language (Ethiosemitic). Cologne: Köppe. Rubin, Aaron D. 2007. On the third person preformative l/n in Aramaic and an Ethiopic parallel. Ancient Near Eastern Studies 44. 1–28. Stump, Gregory Thomas. 2001. Inflectional Morphology: A Theory of Paradigm Structure. Cambridge: Cambridge University Press. Testelets, Yakov G. 2008. Russian works on linguistic typology in the 1960– 1990s. In Martin Haspelmath (ed.), Language Typology and Language Universals: An International Handbook, vol. I, 306–323. Berlin & New York: De Gruyter. Voigt, Rainer M. 1977. Das tigrinische Verbalsystem. Berlin: Reimer. Wagner, Ewald. 2011. Harari. In Stefan Weninger (ed.), Semitic Languages: An International Handbook, 1257–1265. Berlin: De Gruyter Mouton. Wetter, Andreas. 2010. Das Argobba: Eine deskriptive Grammatik der Varietät von Shonke und T’ollaha (Zentraläthiopien). Cologne: Köppe.

The Asymmetry of Verbal Markedness in Libido Joachim Crass Mainz University Abstract This paper is concerned with the mismatch between morphological markedness of affirmative and negative main and subordinate verbs, and clausal finiteness in Libido, a Highland East Cushitic language of the Afroasiatic language family. It will be shown that the verb system of Libido shows a considerable complexity of morphological markedness. In order to describe this complexity, the concept of asymmetry as developed by Bisang (1998; 2001; 2007) is applied. This concept deals with obligatory categories and their occurrence in main and subordinate verbs, respectively. Furthermore, structural asymmetries between affirmative and negative forms of both main and subordinate verbs are described by applying the typology of standard negation as developed by Miestamo (2005). 1 Introduction According to the 2007 Ethiopian Population and Housing Census the number of people who claim to belong to the ethnic group of Libido is reckoned at some 64,000 people (Census 2008:85). Almost 57,000 of them live in south central Ethiopia some 120 km south of Addis Ababa (Census 2008:99). Libido belongs to the Highland East Cushitic (HEC) subgroup of East Cushitic and is most closely related to Hadiyya. Other closely related languages are (i) the members of the Kambaata group (Alaaba, K’abeena, Kambaata, T’imbaaro), (ii) Sidaama, (iii) Gedeo, and (iv) Burji (Hudson 1976; Crass 2001). Libido has 17 affirmative verb paradigms on two levels of clausal finiteness, referred to as main verbs and subordinate verbs. This number of verb paradigms does not include past-marked verbs, nominalized verbs, and focused verbs. In addition, there are three negative paradigms, namely negative imperative, negative hortative/jussive, and negative converb (NEG-converb). The level of clausal finiteness does not correspond to the degree of morphological markedness of both main and subordinate verbs. This situation corresponds with recent findings in formal linguistics that

180

Joachim Crass “the notion of finiteness is essentially decomposed into the syntactic aspect provided with an interpretational foundation, on the one hand, and the morphological aspect, on the other. The nature of the relationship between the two is largely arbitrary and depends on the individual language.” (Nikolaeva 2007:12)

The situation described by Nikolaeva holds true also for Libido. The focus of the article1 is to describe the variety of morphological markedness of main vs. subordinate verbs and affirmative vs. negative verbs using the concepts of asymmetry as developed by Bisang and Miestamo, respectively. The article is organized as follows. In Section 2 I will introduce Bisang’s (1995; 1998; 2001; 2007) and Miestamo’s (2005) concepts of asymmetry followed by a description of the Libido verb morphology in Section 3. Then I will discuss the morphological structure of affirmative main vs. subordinate verbs applying Bisang’s concept of asymmetry. The symmetry/asymmetry of affirmative vs. negative main and subordinate verbs using Miestamo’s concept of asymmetry is dealt with in Section 5, followed by concluding remarks in Section 6. 2 The concept of asymmetry 2.1 Bisang’s concept of asymmetry between main and subordinate verbs The concept of asymmetry was developed by Bisang “in the context of converbs, i.e. verb forms that are specialized for combining clauses sequentially or adverbially [...], but cannot form a sentence on their own, i.e. they cannot occur as main predicates of independent clauses.” (Bisang 2001:1405) A similar binary distinction is made by Hengeveld (1998:339), who distinguishes between “[a]n independent verb form [...] which may be used in main clauses” and “[a] dependent verb form [...] which is used in subordinate constructions only.” Bisang’s and Hengeveld’s binary distinction of main verbs/independent verbs on the one hand and converbs/dependent verbs is applied in this paper using Bisang’s terms main verb and converb. Contrary to Bisang and Hengeveld, however, I discuss also verbs which constitute relative clauses. These verbs are referred to as relative verbs. Therefore my term subordinate verb comprises converbs and relative verbs. 1

I am grateful to Ronny Meyer and Yvonne Treis for useful comments on an earlier version of this paper. Furthermore, I want to thank Yvonne for intense discussions on the complex verb morphology of Highland East Cushitic languages and Walter Bisang and Andrej Malchukov for discussions on asymmetry and related topics.

The Asymmetry of Verbal Markedness in Libido

181

According to Bisang (2001:1405) the concept of asymmetry is relevant only for languages with obligatory grammatical categories because, “[l]anguages with no obligatory categories cannot develop any asymmetry between finite and non-finite clauses, since asymmetry can be produced only by leaving out some categories in the subordinate form that are compulsory in the main clause form or by adding information to the subordinate form which is not necessary in the main clause form.” Bisang calls the former type minus-asymmetry and the latter plus-asymmetry. Categories which can produce minus-asymmetry are tense, illocutionary force, person, and politeness (Bisang 2007:124). Categories which can produce plus-asymmetry are special markers of subordination, markers of case, and markers of person (Bisang 1998:740; 2001:1405; 2007:124). 2.2 Miestamo’s concept of asymmetry between affirmation and negation Miestamo (2005) developed a concept of asymmetry for his typological study of clausal negation. He distinguishes between a formal and a functional level of asymmetry. The latter encompasses cognitive, pragmatic, and semantic phenomena, the former deals with the structural differences of affirmation and negation (Miestamo 2005:7). The concept of structural asymmetry excludes the negation marker, i.e. it deals with “differences between the affirmative and the negative that occur in addition to the presence of the negative marker(s)” (Miestamo 2005:52). Two instances of structural asymmetry are distinguished by Miestamo: paradigmatic asymmetry and constructional asymmetry. Paradigmatic asymmetry refers to the uneven “number of paradigmatic distinctions available in the affirmative and in the negative” (Miestamo 2005:15), e.g. if in a given language the distinction between perfective and perfect made in the affirmation is neutralized in the negation. Miestamo (2005:54) calls this type of paradigmatic asymmetry paradigmatic neutralization. In constructional asymmetry on the other hand an affirmative verb form itself is structurally different from its corresponding negative form “affecting different domains (finiteness, reality status, TAM, etc.)” (Miestamo 2005:15). Whereas Miestamo (2005:1) “focuses on standard negation [...], i.e. the basic way(s) a language has for negating declarative verbal main clauses” (for a more detailed definition, cf. Section 5.1), I will include both declarative and non-declarative main verbs and subordinate verbs, i.e. relative verbs and converbs, into the discussion of Miestamo’s concept of asymmetry between affirmation and negation.

182

Joachim Crass

3

Morphological structure of main and subordinate verb paradigms in Libido 3.1 Introduction Main verbs are divided into two groups: (i) three aspectually marked declarative verbs, and (ii) three non-declarative verbs. The latter group consists of the modally marked categories imperative and hortative/jussive and of the verb which constitutes the apodosis of a hypothetical conditional clause (APOD.HYP-verb). Subordinate verbs fall into two groups: (i) three aspectually marked relative verbs and (ii) eight converbs. Functionally, converbs can be divided into two types, namely converbs with sequential and adverbial function (basically manner) on the one hand and converbs with specific adverbial functions on the other hand. The former are called narrative converbs by Bisang (1995:154 ff.) but narrative or coordinative converbs by Nedjalkov (1995:109), the latter conjunctional converbs by Bisang (1995:154 ff.). Nedjalkov (1995:106 f.), however, distinguishes between specialized converbs if they have one or two adverbial meanings and contextual converbs if they have three or more adverbial meanings, depending on the context. Libido has three narrative and five specialized/contextual converbs. The three narrative converbs are named according to their formal marking: øconverb, la-converb and oonciina-converb. These three converbs do not differ in function but seem to be used interchangeably and often to avoid frequent use of one of these three converbs (for examples of their use, cf. Section 3.3.2). The five specialized/contextual converbs are named according to their respective main adverbial function. These are converb of simultaneity (SML-converb), converb of purpose (PURP-converb), converb of hypothetical condition (HYP-converb), converb of anteriority (ANT-converb), and ‘until’converb (TAQ-converb).2 SIM-converb, PURP-converb, HYP-converb, and TAQconverb are specialized converbs in the sense of Nedjalkov because they have only one specific adverbial meaning. ANT-converb is a contextual converb, because it can constitute temporal adverbial clauses of anteriority, sinceadverbial clauses and real conditional clauses (for examples of the use of these converbs, cf. Section 3.3.3). All verb forms except the affirmative imperative contain subject markers. These subject markers reflect the Afroasiatic pattern of subject marking (Hudson 1976:262 f.; Appleyard 2006:335). The subject marker of 1SG, 3SG.M, and 3PL is a zero morpheme, the subject marker of 2SG, 3SG.F, and 2

The abbreviation TAQ-converb derives from the term terminus ad quem created by Kortmann (1997: 85) for temporal clauses which express an ‘until’-relation.

The Asymmetry of Verbal Markedness in Libido

183

2PL is {t}, and the subject marker of 1PL is {n}. The subject markers of 2PL and 3PL contain the additional plural marker {ak}, which differentiates them from the respective subject markers of the singular, a phenomenon found also in other East Cushitic languages, e.g. Afar (Parker & Hayward 1985:252) and Somali (Saeed 1987:60); cf. also Sim (1988:449). Regarding the morpheme count the following remarks have to be made: The lexical base of the verb, i.e. the “part of a word remaining when all [inflectional] affixes have been removed” (Crystal 1997:40), is included in the morpheme count. Furthermore, the subject marker will always be counted as single morpheme, even if it is a zero morpheme (1SG and 3SG.M) or a morpheme that can be further analyzed (2PL and 3PL). 3.2 Main verbs 3.2.1 Modally marked verbs (imperative and hortative/jussive) Affirmative imperative, which only occurs in the 2SG and 2PL, contains two morphemes: (i) lexical base and (ii) the morpheme that indicates the respective verb form. All other modally marked verbs contain three morphemes: (i) lexical base, (ii) subject marker, and (iii) the morpheme that indicates the respective verb form. Table 4 shows the morphological structure of affirmative and negative imperative and hortative/jussive using the verb baʔ- ‘go’. Table 4: Imperative and Hortative/Jussive AFF

SJ

IMPERATIVE

1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

– baʔ-e – – – baʔ-ette –

ø t ø t n tak ak

NEG

AFF

NEG

IMPERATIVE

HRT/JUSS

HRT/JUSS

– ba-t-itte – – – ba-tak-itte –

baʔ-ø-ona ba-t-ona baʔ-ø-ona ba-t-ona ba-nʔ-ona ba-tak-ona baʔ-ak-ona

baʔ-ø-oone ba-t-oone baʔ-ø-oone ba-t-oone ba-nʔ-oone ba-tak-oone baʔ-ak-oone

In 2SG, 3SG.F and 2PL of the verb baʔ- ‘go’ the glottal stop of the lexical base is deleted. In 1PL the glottal stop and the subject marker {-n} are assimilated to a glottalized nasal, which is transcribed [nʔ]. Affirmative imperative is the only verb form in the whole verbal system of Libido that lacks subject markers. This situation fits well with the fact described by König & Siemund (2007:304) that in morphological marking of (affirmative) imperatives “agreement affixes for person [...] are frequently

184

Joachim Crass

suppressed.” Hortative/jussive paradigms contain all persons, i.e. they include rarely used forms for 2SG and 2PL. This is remarkable because many Afroasiatic languages lack forms for the second person in hortative/jussive paradigms. 3.2.2 Declarative verbs Declarative verbs are aspectually marked. They consist of five morphemes, namely (i) lexical base, (ii) subject marker, (iii) the respective aspect vowel, and two further morphemes, which are most probably the result of a grammaticalization process which Hudson (1976:262) describes as “compounding a verb stem [...] and an auxiliary verb or copula conjugated with person prefixes.” The diachronic origin of this compound verb form is the topic of discussion in several articles (cf. Sim 1988; Tosco 1996; Crass 2013). For the analysis of Libido it is necessary to divide this “auxiliary verb” into two morphemes because they perform different functions synchronically. The consonantal part is analyzed as second subject marker and the vocalic part as declarative marker. Consequently, from now on the subject marker which follows the lexical base will be referred to as first subject marker. Table 5 shows the morphological structure of the three affirmative declarative verb paradigms using the verb ʔit- ‘eat’. Table 5: Declarative Verbs 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

SJ 1 ø t ø t n tak ak

SJ 2 mm tt k ʔ mm ʔ ʔ

IMPERFECTIVE ʔit-ø-aa-mm-o ʔit-t-aa-tt-o ʔit-ø-aa-k-o ʔit-t-a-ʔ-o ʔi-n.t-aa-mm-o ʔit-tak-a-ʔ-o ʔit-ak-a-ʔ-o

PERFECTIVE ʔit-ø-u-mm-o ʔit-t-i-tt-o ʔit-ø-u-k-o ʔit-t-o-ʔ-o ʔi-n.t-u-mm-o ʔit-tak-o-ʔ-o ʔit-ak-o-ʔ-o

PERFECT ʔit-ø-oo-mm-oo ʔit-t-oo-tt-oo ʔit-ø-oo-k-oo ʔit-t-oo-ʔ-oo ʔi-n.t-oo-mm-oo ʔit-tak-oo-ʔ-oo ʔit-ak-oo-ʔ-oo

In this verb paradigm the subject marker of 1PL {n} is metathesized with the non-geminated base-final consonant /t/. This is indicated by a dot between the nasal and the final consonant. The aspect vowel, which follows SJ 1, is {aa} for imperfective, {u} for perfective, and {oo} for perfect. Morphophonologically triggered changes occur in several forms, e.g. in 3SG.F, 2PL and 3PL where – in imperfective – the aspect vowel is shortened or – in perfective – the aspect vowel is assimilated from [u] to [o]. In 2SG of perfective the aspect vowel {u} does not occur

The Asymmetry of Verbal Markedness in Libido

185

and is replaced by [i], which I consider epenthetic. Furthermore, the declarative marker is lengthened in perfect, probably in order to differentiate clearly several forms of perfect from the respective forms of perfective, namely 3SG.F, 2PL and 3PL. It is typologically remarkable that the subject marker is placed closer to the lexical base than the aspect vowel because, according to Hall (2000:542) “it has been proposed that more relevant categories appear closer to the stem than less relevant categories [...]. Thus, affixes expressing aspect tend to appear closer to the verb stem than those expressing subject agreement, and the iconicity hypothesis claims that this is because aspect expresses fundamental properties of the verb meaning, whereas subject agreement refers to an argument of the verb rather than affecting the verb meaning itself.” This phenomenon is common not only to Libido and closely related HEC languages but also to other East Cushitic languages, e.g. Oromo (Mohammed & Zaborski 1990:5 f.). 3.2.3 APOD.HYP-verb APOD.HYP-verb form contains (i) lexical base, (ii) SJ 1, (iii) the vowel {ee}, (iv) SJ 2, and (v) the morpheme {iira}. Table 6 shows the morphological structure of this verb paradigm using the verb ʔag- ‘drink’. Table 6: APOD.HYP-Verb 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

SJ 1 ø t ø t n tak ak

SJ 2 mm tt k ʔ mm ʔ ʔ

APOD.HYP-VERB ʔag-ø-ee-mm-iira ʔag-g-ee-tt-iira ʔag-ø-ee-k-iira ʔag-g-ee-ʔ-iira ʔa-n.g-ee-mm-iira ʔag-gak-ee-ʔ-iira ʔag-ak-ee-ʔ-iira

In this verb paradigm the subject markers of 2SG, 3SG.F and 2PL are assimilated to the final consonant of the lexical base and the subject marker of 1PL is metathesized with the final consonant of the lexical base. The morpheme {iira} replaces the declarative marker of aspectually marked verbs, i.e. the vowel {o(o)}. Furthermore, the vowel [ee] does not function as aspect vowel in Libido but is attested in this function in other

186

Joachim Crass

HEC languages, namely in Alaaba, Kambaata, Gedeo and Sidaama (cf Hudson 1976:262 ff.; Sim 1988:436; Schneider-Blum 2007:202 ff.; Treis 2012:27). 3.3 Subordinate verbs 3.3.1 Relative verbs When comparing relative verbs with their declarative counterparts (cf. Table 5, it becomes clear that relative verbs lack both the declarative marker and SJ 2 in all persons except 1SG and 1PL. Therefore, relative verbs contain three, and in 1SG and 1PL four morphemes: (i) lexical base, (ii) SJ 1, (iii) the respective aspect vowel, and (iv) SJ 2 in 1SG and 1PL. Table 7 shows the morphological structure of the three relative verb paradigms using the verb hin- ‘dig’. Table 7: Relative Verbs 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

SJ 1 ø t ø t n tak ak

SJ 2 m – – – m – –

IMPERFECTIVE hin-ø-aa-m hin-t-aa hin-ø-aa hin-t-a hin-n-aa-m hin-tak-a hin-ak-a

PERFECTIVE hin-ø-u-m hin-t-i hin-ø-u hin-t-o hin-n-u-m hin-tak-o hin-ak-o

PERFECT hin-ø-oo-m hin-t-oo hin-ø-oo hin-t-oo hin-n-oo-m hin-tak-oo hin-ak-oo

The occurrence of SJ 2 leads to a distinction between 1SG and 3SG.M in all aspects. This is remarkable because in converb forms which lack SJ 2 totally, 1SG and 3SG.M on the one hand and 2SG and 3SG.F on the other hand are indistinguishable (cf. Sections 3.3.2 and 3.3.3). Furthermore, 2SG differs from 3SG.F in imperfective due to different vowel quantity and in perfective due to different vowel quality. The only forms which are indistinguishable are 2SG and 3SG.F of perfect. 3.3.2 Narrative converbs The three narrative converbs differ remarkably from each other in their morphological structure. Ø-converb consists of two morphemes: (i) lexical base and (ii) SJ 1. Ø-converb is therefore the least marked subordinate verb form and comparable to the affirmative imperative, which is the least marked main verb form. The difference between the affirmative imperative and ø-converb is the fact that – beside the lexical base – the former contains a morpheme

The Asymmetry of Verbal Markedness in Libido

187

that indicates its function, whereas the latter contains the subject marker (for the identical situation in K’abeena, cf. Crass 2005:176). La-converb consists of four, and in 1SG and 1PL five morphemes: (i) lexical base, (ii) SJ 1, (iii) aspect vowel {u}, (iv) SJ 2 in 1SG and 1PL, and (v) the morpheme {la} that indicates this verb form. Oonciina-converb consists of five morphemes: (i) lexical base, (ii) SJ 1, (iii) aspect vowel {u}, (iv) SJ 2 in all persons, and (v) the morpheme {oonciina} that indicates this verb form. Table 8 shows the morphological structure of narrative converbs using the verb ʔaf- ‘hold’. Table 8: The Three Narrative Converbs SJ 1

Ø-CONVERB

SJ 2

La-CONVERB

SJ 2

1SG

ø

ʔaf-øi

m

ʔaf-ø-u-m-la

mm

2SG

t

ʔaf-fi

–

ʔaf-f-i-la

tt

3SG.M

ø

ʔaf-øi

–

ʔaf-ø-u-la

k

3SG.F

t

ʔaf-fi

–

ʔaf-f-o-la

ʔ

1PL

n

ʔa-n.fi

m

ʔa-n.f-u-m-la

mm

2PL

tak

ʔaf-faki

–

ʔaf-fak-o-la

ʔ

3PL

ak

ʔaf-aki

–

ʔaf-ak-o-la

ʔ

OonciinaCONVERB

ʔaf-ø-u-mmoonciina ʔaf-f-i-ttoonciina ʔaf-ø-u-koonciina ʔaf-f-o-ʔoonciina ʔa-n.f-u-mmoonciina ʔaf-fak-o-ʔoonciina ʔaf-ak-o-ʔoonciina

As stated earlier ø-converb, la-converb and oonciina-converb do not differ in function. Their first function is to describe a sequence of events, as shown in (1) to (3). (1) geeʔmado

fook’aaʔi

lallata

fissameeʔuko.

enclosure.ACC open.3SG.M.CNV1 cattle.F.ACC go_out.CAUS1.3SG.M.PRH.PST

‘[Every day] he opened the enclosures and took out the cattle.’ (2) haraaššeʔuwata ʔollaani ox.PL.F.ACC

ʔuunt’umla

ʔabuullaammo.

neighbor.ABL plead.1SG.CNV2 farm.1SG.IPFV

‘I will ask oxen from our neighbors and I will farm.’

188

Joachim Crass

(3) mooranni tubanni karukoonciina laro bull.GLI

tail.GLI

tie.3SG.M.CNV3

cattle.ACC

ʔamadi

baʔuko.

hold.3SG.M. CNV1

go.3SG.M.PFV

‘He tied [it] to the bull’s tail, took the cattle and went home.’ The second function of the three narrative converbs is to modify another verb adverbially, more specifically manner, as shown in (4) to (6). (4) ʔiittammi

heenʔaammo.

like.RECP.1PL.CNV1

live.1PL.IPFV

‘We will live in peace together.’ (lit. ‘Loving each other we will live.’) (5) ʔisa

mooʔaʔi,

ʔoracco jall=yula

…

3SG.F.GEN body.ACC stare=say.3SG.M.CNV2 see.3SG.M.SML

‘While he stared at her body, …’ (6)

… yukoonciina say.3SG.M.CNV3

murur=yi

wiʔlaʔi

macc’eensummo.

be_sad=say.3SG.M. CNV1

wail.3SG.M.

hear.1PL.PFV

SML

‘We heard him wailing sadly saying, “…”’ A special case of the adverbial function is to mark direct quotes of verba dicendi as shown in (7) to (9). (7) sillik’a

hannu

ʔeebbitto

yiti

t’eʔmitoʔo.

flour.ACC where.ABL bring.2SG.PFV say.3SG.F.CNV1 ask.3SG.F.PFV

‘“From where did you bring the flour?”, they3 asked.’

(8) gassette

yakola

t’eʔmakoʔo.

stay_overnight.CAUS1.IMP.PL say.3PL.CNV2 ask.3PL.PFV

‘“[Please] let us stay overnight”, they asked.’ (9) gasse

yukoonciina

t’eʔmuko.

stay_overnight.CAUS1.IMP.SG say.3SG.M.CNV3 ask.3SG.M.PFV

‘“[Please] let me stay overnight”, he asked.’ 3

The 3SG.F can either refer to a single feminine referent or to several referents. Libido seems to be in the process of changing the plural agreement from 3PL to 3SG.F, which restricts the use of 3PL to an impersonal or polite form. This process is already completed in other Cushitic languages, e.g. K’abeena (cf. Crass 2005:157 f.). Hadiyya, however, restricts the use of 3SG.F to single feminine referents.

The Asymmetry of Verbal Markedness in Libido

189

3.3.3 Specialized converbs Regarding their morphological structure, specialized converbs form two groups, namely specialized converbs with three morphemes and specialized converbs with five morphemes. SML-converb, PURP-converb, HYP-converb and NEG-converb consist of (i) lexical base, (ii) SJ 1 and (iii) the morpheme that indicates the respective verb form. Due to the lack of SJ 2, 1SG is not distinguished from 3SG.M and 2SG is not distinguished from 3SG.F. Table 9 shows the morphological structure of these paradigms using the verb fir- ‘go out, depart’. Table 9: Specialized Converbs with Three Morphemes SJ 1 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

ø t ø t n tak ak

SML-

PURP-

HYP-

CONVERB

CONVERB

CONVERB

CONVERB

fir-ø-aʔi fit-t-aʔi fir-ø-aʔi fit-t-aʔi fil-l-aʔi fit-tak-aʔi fir-ak-aʔi

fir-ø-ena fit-t-ena fir-ø-ena fit-t-ena fil-l-ena fit-tak-ena fir-ak-ena

fir-ø-aaʔni fit-t-aaʔni fir-ø-aaʔni fit-t-aaʔni fil-l-aaʔni fit-tak-aaʔni fir-ak-aaʔni

fir-ø-oʔni fit-t-oʔni fir-ø-oʔni fit-t-oʔni fil-l-oʔni fit-tak-oʔni fir-ak-oʔni

NEG-

In verbs with the non-geminated base-final consonant /r/ two morphophonological processes take place. The final consonant of the lexical base assimilates regressively to the subject marker {t} of 2SG, 3SG.F and 2PL and assimilates reciprocally with the subject marker {n} of 1PL to [ll]. SML-converb expresses a range of closely related semantic functions that Kortmann (1997:87) distinguishes as simultaneity overlap, simultaneity duration and simultaneity co-existensiveness, as in (10) and (11). Further sentences with this converb are found in (5) and (6). (10) anganni

mooʔisaʔi

… yula

hand/arm.GLI see.CAUS1.3SG.M.SML

t’eʔmuko.

say.3SG.M.CNV2 ask.3SG.M.PFV

‘He pointed with his finger [at her] and asked, “…”’ (11) šibaalsiisakaʔi

gassakoʔo.

sing_and_dance.CAUS3.3PL.SML stay_overnight.CAUS1.3PL.PFV

‘They spent the night making them sing and dance.’

190

Joachim Crass

PURP-converb is the head of purpose clauses (12) and HYP-converb the head of the protasis of a hypothetical or counterfactual conditional clause (13). (12) ka mancanniina ʔeddeessena this man.DAT

geettoʔo.

clothe.CAUS2.3SG.F.PURP run.3SG.F.PFV

‘She ran to clothe the man with it.’ (13) sibaattaaʔni

muunteettiira-ššo.

be_hungry.2SG.HYP

belch.2SG.APOD.HYP-NEG

‘If you had been hungry, you would not have belched.’ ANT-converb and TAQ-converb consist of five morphemes: (i) lexical base, (ii) SJ 1, (iii) aspect vowel {u}/vowel {ee}, respectively, (iv) SJ 2 in all persons, and (v) the morpheme that indicates the respective verb form. Table 10 shows the morphological structure of these verb paradigms using the verb y‘say’. In this verb the epenthetic vowel [i] separates the mono-consonantal lexical base from SJ 1, if the latter is not a zero morpheme. In order to show the morphological structure of the verb forms, the epenthetic vowel is not separated from the lexical base by a hyphen. Table 10: Specialized Converbs with Five Morphemes 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

SJ 1 ø t ø t n tak ak

SJ 2 -mm -tt -k -ʔ -mm -ʔ -ʔ

ANT-CONVERB y-ø-u-mm-aaʔlu(unsi) yi-t-i-tt-aaʔlu(unsi) y-ø-u-k-aaʔlu(unsi) yi-t-o-ʔ-aaʔlu(unsi) yi-n-u-mm-aaʔlu(unsi) yi-tak-o-ʔ-aaʔlu(unsi) y-ak-o-ʔ-aaʔlu(unsi)

TAQ-CONVERB y-ø-ee-mm-aaru yi-t-ee-tt-aaru y-ø-ee-k-aaru yi-t-ee-ʔ-aaru yi-n-ee-mm-aaru yi-tak-ee-ʔ-aaru y-ak-ee-ʔ-aaru

ANT-converb may constitute an adverbial clause of anteriority, a ‘since’temporal clause or a real conditional clause. (14) guullitakoʔaaʔluunsi ʔani finish.2PL.ANT

kuraam

luwi

yooko.

1SG.NOM tell.1SG.IPFV.REL thing.NOM exist.3SG.M

‘Since/After you have finished that, I have to tell you something.’

The Asymmetry of Verbal Markedness in Libido (15) ʔísseti 3PL.NOM

faradakoʔaaʔlu

ʔittaatto.

judge.3PL.ANT

eat.2SG.IPFV

191

‘If they judge in your favor, you can eat me.’ TAQ-converb marks a type of temporal clause which is normally introduced in English with ‘until’ (cf. footnote 2). (16) k’amáccoti-m monkey.SG.F.NOM-EMP

waareekaaru

hakk’a

come.3SG.M.TAQ

wood.ACC descend.CAUS2. 3SG.F.PFV.REL

dilliisso …

‘Until [the lion] came back, the monkey brought down some firewood, …’ 4 The symmetry/asymmetry of affirmative main vs. subordinate verbs This section discusses the morphological markedness of affirmative main vs. subordinate verbs in the light of Bisang’s concept of minus- and plusasymmetry. Declarative verbs are the starting point of analysis because they constitute declarative sentences, which are considered functionally unmarked. 4.1 Declarative verbs and converbs with five morphemes Four verb paradigms have the same morphological complexity as declarative verbs: APOD.HYP-verb, oonciina-converb, ANT-converb, and TAQ-converb. All verbs contain five morphemes: (i) lexical base, (ii) SJ 1, (iii) aspect vowel, (iv) SJ 2, and (v) the morpheme which marks the respective verb form. Table 11 contrasts all verbs paradigms, now presenting only 3SG.M. Table 11: Main Verbs and Converbs with Five Morphemes MAIN VERBS DECLARATIVE VERB (Imperfective, Perfective, Perfect) APOD.HYP-VERB

y-ø-aa-k-o y-ø-u-k-o y-ø-oo-k-o(o) y-ø-ee-k-iira

CONVERBS Oonciina- y-ø-u-k-oonciina CNV

ANT-CNV TAQ-CNV

y-ø-u-k-aaʔlu(unsi) y-ø-ee-k-aaru

The main difference between all verb paradigms is the final morpheme which marks the respective verb form. Furthermore, APOD.HYP-verb and TAQconverb contain the vowel {ee} in aspect vowel position. This latter fact will be ignored here because it would trigger a discussion of the status of this vowel. The other two converbs are formed on the basis of perfective.

192

Joachim Crass

The question is whether this situation produces minus- or plus-asymmetry between declarative verbs and the respective converbs. As mentioned in Section 2.1, Bisang (2001:1405) defines minus-asymmetry as “leaving out some categories in the subordinate form that are compulsory in the main clause form” and plus-asymmetry as “adding information to the subordinate form which is not necessary in the main clause form.” In these converbs the respective marker of subordination replaces the declarative marker. Therefore, the declarative marker, i.e. the category illocutionary force, is left out (i.e. minus-asymmetry) and special markers of subordination are added (i.e. plusasymmetry). However, Bisang (2001:1405) states that “[f]orms which exclusively mark clause combining such as markers of sequentiality or adverbial subordination need a special treatment. If they are combined with the omission of a category which is obligatory in the verb form of the main clause I shall subsume them under the heading of minus asymmetry. […] If they are added to a form which could otherwise also occur in the function of a main clause I shall treat them as a case of plus asymmetry.” Therefore, the relation between declarative verbs and converbs with five morphemes is considered minus-asymmetry by leaving out the category illocutionary force. 4.2 Declarative verbs and converbs with three morphemes Four verb forms have three morphemes: hortative/jussive, SIM-, PURP-, and HYP-converbs. The morphemes are (i) lexical base, (ii) SJ 1, and (iii) the marker that constitutes the respective verb form. Table 12 contrasts all verbs paradigms consisting of three morphemes with declarative verbs consisting of five morphemes, again presenting only 3SG.M. Table 12: Main Verbs and Converbs with Three Morphemes MAIN VERBS DECLARATIVE VERB (Imperfective, Perfective, Perfect) HORTATIVE/JUSSIVE

y-ø-aa-k-o y-ø-u-k-o y-ø-oo-k-o(o) y-ø-ona

CONVERBS SIM-CONVERB PURP-CONVERB

y-ø-aʔi y-ø-ena

HYP-CONVERB

y-ø-aaʔni

The relation between declarative verbs and converbs with three morphemes constitutes a clear case of minus-asymmetry. In contrast to converbs with five morphemes, minus-asymmetry is produced here by leaving out three mor-

The Asymmetry of Verbal Markedness in Libido

193

phemes, i.e. (i) the declarative marker, (ii) the aspect vowel, and (iii) SJ 2, and by replacing them with the respective marker of subordination. However, regarding Bisang’s categories which produce minus asymmetry only two categories are involved, namely (i) illocutionary force, and (ii) tense, if the category aspect is subsumed under the category tense in the discussion of asymmetry. The category person, however, is not involved because all verbs contain SJ 1, i.e. all converbs are still marked for the category person. 4.3 Declarative verbs, relative verbs and la-converb In this section I will discuss the relation between declarative verbs, relative verbs, and la-converb. Table 13 contrasts these verbs paradigms, this time presenting 1SG and 3SG.M. Table 13: Declarative Verbs, Relative Verbs and la-Converb IMPERFECTIVE PERFECTIVE PERFECT

DECL VERBS y-ø-aa-mm-o y-ø-aa-k-o y-ø-u-mm-o y-ø-u-k-o y-ø-oo-mm-o y-ø-oo-k-o(o)

REL VERBS y-ø-aa-m y-ø-aa y-ø-u-m y-ø-u y-ø-oo-m y-ø-oo

CONVERBS

LaCONVERB

y-ø-u-m-la y-ø-u-la

The relationship between declarative verbs and la-converb constitutes a clear case of minus-asymmetry. Like in all converbs the declarative marker is replaced by the marker of subordination. Since la-converb, similarly to relative verbs, contains SJ 2 in 1SG and 1PL, first person has to be considered differently. In 1SG and 1PL the minus-asymmetry is of the same type as of converbs consisting of five morphemes: The declarative marker is left out so that the category of illocutionary force remains unmarked (cf. Section 4.1). In all other forms, i.e. second and third persons, minus-asymmetry is produced by leaving out two morphemes, namely (i) declarative marker, and (ii) SJ 2. This is in contrast to converbs with three morphemes, where the aspect vowel is also left out (cf. Section 4.2). However, as stated in Section 3.2.2, declarative verbs have two person markers, SJ 1 and SJ 2. Since SJ 1 is never omitted, there is no minus-aysmmetrie produced by person in the sense of Bisang. Therefore, the category person is given in brackets in Table 14. When considering relative verbs vs. declarative verbs, the category which produces minus asymmetry is again illocutionary force, because the declarative marker is left out. In addition, one has to differentiate again between first

194

Joachim Crass

persons on the one hand and second and third persons on the other hand. In second and third persons SJ 2 is left out but in first persons, SJ 2 is kept. Table 14 summarizes the different types of minus-asymmetry between declarative and subordinate verbs. Table 14: The Three Types of Minus-Asymmetry between Declarative and Subordinate Verbs TYPE OF SUBORDINATE VERB 1. Converbs with five morphemes 2. la-Converb (first persons) 3. Relative verbs (first persons)

MORPHEMES DECL

1. DECL 1. la-Converb (second and third persons) 2. SJ 2 2. Relative verbs (second and third persons) Converbs with three morphemes

CATEGORIES Illocutionary force

1. Illocutionary force (2. Person)

1. Illocutionary force 1. DECL 2. Aspect vowel 2. Aspect (3. Person) 3. SJ 2

Finally, I want to compare the relative perfective with la-converb (cf. Table 13). This discussion exceeds Bisang’s concept of asymmetry, which only takes main verbs and converbs into account. However, by applying Bisang’s definition to relative verbs, the relationship between la-converb and relative verbs is a clear case of plus-asymmetry. La-converb is formed on the basis of relative perfective by adding the morpheme {la}. In the discussion of special negative converbs in Section 5.4 we will see that there are three other (negative) verb paradigms formed on the basis of relative verbs. 5 The symmetry/asymmetry of affirmative vs. negative verbs In the previous two sections three negative verb paradigms were presented, namely negative imperative, negative hortative/jussive and NEG-converb. These paradigms are relevant to the discussion of symmetry/asymmetry of affirmative vs. negative verbs. However, there is another way of negation, namely morphological negation, which is presented in this section. Morphological negation is here defined as a marking strategy in which segmentable negation morphemes are used. This strategy is in contrast with – what I call – negative paradigms in which portmanteau morphemes indicate negation plus

The Asymmetry of Verbal Markedness in Libido

195

other grammatical categories. If morphological negation is applied, it exemplifies a clear split between main and subordinate verbs. 5.1 Negation of main verbs Declarative verbs are negated by the suffix -(V)ššo, which is attached to the affirmative verb. Imperfective is negated by the allomorph -iššo and perfective by -oššo. Perfect is normally negated by -iššo. However, it can exceptionally be negated by -oššo. The exact condition(s) for this distribution are unclear to me. Furthermore, APOD.HYP-verb is also negated by the suffix -(V)ššo, which is realized consistently as -aššo, most probably due to the fact that the affirmative form is marked with iira, which ends in [a]. The suffix (V)ššo is the standard negation marker in the sense of Miestamo (2005:42), who defines standard negation as “a construction whose function is to modify a verbal declarative main clause expressing a proposition p in such a way that the modified clause expresses the proposition with the opposite truth value to p, [... which ] is (one of) the productive and general means the language has for performing this function.” The relationship between affirmation and negation of declarative main verbs on the one hand and APOD.HYP-verb on the other hand is symmetric. With the exclusion of some morphophonological changes, the negative marker does not cause any structural changes in the verb forms, i.e. there is no constructional asymmetry. Furthermore, each declarative affirmative paradigm and the APOD.HYP-verb have negative counterparts, i.e. there is no paradigmatic asymmetry. According to Miestamo’s definition, the paradigms of negative imperative and negative hortative/jussive are a non-standard strategy of negation (cf. Table 1). The relation between affirmative and negative imperative and hortative/jussive is asymmetric. More specifically they represent “[a]symmetry in the marking of grammatical categories” (Miestamo 2005:112 ff., labeled subtype A/Cat). Negative imperative encodes subject agreement while the affirmative does not. This strategy is represented by Miestamo’s “further subtype A/Cat/PNG”4 (cf. Miestamo 2005:128 ff.). Furthermore, the negative marker replaces the affirmative marker. This strategy is represented by Miestamo’s further subtype A/Cat/TAM (2005:116 ff.). In hortative/jussive the 4

Miestamo (2005:60) calls the subtypes of subtypes “further subtypes”, i.e. A/Cat/PNG is, besides A/Cat/TAM, a further subtype of the subtype A/Cat, the latter being one of four subtypes of asymmetric negation.

196

Joachim Crass

affirmative TAM-marker is replaced by a negative marker which includes the meaning of the respective TAM category, i.e. the further subtype of asymmetry is A/Cat/TAM. However, in both paradigms the asymmetry is constructional but not paradigmatic because each affirmative paradigm has a negative counterpart. 5.2 Negation of relative verbs Relative verbs are negated with the morpheme {beekk’i} or its shortened free variant {bee}. This negation marker is derived from the negative existential verb beekk’e/bee’e ‘not exist, not be present’. It occurs in all negative subordinate verb forms except NEG-converb (cf. Table 9). This shows that in negation there is a clear formal distinction between main and subordinate verbs, because declarative main verbs are marked by the standard negation suffix -(V)ššo. The relation between affirmative and negative relative verbs is symmetric with regard to Miestamo’s typology. The negative marker is added to the affirmative form without any structural changes and all affirmative forms have a negative counterpart. For a detailed analysis of asymmetry in negation of HEC languages cf. Treis (2012). 5.3 Negation of converbs with NEG-converb The NEG-converb is used to negate five converbs, namely all three narrative converbs, SML-converb, and TAQ-converb. This represents a clear case of paradigmatic asymmetry because there is only one way to negate these five converbs. In addition, three different types of constructional asymmetry are attested. SML-converb and NEG-converb are in constructional asymmetry because the negative marker of NEG-converb replaces the respective affirmative marker. La-converb and NEG-converb are also in constructional asymmetry, but in a different way because the negative marker causes the following structural changes: Three morphemes which occur in la-converb are not present in the negation: (i) aspect vowel, (ii) SJ 2 of 1SG and 1PL, and (iii) the morpheme {la} that indicates la-converb. Finally, oonciina-converb and TAQconverb are in constructional asymmetry with NEG-converb because the negative marker causes several structural changes. Three morphemes which occur in oonciina-converb and TAQ-converb are deleted: (i) aspect vowel/ vowel {ee}, (ii) SJ 2 in all persons, and (iii) the morpheme that marks the respective converb form. Ø-converb and NEG-converb, however, are constructionally symmetric because the negative marker is added to ø-converb. This is due to the fact that ø-converb is the only converb form without a morpheme that indicates its function (cf. Section 3.3.2).

The Asymmetry of Verbal Markedness in Libido

197

5.4 Special negation of converbs PURP-, HYP-, and ANT-converbs are negated with three specific negative converbs. These negative converbs are formally relative verbs formed with different aspect vowels. Negative HYP-converb contains the perfect aspect vowel {oo} and negative ANT-converb contains the perfective aspect vowel {u}. Negative PURP-converb, however, contains the vowel {ee} in aspect vowel position, similarly to APOD.HYP-verb and TAQ-converb. All specific negative converbs consist of five and in 1SG and 1PL of six morphemes: (i) lexical base, (ii) SJ 1, (iii) the respective aspect vowel, (iv) SJ 2 in 1SG and 1PL, (v) the negative morpheme {beekk’i}, and (vi) the morpheme that indicates the respective verb form. These negative forms are presented in the following three tables, each with its respective affirmative counterpart. Table 15 presents affirmative and negative PURP-converb using the verb hekk- ‘sleep’. Table 15: Affirmative and Negative PURP-Converb 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

SJ 1 ø t ø t n tak ak

AFF PURP-CONVERB hekk-ø-ena hekki-t-ena hekk-ø-ena hekki-t-ena hekki-n-ena hekki-tak-ena hekk-ak-ena

SJ 2 -m – – – -m – –

NEG PURP-CONVERB hekk-ø-ee-m-beekk’-ena hekki-t-ee-beekk’-ena hekk-ø-ee-beekk’-ena hekki-t-ee-beekk’-ena hekki-n-ee-m-beekk’-ena hekki-tak-ee-beekk’-ena hekk-ak-ee-beekk’-ena

Similarly to verb y- ‘say’, the epenthetic vowel [i] separates the lexical base from the following morpheme, if the latter begins with a consonant. Therefore, the epenthetic vowel again is not separated from the lexical base by a hyphen. Affirmative and negative PURP-converb are constructionally asymmetric. The addition of the negative morpheme {beekk’i} causes the occurrence of the vowel {ee} and SJ 2 in 1SG and 1PL. Furthermore, it is noteworthy that the negative morpheme precedes the morpheme {ena} which marks PURPconverb. This morpheme order is also attested in the other two specific negative converbs. Table 16 presents affirmative and negative HYP-converb using the verb lell- ‘play’.

198

Joachim Crass

Table 16: Affirmative and Negative HYP-Converb 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

SJ 1 ø t ø t n tak ak

AFF. HYP-CONVERB lell-ø-aa-ʔni lelli-t-aa-ʔni lell-ø-aa-ʔni lelli-t-aa-ʔni lelli-n-aa-ʔni lelli-tak-aa-ʔni lell-ak-aa-ʔni

SJ 2 -m – – – -m – –

NEG. HYP-CONVERB lell-ø-oo-m-beekk’i-ʔni lelli-t-oo-beekk’i-ʔni lell-ø-oo-beekk’i-ʔni lelli-t-oo-beekk’i-ʔni lelli-n-oo-m-beekk’i-ʔni lelli-tak-oo-beekk’i-ʔni lell-ak-oo-beekk’i-ʔni

Affirmative and negative HYP-converb are constructionally asymmetric. The addition of the negative marker {beekk’i} causes the occurrence of the perfect aspect vowel {oo} and SJ 2 in 1SG and 1PL. In contrast to PURP-converb the negative marker precedes only the consonantal part of the morpheme that marks the affirmative HYP-converb. This must lead to the conclusion that {aa’ni} in fact consists of two morphemes, namely {aa} and {ʔni} and the latter must be considered the main marker of PURP-converb, both affirmative and negative. It is then unclear what the function of the morpheme {aa} in the affirmative should be. Because of the formal identity of the imperfective aspect vowel {aa}, one could claim that these two morphemes are connected. This is not implausible because purpose clauses are related to the category future or, as Treis (2010:2) states, “[p]urpose clauses are inherently future-oriented.” Schmidtke-Bode (2009:42 ff.), who also discusses the link between purpose clause and imperfective, concludes that “[i]mperfective aspect marking in purpose clauses underlines or re-emphasizes the inherent futurity of purposive situations.” Then, however, the question arises why negative PURP-converb contains the perfect vowel {oo}. Another consequence of this analysis is that affirmative PURP-converb generates a separate morphological category because it would be the only verb form consisting of four morphemes. Finally, Table 17 shows affirmative and negative ANT-converb using the verb š- ‘kill’ which is – apart from y- ‘say’ – the only verb with a monoconsonantal base.

The Asymmetry of Verbal Markedness in Libido

199

Table 17: Affirmative and Negative ANT-Converb SJ 2 -m

-tt

AFF ANT-CONVERB š-ø-u-mmaaʔlu(unsi) ši-t-i-tt-aaʔlu(unsi)

ø

-k

š-ø-u-k-aaʔlu(unsi)

–

3SG.F

t

-ʔ

ši-t-o-ʔ-aaʔlu(unsi)

–

1PL

n

-mm

-m

2PL

tak

-ʔ

3PL

ak

-ʔ

ši-n-u-mmaaʔlu(unsi) ši-tak-o-ʔaaʔlu(unsi) š-ak-o-ʔ-aaʔlu(unsi)

1SG

SJ 1 ø

SJ 2 -mm

2SG

t

3SG.M

–

– –

NEG ANT-CONVERB š-ø-u-m-beekk’aaʔlu(unsi) ši-t-i-beekk’aaʔlu(unsi) š-ø-u-beekk’aaʔlu(unsi) ši-t-o-beekk’aaʔlu(unsi) ši-n-u-m-beekk’aaʔlu(unsi) ši-tak-o-beekk’aaʔlu(unsi) š-ak-o-beekk’aaʔlu(unsi)

Affirmative and negative ANT-converb are constructionally asymmetric. The addition of the negative morpheme {beekk’i} causes the deletion of SJ 2 in all persons except 1SG and 1PL. 5.5 Summary of symmetry/asymmetry of affirmative vs. negative verbs Declarative affirmative and negative main verbs are paradigmatically and constructionally symmetric. The same holds true for APOD.HYP-verb. In both cases each affirmative paradigm has a negative counterpart and the negative marker is added without any structural changes. Modally marked affirmative and negative main verbs, however, are in constructional asymmetry because the negative marker replaces the affirmative marker both in imperative and hortative/jussive (Miestamo’s further subtype A/Cat/TAM). Additionally, the negative marker causes the occurrence of subject markers in imperative (Miestamo’s further subtype A/Cat/PNG). Table 18 summarizes the relationship of affirmative and negative main verbs. In the case of constructional asymmetry, the rightmost column shows the difference in number of morphemes when the negative marker is suffixed and – in brackets – the category that is involved.

200

Joachim Crass

Table 18: Symmetry vs. Asymmetry of Affirmative and Negative Main Verbs TYPE OF VERB DECLARATIVE VERBS APOD.HYP-VERB IMPERATIVE HORTATIVE/JUSSIVE

ASYMMETRY PARADIG- CONSTRUCMATIC

TIONAL

no no no no

no no yes yes

DELETED OR ADDED CATEGORIES

+1 (PNG), –1 (TAM) –1 (TAM)

In subordinate verbs the situation is as follows. Affirmative and negative relative verbs are paradigmatically and constructionally symmetric. Converbs, however, are asymmetric with their respective negative counterpart. Three types must be distinguished: 1. Ø-converb is in paradigmatic asymmetry. 2. PURP-, HYP- and ANT-converbs are in constructional asymmetry. 3. La-, oonciina-, SML- and TAQ-converbs are in paradigmatic and constructional asymmetry. Table 19 summarizes the relationship of affirmative and negative subordinate verbs. In the case of constructional asymmetry, the rightmost column shows the difference in number of morphemes when the negative marker is attached. If two different numbers are given, the former refers to the 1SG and 1PL, the latter to all other persons. Table 19: Asymmetry of Affirmative and Negative Subordinate Verbs TYPE OF VERB RELATIVE VERBS Ø-CONVERB PURP-CONVERB HYP-CONVERB ANT-CONVERB SML-CONVERB La-CONVERB Oonciina-CONVERB TAQ-CONVERB

ASYMMETRY PARADIGCONSTRUCMATIC

TIONAL

no yes no no no yes yes yes yes

no no yes yes yes yes yes yes yes

DELETED OR ADDED CATEGORIES

+1 or +2 +1 0 or –1 –1 –2 or –3 –3 –3

The Asymmetry of Verbal Markedness in Libido

201

6 Concluding remarks Libido shows a remarkable degree of asymmetry both in main vs. subordinate verbs and in affirmative vs. negative verbs. Asymmetry in affirmative vs. negative forms of two subordinate verbs includes a shift from converb to a verb form based on a relative verb. The asymmetry of main vs. subordinate verbs is caused by a grammaticalization process that triggered the addition of an auxiliary verb to the main verb. This process is generalized in HEC languages, except Burji, so that all declarative main verbs (and several converbs) contain parts of this auxiliary verb whereas in other East Cushitic languages, this grammaticalized verb form does not occur or is restricted to specific aspects. Abbreviations 1, 2, 3 First, second, third person A Asymmetry ACC Accusative AFF Affirmative ANT Anteriority APOD Apodosis CAT Category CNV1 ø-converb CNV2 la-converb CNV3 oonciina-converb CAUS1 (i)s-causative CAUS2 iis-causative CAUS3 siis-causative DECL Declarative marker EMP Emphasis F Feminine GLI Genitive, locative, instrumental HEC Highland East Cushitic HRT Hortative HYP Hypothetical

IMP JUSS M IPFV NEG NOM PRF PRH PURP PFV PNG PST RECP REL SG SJ SJ SJ

1 2

SML TAM TAQ

Imperative Jussive Masculine Imperfective Negation Nominative Perfect Progressive/Habitual Purpose Perfective Person, number, gender Past Reciprocal Relative Singular Subject marker First subject marker Second subject marker Simultaneity Tense, aspect, mood ‘Until’-converb (cf. fn. 2)

202

Joachim Crass

References Appleyard, David. 2006. Cushitic languages. In Keith Brown (ed.), Encyclopedia of Languages and Linguistics, vol. III, 332–336. Oxford: Elsevier. Bisang, Walter. 1995. Verb serialization and converbs – differences and similarities. In Martin Haspelmath & Ekkehard König (eds.), Converbs in Cross-Linguistic Perspective: Structure and Meaning of Adverbial Verb Forms – Adverbial Participles, Gerunds, 137–188. Berlin & New York: Mouton de Gruyter. Bisang, Walter. 1998. The view from the Far East. In Johan van der Auwera (ed.), Adverbial Constructions in the Languages of Europe, 641– 812. Berlin & New York: Mouton de Gruyter. Bisang, Walter. 2001. Finite vs. non finite languages. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible (eds.), Sprachtypologie und sprachliche Universalien – Ein internationales Handbuch, vol. II, 1400–1413. Berlin & New York: De Gruyter. Bisang, Walter. 2007. Categories that make finiteness: Discreteness from a functional perspective and some of its repercussions. In Irina Nikolaeva (ed.), Finiteness: Theoretical and Empirical Foundations, 115–137. Oxford: Oxford University Press. Census. 2008. Summary and Statistical Report of the 2007 Population and Housing Census. Population Size by Age and Sex. Addis Ababa: Population Census Commission. Crass, Joachim. 2001. The position of K’abeena within Highland East Cushitic. Afrikanistische Arbeitspapiere 67. 5–60. Crass, Joachim. 2005. Das K’abeena. Deskriptive Grammatik einer hochlandostkuschitischen Sprache. Cologne: Köppe. Crass, Joachim. 2013. Some remarks on the compound suffix conjugation in Highland East Cushitic languages. In Marie-Claude SimeoneSenelle & Martine Vanhove (eds.), Proceedings of the 5th International Conference on Cushitic and Omotic Languages, Paris, 16–18 April 2008, 3–20. Cologne: Köppe. Crystal, David. 1997. A Dictionary of Linguistics and Phonetics. Oxford: Blackwell. Hall, Christopher. 2000. Prefixation, suffixation and circumfixation. In Geert Booij, Christian Lehmann & Joachim Mugdan (eds.), Morphologie –

The Asymmetry of Verbal Markedness in Libido

203

Ein internationales Handbuch zur Flexion und Wortbildung, vol. I. 535–545. Berlin & New York: De Gruyter. Hengeveld, Kees. 1998. Adverbial clauses in the languages of Europe. In Johan van der Auwera (ed.), Adverbial Constructions in the Languages of Europe, 335–419. Berlin & New York: De Gruyter. Hudson, Grover. 1976. Highland East Cushitic. In Lionel M. Bender (ed.), The Non-Semitic Languages of Ethiopia, 232–277. East Lansing, Michigan: African Studies Center, Michigan State University. König, Ekkehard & Peter Siemund. 2007. Speech act distinctions in grammar. In Timothy Shopen (ed.), Language Typology and Syntactic Description. Volume I: Clause Structure, 276–324. Cambridge: Cambridge University Press. Kortmann, Bernd. 1997. Adverbial Subordination: A Typology and History of Adverbial Subordinators Based on European Languages. Berlin & New York: Mouton de Gruyter. Miestamo, Matti. 2005. Standard Negation: The Negation of Declarative Verbal Main Clauses in a Typological Perspective. Berlin & New York: Mouton de Gruyter. Mohammed Ali & Andrzej Zaborski. 1990. Handbook of the Oromo Language. Stuttgart: Steiner. Nedjalkov, Vladimir P. 1995. Some typological parameters of converbs. In Martin Haspelmath & Ekkehard König (eds.), Converbs in CrossLinguistic Perspective: Structure and Meaning of Adverbial Verb Forms – Adverbial Participles, Gerunds, 97–136. Berlin & New York: Mouton de Gruyter. Nikolaeva, Irina. 2007. Introduction. In Irina Nikolaeva (ed.), Finiteness: Theoretical and Empirical Foundations, 1–19. Oxford: Oxford University Press. Parker, Enid M. & Richard Hayward. 1985. An Afar-English-French Dictionary (with Grammatical Notes in English). London: SOAS. Saeed, John I. 1987. Somali Reference Grammar. Wheaton, MD: Dunwoody. Schmidtke-Bode, Karsten. 2009. A Typology of Purpose Clauses. Amsterdam & Philadelphia: John Benjamins. Schneider-Blum, Gertrud. 2007. A Grammar of Alaaba. Cologne: Köppe. Sim, Ronald. 1988. The diachronic derivation of the verb in northern Highland East Cushitic. In Marianne Bechhaus-Gerst & Fritz Serzisko

204

Joachim Crass (eds.), Cushitic – Omotic. Papers from the International Symposium on Cushitic and Omotic Languages, Cologne, January 6–9, 1986, 433–452. Hamburg: Buske.

Tosco, Mauro. 1996. The northern Highland East Cushitic verb in an areal perspective. In Catherine Griefenow-Mewis & Rainer M. Voigt (eds.), Cushitic and Omotic languages. Proceedings of the Third International Symposium, Berlin, March 17–19, 1994, 71–99. Cologne: Köppe. Treis, Yvonne. 2010. Purpose-encoding strategies in Kambaata. Afrika und Übersee 91. 1–38. Treis, Yvonne. 2012. Negation in Highland East Cushitic. In Ghil’ad Zuckerman (ed.), Burning Issues in Afro-Asiatic Linguistics, 20–61. Cambridge: Cambridge Scholars Publishing.

The Finite–Infinite Dichotomy in a Comparative Semitic Perspective Lutz Edzard University of Oslo Abstract A clear-cut dichotomy between the categories finite and infinite in the verbal realm is problematic in linguistics in general (cf. Nikolaeva 2007) and in Semitic linguistics in particular – notably in connection with the Akkadian stative/verbal adjective and the conjugated noun in predicative position as well as the Ethio-Semitic converb/gerund. Already the term verbal noun (in Arabic: maṣdar) points to this inherent categorial ambiguity. What is more, in many (Semitic and other) languages, verbal nouns and infinitives can pragmatically adopt a finite function, notably as imperatives. Last but not least, the first element in serial verb constructions tends to lose the morphological features associated with finiteness and to be grammaticalized as an adverb. The same holds for Amharic converbs that are frozen in the 3SG.M and function synchronically as adverbs. All this will be considered in a comparative Semitic perspective, taking into account data mainly from Akkadian, Arabic, and Ethio-Semitic. 1 Introduction A clear-cut dichotomy between the categories finite and infinite in the verbal realm is problematic in linguistics in general (cf. Nikolaeva 2007) and in Semitic linguistics in particular. The grammatical terminology in Semitic linguistics bears testimony to this circumstance. What traditionally has been referred to as the stative (or: permansive) paradigm in Akkadian grammar has been more recently called the (conjugated) verbal adjective and even the conjugated noun in predicative position, notably in the scholarship of J. Huehnergard. The very term verbal noun (in Arabic: maṣdar, literally meaning ‘origin’) points to this inherent categorial ambiguity as well. The Ethio-Semitic converb/gerund,1 which diachronically is related to a verbal noun, is a central point in case here. Functional opaqueness in this context is not restricted to the morphological surface of the respective verb forms but is 1

F

th

h di

i

f th t

b f Shi h H l

(2009)

206

Lutz Edzard

also found on a semantic and especially pragmatic level. In many (Semitic and other) languages, infinitives can pragmatically adopt a finite function, notably when functioning as imperatives, e.g. Modern Hebrew li-sgor ʾet hadelet (to-close.INF ACC DEF-door) ‘to close the door’ (i.e. ‘Close the door!’) or German rechts stehen, links gehen (right stand.INF left go.INF) ‘stand on the right side, go on the left side!’ on many escalators. Incidentally, one also finds far-reaching morphological overlap throughout the biyanim (verbal diatheses) between the unmarked form of the imperative (2SG.M) and the infinitive in the Hebrew verbal system. As an especially interesting feature, one can observe a transition from finiteness to infiniteness, or de-finitization (“deranking” in the linguistic literature) of verb forms in certain syntactic contexts. Amharic converbs can be frozen in the 3SG.M and then function synchronically as adverbs. In other branches of Semitic, the first element in serial verb constructions tends to lose the morphological features associated with finiteness and to be grammaticalized as an adverb. This paper will consider these features in a comparative Semitic and (at least marginally in an) Afroasiatic perspective, taking into account data mainly from Akkadian, Northwest Semitic, Arabic, and Ethio-Semitic, but also from Cushitic and Omotic languages. 2 Comparative Semitic data: Conjugated verbal nouns In a historical perspective, the Akkadian stative is the most obvious case of obfuscation between the poles infinite and finite. The term verbal adjective, as used by J. Huehnergard, underscores this opacity. Indeed, Huehnergard (2005:614) subsumes the verbal adjective plus pronominal subject under the category “Non-Finite Forms”. As is well known, not only (verbal) adjectives, but also common nouns can undergo inflection, e.g. šarrāku ‘I am king’ from šarru(m) ‘king’ (comparable to a regular stative as, e.g. marṣāku ‘I am ill’). Table 20 shows the paradigm based on synchronically underlying paris ‘he is in the state of dividing’, belonging to the root √p-r-s ‘divide’: Table 20: Verbal Adjective/Stative in Akkadian 3SG.M 3SG.F 2SG.M 2SG.F 1SG.C

paris parsat parsāta parsāti parsāku

3PL.M 3PL.F 2PL.M 2PL.F 1PL.C

parsū parsā parsātunu parsātina parsānu

The Finite–Infinite Dichotomy

207

Not only infinitives, but also participles, have emerged as the base of finite conjugations. The best-known examples are found in the history of Aramaic. In modern North-Western and North-Eastern Aramaic, one finds, among other forms, a present tense based on the historical active participle and a (typically) ergative past tense based on the historical passive participle. In Ṭuroyo, a modern Eastern Aramaic language, the so-called predicate (i.e. non-ergative) inflection looks as follows, based on the root √g-r-š ‘pull’.2 The first column reflects the inflectional base C1oC2eC3 (active/present) and the second column the inflectional base C1C2iC3 (passive/resultative past): Table 21: Participle-Based Forms in Ṭuroyo 3SG.M 3SG.F 2SG.M 2SG.F 1SG.M 1SG.F

ACTIVE gorəš goršo gəršət gəršat gorašno gəršono

PASSIVE griš grišo grišət grišat grəšno grišono

3PL.C

ACTIVE gərši

PASSIVE griši

2PL.C

gəršutu

grišutu

1PL.C

gəršina

grišina

Even more idiomatic in modern Aramaic is the expression of past tense by means of an outright ergative paradigm, e.g. Ṭuroyo grəš-l-i ‘I pulled [him]’ (lit. ‘[he] is pulled to me’). Comparable to the scenario in Akkadian, both nouns and adjectives can be followed by clitic pronouns already in Classical Syriac Aramaic (cf. Rubin 2005:31 f.): Clitic Pronouns in Classical Syriac (1) a. wə-ʾena ʿap̄rā (ʾ)nā CONJ-I

dust

I

‘And I am dust’ b. ʾāmar saying

(Gen. 18:27)

(ʾ)nā

lə-ḵōn

I

to-you.PL.M

‘I say to you’

(Matt. 3:9)

The original bound orthography in these cases actually supports the analysis of these examples in terms of a cliticization process.

2

Cf

J t

(2011 701 ff )

214

Lutz Edzard early influence of Cushitic and argue for monogenesis … Some features are not found in all the languages, but they are found in representatives of each branch while not in the closest relatives of these. Such a feature is the use, employing a Semitic form according to a Cushitic pattern (i.e. calqued on Cushitic), of the converb (gerund) instead of sentence coordination. The converbial constructions are common in all the Ethiopian Semitic languages and were already so in Geʾez [sic], but the original converbial forms (based on the Semitic pattern säbir(ä)-) are found today (in addition to Geʾez) in Tigrinya, (but not in Tigre), Amharic, Argobba and, with a limited application, in central and western Gurage (with a modified pattern sïbirtä-) and Gafat.”

5 The converb in a comparative Afroasiatic perspective As already emerged from the quotation by Hetzron (1975) cited above, the pervasive use of the converb construction is an areal feature in the historical core region of Afroasiatic, encompassing Cushitic and Omotic within Afroasiatic in addition to other non-Afroasiatic language families such as NiloSaharan. Regarding the situation in Cushitic, Appleyard (2012:210 f.) summarizes as follows: “All Cushitic languages essentially follow an SOV word order. In complex sentences the main verb is usually final and subordinate clauses precede, each with internal SOV order. Most languages have rich clause-chaining mechanisms, and many have a specific converb form, sometimes called a gerundive, for this purpose, whilst others have different kinds of linking or coordinating devices, for instance involving lengthening of the terminal vowel of the clause-final verb. More specific types of adverbial relations may be indicated by conjunction-like particles, which in some languages may be clause-initial and in others take the form of suffixes added to the verb. This latter device often requires adding this suffix particle to a special form of the verb, different from the main-verb forms, and in some languages has developed through fusion into an extensive range of adverbial subordinate paradigms, such as conditional, temporal, causal, final or complemental forms.” Table 23 below shows two sample paradigms of the converb in Cushitic from Bilin and Sidaama (cf. Appleyard 2012:230; Yri 2012:268):

The Finite–Infinite Dichotomy

215

Table 23: Converbs in Cushitic SIDAAMA ros- ‘learn’

BILIN gɐb ‘refuse’ 1SG, 3SG.M 2SG, 3SG.F 1PL, 3PL 2PL

gɐb-o gɐb-ro gɐb-no gɐb-dǝno

1SG, 3SG.M 2SG, 3SG.F, 3PL 1PL 2PL

ros-e ros-te ros-ne ros-tine

The following two sentences in (9), drawing on the same two Cushitic languages Bilin and Sidaama, illustrate the functional variety (consecutiveness) of the Cushitic converb: Converb constructions in Cushitic (9)

k’wal-gəri=lom

a. jəxrana guinea_fowl

təxwla wolf

see-3SG.F.NEG.CNV=them

səkwər-dɐ

BILIN sɐna

approach-3SG.F.CNV

as

wɐlɐyd-o

ʔɐnt’ɐr

y-o

be_quick-3SG.M.CNV

jump

say-3SG.M.CNV

w

šax-əx =la seize-3M.PST=her

‘As a guinea fowl drew near without seeing them, the wolf leapt up quickly and caught her.’ (Appleyard 2012:231) SIDAAMA b. heeɗ-ɗé stay-CNV.3F

heeɗ-ɗé

y-itú

stay-CNV.3F

say-PFV.3F

geden-s-áá-nni

mereer-ó-nsa

be_later-CAUS-INF.OBL.C-ABL

middle-ACC-their

gíbb-o

kalak’-an-tannó

quarrel-NOM

create-PASS-IPFV.3F

‘After they had stayed together for a long time, a quarrel arose between them.’ (Yri 2012:270) Similar constructions are also found in the Omotic branch of Afroasiatic, e.g. in Wolaitta:

216

Lutz Edzard Converb Construction in Omotic

(10) polísee

mint-í

WOLAITTA kaisóy

ʔoicc-ín

police.DEF.NOM

strong.CAUS-CNV

miʃʃáa

k’ott-ído

sohuwa

ask-DS.CNV

bess-í

thief.M.NOM

money.M.ACC

exist-PFV.REL

place.M.ACC

show-M.CNV

g-iisi say-3SG.M.PFV

‘After the police interrogated him thoroughly, the thief showed them where he had hidden the money.’ (Azeb & Dimmendaal 2006:396) 6 Verbal hendiadys/serial verb constructions in Semitic In a further step, I will have a closer look at such coordinated structures (both ones joined by a conjunction and asyndetic ones) that have been labeled in the literature verbal hendiadys.9 Properly speaking, this term hendiadys is less than fortunate here as it mostly refers to either nouns or verbs on a par that can be understood (or translated) as a complex unit, e.g. Hebrew tōhū wā-ḇōhū ‘deserted and empty’ (Gen. 1:2) or haṣlāḥā(h) ū-ḇrāḵā(h) ‘success and blessing’ (a wish known in German in the distorted form Hals- und Beinbruch). However, and especially in the verbal realm, the term has also been understood as referring to such structures where the first element is subordinated to or modifying the second element, e.g. recently by Chrzanowski (2011). There is not yet consensus among Semiti(ci)sts on whether the term serial verb is appropriate in this context, since it is sometimes the second verb in such constructions (other than Semitic) that is semantically restricted.10 While Woidich (2002) rejects the term for this reason (as applying to Arabic), Versteegh (2003–2005; 2009) explicitly supports it. Here, only such examples will be considered where the two verbs agree in tense and mood. These verbal constructions often semantically resemble complex predicates (or compound verbs), in which the first verb is usually semantically primary, with the second verb (also called vector verb or explicator verb) providing fine distinctions as well as TAM markers, a definition that as a rule works well for Afroasiatic. One finds coordinated constructions joined by a conjunction (syndetic), typically enclitic -ma, as well as asyndetic con-

9 Within Assyriology, cf. notably Kraus (1987), Buccellati (1996:377 ff.), Wasserman (2002:17 ff.), and Huehnergard (2005:125 f.). 10 Cf th t ib ti i J h & Z i k (1990)

The Finite–Infinite Dichotomy

217

structions of this type. To begin with Akkadian, a syndetic example is shown in (11): Syndetic Serial Verb Constructions in Akkadian (11) atūr-ma wardam ana bēli-ya return.PRET.1SG-CONJ

slave.ACC

to

lord.GEN-my

‘I sent the slave to my lord again.’

aṭrud send.PRET.1SG

(Huehnergard 2005:125 f.)

Examples with an asyndetic juncture include the following: Asyndetic Serial Verb Constructions in Akkadian (12) šutebrī šululī ina ibrātim remain.IMP.SG.F

rejoice.IMP.SG.F

in

‘Keep on rejoicing in the shrines!’

shrines

(Wasserman 2002:19 f.)

Wasserman (2002:19 ff.) identifies the following semantic features of the verbal hendiadys/serial verb construction in this context: (1) Duration of action (imperfective – durative/habitative); (2) Repetition of action (imperfective – iterative); (3) Execution of action: speed, total fulfillment (perfective – punctual/terminative); (4) Scope and extent of action: excessiveness; (5) Motivation of agent towards action: willingness or capability; and (6) Multiple action A typical litmus test for the status of verbal hendiadys obviously consists in replacing the first verb in such constructions by an adverb in a European translation. Comparable constructions are also attested in Biblical Hebrew (cf., e.g. Gesenius 1910:§120). Examples for syndetic constructions are given in (13): Syndetic Serial Verb Constructions in Biblical Hebrew (13) a. way-yāšūḇū way-yiḇkū gam bənē CONJ-return.PRET.3PL.M

CONJ-cry.PRET.3PL.M

yiśrāʾēl

also children_of Israel

‘And the Israelites wept again.’ (Num. 11:4) b. way-yōseˉp CONJ-add.PRET.3SG.M

ʾaḇrāhām

way-yiqqaḥ

Abraham

CONJ-take.PRET.3SG.M

‘And Abraham took once again a wife.’

ʾiššā(h) wife

(Gen. 25:1)

218

Lutz Edzard

Asyndetic examples are also attested, mainly chains of imperatives, as in (14): Asyndetic Serial Verb Constructions in Biblical Hebrew (14) šūḇ šəḵāḇ return.IMP.SG.M

lie_down.IMP.SG.M

‘Lie down again!’

(1 Sam. 3:5)

In Arabic dialects, one also encounters several construction types, where two verbs in the same tense are juxtaposed: Asyndetic Serial Verb Constructions in Arabic Dialects (15) rigiʿ hirib tāni return.PRF.3SG.M

flee.PRF.3SG.M

‘He fled a second time.’

second_time

(Woidich 2002:128)

While Fischer (2002) in this context argues for an underlying [XPRF wa-YPRF] structure, Woidich (2002) prefers an explanation in terms of analogy to asyndetic [XPRF YIPF] structures in Arabic, e.g. ǧāʾa yaḍḥaku ‘he came laughing’ (lit. ‘he came – he laughs’). In this context, Versteegh (2009) distinguishes three kinds of verbs in the first position: (1) Verbs expressing Aktionsart (as in (15) above rigiʿ ‘he returned’); (2) Verbs expressing motion or posture (e.g. ʾām ‘he got up’ or rāḥ ‘he went away’ – the latter also occurs in second position, i.e. according to the classical definition of serial verbs); and (3) Certain other fixed expressions The second kind is especially interesting for our purposes, as the verb ʾām ‘he got up’ in first position can be frozen in the 3SG.M.11 This amounts to a process of grammaticalization (or: deranking) towards a non-inflected (or: de-finitized) particle meaning ‘then’; cf. (16): Grammaticalization of qāma in Arabic Dialects (16) a. ʾām inta ʿamalti ʾē then

you.SG.M

do.PRF.2SG.M

‘What did you do?’

11 R

di

thi

h

what

(Woidich 2002:148)

f l

Pi

t (2002)

220

Lutz Edzard

ited number of serial constructions like German and several Afroasiatic languages mentioned in the present paper, in which two or more contiguous finite verbs express a single event. Both types of constructions allow for the expression of logically subordinate events or the expression of notions that would typically be rendered by adverbs in European languages. And in both types one can observe the loss of inflection (or, if one so pleases, finiteness) by means of a process of grammaticalization towards an adverb (converb) and towards a particle (serial verb), respectively. In this context, the functioning of the Ethio-Semitic converb is fascinating insofar as the semantically related gerund in its classical definition designates an infinite verb form. Abbreviations 1, 2, 3 First, second, third person ABL Ablative ACC Accusative ANT Anteriority C Common gender CAUS Causative CNV Converb CONJ Coordinating conjunction DS Different subject F Feminine GEN Genitive IDPH Ideophone IMP Imperative INF Infinitive, verbal noun IPFV Imperfective

JUS M NEG NOM OJ OBL PASS PFV PL PRET PRF PST SG TAM

Jussive Masculine Negation Nominative Object Oblique Passive Perfective Plural Preterite Perfect Past Singular Tempus, aspect, mood

References Appleyard, David. 2012. Cushitic. In Lutz Edzard (ed.), Semitic and Afroasiatic: Challenges and Opportunities, 199–295. Wiesbaden: Harrassowitz. Azeb, Amha & Gerrit Dimmendaal. 2006. Converbs from an African perspective. In Felix K. Ameka, Alan Dench & Nicholas Evans (eds.), Catching Language: Issues in Grammar Writing, 393–440. Berlin: Mouton de Gruyter.

The Finite–Infinite Dichotomy

221

Buccellati, Giorgio. 1996. A Structural Grammar of Babylonian. Wiesbaden: Harrassowitz. Chrzanowski, Jaroslaw. 2011. Verbal Hendiadys Revisited: Grammaticalization and Auxiliation in Biblical Hebrew Verbs. Washington, DC: The Catholic University of America (PhD Dissertation). Cohen, David. 1984. La phrase nominale et l’évolution du système verbal en sémitique. Paris: Société Linguistique de Paris. Diem, Werner. 1997. Suffixkonjugation und Subjektspronomina. Ein Beitrag zur Rekonstruktion des Ursemitischen und zur Geschichte der Semitistik. Zeitschrift der Deutschen Morgenländischen Gesellschaft 147(1). 10–76. Diem, Werner. 2012. Vom Status pendens zum Satzsubjekt. Studien zur Topikalisierung in neueren semitischen Sprachen. Wiesbaden: Harrassowitz. Dillmann, August. 1899. Grammatik der äthiopischen Sprache. Leipzig: Tauchnitz. Fischer, Wolfdietrich. 2002. Unterordnende und nebenordnende Verbalkomposita in den neuarabischen Dialekten und im Schriftarabischen. In Werner Arnold & Hartmut Bobzin (eds.), “Sprich doch mit deinen Knechten aramäisch, wir verstehen es!” 60 Beiträge zur Semitistik. Festschrift für Otto Jastrow zum 60. Geburtstag, 147–164. Wiesbaden: Harrassowitz. Gesenius, Wilhelm. 1910. Gesenius’ Hebrew Grammar as Edited by the Late E. Kautzsch. Second English Edition Revised in Accordance with the 28th German Edition (1909) by A. E. Cowley. Oxford: Clarendon Press. Hetzron, Robert. 1972. Ethiopian Semitic: Studies in Classification. Manchester: Manchester University Press. Hetzron, Robert. 1975. Genetic classification and Ethiopian Semitic. In James Bynon & Theodora Bynon (eds.), Hamito-Semitica: Proceedings of a Colloquium Held by the Historical Section of the Linguistics Association (Great Britain) at the SOAS, University of London, 18–20 March 1970, 103–128. The Hague: Mouton. Huehnergard, John. 2005. A Grammar of Akkadian. Winona Lake, IN: Eisenbrauns.

222

Lutz Edzard

Jastrow, Otto. 2011. Ṭuroyo and Mlaḥsô. In Stefan Weninger (ed.), The Semitic Languages: An International Handbook, 697–707. Berlin: De Gruyter Mouton. Joseph, Brian D. & Arnold M. Zwicky (eds.). 1990. When Verbs Collide: Papers from the 1990 Ohio State Mini-Conference on Serial Verbs. Columbus: Ohio State University, Department of Linguistics. Kapeliuk, Olga. 1997. Reflections on the Ethio-Semitic gerund. In Katsuyoshi Fukui, Eisei Kurimoto & Masayoshi Shigeta (eds.), Ethiopia in a Broader Perspective: Papers of the 13th International Conference of Ethiopian Studies, Kyoto, 12–17 December 1997, vol. I, 492–498. Kyoto: Shokado Book Sellers. Kraus, F. R. 1987. Sonderformen akkadischer Parataxe: Die Koppelungen. (Mededelingen der Koninklijke Nederlandse Akademie van Wetenschappen, Afd. Letterkunde, Nieuwe Reeks, Deel 50, No. 1). Amsterdam: North Holland. Lambdin, Thomas. 1978. Introduction to Classical Ethiopic (Geʿez). Missoula, MT: Scholars Press. Leslau, Wolf. 1995. Reference Grammar of Amharic. Wiesbaden: Harrassowitz. Lipiński, Edward. 2001. Semitic Languages: Outline of a Comparative Grammar. Leuven: Peeters. Lipiński, Edward. 2010. Le gérondif en phénicien. Journal of Semitic Studies 55(1). 1–10. Montgomery, James A. 1928. Origin of the Ethiopic gerund. Journal of the American Oriental Society 48. 283–284. Newman, Paul. 2007. Hausa. In Alan S. Kaye (ed.), Morphologies of Asia and Africa, vol. I, 677–711. Winona Lake, IN: Eisenbrauns. Nikolaeva, Irina (ed.). 2007. Finiteness: Theoretical and Empirical Foundations. Oxford: Oxford University Press. Piamenta, Moshe. 2002. Fossilized and semi-fossilized verbs in Jerusalem Arabic. In Werner Arnold & Hartmut Bobzin (eds.), “Sprich doch mit deinen Knechten aramäisch, wir verstehen es!” 60 Beiträge zur Semitistik. Festschrift für Otto Jastrow zum 60. Geburtstag, 531– 539. Wiesbaden: Harrassowitz. Rubin, Aaron D. 2005. Studies in Semitic Grammaticalization. Winona Lake, IN: Eisenbrauns.

The Finite–Infinite Dichotomy

223

Shisha-Halevy, Ariel. 2009. A note on converbs in Egyptian and Coptic. In Charles G. Häberl (ed.), Afroasiatic Studies in Memory of Robert Hetzron, 95–105. Newcastle upon Tyne, UK: Cambridge Scholars Publishing. Tropper, Josef. 2002. Altäthiopisch. Grammatik des Geʿez mit Übungstexten und Glossar. Münster: Ugarit. Versteegh, Kees. 2003–2005. Some remarks on verbal serialization in Arabic dialects. In Georgine Ayoub & Jérôme Lentin (eds.), Linguistique arabe, 49–69. (Cahiers de Linguistique de l’INALCO 2003–2005 5). Paris: Cercle de Linguistique de l’INALCO. Versteegh, Kees. 2009. Serial verbs. In Kees Versteegh (ed.), Encyclopedia of Arabic Language and Linguistics, vol. IV, 195–199. Leiden: Brill. Wakasa, Motomichi. 2001. Gerund in Amharic. Tokyo University Linguistic Papers (TULIP) 20. 425–451. Wasserman, Nathan. 2002. Style and Form in Old-Babylonian Literary Texts. Leiden, Boston & Cologne: Brill/Styx. Weninger, Stefan. 2011. Old Ethiopic. In Stefan Weninger (ed.), The Semitic Languages: An International Handbook, 1124–1142. Berlin: De Gruyter Mouton. Woidich, Manfred. 2002. Verbalphrasen mit asyndetischem Perfekt im Ägyptisch-Arabischen. Estudios de Dialectologia Norteafricana y Andalusí 6. 121–192. Yri, Kjell Magne. 2012. Sidaama. In Lutz Edzard (ed.), Semitic and Afroasiatic: Challenges and Opportunities, 262–277. Wiesbaden: Harrassowitz. Zaborski, Andrzej. 2005. Tense, aspect and mood categories in Proto-Semitic. In Lutz Edzard & Jan Retzö (eds.), Current Issues in the Analysis of Semitic Grammar and Lexicon I, 11–30. Wiesbaden: Harrassowitz.

Finiteness in Gurage Languages Ronny Meyer Addis Ababa University Abstract Despite their geographical adjacency and genetic relatedness, the various Gurage languages display remarkable variation in marking verbs in independent main clauses vis-à-vis dependent clauses, or, in other words, in distinguishing between finite and nonfinite verbs. A detailed analysis of inflectional features of verbs (marking of aspect/mood, subject indexing, negation, tense, and clausal status) and a comparison of inflectional features of verbs in dependent and independent clauses show that finiteness in Gurage languages is a relational category that can best be described through a combination of morphological and syntactic features. 1 Introduction According to Hetzron (1972a:6 f.), Gurage is best understood as a geographical term for the southernmost pocket of Semitic-speaking peoples in central Ethiopia who are encircled by Cushitic- and Omotic-speaking peoples, as shown in Map 3. Older classifications consider Gurage to be a single language with different dialects (cf. Meyer 2011; Appleyard 2002a). However, Hetzron (1972a; 1977) provides clear evidence for separate Gurage languages belonging to different genetic sub-branches within South Ethio-Semitic (SES) (see Figure 2). The Gurage languages form two major groups: Gunnän Gurage with Northern Gurage (Gälila, Kistane, Dobbi, Muher) and Western Gurage (Mesqan, Central (CWG) and Peripheral (PWG) Western Gurage) vs. Eastern Gurage with Silt’e, Wolane and Zay. Gunnän Gurage is spoken in a relatively contiguous area, while Eastern Gurage is dispersed across the Gurage area. Hetzron’s classification of Ethio-Semitic (ES) is not commonly accepted (cf. Voigt 2009) but, currently, it is “[t]he only really methodologically sound and indeed the definite internal classification of [Ethio-Semitic]” (Appleyard 2002a:409). The classification of the Gurage varieties, which are part of ES, is impeded by the lack of data for some varieties and intense multilingualism among the speakers.

226

Ronny Meyer

Map 3: Geographical Distribution of the Gurage Languages

Figure 2: Genetic Classification of the Gurage Languages

(Adopted from Hetzron 1972a:119; Meyer 2011:1222)

Finiteness in Gurage Languages

227

Due to reasons related to general structural features, available grammatical descriptions and intelligibility,1 I focus in this paper on the Gunnän Gurage languages Chaha (as representative of CWG), Inor (as representative of PWG), Kistane, Mesqan, Muher, and on the Eastern Gurage languages Wolane and Zay.2 2 Finiteness as a linguistic concept The concept of finiteness, traditionally the marking of tense, aspect, mood (TAM), and subject indexing on verbs, goes back to the School Grammar tradition (Nikolaeva 2007:1 ff.). If a verb is marked for any of these categories, it is considered finite; if it lacks them it is nonfinite. From a syntactic perspective, finiteness is used to distinguish between dependent and independent clauses (ICL) in such a way that nonfinite verbs occur in dependent clauses but finite verbs in independent ones. However, cross-linguistically, morphological TAM marking and subject indexes are by no means universal, and nonfinite verbs may also occur in independent clauses. Other points of discord are whether finiteness is a phrasal category connected only with the verb or a clausal category, or whether it is a binary or scalar phenomenon. Recent approaches to finiteness (Anderson 2007; Klein 2006; Maas 2004; etc.) view finiteness as a complex phenomenon that can be defined by a combination of overt morphosyntactic constructions and covert pragmatic/semantic functions. Generally, finiteness entails an independent interpretation of a clause by anchoring the communicated verbal event to a specific time, by establishing the referents of the participants, and by asserting the relationship between the event and the participants (Klein 2006:263 ff.). Consequently, the prototypical finite clause is a simple assertion as found in affirmative, declarative statements in the indicative mood (Anderson 2007:2). Other clause types may lack some of these features but still remain finite.

1

2

Intelligibility between Eastern and Gunnän Gurage is very low. Within Eastern Gurage, Wolane and Silt’e are very similar while Zay diverges. As to Gunnän Gurage, Mesqan is best understood by speakers of all other languages (Fekede 2013:312). Muher and CWG are fairly intelligible to speakers of Kistane and Mesqan whereas PWG diverges from all (cf. Hetzron 1972a:2; Gutt 1980; Ahland 2010:2 ff.). The data is obtained from the following sources if not mentioned otherwise: Kistane (Leslau 1968a; Goldenberg 1968; Elizabeth 2012), Mesqan (Leslau 2004; Meseret 2012), Chaha (Leslau 1983; 2004), Inor (Berhanu & Hetzron 2000; Leslau 1983), Wolane (Meyer 2006a), and Zay (Meyer 2005). The Muher data result from my own field research with Abubakr Sherifo and Sitti Gragn to whom I am very grateful.

228

Ronny Meyer

The description of finiteness in Gurage languages will be based on Bisang (2007), in which a distributional approach based on obligatory marking of certain morphosyntactic categories in dependent or independent clauses was followed. Finiteness is, then, a property of the clause, not the verb (Bisang 2007:134). TAM and subject indexes play an important role for defining finiteness but other categories, like illocutionary force, i.e. assertion, interrogation, etc., and politeness, can also be used to define the independent status of a clause. In this approach, finiteness appears to be a binary phenomenon in individual languages but a scalar phenomenon across languages because not every language uses exactly the same morphosyntactic categories (cf. also Nikolaeva 2007:13). Finiteness is usually discussed as a verbal feature in individual grammars of Ethio-Semitic languages. According to Appleyard (2002b), there is a clear morphological distinction between main vs. subordinate verbs in various Ethio-Semitic languages, whereby verbs in main clause are formed by attaching additional morphemes (for finiteness) to subordinate verbs. This distinction, which constitutes one feature of the Ethiopian linguistic area, is commonly ascribed to Cushitic influence. Appleyard (2002b) distinguishes between three types of main verb markers in Ethio-Semitic: (i) no marker (for North Ethio-Semitic), (ii) the present tense auxiliary (for Transversal South Ethio-Semitic), and (iii) a main verb marker derived from an older copula (for several language of Outer South Ethio-Semitic). This typology fits perfectly Hetzron’s (1972b) genetic classification of Ethio-Semitic (cf. Figure 2 above) but does not faithfully display the complexity of finiteness in Gurage languages. Although marking verbs in ICL is a prominent feature in these languages, there are also other more subtle indicators for finiteness, like specific subject indexes or markers for negation. Often finiteness cannot be ascribed to a specific morphological category but only through paradigmatic relations between them. 3 Inflectional features of verbs 3.1 General remarks All Gurage languages belong to the synthetic-fusional type. Verbs are commonly formed by non-concatenative morphology and consist of three obligatory morphemes: (i) a lexical root (prototypically containing three or four consonants) that interdigitates with (ii) a grammatical template for aspect/mood, i.e. a vocalic prosody that also contains information about the gemination or non-gemination of root consonants, and (iii) a subject index (cf. Table 24).

Finiteness in Gurage Languages

229

Table 24: Formation of Type A Verbs in Mesqan {ROOT √s1b2r3 ‘BREAK’

+ TEMPLATE} PFV C1äCC2äC3IPFV -C1äC2C3SBJV -C1C2C3-

> BASE s1äbb2är3-s1äb2r3-s1b2r3-

VERB (SJ 3PL.M) säbbär-o ‘they broke’ jɨ-säbr-o ‘they break’ jä-sbɨr-o ‘they may break’

A simplex verb may belong to one out of three lexically determined conjugation patterns called (Verb) Type A, B, C. I will restrict my description to verbs of Type A because they are the unmarked type and neatly distinguish between the various aspect/mood forms by separate templates. All Gurage languages distinguish between verb bases for the imperfective aspect, the perfective aspect, and the subjunctive, which encompasses the jussive, imperative and verbal noun (cf. Section 3.2). These bases still need to be specified by subject indexes to form an independent verb. There are two different subject indexes: a suffix subject index for perfective verbs, and a pre-/circumfix subject index for imperfective and jussive verbs (cf. Section 3.3). Usually, the prefix in the latter set denotes person while the suffix marks gender and number.3 Additional morphemes found with verbs in Gurage languages are negative markers (cf. Section 3.4), temporal auxiliary verbs (cf. Section 3.5.1), and markers of clausal status (cf. Section 3.5.2). Object indexes and derivational affixes are also part of the verb morphology but will not be discussed here (cf. Meyer 2011:1239 f. for an overview). 3.2 Templates for aspect and mood Type A verbs in Gurage have minimally three and maximally five templates to form the perfective, imperfective, or subjunctive, as shown in Table 25. With the exception of the negative perfective in Zay whose template is that of the subjunctive, all three verb bases, perfective, imperfective and subjunctive, are always clearly distinguished by a specific vocalization of C2, i.e. no vowel or C2ä in the subjunctive, äC2 in the imperfective but ä(C)C2ä in the perfective. The imperfective template has the most regular pattern across all languages while the perfective and subjunctive templates have two patterns each in some languages.

3

For objections against this generalization, cf. Rose (1996a:339); for a unifying analysis of the two indexes, cf. Degif (2000:241 ff.).

230

Ronny Meyer

Table 25: Conjugational Templates of Type A Verbs PERFECTIVE

IPFV

AFFIRMATIVE NEGATIVE

EASTERN GURAGE Wolane C1äC2äC3 C1äC2äC3 C1C2äC3 ZAY GUNNÄN GURAGE Kistane C1äCC2äC3 C1äC2äC3 Mesqan, Muher C1äCC2äC3 C1äC2äC3 Chaha, Inor C1ä(C)C2äC3

SUBJUNCTIVE TRANSITIVE INTRANSITIVE

C1äC2C3 C1äC2C3

C1C2äC3 C1C2äC3

C1äC2C3 C1äC2C3 C1äC2C3

C1C2(ä)C3 C 1C 2C 3 C1C2äC3 C 1C 2C 3 C1C2äC3

The subjunctive in Gunnän Gurage, except Kistane, morphologically distinguishes between transitive and intransitive verbs, whereas there is only one template in Eastern Gurage. Kistane merges the two transitivity patterns into a single irregular conjugation in the subjunctive: The intransitive template C1C2äC3 occurs in the negative, in the formation of verbal nouns, and with 3SG.F, 1PL and second person subjects in the affirmative whereas the transitive pattern C1C2C3 is used with the remaining subjects (cf. Leslau 1968a:22; Goldenberg 1968:95 f.). The subjunctive template is also used to form verbal nouns by attaching to it the prefix wä-, the suffix -ot, or a combination of them (Hetzron 1977:110). Table 26: Formation of Verbal Nouns SUBJUNCTIVE TRANSITIVE

Wolane Zay, Kistane Mesqan, Muher Chaha Inor

INTRANSITIVE

C1C2äC3-ot wä-C1C2äC3 wä-C1C2äC3 wä-C1C2C3 wä-C1C2äC3 wä-C1C2C3 C1C2C3-ot C1C2äC3-ot ä-C1C2C3-ʷ/ʲ-t ä-C1C2äC3-ʷ/ʲ-t

The verbal noun in Chaha can be optionally formed by using either affix while the Inor verbal noun applies a combination of the two affixes in a phonologically reduced form, whereby the suffix -t is preceded by the non-segmental feature -ʷʲ that triggers labialization and palatalization of root consonants, i.e. Inor *√kft ‘open’ > ä-kufʷʧ-t ‘open’ (Leslau 1983:18). Gemination of C2 as part of the perfective template with Type A verbs is realized phonetically in two different ways in Gunnän Gurage: lengthening or

Finiteness in Gurage Languages

231

devoicing. Lengthening of C2 occurs in Kistane, Muher, and Mesqan.4 These languages are spoken in an almost contiguous area in the north-central part of the Gurage area (cf. Map 3). In the remaining Gunnän Gurage languages except Ezha, gemination is phonetically realized by devoicing underlying voiced consonants in a position that is geminated in the above-mentioned languages, as in Chaha säpärä-m ( *le- >) jä-.12 In Eastern Gurage, the vowel ä of the person prefix was further extended to all third and even 10 The morphophonological changes in the verb triggered by the subject indexes in Inor are not well described (cf. Berhanu & Hetzron 2000:44 ff.). Generally, the indexes for 2SG.F and 2/3PL may labialize or palatalize consonants in the verb base; the final diphthong -ua is sometimes omitted. 11 Hetzron (1977:79) considers *lä- a dative prefix while Wagner (1968:211) relates it to the subjunctive particle la- in Gəʿəz or li- in Arabic (cf. also Lipiński 1997:336). As to Huehnergard (1983:592), the prefix *la- can be reconstructed for Proto-Semitic (PS) where it was used to emphasize the subject of clauses expressing an intention, volition or command. 12 Leslau (1968b) and also Lipiński (1997:368, 373), in contrast, assume that the vowel ä of the subject prefix represents a PS feature.

Finiteness in Gurage Languages

235

first person subjects (Wagner 1968:112), whereby in the first person, the combination of *lä- and original 1SG subject prefix *ʔ- yielded lä-. The 1SG nä- in Kistane is probably related to lä- in Eastern Gurage. Wolane uses a special subject index with negated imperfective verbs in ICL, where the negative marker and the subject index fused into a unique prefix, which clearly distinguishes this clause type from all others (cf. Section 3.4).13 In all other Gurage languages, the subject index used in the negated subjunctive also occurs with affirmative and negative subordinate and negated imperfective verbs in ICL. This might be due to the reason that the subjunctive was used as a subordinate verb form at an earlier stage in these languages, as observable in Gəʿəz (Tropper 2002:90, 192). The subject index of the negated subjunctive also occurs in the affirmative imperfective in ICL, with the exception of the first person subjects. All Gunnän Gurage languages use the prefix ä- to mark 1SG subjects with affirmative imperfective verbs in ICL while the 1SG subject prefix n- is used in all other clause types. Only in Kistane the 1SG subject prefix ä- also occurs in negated ICL and in affirmative subordinate clauses.14 In Eastern Gurage, the first person subjects are marked by the prefix j- in ICL with an affirmative imperfective verb but with the prefix l- in all other clause types. While the first person subject prefixes l(ä)- and n(ä)- result from the fusion of the subjunctive particle *lä- with the older subject prefix,15 the origin of the prefixes ä- and j- for first person subjects remains unclear. To summarize, there is no straightforward relation between subject index and finiteness. The only forms that are clearly mark verbs in ICL are the affirmative imperative, and the affirmative jussive. Further, all Gurage languages have a unique prefix for the affirmative imperfective with 1SG subjects in ICL, which was extended to 1PL subjects in Eastern Gurage. Except Zay, Eastern Gurage languages have a unique subject index for the negative imperfective in ICL. All other subject indexes occur with verbs in dependent and independent clauses.

13 A similar negation for the imperfective in ICL is also found in Silt’e (cf. Gutt 1997:923). 14 The 1SG subject prefix with subordinate verbs is given as ä- for the affirmative but nfor verbs in negative relative clauses (Leslau 1968a:20). 15 Bulakh & Kogan (2010:279) consider the 1SG subject prefix l- ~ n- an innovation in SES, which might not have its origin in the jussive paradigm.

236

Ronny Meyer

3.4 Negation Negation in Gurage is marked on the verb but has scope over the entire clause. Most commonly, it is marked by the negative prefix al-/an- or a-.16 The negative prefix al- occurs with perfective verbs in Eastern Gurage and Kistane but an- in the remaining Gunnän Gurage languages. Table 29: Perfective Verbs in Main Clauses Zay

AFFIRMATIVE säbär-ä-n-u

NEGATIVE al-sɨbär-o [assɨbäro]

break\PFV-3SG.M-FOC-DECL

NEG-break\SBJV-3SG.M.DECL

Wolane

säbär-ä

al-säbär-ä

break\PFV-3SG.M

NEG-break\PFV-3SG.M

säffär-o

al-säfär-ä

break\PFV-3SG.M.MVM

NEG-break\NPFV-3SG.M

säbbär-ä

an-säbär-ä

break\PFV-3SG.M

NEG-break\NPFV-3SG.M

säbbär-ä-m

an-säbär-ä

break\PFV-3SG.M-FOC-MVM

NEG-break\NPFV-3SG.M

säpär-ä-m

an-säpär-ä

break\PFV-3SG.M-MVM

NEG-break\PFV-3SG.M

säpäˈr-ä17

ã-säpär-ä-da

break\PFV-3SG.M.MVM

NEG-break\PFV-3SG.M-KTD

‘he broke’

‘he did not break’

Kistane Mesqan Muher Chaha Inor Gloss

In addition to the negative prefix, Zay, Kistane, Mesqan and Muher use different templates for affirmative and negative perfective verbs (cf. Table 25 above). The negative perfective template in Kistane, Mesqan and Muher lacks the gemination of C2.18 Zay, in contrast, uses the subjunctive template but with the suffix subject index. 16 The negative prefix for perfective verbs in SES is commonly reconstructed as *ʔal-, while the negative marker with imperfective/subjunctive verbs might be *ʔal- or *ʔaj(Bulakh 2012:407 f.). Contrary to Leslau (1952:76 fn. 77), among others, I assume that only the prefix (ʔ)a-, probably a reflex of the negative prefix *ʔaj-, is the negative marker for imperfective/subjunctive verbs in Gurage (cf. also Hudson 2003:213 ff.). 17 The apostrophe in the Inor data indicates an unexpected ultima accent, which is due to the diachronic loss of word-final -m (Hetzron 1977:42 f.). 18 Given that the gemination of C2 with perfective verbs is an innovation in SES (Hetzron 1972a:22 ff.), its lack in the negation must represent an archaism (Goldenberg 1977:487, fn. 123).

Finiteness in Gurage Languages

237

The negative marker a- plus the 3SG.M subject prefix j-, merge to e- in most Gunnän Gurage languages but yield aj- in Eastern Gurage and Inor. As mentioned in Section 3.3, negative imperfective verbs with 1SG subjects use prefixes different from those in the corresponding affirmative forms; negative jussive verbs lack the vowel ä of the corresponding affirmative indexes. Table 30: Imperfective ICL Verbs in the Present Tense AFFIRMATIVE

NEGATIVE EASTERN GURAGE a-j-säbr-u

Zay

jɨ-säbr-ɨn-aa

3SG.M-break\IPFV-FOC-AUX.PRS.DECL

NEG-3SG.M-break\IPFV-DECL

Wolane

jɨ-säbr-an

ʔɨlɨ-säbɨr

3SG.M-break\IPFV-AUX.PRS

NEG.3SG.M[ICL]-break\IPFV

GUNNÄN GURAGE t-i-säbɨr

Kistane

jɨ-säbr-u

3SG.M-break\IPFV-MVM

NEG[ICL]-3SG.M-break\IPFV

Muher

jɨ-säbr-u

e-säbɨr

3SG.M-break\IPFV-MVM

NEG.3SG.M-break\IPFV

jɨ-säbɨr

e-säbɨr

3SG.M-break\IPFV

NEG.3SG.M-break\IPFV

jɨ-säbɨr

e-säbɨr

3SG.M-break\IPFV

NEG.3SG.M-break\IPFV

jɨ-säbɨr

a-j-säbɨr-ka

3SG.M-break\IPFV[ICL]

NEG-3SG.M-break\IPFV-KTD[ICL]

‘he breaks’

‘he does not break’

Mesqan Chaha Inor Gloss

The negative prefixes t- in Kistane19 and ʔl- in Wolane20 occur only with imperfective verbs in ICL, which, thus, become distinct from all other clause 19 Beside Kistane, the negative prefix t- also occurs in Gafat (Leslau 1945:68), but not in Zay as wrongly indicated in Leslau (1952:76), Hudson (2003:215) and probably elsewhere. The negative marker t- is mainly considered a borrowing from Sidaama (Hetzron 1972a:97; Leslau 1952:76 f.). Other suggestions are an origin as phonetic augment (Leslau 1945:68) or a relationship to the Egyptian negative element tm (Lipiński 1997:455). 20 Hetzron (1972a:28) relates ʔl- to the common SES negative marker *ʔal-, which is not sound. Despite the uncommon sound change *ʔal- to ʔɨl- and the deletion of the subject prefix j- in the 3SG.M and 3PL (cf. Table 28), the negative marker for the 1SG in Silt’e has an additional w, as in ilaw-nakt ‘I will not beat’ (Gutt 1997:923). This w does not occur in Wolane but probably left a trace in the gemination of the l in the negative pre-

238

Ronny Meyer

types. In Inor, the negated imperfective in ICL is marked by the KTD-suffix,21 which makes it identifiable as an ICL verb. In Zay, affirmative jussive, and affirmative and negative indicative verbs in ICL are marked by the declarative marker (cf. Section 3.5.2.2). Negative ICL in Kistane and Muher lack the main verb marker (cf. Section 3.5.2.1). Further, in Eastern Gurage, affirmative imperfective verbs in ICL are obligatorily marked by a temporal auxiliary (cf. Section 3.5.1). In negation, however, the present tense auxiliary is always lacking. An ICL with a non-present imperfective verb is ambiguous. It can denote a durative/habitual in the past or an unreal consequence. In negation, however, these two meanings are formally distinguished in the Eastern Gurage languages Zay and Silt’e. The negated non-present durative/habitual in Zay is formed by an affirmative imperfective verb followed by the negated nonpresent auxiliary naar, as in (1a).22 The negated unreal consequence in Zay (and Silt’e), as in (1b), consists of the negated perfective verb (which occurs with the subjunctive template in Zay) preceded by the prefix l-, which marks a real protasis in other contexts (Meyer 2005:285; Gutt 1997:950). (1) a. jɨ-näk’ɨl 1SG-take\IPFV

al-naar-uh

ZAY

NEG-AUX.NPRS-1SG

‘I was not taking’ (durative) b. l-aal-nɨk’äl-uh-u COND-NEG-take\SBJV-1SG-DECL

‘I would not take (it)’ (unreal) In Wolane, another Eastern Gurage language, and in most Gunnän Gurage languages, however, only one negative form for both uses prevails:23

fix and the first radical of the verb, as in ʔɨllɨ-ssäbɨr ‘I do not break’ (Meyer 2006a:110). 21 The KTD-suffix represents a set of allomorphs in complementary distribution (cf. Section 3.5.2.3). 22 Silt’e expresses the negated non-present habitual/durative by the negated imperfective followed by the invariable non-present auxiliary naar (Gutt 1997:923). 23 According to Hetzron (1977:87), the non-present imperfective in ICL in Gunnän Gurage is negated uniformly by using the negated imperfective preceded by b-. At least for Kistane this seems not to be the case because Goldenberg (1968:96) provides an example which follows the negated habitual/durative in Silt’e.

Finiteness in Gurage Languages (2) a. b-a-j-säbɨr

239 WOLANE

COND-NEG-3SG.M-break\IPFV

b. b-e-säbɨr

MUHER

COND-NEG.3SG.M-break\IPFV

i. ‘he did not use to break’ ii. ‘he would not break’ In Gunnän Gurage language, as shown for Muher in (2b), the negated nonpresent imperfective in ICL consists of the negated imperfective to which the prefix b- is attached, which synchronically functions as subordinating conjunction to mark temporal or conditional clauses. As for Wolane in (2a), instead of the ICL negation, the negation for subordinate imperfective verbs is used in this case (cf. Table 28). A separate prohibitive construction exists in Muher, CWG and PWG.24 This construction consists of the invariable prefix ɨn-25 followed by the inflected negative perfective (recall Table 25) (plus main verb markers in Muher): (3) a. ɨnkäfätho

MUHER

ɨn-käfät-hä-u PROH-open\NPFV-2SG.M-MVM

b. ɨn-käfät-xä

INOR

PROH-open\NPFV-2SG.M

‘Don’t open (it)!’ At least for Muher, the prohibitive is exceptional because it co-occurs with the main verb marker that is otherwise restricted to affirmative imperfective verbs in ICL (cf. Section 3.5.2.1). Most negated verbs are marked by the same negative prefixes in dependent and independent clauses. Further, the negated unreal consequence in all Gurage languages, and also the negated non-present durative/habitual in Wolane and Gunnän Gurage, is formed by a construction that resembles a subordinate verb due to the prefixed conditional markers l- or b-. Only in the Eastern Gurage languages Wolane and Silt’e negated imperfective verbs in 24 Contrary to Hetzron (1977:88), the prohibitive construction could not be attested in Mesqan. 25 Leslau (1969a) considers the prefix ɨn- an archaic PS negative marker (cf. also Lipiński 1997:456 f.).

240

Ronny Meyer

ICL form a unique paradigm, which sets them apart from all other clause types. In Gunnän Gurage except Mesqan, a uniquely marked prohibitive construction occurs whose occurrence is restricted to ICL. 3.5 Additional compulsory morphemes Besides the obligatorily marking of aspect/mood, subject indexes, and polarity, there are some recurrent morphemes attached to verbs in Gurage languages that occur only in specific clause types in some of the languages. These morphemes can be divided into two broad groups: tense markers and markers of clausal status. 3.5.1 Tense Affirmative imperfective verbs in ICL in all Gurage languages distinguish between present and non-present events by temporal auxiliary verbs,26 whereby their usage patterns vary across the languages. The most common temporal auxiliaries grammaticalized from the verbs *√hlw ‘exist (present)’ (Hetzron 1972a:18 f.; Goldenberg 1977:494 ff.) and *√nbr ‘(a) live, (b) exist (non-present)’. For the non-present tense, another auxiliary based on the element *bannä of unclear origin is used in some Gunnän Gurage languages (Hetzron 1972a:65 ff.; 1977:106 ff.). The distribution of these auxiliaries and other morphemes co-occurring with affirmative imperfective or subjunctive verbs in ICL is summarized in Table 31.27 The non-present auxiliaries based on *√nbr or *bannä usually occur in an invariable form in ICL, while the present auxiliary, which agrees with the subject in most persons, is restricted to Eastern Gurage. The auxiliary *√hlw underwent stark phonological reduction in Eastern Gurage where it is now a clitic or suffix (Hetzron 1972a:40).

26 Temporal auxiliaries also occur with perfective verbs in ICL. In this case, however, the perfective verb is usually marked as converb. Exceptions, in which a temporal auxiliary and a negated perfective verb are juxtaposed, are reported for Kistane (Goldenberg 1968:96) and Silt’e (Gutt 1997:920). These constructions as well as tense distinctions in dependent clauses will not be considered. 27 The Gurage situation fits Pettersson’s (1994) analysis of tense as a binary category: present (event occurs at the moment of speech) vs. non-present (event does not occur at the moment of speech, i.e. in the past or in the future). This distribution accounts for the use of the non-present auxiliary in temporal and modal constructions.

Finiteness in Gurage Languages

241

Table 31: Additional Morphemes with Imperfective Verbs in ICL

Zay Wolane

NON-PRESENT

PRESENT

EASTERN GURAGE naar-AGR-DECL IPFV närINV

IPFV-[aa-AGR]-DECL

IPFV

IPFV-[a-AGR]

28

29

GUNNÄN GURAGE Kistane Muher Mesqan Chaha Inor

IPFV

IPFV-MVM

IPFV

näbbärINV bannä-MVMINV IPFV bannäINV IPFV [banä ~ ba]INV, SBJV-ʃä, IPFV-KTD IPFV [baanä ~ baan]INV-KTD, SBJV-ˈse, IPFV-KTD

IPFV-MVM IPFV IPFV IPFV

Chaha (CWG) and Inor (PWG) make temporal/modal distinctions in ICL, which are absent in the remaining Gurage languages. Beside a plain imperfective denoting an ongoing situation in ICL, these languages also use an imperfective verb plus the KTD-suffix (see 3.5.2.3) to denote a definite future while the combination subjunctive plus invariable auxiliary -ʃä / -ˈse ‘want’ marks an indefinite future (Hetzron 1977:85 f.). (4) a. jɨ-säbɨr

CHAHA

3SG.M-break\IPFV

‘he breaks/is breaking’ b. jɨ-säbɨr-te

c. jɨ-sbɨr-ʃä

3SG.M-break\IPFV-KDT

3SG.M-break\SBJV-AUX.FUT

‘he will certainly break’

‘he may break’

d. jɨ-säbɨr

INOR

3SG.M-break\IPFV

‘he breaks/is breaking’ e. jɨ-säbɨr-kʷe

f. jɨ-sbɨr-ˈse

3SG.M-break\IPFV-KDT

3SG.M-break\SBJV-AUX.FUT

‘he will certainly break’

‘he may break’

28 In the 3SG.M and 1PL, the auxiliary is optionally -äl(ä) (Meyer 2005:159 f.). 29 In the 3SG.M, 1PL, and 3PL, the present auxiliary occurs invariably as -an (Meyer 2006a:96 f.).

242

Ronny Meyer

The use of temporal auxiliaries with affirmative imperfective verbs in ICL is obligatory in Eastern Gurage (cf. Hetzron 1972a:38 ff.). In negation, the tense distinction is retained but not the respective auxiliary verbs (cf. Table 30). Definite and indefinite future constructions as well as constructions with the non-present auxiliary *bannä occur exclusively in ICL, but never in dependent clauses. These constructions are, thus, a clear indicator for finite verbs. The auxiliaries based on *√hlw and *√nbr, however, can co-occur with verbs in independent and dependent clauses, in particular with relative clause verbs (cf. Section 4). 3.5.2 Clausal status Several Gunnän Gurage languages and Zay distinguish between dependent and independent clauses by additional morphemes. Three main strategies can be distinguished: (a) the use of affirmative indicative main verb markers (MVM), (b) the use of declarative markers (DECL) to distinguish between declaration and interrogation,30 and (c) the use of the KTD-suffix. 3.5.2.1 Affirmative indicative main verb marking The Gurage languages Kistane and Muher, but also Dobbi and partly even Gafat, use a set of suffixes consisting of the allomorphs -u, -i, -n/-t (referred to as MVM -u) attached to verbs to mark affirmative indicative ICL (cf. Hetzron 1968:170 f.; 1972a:37 ff.; 1977:88 ff. for details). As these markers clearly distinguish between affirmative independent vs. negated and dependent clauses, Hetzron (1968; etc.) calls them main verb marker. These markers obligatorily occur with perfective and imperfective verbs in Kistane, cf. (5), but usually only with imperfective verbs in Muher. (5) a. alläfo

b. jalfu

alläf-ä-u go\PFV-3SG.M-MVM

j-alf-u 3SG.M-go\IPFV-MVM

‘He went.’

‘He goes/will go.’

KISTANE

Beside imperfective verbs, the main verb marker in Muher is also suffixed to the non-present tense auxiliary in ICL, cf. (6a), and to a perfective verb in the prohibitive construction (cf. (3a) in Section 3.4). Note that the main verb 30 The terms indicative and declarative denote different clause types. Indicative is in contrast to subjunctive and encompasses verbs in the perfective and imperfective aspect. Declarative stands in contrast to interrogative and encompasses indicative and subjunctive verbs in non-interrogative clauses.

Finiteness in Gurage Languages

243

marker and the non-present tense auxiliary are in complementary distribution in Kistane (cf. also Table 31): (6) a. abdi bethut jef banno abdi Abdi

bet-hut house-POSS.3SG.M

MUHER j-ef 3SG.M-go\IPFV

bannä-ä-u AUX.NPRS-3SG.M-MVM

‘Abdi was going to his house.’ b. käbbädä tä-gäbäja-jän Kebede

to-market-CIRC

j-alf

näbbär

3SG.M-go\IPFV

AUX.NPRS

KISTANE

‘Kebede was going towards the market.’ Perfective verbs in affirmative ICL in Muher are obligatorily marked by the invariant suffix -m. (7) bet-hut house-POSS.3SG.M

ef-ä-m

MUHER

go\PFV-3SG.M-MVM

‘He went to his house.’ Chaha (CWG) and Inor (PWG) also mark perfective verbs in ICL by the suffix -m, but imperfective verbs need not carry additional tense or clause markers in ICL. The suffix -m with perfective verbs in ICL is also considered a main verb marker (short MVM -m) due to its functional overlap with the MVM -u of Kistane.31 The MVM -m is most probably related to the converb marker -m in Gunnän Gurage and Zay (cf. Hetzron 1972a:111 ff.). This is suggested by the situation in Kistane, where the suffix -m basically functions as converb marker, as shown in (8a). Converbs and the non-present tense auxiliary form complex predicates that may denote a past perfect in ICL, as in (8b). Recall that the non-present tense auxiliary is in complementary distribution with the MVM in Kistane.

31 In analogy to Eastern Gurage, Rose (1996b:219 ff.) considers the MVM -u and -m to be present and past tense marker, respectively. However, the evidence from Muher, in which the MVM -u combines with the non-present auxiliary, cf. (6a), does not support such an analysis.

244

Ronny Meyer

(8) a. bällam alläfo bälla-ä-m eat\PFV-3SG.M-CNV1

KISTANE alläf-ä-u go\PFV-3SG.M-MVM

‘He ate and went.’ b. alläf-ä-m go\PFV-3SG.M-CNV1

näbbär AUX.NPRS

‘He had gone (away).’ Perfective verbs marked by -m but without the non-present auxiliary also occur in ICL in Kistane where they are in contrast to verbs marked by the MVM -u, as in (9a) denoting a present perfect vs. the prefective in (9b): (9) a. bällam

b. bällo

bälla-ä-m-ø eat\PFV-3SG.M-CNV1-PRS

bälla-ä-u eat\PFV-3SG.M-MVM

‘He has eaten.’

‘He ate.’

KISTANE

The present perfect in (9a) stands in a paradigmatic relationship to the past perfect in (8b). It can be assumed that the converb in (9a) is followed by a zero morpheme functioning as a present tense auxiliary element in Gunnän Gurage. In analogy to Kistane, the affirmative perfective plus the suffix -m in Muher, Chaha and Inor might actually also go back to a present perfect, which, however, grammaticalized into the only possible construction with perfective verbs in ICL. Consequently, dependent perfective converbs marked by -m became formally homonymous with perfective verbs in ICL. Only syntax, i.e. being in sentence-final position or not, can distinguish between the two. 3.5.2.2 Declarative clause marking Zay (Eastern Gurage) obligatorily marks affirmative imperfective verbs for tense in ICL (cf. Section 3.5.1). In addition, Zay makes use of the declarative marker -u or its allomorph (lengthening of non-central vowels), and obligatory focus marking with -n / -tä in ICL (Meyer 2005:290 ff., 305 ff.). (10) a. ʃäfät’ näk’äl-ä-n-u canoe

take\PFV-3SG.M-FOC-DECL

‘He took the canoe.’

ZAY

Finiteness in Gurage Languages

245

b. ʃäfät’ jɨnäk’lɨnälo ʃäfät’ canoe

j-näk’l-n-älä-ä-u 3SG.M-take\IPFV-FOC-AUX.PRS-3SG.M-DECL

‘He takes the canoe.’ In contrast to the main verb markers discussed in Section 3.5.2.1, the declarative marker in Zay occurs in the affirmative and negative indicative and in the affirmative jussive but not in the imperative, the negative jussive/imperative nor in subordination. Most strikingly, it is never used in questions, as shown in (11) vis-à-vis (10): (11) a. ʃäfät’ näk’äl-ä-n? canoe

ZAY

take\PFV-3SG.M-FOC

‘Did he take the canoe?’ b. ʃäfät’ jɨnäk’lɨnäl? ʃäfät’ canoe

j-näk’l-n-älä-ä 3SG.M-take\IPFV-FOC-AUX.PRS-3SG.M

‘Will he take the canoe?’ The focus marker in Zay distinguishes between affirmative and negative ICL (Meyer 2005:290 ff.), i.e. it is obligatory in affirmative but not in negative ICL. In contrast to the declarative marker, the focus marker can be attached to any constituent of the clause; it need not be the clause-final verb. 3.5.2.3 KTD-suffix Inor (PWG) and to a lesser extend also Chaha (CWG) attach the so-called KTD-suffix to verbs in specific clauses. The three consonants of the suffix are in complementary distribution: -k is attached to a verb base not ending with a subject index, -t follows a long vowel or diphthong, and -d occurs after a short vowel or a consonant belonging to a suffix of the subject index. These consonants are followed by the vowel e in the definite future tense but by the vowel a ~ ä elsewhere (Hetzron 1972a:67 ff.; 1977:92 f.). In Chaha, the KTD-suffix is restricted to the imperfective in the definite future tense (see Section 3.5.1), but in Inor it also occurs with verbs in other clause types: the non-present existential verb in ICL, and negated perfective and imperfective verbs in ICL, as well as with the affirmative imperfective as relative clause verb and in subordinate clauses marked by the conjunction t-:

246

Ronny Meyer

(12) a. a-j-säβɨr-ka

INOR

NEG-3SG.M-break\IPFV-KTD

‘he will not break’ (negative ICL) b. jɨ-säβɨr-kaa-te 3SG.M-break\IPFV-OJ.3PL.F-KTD

‘who (3SG.M) will break them (F)’ (relative clause) c. baanä-dä AUX.NPRS.3SG.M-KTD

‘he was’

(Hetzron 1977:93)

Thus, the KTD-suffix in Inor mainly occurs with negated verbs in ICL but with affirmative imperfective verbs in dependent clauses. This type of asymmetry is unique because a formal feature, the KTD-suffix, is cross-linked with a paradigmatic verbal feature, affirmative vs. negative, to distinguish between dependent and independent clause. Hetzron (1972a:67) is of the opinion that the KTD-suffix in Inor is functionally equivalent to the main verb marker in Kistane and Muher (cf. Section 3.5.2.1). However, the main verb marker is attached to affirmative verbs in ICL but not to subordinate or negated verbs. The KTD-suffix has, thus, a functional range other than the main verb marker, or the declarative marker in Zay. 4 Dependent clauses Dependent clauses are marked by additional subordinating conjunctions attached to verbs but usually lack marking for tense and clausal status occurring with verbs in ICL, as shown in (13): (13) a.

säbbär-hu-m

MUHER

break\PFV-1SG-MVM

‘I broke’ b.

bä-säbbär-hʷ COND-break\PFV-1SG

‘if I break’ c.

al-nɨk’äl-uh-u NEG-take\SBJV-1SG-DECL

‘I did not take’

ZAY

Finiteness in Gurage Languages d.

247

bal-nɨk’äl-uh COND.NEG-take\PFV-1SG

‘if I do not take’ The verbs in the dependent clauses (13b) and (13d) occur in the perfective aspect, but their actual temporal setting depends on the verb in the ICL (Hetzron 1977:101 ff.). Generally, dependent clauses do not anchor the verbal event to a specific time, which was one defining feature of finite verbs mentioned in Section 2. Most common dependent clauses in Gurage languages are relative and converb clauses. Relative clause verbs can be overtly marked by the prefix jä- (with the allomorph j- before vowels) but need not, as shown in Table 32.

Zay

PERFECTIVE AFFIRMATIVE NEGATIVE jä-säbär-(ä) j-al-sɨbär-(ä)

IMPERFECTIVE AFFIRMATIVE jɨ-säbɨr-aal

NEGATIVE j-a-j-säbɨr

REL-break\PFV(3SG.M)

REL-NEG-break\ SBJV-(3SG.M)

3SG.M-break\IPFVAUX.PRS[REL]

REL-NEG-3SG.Mbreak\IPFV

Muher 32 Kistane Wolane

jä-säbär-ä

j-al-säbär-ä

jɨ-säbr-an

j-a-j-säbr-an

REL-break\PFV3SG.M

REL-NEGbreak\PFV-3SG.M

3SG.M-break\IPFVAUX.PRS

REL-NEG-3SG.Mbreak\IPFV-AUX.PRS

jä-säbbär-ä

an-säbär-ä

jɨ-säbɨr

a-j-säbɨr

REL-break\PFV3SG.M

NEG-break\NPFV3SG.M

3SG.Mbreak\IPFV[REL]

NEG-3SG.Mbreak\IPFV[REL]

jä-säbbär-ä

an-säbär-ä

jɨ-säbɨr

e-säbɨr

REL-break\PFV3SG.M

NEG-break\NPFV3SG.M

3SG.M-break\IPFV [REL]Muher only

NEG.3SG.Mbreak\IPFV

Inor

ä-säpär-ä

ã-säpär-ä

jɨ-säbɨr-ka

a-j-säbɨr

REL-break\PFV3SG.M

NEG-break\NPFV3SG.M[REL]

3SG.M-break\IPFVKTD[REL]

NEG-3SG.Mbreak\IPFV[REL]

Gloss

Table 32: Relative Clause Verbs

‘who (3SG.M) ‘who (3SG.M) broke’ did not break’

‘who (3SG.M) breaks’

‘who (3SG.M) does not break’

Imperfective verbs in relative clauses in Zay and Wolane are also marked by the present tense auxiliary. In Zay, this auxiliary occurs only with affirmative 32 The Muher relative verb forms also occur in Mesqan and Chaha.

248

Ronny Meyer

imperfective verbs and merges with the relative prefix jä- yielding *jä-äl(ä) >-aal, which formally differs from the present tense auxiliary -äl(ä) or -aa in ICL (cf. Meyer 2005:177). In Wolane, however, affirmative imperfective verbs in independent and relative clauses are identical. In negation, however, the relative prefix jä- is attached to the imperfective verb in Wolane.33 In Gunnän Gurage, the relative prefix jä- is only attached to affirmative perfective verbs. Consequently, negative perfective verbs in relative clauses and ICL are homonymous in all Gunnän Gurage languages except Inor. In Inor, verbs in relative clauses are clearly separated from ICL by the KTDsuffix, which is attached to negated verbs in ICL but not to negated verbs in relative clauses, and to affirmative imperfective in relative clauses but not in ICL. Negative imperfective verbs in relative and independent clauses in Muher, Mesqan, and Chaha are identical, while Kistane marks imperfective verbs in ICL through the unique negative marker t- (cf. Section 3.4). Further, homonymy between verbs in relative clauses and ICL also occurs in the affirmative imperfective in Mesqan and Chaha, whereas the main verb markers in Kistane and Muher clearly distinguish between them. A similar overlap of verbs in dependent and ICL is found with certain converbs (see Table 33), which are formed from inflected verbs to which various converb markers can be added (cf. Hetzron 1972a:99 ff.; 1977:94 ff.; Leslau 1969b; Meyer 2011:1247). The converb marked by -ä in Wolane but -m elsewhere in Gurage is a general converb (CNV1), which is most frequently used for narration of subsequent events in a macro situation without clear transitions between the individual events, or for causal and manner adverbial modifications of a verb. The converb (CNV2) formed by the suffix -ani in Wolane or -m ta(nnä) ~ -nta(nnä) in Muher, Chaha and Inor is mainly used for narration of subsequent events with different subjects or strong emphasis on the transition between the events. The converb construction (CNV3) based on the subjunctive template to which the suffix -t plus the suffix subject index functions as a specialized narrative converb with selected reference verbs in Chaha and Inor.

33 The relative prefix jä- in Silt’e does not occur with negated imperfective verbs: a-j-säbr-aan ‘who (3SG.M) does not break’.

Finiteness in Gurage Languages

249

Wolane (& Silt’e)

Zay (& Kistane, Mesqan)

Table 33: Converbs in Gurage CONVERB 1 näk’äl-ä-m

CONVERB 2

CONVERB 3

take\PFV-3SG.M-CNV1

jɨ-näk’l-ɨm 3SG.M-take\IPFV-CNV1

jä-nk’äl-ɨm 3SG.M-take\SBJV-CNV1

‘he having taken’ säbär-t-ä

säbär-t-ani

break\PFV-3SG.F-CNV1

break\PFV-3SG.F-CNV2

tɨ-säbr-ä

tɨ-säbr-ani

3SG.F-break\IPFV-CNV1

3SG.F-break\IPFV-CNV2

tä-sbär-ä

tä-sbär-ani

3SG.F-break\SBJV-CNV1

3SG.F-break\SBJV-CNV2

Muher

‘she having broken’ säbbär-ä-m

säbbär-ä-nta(nnä)

break\PFV-3SG.M-CNV1

break\PFV-3SG.M-CNV2

jɨsäbr-ɨm

jɨ-säbr-ɨnta(nnä)

3SG.M-break\IPFV-CNV1

3SG.M-break\IPFV-CNV2

jä-sbɨr-ɨm

jä-sbɨr-ɨnta(nnä)

3SG.M-break\SBJV-CNV1

3SG.M-break\SBJV-CNV2

Chaha (& Inor)

‘he having broken’ säpär-ä-m

säpär-ä-nta(nä)

break\PFV-3SG.M-CNV1

break\PFV-3SG.M-CNV2

jɨ-säbr-ɨm

jɨ-säbr-ɨnta(nä)

3SG.M-break\IPFV-CNV1

3SG.M-break\IPFV-CNV2

jä-sbɨr-ɨm

jä-sbɨr-ɨnta(nä)

3SG.M-break\SBJV-CNV1

3SG.M-break\SBJV-CNV2

sɨbɨr-t-ä break\SBJV-CNV33SG.M

‘he having broken’ Generally, converbs are dependent verbs that always have to co-occur with a reference verb, i.e. they cannot be the only verb in an ICL. The occurrence of CNV1 and CNV2 based on the imperfective or subjunctive templates is restricted to reference verbs in the same aspect/mood, while the CNV3 is re-

250

Ronny Meyer

stricted to negative and non-perfective reference verbs. The converb based on the perfective template co-occurs with reference verbs in all aspects and moods. It is a nonfinite verb because its aspectual/temporal setting depends on that of the reference verb. Further, converbs are subject to certain restrictions with regard to object indexing, and only occur in the affirmative – with the exception of Kistane, Zay, and possibly Silt’e.34 Formally, the perfective converb marked by the suffix -m in Muher, Chaha (CWG) and Inor (PWG) is identical to perfective verbs in ICL, while in Kistane (cf. (9a) above) and Mesqan it is identical to the present perfect in ICL. Similarly, the perfective converb marked by -ä in Wolane is homonymous with certain perfective verbs in ICL (Meyer 2006a:131 ff.). In these cases, only the syntactic position of the verbs distinguishes converbs from verbs in ICL. 5 Independent clauses As shown in the preceding sections, the traditional finiteness features TAM and subject indexing do not plainly distinguish between finite verbs in ICL and nonfinite verbs in dependent clauses. In this function, rather the markers for tense and clausal status (cf. Section 3.5) are of major importance. Their distribution is summarized in Table 34. Table 34: Compulsory Clause Marking in Gurage CLAUSE

MARKER

VERB AFFECTED

GURAGE GROUP

LANGUAGE

Basic Types (1) (2)

No markers AFF indicative

Mesqan (a) (b)

MVM MVM MVM

-u -u -m

PFV / IPFV IPFV / PFV PFV

35

Northern Gurage Northern Gurage

Kistane Muher

34 Although negative converbs are rare, Hetzron’s (1977:98) observation that there are no negated converbs in Gunnän Gurage is not correct, as negative converbs in dependent clauses are attested in Kistane (Bedilu 2010:126 ff.). In Zay (Eastern Gurage), a negated perfective converb followed by the negated present-tense auxiliary forms a complex predicate in ICL (Meyer 2005:171, 158 f.). Based on Gutt (1997:920), a similar construction with a negated perfective CNV1 might also exist in Silt’e, but the data is not entirely conclusive. 35 Only with PROH and AUX.NPRS.

Finiteness in Gurage Languages CLAUSE

VERB AFFECTED

GURAGE GROUP

LANGUAGE

suffix

PFV IPFVFuture AUX.NPRSInor

CWG/ PWG

Chaha Inor (i)

AUX

IPFV

Eastern Gurage

Wolane Zay (ii)

KTDsuffix

PFV / IPFV

PWG

Inor (2c)

Eastern Gurage

Zay (2d)

MARKER (c) (d)

251

MVM KTD-

-m

Additional Sub-Types NEG indicative in i.

ii.

ICL AFF subordinate AFF/NEG nonquestions in ICL

AFF wishes in ICL

DECL

IPFV PFV / IPFV

Jussive

Most Gurage languages distinguish morphologically between dependent and independent clauses only with affirmative indicative verbs. Kistane is the prototypical member of this group. It attaches the main verb marker -u (or one of its allomorphs) to affirmative perfective and imperfective verbs in ICL but not to verbs in dependent clauses. Muher follows the same principle but uses different markers: In affirmative ICL, imperfective verbs are followed by the MVM -u while the MVM -m is attached to perfective verbs.36 Chaha (CWG) only uses the MVM -m attached to affirmative perfective verbs in ICL while imperfective verbs are unmarked. Wolane, in contrast, obligatorily marks affirmative imperfective verbs in ICL, but not perfective verbs, by a temporal auxiliary instead of the main verb marker.37 Inor and Zay represent more complex stages of Chaha and Wolane, respectively. Inor marks the affirmative perfective verb in ICL by an uncommon ultima accent on the verb, which is considered a reflex of the vanished main verb marker -m (Hetzron 1977:42 f.). In addition, Inor obligatorily attaches the KDT-suffix to the non-present auxiliary and to negated indicative verbs in ICL – which is unique within Gurage – but also as a marker of affirmative subordinate imperfective verbs (mainly in relative clauses). Zay, similar to Wolane, marks affirmative imperfective verbs in ICL by a temporal auxiliary, but in addition it has an obligatory declarative marker that 36 Muher also uses the MVM -u with perfective verbs in the prohibitive construction and with the non-present auxiliary verb bannä (cf. Sections 3.4, 3.5.1). 37 Note, however, that the present tense auxiliary in Eastern Gurage is also obligatoriy part of imperfective verbs in relative clauses (cf. Table 32).

252

Ronny Meyer

distinguishes between statements and wishes, on the one hand, and questions, direct orders and negative wishes, on the other. This declarative marker differs functionally from the main verb marker, as it also occurs with negative verbs but not in questions. The approximate geographical distribution of these morphemes is shown in Map 4: Map 4: Distribution of ICL Markers

Mesqan is the only Gurage language that does not use additional markers attached to verbs in ICL. Here either subordinating conjunctions or word order alone distinguishes between verbs in dependent and independent clauses. 6 Conclusion This paper has shown that there are no universal criteria to define finiteness even for genetically closely related languages spoken in a compact area with a high degree of interlingual contacts among their speakers. Subject indexing and aspect/mood marking on verbs are basic for finiteness in all Gurage languages because they enable the independent interpretation of a verbal event regarding its temporal setting and the involved participants. However, verbs marked for perfective or imperfective aspect occur in main and subordinate clauses as well. Although the subjunctive template is used for forming the

Finiteness in Gurage Languages

253

jussive/imperative whose occurrences are restricted to ICL, it is also used to form the nonfinite verbal noun. However, the combination of the subjunctive template with a subject index, which is lacking on verbal nouns, can distinguish between the two. Further, the subject index with affirmative imperfective verbs in ICL for the 1SG in all Gurage languages (and for the 1PL in Eastern Gurage) is a unique morpheme, which does not occur in subordination or negation. The unique subject index with negated verbs in Wolane (and Silt’e), or the unique negative marker in Kistane, clearly distinguishes verbs in ICL from subordinate verbs. Further, subordinating conjunctions and converb markers disable the verb to anchor the expressed event to a specific time. Consequently, these verbs cannot form an ICL on their own. On the other hand, the occurrence of the converb marker -m with perfective verbs in Gunnän Gurage is not sufficient to distinguish between converbs in dependent clauses vs. complex predicates (present perfect) in independent clauses, as they are formally identical. Here only the syntactic position of the verb in a sentence can determine its status. Additional compulsory categories used to mark ICL are tense in Eastern Gurage, the main verb marker in Kistane, Muher, and partly in Chaha and Inor, or the declarative marker in Zay. The KTD-suffix in Inor marks mutual exclusive clauses, namely negated indicative verbs in ICL but also affirmative imperfective verbs in subordination. The compulsory categories to mark ICL are found at geographically distinct places: main verb marking in the north (Kistane), declarative clause marking in the east (Zay) and the KTDsuffix in the south (Inor). Main verb marking dispersed to the south of Kistane into Muher, while the KTD-suffix spread to the north of Inor into Chaha. Today, Muher and Chaha exhibit mixed patterns. Contrary to common opinion (cf. Appleyard 2002b), I do not believe that the different markers for clausal status are diachronically related to each other. Although the main verb marker and the declarative marker contain a morpheme -u, the allomorph of the declarative marker -u is vowel lengthening of non-central vowels, while that of the main verb marker -u are -i, or -t/-n. Vowel lengthening is clearly an innovation in Zay due to Oromo influence (cf. Meyer 2006b), which suggests that declarative marking in Zay might also be an innovation. The distribution of the KTD-suffix might resemble a pattern found in some Cushitic languages, where there is a relation between affirmative subordinate and negative imperfective verbs in ICL with affirmative subordinate verbs. Based on our current knowledge there is no clear answer as to the origin of these markers. The distribution of categories across verbal paradigms, such as aspect/mood templates, subject indexes, subordinating conjunctions, and compulsory morphemes, which are all re-

254

Ronny Meyer

lated to finiteness, suggests that they form interrelated networks with competing patterns. Some of these patterns are restricted to specific clause types, while others may occur in various clauses. Finiteness in Gurage languages is, thus, a relational category. It is the result of various grammaticalization and/or language contact processes, which can best be described through a combination of morphological and syntactic features. Abbreviations ˈ Accented syllable √ Root 1, 2, 3 First, second, third person AFF Affirmative AGR Agreement AUX Auxiliary C1C2C3 First, second, third root consonant CNV Converb COND Conditional marker CWG Central Western Gurage DECL Main declarative clause F Feminine FOC Focus FUT Future ICL Independent clause INV Invariable IPFV Imperfective j Palatalization ʲʷ Combined palatalization and labialization KTD Multifunctional marker of clausal status (cf. Section 3.5.2.3)

M MVM NEG NPFV NPRS OJ PFV PL POSS PROH PRS PS PWG REL SBJV SES SG SJ TAM w

Masculine Main affirmative indicative clause Negative Negative perfective Non-present tense Object Perfective Plural Possessive Prohibitive Present tense Proto-Semitic Peripheral Western Gurage Relative clause Subjunctive South Ethio-Semitic Singular Subject Tense, aspect, mood Labialization

References Ahland, Michael B. 2010. Language Death in Mesmes: A Sociolinguistic and Historical-Comparative Examination of a Disappearing Language. Dallas, TX: SIL International.

Finiteness in Gurage Languages

255

Anderson, John M. 2007. Finiteness, mood, and morphosyntax. Journal of Linguistics 43. 1–32. Appleyard, David. 2002a. New finds in the 20th century: The South Semitic Languages. In Shlomo Isre’el (ed.), Semitic Linguistics: The State of the Art at the Turn of the Twenty-First Century, 401–430. Winona Lake, IN: Eisenbrauns. Appleyard, David. 2002b. The morphology of main and subordinate verb forms in Ethiopian Semitic and Agaw. Afrikanistische Arbeitspapiere 71. 9–31. Bedilu Wakjira Debela. 2010. Morphology and Verb Construction Types of Kistaniniya. Trondheim: NTNU. Berhanu Chamora & Robert Hetzron. 2000. Inor. Munich: Lincom Europa. Bisang, Walter. 2007. Categories that make finiteness: Discreteness from a functional perspective and some of its repercussions. In Irina Nikolaeva (ed.), Finiteness: Theoretical and Empirical Foundations, 115–137. Oxford: Oxford University Press. Bulakh, Maria. 2012. Negative markers *ʔay-, *ʔi- and *ʔal- in EthioSemitic. Babel und Bibel 6. 385–420. Bulakh, Maria & Leonid Kogan. 2010. The genealogical position of Tigre and the problem of North Ethio-Semitic unity. Zeitschrift der Deutschen Morgenländischen Gesellschaft 160. 273–302. Degif Petros Banksira. 2000. Sound Mutations: The Morphophonology of Chaha. Philadelphia & Amsterdam: John Benjamins. Elizabeth Minase. 2012. Tempus, Aspect and Mood in Kəstanəñña. Addis Ababa: Addis Ababa University (MA Thesis). Fekede Menuta Gewta. 2013. Intergroup Communication in Gurage: A Study in Intelligibility, Interlingual Comprehension and Accommodation. Addis Ababa: Addis Ababa University (PhD Dissertation). Goldenberg, Gideon. 1968. Kəstanəñña: Studies in a Northern Gurage language of Christians. Orientalia Suecana 17. 61–102. Goldenberg, Gideon. 1977. The Semitic languages of Ethiopia and their classification. Bulletin of the School of Oriental and African Studies 40. 461–507. Gutt, Ernst-August. 1980. Intelligibility and interlingual comprehension among selected Gurage speech varieties. Journal of Ethiopian Studies 16. 56–84.

256

Ronny Meyer

Gutt, Ernst-August. 1997. Concise grammar of Silt’e. In Eeva H. M. Gutt & Mohammed Mussa Hussein (eds.), Silt’e-Amharic-English Dictionary (with Concise Grammar by Ernst-August Gutt), 896–957. Addis Ababa: Addis Ababa University Press. Hetzron, Robert. 1968. Main verb markers in Northern Gurage. Africa 38(2). 156–172. Hetzron, Robert. 1972a. Ethiopian Semitic: Studies in Classification. Manchester: Manchester University Press. Hetzron, Robert. 1972b. The shape of a rule and diachrony. Bulletin of the School of Oriental and African Studies 35(3). 451–475. Hetzron, Robert. 1977. The Gunnän-Gurage Languages. Napoli: Istituto Orientale di Napoli. Hudson, Grover. 2003. Ethiopian Semitic negative nonpast. In Lionel M. Bender, David Appleyard & Gábor Takács (eds.), Afrasian: Selected Comparative-Historical Linguistic Studies in Memory of Igor M. Diakonoff, 209–218. Munich: Lincom Europa. Huehnergard, John. 1983. Asseverative *la and hypothetical *lu/law in Semitic. Journal of the American Oriental Society 103(3). 569–593. Klein, Wolfgang. 2006. On finiteness. In Veerle van Geenhoven (ed.), Semantics Meets Acquisition, 245–272. Dordrecht: Springer. Leslau, Wolf. 1945. Gafat Documents. Records of a South-Ethiopic Language: Grammar, Text and Comparative Vocabulary. New Haven, CT: American Oriental Society. Leslau, Wolf. 1952. The influence of Sidamo on the Ethiopic languages of Gurage. Language 28(1). 63–81. Leslau, Wolf. 1968a. Ethiopians Speak: Studies in Cultural Background. Part III. Soddo. Berkeley & Los Angeles: University of California Press. Leslau, Wolf. 1968b. An archaic vowel of the jussive in Gurage, Gafat and Harari. Orientalia 37. 90–93. Leslau, Wolf. 1969a. The negative particle ʔin Arabic and ʔǝn in Ethiopic. Annali dell’Istituto Orientale di Napoli (Nuova Serie) 29. 138–145. Leslau, Wolf. 1969b. The pseudo-gerundive in Chaha. Rassegna di Studi Etiopici 23. 27–42. Leslau, Wolf. 1983. Ethiopians Speak: Studies in Cultural Background. Part V. Chaha and Ennemor. Wiesbaden: Harrassowitz.

Finiteness in Gurage Languages

257

Leslau, Wolf. 2004. The Verb in Mäsqan as Compared with Other Gurage Dialects. Wiesbaden: Harrassowitz. Lipiński, Edward. 1997. Semitic Languages: Outline of a Comparative Grammar. Leuven: Peeters and Departement Oosterse Studies. Maas, Utz. 2004. “Finite” and “nonfinite” from a typological perspective. Linguistics 42(2). 359–385. Meseret Eshetu. 2012. Tense, Aspect and Mood in Mesqan. Addis Ababa: Addis Ababa University (MA Thesis). Meyer, Ronny. 2005. Das Zay: Deskriptive Grammatik einer Ostguragesprache (Äthiosemitisch). Cologne: Köppe. Meyer, Ronny. 2006a. Wolane: Descriptive Grammar of an East Gurage Language (Ethiosemitic). Cologne: Köppe. Meyer, Ronny. 2006b. Cultural contact and language change in Eastern Gurage. In Siegbert Uhlig (ed.), Proceedings of the 15th International Conference of Ethiopian Studies, Hamburg July 20–25, 2003, 813–821. Wiesbaden: Harrassowitz. Meyer, Ronny. 2011. Gurage. In Stefan Weninger (ed.), The Semitic Languages: An International Handbook, 1220–1257. Berlin & New York: De Gruyter Mouton. Nikolaeva, Irina. 2007. Introduction. In Irina Nikolaeva (ed.), Finiteness: Theoretical and Empirical Foundations, 1–19. Oxford: Oxford University Press. Petersson, Thore. 1994. Tense. Lund University Department of Linguistics Working Papers 42. 179–196. Rose, Sharon. 1996a. Inflectional affix order in Ethio-Semitic. In Jacqueline Lecarme, Jean Lowenstamm & Ur Shlonsky (eds.), Studies in AfroAsiatic Grammar: Papers from the Second Conference on Afroasiatic Languages, Sophia Antipolis, 1994, 337–359. The Hague: Holland Academic Graphics. Rose, Sharon. 1996b. Allomorphy and morphological categories in Muher. In Grover Hudson (ed.), Essays on Gurage Language and Culture, Dedicated to Wolf Leslau on the Occasion of his 90th Birthday, 205–227. Wiesbaden: Harrassowitz. Rose, Sharon. 2006. Durational conditions on Endegeň gemination. In Siegbert Uhlig (ed.), Proceedings of the 15th International Conference of Ethiopian Studies, Hamburg, July 20–25, 2003, 843–850. Wiesbaden: Harrassowitz.

258

Ronny Meyer

Tropper, Josef. 2002. Altäthiopisch. Grammatik des Geʿez mit Übungstexten und Glossar. Münster: Ugarit. Voigt, Rainer M. 2009. North vs. South Ethiopian Semitic. In Sven Ege, Harald Aspen, Tefera Birhanu & Bekele Shiferaw (eds.), Proceedings of the 16th International Conference of Ethiopian Studies, vol. IV, 1375–1387. Trondheim: Department of Social Anthropology (NUST). Wagner, Ewald. 1968. Drei Miszellen zum südostsemitischen Verbum. Wissenschaftliche Zeitschrift der Universität Halle 17(2/3). 207–215.

Case Marking in Amharic Copular Constructions Mulusew Asratie Addis Ababa University and Utrecht University Abstract In this paper, I discuss the nominative and accusative alternation of NP and AP predicates in Amharic copular constructions. Based on the difference between the accusative and nominative predicates in interpretation, agreement and word order, I argue that accusative predicates must be assigned by a functional element which introduces eventivity. 1 Introduction Amharic copular constructions1 exhibit a case marking system in which the subject is nominative and the predicate2 alternates between nominative (unmarked) and accusative (marked) with the corresponding difference in interpretation: (1) a. lɨǰ-očč-u child-PL-DEF.M

tämari-wočč/

gobäz-očč

n-aččäw

student-PL

clever-PL

COP-3SG.M.SJ;3PL.OJ

‘The children are students/clever.’ b. lɨǰ-očč-u

tämari-wočč-ɨn/ raqut-aččäw-n

child-PL-DEF.M student-PL-ACC

n-aččäw

clever-POSS.3PL-ACC COP-3SG.M.SJ; 3PL.OJ

‘The children are just like students/are naked.’

1

2

I would like to thank Dr. Ora Matushansky and Prof. Martin Everaert for their comments and suggestions. I also thank the audience in the Weak Preferentiality Workshop at Utrecht Institute of Linguistics, 26 March 2012, Belgian Linguistic Day, 19 May 2012, and the 18th International Conference of Ethiopian Studies, 29 October–02 November 2012 in Dire Dawa. By the term predicate I am referring to the noun phrase (NP) or adjectival phrase (AP) that states or affirms something about the subject. In the typological literature, such phrases are referred to as nominal or adjectival complements of the copula. In the generative framework, which I am adopting, the copula is not assumed to take the NP or AP as its complement (see Section 2). As a result, the NP and AP are considered to be predicates.

260

Mulusew Asratie

Note that the present-tense copula in Amharic is an impersonal verb (Getatchew 1974; Baye 2006).3 The focus of this paper is to explain this alternation from the perspective of formal linguistics.4 I show that the nominative/accusative alternation in 3

This is evidenced from its agreement system. In Amharic, personal and impersonal verbs are distinguished by their agreement marking. Personal verbs are marked obligatorily for subject agreement and optionally for object agreement as in (i). Impersonal verbs, on the other hand, are obligatorily marked for subject and object agreement, as in (ii), whereby the subject is always the default 3SG.M. (i) a. lɨǰ-it-u mäs’haf-u-n anäbbäb-äčč child-F-DEF.M book-DEF.M-ACC read\PFV-3SG.F.SJ

‘The girl read the book.’ b. lɨǰ-it-u

mäs’haf-u-n

anäbbäb-äčč-ɨw

child-F-DEF.M book-DEF.M-ACC read\PFV-3SG.F.SJ-3SG.M.OJ

‘The girl read the book.’ (ii) a. *lɨǰ-it-u

rab-äčč

child-F-DEF.M be_hungry\PFV-3SG.F.OJ

b. lɨǰ-it-u

rab-at

child-F-DEF.M be_hungry\PFV-3SG.M.SJ;3SG.F.OJ

4

‘The girl is hungry. (lit. It hungered the girl.)’ In (ii b), the person/thing that is affected by the event denoted by the verb is marked by object agreement. The linguistic theory I am adopting in this study is the Principles and Parameters Theory, specifically the Minimalist Program developed by Chomsky (1995; 2000; 2001). Minimalism assumes that language is based on simple principles that interact to form intricate structures. In Minimalism, it is hypothesized that there is a component in the human cognitive system specialized for language. This cognitive component is assumed to interact with other cognitive systems, namely the phonological-articulatory (P–A) and the conceptual-intentional (C–I) systems which interpret linguistic expressions. Any linguistic expression is acceptable only if it is legible to the P–A and C–I systems. As a result, linguistic representations have only two interfaces, which are known as phonological form (PF) and logical form (LF) that correspond to the P–A and C–I systems, respectively. The PF is an input for P–A, and the LF is an input for C–I. Syntactic constructions are built through three operations known as MERGE, MOVE and AGREE. MERGE is a binary operation which is applied recursively. This means that two constituents merge and form a bigger syntactic constituent, and the bigger constituent may merge with another constituent to form another, bigger, constituent: (iii) Merge (X, Y) → Z MOVE (also known as INTERNAL MERGE) and AGREE are applied to merged constituents. The notion of AGREE and MOVE is closely related to formal features. By formal features, I mean grammatical features such as phi-features (number, gender, and person) and case features. Syntactic constituents are endowed with grammatical (formal) as well as semantic and phonological features. The phonological and semantic features

Case Marking in Amharic Copular Constructions

261

Amharic is related to eventivity as opposed to stativity. Eventive predicates denote a temporary (transient) property as opposed to stative predicates which denote a relatively permanent property. Assuming that eventivity is introduced by a functional projection, I argue, following Matushansky (2008), that accusative must be assigned to Amharic predicates by an extra functional element. In the absence of this functional head, however, I claim following Pereltsvaig (2001) that the predicate receives the default case. The paper is organized as follows. In Section 2, I provide a brief review of how copular clauses are analyzed in the Minimalist Program, the theory I am adopting in this study. In Section 3, I discuss various proposals on predicate case marking and show that most of them are unable to handle the Amharic data. Then, I provide my arguments which show that the accusative vs. nominative alternation in Amharic is related to eventivity and non-eventivity. Section 4 concludes the paper. 2 Structure of copular clauses The Minimalist Program developed in Chomsky (1995; 2000; 2001) assumes that clausal structure is built on a predicational core on which other functional projections which are responsible for checking formal features (case, agreement, and EPP) and information structure are built. In copular constructions the predicational core is known as small clause (Stowell 1981; Bowers 1993; den Dikken 2006). In such clauses, the copula, which is inserted in order to support the affixes bearing tense, aspect and mood (TAM), is assumed to take this small clause as a complement. The syntactic derivation of copular clauses proceeds in such a way that the copula establishes agreement with the subject of the small clause at TP in order to check its uninterpretable phi-features. The subject is assigned nominative case as a byproduct of the are interpretable at PF and LF, respectively, since they are legible to the P–A and C–I systems. Grammatical features, however, can be interpretable or uninterpretable. For example phi-features are interpretable on nouns, but uninterpretable on verbs. Moreover, HEADS have an uninterpretable EPP feature which requires their SPECIFIERS to be filled. In order for a derivation to converge at the interfaces, uninterpretable features have to be deleted. The process of deleting uninterpretable features is done through a feature checking operation which involves AGREE and MOVE. Through AGREE a HEAD which contains uninterpretable features targets another constituent which contains interpretable features. If the features agree, the uninterpretable phi-features of the HEAD are deleted. If not, the derivation crashes. The operation MOVE is used to check the uninterpretable EPP feature of HEADS. That is in order to fulfill the EPP feature of a HEAD, a constituent moves from the COMPLEMENT position to the SPECIFIER position of that HEAD. Fulfilling the EPP feature may also involve inserting an expletive rather than carrying out a movement.

262

Mulusew Asratie

agreement it enters with the copula. The subject, then, moves to the specifier of TP in order to fulfill the EPP feature of To. Since Bowers (1993), the small clause is assumed to be headed by a functional head known as Predo. Accordingly, the syntactic structure of copular clauses such as the English sentence John is a teacher is analyzed as follows: (2)

TP

NPi John

T′ T

o

is

PredP ti

Pred′ Pred

o

NP a teacher

Given this uniform analysis, variation with regard to case marking of predicates, as we saw in (1) for Amharic, raises the following question: If copular constructions have a uniform structure which involves a copula and a small clause construction as in (2), why do languages like Amharic show variation in case marking patterns of the NP/AP predicates in copular constructions? This question has been addressed by a number of researchers based on data from different languages. In the sections that follow, I review some of the solutions proposed and show that most of them are unable to explain the Amharic data. 3 Predicate case marking Variation in predicate case marking is not unique to Amharic. In many languages, non-verbal (NP/DP and AP) predicates can be marked for the same case as the subject or for a different case (see Maling & Sprouse 1995; Comrie 1997; Bailyn 2001; Pereltsvaig 2001; Matushansky 2008; Citko 2008 among others). The following examples from Pereltsvaig (2001:1) demonstrate this phenomenon in Russian: (3) a. čexov

byl pisatel’

Chekhov.NOM was writer.NOM

‘Chekhov was a writer.’ b. čexov

byl pisatelem

Chekhov.NOM was writer.INST

‘Chekhov was a writer.’

RUSSIAN

Case Marking in Amharic Copular Constructions

263

Why predicates show such case alternation, however, remains an issue of debate. There are three syntactic proposals with regard to this. The first is that of Maling & Sprouse (1995) and Comrie (1997) who claim that case marking on the predicate is determined by whether the copula assigns case or not. This means that the predicate is marked for a different case from the subject when the copula assigns case to it while it is marked for the same case as the subject when the copula does not assign case. The second proposal is that of Bailyn (2001), Matushansky (2008) and Citko (2008), who argue that case alternation is determined by whether or not there is a case-assigning head in the small clause: The predicate is marked for a different case if the small clause contains a case-assigning head while it is marked for the same case as the subject if the small clause does not contain a case-assigning head. The third proposal has been made by Pereltsvaig (2001), who attributes predicate case marking to the category of the predicate and the dual nature of the copular verb (that it is a lexical and a functional element). According to her, the predicate is assigned a different case from the subject when it is AP or NP and the verb is a functional element, but it is assigned the same case as the subject when it is DP and the verb is lexical. The three proposals also vary regarding how the predicate gets the same case as the subject. According to Maling & Sprouse (1995), Bailyn (2001), Matushansky (2008) and Citko (2008), the predicate receives the same case as the subject because it is assigned case by the functional element which also assigns case to the subject. For Comrie (1997), the predicate agrees in case with the subject while for Pereltsvaig (2001) the predicate receives the default case. The question then emerges as to what determines case alternation in Amharic? Is it the copula as postulated by Comrie (1997) and Maling & Sprouse (1995)? Is it the category of the predicate (that it is AP/NP vs. DP) as proposed by Pereltsvaig (2001)? Is it the presence of a case-assigning head in the small clause as argued by Bailyn (2001), Matushansky (2008) and Citko (2008)? Or is it something else? Obviously, contra Maling & Sprouse (1995) and Comrie (1997), the copula cannot be responsible for the case alternation in Amharic unless one assumes that the copula is ambiguous between two types of verbs: one which assigns accusative case to the predicate and the other which does not, as claimed in Pereltsvaig (2001). However, an ambiguous copula can also not explain the case alternation since the same alternation is also found with the past tense copula näbbär- and with lexical verbs like adärräg- ‘make’ (cf. (4) and (5)).5 Thus, the claim that the copula is ambiguous between a case5

Other verbs which behave like adärräg- include mässäl- ‘seem’ and hon- ‘become’.

264

Mulusew Asratie

assigning verb and a non-case assigning verb would have to be extended to these verbs too: (4)

NOMINATIVE PREDICATE a. lɨǰ-očč-u tämari-wočč/ child-PL-DEF.M student-PL

gobäz-očč

näbbär-u

clever-PL

AUX.PST-3PL.SJ

‘The children were students/clever.’ ACCUSATIVE PREDICATE b. lɨǰ-očč-u tämari-wočč-ɨn / gobäz-očč-ɨn näbbär-u child-PL-DEF.M student-PL-ACC

clever-PL-ACC

AUX.PST-3PL.SJ

‘The children were just like students/were like clever ones.’ (5)

NOMINATIVE PREDICATE a. yonas ɨňňa-n tämari-očč Jonas

we-ACC

student-PL

adärräg-ä-n make\PFV-3SG.M.SJ-1PL.OJ

‘Jonas considered (lit. made) us to be students.’ [We may be or not be students.] ACCUSATIVE PREDICATE b. yonas ɨňňa-n tämari-očč-ɨn Jonas

we-ACC

student-PL-ACC

adärräg-ä-n make\PFV-3SG.M.SJ-1PL.OJ

‘Jonas considered (lit. made) us to be students.’ [We cannot be students.] Similarly, Pereltsvaig’s (2001) claim that predicates which are marked for a different case from the subject are APs/NPs, while those which are marked for the same case as the subject are DPs does not work for Amharic since DPs containing an overt definite article can be nominative or accusative: (6) a. saba tämari-wa/ Saba

student-DEF.F/

ityop’yawi-t-wa

n-at

Ethiopian-F-DEF.F

COP-3SG.M.SJ;3SG.F.OJ

‘Saba is the student/the Ethiopian.’ b. saba tämari-wa-n/ Saba

ityop’yawi-t-wa-n

n-at

student-DEF.F-ACC Ethiopian-F-DEF.F-ACC COP-3SG.M.SJ;3SG.F.OJ

‘Saba is just like the student/the Ethiopian.’ The claim that predicate case marking is related to the presence and absence of a case-assigning head in small clauses has two versions. In the first version advocated by Citko (2008) two types of small clause heads are assumed: one

Case Marking in Amharic Copular Constructions

265

non-eventive with incomplete phi-features and the other eventive with complete phi-features.6 The former is found in identity and specificational as well as individual-level predicational clauses7 and does not assign case to the predicate. The latter is found only with predicational clauses and assigns case to the predicate. In the second version advocated by Matushansky (2008), the case feature is assumed to be assigned by a particular head to its complement with the consequence of one case feature being assigned to more than one NP/DP or AP, and one NP/DP or AP being assigned more than one case. As a result, it is claimed that predicates are assigned a different case from the subject when another case assigner interferes between the subject and the predicate. In this instance, case marking on the predicate is the spellout of the composite case features assigned by the case assigner of the subject and the interfering case assigner. Both versions seem to be equally applicable to the Amharic data until compelling evidence is found in favor of one or the other proposal. In the remaining sections of this paper, I show that the nominative/ accusative alternation in Amharic is related to the presence and absence of eventivity. Assuming that eventivity is introduced by a functional projection, I argue, following Matushansky (2008), that variation in predicate case marking is related to the presence and absence of an extra functional head. In the absence of this functional head, however, I claim following Pereltsvaig (2001) that the predicate receives the default case. Evidence in support of my claim comes from the difference between nominative and accusative predicates with regard to predicate selection, interpretation, subject-predicate agreement and word order. In the sections that follow, I elaborate on how these can support my claim. In Section 3.1, I discuss nominative predication. In Section 3.2, I discuss accusative predication. Section 3.3 focuses on unmarked predicates.

6

7

The notion of (in)completeness regarding phi-features is mainly related to the person, number and gender features (cf. Higgins 1979). A head is known as phi-complete if it is specified for all of these features and it is incomplete if it is unspecified or only partially specified for these features. The classification of copular clauses as predicational, specificational and identificational is based on their semantics. Predicational clauses attribute a property to the subject as in The hat is big. Identificational clauses equate two referents as in Cicero is Tully, and specificational clauses involve valuing a variable introduced by the subject as in Whom I met was Peter.

266

Mulusew Asratie

3.1 Nominative predicates Nominative predicates are allowed with NP and individual-level AP predicates.8 These clauses have the standard predicational interpretation. That is, with NP predicates they have a set membership interpretation and with AP predicates a property ascription interpretation: (7) a. saba tämari/ gobäz n-at Saba

student

clever

COP-3SG.M.SJ;3SG.F.OJ

‘Saba is a student/clever.’ b. lɨǰ-očč-u child-PL-DEF.M

tämari-wočč/

gobäz-očč n-aččäw

student-PL

clever-PL

COP-3SG.M.SJ;3PL.OJ

‘The children are students/clever.’ c. *lɨǰ-u

rak’ut n-ä-w

child-DEF.M

naked

COP-3SG.M.SJ-3SG.M.OJ

Intended meaning: ‘The boy is naked.’ Example (7c) shows that a stage-level predicate cannot be expressed by a nominative AP. Copular clauses with two nominative DPs give rise to an identity interpretation: (8) a. saba tämari-wa/ Saba

student-DEF.F

ityop’yawi-t-wa

n-at

Ethiopian-F-DEF.F

COP-3SG.M.SJ;3SG.F.OJ

‘Saba is the student/the Ethiopian.’ b. finfinne Finfinne

addis abäba

n-at

Addis Ababa

COP-3SG.M.SJ;3SG.F.OJ

‘Finfinne is Addis Ababa.’ With regard to agreement, nominative predicates remain either unmarked, as in (9), or they agree in number and gender with the subject, as in (10) and (11): 8

Predicates are classified into individual level (IL) and stage level (SL) depending on their temporal span. IL predicates denote properties which are true of the subject throughout his/her/its lifespan. In the sentence John is smart, the predicate smart is an IL predicate since John’s smartness is true regardless of any point of time in his life time. SL predicates on the other hand denote a temporary property of the subject. The predicate hungry in John is hungry is an SL predicate. It denotes the property of John only at some point of time.

Case Marking in Amharic Copular Constructions

267

(9) a. saba tämari/ gobäz n-at Saba

student

clever

COP-3SG.M.SJ;3SG.F.OJ

‘Saba is a student/clever.’ b. yonas tämari/ gobäz n-ä-w Jonas

student

clever

COP-3SG.M.SJ-3SG.M.OJ

‘Jonas is a student/clever.’ c. lɨǰ-očč-u

tämari/ gobäz n-aččäw

child-PL-DEF.M

student

clever

COP-3SG.M.SJ;3PL.OJ

‘The children are students/clever.’ (10) a. lɨǰ-očč-u

tämari-očč/ gobäz-očč n-aččäw

child-PL-DEF.M

student-PL

clever-PL

COP-3SG.M.SJ;3PL.OJ

‘The children are students/clever.’ b. *yonas tämari-očč n-ä-w Jonas

student-PL

(11) a. saba Saba

COP-3SG.M.SJ-3SG.M.OJ

tämari-wa

n-at

student-DEF.F

COP-3SG.M.SJ;3SG.F.OJ

‘Saba is the student.’ b. *saba Saba

tämari-w

n-at

student-DEF.M

COP-3SG.M.SJ;3SG.F.OJ

As a result, English sentences such as The problem is your parents, The Walias are the best team cannot be translated into Amharic unless the two DPs agree: (12) a. *čɨgɨr-u problem-DEF.M

b. *walya-očč Walia-PL

betäsäb-očč-ɨh

n-ä-w

parent-PL-POSS.2SG.M

COP-3SG.M.SJ-3SG.M.OJ

mɨrt’-u

budɨn

n-aččäw

best-DEF.M

team

COP-3SG.M.SJ;3PL.OJ

With regard to word order, the subject and the nominative predicate can occur in subject-predicate or predicate-subject order. The subject-predicate order (13b) has a predicational interpretation. The predicate-subject order (14b) on the other hand has what is known as a specificational interpretation (Higgins 1979; Line Mikkelsen 2005; Mike Mikkelsen 2011):

268

Mulusew Asratie

(13) a.

ɨnnantä mɨndɨn n-aččɨhu? you.PL

9

what

COP-3SG.M.SJ;2PL.OJ

‘What are you?’ b. ɨňňa tämari-wočč(-u) we

student-PL(-DEF.M)

n-ä-n COP-3SG.M.SJ-1PL.OJ

‘We are (the) students.’ (14) a.

tämari-wočč(-u)

ɨnnäman

n-aččäw?

student-PL(-DEF.M)

who.PL

COP-3SG.M.SJ;3PL.OJ

‘Who are (the) students.’ b. tämari-wočč(-u) student-PL(-DEF.M)

ɨňňa

n-ä-n

we

COP-3SG.M.SJ-1PL.OJ

‘The students are us.’ I propose that the properties we saw above suggest a lack of eventivity in clauses with nominative predicates. Firstly, the reason why nominative predicates are allowed only with NP and individual-level AP predicates but not with stage-level predicates is because the latter are eventive. That is, SL predicates denote a temporary property which is eventive. NP and IL adjectival predicates, on the other hand, denote a lifelong property of the subject. Thus, they do not involve any event.10 Similarly, the subject-predicate agreement in number and gender follows from the lack of any other element that would control the formal properties of the predicate. Since there is no intermediate head which assigns case to the 9 Amharic does not distinguish between masculine and feminine gender in the plural. Thus, all plural forms in Amharic are in common gender. 10 An anonymous reviewer suggested that stage-level predicates always contain an event role while individual-level predicates do not. But this is arguable. In fact, it is not always the case that individual- and stage-level predicates differ due to the presence and absence of an event role. As we will see in Section 3.2, for example, individual-level predicates can also be eventive if they are marked for accusative. On the other hand, some stage-level adjectives can also be marked for nominative if they are coerced to give an individual-level interpretation as in (iv a) as opposed to the stage-level interpretation (iv b): (iv) a. sähara bäräha däräk’/*däräk’-u-n n-ä-w Sahara

desert

dry/*dry-POSS.3SG.M-ACC

COP-3SG.M.SJ-3SG.M.OJ

‘The Sahara desert is dry.’ b. dabbo-w

däräk’/*däräk’-u-n

bread-DEF.M dry/*dry-POSS.3SG.M-ACC

tä-bälla PASS-eat\PFV.3SG.M.SJ

Lit. ‘The bread is eaten dry [i.e. it is on its own, without anything else].’

Case Marking in Amharic Copular Constructions

269

predicate and controls its formal properties, the formal properties of the predicate are controlled by the subject unless the predicate remains unmarked for phi-features. Moreover, the free word order can be attributed to the optional movement of the subject or the predicate to the specifier position of TP, as proposed in Moro (1997). According to him, in the derivation of copular clauses, either the subject or the predicate of the small clause can raise to the specifier position of TP. When the subject raises to the specifier position of TP we get predicational clauses. When the predicate raises, we get specificational clauses. The syntactic structure of clauses with nominative predicates therefore constitutes canonical small clauses headed by Predo (Bowers 1993; den Dikken 2006). The copula is inserted to indicate tense. How does the predicate get the nominative case then? As mentioned above, three proposals have been brought forward. The first is that the predicate is assigned case by the same functional element that assigns case to the subject, as argued by Maling & Sprouse (1995), Bailyn (2001), Citko (2008), Matushansky (2008). The second is that the predicate is assigned case through case agreement with the subject, as proposed by Comrie (1997); and the third is that the predicate receives the default case, as argued in Pereltsvaig (2001). It is difficult to choose one of these three alternative analyses based on the data presented so far. However, the distribution of accusative and nominative predicates in secondary predication, as I show below, suggests that the last option better explains the Amharic data. That is, in the absence of the functional projection which assigns case to the predicate, the predicate receives the default case. The two types of predicate case we saw in primary predication are also found in secondary predication. ECM (15) a. yonas ɨňňa-n Jonas

we-ACC

tämari-wočč adärräg-ä-n student-PL

make\PFV-3SG.M-SJ-1PL.OJ

‘Jonas considered us to be students.’ [We may be or not be students.] b. yonas Jonas

ɨňňa-n

tämari-wočč-ɨn

adärräg-ä-n

we-ACC

student-PL-ACC

make\PFV-3SG.M-SJ-1PL.OJ

‘Jonas considered us just like students.’ [We cannot be students.]

270

Mulusew Asratie

RAISING (16) a. saba Saba

tämari-wa

mässäl-äčč-ɨň

student-DEF.F

seem\PFV-3SG.F.SJ-1SG.OJ

‘Saba seems to me to be the student.’ [She may be or not be a student.] b. saba Saba

tämari-wa-n

mässäl-äčč-ɨň

student-DEF.F-ACC

seem\PFV-3SG.F.SJ-1SG.OJ

‘Saba seems to me to be the student.’ [She cannot be the student.] (17) a. yonas wättadär-u Jonas

soldier-DEF.M

hon-ä become\PFV-3SG.M.SJ

‘Jonas has become the soldier.’ b. yonas wättadär-u-n Jonas

hon-ä

soldier-DEF.M-ACC

become\PFV-3SG.M.SJ

‘Jonas has become just like the soldier.’ DEPICTIVES (18) a. saba bunna-w-n Saba

tɨkkus-u-n

t’ätt’a-čč-ɨw

coffee-DEF.M-ACC hot-POSS.3SG.M-ACC drink\PFV-3SG.F.SJ-3SG.M.OJ

‘Saba drank the coffee hot.’ b. *saba bunna-w-n Saba

coffee-DEF.M-ACC

(19) a. bunna-w coffee-DEF.M

tɨkkus-u

t’ätt’a-čč-ɨw

hot-POSS.3SG.M

drink\PFV-3SG.F.SJ-3SG.M.OJ

tɨkkus-u-n

mätt’a

hot-POSS.3SG.M-ACC

come\PFV.3SG.M.SJ

‘The coffee came hot.’ b. *bunna-w coffee-DEF.M

tɨkkus-u

mätt’a

hot-POSS.3SG.M-

come\PFV.3SG.M.SJ

Nominative predicates are found in Exceptional Case Marking (ECM)11 and raising constructions whereas accusative predicates are found in ECM, raising and depictive constructions.12

11 ECM construction denotes a phenomenon where the subject of an embedded clause is marked with an object case rather than the subject case as in (v). Clauses like (15) are also considered as ECM constructions which contain an embedded small clause whose subject is accusative-marked like objects:

Case Marking in Amharic Copular Constructions

271

The occurrence of nominative predicates in those constructions would not follow if nominative case were assigned by the functional element which assigns case to the subject or if case were assigned by case agreement with the subject. If case were assigned to the predicate through either of these ways, the nominative case of the small clause predicate in ECM constructions would have been impossible since the subject of the small clause would be assigned accusative case.13 For this reason, I propose that the nominative case on the secondary predicates is the default case. This means that predicates are assigned the default case unless they are assigned case by the functional head which introduces eventivity. 3.2 Accusative predicates The properties of clauses with accusative predicates with regard to distribution and interpretation, agreement and word order also suggest the presence of an extra functional projection which introduces eventivity. To begin with, accusative predicates are primarily found with stage-level (SL) predicates: (20) a. bunna-w

tɨkkus-u-n/*tɨkkus-u

n-ä-w

coffee-DEF.M hot-POSS.3SG.M-ACC/hot-POSS.3SG.M COP-3SG.M.SJ-3SG.M.OJ

‘The coffee is hot.’ b. ɨňňa rak’ut-aččɨn-n/*rak’ut-aččɨn we

naked-POSS.1PL-ACC/naked-POSS.1PL

n-ä-n COP-3SG.M.SJ-1PL.OJ

‘We are naked.’

(v)

a. John believed him to be innocent. b. He wants us to be respectful. Similarly, raising denotes a phenomenon in which a given predicate or verb takes the argument of an embedded predicate as its argument as in (vi). (vi) a. It seems that they are writing a letter. b. They seem to be writing a letter. The subject they in (vi) is the argument of the embedded verb writing as we see in (vi a). In (vi b), however, it is the subject of the matrix verb seem by raising. 12 Other secondary predicates that could be mentioned here are resultatives. However, resultative constructions in Amharic involve a complex clause: (vi) bunna-w-n ɨsk-i-däkk’ wäkk’ät’-äčč-ɨw coffee-DEF.M-ACC

until-3SG.M.SJ-be_fine\IPFV

pound\PFV-3SG.F.SJ-3SG.M.OJ

‘She pounded the coffee until it became fine.’ 13 Note that the subject of the small clause is not necessarily the subject of the clause. For example, in the ECM construction (15), the subject of the matrix clause is Jonas. The subject of the small clause (secondary predicate tämariwočč), however, is ɨňňa.

272

Mulusew Asratie c. ɨne I

bado-ye(*-n)

n-ä-ň

empty-POSS.1SG(*-ACC)

COP-3SG.M.SJ-1SG.OJ

‘I am empty [i.e. I don’t have anything at the moment].’ NP and individual-level (IL) adjectival predicates can also be marked for accusative. In this case, they have a stage-level ‘be just like’ interpretation as opposed to set membership or property ascription; see (21) and (22): (21) a. lɨǰ-očč-u child-PL-DEF.M

tämari-wočč-ɨn

n-aččäw

student-PL-ACC

COP-3SG.M.SJ;3PL.OJ

‘The children are just like students.’ [They are not real students]. b. lɨǰ-očč-u

tämari-wočč n-aččäw

child-PL-DEF.M

student-PL

COP-3SG.M.SJ;3PL.OJ

‘The children are students.’ [They are real students]. (22) a. yohannɨs John

ɨnna

saba

ityop’yawy-an-ɨn

n-aččäw

and

Saba

Ethiopian-PL-ACC

COP-3SG.M.SJ;3PL.OJ

‘John and Saba are just like Ethiopians.’ [They are not real Ethiopians]. b. yohannɨs ɨnna saba ityop’yawy-an n-aččäw John

and

Saba

Ethiopian-PL

COP-3SG.M.SJ;3PL.OJ

‘John and Saba are Ethiopians.’ The difference between (21a) and (21b) is that while the children are real students in the latter, they are not in the former. In (21a), the children are said to be students not because they are real students, but because they show some properties associated with students. Similarly, (22a) and (22b) are different in that John and Saba are Ethiopian citizens in the latter, but they are not in the former. In (22a), they are said to be Ethiopians because they behave like Ethiopians. Copular clauses with accusative DP predicates also give rise to the ‘behave like’ interpretation as opposed to the identity interpretation of nominative DPs. Suppose someone says the clauses in (23) to a group of children. (23a) means that the children and the students refer to the same individuals whereas (23b) means that the children are behaving like the (other) students. This is true even if the children are not students at all. Similarly, (24a) is false since Bahirdar and Hawassa are different cities whereas (24b) can be true if

Case Marking in Amharic Copular Constructions

273

the two cities are similar in some respects, for example, in their architecture or beauty. (23) a. tämari-wočč-u

n-aččɨhu

student-PL-DEF.M

COP-3SG.M.SJ;2PL.OJ

‘You are the students.’ b. tämari-wočč-u-n

n-aččɨhu

student-PL-DEF.M-ACC

COP-3SG.M.SJ;2PL.OJ

‘You are just like the students.’ (24) a. bahɨrdar Bahirdar

hawassa

n-at

Hawassa

COP-3SG.M.SJ;3SG.F.OJ

‘Bahirdar is Hawassa.’ [false] b. bahɨrdar Bahrdar

hawassa-n

n-at

Hawassa-ACC

COP-3SG.M.SJ;3SG.F.OJ

‘Bahirdar looks like Hawassa.’ [potentially true] The facts that stage-level AP predicates are accusative-marked and that accusative individual-level AP and NP/DP predicates have a ‘behave like’ interpretation suggest that accusative case marking is related to eventivity. However, this does not show whether the eventivity is introduced by the small clause head which is responsible for predication, as proposed by Citko (2008), or by some extra functional projection. An interesting piece of evidence in favor of the latter proposal comes from the interaction of accusative case and prepositions. Accusative case on predicates is omitted if a preposition is used. Compare (25) with (21a) and (22a): (25) a. lɨǰ-očč-u

ɨndä-tämari-wočč n-aččäw

child-PL-DEF.M

like-student-PL

COP-3SG.M.SJ;3PL.OJ

‘The children are just like students.’ b. yohannɨs ɨnna John

and

saba ɨndä-ityop’yawy-an

n-aččäw

Saba

COP-3SG.M.SJ; 3PL.OJ

like-Ethiopian-PL

‘John and Saba are just like Ethiopians.’ This kind of complementary distribution between accusative case and prepositions would be unexpected if the case assigner were the head of the small clause which is responsible for predication. This means that the small clause

274

Mulusew Asratie

head must be available even with the PP. Associating accusative case with the small clause head would predict that the latter is absent with PPs. In fact, one may assume that the preposition is also a manifestation of the small clause head, as proposed by Aarts (1992), Bailyn & Rubin (1991), Bowers (1993), Bailyn (2001; 2002), and claim that the alternation between prepositions and accusative marking is the result of alternating two types of small clause heads. However, this assumption is ruled out because the alternation between prepositions and accusative case is not restricted to copular clauses. Prepositions and accusative marking alternate freely in other environments where small clauses cannot be available, as can be seen from the examples below: (26) a. bä-mäkina-w mätt’a-hu-bb-ät by-car-DEF.M

come\PFV-1SG.SJ-INSTR-3SG.M.OJ

‘I came by car.’ b. mäkina-w-n car-DEF.M-ACC

mätt’a-hu-bb-ät come\PFV-1SG.SJ-INSTR-3SG.M.OJ

‘I came by car.’ (27) a. lä-saba mäs’haf sätt’ä-hu-at to-Saba

book

give\PFV-1SG.SJ-3SG.F.OJ

‘I gave a book to Saba.’ b. saba-n Saba-ACC

mäs’haf sätt’ä-hu-at book

give\PFV-1SG.SJ-3SG.F.OJ

‘I gave Saba a book.’ Such an alternation can rather be explained as the result of replacing prepositions with some kind of functional projection that assigns case to the NP complements of the prepositions. I therefore propose that the functional projection which is responsible for accusative case marking must be the eventivity projection, which is different from the head of the small clause which is required for predication. The agreement phenomenon in accusative predicates also provides additional support to my claim. Recall that nominative predicates, unless they remain unmarked for phi-features, agree in number and gender with the subject (cf. (9), (10) and (11)). Accusative predicates behave differently from nominative predicates with regard to subject-predicate agreement. Unlike their nominative counterparts, accusative NP/DP predicates need not agree with the subject even though both the subject and the predicate are marked

Case Marking in Amharic Copular Constructions

275

for phi-features, as shown in (28). Given that the predicate is assigned case by the functional element which introduces eventivity, such an agreement pattern is explained straightforwardly. The phi-features of the predicate are controlled by the case-assigner and thus the predicate need not agree with the subject: (28) a. yonas tämari-očč-ɨn n-ä-w Jonas

student-PL-ACC

COP-3SG.M.SJ-3SG.M.OJ

‘Jonas behaves like students.’ b. lɨǰ-očč-u child-PL-DEF.M

tämari-w-n

n-aččäw

student-DEF.M-ACC

COP-3SG.M.SJ;3PL.OJ

‘The children behave like the student.’ c. saba tämari-w-n Saba

student-DEF.M-ACC

n-at COP-3SG.M.SJ;3SG.F.OJ

‘Saba behaves like the student (male).’ Stage-level predicates, on the other hand, are always marked for possessive agreement which indicates the number, gender and person features of the subject, as in (29): (29) a. ɨňňa rak’ut-aččɨn-ɨn we

naked-POSS.1PL-ACC

n-ä-n COP-3SG.M.SJ-1PL.OJ

‘We are naked.’ b. ɨne bado-ye-n I

empty-POSS.1SG-ACC

n-ä-n COP-3SG.M.SJ-1SG.OJ

‘I am empty [i.e. I don’t have anything at the moment].’ How does the subject trigger agreement twice (once with the predicate and once with the copula)? There are two possibilities to explain this: raising or control. In the raising explanation, the subject has to originate from the embedded position and agrees with the predicate. Then it raises to agree with the copula. In the control assumption, the subject originates in a higher position and controls another position in the stage-level adjective. In this case the possessive agreement is a manifestation of PRO/pro argument of the stagelevel adjective. The second assumption that the subject controls another position in the stage-level adjective is unlikely to account for the data for reasons associated with the theta role. If stage-level predicates assign the theta role to PRO/pro,

276

Mulusew Asratie

what assigns the theta role to the subject? The copula cannot be a theta role assigner since it is a functional element. The possessive agreement in stagelevel predicates, therefore, must be the result of raising. This means that the apparent possessive agreement indicates that the subject of stage-level predicates generates one step lower than individual-level predicates. Support for this comes from the fact that agreement with the copula is not obligatory. Subjects of stage-level predicates need not agree with the copula. The clauses in (29) are also acceptable with the subject establishing only possessive agreement with the adjective. In this case the impersonal copula takes two third person singular masculine agreement markers which license two expletive pros: (30) a. ɨňňa rak’ut-aččɨn-ɨn we

naked-POSS.1PL-ACC

n-ä-w COP-3SG.M.SJ-3SG.M.OJ

‘We are naked (lit. It is the case that we are naked).’ b. ɨne bado-ye-n I

empty-POSS.1SG-ACC

n-ä-w COP-3SG.M.SJ-3SG.M.OJ

‘I am empty (lit. It is the case that I am empty).’ Note that under the analysis that the copula is a raising verb (Stowell 1983), the subject of the small clause agrees with the copula by raising. The fact that the impersonal copula14 shows up with two default 3SG.M subject agreement markers with stage-level predicates in clauses like (29) suggests the presence of three different positions where two expletive pros and the overt subject originate. This supports the claim that such constructions contain an extra functional projection to generate each of them. I claim that this extra functional position from which the subject of stage-level predicates generates is the specifier of a functional projection which introduces eventivity and assigns accusative case to the predicate. Again further evidence comes from word order of clauses with accusative predicates. Unlike nominative predicates, accusative predicates cannot precede the subject. This means that the predicate–subject–copula order which provides a specificational interpretation in nominative predicates is not allowed with accusative predicates:

14 Recall footnote 3.

Case Marking in Amharic Copular Constructions (31) a. ɨňňa tämari-wočč(-u)-n we

277

n-ä-n

student-PL(-DEF.M)-ACC

COP-3SG.M.SJ-1PL.OJ

‘We are just like the students.’ b. *tämari-wočč(-u)-n

ɨňňa n-ä-n

student-PL(-DEF.M)-ACC

c. *rak’ut-aččɨn(-ɨn) naked-POSS.1PL(-ACC)

we

COP-3SG.M.SJ-1PL.OJ

ɨňňa n-ä-n we

COP-3SG.M.SJ-1PL.OJ

This restriction, however, applies only if the copula stays clause-finally. If the copula is between the subject and the predicate, the subject and the predicate can exchange positions. In this case, the clauses have a cleft interpretation: (32) a. ɨňňa n-ä-n

tämari-wočč(-u)-n15

we

student-PL(-DEF.M)-ACC

COP-3SG.M.SJ-1PL.OJ

‘It is us who are just like students.’ b. tämari-wočč(-u)-n student-PL(-DEF.M)-ACC

n-ä-n

ɨňňa

COP-3SG.M.SJ-1PL.OJ

we

‘It is just like the students who we are.’ Assuming that the specificational interpretation involves argument-movement (A-movement) of the predicate to the subject position of the clause (Moro 1997), a movement restriction of accusative predicates in (31) clearly suggests the presence of an extra functional projection which blocks such a movement due to the minimal link condition.16 This means that, as such

15 This sentence may seem a bit odd, but it is used in expressions such as (vii): (vii) ɨňňa n-ä-n tämari-wočč(-u)-n ɨnnantä we

COP-3SG.M.SJ-1PL.OJ

student-PL(-DEF.M)-ACC

you.PL

aydäll-aččɨhu-m NEG.COP-2PL.SJ-NEG

‘It is us not you who are just like student.’ 16 A-movement refers to the phenomenon of displacing a phrase into a position where a fixed grammatical function is assigned, e.g. the movement of the object to the subject position in passives: (viii) a. I read the book. b. The book was read ___ (by me). A-bar movement (i.e. non-argument movement), in contrast, refers to displacing a phrase into a position where a fixed grammatical function is not assigned, e.g. the movement of a subject or object NP to a pre-verbal position in interrogatives:

278

Mulusew Asratie

clauses contain an extra functional projection, the predicate cannot move since such a movement would violate the minimal link condition (Chomsky 1995). The clauses in (32), on the other hand, are acceptable because they involve an A-bar movement. This means that the copula first moves to the CP layer and the subject or the predicate move to the position preceding the copula. 3.3 Unmarked predicates Accusative case marking in Amharic requires the NP to be marked for gender, number or definiteness. This predicts that unmarked individual-level predicates are ambiguous between the nominative and the accusative interpretations, as in (33a) and (33b), while stage-level predicates have only one interpretation, as in (34). This indicates that the functional projection which introduces eventivity is also available with unmarked predicates, despite the morphological constraint that accusative marking requires the NP to be marked for number, gender or definiteness: (33) a. yonas tämari n-ä-w Jonas

student

COP-3SG.M.SJ-3SG.M.OJ

‘Jonas is a student.’ [He is real student.] b. yonas (huneta-w Jonas

situation-POSS.3SG.M

hullu) tämari n-ä-w all

student

COP-3SG.M.SJ-3SG.M.OJ

‘Jonas is just like a student (in all his behavior).’ (34) bunna-w

tɨkkus n-ä-w

coffee-DEF.M hot

COP-3SG.M.SJ-3SG.M.OJ

‘The coffee is hot.’ 4 Conclusion In this paper, I have shown that the predicate case alternation in Amharic is not related to the copula, as suggested by Maling & Sprouse (1995) and Comrie (1997), nor to the category of the predicate, as suggested by Pereltsvaig (2001). I have shown, following Matushansky (2008), that it is rather determined by the presence of an extra functional projection which (ix) a. You think John loves Mary. b. Who do you think ___ loves Mary? Under the minimal link condition, an A-movement is allowed to the next higher argument position while A-bar movement is allowed to the next higher non-argument position.

Case Marking in Amharic Copular Constructions

279

introduces eventivity. In the absence of this functional projection, however, I follow Pereltsvaig (2001) and claim that the predicate receives the default case. Abbreviations 1, 2, 3 First, second, third person ACC Accusative AGR Agreement AP Adjectival phrase AUX Auxiliary verb COP Copula CP Complementizer phrase DEF Definiteness DP Determiner phrase ECM Exceptional case marking EPP Extended Projection Principle F Feminine GEN Genitive INST Instrumental

IPFV M NEG NOM NP OJ PFV PL POSS PP PST SG SJ TAM TP VP

Imperfective Masculine Negation Nominative Noun phrase Object Perfective Plural Possessive Prepositional phrase Past tense Singular Subject Tense, aspect and mood Tense phrase Verb phrase

References Aarts, Bas. 1992. Small Clauses in English: The Nonverbal Types. Berlin: Mouton de Gruyter. Bailyn, John. 2001. The syntax of Slavic predicate case. In Anatolij Strigin & Assinja Demjjanow (eds.), Secondary Predication in Russian, 1–26. (ZAS Papers in Linguistics 25). Berlin: ZAS. Bailyn, John. 2002. Overt predicators. Journal of Slavic Linguistics 10. 23– 52. Bailyn, John & Edward J. Rubin. 1991. The unification of instrumental case assignment in Russian. Cornell Working Papers in Linguistics 9. 99–126. Baye Yimam. 2006. The interaction of tense, aspect and agreement in Amharic syntax. In John Mugane, John P. Hutchison & Dee A. Worman

280

Mulusew Asratie (eds.), Selected Proceedings of the 35th Annual Conference on African Linguistics, 193–202. Somerville, MA: Cascadilla Proceedings Project.

Bowers, John. 1993. The syntax of predication. Linguistic Inquiry 24. 591– 656. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, Juan Uriagereka & Samuel Jay Keyser (eds.), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, 89–155. Cambridge, MA: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz (ed.), Ken Hale: A Life in Language, 1–52. Cambridge, MA: MIT Press. Citko, Barbara. 2008. Small clauses reconsidered: Not so small and not all alike. Lingua 118. 261–295. Comrie, Bernard. 1997. The typology of predicate case marking. In Joan Bybee, John Haiman & Sandra A. Thompson (eds.), Essays on Language Function and Language Type Dedicated to T. Givón, 39–50. Amsterdam & Philadelphia: John Benjamins. Dikken, Marcel den. 2006. Relators and Linkers: The Syntax of Predication, Predicate Inversion and Copulas. Cambridge, MA: MIT Press. Getatchew Haile. 1974. The copula ነው (näw) in Amharic. IV Congresso Internazionale di Studi Etiopici, Tomo II (Sezione Linguistica), 139– 154. Roma: Accademia Nazionale dei Lincei. Higgins, Roger Francis. 1979. The Pseudo-cleft Construction in English. New York: Garland. Maling, Joan & Rex A. Sprouse. 1995. Structural case, specifier-head relations, and the case of predicate NPs. In Hubert Haider, Susan Olsen & Sten Vikner (eds.), Studies in Comparative Germanic Syntax, 167–186. Dordrecht: Kluwer. Matushansky, Ora. 2008. A case study of predication. In Franc Marusic & Rok Zaucer (eds.), Studies in Formal Slavic Linguistics: Contributions from Formal Description of Slavic Languages 6.5 Held at the University of Nova Gorica, December 1–3, 2006, 213–239. Frankfurt/Main: Lang. Mikkelsen, Line. 2005. Copular Clauses: Specification, Predication and Equation. Amsterdam & Philadelphia: John Benjamins.

Case Marking in Amharic Copular Constructions

281

Mikkelsen, Mike. 2011. Copular clauses. In Claudia Maienborn, Klaus von Heusinger & Paul Portner (eds.), Semantics: An International Handbook of Natural Language Meaning, vol. II, 1805–1829. Berlin & New York: Mouton de Gruyter. Moro, Andrea. 1997. The Raising of Predicates: Predicate Noun Phrases and the Theory of Clause Structure. Cambridge: Cambridge University Press. Pereltsvaig, Asya. 2001. On the Nature of Intra-Clausal Relations: A Study of Copular Sentences in Russian and Italian. Montreal: McGill University (PhD Dissertation). Stowell, Tim A. 1981. Origins of Phrase Structure. Cambridge, MA: Massachusetts Institute of Technology (PhD Dissertation). Stowell, Tim A. 1983. Subjects across categories. The Linguistic Review 2. 285–312.

Wandering along the Border of Finiteness: The Gəʿəz and Tigrinya Converb(s) in a Diachronic Perspective Stefan Weninger University of Marburg Abstract Studies on the grammar of converbs often have individual languages in their focus. This paper instead intends to study the converb in a comparative and diachronic perspective. In this way, we can better identify where finiteness is gained or lost and under what conditions this happens. The grammar of the converb in Classical Ethiopic (Gəʿəz) is compared with its counterpart in modern Tigrinya, the one modern Ethio-Semitic language that is the closest cognate to Gəʿəz. In some respects the Tigrinya converb loses features of finiteness, but in others it behaves more like a finite verb. On the other hand, we can also identify elements of stability. During the entirety of Ethio-Semitic’s lengthy history, the (former) Gəʿəz converb was neither grammaticalized to a finite verb form nor to a fully infinite category, but rather remained in the border region between finiteness and infiniteness, thus underpinning the necessity to develop a scalar definition of finiteness. 1 Introduction The present paper treats the Ethio-Semitic converb1 in a diachronic perspective. Studies of a given linguistic phenomenon can be carried out in different ways: as an in-depth, synchronic examination of the grammar of a particular phenomenon in a single variety,2 as a typological comparison of the phenomenon over many unrelated languages, or as a typological study restricted to a certain area (e.g. the Ethiopian area). But it is also desirable to study such phenomena in a diachronic perspective, to look at the history of the linguistic phenomenon in question. In this way, we can better identify where finiteness is gained or lost and under what conditions this happens. For most African languages this is rather difficult due to the lack of historical data. But the case 1 2

On the notion converb, as used here, see Haspelmath (1995), who mentions the Gəʿəz converb as an example for the nonfinite character of converbs (Haspelmath 1995:5 f.). As e.g. in the papers collected in Ebert, Mattissen & Suter (2008).

284

Stefan Weninger

of Ethio-Semitic is different. With Gəʿəz, the ecclesiastical language of the Ethiopian Orthodox church,3 we have access to data from the fourth century A.D. onward. Ideally, a study like this would be conducted on a language with a fully documented language history from an old stage step-by-step to a modern daughter-language. This is not quite the case with Gəʿəz. The core grammar of Gəʿəz remains basically the same throughout its history from the first attestation until the 19th century. Although there is some variation and development in pronunciation, phraseology, syntax, and lexicon due to language contact, the morphology is identical. But the following strategy seems promising: The starting point is the grammar of the converb in Classical Ethiopic (Gəʿəz), as established in Weninger (2001).4 The converb can be conceived as a form that is neither fully finite nor fully infinite. After a glance on its etymology, established through the comparison with the Asian Semitic languages and in the light of grammaticalization theory, the Gəʿəz system is compared with its counterpart in Tigrinya (described especially in Voigt 1977), the one Ethio-Semitic language that is the closest modern cognate to Gəʿəz (Bulakh & Kogan 2010). While Gəʿəz was a literary language already in the fourth century A.D., Tigrinya evolved from a vernacular that must have been quite close to Gəʿəz. 2 The grammar of the Gəʿəz converb Gəʿəz converbs are inflected according to person (1, 2, 3), number (SG, PL), and gender (M, F) and thus they share properties of the finite verb. The converb is a subordinate verbal form. It is always dependent on a verb in the matrix clause. Verbless clauses involving nominal predicates are not uncommon in Gəʿəz,5 but a converbial clause can never depend on a verbless matrix clause. A typical example of a narrative sentence with a converbial clause is (1):

3 4 5

On Gəʿəz, its position within the Semitic languages and the Ethiopian language area, its significance and history, and an outline of the prominent grammatical features, see Weninger (2011a; 2011b). Quite often this form is called gerund, especially by authors of the pre-1990s (but also e.g. Voigt 2010; 2011). The reasons, why the present author prefers the term converb against gerund or other terms, are explained in Weninger (2001:217 ff.). These are sentences with nominal predicates, sometimes involving a pronoun as a copula. Verbless sentences are a feature that Gəʿəz shares with the other classical Semitic languages like Akkadian, Hebrew, Classical Arabic, etc.

Wandering along the Border of Finiteness

285

(1)

GƎʿƎZ wä-ʾanbibomu

wäṣ́ʾu

wəstä däbrä

and-sing\CNV.3PL.M go_out\PRF.3PL.M in

zäyt

mountain.CST olive

‘And when they had sung, they went out to the Mount of Olives.’ (MK 14:26) A converbial clause is never introduced by a conjunction. The wä- at the beginning of (1) belongs to the whole sentence, not to the converbial clause.6 In the majority of cases, the converbial clause precedes the main clause, as in (1), but cases as in (2), where the converbial clause follows the predicate of the main clause, are also not rare: (2)

GƎʿƎZ wä-mäṣʾu

śäniḳomu

wä-ʾastädaliwomu

and-come\PRF.3PL.M prepare_provisions\CNV.3PL.M and-be_prepared\CNV.3PL.M

‘They came after they had prepared provisions and equipped themselves.’ (JOS 9:2) There seems to be no semantic or pragmatic difference between the two possible sequences. Converbs cannot be coordinated with the conjunction wäwith a finite verb form, but only with another converb, as in (2), where the converbs śäniḳomu ‘having prepared provisions’ and ʾastädaliwomu ‘having prepared themselves’ are coordinated by the conjunction wä- ‘and’. The subject of the converbial clause may be identical with the subject of the matrix clause, as in (1) and (2), which is more common. But subjects can also differ without the subject shift being indicated by a special marker, as in (3): (3)

6

GƎʿƎZ wä-ʾəm-zə

ʾalṣiḳomu

ʾəyäriḥo hallo

and-from-DEM

draw_near\CNV.3PL.M

Jericho

be\PRF.3SG.M

On some seeming exceptions that can be explained as anacolutha see Weninger (2001:225 ff.). Kapeliuk (2003:262) adds further examples from texts she calls “more reliable”. However, also these examples are not without problems. The verse MT 4:2 shows the wording quoted by Kapeliuk only in the B-Text, but not in the older A-Text (no wä-) nor in the D- and E-Texts (no converbs at all). The verse GEN 19:3 shows several variants that do not show this grammatical problem. So these examples are also not really “reliable”.

286

Stefan Weninger ʾaḥadu ʿəwur yənäbbər

wəstä fənot

one

on

blind

sit\IPFV.3SG.M

way

‘And after that, when they were approaching Jericho, there was a blind man sitting on the way.’ (LK 18:35) The converb can be combined with any finite verb, with the perfect,7 as in the preceding examples (1), (2), and (3), with the imperfect, as in (4), with the jussive, as in (5), and also with the imperative, as in (6). (4) ʾəsmä because

8

tämäkkiro

yənäśśəʾ

ʾaklilä

GƎʿƎZ ḥəywät

tempt\PASS.CONV.3SG.M

take\IPFV.3SG.M

crown.CST

life

‘For when he has stood the test, he will receive the crown of life.’ (Jas 1:12) (5)

GƎʿƎZ lä-yəgbəʾu VOL-turn_back\ JUS.3PL.M

täḫafiromu

dəḫre=homu

bä-gize=ha

after=them

in-time=her

ʾəllä

shame\PASS.CONV.3PL.M REL

yəbəlu=ni

...

say\IPFV.3PL.M=me

‘Let those be turned back ashamed who say to me: ...’ (6)

(PS 69:4) GƎʿƎZ

wä-bäwiʾakəmu

wəstä bet

täʾamməḫəwomu

and-enter\CNV.2PL.M

in

bless\IMP.PL.them

house

‘And when you enter a house, bless them!’

(MT 10:12)

The converb is hardly ever negated. Example (7) is an extremely rare exception:

7 8

Note that the term perfect in this contribution denotes an aspectual category of the verb expressing completion or relative anteriority of the verbal event. The scope of the conjunction ʾəsmä is on the whole sentence and is not restricted to the converbial clause. So the rule that converbs are not introduced by conjunctions is not violated by this example.

Wandering along the Border of Finiteness

287

(7)

GƎʿƎZ wa-ʾi-yaʾmiromu

ʾəllä

and-NEG-know\CNV.3PL.M REL

yäʿaḳḳəbu

rədät=o

yəbelu

care\IPFV.3PL.M

descent=his say\PRF.3PL.M

‘And those who cared for his descent without knowing spoke: ...’ (PHYSIOLOGUS 2, 8) The few attested cases of negated Gəʿəz converbs always have the verb ʾaʾmärä ‘know’;9 so the violation of the constraint on negated converbs is only minimal. The converb refers to preceding actions, as in (1), (2), (4) and (6), but also to concomitant states as in (3), or to an expression of manner (adverbial function), as in (5). Time reference depends on the matrix clause. In narrative contexts, as in (1) or (2) it refers to past situations. When the matrix clause refers to a situation in the future, the converb can refer to future situations, as we have seen in (4). The converbial clause can have the same illocutionary force as the matrix clause, as in (5), where the volitive modality of the converbial clause täḫafiromu ‘brought to shame’ is identical with the modality of the matrix clause. But this is not necessarily so. In (6) we have a case where the converbial clause is not part of the proposition that is in the imperative. So to sum up: Concerning the categories of person, gender and number, the Gəʿəz converb is certainly a finite category. Concerning tense and mood it is infinite. 3 Etymology The Gəʿəz converb originates from a nominal, infinitive-like form in the adverbial accusative. In ancient Semitic, the accusative marked not only the direct object, but also adverbial expressions, as can be seen in an example from Classical Arabic (Wright 1896–1898:II 106 ff.):

9 To the cases cited in Weninger (2001:230 f.), the cases SAP 12:10 and PHYSIOLOGUS 16 paen. have to be added (quoted in Praetorius 1879:373). Goldenberg (2013:216) raises the question “as to possible interference of later vernaculars in the transmission of Gəʿəz, since negative Gerund is typical of some Amharic dialects (of Goǧǧam) but intolerable in the standard variety” and adds a further example (LK 2:45) that has indeed problems of textual criticism. However, since the other cases of negative Gəʿəz converbs uniformly show the verb ʾaʾmärä ‘know’, the present author is hesitant to dismiss them all as errors of textual transmission.

288

Stefan Weninger

(8) allaḏīna yaḏkurūna REL.PL

llāha

think\IPFV.3PL.M God.ACC

qiyāman

CLASSICAL ARABIC wa-quʿūdan

stand\INF.ACC.INDF and-sit\ INF.ACC.INDF

‘those who think of God while standing or sitting’

(KORAN 3:191)

The Arabic accusative marker a is etymologically identical with the ä-accusative in Gəʿəz.10 The infinitive of the Gəʿəz simplex stem has the vowel i in the last syllable. In the other derived stems, the infinitive has the vowel ə instead. In the converb, the vowel i of the simplex stem has obviously been generalized, as shown in Table 35. Table 35: Infinitives and Converbs STEM 01 02 03 A1 T1 AST1 etc.

INFINITIVE ḳätilo(t) ḳättəlo(t) ḳatəlo(t) ʾaḳtəlo(t) täḳätəlo(t) ʾastäḳätəlo(t)

CONVERB (3SG.M) ḳätilo ḳättilo ḳatilo ʾaḳtilo täḳätilo ʾastäḳätilo

The termination of the converb is the product of the merger of the accusative marker with the enclitic possessive pronoun. The converb shows the same contractions as the noun, as can be seen in Table 36:

10 Classical Arabic, although attested only in the 6th ct. A.D., is rather conservative and shows many archaic elements that can be attributed to Proto-Semitic. Although today’s Semitists agree that this principle should not be used too simplistically (Arabocentrism), in this case the reconstruction of a Proto-Semitic accusative *=a (> Gəʿəz =ä) with an adverbial function is secure.

Wandering along the Border of Finiteness

289

Table 36: Suffixes of the Gəʿəz Converb Compared with Possessive Pronouns NOUN IN THE ACCUSATIVE

CONVERB

WITH POSSESSIVE PRONOUNS

SG

PL

1 2M 2F 3M 3F 1 2M 2F 3M 3F

‘my house’, etc. bet=əyä bet=äkä bet=äki bet=o bet=a bet=änä bet=äkəmu bet=äkən bet=omu bet=on

Reconstructed forms *bet=ä=yä *bet=ä=kä *bet=ä=ki *bet=ä=hu *bet=ä=ha *bet=ä=nä *bet=ä=kəmu *bet=ä=kən *bet=ä=homu *bet=ä=hon

‘I killing’, etc. Reconstructed forms ḳätil=əyä *ḳätil=ä=yä ḳätil=äkä *ḳätil=ä=kä ḳätil=äki *ḳätil=ä=ki ḳätil=o *ḳätil=ä=hu ḳätil=a *ḳätil=ä=ha ḳätil=änä *ḳätil=ä=nä ḳätil=äkəmu *ḳätil=ä=kəmu ḳätil=äkən *ḳätil=ä=kən ḳätil=omu *ḳätil=ä=homu ḳätil=on *ḳätil=ä=hon

So there is abundant evidence that the Gəʿəz converb evolved from what was originally an adverbial construction. 4 The Situation in Tigrinya 4.1 Continuity When we now compare the Gəʿəz situation with the converb in modern Tigrinya, we find that several properties remain the same. Just as in Gəʿəz, the Tigrinya converb can also refer to anterior actions, as we see in (9): (9)

TIGRINYA mahdawəyan

ʾəwwən

ʾənkab

sudan

tälaʿilom

Mahdist.PL

then

from

Sudan

depart\CNV.3PL.M

nə-ʾityop̣p̣ya

wåräru=wa,

bəwågän

bägemədr

to-Ethiopia

invade\PRF.3PL.M=her

along

Bägemədr

ʾatəyom

käʾa nə-gʷåndär ʾaḳḳaṣäluwa

come\CNV.3PL.M

then

to-Gondar

burn_down\PRF.3PL.M

‘But the Mahdists departed from the Sudan and invaded Ethiopia, they came over Bägemədr and burnt Gondar down.’ (Voigt 1977:150)

290

Stefan Weninger

The Tigrinya converb can also refer to concomitant states, as in (10), where the state of being on the knees – concomitant with the act of thanking – is a result of having knelt down immediately before: (10)

TIGRINYA tämärkikä nə-ʾamlaḵ ʾamäsgänku kneel\CNV

to-God

thank\PRF.1SG

‘I thanked God on my knees.’

(Voigt 1977:153)

The Tigrinya converb can be used with the perfect, as we saw in (9) and (10), but also with other tenses, as well as with the imperative (11). (11)

TIGRINYA ʾasfiḥka

ʾazäntu

extend\CNV.2SG.M recount\IMP.2SG.M

‘Tell extensively!’

(Voigt 1977:153)

As in Gəʿəz, the Tigrinya converb can also be used in adverbial function for expressions of manner. (11) was one example of this. A further example is (12): (12)

TIGRINYA tähaggʷisom

täḳäbbäləw=wo

be_happy\CNV.3PL.M

receive\PRF.3PL.M=him

‘Happily they received him.’

(Voigt 1977:154)

There are many lexicalized converbs in this function, words like ʾaʿmiḳu ‘deeply’, ʾaḳäddimu ‘first; earlier’ or dägagimu ‘repeatedly’, etc. That these converb forms are indeed lexicalized, i.e. not fully explainable by their root and grammatical form, can be seen from (13). The Tigrinya verb ʾamrärä means ‘make bitter, sour, hot (food by adding pepper), embitter, sadden, grieve, distress, cause bitterness or displeasure’ (Kane 2000:362). This causative or factitive meaning is just what is to be expected by an A-stem from the common Semitic root m-r-r ‘bitter’. So the converb should mean *‘after causing bitterness’ or *‘while causing bitterness’ or the like. But instead, it just means ‘bitterly’, as we can see in (13):

Wandering along the Border of Finiteness (13)

291 TIGRINYA

ʾamrirom

bäḵäyu=la

bitter\CNV.3PL.M cry\PRF.3PL.M=for_her

‘They cried bitterly for her.’

(Kane 2000:362)

The causative element is totally lost in this usage. Its meaning cannot be inferred from the meaning of the simplex stem. Hence the term is lexicalized. In this usage the converb can still be fully inflected in terms of person, number, and gender, but this is restricted to manner-expressions. Here we can see a certain loss of finiteness. A further feature that Tigrinya shares with Gəʿəz is the fact that the converb is hardly ever negated (Voigt 1977:169). 4.2 Innovations Not everything, however, remained the same. The use of the Tigrinya converb also shows significant differences in respect to that in Gəʿəz. To begin with and most notably: While the Gəʿəz converbial clause always depends on a verb in the matrix sentence, the Tigrinya converb can be the nucleus of a main clause. We can see this in (14): (14)

TIGRINYA ʾəskä gälä-do INTJ

sämiʿki

something-Q hear\CNV.2SG.F

‘Hey, did you hear something?’

(Voigt 1977:143)

This usage is, according to Voigt’s (1977:143) description, restricted to direct speech and to related kinds of utterances, like the beginning of letters, as in (15): (15)

TIGRINYA mälʾəḵti selä letter

zə-ṣäḥafkə=lläy

because REL-write\PRF.2SG.M=to_me

täḥaggʷisä be_happy\CNV.1SG

‘Because you have written a letter to me, I am happy.’ (Voigt 1977:143) In spite of this restriction, it is safe to say that the converb in Tigrinya has gained an important feature of finite verbs. It has increased its degree of finiteness.

292

Stefan Weninger

Secondly, the converb can be used as the complement of expressions of possibility or perception, as in (16): (16)

TIGRINYA tåwåṣṣiʿu

məs

räʾäytto

wound\PASS.CNV.3SG.M

when

see\PRF.3SG.F

‘when she saw him wounded’

(Voigt 1977:165)

This is a new development. Gəʿəz converbs could never be used as complements. Thirdly, we saw that the Gəʿəz converb is never introduced by a conjunction. In Tigrinya, this restriction is loosened. The Tigrinya converb can be employed in several types of embedded clauses that are introduced by conjunctions. As an example, we can see this in a temporal clause in (17): (17)

TIGRINYA dəḥri täḥaṣibka

ḳäyyər

after

change_clothes\IMP

wash\REFL.CNV.2SG.M

‘When you have washed, change your clothes!’

(Voigt 1977:169)

This is a field where the Tigrinya converb behaves more like a finite verb. So we have here another factor that sets the Tigrinya converb apart from the Gəʿəz converb with its more adverbial-like behavior. Finally, Tigrinya converbs can be employed with the copula ʾəyyu and with several auxiliary verbs, like konä ‘be(come)’ (cf. (18)), ʾallo ‘exist’, näbärä ‘was’ or with descriptive auxiliaries. (18)

TIGRINYA ṣəbaḥ

ʾəzi säʿat ʾəzi näḳilkum

morning DEM hour

DEM

leave\CNV.2PL.M

‘Tomorrow at this time you will be gone.’

təkonu be\IPFV.2PL.M

(Voigt 1977:178)

This is a vast field that cannot be dealt with here. Generally it can be said that the finiteness of the converb is strengthened in these combinations because they allow several types of usage in independent clauses. 5 Conclusion By comparing the Gəʿəz and Tigrinya converbs we found that several features remain stable, e.g. the usage for anterior actions, concomitant states and

Wandering along the Border of Finiteness

293

adverbial function for expressions of manner. Both in Gəʿəz and in Tigrinya, the converb is negated only in exceptional cases. Other features changed. The most important is that in Tigrinya the converb can be the nucleus of a main clause, thus gaining a certain degree of finiteness. So what do we learn from all this? In a diachronic perspective finiteness is certainly not a feature of a verb form that is either present or absent. Instead, we must conceptualize it in terms of degrees. The fact that, during the entirety lengthy history of Gəʿəz or from Proto-Tigrinya to modern Tigrinya, the (former) Gəʿəz converb was neither grammaticalized to a finite verb form nor to a fully infinite category, but rather remained in the border region between finiteness and infiniteness, underpins the necessity to develop a scalar definition of finiteness. Finiteness is not an absolute, but a relational category. And perhaps most importantly: the presence of forms that are neither fully finite nor fully infinite can be very stable in a diachronic perspective. Abbreviations ACC Accusative CNV Converb CST Construct state F Feminine IMP Imperative IPFV Imperfect INDF Indefinite INF Infinitive INTJ Interjection JUS Jussive

M NEG PASS PL PRF Q REFL REL SG VOL

Masculine Negation Passive Plural Perfect Question Reflexive Relative Singular Volitive

References Bulakh, Maria & Leonid Kogan. 2010. The genealogical position of Tigre and the problem of North Ethio-Semitic unity. Zeitschrift der Deutschen Morgenländischen Gesellschaft 169. 273–302. Ebert, Karen H., Johanna Mattissen & Rafael Suter (eds.). 2008. From Sibiria to Ethiopia – Converbs in a Cross-Linguistic Perspective. (Arbeiten des Seminars für Allgemeine Sprachwissenschaft 20). Zürich: Universität Zürich. Goldenberg, Gideon. 2013. Semitic Languages: Features, Structures, Relations, Processes. Oxford: Oxford University Press.

294

Stefan Weninger

Haspelmath, Martin. 1995. The converb as a cross-linguistically valid category. In Martin Haspelmath & Ekkehard König (eds.), Converbs in Cross-Linguistic Perspective: Structure and Meaning of Adverbial Verb Forms – Adverbial Participles, Gerunds, 1–55. Berlin: Mouton de Gruyter. Kane, Thomas Leiper. 2000. Tigrinya-English Dictionary. 2 vols. Springfield: Dunwoody Press. Kapeliuk, Olga. 2003. Review of [Weninger 2001]. Aethiopica 6. 259–263. Praetorius, Franz. 1879. Die amharische Sprache. Halle: Verlag der Buchhandlung des Waisenhauses [Reprint: Hildesheim: Olms, 1970]. Voigt, Rainer Maria. 1977. Das tigrinische Verbalsystem. Berlin: Reimer. Voigt, Rainer Maria. 2010. Das oromonische Konverb im Vergleich mit dem Gerundium im Amharischen, Tigrinischen und Türkischen. Aethiopica 13. 152–162. Voigt, Rainer Maria. 2011. Tigrinya. In Stefan Weninger (ed.), The Semitic Languages, 1153–1169. (Handbücher zur Sprach- und Kommunikationswissenschaft 36). Berlin: De Gruyter. Weninger, Stefan. 2001. Das Verbalsystem des Altäthiopischen: Eine Untersuchung seiner Verwendung und Funktion unter Berücksichtigung des Interferenzproblems. Wiesbaden: Harrassowitz. Weninger, Stefan. 2011a. Ethio-Semitic in general. In Stefan Weninger (ed.), The Semitic Languages, 1114–1123. (Handbücher zur Sprach- und Kommunikationswissenschaft 36). Berlin: De Gruyter. Weninger, Stefan. 2011b. Old Ethiopic. In Stefan Weninger (ed.), The Semitic Languages, 1124–1142. (Handbücher zur Sprach- und Kommunikationswissenschaft 36). Berlin: De Gruyter. Wright, William. 1896–1898. A Grammar of the Arabic Language. Third Edition, Rev. by W. R. Smith and M. J. de Goeje, I-II. Cambridge: Cambridge University Press.

Abbreviations for Gəʿəz text editions (in alphabetical order) GEN Boyd, James Oscar (ed.). 1909. The Octateuch in Ethiopic. Part I: Genesis. Leiden: Brill. JAS

Hofmann, Josef & Siegbert Uhlig (eds.). 1993. Novum Testamentum Aethiopice: Die katholischen Briefe. Stuttgart: Steiner.